Unique Microbial Catabolic Pathway for the Human Core N-Glycan Constituent Fucosyl-α-1,6-N-Acetylglucosamine-Asparagine

The gastrointestinal tract accommodates more than 1014 microorganisms that have an enormous impact on human health. The mechanisms enabling commensal bacteria and administered probiotics to colonize the gut remain largely unknown. The ability to utilize host-derived carbon and energy resources available at the mucosal surfaces may provide these bacteria with a competitive advantage in the gut. Here, we have identified in the commensal species Lactobacillus casei a novel metabolic pathway for the utilization of the glycoamino acid fucosyl-α-1,6-N-GlcNAc-Asn, which is present in the core-fucosylated N-glycoproteins from mammalians. These results give insight into the molecular interactions between the host and commensal/probiotic bacteria and may help to devise new strategies to restore gut microbiota homeostasis in diseases associated with dysbiotic microbiota.

M any investigations have recently highlighted the importance of the gut microbiota in the onset and progression of a number of human diseases, including gastrointestinal disorders (1,2), inflammatory diseases (3), respiratory tract infections (4), and allergies (5). The functional impact of commensal gut microorganisms depends on their ability to survive in the gastrointestinal tract, adhere to epithelial mucus, and obtain energy from nondigestible dietary substrates and host mucosal secretions (6). More than half of all proteins in nature have been estimated to be glycosylated through O-glycosidic or N-glycosidic bonds (7). O-Glycans are linked to a serine or threonine residue via an N-acetylgalactosamine, which is elongated by additional sugars (8). N-Glycosylation is a common modification of extracellular membrane proteins present at the gastrointestinal epithelium, the secreted proteins of human breast milk, and many dietary proteins (7,(9)(10)(11). N-Linked glycans are attached via the core N,N=diacetylchitobiose disaccharide (ChbNAc) (GlcNAc-␤1,4-GlcNAc) to an asparagine residue of proteins containing the Asn-Xxx-Ser/Thr (with Xxx being any amino acid except Pro) motif (12). In N-glycans from mammals, the inner GlcNAc moiety bound to Asn is often fucosylated through an ␣1,6 linkage, named core fucose. Protein N-glycosylation plays a crucial role in a variety of cellular processes, such as cell adhesion (13), immune pathway signaling (14), and bacterial recognition (15). Some intestinal microorganisms have the ability to process the carbohydrate moieties of N-glycosylated proteins (16) so that the type, abundance, and location of these glycans contribute to shaping the composition and distribution of the gut microbiota (17). Some bacterial pathogens possess endo-␤-N-acetylglucosaminidase enzymes that cleave the ␤-1,4 linkage of the core ChbNAc present in all N-glycoproteins, releasing the N-glycan moiety (18). The activities of these enzymes have been associated with the modification of the biological function of host defense glycoproteins, such as immunoglobulins and lactoferrin (19)(20)(21), and with the use of the glycans as nutrients (21), for which they are considered virulence factors (22). In commensal bacteria, the ability to remove N-glycans from glycoproteins has been described in a few Bifidobacterium species (23,24). Recently, the importance of core-fucosylated N-glycans from human milk in promoting the intestinal growth of Bifidobacterium and Lactobacillus species has been demonstrated in lactating infants from mothers carrying different alleles of the fucosyltransferase Fut8, responsible for core fucosylation (25). This provides the first in vivo evidence of the importance of this core structure in feeding intestinal commensals. However, there is little information about the fate of the fucosyl-␣-1,6-GlcNAc bound to proteins through the Asn residue (6=FN-Asn). This glycoamino acid possibly results from the combined action of endo-␤-N-acetylglucosaminidase enzymes and proteases on N-glycosylated proteins (18,23,24). The amide bond between the amino acid and the GlcNAc residue is hydrolyzed by two different enzymes, peptide-N(4)-(␤-N-acetylglucosaminyl)-Lasparaginases (glycopeptide N-glycosidase) (PNGase) (EC 3.5. 1.52) and N(4)-(␤-Nacetylglucosaminyl)-L-asparaginases (glycosylasparaginase) (EC 3.5.1.26). Both types of enzymes are produced as precursors that undergo intramolecular autoproteolysis to produce the mature active proteins (26,27), but PNGases require the presence of more than 2 amino acid residues in the substrate (28), whereas glycosylasparaginases act only in asparagine-oligosaccharides containing 1 amino acid (29). Currently, bacterial PNGases have been characterized only from the human pathogens Elizabethkingia meningoseptica and Elizabethkingia miricola (30,31) and from the soil bacterium Terriglobus roseus (32). In E. meningoseptica, a glycosylasparaginase has also been characterized (33,34).
Lactobacillus casei is a lactic acid bacterium able to survive in the gastrointestinal tract (35,36), which has been isolated from a wide variety of habitats, including feces of breastfed infants (37,38), and several strains are commonly used as probiotics in functional foods (39,40). Oligosaccharides present in human milk, such as lacto-N-biose and N-acetyllactosamine, and derived from mucins, such as galacto-N-biose, can support the growth of L. casei (41,42). This species is also able to catabolize lacto-N-triose (43) and fucosyl-␣-1,3-N-acetylglucosamine (44), which are abundant carbohydrates that form part of larger glycan structures from human gut mucosas and human milk.
To determine whether L. casei BL23 is able to use 6=FN-Asn, GlcNAc-Asn, N2F N-glycan, 6=FucGal, or 6=FucGlc as a carbon source, its growth profiles in MRS basal medium supplemented with each compound were analyzed. L. casei grew only in the presence of the glycoamino acid 6=FN-Asn (Fig. 3A). The analysis of the carbohydrate content of the growth medium showed that BL23 did not degrade any of the assayed glycans (Fig. 3C). However, the AlfR2-deficient strain (BL405) could degrade all the synthesized 6=-fucosyl oligosaccharides, as evidenced by the quantitative accumulation of L-fucose in the supernatant and the disappearance of the peaks corresponding to the fucosylated substrates (Fig. 3C). The presence of L-fucose in the culture supernatants was due to the fact that L. casei does not metabolize L-fucose and excretes it into the culture medium (44). Nevertheless, growth with these carbon sources was very poor compared to that with 6=FN-Asn (Fig. 3B). Interestingly, like the wild type (WT), the alfR2 mutant strain did not use GlcNAc-Asn (Fig. 3B). These results suggested that AlfR2 represses the expression of alf-2 genes and that the presence of 6=FN-Asn is required to relieve repression. Furthermore, the existence of an L-fucose moiety with an ␣1,6 linkage configuration is possibly necessary for the uptake of the tested fucosyl oligosaccharides, including 6=FN-Asn.
Transcription of the alf-2 gene cluster is induced by 6=FN-Asn and repressed by AlfR2. To find out whether the transcription of the alf-2 genes is regulated by the glycoamino acid 6=FN-Asn or its glycan moiety, reverse transcription-quantitative PCR (RT-qPCR) experiments were performed with RNA isolated from L. casei BL23 grown with 6=FN-Asn, 6=FN, GlcNAc, or glucose (Fig. 4). The results showed that the alf-2 operon is induced by the glycoamino acid 6=FN-Asn and not by the presence of 6=FN or GlcNAc, indicating that the glycoamino acid and not the glycan moiety is responsible for the induction of these genes and explaining the lack of growth of the wild type on N2F N-glycan, 6=FucGal, and 6=FucGlc. In order to obtain direct evidence for the regulation of AlfR2 on these genes, RNA was isolated from the alfR2 deletion strain BL405 cultured under the same conditions as those for the wild type (Fig. 4). Regardless of the substrate tested, the expression levels of alf-2 genes were higher than those in the wild type growing on glucose, indicating that AlfR2 indeed acts as a transcriptional repressor.
The AlfC ␣-L-fucosidase and the AlfH permease are necessary for the metabolism of 6=FN-Asn and 6=-fucosyl-oligosaccharides. To assess the role of alfC and alfH in the catabolism of 6=FN-Asn in L. casei, the growth patterns of mutant strains disrupted in alfC (BL415) and alfH (BL372) were analyzed in MRS basal medium supplemented with the fucosylated glycoamino acid ( Fig. 5A and B). The profiles showed that both mutants failed to grow on 6=FN-Asn. To further confirm that the glycoamino acid was not fermented, culture supernatant analyses were performed, and they showed that 6=FN-Asn remained in BL415 and BL372 supernatants, without the appearance of L-fucose (Fig. 5E). The amount of L-fucose (1.12 mM) present in both supernatants corresponded to the amount of L-fucose copurified with the synthesized 6=FN-Asn ( Fig. 2 and Fig. 5E). These results demonstrated that the ␣-L-fucosidase AlfC (Continued on next page) Glycoamino Acid Catabolism by Lactobacillus casei ® is involved in the metabolism of 6=FN-Asn and that the permease AlfH participates in its transport. To determine the role of AlfC and AlfH in the utilization of the rest of the fucosyl-glycans synthesized here (6=FN, 6=FucGlc, 6=FucGal, and N2F N-glycan), two double mutant strains, BL406 (alfR2 alfC) and BL407 (alfR2 alfH), were constructed. Both mutants showed growth patterns with 6=-fucosyl-glycans similar to that of the negative-control culture without an added carbohydrate ( Fig. 5C and D). In addition, the results showed a final culture optical density (OD) of strain BL406 or BL407 that was significantly lower than that of strain BL405 (alfR2) on these sugars (Table S2). These results indicated that the ␣-L-fucosidase AlfC and the permease AlfH are also involved in their metabolism and transport, respectively. 6=FN-Asn was also tested as a carbon source in the culture medium with these double mutants, confirming the requirement of a functional ␣-L-fucosidase and permease for its metabolism ( Fig. 5C and D).
6=FN-Asn catabolism requires AsdA, AsnA2, and SugK but not PepV. An asdA mutant strain (BL416) showed impaired growth in MRS basal medium supplemented with 6=FN-Asn (Fig. 6A). Curiously, the glycoamino acid was completely depleted from the culture supernatant (Fig. 5E); however, the mutant strain reached a lower optical density than the wild-type strain, suggesting that asdA is involved in 6=FN-Asn metabolism. A BLAST search using the deduced amino acid sequence of AsdA against the genomic sequence of L. casei BL23 did not reveal the presence of other AsdA paralogues. AsdA belongs to the Asp aminotransferase family (cd00609), which also includes a number of enzymes with decarboxylase or racemase activities. The BL23 genome encoded 11 hypothetical carboxylases (https://www.ncbi.nlm.nih.gov/genome/proteins/ 652?genome_assembly_idϭ300266); whether AsdA is a decarboxylase and whether any of these enzymes would complement its activity need to be investigated. A pepV mutant strain (BL417) showed a growth pattern similar to that of the wild-type strain (Fig. 6B), indicating that the hypothetical peptidase encoded by this gene is not essential for the metabolism of 6=FN-Asn. The two genes asnA2 and sugK, present downstream of pepV, were, however, necessary for the utilization of 6=FN-Asn by L.
casei. An asnA2 (BL418) mutant strain constructed here and a sugK (BL392) mutant characterized previously (41) failed to grow on 6=FN-Asn ( Fig. 6C and D). These results indicate that both the putative glycosylasparaginase AsnA2 and the putative sugar kinase SugK are required for the catabolism of 6=FN-Asn.
AsnA2 has glycosylasparaginase activity. The capacity to catabolize 6=FN-Asn by L. casei indicated that AsnA2 probably encodes a glycosylasparaginase that would cleave the linkage between GlcNAc and Asn. To confirm this hypothesis, AsnA2 was overexpressed as a His-tagged protein and purified (Fig. S2). The molecular weight of His-tagged AsnA2 was estimated to be 55.6 kDa using size exclusion chromatography, which does not coincide with the theoretical molecular mass (36.071 kDa), suggesting that the enzyme probably forms dimers. Interestingly, when the purified AsnA2 fraction was subjected to SDS-PAGE analysis, three bands were observed: a very faint band at about 38.0 kDa and two other bands with estimated molecular weights of 18.2 and 17.8 kDa. Mass spectrometry (MS) analysis showed that all three protein bands were derived from the glycosylasparaginase AsnA2 (data not shown). The 38-kDa band corresponds to the full-length protein, whereas the 18.2-and 17.8-kDa fragments correspond to the N-terminal and C-terminal domains, respectively. Therefore, AsnA2 is probably a zymogen that is processed during purification. A mechanism of intramo- lecular autoproteolysis has been previously described for glycosylasparaginases and PNGases (26,27). The self-processing of the precursor protein occurs at a Thr residue, and the two fragments form a noncovalent heterodimeric complex (26). A BLAST search with the amino acid sequence of L. casei AsnA2 evidenced 28% sequence identity to the glycosylasparaginase of E. meningoseptica (26). The sequence alignment between both proteins revealed that all residues involved in autoproteolytic processing, including the catalytic Thr residue (Thr-154 in AsnA2), are conserved. We analyzed the activity of the purified AsnA2 by measuring the release of GlcNAc from GlcNAc-Asn, which confirmed that AsnA2 is an N(4)-(␤-N-acetylglucosaminyl)-L-asparaginase (Fig. S2). This exhibited a specific activity of 20.91 mol mg protein Ϫ1 min Ϫ1 for GlcNAc-Asn. The activity of AsnA2 on the fucosylated glycoamino acid was also tested (Fig. S2). In contrast to the human glycosylasparaginase, which does not act on 6=FN-Asn (48), AsnA2 was able to degrade 6=FN-Asn, releasing 6=FN with a specific activity of 0.26 mol mg protein Ϫ1 min Ϫ1 . These results indicated that AsnA2 preferentially acts on GlcNAc-Asn over the fucosylated substrate 6=FN-Asn.
AsdA displays aspartate 4-decarboxylase activity, and SugK displays kinase activity. As described below, the Asp resulting from the activity of AsnA2 on GlcNAc-Asn might be the substrate for the AsdA enzyme. To test this hypothesis, AsdA was overexpressed with a His tag and purified. Recombinant AsdA showed decarboxylation activity on L-Asp, with a specific activity of 10.97 Ϯ 2.27 nmol mg protein Ϫ1 min Ϫ1 . This weak activity on L-Asp might indicate that this amino acid is not the preferred substrate for AsdA. Regarding SugK, the proposed pathways for 6=FN-Asn and 6=-fucosyl-glycans (Fig. 1B) involved the release of the monosaccharides GlcNAc, Glc, and Gal, which should be phosphorylated by specific kinases before entering glycolysis. In order to Glycoamino Acid Catabolism by Lactobacillus casei ® prove whether SugK showed kinase activity on these sugars, it was overexpressed as a His-tagged protein and purified. SugK displayed kinase activity on GlcNAc (1.44 mol mg protein Ϫ1 min Ϫ1 ) but could not phosphorylate Glc and Gal. Activity on N-acetylgalactosamine (GalNAc) was also assayed, as this N-acetylhexosamine is also very abundant on mucosa-associated glycans, and no activity on this sugar was detected.
Proposed pathway for 6=FN-Asn catabolism in L. casei. The genetic and biochemical evidence reported here allows the proposal of a catabolic pathway for 6=FN-Asn in L. casei (Fig. 1B). The glycoamino acid is internalized by the permease AlfH, which has broad specificity over a range of 6=-fucosylated substrates, and is subsequently defucosylated by the ␣-L-fucosidase AlfC, generating L-fucose and GlcNAc-Asn. The released L-fucose is excreted from the cells by an as-yet-undetermined mechanism, while GlcNAc-Asn is split by AsnA2 into Asp and 1-amino-GlcNAc, which is unstable, and it is nonenzymatically converted to GlcNAc and ammonium (49,50). The requirement of AlfC for 6=FN-Asn utilization and the lower affinity of AsnA2 for 6=FN-Asn than for GlcNAc-Asn support that the catabolism of 6=FN-Asn occurs through the consecutive actions of AlfC and AsnA2. GlcNAc produced by the action of AsnA2 on GlcNAc-Asn would then be the substrate for the sugar kinase SugK (Fig. 1). The other resulting product of AsnA2 activity, Asp, might be the substrate of AsdA, although the weak aspartate 4-decarboxylase activity could indicate that AsdA may play an as-yetundetermined role in this pathway. Regarding the function of pepV, which encodes a putative peptidase, no signal peptide was evidenced for it, suggesting that it is a cytoplasmic protein. The permease AlfH might also transport more complex substrates, like N-glycosylated peptides derived from the proteolysis of host-and food-derived proteins. Furthermore, the fact that AlfC is able to release L-fucose from core fucosylation of the Fc fragment from immunoglobulins (51) shows that this enzyme has the capacity to hydrolyze ␣-1,6 linkages in polypeptidic substrates and could act on core-fucosylated N-glycopeptides. In this case, after the release of L-fucose by AlfC, amino acids would be removed by PepV to liberate GlcNAc-Asn (Fig. 1B).

DISCUSSION
N-Glycosylated proteins are present at human mucosal surfaces and in breast milk (9,10,52), and therefore, they can be accessible to the gut microbiota. Most studies about energy sources for gut-beneficial microbes have been focused on carbohydrates present in the diet or added as prebiotics (53), whereas knowledge about the catabolism of host-derived carbon and nitrogen sources is scarce (16). We have demonstrated that L. casei is able to metabolize the glycoamino acid 6=FN-Asn, which is the core structure of the N-glycan sites of ␣1,6-fucosylated glycoproteins. The relevance of the metabolism of this core structure for intestinal microbial ecology was recently established in vivo in humans and in mouse models with reduced ␣1,6-fucosylation of core structures (25). In this work, we filled the existing gap relative to the metabolic pathways involved in the utilization of core-fucosylated N-glycopeptides by bacteria. The backbone of the L. casei 6=FN-Asn pathway consists of the MFS transporter AlfH; the ␣-L-fucosidase AlfC, which removes the ␣1,6-fucosyl residue; and the glycosylasparaginase AsnA2, which processes the resulting GlcNAc-Asn to 1-amino-GlcNAc and Asp (Fig. 1B). The generated 1-amino-GlcNAc is not metabolizable, but this compound is unstable at acidic pH, and it degrades into GlcNAc and ammonium (49). As lactobacilli do not maintain a constant internal pH, but it decreases as the external pH drops (54), intracellular conditions may allow the spontaneous degradation of 1-amino-GlcNAc.The resulting GlcNAc would then be the substrate for the sugar kinase SugK, allowing its channeling through glycolysis.
The genetic organization of the L. casei alf-2 gene cluster is well conserved in gene content and gene order across the L. casei-Lactobacillus paracasei-Lactobacillus rhamnosus phylogenetically related group of lactobacilli, excepting the absence of asdA in L. rhamnosus (Fig. 7). Homologues of alfC are present in only a few Lactobacillus species, and the closest homologues are harbored by some bifidobacteria isolated from Hyme-  Fig. S3A in the supplemental material). Interestingly, many lactobacilli carrying alfC have also been isolated from insects (58,59). Phylogenetic analyses of these genes show that they constitute a well-supported cluster together with the alfC gene of Vagococcus humatus. More distant homologues are mostly present in gut anaerobic bacteria (Fig. S3A). Homologues of L. casei alfH are mostly found in the same set of species, although this gene is also present in other bifidobacteria and in Lactobacillus kisonensis (Fig. S3B). The genetic association of these genes in these species and their close phylogenetic relationships suggest that they share an origin and an evolutionary history. Homologues of AsnA2 (Fig. S4), AsdA (Fig. S5), and PepV (Fig. S6) are present in numerous lactobacilli. Indeed, Lactobacillus gasseri ATCC 33323, which carries a gene cluster with asnA2, pepV, and alfR2 (Fig. 7), was recently shown to grow with 6=FN and N2F N-glycan (25). However, in the same work, L. casei ATCC 334 was reported to use L-fucose, which is contradictory to the fact that no fuc genes are present in its genome (60). In addition, L. casei ATCC 334 was also shown to metabolize 6=FN and N2F N-glycan (25), which, according to our results with L. casei BL23, can be possible only if the repressor alfR2 is inactivated.
In bifidobacteria, AsnA2 homologues are present only in Bifidobacterium actinocoloniiforme, Bifidobacterium asteroides, and Bifidobacterium xylocopae (Fig. S4), and as for their AlfC and AlfH counterparts, they are most closely related to Lactobacillus sequences. The limited presence in bifidobacteria and the phylogenetic clustering of the three bifidobacterial genes within Lactobacillus sequences suggest that they were transferred from Lactobacillus to Bifidobacterium. Genes encoding PepV homologues are absent in bifidobacteria, whereas asdA genes have been detected only in Bifidobacterium magnum and Bifidobacterium gallicum. The AsdA bifidobacterial sequences constitute a well-supported cluster with sequences from the Bacteroidetes and are not closely related to their Lactobacillus counterparts (Fig. S5), indicating that they had a different evolutionary origin. Interestingly, the closest relatives of L. casei pepV are genetically linked to asnA2 homologues (Fig. 7). This observation suggests a functional link between PepV and AsnA2.
Glycosylasparaginases (EC 3.5.1.26) are essential for the removal of the sugar moiety from the Asn in the GlcNAc-Asn structures derived from N-glycoproteins in humans (29,61). These enzymes cleave the ␤-aspartylglucosamine linkage, and they require both a free ␣-amino and ␣-carboxy group on the Asn substrate (29). In lactobacilli, an asnA2 homologue had been previously identified in Lactobacillus sakei as a gene induced during meat (sausage) fermentation, and a knockout mutant resulted in reduced growth on meat, but its activity was not ascertained (62). In contrast to the Elizabethkingia glycosylasparaginase, AsnA2 from L. casei characterized here lacks a signal peptide, strongly suggesting that it is an intracellular enzyme. As described above for glycosylasparaginases, AsnA2 suffers a self-processing proteolytic process to render an active enzyme (Fig. S2). The mature AsnA2 enzyme did not show activity on glycosylated proteins (Fig. S2D), but it showed in vitro activity on 6=-fucosylated and nonfucosylated GlcNAc-Asn. Curiously, the presence of the linked L-fucose blocks the hydrolysis of 6=FN-Asn by the human glycosylasparaginase (29). However, for the related PNGases, it has been demonstrated that the size of the carbohydrate moiety in the substrate has little effect on their activity (63). Therefore, AsnA2 would exhibit intermediate characteristics between both types of enzymes: ␣-1,6-linked L-fucose does not block hydrolysis, and the enzyme does not require a glycosylated peptide as the substrate.
We previously characterized the ␣-L-fucosidase AlfC from L. casei (45) and demonstrated that it displays high regiospecific transglycosylation activity that produces 6=FN disaccharide (47). This enzyme constitutes the only characterized bacterial ␣-Lfucosidase acting on ␣-1,6 linkages in core fucosylation structures (core fucosidase), and it was recently employed as a tool for IgG glycoengineering to obtain defucosylated immunoglobulins with enhanced antibody cell-mediated toxicity (51). Core fucosylation of N-glycopeptides and N-glycoproteins has also been attained using L. casei AlfC mutant enzymes (64). In addition to GlcNAc, we show here that AlfC is able to use GlcNAc-Asn, ChbNAc, galactose, and glucose as acceptor substrates in transfucosylation reactions. All the synthesized compounds (6=FN-Asn, N2F N-glycan, 6=FucGal, and 6=FucGlc) have exclusively ␣-1,6-fucosidic bonds, confirming the high linkage specificity of AlfC. However, the recognition of different acceptors indicates a relaxed substrate specificity.
Wild-type L. casei did not metabolize the 6=-fucosyl-glycans 6=FN, N2F N-glycan, 6=FucGal, and 6=FucGlc, probably due, as shown for 6=FN, to their inability to induce the alf-2 operon. This was confirmed by inactivating alfR2, which resulted in the constitutive expression of alf-2 genes and the subsequent catabolism of these oligosaccharides. Although alf-2 operon induction required the complete glycoamino acid 6=FN-Asn, the coexistence of this and other 6=-fucosyl-glycans in environments such as the gastrointestinal tract might allow their simultaneous utilization. Complex networks of crossfeeding between bacteria exist in the gastrointestinal niche, and these substrates are probably released into the gut by the concerted action of different microbial enzymes on glycoproteins and other glycocomplexes (21,23,24,62). The results presented here describe the first catabolic route for the utilization of 6=-fucosyl-related compounds in bacteria. They support previous works that assign to N-glycoproteins a role in nourishing beneficial bacteria in the gut. In addition, they show that commensals and pathogens share related mechanisms to take advantage of host molecules.
To determine the glycoamino acids and carbohydrates present in the supernatants from the Lactobacillus strain cultures, the bacterial cells were removed by centrifugation, and the cultures were analyzed with an ICS3000 chromatographic system (Dionex) using a CarboPac PA100 column with pulsed amperometric detection. A gradient of 10 mM to 100 mM NaOH was used for 16 min at a flow rate of 1 ml/min.
Nuclear magnetic resonance spectroscopy. Samples for nuclear magnetic resonance (NMR) spectroscopy were prepared as previously described (66). NMR spectra were also recorded as previously indicated (66), with some modifications. 1 H-13 C HSQC data were acquired with 360 transients over a spectral width of 10 ppm (for 1 H) and 160 to 220 ppm (for 13 C) and 128 points in the indirect dimension. Total correlation spectroscopy (TOCSY) data were acquired with 64 transients over a spectral width of 10 ppm in both dimensions and 128 points in the indirect dimension. NMR spectra were processed using the program MesReNoeva 8.1 (Mestrelab Research SL).
Bacterial strains and culture conditions. Lactobacillus strains (Table 1) were grown at 37°C under static conditions in MRS medium (Difco). Escherichia coli was used as a cloning host, and it was routinely grown in Luria-Bertani medium (Oxoid) with shaking at 37°C. The corresponding solid media were prepared by adding 1.8% agar. L. casei growth assays with different carbon sources were carried out in MRS basal medium as previously described (41). 6=FN-Asn, 6=FucGal, 6=FucGlc, 6=FN, GlcNAc-Asn, GlcNAc, or glucose was added to MRS basal medium at a concentration of 4 mM, and N2F N-glycan was added at 2 mM. Bacterial growth was determined in microtiter plates with a POLARstar Omega plate reader (BMG Labtech). At least three independent biological replicates for each growth curve were obtained, and a representative growth curve is shown. For each biological replicate comparing the wild type and a mutant strain, the same batch of MRS basal medium was used. The concentrations of glycoamino acids Glycoamino Acid Catabolism by Lactobacillus casei ® January/February 2020 Volume 11 Issue 1 e02804-19 mbio.asm.org 13 and carbohydrates in the culture supernatants at the end of fermentation were determined by HPLC as described above. E. coli strains were transformed by electroporation with a Gene Pulser apparatus (Bio-Rad), as recommended by the manufacturer. E. coli DH10B transformants were selected with ampicillin (100 g/ ml), and E. coli BE50 transformants were selected with ampicillin (100 g/ml) and kanamycin (50 g/ml). L. casei strains were transformed as described previously (67). L. casei transformants were selected with erythromycin (5 g/ml).
Construction of L. casei mutants in alf-2 genes. L. casei BL23 chromosomal DNA was isolated as previously described (67) and used as the template for PCRs, which were performed with the Expand high-fidelity PCR system (Roche). The alfR2 gene was amplified by PCR using the primer pair AspDecaFor/ XhoPeptVRev2 (see Table S3 in the supplemental material). The resulting 1,289-bp fragment was ligated into pRV300 (68) digested previously with EcoRV/KpnI and treated with the Klenow fragment of DNA polymerase I. The obtained plasmid was digested with SalI, ligated, and transformed. A clone with a 466-bp deletion in alfR2 (pRValfR2) was selected. DNA fragments containing parts of asdA, pepV, and asnA2 were obtained by PCR using the oligonucleotides pairs AsdFor/AsdRev, PepVFor/PepVRev, and AsnFor/AsnRev, respectively (Table S3), and cloned into pRV300 digested with EcoRV. The resulting plasmids, pRVasdA, pRVpepV, and pRVasnA, were cleaved at the unique HindIII, NcoI, and BclI restriction sites present in the asdA, pepV, and asnA2 coding regions, respectively, to introduce frameshifts into their corresponding coding sequences ( Table 1). The three digested plasmids were then treated with Klenow fragment, ligated, and transformed into E. coli DH10B. The resulting plasmids were transformed into L. casei BL23, and frameshifts were introduced into the corresponding genes by a double-recombination strategy (41) ( Table 1). To construct an alfC mutant, an internal DNA fragment of alfC was obtained by PCR using the oligonucleotides AlfCFor and AlfCRev. The PCR product was cloned into pRV300 digested with EcoRV. The resulting plasmid, pRValfC, was used to transform L. casei BL23, and single-crossover integrants were selected by resistance to erythromycin and confirmed by PCR analysis and DNA sequencing. One mutant was selected and named BL415. The same procedure was used to inactivate alfC in the mutant strain BL405 (alfR2), obtaining the alfR2 alfC double mutant (strain BL406). To construct the alfR2 alfH double mutant (strain BL407), the plasmid pRValfH was generated (Table 1). A DNA fragment containing part of alfH was obtained with the oligonucleotide pair FucPerFor/FucPerRev and cloned into pRV300 digested with EcoRV. The resulting plasmid, pRValfH, was cleaved at the unique BclI restriction site present in the alfH coding region, treated with Klenow fragment, ligated, and transformed. A construct in which a frameshift was introduced at the BclI site in alfH (pRValfH) was selected. pRValfH was transformed into the mutant strain BL405 (alfR2). A clone having a second recombination event was selected to obtain the alfR2 alfH double mutant (strain BL407). Sequence analysis. DNA sequencing was carried out by the Central Service of Research Support of the University of Valencia (Spain). M13 universal and reverse primers or custom primers hybridizing within the appropriate DNA fragments were used for sequencing. Sequence analyses were carried out with DNAMAN 4.03 for Windows (Lynnon BioSoft), and sequence similarities were analyzed with the BLAST program (69). Genomic context analysis was performed at the Microbial Genome Database for Comparative Analysis (MBGD) (http://mbgd.genome.ad.jp/) (70).
Expression and purification of AsnA2, AsdA, and SugK. The coding regions of asnA2, asdA, and sugK were amplified by PCR using the primer pairs AsnHindIIFor/AsnBamHIRev, AsdAHCt-F/AsdAHCt-R, and SugKBamH1For/SugKRev, respectively (Table S3). AsnA2 was cloned into the HindIII-BamHI sites and SugK was cloned into the BamHI-SmaI sites of the pQE80 vector and transformed into E. coli BE50 coexpressing the chaperones groES and groEL. pETasdA was constructed using the Gibson assembly kit (NEB) with the asdA PCR fragment and pET28a(ϩ) digested with NcoI and XhoI. pETasdA was transformed into E. coli BL21(DE3)/pLys. For purification, cells were grown to an OD at 600 nm (OD 600 ) of 0.8, 1 mM isopropyl-␤-D-thiogalactopyranoside was added, and incubation was continued at 25°C for 10 h. Bacterial cell extracts were loaded onto a HisTrap column (GE Healthcare), and His-tagged protein was purified according to the supplier's recommendations. The native molecular weight of the AsnA2 protein was estimated by size exclusion chromatography (HiPrep 16/60 Sephacryl S-300 HR column).
Mass spectrometry analysis. Protein bands from AsnA2 purification were excised from the gel, trypsinized, and analyzed with a 5800 matrix-assisted laser desorption ionization-tandem time of flight (MALDI-TOF/TOF) system (AB Sciex) at the proteomics service of the University of Valencia. MS and tandem MS (MS/MS) data were analyzed with the Mascot server.
AsnA2 enzyme activity. The activity of the purified AsnA2 enzyme was assayed at 37°C with GlcNAc-Asn and 6=FN-Asn as the substrates. The released GlcNAc and 6=FN, respectively, were quantified by analyzing the reaction mixtures by ionic chromatography (Dionex) as described above. Reactions using mixtures (20 l) containing 5 mM the substrate in 100 mM Tris-HCl buffer (pH 7.0) were initiated by adding 0.2 g of the enzyme. Protein deglycosylation assays were performed with reaction mixtures (15 l) containing 3.75 g of ovalbumin or lactoferrin in 100 mM Tris-HCl buffer (pH 7.0). The reactions were initiated with 0.2 g of the enzyme, and the mixtures were incubated at 37°C overnight.
AsdA enzyme activity. AsdA decarboxylase activity on L-Asp was measured as previously described, with some modifications (73). The resulting L-Ala was detected by using a coupled enzyme assay. Reactions using mixtures (100 l) containing 100 mM K-acetate buffer (pH 5.5), 0.2 mM pyridoxal 5-phosphate, 1 mM ␣-ketoglutarate, 0.4 mM NAD ϩ , 0.32 U of L-alanine dehydrogenase (Sigma), and 10 g of AsdA were initiated by the addition of 40 mM L-Asp. Reaction mixtures were incubated for 60 min at 37°C, and the production of NADH was monitored at 340 nm.
SugK enzyme activity. SugK phosphorylation activity on different sugars was determined by using a coupled enzyme assay as previously described (74). Reactions using mixtures (100 l) containing 100 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 5 mM ATP, 1 mM phosphoenolpyruvate, 0.2 mM NADH, 10 U of pyruvate kinase, 7 U of lactate dehydrogenase, and various substrates (GlcNAc, Glc, Gal, or GalNAc) at a final concentration of 1 mM were initiated by the addition of 0.75 g of SugK. Reaction mixtures were incubated for 15 min at 37°C, and NADH formation was monitored at 340 nm.
Phylogenetic analysis. Sequences of alfC, alfH, asnA2, asdA, and pepV homologues were retrieved from the microbial genome repository at the NCBI by BLAST (75) using their corresponding L. casei BL23 protein sequences as query sequences. For asnA2, only sequences from the Lactobacillaceae and Bifidobacteriaceae were included. Domain analysis was performed using the tools implemented at the NCBI BLAST site (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The resulting data sets were aligned with M-Coffee (76) at the T-Coffee server (http://tcoffee.crg.cat), with default settings. Positions of uncertain homology and gaps were removed using GBLOCKS (77) at the Gblocks server (http://molevol.cmima .csic.es/castresana/Gblocks_server.html), allowing smaller final blocks and less strict flanking positions. Redundant sequences were removed by using the EMBOSS suite Skipredundant tool (78) with a percent sequence identity redundancy threshold of 98%. The best-fit models of amino acid substitution and maximum likelihood trees were obtained using PhyML version 3.0 (79) at the PhyML server (http://www.atgc-montpellier.fr/phyml). Bootstrap support values were obtained from 1,000 pseudorandom replicates.
Statistical analysis. Student's t test was performed using Statgraphics Plus version 2.1 (Statistical Graphics Corp., USA), and it was used to detect statistically significant differences between the final OD values reached by L. casei BL405 (alfR2) cultures and the mutant strains BL406 (alfR2 alfC) and BL407 (alfR2 alfH). Statistical significance was accepted at a P value of Ͻ0.05.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.