Two complementary α-fucosidases from Streptococcus pneumoniae promote complete degradation of host-derived carbohydrate antigens

An important aspect of the interaction between the opportunistic bacterial pathogen Streptococcus pneumoniae and its human host is its ability to harvest host glycans. The pneumococcus can degrade a variety of complex glycans, including N- and O-linked glycans, glycosaminoglycans, and carbohydrate antigens, an ability that is tightly linked to the virulence of S. pneumoniae. Although S. pneumoniae is known to use a sophisticated enzyme machinery to attack the human glycome, how it copes with fucosylated glycans, which are primarily histo-blood group antigens, is largely unknown. Here, we identified two pneumococcal enzymes, SpGH29C and SpGH95C, that target α-(1→3/4) and α-(1→2) fucosidic linkages, respectively. X-ray crystallography studies combined with functional assays revealed that SpGH29C is specific for the LewisA and LewisX antigen motifs and that SpGH95C is specific for the H(O)-antigen motif. Together, these enzymes could defucosylate LewisY and LewisB antigens in a complementary fashion. In vitro reconstruction of glycan degradation cascades disclosed that the individual or combined activities of these enzymes expose the underlying glycan structure, promoting the complete deconstruction of a glycan that would otherwise be resistant to pneumococcal enzymes. These experiments expand our understanding of the extensive capacity of S. pneumoniae to process host glycans and the likely roles of α-fucosidases in this. Overall, given the importance of enzymes that initiate glycan breakdown in pneumococcal virulence, such as the neuraminidase NanA and the mannosidase SpGH92, we anticipate that the α-fucosidases identified here will be important factors in developing more refined models of the S. pneumoniae–host interaction.

The structural repertoire of glycans present in the human glycome is diverse, with the glycoconjugates that carry these glycans also being highly abundant. Approximately 1% of the human genome is dedicated to the synthesis and modification of glycans, and most human proteins are thought to be glyco-sylated (1,2). Common human glycans include N-and O-linked glycans, histo-blood group antigens, glycosaminoglycans, glycogen, and the glycan families attached to glycosphingolipids (3). These glycans, both secreted and conjugated, have numerous important and varied functions. These include cell-cell interactions and cellular signaling; glycans also influence folding, stability, and function of glycoproteins (4). Commensurate with the importance and abundance of the human glycome, many commensal and pathogenic organisms have evolved strategies to degrade, transport, and process human glycans (e.g. Refs. [5][6][7]. One bacterium that is particularly adept at harvesting human glycans is Streptococcus pneumoniae (8 -10). This commensal bacterium frequently inhabits the human nasopharynx and upper respiratory tract; however, it is also an important etiological agent of a number of serious and potentially life-threatening diseases such as pneumonia, bacteremia, and meningitis (11). Among respiratory pathogens, S. pneumoniae is unique in its capacity to degrade, transport, and metabolize a wide range of complex carbohydrates, most of which are host-derived (9,12). This ability to break down and transport human glycans has been identified as a key virulence mechanism within this bacterium (13)(14)(15), contributing to nutrient acquisition, the uncovering of host receptors required for adherence and invasion, and immune modulation through the deglycosylation of host glycoproteins (16 -18). The human respiratory tract is rich in functionally important glycoconjugates that bear a range of glycans, including complex, high-mannose, and hybrid N-linked glycans; core 1-and core 2-type O-linked glycans; and histo-blood group capping motifs (19). These glycan structures and those present on other glycoconjugates, such as glycosphingolipids, are also found disseminated throughout the human body and at sites of invasive pneumococcal disease, for example on the surface of erythrocytes and immune cells, and in the brain (20). Together, these glycans contain more than 20 different linkages among the monosaccharides N-acetylneuraminic acid (sialic acid), D-galactose, N-acetyl-D-glucosamine (GlcNAc), 3 N-acetyl-D-galactosamine (GalNAc), D-man-nose, D-glucose, and L-fucose. The S. pneumoniae genome encodes for more than 40 known or predicted proteins that break glycosidic bonds, the majority of which are glycoside hydrolases (GHs) (10). Many of these GHs have now been functionally characterized and found to cleave one or more of the above linkages as well as contribute directly to the virulence of this pathogen (10). Therefore, a comprehensive picture of human glycan degradation by S. pneumoniae is now emerging.
Fucose is attached to human glycans via ␣-(132), ␣-(133), ␣-(134), and ␣-(136) linkages, with the latter found as decorations on the core GlcNAc of N-glycans (33). Fucose residues attached via linkages other than ␣-(136) are typically found in the histo-blood group antigens, which comprise the A, B, and O antigens, and four Lewis antigens, Lewis A , Lewis B , Lewis X , and Lewis Y . ABO and Lewis antigens are commonly observed as capping motifs on the arms of N-and O-glycans, as well as glycosphingolipids, in a wide variety of tissues (19,34). As the majority of pneumococcal GHs are exoglycosidases, the presence of fucose as a capping residue on these glycans would likely necessitate the deployment of an enzyme, or enzymes, that can remove fucose residues, thereby allowing other pneumococcal GHs to access the main glycan.
We have recently identified a highly conserved carbohydrate-processing locus in S. pneumoniae (26). Located within this locus is SP_2146 (TIGR4 locus tag), a gene encoding for a putative ␣-fucosidase belonging to GH family 29. This gene (and its protein product, herein referred to as SpGH29 C ; super-script "C" for belonging to the core genome) has been identified as a putative virulence factor in multiple signature-tagged mutagenesis studies of pneumococcal disease and is a component of the core pneumococcal genome (13,14,35). A second putative ␣-fucosidase belonging to GH family 95 is also encoded by the core genome (locus tag SP_1654, herein referred to as SpGH95 C ). Like SpGH29 C , SpGH95 C has been identified as a putative virulence factor in multiple signaturetagged mutagenesis studies (13,37). SpGH95 C does not reside within an operon or functional locus; the only protein predicted to be functionally associated with SpGH95 C via STRING analysis with a score of 0.95 (38) is SpGH29 C . Given the classification of SpGH29 C and SpGH95 C into GH families 29 and 95, respectively (39), and their predicted functional association, we hypothesized that these two enzymes are ␣-fucosidases with complementary linkage specificities. Here, we show that SpGH29 C and SpGH95 C are indeed ␣-fucosidases with differing linkage specificities, that they are active against histo-blood group antigens, and that together they act as keystone enzymes to "uncap" fucosylated human glycans, enabling complete depolymerization by other enzymes. By recapitulating pneumococcal glycan degradation pathways in vitro, we also demonstrate the competence of S. pneumoniae to degrade a wide range of human glycans.

Pneumococcal ␣-fucosidases active on histo-blood group antigens
To test our hypothesis that SpGH29 C and SpGH95 C are ␣-fucosidases with complementary linkage specificities, we produced the proteins recombinantly in Escherichia coli. Following purification, neither enzyme exhibited activity against 4-nitrophenyl ␣-L-fucopyranoside with substrate concentrations in the mM range (data not shown); however, a screen against histoblood group antigens and other ␣-fucosylated glycans by TLC revealed that SpGH29 C and SpGH95 C have ␣-fucosidase activity (Table 1 and Fig. S1). SpGH29 C displayed activity against substrates containing ␣-(133)and ␣-(134)-linked fucose  Fig. S1 for TLC images. ϩ/Ϫ indicate presence/absence of activity.

Fucosylated glycan degradation by S. pneumoniae
units, including 3-fucosyllactose and the four Lewis antigens; however, it was unable to cleave fucose from substrates smaller than a trisaccharide. In contrast, SpGH95 C exhibited exclusive activity against substrates containing ␣-(132)-linked fucose units (namely 2-fucosyllactose, blood group H-antigens, Lewis Y , and Lewis B ), and it was able to cleave a disaccharide substrate. Activity on the H-disaccharide (Fuc-␣-(132)-Gal) but lack of activity on 4-nitrophenyl ␣-L-fucopyranoside suggests quite strict accommodation of the residue preceding the terminal fucose. Likewise, despite the presence of ␣-(132)linked fucose units, the blood group A-and B-antigens were resistant to SpGH95 C activity (Table 1 and Fig. S1). We followed up this initial activity screen by determining the kinetic parameters of SpGH29 C and SpGH95 C against relevant ␣-fucosylated substrates using an enzyme-coupled fucose detection assay (40) (see "Experimental procedures" for details). Both SpGH29 C and SpGH95 C exhibited a linear increase in initial velocity with increasing substrate concentration for a number of substrates; therefore, precise K m values could not be determined. However, k cat /K m values for each substrate-enzyme combination were determined (Table 1). SpGH29 C exhibited very similar k cat /K m values for all four Lewis antigens; therefore, it demonstrated no significant preference for glycan size (trisaccharide or tetrasaccharide) or fucose linkage (␣-(133) or ␣-(134)). Conversely, SpGH95 C exhibited k cat /K m values that varied by up to 17-fold among substrates depending on the size and configuration of the glycan. All of the substrates for SpGH95 C contained the same core H-motif. The H-disaccharide acted as a substrate for SpGH95 C with a k cat /K m of 10.0 Ϯ 0.2 min Ϫ1 mM Ϫ1 (ϮS.E.; Table 1). The linkage of a glucose or GlcNAc unit to the galactose on this H-disaccharide motif resulted in a 2-10-fold increase in k cat / K m . Specifically, addition of a GlcNAc residue via a ␤-(133) linkage (H-trisaccharide type I) resulted in an ϳ2-fold increase in k cat /K m , whereas addition of this same residue via a ␤-(134) linkage (H-trisaccharide type II) resulted in a Ͼ10-fold increase in catalytic efficiency. SpGH95 C also exhibited higher catalytic efficiency when the H-disaccharide was modified by the addition of a GalNAc-␤-(133)-Gal disaccharide to the galactose via a ␤-(133) linkage (H-tetrasaccharide type IV). Despite the fact that the Lewis B and Lewis Y tetrasaccharides contain the H-trisaccharide type I and II antigens, respectively, the catalytic efficiency of SpGH95 C against these substrates was similar to that observed with the H-disaccharide (Table 1).

Structural analysis of SpGH29 C
The activity of SpGH29 C reveals it to be of the "B" group of GH29 fucosidases, which are defined as having little/no activity on pNP-␣-L-fucopyranoside (where pNP is p-nitrophenyl) and specificity for terminal ␣-(1,3/4)-fucosidic linkages (40). Furthermore, SpGH29 C displays an absolute requirement for a more complex glycan substrate than a simple disaccharide, which is similar to the GH29 BiAfcB enzyme from Bifidobacterium longum subsp. infantis (41). We used X-ray crystallography to probe the molecular basis for the ability of SpGH29 C to recognize complex glycans and, specifically, accommodate substrates with both type I and type II core motifs (e.g. Lewis X versus Lewis A ). Initially, a single crystal of SpGH29 C was obtained, but subsequent trials failed to reproduce the crystals. This crystal yielded a good diffraction data set to a resolution of 1.72 Å, allowing the structure to be solved by molecular replacement.
The final refined structure comprised two molecules per asymmetric unit with each polypeptide chain unexpectedly terminating at residue 452 (of 559 expected residues). This C-terminally truncated form of the protein, which was presumably generated by degradation during the crystallization experiment, had an overall fold containing two domains that is typical of several GH29 enzymes (Fig. 1). The C-terminal domain is a ␤-sandwich domain made up of three and five antiparallel ␤-strands arranged in ␤-sheets (Fig. 1). The N-terminal (␣/␤) 8barrel domain houses the catalytic machinery, which on the basis of similarity to other GH29 enzymes can be identified as Asp-171 for the nucleophile and Glu-215 for the acid/base (Fig. 1).
To enable reproducible crystallization of SpGH29 C , we used the native structure to inform the generation of a shorter construct (amino acids 1-451; SpGH29 C T) into which we also introduced a D171N/E215Q double mutation to catalytically inactivate the enzyme. This protein crystallized easily and showed no hydrolytic activity, allowing us to determine the structure of the protein in complex with intact Lewis A , Lewis X , and Lewis Y antigen substrates. In all three cases, clear electron density for the complete glycans in the active site was present, allowing us to model these substrates (Fig. S2).
SpGH29 C T interacts with the Lewis A antigen in a manner that is largely indistinguishable from the interaction of BiAfcB with the same antigen structure ( Fig. 2A) (41). The terminal Figure 1. Overall structure of SpGH29 C . The X-ray crystal structure of SpGH29 C is represented as a cartoon colored from blue (N terminus (Nter)) to red (C terminus (Cter)). Both catalytic residues and a Bis-Tris molecule observed to be bound in the active site are shown as gray sticks. The numbering of helices and ␤-strands comprising the (␣/␤) 8 catalytic module is indicated. The strand numbering of the ancillary module is also indicated.

Fucosylated glycan degradation by S. pneumoniae
fucose residue, which is in a standard 1 C 4 chair conformation, sits in the Ϫ1 subsite, making a series of hydrogen-bonding interactions and a classical CH-interaction with Trp-264. This poise for the fucose residue places the oxygen of its glycosidic bond in proximity to Gln-215, which in the unmutated enzyme would be a glutamate residue, thus indicating the appropriate positioning of this residue to act as the catalytic acid/base ( Fig. 2A). Asn-171, which in the unmutated enzyme would be an aspartate residue, is placed ϳ3.5 Å beneath C1 of the fucose, consistent with its role as a nucleophile in the active enzyme ( Fig. 2A).
The GlcNAc residue that precedes the fucose residue and is in the type I motif of the Lewis A antigen does not appear to make any interactions with the enzyme active site, and thus we cannot structurally define a distinct ϩ1 subsite. However, this residue must be present in the minimal trisaccharide substrate of the enzyme, so we consider this as a pseudo ϩ1 subsite (referred to as ϩ1*). The terminal galactose residue of the antigen, however, sits in a subsite, which we refer to as a ϩ2Ј subsite, where the plane of C3-C4 -C5 packs against Trp-211 and the C6, C3, and, notably, C4 hydroxyl groups make a series of hydrogen bonds with the active site. This particular constellation of interactions thereby provides specificity for galactose in this subsite.
The structures of SpGH29 C T D171N/E215Q in complex with the Lewis X (Fig. 2B) and Lewis Y (Fig. 2C) antigens revealed the molecular basis for accommodation of the type II core motif as well as the recognition of the additional ␣-(1,2)-linked terminal fucose residue in the Lewis Y antigen (Figs. 2C and S2). In both complexes, the fucose and galactose residues in the Ϫ1 and ϩ2Ј subsites, respectively, employ an identical set of interactions to those described for the Lewis A complex. Likewise, the GlcNAc is positioned in the ϩ1* subsite; however, revealing the plasticity of this pseudo-subsite, the GlcNAc is flipped 180°i n accordance with accommodating the altered linkages to the fucose and galactose residues. The terminal ␣-(132)-linked fucose of the Lewis Y antigen makes only a water-mediated hydrogen bond and thus is largely just accommodated by the active site of the enzyme rather than appearing to act as a key recognition determinant. Presumably, the terminal ␣-(132)linked fucose of the Lewis B antigen, with its type I core motif, would be accepted in a similar manner.
Overall, therefore, the specificity of SpGH29 C is determined by the unique spatial arrangement of the Ϫ1 and ϩ2Ј subsites and the occupation of these subsites by the appropriately positioned fucose and galactose residues, respectively, in the nonsialylated series of Lewis antigens. The accommodation of both the type I and II motifs in these antigens is enabled by the lack of specific interactions between the ϩ1* subsite and the GlcNAc residue in these glycans. Notably, this distinctive set of interactions legislates against recognition and hydrolysis of ␣-(132)fucosidic bonds, necessitating the presence of SpGH95 C to process glycans with this modification.

SpGH29 C and SpGH95 C initiate a cascade of histo-blood group degradation
SpGH29 C and SpGH95 C are ␣-fucosidases with differing linkage specificities and therefore have the potential to uncap

Fucosylated glycan degradation by S. pneumoniae
fucosylated glycans to expose potential substrates for other pneumococcal exoglycosidases. We tested the ability of SpGH29 C and SpGH95 C to initiate complete degradation of human histo-blood group antigens into monosaccharides by other known pneumococcal GHs using fluorophore-assisted carbohydrate electrophoresis (FACE). This is illustrated, as an example, by the sequential depolymerization of the type IV H-tetrasaccharide (Fig. 3). This glycan is resistant to depolymerization by pneumococcal enzymes unless first treated with SpGH95 C . Uncapping of this glycan by SpGH95 C exposes a terminal Gal-␤-(133)-GalNAc motif, which could be hydrolyzed by the exo-␤-(133)-galactosidase BgaC (23) to release galactose. The sequential action of SpGH95 C and BgaC then allowed GH20C, a known exo-␤-N-acetylhexosaminidase (30), to cleave the remaining GalNAc-␤-(133)-Gal disaccharide. This general approach was used to examine the depolymerization of a wider range of glycans.
The lacto-N-biose (Gal-␤-(133)-GlcNAc) and LacNAc (Gal-␤-(134)-GlcNAc), which are found in type I H-trisaccharide/Lewis A /Lewis B and type II H-trisaccharide/Lewis X / Lewis Y , respectively, are known targets for the characterized pneumococcal exoglycosidases BgaC (23) and BgaA (16,42). In the absence of SpGH95 C , these ␤-galactosidases are unable to degrade the H-trisaccharides (Fig. S3, A and B). SpGH29 C is required to uncap the Lewis A and Lewis X antigens (Figs. 4 and S3, C and D). Both SpGH95 C and SpGH29 C are required to remove the capping fucose residues from Lewis B and Lewis Y and to allow degradation by BgaC or BgaA, respectively (Figs. 4 and S3, E and F). We observed that either fucosidase is able to initiate the degradation of these glycans (Figs. 4 and S3, E and F). Degradation of Lewis Y can proceed either via SpGH29 C , which generates the type II H-trisaccharide, or via SpGH95 C , which generates Lewis X . These trisaccharides are then acted on by the complementary fucosidase and converge at LacNAc, a substrate for BgaA. A parallel degradation pathway takes place for Lewis B , with type I H-trisaccharide, Lewis A , and lacto-Nbiose acting as intermediates, followed by BgaC activity. Lewis X and Lewis A are sometimes sialylated; therefore, we also determined the order of enzymatic degradation of sialyl-Lewis X and sialyl-Lewis A (Figs. 4 and S3, G and H). The presence of the ␣-(233)-linked sialic acid on both antigens influenced the activity of SpGH29 C by abrogating it on sialyl-Lewis A and limiting activity on sialyl-Lewis X (Fig. S3, E and F). However, desialylation of sialyl-Lewis X or sialyl-Lewis A by the exo-␣-sialidase NanA (21) allowed the activity of SpGH29 C and the other pneumococcal GHs to proceed to full depolymerization of the glycans.

Cellular localization of SpGH29 C
Neither SpGH29 C nor SpGH95 C possesses an LPXTG cell wall-anchoring motif, and protein localization prediction software (43) did not identify any signal peptides. However,  GHs are indicated in bold next to the arrow for the reaction they catalyze. A, degradation of Lewis Y can be initiated either by SpGH29 C , which yields the type II H-trisaccharide (H-Tri), or by SpGH95 C , which yields Lewis X . The complementary ␣-fucosidase then acts to produce N-acetyllactosamine (LacNAc), which is cleaved into its constituent monosaccharides by BgaA. Sialyl-Lewis X must be desialylated by NanA prior to SpGH29 C activity. B, degradation of Lewis B can be initiated either by SpGH29 C , which yields the type I H-trisaccharide, or by SpGH95 C , which yields Lewis A . The complementary ␣-fucosidase then acts to produce lacto-N-biose, which is cleaved into its constituent monosaccharides by BgaC. Sialyl-Lewis A must be desialylated by NanA prior to SpGH29 C activity. See

Fucosylated glycan degradation by S. pneumoniae
S. pneumoniae is known to export many of its carbohydrateactive enzymes, both classically and nonclassically, either into the supernatant or to be associated with the cell wall (10,44). Given the initiating role SpGH29 C and SpGH95 C play in the degradation of fucosylated glycans and the fact that BgaA, BgaC, and GH20C are all known or strongly suspected to be exported (23,30,45,46), we hypothesized that SpGH29 C and SpGH95 C function extracellularly. To experimentally test our hypothesis, we assayed isolated cellular fractions of S. pneumoniae TIGR4 for SpGH29 C activity. Exposure of Lewis X to the cell wall-associated fraction (CWF) and total soluble fraction (TSF) resulted in loss or significant reduction of the band corresponding to Lewis X on a FACE gel, indicating processing of the glycan (Fig. 5A). Production of bands corresponding to monosaccharides could also be seen in the TSF-treated sample, but the CWF-treated sample contained a contaminating species that migrated the same distance as the monosaccharides, which prevented conclusive visualization of monosaccharides in this sample. Notably, neither the TSF-treated sample nor the CWF-treated samples displayed the presence of a LacNAc intermediate, as seen in the sample of Lewis X treated with recombinant SpGH29 C . LacNAc is the substrate of BgaA, which is cell wall-associated via its N-terminal signal peptide and C-terminal LPXTG motif (46). The absence of LacNAc in the CWF-treated sample, therefore, most likely indicates that SpGH29 C and BgaA are localized together in this fraction. To confirm that the degradative activity against Lewis X observed with the CWF and TSF was initiated by SpGH29 C , we repeated this experiment with a deletion mutant of SpGH29 C (⌬spgh29 C ;  Fig. 5B). In this experiment, none of the cellular fractions exhibited activity against Lewis X , and no band corresponding to fucose was observed in the TSF-treated sample. These results are most consistent with SpGH29 C being associated with the bacterial cell wall, placing it as another likely example of a nonclassically secreted pneumococcal protein.
Similar attempts were made to determine the localization of SpGH95 C by testing cellular fractions for activity against the type II H-trisaccharide (the substrate against which SpGH95 C exhibited the highest k cat /K m ; Table 1). However, no degradative activity was observed in any of the fractions, including the TSF (data not shown). Therefore, we suggest that SpGH95 C is not expressed under typical laboratory growth conditions.

The ability of pneumococcal GHs to degrade important human glycans
We have demonstrated the ability of SpGH29 C and SpGH95 C , together with other pneumococcal GHs, to completely degrade H-and Lewis blood group antigens into their monosaccharide constituents. As previously mentioned, these antigens are frequently observed as capping motifs on more complex glycans (19,34). Therefore, we set out to assess the overall ability of the S. pneumoniae glycan-processing machinery to depolymerize important human glycans. Trifucosyllacto-N-hexaose (TFLNH) is a human milk oligosaccharide, but it mimics a complex O-glycan containing many of the linkages and motifs that S. pneumoniae likely encounters during colonization and infection, including terminal Lewis X and Lewis B motifs as well as an internal lacto-N-tetraose motif (Gal-␤-(133)-GlcNAc-␤-(133)-Gal-␤-(134)-Glc), which forms the core of the lacto series of glycosphingolipids (20). Thus, this complex glycan makes an excellent model glycan, and therefore we used it as a substrate to demonstrate the capacity of the pneumococcal GH arsenal to depolymerize a highly modified glycan (Fig. 6). Using FACE analysis, we observed the ability of pneumococcal GHs to cleave all eight differentlinkagespresentinTFLNHandthefundamentaldependence on SpGH29 C and SpGH95 C for initiation of this process (Figs. 6 and S4). SpGH29 C was able to remove both the ␣-(133)-linked fucose residue from the arm bearing a Lewis X motif and the ␣-(134)-linked fucose from the Lewis B arm of TFLNH without prior action of SpGH95 C . In contrast, SpGH95 C exhibited only partial activity against TFLNH, and the presence of SpGH29 C was required to facilitate complete removal of the ␣-(132)-linked fucose. In the absence of SpGH95 C , SpGH29 C was able to initiate degradation of the against Lewis X as detected by fluorophore-assisted carbohydrate electrophoresis. The activities of recombinant SpGH29 C and BgaA against Lewis X are shown as controls, and fucose is shown as a standard. C, background labeling of the different cellular fractions in the absence of Lewis X . Lewis X and Lewis X treated with total soluble protein are shown for comparison. Le X , Lewis X ; EF, extracellular fraction; CF, cytoplasmic fraction; MF, membrane fraction. The 8-aminonaphthalene-1,3,6-trisulfonic acid (ANTS) lane indicates background labeling due to the fluorophore alone. Due to the background labeling of the cell wallassociated fraction, SpGH29 C activity can be observed as a disappearance of Lewis X rather than an appearance of fucose.

Fucosylated glycan degradation by S. pneumoniae
Lewis X arm of TFLNH by BgaA and GH20C; however, the Lewis B arm could not be degraded. Therefore, both fucosidases were required to uncap the two arms of TFLNH. The complete degradation of TFLNH by BgaA, BgaC, and GH20C following defucosylation is consistent with their published activities (23,30,42).

Discussion
S. pneumoniae is considered an accomplished degrader of human glycans, with known capacity to depolymerize complex and high-mannose N-linked glycans as well as some O-linked glycans (e.g. Refs. 20, 24, 27, 28, and 46). The bacterium also has the ability to metabolize the glycosaminoglycan hyaluronan (e.g. Refs. 47 and 48) and glycogen (e.g. Refs. 49 and 50). The activities of some of the pneumococcal enzymes are also consistent with depolymerization of glycosphingolipid glycans (23,30). Here, we have focused on the previously uncharacterized capacity of S. pneumoniae to degrade the full complement of fucosylated blood group H-and Lewis antigens and the underlying glycans that can bear these motifs.
Fucose is an important and common monosaccharide that often decorates, and more frequently terminates, a number of human glycans (52). We have previously reported that all sequenced strains of S. pneumoniae carry one of two types of fucose utilization operon (24,53,54). Both operon types encode for a set of intracellular enzymes dedicated to processing free fucose to dihydroxyacetone phosphate and lactaldehyde, whereas the transporter systems and GHs that process the gly-cans vary between the operons (14,55). The type 1 operon is found in the majority of pneumococcal strains, including TIGR4, and encodes for a member of GH family 98, Sp4GH98, which is an extracellular endo-␤-galactosidase that cleaves the type II linkage of Lewis Y . This action releases a free H-disaccharide, whereas the GlcNAc and ␣-(133)-linked fucose remain attached to the glycoconjugate. The released H-disaccharide is thought to be imported by a phosphotransferase system transporter and then degraded by a putative intracellular GH95 (encoded by a gene distinct from the gene encoding SpGH95 C ). The type 2 operon, which was originally identified in a serotype 3 strain of S. pneumoniae, also encodes for a GH98 endo-␤galactosidase, Sp3GH98, that cleaves type II linkages; however, this enzyme is specific for blood group A-and B-antigens. Sp3GH98 releases soluble A/B-trisaccharides, which are then imported by an ABC transporter into the cytoplasm where they are degraded by a putative GH29 (encoded by a gene distinct from the gene encoding SpGH29 C ) and two putative GH family 36 members (10). Thus, there is evidence that S. pneumoniae can harvest fucosylated glycans from host tissues. Indeed, in TIGR4, the presence of the type 1 fucose operon is strongly linked to the full virulence of the microbe (56). However, by virtue of the well-characterized endo-acting enzymes that initiate Lewis Y or A/B-antigen harvesting, the models of the pathways encoded by these operons indicate highly specific glycan targets, which do not include Lewis A , Lewis B , Lewis X , or H-antigens.  Fig. S4. SpGH95 C and SpGH29 C are required to remove the capping fucose residues from TFLNH and allow access to the oligosaccharide by other GHs. Treatment of TFLNH with SpGH29 C results in removal of the ␣-(133)and ␣-(134)-linked fucose units and allows BgaA and GH20C to degrade the arm proximal to the reducing end; however, without SpGH95 C , the distal arm cannot be degraded. Treatment of TFLNH with SpGH95 C results in partial removal of the ␣-(132)-linked fucose unit and a difucosylated oligosaccharide that cannot be acted upon by other GHs (except SpGH29 C ). If the ␣-(133)and ␣-(134)-linked fucose units are removed by SpGH29 C first, SpGH95 C is able to fully remove the ␣-(132)-linked fucose unit from the distal arm to produce lacto-N-hexaose. This hexasaccharide can then be fully degraded into galactose and glucose by the combined actions of BgaA, BgaC, and GH20C. See Fig. S4 for experimental validation of this pathway.

Fucosylated glycan degradation by S. pneumoniae
The presence of SpGH29 C and SpGH95 C as part of the core arsenal of GHs deployed by S. pneumoniae indicates that all strains of this bacterium likely have an innate capacity to target the H(O)-blood group antigen and all Lewis antigens, again suggesting the importance of fucosylated glycan degradation to the host-adapted lifestyle of S. pneumoniae. However, it also reveals potential redundancy, and even competition, between the functions of the "core" fucosidases and the fucose utilization pathways. For example, the processing of Lewis Y by SpGH29 C or SpGH95 C would prevent the action of Sp4GH98, which is unable to cleave the type II H-trisaccharide or Lewis X products, respectively, that would be left by exo-␣-fucosidase activity (24). Conversely, the cleavage of Lewis Y by Sp4GH98 leaves a glycoconjugate terminating in Fuc-␣-(133)-GlcNAc, which is not a substrate for any of the known pneumococcal enzymes. Therefore, unless these enzymes are competing for substrates, they are likely expressed under different conditions in vivo, which are yet to be uncovered.
Although SpGH95 C was able to cleave ␣-(132)-linked fucose residues found on histo-blood group antigens, the blood group A-and B-antigens were resistant to defucosylation by this enzyme. We were unable to obtain the X-ray crystal structure of SpGH95 C ; however, this enzyme is clearly unable to accommodate the additional terminal GalNAc/galactose residue found on blood group A/B-antigens. The lack of this activity is consistent with the well-characterized GH95 enzyme from Bifidobacterium bifidum (57). One potential mechanism for the degradation of the A/B-antigens could involve removal of the terminal GalNAc/galactose by an exo-␣-N-acetylgalactosaminidase/galactosidase, which would allow degradation of the resulting H-antigen by SpGH95 C and additional enzymes, depending on the glycan core type. The core S. pneumoniae genome, however, encodes for only a single GH having this possible activity, Aga, which is a member of GH36. This enzyme exhibits ␣-(136)-galactosidase activity against the plant oligosaccharide raffinose (14,58). Furthermore, in direct tests, we failed to find activity for Aga on blood group A/B-antigens (not shown). Thus, deconstruction of the H(O)-blood group antigen and all Lewis antigens is a conserved feature in the glycan-degrading capacity of all S. pneumoniae strains, but targeting the A/B-antigens is not. Nevertheless, the type 2 fucose utilization operon found in some strains of S. pneumoniae is specific for the blood group A/B-antigens; therefore, at least a subset of strains has the ability to target these glycans. As we have inferred previously, the apparent differential ability of particular S. pneumoniae strains to degrade A/B-antigens may have implications for host susceptibility to infection (24).
A key distinction between the fucosylated glycan degradation pathways described here and those encoded by the type 1/2 operons is the cellular location in which defucosylation occurs. S. pneumoniae is able to import galactose and GlcNAc, which would be released from histo-blood group antigens extracellularly by BgaA and BgaC, and use them as a carbon source for growth (12,59,60); however, it is unable to grow on exogenous fucose (54,56). Both type 1 and 2 fucose utilization operons are known or predicted to import fucosylated glycans and utilize intracellular ␣-fucosidases. Therefore, the released fucose can then be processed by the other components of the operon and feed into central metabolism (54). In contrast, we have shown that SpGH29 C is cell wall-associated. Likewise, based on its uncapping function and the cellular location of the enzymes that act after it, we predict that SpGH95 C is also extracellular. Therefore, the fucosylated glycan degradation pathways described here would release free fucose that apparently cannot be utilized by S. pneumoniae. As such, the bacterium may view fucose as a capping residue that has to be removed for the pneumococcus to release other monosaccharides that it can import. This apparent "waste" of fucose may point more importantly toward the functional significance of fucose in the context of human glycoconjugates and the importance of defucosylation to other aspects of the host-pneumococcus interaction rather than simple nutrition.
SpGH29 C and SpGH95 C possess complementary linkage specificities that, together, allow them to expose a wide range of human glycans to the action of other pneumococcal GHs. It is common for deletion mutants of pneumococcal initiating enzymes, such as NanA and the high-mannose N-glycan degradation initiator SpGH92, to display strong virulence phenotypes (10). Therefore, it is consistent that SpGH29 C and SpGH95 C have been identified as putative virulence factors in multiple animal models of disease (13,35,37). Given the known role of the type 1 operon in pneumococcal virulence (56) and the uncapping function of SpGH29 C and SpGH95 C , we hypothesize that directed studies into the contributions of these fucosidases to the host-pathogen interaction would confirm their roles as important virulence factors.
During our exploration of glycan degradation by the enzymes of S. pneumoniae, we unexpectedly uncovered a previously unknown activity for BgaC. Jeong et al. (23) previously reported that BgaC is unable to cleave the Gal-␤-(133)-GalNAc motif in the context of the ganglioside GA1 (Gal-␤-(133)-GalNAc-␤-(134)-Gal-␤-(134)-Glc; also known as asialo GM1). However, we observed that BgaC cleaved the terminal Gal-␤-(133)-Gal-NAc motif in the type IV H-tetrasaccharide after it was uncapped by SpGH95 C . This suggests that the substrate repertoire for BgaC is broader than previously suspected, which is notable because this linkage also occurs in the core of O-linked glycans as well as the globoside series of glycosphingolipids.
S. pneumoniae possesses a considerable ability to degrade distinct linkages found in human glycans. Of the Ͼ20 linkages commonly found in N-glycans, O-glycans, histo-blood group antigens, and glycosphingolipids, many are now associated with the activity of a characterized pneumococcal GH. Overall, our characterization of two complementary ␣-fucosidases and the in vitro recapitulation of glycan degradation pathways employed by S. pneumoniae expands the known capacity of this pathogen to degrade human glycans and highlights the comprehensive nature of its ability to target the human glycome.

Fucosylated glycan degradation by S. pneumoniae
type II A-and B-tetrasaccharides, and lacto-N-tetraose were obtained from Elicityl (Crolles, France). TFLNH was purchased from ProZyme (Hayward, CA). All other materials were from Millipore-Sigma unless otherwise stated.

Cloning and mutagenesis
The gene encoding for full-length SpGH29 C (amino acids 1-559) from TIGR4 (locus tag SP_2146) was amplified by PCR with the primers GH29-F and GH29-R (Table S1) and cloned into pET28a between the NdeI and SalI sites to produce pET28a-SpGH29 C . A truncated version of SpGH29 C (amino acids 1-451) was also cloned into pET28a using the primers GH29-F and GH29T-R to produce pET28a-SpGH29 C T. The gene encoding for full-length SpGH95 C (locus tag SP_1654) was codon-optimized for expression in E. coli and synthesized by GenScript (Piscataway, NJ). This synthetic gene was then cloned into pET28a between the NdeI and XhoI sites to produce pET28a-SpGH95 C . BgaC (locus tag SP_0060) and the catalytic domain of NanA (amino acids 303-777; locus tag SP_1693) were amplified using the primers BgaC-F, BgaC-R, NanA-F, and NanA-R and cloned into pET28a between the NheI and NotI or NdeI and XhoI sites to produce pET28a-BgaC and pET28a-NanA, respectively. Cloning of BgaA and GH20C has been reported previously (16,30). Mutagenesis of pET28a-SpGH29 C T to generate the SpGH29 C T D171N/E215Q double mutation was performed using the "megaprimer" PCR method (61). All mutagenic primers are listed in Table S1. The integrity of all constructs was confirmed by bidirectional sequencing.

Protein expression and purification
Protein expression constructs were transformed into BL21(DE3) (or Tuner TM (DE3) for expression of ␤-galactosidases). Expression of SpGH29 C , SpGH29 C T, and BgaC was performed in LB broth with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside induction at 16°C for 18 h; SpGH95 C and NanA were expressed in autoinduction medium at 16°C for 4 days. Expression of BgaA and GH20C has been reported previously (16,30). Standard procedures, as detailed previously (62), were used to lyse cells and purify the released proteins by immobilized metal-affinity chromatography and size-exclusion chromatography using either an S200 or S300 HiPrep 16/60 Sephacryl column (GE Healthcare) as appropriate. Protein purity was judged by SDS-PAGE analysis, and protein concentrations were determined using extinction coefficients calculated by ProtParam on the ExPASy server (63).

␣-Fucosidase assays
The activity of SpGH29 C and SpGH95 C on ␣-fucosylated glycans was assayed by TLC and the detection of liberated fucose using an L-fucose assay kit that contains an NADP ϩ -dependent fucose dehydrogenase (Megazyme Inc., Chicago, IL). TLC reactions contained 5 mM substrate and 1 M enzyme in 20 mM Tris, pH 8.0, and were incubated at 37°C for 1 h. Reactions were spotted onto precoated POLYGRAM SIL G/UV 254 TLC sheets (Thermo Fisher Scientific, Waltham, MA), separated in a solvent of 7:2:1 propanol:H 2 O:ethanol, and visualized with 5% (v/v) H 2 SO 4 in ethanol followed by heating at 90°C. For the determination of kinetic parameters, the fucose detection kit method was adapted to allow both the ␣-fucosidase and fucose dehydrogenase reactions to occur simultaneously. Conditions were optimized to ensure that neither the fucose dehydrogenase nor NADP ϩ were limiting. Reactions (100 l) contained 5 l of substrate (at varying concentrations), 4 l of NADP ϩ (kit supply), 2 l of fucose dehydrogenase, and 1 M ␣-fucosidase in 100 mM Tris, 50 mM NaCl, pH 8.0. Reactions (in triplicate) were incubated at 37°C in a SpectraMax M5 plate reader (Molecular Devices, San Jose, CA), and the absorbance at 340 nm was read every 5 s. Slopes for each substrate concentration were converted into NADPH concentrations using an extinction coefficient of 6220 M Ϫ1 cm Ϫ1 . The k cat /K m for each substrate-enzyme combination was calculated by linear regression of the initial velocities versus substrate concentration using GraphPad Prism 6.0.7.

General crystallography procedures
Crystals were obtained using sitting-drop vapor diffusion for screening and hanging-drop vapor diffusion for optimization at 18°C. Prior to data collection, single crystals were flash-cooled with liquid nitrogen in crystallization solution supplemented with 20% (v/v) ethylene glycol as cryoprotectant. Diffraction data were collected either on beamline 9-2 or 11-1 at the Stanford Linear Accelerator Center (SLAC, Stanford Synchrotron Radiation Lightsource (SSRL), CA) or beamline 08B1-1 at the Canadian Light Source (CLS, Saskatoon, Saskatchewan, Canada) as indicated in Table 2. All diffraction data were processed using MOSFLM and SCALA (64 -66). Data collection and processing statistics are shown in Table 2. For all structures, manual model building was performed with Coot (67), and refinement of atomic coordinates was performed with REFMAC (68). Water molecules were added in Coot with Find Waters and manually checked after refinement. In all data sets, refinement procedures were monitored by flagging 5% of all observations as "free" (69). Model validation was performed with MolProbity (70).

SpGH29 C and SpGH29 C T D171N/E215Q structure determinations
A unique crystal of SpGH29 C (25 mg ml Ϫ1 ) was obtained in 16% (w/v) polyethylene glycol (PEG) 3350, 0.15 M potassium chloride, 1 mM DTT, 0.1 M Bis-Tris, pH 6.0. This crystal was flash frozen in liquid nitrogen using the crystallization solution supplemented with 20% (v/v) ethylene glycol. After data collection, initial phases for SpGH29 C were determined by molecular replacement using Phaser (71) and the structure of an ␣-L-fucosidase from Bacteroides thetaiotamicron as the search model (Protein Data Bank (PDB) code 3EYP). An initial model of SpGH29 C was generated by automatic model building using the program Buccaneer (72). SpGH29 C T D171N/E215Q (35 mg ml Ϫ1 ) was cocrystallized in the presence of an excess of Lewis X or Lewis Y in 21-23% (w/v) PEG 4000, 0.22 M sodium acetate, 1 mM DTT, 0.1 M Tris, pH 8.5. Cocrystals of SpGH29 C T D171N/ E215Q with Lewis A were obtained in 20 -24% (w/v) PEG 3350, 0.18 -0.22 M sodium chloride, 1 mM DTT, 0.1 M Tris, pH 8.5. All complexes were solved by molecular replacement using Phaser and the SpGH29 C crystal structure.

Generation of SpGH29 C deletion mutant
A PCR ligation technique was used to replace sp_2146 with a chloramphenicol resistance cassette as described previously (30). Briefly, the chloramphenicol resistance cassette was amplified with the primers CAM-F and CAM-R (Table S1), which introduced a 5Ј NheI site and a 3Ј XhoI site. Up-and downstream regions flanking sp_2146 were amplified using the primers Upstream-F, Upstream-R, Downstream-F, and Downstream-R, which introduced a 3Ј NheI site into the upstream flank and a 5Ј XhoI site into the downstream flank. Following digestion, all three amplicons were ligated together, and this ligation mixture was used to transform S. pneumoniae TIGR4 as described previously (30). The presence and location of the chloramphenicol resistance cassette and absence of sp_2146 were confirmed by multiple PCR analyses and bidirectional DNA sequencing.

Localization of SpGH29 C
S. pneumoniae TIGR4 and ⌬spGH29 were grown in 50 ml of AGCH medium (73) with 1% glucose at 37°C in a candle jar to an OD 600 of 0.6, then pelleted, and resuspended in AGCH medium containing no carbohydrate for a further 30 min (in an attempt to induce expression of GHs). The cells were then pelleted again, and the supernatant was retained as the extracellular fraction and concentrated 100-fold using an Amicon ultrafiltration cell fitted with a 10-kDa molecular-mass-cutoff membrane. The pelleted cells were split into two samples: one was used to produce protoplasts and obtain the cell wall, cytoplasmic, and membrane fractions, whereas the other was resuspended in 50 mM Tris-HCl, pH 7.5; sonicated on ice; and centrifuged to obtain the total protein fraction. The cell pellet intended for protoplast production was washed with 50 mM Tris-HCl, pH 7.5; resuspended in cell wall digestion buffer (74); and incubated at 37°C with gentle shaking for 2 h. The protoplasts were then pelleted, and the supernatant was retained as the cell wall fraction. The cytoplasmic fraction was obtained by gently washing the protoplasts with 50 mM Tris-HCl, pH 7.5, 30% sucrose; resuspending and lysing them in 50 mM Tris-HCl, pH 7.5; pelleting the protoplast membranes at 20,000 rpm for 30 min; and retaining the supernatant. Finally, the membrane fraction was obtained by solubilizing the membranes in 50 mM Tris-HCl, pH 7.5, 0.05% Triton as described previously (75). The different fractions were kept on ice, and 5 l of each was used to set up reactions with Lewis X . Reactions were incubated at 37°C for 48 h and then processed for fluorophore-assisted carbohydrate electrophoresis as described below.

FACE
FACE reactions contained 10 g of glycan substrate and 1 M enzyme in 50 mM sodium phosphate buffer, pH 6.5, 45 mM ␤-mercaptoethanol and were incubated at 37°C for 20 h. Reactions were stopped by the addition of ethanol, dried in a SpeedVac for 4 h, and then labeled overnight with 5 l of 0.2 M 8-aminonaphthalene-1,3,6-trisulfonic acid (Thermo

Fucosylated glycan degradation by S. pneumoniae
Fisher Scientific) and 5 l of 1 M sodium cyanoborohydride at 37°C as described previously (36). Labeled reaction products were separated on a 35% polyacrylamide gel, and labeled glycans were visualized under UV light.