Complex Oligosaccharide Utilization Pathways in Lactobacillus

Lactobacillus is the bacterial genus that contains the highest number of characterized probiotics. Lactobacilli in general can utilize a great variety of carbohydrates. This characteristic is an essential trait for their survival in highly competitive environments such as the gastrointestinal tract of animals. In particular, the ability of some strains to utilize complex carbohydrates such as milk oligosaccharides as well as their precursor monosaccharides, confer upon lactobacilli a competitive advantage. For this reason, many of these carbohydrates are considered as prebiotics. Genome sequencing of many lactobacilli strains has revealed a great variety of genes involved in the metabolism of carbohydrates and some of them have already been characterized. In this chapter, the current knowledge at the biochemical and genetic levels of the catabolic pathways of complex carbohydrates utilized by lactobacilli will be summarized. Introduction All animals establish symbiotic associations with microbes. Paramount among these is the establishment of complex microbial communities in the gastrointestinal tract. These communities may comprise thousands of species and are particular to each individual (Eckburg et al., 2005; Ley et al., 2006; Turnbaugh et al., 2009). The composition and distribution of microbiota changes along the gastrointestinal tract reflecting the differing physicochemical conditions and epithelial surfaces that the microbes find in the different compartments of the gastrointestinal tract (Gu et al., 2013; Kim and Isaacson, 2015; Stearns et al., 2011; Tropini et al., 2017). For example, Lactobacillaceae is predominantly found in the stomach and small intestine of mice whereas anaerobes such as Bacteroidaceae, Prevotellaceae, Rikenellaceae, Lachnospiraceae and Ruminococcaceae are mainly found in the large intestine and faeces (Gu et al., 2013). In the oral cavity of humans, a neutral pH and aerobic conditions prevail although anaerobic niches are also present. The environment in the stomach is acidic and microaerophilic whereas pH increases and oxygen availability decreases along the small intestine and colon. Nutrient availability and immune effectors also vary along the gastrointestinal tract. Furthermore, microbes are not uniformly distributed along the transverse axis of the gut (Tropini et al., 2017). Species such as Akkermansia muciniphila and some Bacteroides spp. are thought to be predominantly associated with the mucus layer whereas closer to the mucosa, aerotolerant taxa such as Proteobacteria and Actinobacteria are more abundant due to the radial oxygen gradient established across the intestinal wall (Albenberg et al., 2014). Numerous studies have demonstrated that gut microbiota has a great influence on manifold host physiology aspects, from metabolism (Li et al., 2008) to behaviour (Ezenwa et al., 2012). Perturbation of the gut microbiota may markedly affect the host health (Claesson et al., 2012; Cho and Blaser, 2012) and has been related to a number of diseases


Introduction
All animals establish symbiotic associations with microbes. Paramount among these is the establishment of complex microbial communities in the gastrointestinal tract. These communities may comprise thousands of species and are particular to each individual (Eckburg et al., 2005;Ley et al., 2006;Turnbaugh et al., 2009). The composition and distribution of microbiota changes along the gastrointestinal tract reflecting the differing physicochemical conditions and epithelial surfaces that the microbes find in the different compartments of the gastrointestinal tract (Gu et al., 2013;Kim and Isaacson, 2015;Stearns et al., 2011;Tropini et al., 2017). For example, Lactobacillaceae is predominantly found in the stomach and small intestine of mice whereas anaerobes such as Bacteroidaceae, Prevotellaceae, Rikenellaceae, Lachnospiraceae and Ruminococcaceae are mainly found in the large intestine and faeces (Gu et al., 2013). In the oral cavity of humans, a neutral pH and aerobic conditions prevail although anaerobic niches are also present. The environment in the stomach is acidic and microaerophilic whereas pH increases and oxygen availability decreases along the small intestine and colon. Nutrient availability and immune effectors also vary along the gastrointestinal tract. Furthermore, microbes are not uniformly distributed along the transverse axis of the gut (Tropini et al., 2017). Species such as Akkermansia muciniphila and some Bacteroides spp. are thought to be predominantly associated with the mucus layer whereas closer to the mucosa, aerotolerant taxa such as Proteobacteria and Actinobacteria are more abundant due to the radial oxygen gradient established across the intestinal wall (Albenberg et al., 2014).
Numerous studies have demonstrated that gut microbiota has a great influence on manifold host physiology aspects, from metabolism (Li et al., 2008) to behaviour (Ezenwa et al., 2012). Perturbation of the gut microbiota may markedly affect the host health (Claesson et al., 2012;Cho and Blaser, 2012) and has been related to a number of diseases such as metabolic disorders (Sonnenburg and Bäckhed, 2016), inflammatory diseases (Blander et al., 2017), diabetes (Membrez et al., 2008;Vaarala et al., 2008), coeliac disease (Collado et al., 2007), etc. Although the microbiota associated with an individual may display some resilience to perturbation (Lozupone et al., 2012;Sommer et al., 2017), external environmental factors such as food intake (Kohl et al., 2014) and composition (David et al., 2014) can alter it substantially. This aspect is pivotal in the relationship between microbiota and host nutrition and health as the gastrointestinal microbiota provide important metabolic capabilities, including the ability to obtain energy from indigestible dietary polysaccharides (Goh and Klaenhammer, 2015). The human genome encodes only 17 glycosidases, enabling the utilization of a very limited set of polysaccharides (Cantarel et al., 2012;Goh and Klaenhammer, 2015). Therefore, most complex carbohydrates are available to gut microbes able to utilize these compounds.
Carbohydrates, as oligo-and polysaccharides or as glycoconjugates constitute an amazingly diverse group of molecules due to the structural diversity and numerous bonding sites of their constituent monosaccharides that allow their assembly among themselves or to almost any other organic molecule in a wide array of architectures. Because of this structural versatility, carbohydrates fulfil a wide variety of functions in organisms as structural polymers, energy reserve, signalling, etc. Furthermore, as a major component of human diet, carbohydrates also have a determining influence on the interactions between host and associated microbiota (Hooper et al., 2002). In addition to dietary carbohydrates, host-derived glycans constitute a secondary source of carbohydrates for gut microbiota (Hooper et al., 2002).
Although the gut microbiota taken as a whole can utilize a wide variety of glycans, cells belonging to individual taxa can usually metabolize a limited set of them. From this fact stems the concept of prebiotic, defined as a non-digestible food ingredient that beneficially affects the host by selectively stimulating the growth and/or activity of one or a limited number of bacteria in the colon, and thus improves host health (Gibson and Roberfroid, 1995). Prebiotics were first thought of as a means to selectively enrich probiotic gut microbes, specifically Lactobacillus and Bifidobacterium (Gibson and Roberfroid, 1995;Goh and Klaenhammer, 2015). Although not limited by the definition, all prebiotics are glycans. The definition of prebiotic has been revised and the specific stimulatory effect of prebiotics on lactobacilli and bifidobacteria has been challenged (Hutkins et al., 2016). Notwithstanding, the original idea gave a strong boost to the study of the glycan catabolic pathways of these organisms.
Lactobacillus is a large genus currently comprising over 200 species that have been isolated from a wide variety of habitats. They are Gram-positive, microaerophilic or anaerobic obligate fermentative organisms that produce lactic acid as the major end product of sugar fermentation. Together with genera Paralactobacillus, Pediococcus and Sharpea they constitute the family Lactobacillaceae within the order Lactobacillales. Lactobacillus strains play a major role in the production of a wide variety of fermented products. Others are naturally associated with mucosal surfaces of humans and animals and have been considered as probiotics (Tannock, 2004). Either as foodstuff fermenters or as probiotics, their capacity to utilize glycans is an important trait for their performance. This chapter focuses on our current knowledge on the pathways of complex glycan dissimilation identified in species of Lactobacillus.

Fructan and fructooligosaccharide catabolic pathways
Structural characteristics of fructans Fructans are linear or branched fructose polymers and can be broadly divided into inulins (β-2,1-linked) and levans (β-2,6-linked) (Fuchs, 1991) (Fig. 3.1). Fructans are usually synthesized from sucrose by repeated fructosyl transfer so that they have a terminal glucose unit. Levans are produced by many microorganisms, including some lactobacilli (Bello et al., 2001;Tieking et al., 2003;Van Geel-Schutten et al., 1999), and a few plant species (Öner et al., 2016). In contrast, inulins are relatively common in plants, especially Asteraceae, but only a few bacterial species produce them, among them some Lactobacillus and Leuconostoc species (Anwar et al., 2008;Olivares-Illana et al., 2003;van Hijum et al., 2002). While bacterial fructans have a very high degree of polymerization (DP) up to 10 5 fructose units, the DP of plantderived fructans does generally not exceed DP 100. High DP of bacterial fructans is possibly related to their function as structural components of biofilms, but they also constitute an extracellular nutrient reservoir (Öner et al., 2016). In plants, fructans serve essentially as reserve carbohydrates.

Metabolic pathways for fructans utilization
The interest on fructans in relation to the intestinal microbiota stemmed from the search of carbohydrate sources that reached the large intestine and were selectively used by beneficial microbes such as bifidobacteria. Several studies noted that fructooligosaccharides derived from inulin hydrolysis (usually referred to as FOSs) were not hydrolysed by the host endogenous enzymes but efficiently used by bifidobacteria (Hidaka et al., 1986;Yazawa and Tamura, 1982). FOS naturally occur in many kinds of plants but they are commercially produced from the hydrolysis of inulin or synthesized from sucrose by transfructosylation by β-fructofuranosidases or β-d-fructosyltransferases (Goh and Klaenhammer, 2015). The most utilized natural source of inulins is chicory and depending on the method of extraction the product obtained may be almost exclusively FOS of the GF n type (a glucose monomer linked α-1,2 to two or more β-2,1-linked fructosyl units) or a mixture of GF n and F m type (two or more β-2,1-linked fructosyl units) oligomers (Roberfroid et al., 1998). However, the first studies that reported the utilization of fructans by lactobacilli were focused on the usage of lactic acid bacteria for ensilage and showed that some lactobacilli could utilize fructans for growth (Kleeberger and Kühbauch, 1976). In a later study, it was shown that 16 strains out of 712 were able to degrade levan and eight of them could also degrade inulin (Müller and Lier, 1994). Subsequent studies confirmed the ability of some lactobacilli to degrade fructans (Merry et al., 1995;Müller and Steller, 1995;Winters et al., 1998) although the enzymes and pathways involved were not determined. Other studies also established that some lactobacilli could grow on FOSs as carbon sources (Kaplan and Hutkins, 2000;Sghir et al., 1998). The characterization of an extracellular fructan hydrolase from Lactobacillus paracasei subsp. paracasei P 4134 provided the first clues on the fructan degradative pathways of lactobacilli (Müller and Seyfarth, 1997). The enzyme hydrolysed β-2,6-linked fructans more rapidly than β-2,1 linked fructans and the main product of hydrolysis was fructose, suggesting that the enzyme is an exofructanase (Müller and Seyfarth, 1997). Later studies on other L. casei/paracasei have confirmed these conclusions (Kuzuwa et al., 2012;Velikova et al., 2017). Subsequently, another extracellular fructanhydrolase purified from Lactobacillus pentosus B235 was characterized (Paludan-Müller et al., 2002). The purified enzyme had the highest activity for levan, but also hydrolysed garlic extract, a β-2,1-linked fructan with β-2,6-linked fructosyl sidechains, 1,1,1-kestose, 1,1-kestose, 1-kestose, inulin and sucrose at 60, 45, 39, 12, 9 and 3%, respectively, of the activity observed for levan (Paludan-Müller et al., 2002).
Genome sequencing and transcriptomic analyses paved the way for the identification and characterization of genes involved in the utilization of fructans. The studies carried out so far have characterized three different fructan utilization pathways in Lactobacillus acidophilus, Lactobacillus casei and Lactobacillus plantarum that differ in their transport systems and fructofuranosidase enzymes ( Fig. 3.2).
In 2003, Barrangou et al. (2003) identified an operon (msm) involved in FOS utilization by Lactobacillus acidophilus NCFM. The msm operon consisted of genes encoding for a transcriptional regulator of the LacI family (msmR), an ABC transport system (msmEFGK), a fructosidase (bfrA) and a sucrose phosphorylase (gtfA; Fig.  3.3) (Barrangou et al., 2003). Similar operons can be detected in a limited number of strains of other species of lactobacilli ( Fig. 3.3). In Lactobacillus delbrueckii, Lactobacillus perolens, Lactobacillus saniviri and Lactobacillus concavus strains, a gene encoding a putative fructokinase is associated with the msm   Fig. 3.3). The transcriptional analysis of the msm operon of L. acidophilus NCFM showed that all genes were transcribed in a single transcriptional unit (Barrangou et al., 2003). Sucrose and oligofructose (both GF n and F n types) induced the expression of the operon whereas glucose and fructose did not. The expression of the operon was repressed by glucose suggesting that it is subjected to carbon catabolite repression (CCR). This hypothesis was further supported by the presence of several CRE-like sites in the msm promoter region. The functionality of the operon was shown by inactivation msmE and bfrA that resulted in defective growth on FOS-F n (Barrangou et al., 2003). A comparative analysis of L. ruminis strains of human and bovine origin led to the identification of an operon possibly involved in FOS utilization consisting of a β-fructan hydrolase and an oligosaccharide H + symporter (O'Donnell et al., 2011) although experimental evidence is still lacking.
In Lactobacillus plantarum WCFS1, a gene cluster consisting of a putative fructokinase (sacK1), a putative phosphoenolpyruvate-dependent phosphotransferase transport system (PTS) of the β-glucoside family (pts1BCA), a β-fructofuranosidase (sacA), a LacI family transcriptional regulator (sacR) and a putative α-glucosidase (agl2) (Fig. 3.3), was induced when this strain was grown in the presence of low molecular weight FOS (Saulnier et al., 2007). The biochemical characterization of the L. plantarum ST-III SacA demonstrated that the enzyme has exofructofuranosidase activity with preference for β-2,1 linkages between two fructose moieties in fructans with low DP . The heterologous expression of SacA in Lactobacillus rhamnosus GG, an organism that can utilize fructose but not FOS, enabled this strain to grow on FOSs, thus demonstrating the functional role of this enzyme . Furthermore, inactivation of sacA in L. plantarum ST-III severely impaired the growth of this strain on FOSs (Chen et al., 2015). In contrast, a mutant defective in pts1BCA still could grow with FOSs although at a lower growth rate than the wild-type strain (Chen et al., 2015). A transcriptomic analysis of this strain had detected a second putative sac (sacPTS26)  gene cluster constituted by a β-glucoside PTS (PTS26), an α-glucosidase (Agl4), and a transcriptional regulator (SacR2), that was also induced in the presence of FOSs (Chen et al., 2015). A double mutant pts1BCA/pts26 was unable to grow on FOS, indicating that both transporters are required for optimal FOS uptake and utilization (Chen et al., 2015). This second sac cluster is also present in L. plantarum WCFS1 (Fig. 3.3) but it was not detected as up-regulated in the presence of FOS in this strain (Saulnier et al., 2007) and its involvement in FOS utilization in WCFS1 remain to be determined (Chen et al., 2015). The presence of α-glucosidase-encoding genes in sac clusters is also intriguing. The functional role of these genes remains to be established. As mentioned above, L. paracasei utilizes an extracellular fructanhydrolase. Goh et al. (2006) determined that the genes required for FOS utilization by L. paracasei 1195 are organized in a cluster (fosRABCDXE) encoding a putative mannose family PTS transporter (fosABCDX), a β-fructosidase (fosE) and, divergently transcribed, a transcriptional antiterminator (fosR) (Fig. 3.3). Homologous clusters are found in other L. casei/ paracasei but considerable variability is observed. For example, L. paracasei ATCC 334 fosE homologue (LSEI_0564) lacks the C-terminal part including the cell wall anchor motif whereas this gene is absent in strain BL23 (Fig. 3.3). This variability may account for the conflicting observations of extracellular or cell wall anchored fructanhydrolase activity in different L. paracasei strains. Inactivation of fosE led to the loss of the ability to grow on sucrose, FOS, oligofructose (FF n type), inulin and levan, thus demonstrating the functionality of the operon (Goh et al., 2006). Furthermore, introduction of fosE into Lactobacillus rhamnosus GG enabled this strain to utilize FOS (Goh et al., 2007). The analysis of the FosE encoding sequence revealed an N-terminal signal peptide sequence and an LPQAG cell wall anchor motif at the C-terminal region, suggesting its localization at the cell wall. Cell fractionation assays confirmed this hypothesis as FOS hydrolysis activity was present exclusively in the cell wall extract of L. paracasei previously grown on FOS (Goh et al., 2007). In agreement with a previous biochemical characterization of a fructanhydrolase of L. paracasei (Müller and Seyfarth, 1997), the analysis of the degradation products of L. paracasei 1195 FosE indicated that it is an exofructanhydrolase (Goh et al., 2007). The transcriptional regulation of the fos cluster has also been studied. Expression of fos genes is induced in the presence of FOS, inulin and, to a lesser extent, sucrose and fructose, but repressed by glucose (Goh et al., 2007;Goh et al., 2006). A CRE sequence is present in the lev promoter region, suggesting that the operon is subjected to CCR via the P-Ser-HPr/ CcpA complex (Goh et al., 2006).
The role of FosR has not been addressed in L. paracasei 1195; however, the functional role of the homologous LevR of strain BL23 has been studied (Mazé et al., 2004). The intergenic regions of fosR-fosA and levR-levA in strain BL23 differ only in a single nucleotide and the FosABCD proteins shared more than 99% identity with their BL23 counterparts (Goh et al., 2006). LevR is homologous to the Bacillus subtilis LevR transcriptional regulator, which controls the expression of a mannose-class PTS transporter and a levanase involved in levan utilization. B. subtilis LevR interacts with the σ 54 factor and its activity is modulated via phosphorylation by P-His-HPr and P-EIIB Lev (Martin-Verstraete et al., 1998). In contrast, BL23 strain LevR do not require σ 54 although the regulation of its activity by phosphorylation still occurs by dual PTS-catalysed phosphorylation at conserved histidine residues in the EIIA and PRD2 domains of LevR by P-His-HPr and P-His-EIIB Lev , respectively (Mazé et al., 2004). When the PTS Lev transporter is active, P-His-EIIB Lev preferably donates its phosphoryl group to the transported sugar, leading to dephosphorylation of LevR at His-776 by P-His-EIIB Lev and LevR activation and thereby induction of the lev PTS. On the other hand, when metabolically preferred PTS sugars, such as glucose, are present, the phosphoryl group of P-His-HPr is used for sugar phosphorylation. Poor phosphorylation at His-488 by P-His-HPr renders LevR less active and down regulates expression of the lev PTS.
The different strategies of fructan utilization by lactobacilli possibly determine their abilities to utilize different fructans. Internalization and subsequent hydrolysis possibly limits the ability of fructan utilization to low DP oligosaccharides whereas extracellular degradation would enable the utilization of high DP fructans. Experimental evidence available supports this view. Makras et al. (2005) assayed the capacity of ten strains of lactobacilli to degrade inulin-type fructans, observing that L. acidophilus only degraded oligofructose whereas L. paracasei could also degrade long-chain inulin. L. plantarum could utilize short-chain fructooligosaccharides but grew poorly with FOS (Saulnier et al., 2007). It has been proposed that the different strategies of fructan utilization in lactobacilli may respond to different ecological strategies. Internalization and subsequent degradation eliminates cross-feeding while conferring an advantage in a nutrient-competitive environment. On the other hand, extracellular degradation let other organisms profit of the hydrolytic products thereby allowing the establishment of symbiotic relationships with other members of the community (Goh and Klaenhammer, 2015).

Metabolism of glucans and glucooligosaccharides
Structural characteristics of glucans Lactobacillus species are equipped with the enzymatic machinery to utilize multiple glucan structures, which consist of glucose homopolymers with different linkages and branching types. However, the presence of these capacities is species-and strain-specific and usually linked to particular niche adaptations. Glucans can be classified into α-and β-glucans depending on the type of glycosidic bond present in the molecules ( Fig. 3.4). Starch is the main example of α-glucan and it represents the major carbon storage polysaccharide in plants. It is made of linear glucose chains with α-1,4 linkages with high DP (amylose) and shorter chains which in addition to α-1,4 bonds possess around 5% of α-1,6 branching of 18 to 25 glucoses (amylopectin). Other α-glucans can be linear or branched and carry diverse bonds (α-1,2; α-1,3) in addition to α-1,4 and α-1,6. Most of these glucans are produced by bacteria and fungi, such as dextrans (α-1,6 with α-1,3 branching; Fig. 3.4) produced by strains of lactobacilli such as Leuconostoc mesenteroides (Chen et al., 2016), Weisella cibaria (Malang et al., 2015) or Lactobacillus sakei (Nácher-Vázquez et al., 2017), reuteran (α-1,4 with α-1,6 branching), synthesized by Lactobacillus reuteri (Chen et al., 2016) or pullulan (maltotriose units linked by α-1,6 bonds; Fig. 3.5), produced by Aureobasidium pullulans (Cheng et al., 2011). Isomaltooligosaccharides (IMO) are a type of α-glucans that are gaining interest due to their prebiotic effects (Ketabi et al., 2011;Leemhuis et al., 2014;Yen et al., 2011). They are short α-1,6-linked glucans (e.g. isomaltose , isomaltotriose, isomaltotetraose and isomaltopentaose) that are present in some foods or can be commercially prepared from starch. Their prebiotic effect derives from the fact that humans and other monogastric animals generally lack IMO-degrading enzymes. Glycogen is an α-glucan equivalent to starch but present in animals and it is characterized by being more extensively branched. Many bacteria, including lactobacilli (Goh and Klaenhammer, 2013), can synthesize and utilize this molecule as carbon storage.
Cellulose is the most abundant β-glucan present in nature. It is a key component of plant cell walls so that it accounts for the major proportion of fixed carbon in living organisms. Plant as well as fungal cell walls may contain other β-glucans with β-1,3 or multiple alternating β-1,3 and β-1,4 linkages and β-1,6 branching linkages. Similar to α-glucans, some β-glucans can be produced by bacteria, including lactobacilli. Examples of these are the β-glucans produced by Lactobacillus suebicus (β-1,3 linked) (Garai-Ibabe et al., 2010) and L. brevis (Fraunhofer et al., 2017), where the glycosyl transferases catalysing the synthetic process have been characterized.
Lactobacilli usually encode multiple α-glycosidases in their genomes, mainly from the glycosyl hydrolase (GH) 13 family (CAZy, Carbohydrate Active enZYmes classification; www. cazy.org), the family to which α-amylases belong, although their specificities remain to be investigated in most cases. Among α-glucans, starch degradation by lactobacilli is relatively well characterized. The ability to degrade starch was noticed after the isolation and phenotypic characterization of new Lactobacillus strains from waste maize fermentations such as Lactobacillus amylophilus (Nakamura and Crowell, 1979) and Lactobacillus amylovorus (Nakamura, 1981). However, extracellular amylases are not common in this genus (less than 2% of all the GH13 glycosyl hydrolases present in lactobacilli). They are mainly concentrated in five species: L. acidophilus, L. amylovorus, Lactobacillus fermentum, L. plantarum and Lactobacillus manihotivorans (Petrova et al., 2013), although some particular strains belonging to different species can also degrade starch (e.g. Lactobacillus paracasei B41 (Petrova and Petrov, 2012)). In addition to amylases, pullulanases and amylopullulanases have been found in lactobacilli. A thermostable pullulanase (endo-α-1,6-glucosidase, GH13_14 subfamily) encoded by the LBA_1710 gene from L. acidophilus NCFM has been characterized. This enzyme preferentially acts on β-limit dextrins (amylopectins digested by β-amylases) over amylopectin (Møller et al., 2017). The products of LBA_1710 can also act on the linear polymer pullulan [(maltotriose-α1,6-maltotriose) n ] and possess a very low K m and very high specific activity for this polysaccharide. Notwithstanding, pullulan and amylopectin do not support growth of this strain. This would reflect the lack of α-amylase activity and

A B
Dextran a non-efficient transport system for the resulting oligosaccharides after pullulanase digestion in this strain. The enzyme is secreted due to presence of a signal peptide and possesses two starch-binding modules (CBM48), a catalytic domain (GH13_14 subfamily) and an additional surface-layer associated domain (SLAP) which warrants its retention at the cell surface. Homologues of this enzyme are present in many lactobacilli from intestinal origin. The enzyme may act on short-branched α-glucans derived from the degradation of dietary starch and glycogen by the human enzymes, supporting that these debranching enzymes play a role in the adaptation of these lactobacilli to the gut niche (Møller et al., 2017). An extracellular amylopullulanase from L. plantarum L137, an isolate from a traditional fermented food containing fish and rice with high hydrolytic activity towards starch, has also been characterized. The enzyme degrades soluble starch to maltotriose and maltotreaose whereas it produces only maltotriose from pullulan (Kim et al., 2008). The encoding gene (apuA) is located in the endogenous plasmid pLTK13 (Kim et al., 2008). The enzyme contains a number of amino acid repeats at the N-and C-terminus that are derived from the same repeated DNA sequence (5′-ACCGACGCAGCCAACTCA-3′) but translated in different frames. The C-terminal repeats are similar to mucin-binding domains present in bacterial peptidoglycan-bound proteins and most probably participate in substrate binding (Kim et al., 2008). Domains with amino acid repeats are typically found in carbohydrate-degrading enzymes. The amylases from L. amylovorus, L. plantarum and L. manihotivorans carry a C-terminal starch-binding domain (SBD) of almost 500 amino acids consisting of tandem repeat units of 91 amino acids in variable numbers (Morlon-Guyot et al., 2001). SBDs promote attachment to the substrate and allow degradation of non-soluble starch (Rodriguez Sanoja et al., 2000). A neopullulanase has been cloned and characterized from Lactobacillus mucosae LM1 (Balolong et al., 2016). A homologous gene had been previously shown to be induced in the amylolytic L. plantarum A6 strain during pearl millet fermentation and neopullulanase activity detected (Humblot et al., 2014). A number enzymes involved in the degradation of IMOs encoded by lactobacilli have also been characterized. A glucan-α-1,6-glucosidase (GH13_31 subfamily) encoded by the L. acidophilus NCFM gene LBA_0264 is involved in the catabolism of IMO (Møller et al., 2012). The participation of glucan-α1,6-glucosidase and maltose phosphorylase (see below) have also been implicated in the catabolism of IMO in L. brevis (Hu et al., 2013). The L. acidophilus NCFM enzyme is induced after growth in IMO mixtures, displays a high activity on panose and prefers IMO with more than two glucoses, as it acts preferentially on isomaltotriose and isomaltotetraose compared to isomaltose. The LBA_0264 genes and its homologues in other lactobacilli are not clustered with other sugar catabolism or sugar transporter genes except for Lactobacillus johnsonii ATCC 33200 and Lactobacillus gasseri JV-V03, which belong to the acidophilus group and where the IMO-utilizing gene forms part of the maltodextrin operon (Møller et al., 2012), and in other Lactobacillus species such as Lactobacillus casei (Monedero et al., 2008). Another GH13_13 enzyme has been characterized in L. plantarum LL441. It displays activity on isomaltose and isomaltulose [α-d-glucopyranosyl-(1→6)-d -fructofuranoside], but not on panose or isomaltotriose, and its gene is clustered with genes encoding EIIABCD components of a mannose-class PTS (Delgado et al., 2017). Whether the enzyme might be acting on phosphorylated disaccharides needs to be proven.

Metabolism of maltodextrins
Utilization of maltodextrins (linear oligosaccharides derived from starch hydrolysis) is a characteristic more extended in lactobacilli. Species adapted to starch-rich environments (e.g. plant material fermentations where the endogenous plant amylases release maltose and maltodextrins) such as Lactobacillus sanfranciscensis, participating in sourdough fermentations during bread making, are particularly efficient in maltose and maltodextrin utilization. The enzymatic machinery for the utilization of these carbohydrates in lactobacilli soon attracted attention and it has been thoroughly investigated. In the 90´s maltose catabolic proteins were partially characterized in L. sanfranciscensis DSM20451 (Ehrmann and Vogel, 1998) and Lactobacillus brevis ATCC 8287 (Hüwel et al., 1997). These microorganisms relied on an intracellular maltose phosphorylase (MapA [EC 2.4.1.8]) non requiring pyridoxal 5′-phosphate and belonging to the GH65 family, which also includes trehalose phosphorylases (EC 2.4.1.64), and kojibiose phosphorylases (EC 2.4.1.230), for catalysing a phosphorolysis of the disaccharide that involves an inversion of the anomeric configuration of the C-1 atom, giving β-glucose 1-phosphate and glucose. β-glucose 1-phosphate is further converted to glucose 6-phosphate for its incorporation into glycolysis via β-phosphoglucomutase (Pgm), whose gene is co-localized in the chromosome with mapA ( Fig. 3.6). MapA displays a high specificity for maltose, but it is not active on maltodextrins such as maltotriose or maltotetraose and it cannot phosphorolyse disaccharides with other α-linkage configurations such as isomaltose (α-1,6), nigerose (α-1,3), kojibiose (α-1,2) or trehalose (α-1,1). Furthermore, it differs in sequence from other maltose (maltodextrin) phosphorylases such as that of E. coli (GT35 family). MapA is a dimer in solution and its structure has been solved for the L. brevis enzyme (Egloff et al., 2001). The structure consists of a β-sandwich domain linked to an (α/α) 6 barrel catalytic domain, and a C-terminal β-sheet domain. The (α/α) 6 barrel domain displays striking structural and functional similarities to the catalytic domain of a glucoamylase from Aspergillus awamori, suggesting evolution from a common ancestor (Egloff et al., 2001). In the structures of these enzymes the catalytic conserved Glu residue from the glucoamylase superposes onto a conserved Glu of MapA that likely acts as the acid catalytic residue that promotes the nucleophilic attack of phosphate on the glycosidic bond. Modelling of the L. acidophilus MapA protein with different substrates allowed understanding why this enzyme does not accommodate maltotriose or maltotetraose in its active site (Nakai et al., 2009). The protein adopts a configuration where loop His413-Glu421 between α3 and α4 of the (α/α) 6 barrel domain blocks the binding of longer malto-oligosaccharides. However, this enzyme has been applied in reverse phosphorolysis reactions for the synthesis of α-1,4-linked disaccharides with β-glucose 1-phosphate as donor and glucose, glucosamine, N-acetyl glucosamine, l-fucose, mannose or xylose as acceptors; being unable to use other sugars with axial hydroxyls at C-3 and C-4 positions or disaccharides/trisaccharides (Nakai et al., 2009).
A genome survey of 38 Lactobacillus strains revealed that the presence of a set of genes for maltodextrins utilization, including mapA and pgm, is widespread in lactobacilli. These gene clusters usually include different α-glycosidases, although the presence of amylases is scarce (Gänzle and Follador, 2012). The maltodextrin operon which contain the mapA and pgm genes have been genetically characterized in L. acidophilus (Nakai et al., 2009) and L. casei (Monedero et al., 2008), showing that in these species maltodextrin transport is carried out by an ABC transporter (MalEFGK 2 ) homologous to that of the well-studied maltose/ maltodextrin transporter of E. coli (Fig. 3.6). In L. casei BL23, ten mal genes are clustered and cotranscribed in a single mRNA whose expression is regulated by MalR, a transcriptional regulator of the LacI/GalR family and repressed by the presence of  L. ruminis glucose via the general CCR mechanism mediated by the global regulator CcpA. According to studies conducted with Streptococcus pneumoniae MalR, this regulator acts as a transcriptional repressor and maltose strongly inhibits its DNA-binding capacity (Puyet et al., 1993). Cis-acting sequences recognized by MalR can be identified adjacent to the −10 and −35 promoters of the L. acidophilus and Lactococcus lactis mal operons (Nakai et al., 2009). In addition to the maltose phosphorylase and phosphoglucomutase genes, three α-glucosidase-encoding genes (GH13) are clustered together with the L. casei maltose-catabolic genes: malL, encoding a putative oligo-α1,6-glucosidase (putatively acting on short IMO like isomaltose); malM, encoding a maltogenic α-amylase (cuts α-1,4 linkages from dextrins yielding maltose) and dexB, coding for a second α-1,6 glucosidase (Monedero et al., 2008). This last enzyme belongs to the GH13_31 subfamily (glucan-α-1,6-glucosidase) and, as described before, based on the studies of its L. acidophilus counterpart is able to degrade IMO. Therefore, it is postulated that in addition to maltodextrins, the ABC transporter MalEFGK 2 would be able to transport IMO. A second operon located in opposite direction encodes an additional ABC transporter with high homology to MalEFGK 2 but its function is unknown and mutations in their genes have no phenotypic effects on maltose or maltotriose growth. In other L. casei strains, such as ATCC334, maltodextrin utilization is impaired by a large deletion in the mal cluster (Monedero et al., 2008). The mal operon of L. acidophilus NCFM consists of nine genes and a malR regulator (Nakai et al., 2009). In addition to the maltodextrin ABC transporter, MapA and Pgm, L. acidophilus encodes two glycosidases: a maltogenic α-amylase (MalM) and an oligo α-1,6-glucosidase (MalL) that are homologous to their L. casei counterparts. The L. acidophilus mal operon also carries an acetate kinase gene (ackA) involved in pyruvate metabolism during glycolysis. Although ABC transporters can be identified as the main maltodextrin transport systems in lactobacilli, in some species that efficiently use maltose as a carbon source a genetic association of a gene encoding a permease of the major facilitator superfamily (MFS) is found with mapA and pgm genes (Fig. 3.6). This suggests that these species (e.g. L. sanfranciscensis, L. salivarius, L. brevis) make use of a maltose-H + symport system for the uptake of the disaccharide. In contrast, no PTS systems for the transport of maltose similar to those described in Bacillus subtilis have been identified in lactobacilli. Studies on the maltose uptake in L. sanfranciscensis LTH2581, a strain which only ferments maltose and glucose, confirmed the presence of a maltose-H + symport system. When maltose is taken up by this strain the intracellularly generated glucose exceeds the metabolic capacity of the cells, which results in glucose expulsion through a glucose uniport system (Neubauer et al., 1994). In lactobacilli utilizing maltose through a MFS permease, a genetic association of the mal cluster with a gene encoding an aldose 1-epimerase (EC 5.1.3.3) can be found (Fig. 3.6). This enzyme is involved in the anomeric conversion of d-glucose between the α and β forms, which possibly speeds up the entry of glucose into the glycolytic pathway via its phosphorylation by glucokinase. Strains of L. plantarum are remarkable by the fact that they carry two mapA and pgm genes. One couple is linked to an ABC transporter (MalEFGK 2 ) and α-glucosidases (mal cluster 1), whereas the other forms a cluster (mal cluster 2) with a MFS permease (Fig. 3.6). Therefore, strains of this species possess the capacity to use maltose and maltodextrins by using two separated sets of genes. Unlike the rest of lactobacilli, the L. plantarum mal cluster 1 contains two transcriptional regulators (genes Lp_0172 and Lp_0173) with homology to MalR (Muscariello et al., 2011). Expression of malE from this cluster is induced by maltose and repressed by glucose via CcpA. Mutational analysis suggested that only the product of Lp_0173 participated in malE regulation. A mutant in this gene showed a glucose-insensitive expression of malE together with a lack of induction by maltose (Muscariello et al., 2011). Remarkably, an in silico approach for the study of LacI-GalR transcriptional regulators in L. plantarum WCFS1 identified five operons putatively controlled by the products of Lp_0172 and Lp_0173 which include the mal1 and mal2 clusters, an operon for β-glucosides utilization, an operon carrying the genes for teichoic acid synthesis tagB1 and tagB2 and an amino acid permease (Francke et al., 2008).
Transport studies with 14 C-maltose has revealed an unusually high K m for the L. casei maltose ABC transporter (around 0.3 mM; K m for ABC transporters and their substrates are usually in the µM range), which suggested that maltose is not the preferred substrate and points to maltodextrins as the natural oligosaccharides taken up by this transporter (Monedero et al., 2008). This notion is further substantiated by the fact that the three glycosidase enzymes encoded by the mal cluster are intracellular. Therefore, in this microorganism maltodextrins are preferentially metabolized over maltose, and they are hydrolysed in the cytoplasm to render maltose and glucose. This characteristic is probably shared by the rest of lactobacilli harbouring maltodextrin clusters with ABC transporters. In vitro studies on the binding capacity of the solutebinding component of the maltodextrin ABC transporter from L. casei (MalE) revealed that it is able to interact with maltotriose, maltotetraose, maltopentaose and with α, β and γ-cyclodextrins, which carry six, seven and eight glucose molecules, respectively, with K d values that were in the µM range, albeit showing a preference for linear maltodextrins over cyclodextrins (Homburg et al., 2017). However, contrarily to E. coli MalE, the solute-binding component of the L. casei maltose system does not interact with maltose. The fact that mutants in the L. casei malK gene are not able to ferment maltose (Monedero et al., 2008) suggests that other solute-binding proteins may be responsible to recognize and deliver maltose intracellularly via the MalFG permease component of the ABC system. Alternatively, it cannot be excluded that the MalK ATPase component of the transporter can be shared by an as yet unidentified and incomplete maltose-specific ABC system lacking a cognate ATPase unit (Homburg et al., 2017). Crystallographic data of L. casei MalE complexed with maltotriose, maltotetraose and cyclodextrins provided structural clues for its lack of interaction with maltose (Homburg et al., 2017). The globular carbohydrate-binding protein MalE consists of two N-and C-terminal domains which form a ligand-binding pocket situated between them, a characteristic shared by other MalE homologues. However, in L. casei MalE three aromatic residues from the C-terminal domain (W234, Y164 and W353) stack against a specific glucose moiety of the bound substrate and create three distinct subpockets, where the position of three glycosidic moieties is fixed with additional hydrogen bonds from the N-terminal MalE domain. Thus, the disaccharide maltose cannot adequately accommodate the three subpockets, preventing MalE to adopt the closed conformation that is achieved after interaction with linear and cyclic dextrins. This observation is also confirmed by the fact that, contrarily to maltotetraose, maltose does not stimulate ATPase activity of the MalEFGK 2 complex (Homburg et al., 2017).
Transcriptomic data under laboratory or natural fermentation conditions has shed some light on the regulation of the expression of glucans/starch utilizing enzymes and their concerted action during the degradation of these carbohydrates by lactobacilli. Experiments with L. acidophilus NCFM show that this strain induces preferentially the expression of PTS transport systems in the presence of prebiotic glucans such as cellobiose [β-d-glucopyranosyl -(1→4)-d-glucopyranoside], isomaltose, panose or gentiobiose [β-d-glucopyranosyl-(1→6)-d -glucopyranoside], whereas polydextrose (a synthetic glucose polymer consisting of a mixture of different α-glycosidic linkages) induces ABC transporters (Andersen et al., 2012). Transport via the PTS results in intracellular phosphorylated sugars that can be cleaved by different phosphoglucosidases. As expected, MapA was also induced by polydextrose. Of note, MalH, a isomaltose 6-phosphate hydrolase (GH4) encoded by LBA_1689 was induced by isomaltose, isomaltulose, panose and polydextrose. This enzyme participates in the formation of glucose 6-phosphate and glucose from isomaltose 6-phosphate but also glucose 6-phosphate plus fructose from isomaltulose internalized via a PTS encoded by the LBA_0606-LBA_0609 locus (Andersen et al., 2012). MalH has been found in lactobacilli associated with the gastrointestinal tract and it could be a good indicator of prebiotic activity of α-1,6 glucosides such as panose and polydextrose that can in addition be degraded by the activity of the product of LBA_0264, a glucan-α-1,6-glucosidase (GH13_31) which is induced by IMO (Andersen et al., 2012). Expression of genes involved in starch metabolism in L. plantarum A6 has been studied during a natural fermentation of pearl millet porridge (Humblot et al., 2014). This highly amylolytic strain expresses α-amylases (intracellular and extracellular), α-glucosidase, neopullulanase, amylopectin phosphorylase and MapA when growing in this natural substrate. The ability of this strain to liquefy the pearl millet gruel compared to other L. plantarum strains that do not grow in this substrate is attributable to the presence of the extracellular α-amylase (amyA). Metatranscriptomic analyses during the spontaneous fermentation by natural microbial consortia of different sourdoughs (wheat and spelt) with back-slopping for ten days have been carried out (Weckx et al., 2011). By using a DNA microarray carrying genes from several lactic acid bacteria, a high expression of glycolytic enzymes was observed during these fermentations. However, the expression of mapA and maltose/ maltodextrins ABC transporter encoding genes was low in both sourdough types and it corresponded mainly to the L. plantarum and Lactococcus lactis genes, respectively; even although the microarray also carried genes from L. sakei, L. curvatus, L. brevis, and L. fermentum (Weckx et al., 2011). These expression levels were related to the concentration of maltose throughout the back-slopping process. The sourdough microbiota was capable of degrading other carbohydrates important in sourdough (e.g. saccharose and fructose) and their metabolism could cause CCR of the utilization of other carbon sources mediated by CcpA, whose gene is highly expressed during sourdough fermentations (Weckx et al., 2011).

Metabolism of glycogen
Lactobacilli carry in their genomes gene clusters for the biosynthesis of the storage polysaccharide glycogen. In some lactobacilli that dwell in specific mucosal niches, such as the vagina, glycogen metabolism has been associated with their proliferation (Miller et al., 2016). Glycogen metabolic genes have been characterized in L. acidophilus and it is postulated that they play a role in the persistence of this bacterium in the gut (Goh and Klaenhammer, 2013). The bacterial glycogen synthesis starts by the synthesis of ADP-glucose from glucose 1-phosphate by GlgC or GlgD enzymes. Then the glycogen synthase (GlgA) transfers glucose from ADP-glucose to a chain of α-1,4-glucan, whereas GlgB is involved in the formation of the α-1,6 branching points. For its catabolism, glycogen phosphorylase (GlgP) catalyses the breakdown of α-1,4 linkages and GlgX participates in the debranching at the α-1,6 bonds in dextrins that cannot be further processed by GlgP. In L. acidophilus NCFM the glycogen cluster encompasses 11.7 kb and carry glgBCDAP together with two genes coding for a α-amylase and β-phosphoglucomutase (Goh and Klaenhammer, 2013). Similar genetic structures are found in approximately one third of the sequenced Lactobacillus, being mainly present in strains associated with the gastrointestinal tract of mammals and other animals. As an example, although the cluster is present in human intestinal isolates of L. bulgaricus and L. helveticus, it is not found in dairy isolates of these species (Goh and Klaenhammer, 2013). Expression of the L. acidophilus glg genes depends on the carbon source and the growth phase, showing maximal expression with raffinose [α-d-galactopyranosyl-(1→6)-α-d -glucopyranosyl-(1→2)-β-d-fructofuranoside] and repression by glucose. As the enzymes for the synthesis and degradation are produced in parallel, the regulation of intracellular glycogen levels depending on the carbon sources may rely on the carbon fluxes. Mutants impaired in glgA or glgB present a reduced growth on raffinose and a mutant in glgB and in the gene encoding the catabolic glycogen phosphorylase (glgP) grows slower in MRS medium (containing glucose) and are less resistant to simulated gastrointestinal conditions (Goh and Klaenhammer, 2013). This suggests that the accumulation of α-glucan polymers with sequestered glucose that cannot be further metabolized results in impaired growth. This highlights the need for co-ordinated glycogen synthesis and degradation for retrieving glucose from glycogen storage under normal and stress conditions. In vivo experiments in a germ-free mice model in which wild-type L. acidophilus and its glgA mutant were delivered to animals by intragastric gavage showed that the wild-type strain was able to compete and to displace the glgA mutant in monocolonized mice. This demonstrated the role of glycogen synthesis and utilization in lactobacilli in competitive fitness in the gut (Goh and Klaenhammer, 2014).

β-Glucan metabolic pathways in Lactobacillus
Despite the fact that lactobacilli are usually associated with the microbiota of plant decaying material, cellulases (endo-β-1,4-d-glucanases) have not been described for this genus and in general, for the lactic acid bacteria. Notwithstanding, the prebiotic activity of some β-glucans has been established but the metabolism of these polymers by lactobacilli has been rarely demonstrated. Very few examples of growth stimulation effects of β-glucans in lactobacilli are found in the literature, and they have mainly proved by using β-glucan hydrolysates (Dong et al., 2017). Important efforts have been made to use lactobacilli for the conversion of lignocellulosic material by applying saccharolytic processes prior fermentation or by the use of engineered strains expressing β-d-glucanases from other microbial sources (Moraïs et al., 2014;Okano et al., 2010;Overbeck et al., 2016). Similarly, lactobacilli expressing β-glucanases from other sources have been engineered for fermentation or health promoting effects (Liu et al., 2005;Wang et al., 2014a).

Metabolism of xylooligosaccharides
Structural characteristics of xylooligosaccharides Xylooligosaccharides (XOS) are plant-derived oligosaccharides with β-1,4 linkages between xylose (a pentose) monomers that can be decorated with residues of the pentose arabinose. These residues can be linked by α-1,2 or α-1,3 bonds to xylose molecules along the chain (arabinoxylan oligosaccharides) where one or two arabinose residues can be found per xylose. These polymers are abundant in plant cell walls and together with other heteropolysaccharides form the hemicellulose component in plants.

Xylooligosaccharides metabolic pathways in Lactobacillus
In vitro growth assays and studies of the microbiota of humans concluded that these polysaccharides possess a prebiotic effect that stimulates the growth of bifidobacteria (Childs et al., 2014;Lin et al., 2016). This is supported by the characterization of multiple ABC transporters for XOS and XOSdegrading enzymes in species of Bifidobacterium (Ejby et al., 2013). In vitro growth assays and human trials which explored changes in gut microbiota composition after XOS intake also pointed to XOS as prebiotic polysaccharides that stimulate growth of certain lactobacilli (Lin et al., 2016). Notwithstanding, the capacity to ferment XOS by lactobacilli seems to be limited (Ananieva et al., 2014). In accordance to this, the information about enzymes degrading XOS and arabinoxylans in lactobacilli is scarce. Two different enzymatic activities are needed to completely degrade arabinoxylans: arabinofuranosidase (liberating the arabinose residues that decorate the xylooligosaccharide backbone) and β-xylosidase (acting on the β-1,4 linkage between xylose molecules). Enzymes with this activity are classified into GH43 and GH51 glycosyl hydrolase families. L. brevis is thus far the only Lactobacillus species in which these activities have been studied (Michlmayr et al., 2013;Michlmayr et al., 2011). This species can be found in a wide variety of habitats including fermentations of hemicellulose-rich plant materials. Three GH43 β-xylosidases and two GH51 arabinofuranosidases have been found during the study of the genomes of several strains. The sequences of these enzymes show a high level of amino acid identity to enzymes from typical intestinal bacteria (e.g. bifidobacteria), suggesting events of horizontal gene transfer at the intestinal niche. The GH43 enzymes from L. brevis DSM 20054, annotated as β-xylosidases, have been thoroughly characterized (Michlmayr et al., 2013). The β-xylosidase encoded by LVIS_0375 (xynB1) gene exhibited activity towards β-1,4-xylobiose and β-1,4-xylotriose. LVIS_2285 (xynB2) showed low activity with p-nitrophenyl-β-d-xylopyranoside and no activity with β-1,4-xylooligosaccharides, whereas the β-xylosidase encoded by LVIS_1748 (abf3) exhibited activity for α-1,5-arabinooligosaccharides. XynB1 and XynB2 are 32% identical and are also present in strains of Lactobacillus buchneri, L. fermentum, Lactobacillus hilgardii, L. pentosus and L. reuteri. These species belong to the group of heterofermentative lactobacilli so that it has been postulated that the capacity to degrade XOS and arabinoxylans is restricted to this particular group within lactobacilli (Michlmayr et al., 2013). Unlike the arabinofuranosidases Abf1 and Abf2 characterized from L. brevis DSM 20054 (GH53) (Michlmayr et al., 2011), Abf3 (and XynB1 and XynB2) cannot release arabinose from arabinoxylans with a composition of 65% α-1,3-linked arabinose, 8% α-1,2-linked arabinose and 26% doubly substituted xylose (two arabinose linkages per xylose monomer) indicating that these enzymes do not act as arabinofuranosidases. Furthermore, Abf1 and Abf2 are selective for α-1,3-linked arabinose residues of monosubstituted xylose (Michlmayr et al., 2011).
The degradation of other hemicelluloses such as xyloglucan (a β-1,4 glucan backbone with xylose residues linked to glucose via α-1,6 bonds) by lactobacilli has received less attention. L. pentosus MD353, isolated from cucumber fermentation, carries a xylose operon (xylAB) involved in the metabolism of this pentose and encoding a xylose isomerase (xylA) and xylulose kinase (xylB) (Lokman et al., 1991). These enzymes convert cytoplasmic d-xylose to d-xylulose 5-phosphate, an intermediate of the pentose phosphate pathway. Adjacent to this operon two genes are present, xylPQ, which are also induced by xylose via the xylose repressor XylR. XylQ is a GH31 α-xylosidase which has been demonstrated to act on isoprimeverose [α-d-xylopyranosyl-(1→6)-d -glucopyranoside] and it is also able to liberate with very low efficiency small amounts of xylose from xyloglucan oligosaccharides with different linkage configurations (Chaillou et al., 1998). Isoprimeverose is usually released from xyloglucan by cellulolytic microorganisms producing endoglucanases and it can be taken up by L. pentosus via the product of xylP. This gene encodes a galactosidepentoside-hexuronide family transporter that catalyses the transport of isoprimeverose, but not xylose, through a proton motive force-driven process (Heuberger et al., 2001).

Metabolism of galactooligosaccharides
Structural characteristics of galactooligosaccharides β-Galacto-oligosaccharides (GOS) are nondigestible carbohydrates usually composed of lactose at the reducing end and one to ten galactose units linked by β-1,3, β-1,4 or β-1,6 bonds (Macfarlane et al., 2008). They can be acquired naturally through the diet from the degradation of galactan side chains of the rhamnogalacturonan I fraction of pectin ( Jones et al., 1997). In addition, they are also incorporated as prebiotics after their synthesis by the transgalactosylation activity of β-galactosidases on lactose, which acts both as donor and as acceptor of the galactose moiety (Vera et al., 2016). Analysis of some GOS mixtures revealed the presence of oligosaccharides with galactose at the reducing end instead of glucose, and they may also contain disaccharides different from lactose that are considered GOS as well. Indeed, commercial GOS are typically mixed-length galactosylated compounds with a DP ranging from 2 to 12 (Coulier et al., 2009). The type of linkage, mostly β-1,3, β-1,4 and/ or β-1,6, and to a lesser extent, β-1,2, is determined by the enzyme source. Commercial enzymes used for GOS synthesis belong to the CAZy glycosyl hydrolase family 2 (GH2) and they are obtained from Bifidobacterium bifidum and Bacillus circulans for GOS with β-1,3 and β-1,4 bonds, respectively, and from Kluyveromyces lactis and Aspergillus oryzae to obtain β-1,6-linked GOS (Rodriguez-Colinas et al., 2011). Most of the published studies use the term GOS when referring to β-GOS, but there are also GOS with α-configuration, which are produced by transgalactosylation reactions with α-galactosidases (Wang et al., 2014b). The oligosaccharides 3′-, 4′-and 6′-galactosyllactose have been found in colostrum and human milk, and they constitute the only oligosaccharides contained in common between human milk oligosaccharides (HMOs) and GOS.
GOS are metabolized by specific bacteria in the gastrointestinal tract, and they have been found to modulate the gut microbiota by stimulation of beneficial bacteria such as bifidobacteria and lactobacilli, and inhibition of pathogenic bacteria (Macfarlane et al., 2008;Rastall et al., 2005). The fermentation of GOS in the gastrointestinal tract leads to an increased production of specific shortchain fatty acids (SCFA), which are known for their health benefits including reduction of the risk of developing cancer and intestinal disorders (Cardelle-Cobas et al., 2009;Sangwan et al., 2011).

GOS metabolic pathways in Lactobacillus
Lactobacillus species in general can efficiently utilize GOS, although their utilization is a straindependent character and it may also vary depending on GOS DP (Endo et al., 2016;Thongaram et al., 2017). A study on GOS utilization by different species of Lactobacillus showed that 9 out of 10 tested species metabolized the galactosyllactose fraction (35%) of a GOS mixture to a different degree (Endo et al., 2016). All tested strains of L. delbrueckii, L. plantarum, L. fermentum, L. reuteri, L. johnsoni and L. acidophilus metabolized galactosyllactose whereas only some strains of L. rhamnosus, L. paracasei and L. sakei did it. As with fructans, the availability of genomic sequences allowed the determination of the genetic basis of GOS metabolism by lactobacilli. In this way, transcriptomic analyses of L. acidophilus NCFM with whole-genome DNA microarrays, revealed that GOS induce the lac-gal gene cluster, which encodes a galactoside-pentosehexuronide permease (LacS), two β-galactosidases belonging to the GH family 42 (LacA) and GH family 2 (LacLM), and enzymes of the Leloir Pathway (GalM, GalT, GalK and GalE) involved in the metabolism of galactose (Andersen et al., 2011). Inactivation of LacS impaired growth on lactose, lactitol and GOS (Andersen et al., 2011). Phylogenetic analysis showed that lacS is mainly found in human gut-associated Lactobacillus species, suggesting that transport and catabolism of those carbohydrates could be a significant energy source for lactobacilli in the gut (Andersen et al., 2011). Possibly, GOS are transported into the cells by the LacS permease, hydrolysed by the β-galactosidases LacA and LacLM into galactose and glucose, which would be directed to the Leloir Pathway and glycolysis, respectively. Interestingly, the lac-gal cluster is also induced by bile acids (Barrangou et al., 2006;Pfeiler et al., 2007), suggesting that bile may act as a location signal in the gut environment where GOS and related carbohydrates would be readily available.
Human milk also contains highly glycosylated proteins, including mucins, that have attached oligosaccharide moieties with structures that resemble those of the free HMOs (Liu and Newburg, 2013). Both HMOs and the glycan moieties of the proteins are synthesized by the same glycosyltransferases. N-glycans are linked to an asparagine residue through an GlcNAc, that is elongated by an additional GlcNAc residue through a β-1,4 linkage and three mannose residues. The GlcNAc residue linked to Asn can be modified via α-1,6-fucosylation and the mannose residues with other monosaccharides, including l-fucose and Sia, and it becomes a complex structure . O-glycans usually contain an N-acetylgalactosamine (GalNAc) linked to a serine or threonine residue. In the type-1 sugar chain found in mucins, the GalNAc is extended with Gal, linked via a β-1,3 bond, forming the disaccharide galacto-N-biose [GNB; β-d-galactopyranosyl-(1 →3)-N-acetyl-d-galactosamine] (Fig. 3.7). GNB is also present in glycosphingolipids and in bioactive sugar structures like the T-antigen disaccharide (Liu and Newburg, 2013;Moran et al., 2011).
In the last years, many studies have suggested that HMOs act as anti-adhesins against pathogens. HMOs are structurally similar to host receptors for pathogens since they are synthesized by the same glycosyltransferases that synthesize cell surface glycoproteins and glycolipids. As soluble receptor analogues, HMOs can act as decoys protecting infants against infections. Some HMOs inhibit the attachment of norovirus and bacterial pathogens such as Listeria monocytogenes and pathogenic E. coli strains (Newburg et al., 2005), Campylobacter jejuni (Ruiz-Palacios et al., 2003), Helicobacter pylori (Mysore et al., 1999) and parasites such as Entamoeba histolytica ( Jantscher-Krenn et al., 2012), explaining the fact that breast-fed infants are at lower risk to acquire E. histolytica infections than formulafed infants (Islam et al., 1988). Two disaccharides, α-l-fucosyl-(1→3)-N-acetyl-d-glucosamine (3FN) and α-l-fucosyl-(1→6)-N-acetyl-d-glucosamine (6FN) (Rodríguez-Díaz et al., 2013) that form part of the structure of many HMOs, either free or   (Becerra et al., 2015a). Recently, it has been shown that HMOs function as antimicrobial and antibiofilm agents against Streptococcus agalactiae, an invasive pathogen in both children and adults (Ackerman et al., 2017;Lin et al., 2017). Interestingly, specific neutral HMOs directly inhibit the growth of this bacterium and a mutant impaired in a putative glycosyltransferase is resistant to those HMOs (Lin et al., 2017).

Metabolic pathways for HMOs in Lactobacillus
Convincing evidence supports that HMOs favour the growth of beneficial bacteria present in the gastrointestinal tract of breastfed infants. HMOs were first identified as the prebiotic 'bifidus factor' described for human milk. Bifidobacteria constitute a considerable proportion of the intestinal microbiota of infants (Gomez-Llorente et al., 2013), and they are highly adapted to use HMOs as a carbon source (Garrido et al., 2013). Genome analyses have revealed that strains of Bifidobacterium longum subsp. infantis, Bifidobacterium longum subsp. longum, Bifidobacterium breve and B. bifidum encode a battery of enzymes involved in HMOs catabolism (Kwak et al., 2016;LoCascio et al., 2010). Unlike Bifidobacterium, species of the genera Lactobacillus, that are often isolated from breast-fed infant faeces (Albesharat et al., 2011;Martín et al., 2007;Rubio et al., 2014), usually showed a limited capacity for HMOs utilization. The only exception is represented by members of the L. casei/ paracasei/rhamnosus group, which contain several genes encoding enzymes involved in the hydrolysis of fucosyl-oligosaccharides (Rodríguez-Díaz et al., 2011) and in the metabolism of the type-1 (Bidart et al., 2014) and type-2 core structures from HMOs (Bidart et al. unpublished) (Fig. 3.8). Three α-l-fucosidases (AlfA, Alf B and AlfC) encoded in the L. casei BL23 genome have been characterized and they were able to hydrolyse in vitro fucosylated HMOs (Rodríguez-Díaz et al., 2011). All three enzymes are possibly intracellularly located as they lack secretion signals, suggesting that L. casei must transport the fucosylated substrates into the cytoplasm before their hydrolysis. This notion was demonstrated for the disaccharide 3FN that is transported into the cells by the mannose-class PTS encoded by the genes alfEFG, without being phosphorylated: l-fucose is a 6-deoxy-galactose and therefore lacks a phosphorylatable hydroxyl group at the carbon in the sixth position. These genes are divergently oriented from the gene cluster alfBR, encoding the α-l-fucosidase Alf B and the transcriptional repressor Alf R (Rodríguez-Díaz et al., 2012). Alf B digested the disaccharide within the cells into l-fucose and GlcNAc. The latter is metabolized by L. casei, whereas the l-fucose moiety is excreted to the medium (Fig. 3.8) because, in contrast to L. rhamnosus GG (Becerra et al., 2015b), L. casei lacks l-fucose catabolic genes. The release of l-fucose and Sia from the nonreducing ends is the first step to degrade the HMOs core structures. Bifidobacterium spp. and Bacteroides spp. generally are good consumers of fucosylated and sialylated HMOs as they usually possess fucosidase and sialidase activities. The last activity is not a common feature among lactobacilli, although L. delbrueckii ATCC7830 showed sialidase activity when cultured in the present of 6′-sialyllactose (Yu et al., 2013). A number of lactobacilli contain genes enabling the use of l-fucose or Sia moieties released from HMOs. L. rhamnosus GG can utilize l-fucose since it contains an operon encoding a specific catabolic pathway similar to that of E. coli (Becerra et al., 2015b). L. sakei 23K contains two gene clusters, nanTEAR and nanKMP involved in the catabolism of Neu5Ac (Anba- Mondoloni et al., 2013), and some strains of L. plantarum, L. salivarius (Almagro-Moreno and Boyd, 2009) and L. paracasei (Hammer et al., 2017) also contain genes for Sia metabolism. Extracellular sialidase and fucosidase activities have not been described in those species; therefore, the utilization of these carbohydrates as well as the glycan core structures by lactobacilli probably depends on their release from HMOs and mucin by other members of the intestinal microbiota. This is the case of L. casei, which has a complete machinery to metabolize HMOs and O-glycans core structures such as LNB, GNB (Bidart et al., 2014), LacNAc (Bidart et al., unpublished) and lacto-N-triose II (LNTII; β-N-acetyl-d-glucosamine-(1→3)-β-d-ga lactopyranosyl-(1→4)-d-glucopyranoside) (Bidart et al., 2016) (Fig. 3.8). LNB and GNB utilization relies on the gnb operon, which contains genes encoding a transcriptional repressor (gnbR), a galactosamine 6-phosphate isomerizing deaminase (gnbE), a GalNAc 6-phosphate deacetylase (gnbF), a phospho-β-galactosidase (gnbG) and four genes (gnbBCDA) encoding the EIIB, EIIC, EIID and EIIA components of a mannose-class PTS system (PTS Gnb ) (Bidart et al., 2014). LNB, GNB and also GalNAc are transported and phosphorylated by the PTS Gnb and then, both disaccharides are hydrolysed by the specific β-1,3-galactosidase GnbG (GH family 35) into galactose 6-phosphate and the corresponding N-acetylhexosamines (GlcNAc and GalNAc). Galactose 6-phosphate is metabolized through the tagatose 6-phosphate pathway, whereas GlcNAc and GalNAc would be phosphorylated by as yet unknown kinases before entering different catabolic routes (Fig. 3.8). GlcNAc 6-phosphate is converted by the NagA deacetylase to glucosamine 6-phosphate, which enters the glycolysis pathway via conversion to fructose 6-phosphate by the NagB deaminase. GalNAc 6-phosphate would be deacetylated and deaminated to tagatose 6-phosphate by the products of the genes gnbF and gnbE. Therefore, all gnb genes would participate in GNB and GalNAc metabolism while LNB utilization by L. casei would not require GnbE and GnbF activities. It is worth noting that gnb genes are highly induced by GNB and GalNAc which relieve repression by GnbR, whereas the presence of LNB barely induces the gnb operon (Bidart et al., 2014). According to this, the gnb operon would be primarily adapted to catabolize GNB and GalNAc. The coexistence of these sugars with LNB in environments as the gastrointestinal tract might account for the utilization of LNB by this pathway although this sugar would not induce the expression of gnb genes. The gnb gene cluster is conserved in the L. casei/paracasei/ Figure 3.8 HMOs catabolic pathways identified from Lactobacillus casei. LNB, lacto-N-biose; GNB, galacto-N-biose; LacNAc, N-acetyllactosamine; Lac, lactose; GalNAc, N-acetylgalactosamine; GalN, galactosamine; Gal, galactose; Glc, glucose, GlcNAc; N-acetylglucosamine; GlcN, glucosamine; Tag, tagatose; 3FN, fucosyl-α1,3-N-acetylglucosamine; Fuc, fucose; IICB Lac and IIA Lac , lactose-specific domains of the phosphoenolpyruvate: phosphotransferase system (PTS); IIC Gnb , IID Gnb , IIA Gnb and IIB Gnb , LNB/GNB/ GalNAc-specific domains of the PTS; IIC 3FN , IID 3FN and IIAB 3FN , 3FN 3FN 3FN-specific domains of the PTS; BnaG, beta-N-acetylglucosaminidase; LacG, phospho-β-galactosidase; GnbG, phospho-β-galactosidase; GnbF, N-acetylgalactosamine 6-phosphate deacetylase; GnbE, galactosamine 6-phosphate isomerase/deaminase; LacAB, galactose 6-phosphate isomerase; LacC, tagatose 6-phosphate kinase; NagA, N-acetylglucosamine 6-phosphate deacetylase; NagB, glucosamine 6-phosphate deaminase. rhamnosus/zeae group, and it has been shown that both GNB and LNB are fermented by several strains of L. casei, L. rhamnosus and L. zeae species (Bidart et al., 2017). Strains belonging to L. gasseri and L. johnsonii species are also consumers of LNB and GNB, although they do not have a gnb operon (Bidart et al., 2017). Therefore, at least another catabolic system for those disaccharides remains to be discovered in lactobacilli. Lactose metabolism in lactic acid bacteria has been widely studied due to the economic relevance of lactose fermentation in the dairy industry (Cavanagh et al., 2015;de Vos and Vaughan, 1994;Lapierre et al., 2002;Stefanovic et al., 2017). Lactose can be transported by lactose/galactose antiport permeases, proton symport permeases or through a PTS transporter (Alpert and Chassy, 1990;de Vos and Vaughan, 1994;Francl et al., 2012;Gosalbes et al., 1997;Leong-Morgenthaler et al., 1991). Recently, it has been shown that the lac operon from L. casei is also responsible of the utilization of LacNAc (Bidart et al. unpublished). This carbohydrate is transported and phosphorylated by the PTS Lac (Gosalbes et al., 2002;Gosalbes et al., 1997;Gosalbes et al., 1999), and then is hydrolysed by the phospho-β-galactosidase LacG (GH family 1) into galactose 6-phosphate and GlcNAc. In fact, the lac operon of L. casei showed higher induction levels in the presence of LacNAc than with lactose, suggesting that LacNAc may be the preferential substrate of this transporter. Indeed, this carbohydrate is present in the human gastrointestinal tract through all stages of life (Marionneau et al., 2001;Moran et al., 2011) whereas lactose would be only present during the lactating period since the introduction of dairy farming is a very recent event in the evolutionary history of humankind. Genome sequence analyses (http://www.ncbi. nlm.nih.gov/genomes) showed that many species belonging to the genus Lactobacillus contain genes encoding PTS transporters homologous to the PTS Lac from L. casei BL23, suggesting that lactosespecific PTS transporters are quite common among lactobacilli. Lactose is also the product resulting from the metabolism of LNTII by the action of the exoglycosidase β-N-acetylglucosaminidase BnaG (GH family 20) from L. casei (Bidart et al., 2016). Unlike the other glycosidases active on HMOs characterized in this species, BnaG is a cell wall -anchored extracellular protein. This enzyme shows high specificity for N-acetylhexosaminyl-β-1 ,3-linked sugars as it releases GlcNAc not only from LNTII but also from β-N-acetyl-d-glucosamin e-(1→3)-d-mannopyranoside, a disaccharide that forms part of glycoproteins (Garrido et al., 2012). As well, BnaG liberates GalNAc from β-N-acety l-d-galactosamine-(1→3)-d-galactopyranoside, which forms part of globotetraose, a glycan moiety of human glycosphingolipids present at cell surfaces (Schnaar et al., 2009). The oligosaccharide part of these lipids have recently been described as substrates for lacto-N-biosidases isolated from B. longum subsp. longum, which, in contrast to BnaG, are endoglycosidases and release GNB (Gotoh et al., 2015). The possession of cell wall-attached glycosidases may provide a competitive advantage by allowing cleavage and consumption of complexlinked sugars. In addition, there is also evidence suggesting that these enzymes might also modulate the activity of host glycoproteins and pathogenhost receptor interactions through modification of the surface-exposed host glycans structures (Garbe and Collin, 2012;Garrido et al., 2012;Kobata, 2013).
The presence of α-l-fucosidases and catabolic pathways for the utilization of LNB, GNB, LacNAc and LNTII in L. casei shows the capacity of this species for the exploitation of human milk and mucosa-associated glycans. This feature probably constitutes an adaptation of these bacteria to survive in the gastrointestinal tract of breastfed infants.

Concluding remarks
Lactobacilli play a major role in the production of many fermented foods and, in human and animal health as components of the microbial communities associated with different mucosal surfaces or as health-improving food supplements. These organisms rely on sugar utilization for growth so that knowing their sugar utilization pathways is a must to understand their role in microbial communities and their performance in food production. Due to their economic relevance, some species of lactobacilli have been extensively studied (e.g. specific strains of L. plantarum, L. casei or L. acidophilus) and the pathways of utilization of many monosaccharides and disaccharides have been elucidated. However, there are still great gaps in our knowledge of the utilization of complex glycans, especially host mucosal glycans, even in well-characterized strains of lactobacilli. Also, information is partial or simply lacking for many other species that inhabit mucosal and food/feed niches. Genomic sequencing in lactobacilli has revealed an enormous variety of putative glycan transport systems and glycosyl hydrolases belonging to different glycosyl hydrolase (GH) families. However, the function of most of these transporters and enzymes remains to be elucidated. Furthermore, most studies have been carried out using pure cultures in laboratory conditions. However, utilization of complex glycans in food matrixes or microbial communities is possibly a key factor for growth and survival in these environments. In the natural niches where lactobacilli dwell, complex ecological relationships are found and cross-feeding is established between different microbial groups, where sequential and cooperative degradation of complex glycans probably takes place. Unravelling these associations will require the use of different 'omic' technologies that include metagenomics, transcriptomics and metabolomics. This constitutes a future challenge in the study of carbohydrate metabolism in lactobacilli and its functional and ecological relevance.