Analysis of a New Family of Widely Distributed Metal-independent α-Mannosidases Provides Unique Insight into the Processing of N-Linked Glycans*

The modification of N-glycans by α-mannosidases is a process that is relevant to a large number of biologically important processes, including infection by microbial pathogens and colonization by microbial symbionts. At present, the described mannosidases specific for α1,6-mannose linkages are very limited in number. Through structural and functional analysis of two sequence-related enzymes, one from Streptococcus pneumoniae (SpGH125) and one from Clostridium perfringens (CpGH125), a new glycoside hydrolase family, GH125, is identified and characterized. Analysis of SpGH125 and CpGH125 reveal them to have exo-α1,6-mannosidase activity consistent with specificity for N-linked glycans having their α1,3-mannose branches removed. The x-ray crystal structures of SpGH125 and CpGH125 obtained in apo-, inhibitor-bound, and substrate-bound forms provide both mechanistic and molecular insight into how these proteins, which adopt an (α/α)6-fold, recognize and hydrolyze the α1,6-mannosidic bond by an inverting, metal-independent catalytic mechanism. A phylogenetic analysis of GH125 proteins reveals this to be a relatively large and widespread family found frequently in bacterial pathogens, bacterial human gut symbionts, and a variety of fungi. Based on these studies we predict this family of enzymes will primarily comprise such exo-α1,6-mannosidases.

A feature of emerging importance to bacteria that colonize or infect humans is their capacity to process host glycans. Streptococcus pneumoniae is one notable human pathogen that relies on this ability for its full virulence (1). Among its known carbohydrate active virulence factors are NanA, StrH, BgaA, and EndoD. NanA, StrH, and BgaA are a sialidase, an exo-␤-D-N-acetylglucosaminidase, and an exo-␤-D-galactosidase, respectively, that sequentially remove the terminal sugars from the distal arms of complex N-linked glycans. EndoD is an endo-␤-D-N-acetylglucosaminidase that cleaves the chitobiose core of N-linked glycans smaller than Man 5 GlcNAc 2 to remove the glycan from the protein scaffold. Despite increasing knowledge in this area it remains unclear how bacteria process the core mannose component of N-linked glycans.
Glycoside hydrolases, enzymes that break glycosidic bonds through a hydrolytic mechanism, are presently classified into 123-amino acid sequence based families (2). ␣-Mannosidases known to process N-glycans are found in families 38, 47, 76, 92, and 99. Very recent studies have shown the bacterial family 38 ␣-mannosidase from Streptococcus pyogenes (SpyGH38) to be a specific exo-␣1,3-mannosidase that is tolerant of the ␣1,6-branches in N-glycans (3). Analysis of family 92 glycoside hydrolases from the human gut symbiont Bacteroides thetaiotaomicron revealed an expanded repertoire of ␣-mannosidases (4). These enzymes displayed activity primarily toward ␣1,2and ␣1,3-mannosidic linkages with some having low ␣1,6-mannosidase activity. In addition to the established ability of S. pneumoniae to exo-hydrolytically process the distal arms of complex glycans, which comprise sialic acid, galactose, and N-acetylglucosamine, consideration of additional putative carbohydrate-active enzymes found in this organism suggests it can partly degrade the mannose component of N-glycans using enzymes similar to those found in S. pyogenes and B. thetaiotaomicron. Through these observations it has become clear that some bacteria, possibly including S. pneumoniae, have the capacity to process the mannose component of N-glycans. A noteworthy gap, however, in the known bacterial N-glycan degradation pathway is that efficient ␣1,6-mannosidases that would be required for complete N-glycan depolymerization have not yet been found. Thus, there is a clear possibility that further study of the bacterial glycan processing machineries might reveal new catabolic enzymes and additional insight into bacterial N-glycan degradation.
The gene encoding the hypothetical protein SP_2144 from Streptococcus pneumoniae TIGR4 resides in a putative carbohydrate processing locus containing ␣-mannosidases, an ␣-fucosidase and a ␤-hexosaminidase. SP_2144 shows ϳ40% amino acid sequence identity to four proteins whose structures were deposited in the Protein Data Bank (PDB code 2nvp from Clostridium perfringens strain 13; 2p0v from Bacteroides thetaiotaomicron; 3p2c and 3on6, both from Bacteroides ovatus). These structures display an (␣/␣) 6 -fold and have substantial structural identity with the family 15 glycoside hydrolases, which are primarily ␣-glucanases. The genomic context of SP_2144 and its similarity to proteins having structural identity to carbohydrate-active enzymes led to the hypothesis that SP_2144, and by extension its homologs, would have activity on carbohydrates.
Here the analysis of SP_2144 from S. pneumoniae TIGR4 and a homolog from Clostridium perfringens ATCC 13124 (locus tag CPE0426) is described. Through structural and functional studies both of these proteins, called SpGH125 and CpGH125, are shown to be strict exo-␣1,6-mannosidases that operate via a metal-independent inverting catalytic mechanism. These two proteins are thus the founding members of a new glycoside hydrolase family, GH125. Phylogenetic analyses reveal GH125 proteins to be quite widespread in prokaryotes and eukaryotes. Significantly, among the Firmicutes bacteria, GH125 encoding genes co-occur at a very high frequency with adjacent genes encoding GH38 mannosidases; the role of these enzymes in N-glycan degradation is discussed.

EXPERIMENTAL PROCEDURES
Cloning, Production, and Purification of GH125-The fulllength spgh125 gene (locus tag SP_2144) was PCR amplified from S. pneumoniae TIGR4 genomic DNA (ATCC BAA-334D) using the following forward and reverse oligonucleotide primers, 5Ј-GGC AGC CAT ATG ATG GTT TAT TCG AAA G-3Ј and 5Ј-GTG GTG CTC GAG TTA GCG AAT ATC CAA GTA ATC C-3Ј, respectively. The full-length cpgh125 gene (locus tag CPE0426) was PCR amplified from C. perfringens ATCC 13124 genomic DNA using the following forward and reverse oligonucleotide primers, 5Ј-TAT ACC ATG GCC AGT TTA TCA  ACT AAC GAA TT-3Ј and 5Ј-GTG GTG CTC GAG TTT TTT  ATT TAT AAC TTT CTC-3Ј, respectively. The PCR amplified gene fragments were obtained using standard PCR methods with Phusion High-Fidelity DNA polymerase (New England Biolabs). The amplified spgh125 gene was cloned into pET-28a(ϩ) (Novagen) via engineered 5Ј and 3Ј NdeI and XhoI restriction sites, respectively. The amplified cpgh125 gene was similarly cloned with 5Ј and 3Ј NcoI and XhoI restriction sites, respectively. Standard cloning procedures were used. The plasmid containing the spgh125 gene encoded the polypeptide preceded by an N-terminal, thrombin cleavable His 6 tag. The plasmid containing the cpgh125 gene encoded the polypeptide followed by a C-terminal His 6 tag. The DNA sequences of the constructs were verified by bidirectional sequencing.
The appropriate plasmids were transformed into chemically competent E. coli BL21 STAR (DE3) cells (Novagen). SpGH125 and CpGH125 proteins were produced in Luria-Bertani medium containing 50 g ml Ϫ1 of kanamycin (Sigma). The cells were grown at 37°C to an optical density of 0.5 and induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside at 18°C for 14 h. Cells were harvested by centrifugation at 27,000 ϫ g and chemically lysed (5). Cell debris was pelleted using centrifugation at 27,000 ϫ g for 45 min. The polypeptides were purified from a cell-free extract using immobilized metal affinity chromatography using previously described methods (6). The purity of fractions was assessed using SDS-PAGE and those deemed to be greater than 95% pure were pooled, concentrated, and buffer exchanged into 20 mM Tris-HCl, pH 8.0, in a stirred ultrafiltration unit (Amicon) using a 10,000 molecular weight cut-off membrane (Filtron). Protein was further purified by size exclusion chromatography using Sephacryl S-200 (GE Biosciences) in 20 mM Tris-HCl, pH 8.0. The concentration of purified protein was determined by UV absorbance at 280 nm using calculated molar extinction coefficients of 103,625 and 91,010 M Ϫ1 cm Ϫ1 for SpGH125 and CpGH125, respectively (7).
GH125 Activity Assay-The activities of SpGH125 and CpGH125 were initially screened on a variety of chromogenic arylglycosides (see supplemental Table S1). These assays were performed with 500 nM enzyme and 0.5 mM substrate, all in phosphate-buffered saline (PBS), pH 7.4, at 37°C. The reaction was monitored spectrophotometrically at 400 nM. Kinetic studies using the 2,4-dinitrophenyl ␣-D-mannopyranoside (DNP-Man) 5 were carried out at 25°C using a Spectramax Plus 384 microplate reader (Molecular Devices). Standard reaction mixtures comprised 200-l reactions in PBS, pH 7.5, containing 0.1% BSA (Sigma), 2 M enzyme, and 100 M to 2.5 mM DNP-Man. The production of 2,4-dinitrophenolate was monitored at 400 nm over 30 min. The reaction velocity was determined by linear regression of the data within the early time period of the assay. Absorbance values were converted to concentration terms using an extinction coefficient (at ϭ 400 nm) of 10.91 mM Ϫ1 cm Ϫ1 for the dinitrophenolate at pH 7.4. To determine whether the presence of EDTA would eliminate or reduce the catalytic activity of SpGH125 and CpGH125 toward DNP-Man, kinetic studies were carried out as described above in the presence of 10 mM EDTA with 1 mM DNP-Man as the substrate.
Non-chromogenic substrates (see supplemental Table S1) were tested for activity using high performance anion exchange chromatography using pulsed amperometric detection (HPAEC-PAD) to monitor glycoside cleavage. Assays were performed in 20-l volumes at 37°C in PBS at pH 7.5. Reactions were initiated by the addition of 500 nM enzyme to 0.5 mM of each carbohydrate. Reactions were stopped by the addition of 140 l of 100 mM sodium hydroxide. After centrifugation at 5,000 rpm for 5 min, 20 l of the samples were analyzed by HPAEC-PAD using a Dionex ICS 3000 HPLC equipped with an ASI 100 Automated sample injector and an ED 50 electrochemical detector (Dionex) with a gold working electrode and an Ag/AgCl reference electrode. Products were analyzed using a PA-20 column set (analytical plus guard column) using an isocratic gradient of 100 mM NaOH.
Capillary Electrophoresis-The oligosaccharide standards Man9, Man3a, and Man1 were obtained from V-Labs Inc., whereas Man5 was purchased from Glyko/Prozyme. All standards were dissolved in H 2 O and stored at Ϫ20°C until use.
Lyophilized oligosaccharide standards, Man-9, Man-5, Man-3a, and Man-1 were labeled with APTS by reductive amination exactly as previously described (8). Excess labeling reagents and impurities were removed from APTS-labeled standards by fluorophore-assisted carbohydrate electrophoresis as described previously (9) with minor modifications (supplemental Fig. S2). Briefly, labeled material was resolved by gel electrophoresis at a constant voltage (200 V) on 0.75-mm thick, 20% polyacrylamide gels, at 4°C in the dark. Bands corresponding to N-glycan standards, or their mannosidase-digested products, were excised, transferred to separate 15-ml centrifuge tubes, mixed with 0.4 ml of H 2 O, and sonicated in a water bath sonicator for 2 h. The H 2 O was removed, and the gel slices were extracted a second time. All extracts were lyophilized, dissolved in H 2 O, and analyzed by capillary electrophoresis (CE) before they were desalted on Hypercarb SpinTips as described by Whitworth et al. (9). APTS-labeled material eluting in 50% (v/v) CH 3 CN ϩ 25 mM TFA was immediately flash-frozen in liquid nitrogen and lyophilized.
Digestion of APTS-labeled glycans with commercial ␣(1,2)/ (1,3)-mannosidase was performed following the manufacturer's protocols. SpGH125 and CpGH125, both in PBS, were tested at 5 mg/ml. All digests contained 0.1% BSA and were incubated for ϳ24 h. Double digested samples were pre-treated with ␣(1,2)/(1,3)-mannosidase for 3 h prior to adding SpGH125 or CpGH125 for an additional 18 h. All reactions were quenched by the addition of 3 volumes of Ϫ20°C ethanol. After further incubation at Ϫ20°C for 1 h, samples were centrifuged (4°C, 10,000 ϫ g, 20 min) to pellet the precipitated protein, and then analyzed directly by CE.
All CE separations were carried out using a ProteomeLab PA800 (Beckman Coulter) equipped with a laser-induced fluorescence detector and a 488-nm argon ion laser. Separations were performed on PVA-coated capillaries (Beckman Coulter) filled with NCHO separation buffer (Beckman Coulter) under reversed polarity conditions as described previously (10).
NMR Experiments-1 H NMR spectroscopy (600 MHz Bruker AMX spectrometer) was used to follow the progress and identify the products of the SpGH125-and CpGH125-catalyzed reactions. The reactions were carried out in ϳ0.2 ml (21°C, PBS buffer (pH 6.5)) containing 7.0 mM methyl 6-O-(␣-D-mannopyranosyl)-␤-D-mannopyranoside. Initiation of the reaction was done by the addition of 15 l of a 10 mg/ml stock of recombinant SpGH125 or CpGH125. The hydrolysis of the disaccharide was monitored until the reaction reached equilibrium. An initial spectrum (referred to as time 0) containing substrate and buffer was acquired before the addition of enzyme.
Crystallization and X-ray Data Collection-All crystals were obtained using hanging or sitting-drop vapor diffusion at 18°C. Prior to crystallization, SpGH125 and CpGH125 were concentrated to 15 mg ml Ϫ1 in 20 mM Tris, pH 8.0. SpGH125 native crystals were obtained in 0.1 M Hepes, pH 8.0, and 1.2 M LiSO 4 . CpGH125 native crystals were obtained in 0.1 M MES, pH 6.5, and 12% polyethylene glycol 2000 (Hampton Research).
All crystals were cryoprotected with crystallization solution supplemented with 25% ethylene glycol (Hampton Research) and flash cooled directly in a nitrogen gas stream at 113 K. Data sets were collected either on a "home beam" comprising a Rigaku R-AXIS 4ϩϩ area detector coupled to a MM-002 x-ray generator with Osmic "blue" optics and an Oxford Cryostream 700, BL9-2 at the Stanford Synchrotron Radiation Laboratories, or CMCF1 at the Canadian Light Source as indicated in Table 1. All diffraction data were processed using MOSFLM/SCALA in the CCP4 suite of programs (11,12). Data collection and processing statistics are shown in Table 1.
Structure Solution and Refinement-The structure of native SpGH125 was determined by molecular replacement using MOLREP (13) to find two molecules of SpGH125 in the asymmetric unit using the coordinates of CPE0426 from C. perfringens strain 13 (PDB code 2nvp) as a search model. The initial model was corrected and completed manually by multiple rounds of building using COOT (14) and refinement using REFMAC (15). The same approach was used to solve the structure of CpGH125 in complex with methyl-S-(␣-D-mannopyranosyl)-(1-6)-␣-D-mannopyranose. The completed model of native SpGH125 was used to solve the structures of SpGH125 in complex with dMNJ and mannobiose. These models were manually corrected and refined as above. Water molecules were added using COOT:FINDWATERS and manually inspected after refinement. In all data sets, 5% of the observations were flagged as "free" and used to monitor refinement procedures (16). Model validation was performed with SFCHECK (17), PROCHECK (18), and MOLPROBITY (19). Structure and refinement statistics are shown in Table 1.
Phylogeny-Searches for protein sequence similarity to C. perfringens ATCC 13124 CpGH125 were performed with BLASTp (20) and the sequences with significant similarity (E-value cutoff of 10 Ϫ5 ) were extracted from the National Center for Biotechnology Information (NCBI) via blast explorer (www.phylogeny.fr) (21). For the multiple sequence align-ments, sequences were restricted to one paralogue per species and selected from Firmicutes, the Bacteroidetes/Chlorobi group, and Eukaryotes with the aim to select an overall representative data set. All initial multiple sequence alignments were carried out with ClustalW3D using SpGH125 and CpGH125 as lead structures to verify occurrence of the catalytic residues and the key residues involved in substrate recognition. Alignments were manually edited with Bioedit based on the structural information of the catalytic residues and residues involved in substrate binding in SpGH125 (S. pneumoniae TIGR4) and CpGH125 (C. perfringens ATCC 13124) (22). For the final tree construction the multiple sequence alignment was calculated with Mafft (23, 24) using Blosum62 and the trees were constructed by a minimum evolution method in Mega4 (25). The reliability of the trees was always tested by bootstrap analysis using 100 resamplings of the data set. To calculate the conservation of amino acids the Consurf server was used (26,27). The conservation scores were calculated using the maximum likelihood algorithm based on the same multiple sequence alignment that was used to compute the phylogenetic tree. The conservation scores were plotted onto the crystal structure of SpGH125 using the B-factor coloring mode in PyMOL.

RESULTS AND DISCUSSION
GH125 from S. pneumoniae and C. perfringens Are ␣1,6-Mannosidases-Guided by both the genomic context of SpGH125 and its distant relationship to glucose processing enzymes to choose potential substrates, recombinant SpGH125 was screened against a variety of synthetic arylglycosides and unmodified disaccharides (see supplemental Table S1). Of the arylglycosides tested, SpGH125, displayed activity only on DNP-Man. Assessment of the kinetic parameters describing DNP-Man hydrolysis was complicated by an inability to reach substrate saturation making it possible only to determine the second-order rate constant, k cat /K m (28). The k cat /K m value was found to be 0.13 (Ϯ 0.01) min Ϫ1 mM Ϫ1 for SpGH125 (supplemental Fig. S1). To provide more general insight into the activity of this family of enzymes, the related protein CpGH125 was also assayed on DNP-Man with the k cat /K m determined to be 0.58 (Ϯ 0.01) min Ϫ1 mM Ϫ1 (supplemental Fig. S1). Thus, both SpGH125 and CpGH125 exhibit relatively low activity on this substrate. The pattern of these enzymes having no observable activity on para-nitrophenyl ␣-D-mannopyranoside but measurable activity on DNP-Man is similar to the class I ␣-mannosidases, although this latter class of enzyme is specific for ␣1,2-mannose linkages (29,30). Unlike all of the currently characterized mannosidases, the activity of SpGH125 and CpGH125 displayed no sensitivity to the addition of EDTA (not shown), indicating a catalytic mechanism that does not rely on common metal ions used by other currently characterized mannosidases. Thus, SpGH125 and CpGH125 represent the founding members of a new family of glycoside hydrolases, family 125, that appears to possess properties distinct from other known ␣-mannosidases.
To confirm the ability of SpGH125 and CpGH125 to act on carbohydrates representing more complete N-glycans and more firmly establish the specificity of the GH125 enzymes CE was employed as a sensitive method to examine activity on high-mannose glycans (Fig. 1B). Treatment of Man9, Man5, and Man3a with SpGH125 or CpGH125 revealed the resistance of these glycans to the activity of these enzymes. The use of a commercial ␣(1,2)/(1,3)-mannosidase to first remove the terminal ␣1,3-linked mannose residues from Man3a resulted in species having a CE mobility between that of Man1 and Man3a and was thus inferred to be Man2a; surprisingly this enzyme appeared to have a low level of ␣1,6-mannosidase activity giving some of the Man1 species. After co-treatment of Man3a with the ␣(1,2)/(1,3)-mannosidase, SpGH125 and CpGH125 were able to completely convert the glycan into Man1 indicating that the ␣1,6-linkage can only be efficiently hydrolyzed by these enzymes after removal of the ␣1,3-linked mannose by the commercial enzyme.
Taken together these results clearly reveal that the GH125 enzymes are specific for ␣1,6-mannosides; however, these enzymes are only able to process this linkage in the mannose core of N-glycans after the ␣1,3-linked mannose is removed. The activity of these enzymes on DNP-Man and the tested glycans is suggestive of a exo-mode of action but this is not unequivocally demonstrated by these experiments. To provide greater insight into the molecular features that govern the activity of SpGH125 and CpGH125 the structures of these proteins were determined using x-ray crystallography.
The Structural Basis of ␣1,6-Mannoside Recognition-The structures of SpGH125 and CpGH125 were initially solved by molecular replacement in their uncomplexed forms to 2.15 and 2.35 Å, respectively. These structures reveal the expected (␣/␣) 6 -fold previously observed for the homologous hypothetical proteins from C. perfringens (PDB code 2nvp), B. thetaiotaomicron (PDB code 2p0v), and B. ovatus (PDB codes 3p2c and 3on6) (Fig. 2). In keeping with their high amino acid sequence identity of ϳ40%, SpGH125 and CpGH125 displayed a root mean square deviation of 1.2 Å. Not surprisingly, the structure of CpGH125 showed a quite low root mean square deviation of 1.2 Å with the structure of the deposited homolog from C. perfringens strain 13 (PDB code 2nvp), which has 100% sequence identity with CpGH125 but was crystallized in a different space group. Insight into how SpGH125 and CpGH125 both recognize and hydrolyze mannooligosaccharides was gained through analysis of three structures in complex with different ligands: SpGH125 in complex with the inhibitor dMNJ (31), a non-productive complex of SpGH125 with ␣1,6mannobiose, and a complex of CpGH125 with a non-hydrolyzable methyl-S-(␣-D-mannopyranosyl)-(1-6)-␣-D-mannopyranose (32).
The structure of SpGH125 in complex with dMNJ was determined to 1.75-Å resolution and revealed clear electron density for the inhibitor (Fig. 3A). All of the hydroxyl groups of the inhibitor are engaged in hydrogen bonds with residues in the SpGH125 active site (Fig. 3A). Notably, the inhibitor hydroxyl groups that would be the equivalent of the O2, O3, and O4 hydroxyls on mannose interact with Arg 62 , Asp 63 , Asn 302 , and with the Pro 216 main chain carbonyl and are buried deep in the active site (Fig. 3, A and B). This arrangement is consistent with occupation of the Ϫ1 subsite by a terminal mannose residue, it also precludes the recognition of mannose residues that are further modified on their non-reducing end and thus supports the proposal that the GH125s are exo-mannosidases.
Aglycon recognition appears to be a critical feature of the GH125 enzymes as they lack activity on the para-nitrophenyl-Man and have very poor activity on DNP-Man, despite these substrates possessing chemically activated leaving groups. Furthermore, these enzymes appear to show selectivity for terminal, unbranched ␣1,6-dimannose motifs, again indicating a strict aglycon requirement. To better understand the aglycon features required by the GH125 enzymes, a complex was obtained of CpGH125 bound to methyl-S-(␣-D-mannopyranosyl)-(1-6)-␣-D-mannopyranose. This 2.05-Å resolution structure revealed unambiguous electron density for the sugar residues occupying the Ϫ1 and ϩ1 subsites of the CpGH125 active site but the methyl group was too disordered to be modeled (Fig. 3C). The non-reducing end of this disaccharide fits into the pocket of the Ϫ1 subsite and makes a range of interactions with the protein that are identical to those observed in the SpGH125-dMNJ complex. The reducing terminal mannose residue sits in the ϩ1 subsite and engages the side chains of Glu 129 , Lys 131 , Arg 188 , and Thr 194 in an extensive hydrogen bond network; this subsite is also absolutely conserved with SpGH125. The Ϫ1 subsite is separated from the ϩ1 subsite by ϳ6.5 Å and, like the Ϫ1 subsite, the ϩ1 subsite almost fully encloses the mannose residue, leaving only room for additional substituents linked through the reducing terminal O1 group (Fig. 3D). Thus the architecture of the ϩ1 subsite clearly excludes modification of this portion of the aglycon with an ␣1,3-linked mannose branch. Furthermore, the separation of the subsites is roughly 1.5 Å larger than expected to separate subsites accommodating ␣1,3-linked or ␣1,2-linked mannobiose and thus confers specificity toward the more extended ␣1,6-linkage that positions the two mannose residues further apart.
The structure of the GH125 active site is uniquely tailored to accept a terminal ␣1,6-linked dimannose motif in its Ϫ1 and ϩ1 subsites. However, the demonstrated activity of these enzymes on a larger glycan, i.e. Man2a, reveals the ability of the enzymes to tolerate additional reducing end sugars and the possibility of additional subsites on the aglycon side of the cleaved bond. Consideration of the CpGH125-thio-sugar complex shows that the O1 of the mannose bound to the ϩ1 subsite (Fig.  3D) can likely accommodate a sugar residue such as the N-acetylglucosamine that is ␤1,4-linked to the first mannose in the N-linked glycan core. A catalytically non-productive complex of SpGH125 with ␣1,6-mannobiose, fortuitously trapped by raising the pH of the ␣1,6-mannobiose containing solution used to soak the crystal, provides some additional insight into recognition of additional saccharide residues. In this 2.1-Å resolution structure the ␣1,6-mannobiose was well ordered with the non-reducing end mannose occupying the ϩ1 subsite where the interactions were identical to those seen in the CpGH125 thio-sugar complex (Fig. 3E). Although well ordered and in the same conformation for both molecules in the asymmetric unit, the reducing end mannose residue made no direct interactions with the protein (Fig. 3, E and F). This mode of accommodating this disaccharide indicates that various structures terminating with Man␣1-6Man␤1-OR (where R could be various structures) such as Man␣1-6Man␣1-6Man␤1-4GlcNAc2-Asn or Man␣1-6Man␣1-6(Man␣1-3)Man␤1-4GlcNAc2-Asn. Thus, the S. pneumoniae and C. perfringens GH125 enzymes appear to lack a defined ϩ2 subsite, which should provide some flexibility in accommodating various structures.
Comparison of GH125 Structures-Comparison of SpGH125 and CpGH125 structures with the homologs from B. ovatus and B. thetaiotaomicron gives root mean square deviations of ϳ1.5 Å for all of the comparisons, consistent with the ϳ40% amino acid sequence identity they all share (the example from C. perfringens strain 13, PDB code 2nvp, was omitted due to its 100% sequence identity with CpGH125). All of these proteins have absolutely conserved Ϫ1 and ϩ1 subsites indicating that the structurally related Bacteroides proteins are almost certainly also ␣1,6-mannosidases (supplemental Fig. S3).
␣-Glycoside Hydrolysis on a Conserved Platform-All of the GH125 enzymes have a high degree of structural similarity (root mean square deviations of ϳ2.7-3.0 Å) but low amino acid sequence identity (Ͻ14%) with members of glycoside hydrolase family 15. A comparison of the active site of CpGH125 with the GH family 15 Arthrobacter globiformis glucodextranase (G1d; PDB code 1ulv) (33) shows not only the fold-similarity but the conserved location of the active site (Fig.   4A). G1d and CpGH125 recognize ␣-configured glucose and mannose, respectively, in their Ϫ1 subsites; nevertheless, the Ϫ1 subsites of the two enzymes show remarkable conservation revealing the shared molecular features involved in recognition of an ␣-configured pyranose ring having equatorial groups at C3, C4, and C5 (Fig. 4B). The primary difference between the two different families of enzymes in this subsite is in regards to their recognition of O2, the orientation of which distinguishes glucose from mannose. In CpGH125 the axial O2 of mannose makes a hydrogen bond with Asn 302 . This interaction is not present in G1d but instead Arg 567 of this protein makes a hydrogen bond with the equatorial O2 of glucose. In contrast, there is virtually no conservation between CpGH125 and G1d in the ϩ1 subsites, which reflects the quite different aglycon requirements of these two families of enzymes (Fig. 4C).
In addition to conservation in the Ϫ1 subsite, the known catalytic residues in GH family 15, as represented by G1d, are conserved in the GH125 enzymes, as is the position of a putative nucleophilic water molecule. In G1d, Glu 430 acts as the catalytic acid, whereas Glu 628 acts as the catalytic base that activates a strategically positioned water molecule. In the CpGH125 structure, the ␣1,6 S-linkage spans the Ϫ1 and ϩ1 subsites where the carboxylate groups of Asp 220 and Glu 393 are structurally conserved with the GH15 catalytic residues, suggesting roles as catalytic acid and base, respectively, in the GH125 mechanism. The carboxylate group of Asp 220 in CpGH125 is 3.0 Å from the sulfur of the S-linkage and appropriately oriented to donate a proton, consistent with a predicted role as the general acid (Fig.  4D). In the Ϫ1 subsite Glu 393 coordinates a water molecule that sits 3.2 Å directly beneath C1 of the mannose residue, which itself adopts a 4 C 1 conformation. This water molecule is perfectly positioned to attack the anomeric center and displace the leaving group; an identically positioned water is present in the SpGH125-dMNJ complex (Fig. 4D). Thus, elements of the GH125 active sites, including the catalytic machinery, are highly conserved within the active sites of GH15 enzymes, strongly suggesting that GH15 and GH125 enzymes have in common an inverting catalytic mechanism. Indeed, the shared catalytic features of the GH125 enzymes and GH15 indicate that family GH125 is part of GH-L in the glycoside hydrolase clan classification (2).

The GH125 Enzymes Use an Inverting Catalytic Mechanism-
To provide clear support for the structure-based analysis of the GH125 catalytic mechanism, 1 H NMR spectroscopy was used to monitor SpGH125 (Fig. 5) and CpGH125 (see supplemental Fig. S4) catalyzed cleavage of methyl 6-O-(␣-D-mannopyranosyl)-␤-D-mannopyranoside as a function of time. This analysis revealed that formation of the ␤-hemiacetal of mannose preceded the appearance of the ␣-hemiacetal (Fig. 5). This confirmed a catalytic mechanism wherein the ␣-glycosidic linkage is hydrolyzed via an inverting mechanism. In light of the observed architecture of the CpGH125 and SpGH125 active sites, and the similarity of this to the GH15 active sites, the general acid catalytic residue can be assigned as Asp 218 in SpGH125 and Asp 220 in CpGH125. In this enzyme family this residue is in a position enabling it to deliver a proton from the anti-trajectory (34). Likewise, the general base residue can be assigned as Glu 393 in both SpGH125 and CpGH125.
In none of the five SpGH125 and CpGH125 structures was a metal ion observed in the active site, nor was there any effect of EDTA on the activity of either enzyme. Thus, a notable feature of the GH125 catalytic mechanism is its lack of a participating metal ion, which distinguishes it from currently well characterized ␣-mannosidases. Mannosidases in families GH38, GH47, and GH92, which are structurally and mechanistically distinct and hydrolyze a variety of ␣-mannosidic linkages, all require the coordination of a divalent metal ion in the catalytic site for activity. The divalent metal ion in the active site is thought to aid in distorting the substrate and stabilize the transition state by bridging the O2 and O3 hydroxyl groups of the mannoside bound in the Ϫ1 subsite (3,4,35). The absence of a participating metal ion in the GH125 mechanism reveals that divalent cation involvement is not a general requirement for efficient hydrolysis of ␣-mannosides.
GH125s Are Widely Distributed-The distribution of SpGH125 and CpGH125 homologues was tested by BLAST extraction of related sequences from the non-redundant protein data base at NCBI. 360 related proteins were identified with significant amino acid sequence similarity to SpGH125 and CpGH125 (E-value Ͻ10 Ϫ5 ). These sequences fall into a wide variety of prokaryotic and eukaryotic phyla.
A multiple sequence alignment was generated using 99 amino acid sequences that were chosen to be representative of the distribution across the Firmicutes, Bacteroidetes, and Fungi, which were the most highly represented phyla. Overall, the sequences are quite highly conserved with pairwise amino acid sequence identities between 26 and 94%. A phylogenetic analysis, based on this data set, shows that the GH125 family forms distinct clades separating the sequences from Bacteroidetes, Firmicutes, and Fungi (Fig. 6A). Using a structurebased sequence alignment we found conservation of catalytic residues Asp 218 and Glu 393 in SpGH125 (Asp 220 and Glu 393 in CpGH125) throughout all of the sequences. More general conservation of surface-exposed amino acid residues was analyzed by mapping conservation scores for each amino acid onto the structure of SpGH125 (Fig. 6, B and C). This revealed near absolute conservation of residues in the binding subsites. In contrast, surface residues distal to the active site are poorly conserved. This combined phylogenetic and structural analysis predicts that specificity for the ␣1,6-mannosidic linkage is very highly conserved among the GH125 family members.
Microbial N-Glycan Deconstruction-In N-glycans, ␣1,6mannose branches always co-occur with ␣1,3-mannose branches. GH125 enzymes, however, do not appear to tolerate the ␣1,3-mannoside branches generating the requirement for , the substrate monitored by 1 H NMR spectroscopy. B, mannose, the product monitored by 1 H NMR spectroscopy. C, the SpGH125 catalyzed cleavage of Man␣1-6Man␤1-OMe was monitored as a function of time by 1 H NMR spectroscopy. A stacked plot shows hydrolysis of the substrate, with S H1␣ representing the resonance of the anomeric proton of the non-reducing, terminal mannose unit of Man␣1-6Man␤1-OMe, to first form the mannose hemiacetal having the ␤-configuration at the anomeric center, indicated by P H1␤ . P H1␤ represents the anomeric proton of the ␣-anomer of the mannose product and arises from spontaneous mutarotation of the first formed ␤-anomer. D, graphical representation of the anomeric ratio of S H1␣ (OE), P H1␤ (f), and P H1␣ (ࡗ) with respect to time illustrate that the ␤-anomer is formed first.
pre-processing of these substrates with an ␣1,3-mannosidase prior to GH125 action. This suggests the need for the evolutionary co-retention of an ␣1,3-mannosidase with a GH125. Indeed, in the Firmicutes, except for C. perfringens, all of the bacteria possessing a gene encoding a GH125 also have a gene encoding a GH38 mannosidase immediately adjacent to the GH125-encoding gene (C. perfringens does have two GH38 encoding genes that are elsewhere in the genome). SpyGH38 from the Firmicute S. pyogenes was demonstrated to be a specific exo-␣1,3-mannosidase that is tolerant of substrates having the ␣1,6-branches found in N-linked glycans (3). The GH125linked GH38s in the Firmicutes share no less than 42% overall amino acid sequence identity to SpyGH38 and have highly conserved active site residues suggesting that these enzymes have similar, if not identical, specificities to SpyGH38 (not shown). Therefore, in the N-glycan degrading Firmicutes, this implies a model whereby the ␣1,3-branches in N-glycans are necessarily first processed by a GH38 followed by the activity of a GH125 to remove exposed and terminal ␣1,6-mannose residues. This model is also supported by our studies of N-glycan processing using GH125 in combination with another mannosidase. This particular duo of enzymes is ideally suited to process the Man3GlcNAc2 core of N-glycans. The biological relevance of this remains largely un-investigated; however, it is known that SpGH125 is a pneumococcal virulence factor, as is its accompanying putative GH38 (SP_2143), indicating that N-glycan degradation is important to the pneumococcus-host interaction (36,37). The occurrence of these enzymes among other notable pathogens, such as Listeria monocytogenes, C. perfringens, S. pyogenes, and other streptococci, suggest that N-glycan depolymerization is of relatively widespread importance to pathogens that infect a broad variety of human tissues.
The Bacteroides sp., with B. thetaiotaomicron being the primary model at present, are notable in that many are symbiotic inhabitants of the human gut. B. thetaiotaomicron is known to metabolize human glycans lining the gut under conditions of limited dietary polysaccharide intake by the host (38). The large expansion of ␣-mannosidases in this bacterium is thought to contribute to this ability by enabling the metabolism of N-glycans (4). These enzymes, however, lack substantial ␣1,6-mannosidase activity, which would be required for complete N-glycan deconstruction. GH125s are relatively common among the Bacteroides sp., including three GH125s in B. thetaiotaomicron, suggesting these enzymes play a role in completing the machinery required for this bacterium, and possibly related gut symbionts, to completely depolymerize N-glycans for use as a carbon source. FIGURE 6. Phylogenetic analysis of GH family 125. A, the phylogenetic tree for family GH125 was inferred using the Minimum Evolution method. Phylogenetic analyses were conducted in MEGA4 (42)(43)(44)(45). B and C, conservation of surface residues in family GH125. The SpGH125 structure is shown in surface representation from back (B) to front (C) with the substrate in the active site, which was modeled based on the SpGH125-dMNJ complex and the CpGH125substrate complex. The surface is colored according to conservation scores for the surface residues (variable in blue to highly conserved in purple; see legend below panel C). Conservation of residues was scored based on the alignment used for phylogenetic tree construction and calculated using Consurf (26,27).
The function of GH125 enzymes in the fungi that contain them is less clear. These organisms can extensively post-translationally glycosylate their secreted proteins; both the N-and O-glycans can have high-mannose content. Thus, the GH125 enzymes in these organisms may play a role in the maturation of their glycans during synthesis or in degrading their own glycans as a mechanism of recycling. Alternatively, they may aid in foraging for mannose from the environment.
Conclusion-A surprisingly small number of enzymes with specific ␣1,6-mannosidase activity have been identified. A human core-specific lysosomal ␣1,6-mannosidase that removes the ␣1,6-mannose residue from Man3GlcNAc but not Man3GlcNAc2 indicates that it cannot be active on N-glycans still linked to proteins (39). A mannosidase from B. thetaiotamicron (Bt3994) hydrolyzes the distal ␣1,6-mannosidic linkage in high-mannose N-glycans irrespective of the presence of the ␣1,3-mannose arm; however, this enzyme also shows some activity toward ␣1,3-mannobiose (4). These examples highlight the unique specificity of GH125, which is made more remarkable as this specificity appears to be shared among the large number of family members. Furthermore, the catalytic platform used by GH125 closely resembles that used by glucoamylases, possibly indicating a common ancestry; however, future mechanistic studies will be required to clarify the role of active site residues in the GH125 family. Nevertheless, the active site architecture clearly indicates that the GH125 enzymes utilize a catalytic mechanism that does not rely on the participation of a metal ion, making GH125 the only presently characterized metal-independent ␣-mannosidase family. Significantly, this reveals that metal-assisted catalysis is not a required feature of ␣-mannoside hydrolysis.