A unique combination of glycoside hydrolases in Streptococcus suis specifically and sequentially acts on host-derived a Gal-epitope glycans

Infections by many bacterial pathogens rely on their ability to degrade host glycans by producing glycoside hydrolases (GHs). Here, we discovered a conserved multifunctional GH, SsGalNagA, containing a unique combination of two family 32 carbo-hydrate-binding modules (CBM), a GH16 domain and a GH20 domain, in the zoonotic pathogen Streptococcus suis 05ZYH33. Enzymatic assays revealed that the SsCBM-GH16 domain displays endo -( b 1,4)-galactosidase activity specifically toward the host-derived a Gal epitope Gal( a 1,3)Gal( b 1,4)Glc(NAc)-R, whereas the SsGH20 domain has a wide spectrum of exo - b - N acetylhexosaminidase activities, including exo -( b 1,3)- N -acetyl-glucosaminidase activity, and employs this activity to act in tandem with SsCBM-GH16 on the a Gal-epitope glycan.

Infections by many bacterial pathogens rely on their ability to degrade host glycans by producing glycoside hydrolases (GHs). Here, we discovered a conserved multifunctional GH, SsGal-NagA, containing a unique combination of two family 32 carbohydrate-binding modules (CBM), a GH16 domain and a GH20 domain, in the zoonotic pathogen Streptococcus suis 05ZYH33. Enzymatic assays revealed that the SsCBM-GH16 domain displays endo-(b1,4)-galactosidase activity specifically toward the host-derived aGal epitope Gal(a1,3)Gal(b1,4)Glc(NAc)-R, whereas the SsGH20 domain has a wide spectrum of exo-b-Nacetylhexosaminidase activities, including exo-(b1,3)-N-acetylglucosaminidase activity, and employs this activity to act in tandem with SsCBM-GH16 on the aGal-epitope glycan. Further, we found that the CBM32 domain adjacent to the SsGH16 domain is indispensable for SsGH16 catalytic activity. Surface plasmon resonance experiments uncovered that both CBM32 domains specifically bind to aGal-epitope glycan, and together they had a K D of 3.5 mM toward a pentasaccharide aGal-epitope glycan. Cell-binding and aGal epitope removal assays revealed that SsGalNagA efficiently binds to both swine erythrocytes and tracheal epithelial cells and removes the aGal epitope from these cells, suggesting that SsGalNagA functions in nutrient acquisition or alters host signaling in S. suis. Both binding and removal activities were blocked by an aGal-epitope glycan. SsGalNagA is the first enzyme reported to sequentially act on a glycan containing the aGal epitope. These findings shed detailed light on the evolution of GHs and an important hostpathogen interaction.
The glycosylation of macromolecules, such as proteins and lipids, to form glycoconjugates is one of the most abundant modifications of biomolecules in nature (1). In mammals, the glycoconjugates displayed on the cell surface form the complex carbohydrate-rich glycocalyx that surrounds cells and play key roles in important cellular functions, such as cell-cell interactions and cell signaling (2)(3)(4). The N-glycosylation of Asn residues in proteins is the most frequent form of protein glycosylation (4,5). Correspondingly, many commensal and pathogenic microbes have developed mechanisms to depolymerize N-glycans to take advantage of the carbohydrate-rich environment in the host and successfully colonize and/or infect the host (6)(7)(8)(9). Recently, a model was developed for N-glycan depolymerization in the human pathogen Streptococcus pneumoniae. In this model, an exo-a-sialidase (NanA) and an exo-(b1,4)-galactosidase (BgaA), together with an exo-b-N-acetylglucosaminidase (StrH), all of which are anchored to the cell surface of S. pneumoniae, were shown to have sequential activity on complex N-glycans, thereby contributing to nutrient acquisition and host immune response suppression (10,11).
The aGal-epitope glycan is absent from humans. Instead, humans naturally generate a large amount of anti-aGal epitope antibody, accounting for ;1% of total Igs (21,22) and causing hyperacute rejection of pig xenograft organs in humans (23). Therefore, the aGal epitope-containing glycans have mainly been investigated in the context of xenotransplantation (20). Under these circumstances, endo-b-gal C (EndoGalC), characterized from Clostridium perfringens, and its potential application in removing the aGal epitope for xenotransplantation in humans were studied (24,25). Recently, the host-derived aGal epitope was identified as a receptor for bovine norovirus New-bury2, revealing an expanded role of the aGal-epitope glycan in This article contains supporting information. * For correspondence: Yueling Zhang, zhangyueling@caas.cn or zhang.yl@ foxmail.com; Siguo Liu, liusiguo@caas.cn.
host-pathogen interactions (26). However, its potential as a target for bacterial pathogens has not been elucidated (27). Streptococcus suis is a major swine bacterial pathogen that is responsible for severe economic losses in the porcine industry and represents a significant threat to human health (28,29). The natural habitat of S. suis is the upper respiratory tract of pigs. Healthy pigs asymptomatically colonized with S. suis form a reservoir for this disease and play a major role in its epidemiology (29). The successful colonization of S. suis in upper respiratory tracts suggested that this microorganism can utilize the glycans enriched in these niches. Therefore, in this study, we identified a multifunctional glycoside hydrolase, SsGalNagA, from S. suis. This enzyme possesses a unique combination of activities to interact with and cleave the aGal epitope, which has been proven to be abundantly present on swine tracheal epithelial cells, potentially playing roles in upper respiratory tract colonization of S. suis. The results presented here represent the first report of the roles of the aGal epitope in the interactions of host and bacterial pathogens.

Results
S. suis does not have the same complex N-glycan degradation enzymatic system as S. pneumoniae Previously, a complex N-glycan degradation enzymatic system, containing NanA, BgaA, and StrH, was revealed in S. pneumoniae (10,11). To examine whether this system was present in a related bacterium, the zoonotic streptococcus S. suis, homologs for these three proteins were searched among S. suis genomes that were sequenced. However, no homolog was found for NanA and BgaA, whereas one homologous protein was found for StrH, in S. suis strain 05ZYH33, which was annotated as an N-acetyl-b-hexosaminidase (NCBI accession no. ABP89596; gene locus tag SSU05_0630). We named this protein SsGalNagA on the basis of the observations that we report in this study, representing an enzyme with both galactosidase and N-acetylglucosaminidase activity. The absence of NanA and BgaA homologs indicated that S. suis does not have the same complex N-glycan degradation enzymatic systems as S. pneumoniae.
SsGalNagA contains a unique and novel combination of GH16 and GH20 catalytic domains that is conserved in S. suis As shown in Fig. 1A, the 1419-residue SsGalNagA is composed of six domains: sequentially, a 32-residue N-terminal signal peptide (SP), a tandem repeat of two family 32 carbohydrate-binding modules (CBM32), two tandem glycosidase hydrolase (GH) catalytic domains, namely, family GH16 and GH20 domains, followed by a C-terminal canonical LPxTG motif (CW) (30). The presence of the LPxTG motif suggested that SsGalNagA is also a cell wall-anchored protein, similar to NanA, BgaA, and StrH in S. pneumoniae.
No homologous proteins possessing both GH16 and GH20 catalytic domains have been revealed from a species other than S. suis, indicating that the combination of GH16 and GH20 catalytic domains in one protein is unique for SsGalNagA and novel for GH. To assess whether the combination of GH16 and GH20 is conserved in S. suis, the 43 completely sequenced S. suis genomes were searched using full-length SsGalNagA as a probe. The search revealed that full-length SsGalNagA is highly conserved in S. suis from different serotypes, with 99-100% full-length coverage and 82-100% identities on the amino acid sequence level (Table S1). In summary, SsGalNagA is a multidomain protein containing a unique combination of GH16 and GH20 catalytic domains. This GH domain combination is novel for GH and conserved in S. suis.

SsGH20 has exo-(b1,2)-N-acetylglucosaminidase activity
To explore the biological function of the unique domain combination of SsGalNagA, full-length SsGalNagA and its domain truncations were cloned as summarized in Fig. 1B. The proteins of the domain variant were expressed and purified (Fig. S1).
As SsGH20 showed high similarity to GH20A and GH20B of StrH, it was hypothesized that SsGH20 had the same activities as GH20A and GH20B, with catalytic activity toward 4-nitrophenyl N-acetyl-b-D-glucosaminide (pNp-NAG), and, further, would release (b1,2)-linked GlcNAc (GlcNAc) from the a1,3 or a1,6 arm of complex N-linked glycans. We first measured the activity of SsGH20 toward pNp-NAG. Purified SsGH20 demonstrated hydrolysis of pNp-NAG with a K m of 0.69 mM, a k cat of 2.95 s 21 , and a k cat /K m of 4.28 s 21 mM 21 ( Fig. 2A).
To determine the activity of SsGH20 on (b1,2)-linked GlcNAc from the a1,3 or a1,6 arm of a complex biantennary N-linked glycan, the deglycosylation of transferrin, which possesses a typical complex N-linked glycan (33), was performed. As shown in Fig. 2B, SsGH20 was unable to remove terminal (a2,6)-linked sialic acid (Sia) or expose additional mannose (Man) from complex N-linked glycan (Fig. 2B, lane 2), indicating that it could not release (b1,2)-linked GlcNAc through an endo-type activity. However, after sequential removal of terminal Sia and the second galactose (Gal) by sialidase and galactosidase, SsGH20 catalyzed the release of the exposed terminal (b1,2)-linked GlcNAc through an exo-type activity, exposing much more terminal Man recognized by the lectin Galanthus nivalis agglutinin (GNA) (Fig. 2B, lane 5). The results confirmed that SsGH20 has exo-(b1,2)-acetylglucosaminidase activity, as GH20A and GH20B of intact StrH do.
SsCBM-GH16 acts on the aGal epitope on erythrocytes from pigs, rabbits, and cows but not on the human blood type B epitope SsCBM-GH16 showed high similarity to EndoGalC (Fig. 1A), which cleaves the (b1,4)-linkage between Gal(b1,4)GlcNAc in the aGal-epitope glycan to release Gal(a1,3)Gal. To detect whether SsCBM-GH16 showed the same activity as EndoGalC, based on the facts that the aGal epitope is abundant on erythrocytes of nonprimate mammals and human serum contains a large amount of anti-aGal epitope antibodies, a simple hemagglutination test was developed in this study. As shown in Fig.  3A, erythrocytes from pigs, cows, and rabbits agglutinated after the addition of human serum (Fig. 3A, lane 2) confirming the existence of aGal epitopes on these erythrocytes. In contrast, mouse erythrocytes did not show clear hemagglutination after the addition of human serum, indicating there was little or no aGal epitope on mouse erythrocytes.
These results were extended using flow cytometric analysis (FCA) with detection by FITC-conjugated Bandeiraea simplicifolia isolectin B4 (FITC-BSI-B4), which specifically binds to the terminal aGal residue. The results showed that the aGal epitope was indeed present on the erythrocytes from pigs, cows, and rabbits but absent from the erythrocytes from mice (dotted line in Fig. 3B). Further, after treatment with SsCBM-GH16, no additional hemagglutination was observed for these same erythrocytes after the addition of human serum (Fig. 3A, lane 4). FCA additionally showed that the aGal epitopes were removed by SsCBM-GH16 (Fig.  3B, solid line).
However, different degrees of activity have been observed by FCA. The removal of aGal epitopes from pig erythrocytes was more complete than that for cow and rabbit erythrocytes. Considering that Sia-containing aGal epitopes have been detected on bovine erythrocytes (18), the residual signal likely originates from FITC-BSI-B4 binding to this type of aGal epitope, which was not removed by SsCBM-GH16. Furthermore, compared with the human blood type O epitopes on erythrocytes, human blood type B epitopes were also recognized by BSI-B4 (Fig. 3B, dotted line) but not by SsCBM-GH16 (Fig. 3B, solid line). This suggested that SsCBM-GH16 could not catalyze cleavage of the fucosylated (a1,3)-galactobiose from human blood type B epitopes.
SsCBM-GH16 does not synergistically act with SsGH20 on complex N-glycans We considered a rationale for S. suis combining SsCBM-GH16 and SsGH20 in one protein. The activity of SsGH20 on complex N-glycans requires removal of the terminal Sia and the downstream Gal. S. pneumoniae achieved this by sequential cleavage of (a2,6)-linked Sia and (b1,4)-linked Gal by NanA and BgaA, respectively. S. suis does not possess the same complex N-glycan degradation system as S. pneumoniae (vide supra), because homologs of NanA and BgaA are absent from S. suis. Considering that SsCBM-GH16 shows endo-type activity on (b1,4)-linked Gal in Gal(a1,3)Gal(b1,4)GlcNAc, except that a different disaccharide Gal(a1,3)Gal is released, we considered whether SsCBM-GH16 could act on the (b1,4)-linked Gal in canonical complex N-glycans through an endo-type action to directly trim the disaccharide, Sia(a2,6)Gal, and subsequently form the substrate for SsGH20. However, incubation of transferrin with SsCBM-GH16 and SsGH20 did not remove additional terminal Sia from transferrin (Fig. 4E, lane 2) compared to that from transferrin incubated with SsGH20 alone, which does not remove terminal Sia (Fig. 4E, lane 1). Therefore, SsCBM-GH16 did not act on the (b1,4)-linked Gal in complex N-glycans through an endo-type activity. Similar assays were performed as above, except that sialidase was added. Here, SsCBM-GH16 and SsGH20 did not remove any more Gal (Fig. 4E, lane 4) than SsGH20 alone (Fig. 4E, lane 3). This result additionally indicated  Bacterial enzyme acts on host-derived aGal-epitope glycans that SsCBM-GH16 did not act on the (b1,4)-linked Gal in complex N-glycans through exo-type action. Thus, SsCBM-GH16 does not synergistically act on complex N-glycans with SsGH20 via its (b1,4)-galactosidase activity.
SsGH20 shows exo-(b1,3)-acetylglucosaminidase activity, which synergistically acts on the aGal-epitope glycan with SsCBM-GH16 Whereas the endo-(b1,4)-galactosidase activity of SsCBM-GH16 does not provide synergy with the activity of SsGH20 on complex N-glycans, we next focused on their synergistic actions on the aGal-epitope glycan. As SsCBM-GH16 cleaved the Gal(b1,4)GlcNAc linkage in the aGal-epitope glycan and exposed a terminal (b1,3)-linked GlcNAc, we considered whether SsGH20, which was shown to cleave (b1,2)-linked GlcNAc, could act on this (b1,3)-GlcNAc linkage, the activity of which has not been revealed for StrH. To test this concept, SsGH20 was incubated with Gap in the absence and presence of SsCBM-GH16. As shown in Fig. 4B, lane 3, in the absence of SsCBM-GH16, SsGH20 alone does not act on Gap. However, in the presence of SsCBM-GH16, SsGH20 could further . Activity assays for synergy and specificity of SsCBM-GH16 and SsGH20. A, the structure of the aGal-epitope glycan, Gap. The cleavage sites for SsCBM-GH16 and SsGH20 are indicated by arrows. B, TLC analysis of the products from Gap hydrolysis. Gal (G), lactose (L), and raffinose (R) were used to indicate the chromatographic mobilities of monosaccharide, disaccharide, and trisaccharide, respectively. C, D, and F, UPLC-HRMS analysis of the products from Gap hydrolyzed by no protein (C), SsCBM-GH16 (D), and SsCBM-GH16 plus SsGH20 (F). E, deglycosylation of the complex N-glycan on transferrin by SsCBM-GH16 and SsGH20 without (lanes 1 and 2) and with (lanes 3 and 4) sialidase. Terminal sialic acid (Sia) and galactose (Gal) of the complex N-glycan on transferrin were detected by lectins SNA and DSA, respectively, using the DIG glycan differentiation kit (Sigma). G and H, TLC analysis of the products from the indicated oligosaccharides hydrolyzed by SsGH20 (G) or SsCBM-GH16 (H). S2-S13 oligosaccharides and their structural compositions are listed in Table 1 hydrolyze Gap (Fig. 4B, lane 4). Compared with the products released by hydrolysis by SsCBM-GH16 alone (Fig. 4B, lane 2), the disappearance of trisaccharide, the increased levels of disaccharide, and the appearance of monosaccharides, after synergistic action of SsCBM-GH16 and SsGH20, strongly indicated that the trisaccharide GlcNAc(b1,3)Gal(b1,4)Glc produced by SsCBM-GH16 was further cleaved by SsGH20, resulting in the monosaccharide, GlcNAc, and the disaccharide, Gal(b1,4)Glc. This was confirmed by UPLC-HRMS analysis, which revealed that the hydrolysis products showed molecular masses of 244 Da and 365 Da (Fig. 4F), corresponding to that of GlcNAc and Gal(a1,3)Gal/Gal(b1,4)Glc, respectively. These results clearly revealed that SsGH20 acts on the terminal (b1,3)-linked GlcNAc. Thus, in addition to exo-(b1,2)-acetylglucosaminidase activity, SsGH20 also showed exo-(b1,3)-acetylglucosaminidase activity, and, with this activity, was able to synergistically act on the aGal-epitope glycan with SsCBM-GH16.
To further confirm that this synergistic action is combined in one protein, the full length of SsGalNagA ( Fig. 1B and Fig.  S1) was tested with Gap as the substrate. Not surprisingly, fulllength SsGalNagA hydrolyzed Gap to GlcNAc and Gal(a1,3) Gal/Gal(b1,4)Glc in the same manner as the combination of the individual domains, SsCBM-GH16 and SsGH20 (Fig. 4B, SsGH20 has a broad substrate spectrum, whereas SsCBM-GH16 specifically targets aGal-epitope glycan We show above that SsGH20 has both exo-(b1,2)-and exo-(b1,3)-acetylglucosaminidase activities, indirectly or directly, whereas SsCBM-GH16 displays endo-b-galactosidase activity releasing terminal (b1,4)-linked Gal(a1,3)Gal from aGal-epitope glycans. To better explore the potential substrates of SsGal-NagA, activities of SsGH20 and SsCBM-GH16 were screened on a variety of biologically relevant sugars using TLC. The results are shown in Fig. 4G and H and summarized in Table 1.

The SsCBM32-2 domain is indispensable for SsGH16 catalytic activity
SsCBM-GH16 and its homolog contain both the tandem CBM32 domains (designated SsCBM32-1 and SsCBM32-2; Fig.  1B) and the GH16 domain. In addition, it was reported that more than 75% of enzymes from subfamily GH16_8 were linked to CBM32 (32). Therefore, we examined the contributions of the CBM32 domains to the catalytic activity of SsGH16. Three truncated proteins, viz., SsGH16, SsGH16-2, and SsGH16-3, were prepared, as shown in Fig. 1B and Fig. S1. Their activities on both of the SsCBM-GH16 substrates, S6/Gap and Gal(a1,3)Gal(b1,4) Glc (S7), were examined. As shown in Fig. 5, A and B, neither SsGH16, with no CBM32 domain, nor SsGH16-2, with a partial domain of SsCBM32-2, was active on the substrates. However, SsGH16-3, which contains the SsGH16 domain and the complete SsCBM32-2 domain, recapitulated the activities on both substrates. These findings indicated that SsCBM32-2, the CBM32 domain adjacent to SsGH16, is indispensable for the catalytic activity of the SsGH16 catalytic domain.
In summary, SsCBM32-1, SsCBM32-2, and SsCBM bind aGal-epitope glycan with high specificity, and the terminal Gal (a1,3)Gal(b1,4)GlcNAc trisaccharide structure of aGal-epitope glycan plays an important role in the specific binding. With this specificity, it is unsurprising that the proteins did not bind human blood type A epitope (S12).
The binding kinetics of SsCBM to the pentasaccharide Gap was measured by SPR and are the following: k on , 0.6746 6 0.024 M 21 s 21 ; k off , (2.334 6 0.008) 3 10 23 s 21 . The two CBM32 together interact with Gap with a K D of 3.46 mM.

SsGalNagA binds and releases aGal epitopes on swine erythrocytes
Further elucidation of the function of SsGalNagA in the removal of aGal epitopes on erythrocytes in the natural host of S. suis will assist in understanding the mechanism of diseases, such as septicemia, caused by S. suis. The aGal epitope removal and binding activities of the full-length SsGalNagA and domain variants were detected by FCA, using FITC-BSI-B4 and corresponding antibodies, respectively. As expected, neither SsCBM nor the exo-acting SsGH20 shows aGal epitope removal activity, whereas both SsCBM-GH16 and full-length SsGal-NagA efficiently removed the aGal epitope from swine erythrocytes (Fig. 7A). Binding assays revealed that SsGH20 did not interact with swine erythrocytes, which was not unexpected. Surprisingly, among SsCBM, SsCBM-GH16, and SsGalNagA, all of which contain the SsCBM domain and are predicted to bind to the aGal epitope, only SsCBM binds to swine erythrocytes. Neither SsCBM-GH16 nor SsGalNagA showed any such binding activity (Fig. 7B).
Considering the efficient aGal-epitope cleavage activities of SsCBM-GH16 and SsGalNagA (Fig. 4B and 7A), we predicted that the cleavage of the aGal epitope would attenuate the binding target of SsCBM. To address this issue, mutations at the catalytic sites of SsGH16 were performed. We hypothesized, by structural simulations and alignments with GH16 laminarinase from Thermotoga maritima MSB8 (PDB entry 3b01) (Fig. 7C), that E561 and E566 are essential for the activity of SsGH16. Mutations of both E561 and E566 to alanines in SsCBM-GH16 and SsGalNagA were performed, resulting in the variants SsCBM-GH16m and SsGalNagAm, respectively ( Fig. 1B and  Fig. S1). The loss of catalytic activities of SsCBM-GH16m and SsGalNagAm were confirmed both by Gap hydrolysis (Fig. 7D,  lanes 2 and 3) and aGal epitope removal from swine erythrocytes (Fig. 7E). Finally, as predicted, abundant binding of SsCBM-GH16m and SsGalNagAm to swine erythrocytes was observed after loss of the catalytic activity (Fig. 7F).
In summary, SsGalNagA binds to the aGal epitope on swine erythrocytes through SsCBM, after which the aGal epitope is cleaved by SsGH16, followed by b(1,3)-GlcNAc cleavage by SsGH20. The steps also occur with Gap as a substrate.

SsGalNagA binds to STEC and removes aGal epitopes
The natural niche of S. suis is the upper respiratory tract of swine. Swine tracheal epithelial cells (STEC), the main cell type at this location, provide the first contact point for colonization of S. suis. To assess whether SsGalNagA plays a role in the interaction of STEC and S. suis, the aGal epitope on STEC was analyzed by confocal laser scanning microscopy (CLSM) using FITC-BSI-B4. As shown in Fig. 8A, aGal epitopes were detected on STEC by FITC-BSI-B4. Incubation of STEC with either SsGalNagA or SsCBM-GH16 significantly reduced the amount of aGal epitopes on STEC (Fig. 8A). The reduction of aGal epitopes was significantly dependent on the concentration of SsGalNagA added (Fig. 8B), clearly indicating that the reduction of aGal epitopes was due to the action of SsGalNagA. After inactivation of the catalytic activity of GH16, neither SsGalNagAm nor SsCBM-GH16m catalyzed the cleavage of the aGal epitope from STEC (Fig. 8, A and C), further confirming that the removal of the aGal epitope from STEC was attributable to the catalytic activity of SsGalNagA.
Binding of SsGalNagA and SsCBM-GH16 was also investigated by CLSM. As shown in Fig. 9, A and B, similar to swine erythrocytes, little binding was detected for catalytically active SsCBM-GH16 and SsGalNagA, whereas high levels of binding were detected for the inactive mutants, SsCBM-GH16m and SsGalNagAm. The results indicated that, similar to the action of aGal epitopes on erythrocytes, SsGalNagA binds to the aGal epitope and removes this terminal aGal from STEC.
The competitive effects of Gap on the removal of the aGal epitope from STEC and binding to the aGal epitope on STEC by SsGalNagA were examined. As shown in Fig. 8, A and D, significantly more residual aGal epitopes were revealed on STEC with the addition of 200 mM and 400 mM of Gap than that without the addition of Gap. These results indicate that the competition by Gap with the aGal epitope on STEC decreased the cleavage of the aGal epitope on STEC by SsGalNagA. Correspondingly, the binding of SsGalNagA to the aGal epitope on STEC was greatly attenuated upon the addition of Gap in a concentration-dependent manner (Fig. 9, A and C). This demonstrated that Gap competed with the aGal epitope for binding of SsGalNagA to STEC.

Discussion
In this study, a unique and conserved multifunctional SsGal-NagA has been discovered and characterized from the bacterial pathogen S. suis. Its singular combination of GH16 and GH20 Bacterial enzyme acts on host-derived aGal-epitope glycans catalytic domains sheds new light on the evolution of GH. The sugar components and linkages in host glycans are complicated in structure, and their degradation requires the combined activities of multiple GHs with different activities. Commensal and pathogenic bacteria have evolved different strategies to degrade the complex host glycans. One common strategy found in many bacteria is the formation of polysaccharide utilization loci or the carbohydrate-processing locus, which encodes an array of glycoside hydrolases (GH) required to degrade complex carbohydrates at a single genomic locus (37,38). Another strategy is to locate the required GHs at the same cellular location, although they are encoded by different gene loci. A representative example is the complex N-glycan degradation enzymes of S. pneumoniae, viz., NanA, BgaA, and StrH. All of these enzymes are equipped with the LPxTG motif, recognized by sortase A; consequently, all are likely anchored on the cell surface (11). Revealing the roles of S. suis SsGalNagA in this study represents a rare evolutionary strategy for GH by combining different GH catalytic domains needed into a single protein that targets different linkages in one glycan. Because the capacity of one protein is limited, this strategy may have developed for the glycans that contain relatively simple linkages.
In this study, a dependence of activity on the adjacent CBM32 domain has been revealed for a GH16 enzyme. This is unexpected, as GH16 enzymes are known to adopt a welldefined b-sandwich fold and have not been reported to rely on an additional ancillary module. However, this is not totally unexpected. First, until now, the activity and specificity toward aGal-epitope glycans were only reported for the members of the GH16_8 subfamily, indicative of a special substrate recognition pattern for this subfamily. Second, most of the enzymes from this subfamily are linked to CBM32, suggesting that an important role is performed by this CBM for the enzymes in this subfamily. Third, although rare, CBM has been reported to contribute to a conformational change and subsequent activesite formation for GH (35,36). As SsCBM32-2 specifically binds the aGal epitope, the catalytic activity of SsGH16 seemingly depends on the binding ability of SsCBM32-2. However, SsCBM32-2 was also indispensable for the hydrolysis of Gal (a1,3)Gal(b1,4)Glc (Fig. 5B), to which it did not show significant binding activity (Fig. 6). This suggested that the indispensability is not due to the binding ability of SsCBM32-2 but because of its contribution to another process, e.g. conformational change, substrate recognition, or active-site formation. Nevertheless, more evidence from activity and structural studies on SsGalNagA and its GH16_8 homologs are needed to further address this question.
Before now, the potential of the host-derived aGal epitope as a target for bacterial pathogens has not been revealed (27). In this study, it was shown that SsCBM-GH16 of SsGalNagA specifically binds and acts on aGal-epitope glycans. Whereas SsGH20 has a broad substrate spectrum and potentially acts on complex N-glycans, the lack of exo-sialidase and exo-(b1,4)-galactosidase partners in S. suis suggests that this is unlikely. Thus, the putative surface protein, SsGalNagA, from S. suis, specifically and sequentially cleaved terminal sugar residues from the aGal-epitope glycan. Furthermore, aGal epitopes are present on the surface of STEC, the first contact point for the colonization of S. suis, and can be removed by SsGalNagA. Based on these results, we propose that SsGalNagA from S. suis targets the aGal-epitope glycans on STEC in the pig upper respiratory tract, through which it acquires carbohydrate nutrients for the bacterium by degradation of aGal-epitope glycan, thereby facilitating the colonization of S. suis at this niche. Further, signal transduction functions have been reported for ceramide-based glycolipids (39), and the aGal epitope-depleted glycan structure has been identified as a regulator of transforming growth factor b receptor (TbR) (2). Therefore, SsGalNagA may also play a role in host cell signaling by altering the structure of this potential glycan regulator.
In conclusion, a multifunctional SsGalNagA was identified from S. suis 05ZYH33. This protein contains a unique GH16 and GH20 catalytic domain combination, enabling it to specifically and sequentially act on host-derived aGal-epitope glycans. SsGalNagA efficiently binds to swine erythrocytes and STEC and removes aGal epitopes, indicating that it plays roles in host colonization and host cell signaling in S. suis.

Bacterial strains, plasmids, cells, and growth conditions
Streptococcus suis strain 05ZYH33, serotype 2, which caused severe human infection in China in 2005 (40), was used to clone cDNAs of ssGalNagA and its truncation fragments. The S. suis strain was cultured in Todd-Hewitt broth (BD Bacto) supplemented with 1% yeast extract (BD Bacto) (THY) at 37°C and 5% CO 2 . Escherichia coli strains DH5a (Invitrogen) and BL21 (DE3) and plasmid pET22b (Novagen) were used for gene cloning and protein expression. The E. coli strains were grown at 37°C in Luria-Bertani medium, and 100 mg/ml ampicillin was added for transformant selection. Fresh normal erythrocytes were obtained by drawing heparinized blood from healthy human and other animal donors. Swine tracheal epithelial cells (STEC) were obtained from Shanghai Fusheng (China) and routinely maintained in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. The cells were passaged at confluence after dispersal with 0.25% trypsin-EDTA (Gibco) (41).

Sequence and structure analysis
Homologs of NanA (NCBI accession no. NP_359129), BgaA (NCBI accession no. NC_003098), and StrH (NCBI accession no. NP_357651) from S. pneumoniae strain R6 were identified from the genome of S. suis 05ZYH33 using the protein Basic Local Alignment Search Tool (BLAST) (42). The domain architecture was searched by CD-Search (43), and the annotation for each domain was obtained from the dbCAN-seq database (34). Sequences were aligned with ClustalX (44). Signal peptide predictions were performed using SignalP 5.0 (RRID: SCR_015644).
Structural neighbors were searched against the Protein Data Bank (PDB) using SWISS-MODEL (45). Three-dimensional molecular visualizations, structure superimpositions, and figure preparations were performed with PyMOL (46).

Cloning, protein expression, and purification
Gene fragments encoding full-length SsGalNagA and its truncated proteins, shown in Fig. 1B, were amplified from the genomic DNA of S. suis 05ZYH33 by PCR with the primers listed in Table S3. The PCR products were cloned into an NdeIand XhoI-linearized pET22b vector by homologous recombination using the Trelief SoSoo cloning kit Ver.2 (TsingKe, Beijing, China). The constructions were verified by DNA sequenc-ing and transformed into E. coli BL21(DE3) to express Cterminally His-tagged SsGalNagA and domains. His-tagged proteins were expressed and purified as described previously (47) (Fig. S1). The purified protein was finally maintained in 50 mM sodium phosphate, pH 7.4. The purity of proteins was assessed with 12% (w/v) SDS-PAGE. Protein concentrations were measured by the absorbance at 280 nm using the calculated extinction coefficients (RRID:SCR_018087) listed in Table S4. The final solutions of purified protein were stored in aliquots at 280°C for later use.

Site-directed mutagenesis
These steps were performed using the QuikChange sitedirected mutagenesis kit (Stratagene) with the plasmids encoding the WT ssCBM-GH16 or ssGalNagA gene as the template. The mutant proteins were expressed and purified similarly to the WT protein.

b-N-acetylglucosaminidase activity and kinetics
The b-N-acetylglucosaminidase activity and kinetic parameters of SsGH20 were measured using the b-N-acetylglucosaminidase assay kit (Sigma). In summary, the activity was detected using 4-nitrophenyl N-acetyl-b-D-glucosaminide (pNp-NAG; Sigma) as the substrate compared with a standard curve of 4nitrophenol. All of the assays were performed at 37°C in 0.09 M citrate buffer, pH 4.7, containing 64 nM SsGH20. For kinetic assays, substrate concentrations from 0.04 mM to 5 mM were used. The increase in absorption at 405 nm was monitored using an Enspire multimode plate reader (PerkinElmer). Kinetic parameters (V max and K m ) were extracted from these data by nonlinear fitting to the Michaelis-Menten equation using GraphPad Prism software.

Hemagglutination assay
Hemagglutination assays were carried out using fresh normal pig, cow, rabbit, and mouse erythrocytes with addition of normal human serum (NHS). Fresh normal erythrocytes were obtained by drawing heparinized blood from healthy human and other animal donors and then washed 33 in PBS containing 2% BSA (PBS-BSA). Erythrocyte solutions (10%) were prepared in PBS-BSA. NHS was isolated from healthy human whole blood. Blood was drawn into a gel vacuum tube, allowed to clot at 37°C for 0.5 h, and then centrifuged at 1000 3 g for 20 min at 4°C. The serum layer was collected, aliquoted, and frozen at 280°C until use. PBS-BSA (5 ml) or protein (5 mM) was added to 50 ml of 10% erythrocyte suspension and incubated at 37°C for 1 h. The erythrocytes were collected by centrifugation and washed twice with PBS-BSA and resuspended in 50 ml of PBS-BSA. Finally, 50 ml of 50% NHS was added and mixed. The solution was transferred to a round-bottom 96-well dish to observe the hemagglutination.

Removal of the aGal epitopes from erythrocytes
Removal of the aGal epitope from the erythrocytes prepared as described above was carried out, and the residual aGal epitope was labeled with FITC-BSI-B4 and detected using FCA. Erythrocyte suspensions (100 ml) were incubated with 10 ml of PBS-BSA or protein (40 mM) for 2 h. The erythrocytes were collected and washed, resuspended in 400 ml PBS-BSA containing 1 mg/ml of FITC-BSI-B4, and incubated in the dark for 2 h. The erythrocytes were collected, washed, and then resuspended in PBS-BSA. The terminal aGal epitope on the erythrocytes was assessed by FITC detection by FCA using a 488-nm laser. All incubations were performed at room temperature with gentle rotation with PIC added.

Hydrolysis of oligosaccharides and detection of the hydrolytic products
The hydrolysis of oligosaccharides was performed using the oligosaccharides listed in Table 1. The hydrolytic products were analyzed by TLC and UPLC-HRMS. Procedurally, in a 10ml reaction, 2.5 mM the indicated oligosaccharide was incubated with 4 mM protein in 50 mM phosphate buffer, pH 7.4, at 37°C for 1 h. After incubation, 2 ml of samples was spotted onto a GF254 silica gel TLC plate (Haiyang, China). The plate was developed in n-butanol-acetic acid-H 2 O (2:1:1), and sugar spots were visualized by spraying with acetone-diphenylamine-aniline-phosphoric acid reagent (100 ml-2 g-2 ml-10 ml) and heating. For MS, the resulting hydrolytic products from Gap were characterized by UPLC-HRMS. Chromatography was performed on a Shimidzu LC30A UPLC system. The mobile phase was run at 0.3 ml/min using 0.1% formic acid in H 2 O and CH 3 CN. MS was performed on a Triple TOF 4600 (AB SCIEX) operated in high-resolution mode in positive polarity.

SPR
Surface plasmon resonance (SPR) was carried out with a BIAcore T200 Biosensor (GE Healthcare) to measure the bind-ing specificities of individual SsCBM32-1 and SsCBM32-2 domains and SsCBM, containing both SsCBM32 domains, toward different oligosaccharides. The proteins were immobilized onto a CM5 sensor chip (GE Healthcare) using the amine coupling method. For immobilization, 40 mg/ml of SsCBM32-1, SsCBM32-2, or SsCBM in 10 mM NaOAc, pH 4.0, was used to attain 5781 response units (RU), 4143 RU, and 3869 RU, respectively. For each protein, a similarly treated flow cell, except that protein was not injected, was used as a control surface for the subtraction of the nonspecific binding. Binding was measured for each protein and the indicated oligosaccharides in HBS-P buffer (10 mM HEPES, 150 mM NaCl, 0.005% P20, pH 7.4). One millimolar each oligosaccharide was used. The oligosaccharide solutions were injected at a flow rate of 50 ml/min for 60 s. Postinjection phase dissociation was monitored in the running buffer for 60 s. The surface was regenerated between injections using 10 mM glycine-HCl, pH 2.0, at a flow rate of 50 ml/min for 30 s. The background was subtracted by passing the oligosaccharide solution over the control chip surface.
To measure the binding kinetic for SsCBM and Gap, the same immobilization strategy was used for SsCBM, except that 10,000 RU was reached. Concentrations from 31 mM to 500 mM Gap were used. Gap was injected at a flow rate of 50 ml/min with association and dissociation times of 120 and 180 s, respectively. The kinetic constants were calculated from the sensorgrams by nonlinear fitting of the association and dissociation curves according to a 1:1 Langmuir binding model using BIAevaluation software (GE Healthcare).

Protein binding to swine erythrocytes
Protein binding to erythrocytes was performed with fresh normal erythrocytes, and the binding proteins were detected by indirect immunofluorescence using FCA. Protein or PBS-BSA (10 ml) was added to 150 ml of 10% swine erythrocyte suspension and incubated for 1 h. The erythrocytes then were collected by centrifugation, washed with PBS-BSA, and incubated with corresponding rabbit polyclonal antibody. Anti-SsCBM was used for SsCBM, SsCBM-GH16, SsGalNagA, SsCBM-GH16m, and SsGalNagAm, and anti-SsGH20 was used for SsGH20. Next, erythrocytes were collected and washed with PBS-BSA, resuspended in 1 ml PBS-BSA with Alexa Fluor 647-goat-anti-rabbit IgG, and incubated in the dark for 1 h. The erythrocytes were then collected, washed, and resuspended in PBS-BSA. All incubations were performed at room temperature with gentle rotation with PIC added. The extent of the binding of proteins to erythrocytes was detected by the fluorescence signal of Alexa Fluor 647 on erythrocytes using FCA.

FCA
FCA was performed with a Cytomics TM FC500 cytometer (Beckman) using 488-nm and 633-nm lasers for the detection of FITC and Alexa Fluor 647, respectively. The cells were fixed with 4% paraformaldehyde, washed, and resuspended in PBS-BSA to yield a cell concentration of 1 3 10 6 cells/ml. Acquisition and analysis were performed by gating on side scatter and fluorescence. Cells in suspension were analyzed at a flow rate of 10 ml/min, and 50,000 events were recorded for analysis. Each histogram was analyzed using FlowJo_V10 software (BD Biosciences).

Removal of the aGal epitope from STEC
The removal of the terminal aGal epitope from STEC was detected by FITC-BSI-B4 using confocal laser scanning microscopy (CLSM). STEC were plated at 5 3 10 5 cells/disc in 15mm glass-bottom cell culture dishes 2 days prior to the assay. On the day of the assay, STEC were washed twice with Hank's balanced salt solution (HBSS), followed by incubation for 30 min with PBS-BSA (control) or the indicated concentrations of proteins in DMEM in a humidified 37°C, 5% CO 2 , incubator. After washing with HBSS, the cells were fixed with 4% paraformaldehyde, followed by sequential incubation with 1 mg/ml FITC-BSI-B4 for 2 h, 5 mg/ml Alexa Fluor 633-WGA for 10 min, and 2 mg/ml DAPI for 5 min. The incubations all were performed in HBSS at room temperature in the dark, with addition of PIC. The cells were then washed twice with HBSS between additions, and fluorescent cells were observed by CLSM. The fluorescence was observed by a 640-nm laser for Alexa Fluor 633, for analysis of the cell membrane, and a 405-nm laser for DAPI, for cellular DNA. The residual aGal epitopes were indicated by fluorescence observed by a 488-nm laser (for FITC).
To quantify the residual aGal epitope, CLSM images were analyzed by ZEN Lite microscope software (Zeiss) to obtain the mean fluorescence intensity (MFI). To correct the errors introduced to MFI by different amounts of cell membrane, the MFI of Alexa Fluor 633-WGA was used to normalize cell counts. The resulting MFI FITC,488 /MFI WGA,633 was designated the relative residual aGal epitope level and used to compare the number of residual aGal epitopes per unit of cell membrane. The removal of the aGal epitope from STEC was evaluated by reduction of the relative number of residual aGal epitopes compared with controls. The statistical significance was tested by the nonpaired t test using GraphPad.
To assess the competitive effects of free aGal epitope on its removal from STEC, the experiment was carried out as above, except that the indicated concentrations of Gap were added.

Protein binding to STEC
The binding of proteins to STEC was performed by immunofluorescence using CLSM. STEC was prepared as above 2 days prior to the assay. The cell suspensions were prepared and fixed as above. Incubation of the cells proceeded with anti-SsCBM, Alexa Fluor 647-goat anti-rabbit IgG, Alexa Fluor 488-WGA, and DAPI for 5 min in HBSS. The incubations all were performed at room temperature in the dark, with the addition of PIC. The cells were washed twice with HBSS between additions. Protein binding to the cells was observed by fluorescence with a 640-nm laser as in the previous section.
To quantify the protein bound, CLSM images were analyzed by the ZEN Lite microscope software (Zeiss) to obtain the MFI. Similarly, MFI Alexa, 633 /MFI WGA,488 was designated the relative amount of bound protein and used to compare the protein bound per unit of cell membrane. Protein binding levels were evaluated by comparing the relative amount of bound protein to the control. Statistical significance was tested by nonpaired t test using GraphPad.
To test the effect of free aGal epitope on binding, the experiment was carried out as above, except that prior to incubation with STEC, protein was incubated with the indicated concentrations of Gap for 10 min.

CLSM
CLSM was conducted with an LSM800 confocal laser scanning microscope with an Airyscan (Zeiss) microscope equipped with a 633 oil objective and 405-nm, 488-nm, and 640-nm lasers. Identical parameters were set for all the images taken. Image preparation and analysis were performed using the ZEN Lite microscope software (Zeiss).

Data availability
All data are contained within the manuscript.