Molecular mechanism of Aspergillus fumigatus biofilm disruption by fungal and bacterial glycoside hydrolases

During infection, the fungal pathogen Aspergillus fumigatus forms biofilms that enhance its resistance to antimicrobials and host defenses. An integral component of the biofilm matrix is galactosaminogalactan (GAG), a cationic polymer of α-1,4-linked galactose and partially deacetylated N-acetylgalactosamine (GalNAc). Recent studies have shown that recombinant hydrolase domains from Sph3, an A. fumigatus glycoside hydrolase involved in GAG synthesis, and PelA, a multifunctional protein from Pseudomonas aeruginosa involved in Pel polysaccharide biosynthesis, can degrade GAG, disrupt A. fumigatus biofilms, and attenuate fungal virulence in a mouse model of invasive aspergillosis. The molecular mechanisms by which these enzymes disrupt biofilms have not been defined. We hypothesized that the hydrolase domains of Sph3 and PelA (Sph3h and PelAh, respectively) share structural and functional similarities given their ability to degrade GAG and disrupt A. fumigatus biofilms. MALDI-TOF enzymatic fingerprinting and NMR experiments revealed that both proteins are retaining endo-α-1,4-N-acetylgalactosaminidases with a minimal substrate size of seven residues. The crystal structure of PelAh was solved to 1.54 Å and structure alignment to Sph3h revealed that the enzymes share similar catalytic site residues. However, differences in the substrate-binding clefts result in distinct enzyme-substrate interactions. PelAh hydrolyzed partially deacetylated substrates better than Sph3h, a finding that agrees well with PelAh's highly electronegative binding cleft versus the neutral surface present in Sph3h. Our insight into PelAh's structure and function necessitate the creation of a new glycoside hydrolase family, GH166, whose structural and mechanistic features, along with those of GH135 (Sph3), are reported here.

solved to 1.54 Å and structure alignment to Sph3 h revealed that the enzymes share similar catalytic site residues. However, differences in the substrate-binding clefts result in distinct enzyme-substrate interactions. PelA h hydrolyzed partially deacetylated substrates better than Sph3 h , a finding that agrees well with PelA h 's highly electronegative binding cleft versus the neutral surface present in Sph3 h . Our insight into PelA h 's structure and function necessitate the creation of a new glycoside hydrolase family, GH166, whose structural and mechanistic features, along with those of GH135 (Sph3), are reported here.
Aspergillus fumigatus is a ubiquitous filamentous fungus that causes invasive pulmonary infections in patients who are immunosuppressed due to cytotoxic chemotherapy, organ and stem cell transplantation, or biological therapies such as tumor necrosis factor inhibitors (1)(2)(3). During pulmonary infection, filamentous hyphae of A. fumigatus grow within biofilms: multicellular communities of organisms embedded in a self-produced extracellular matrix (4). This biofilm extracellular matrix plays several roles in the pathogenesis of invasive aspergillosis including mediating the adherence of hyphae to host tissues and enhancing resistance to antifungal drugs and host immune defenses (5)(6)(7)(8).
Solid-state NMR spectroscopy analyses have revealed that polysaccharides are the most abundant component of the A. fumigatus extracellular matrix (9). Immunohistochemical and EM studies of Aspergillus biofilms have identified ␣-1,3glucan, galactomannan, and galactosaminogalactan (GAG) 6 as key polysaccharide components of the matrix (10). GAG is a linear cationic polymer of ␣-1,4-D-galactose and partially deacetylated ␣-1,4-N-acetyl-D-galactosamine (GalNAc) that is produced by activity of the products of a five-gene cluster located on chromosome 3 (11). GAG synthesis is thought to be initiated by the synthesis of UDP-GalNAc and UDP-Gal by the glucose-4-epimerase Uge3 (12). These sugars are then linked and exported through the action of the predicted glycosyltransferase Gtb3, and the resulting polymer is partially deacetylated in the extracellular space by secreted Agd3 (13). Cleavage of the emerging polymer is thought to be mediated by two glycoside hydrolases encoded within this gene cluster, Sph3 and Ega3 (11). GAG, plays a central role in biofilm formation and virulence (14). GAG-deficient strains are markedly impaired in their ability to form biofilms and adhere to pulmonary epithelial cells in vitro (15). Hyphae-associated GAG enhances resistance to neutrophil extracellular traps (16) and conceals pathogen-associated molecular patterns such as ␤-glucan from immune detection (17). Secreted GAG also promotes infection by inducing neutrophil apoptosis and the production of the anti-inflammatory cytokine IL-1 receptor antagonist (18). Consistent with these observations, GAG-deficient strains of A. fumigatus exhibit attenuated virulence in mouse models of invasive aspergillosis (17).
The importance of GAG in biofilm formation and virulence suggest that this exopolysaccharide is a promising therapeutic target. We recently reported that the recombinant glycoside hydrolase domains from two microbial proteins, A. fumigatus Sph3 and Pseudomonas aeruginosa PelA, degrade GAG and disrupt A. fumigatus biofilms (19). PelA is a multidomain protein with both deacetylase and glycoside hydrolase activity that is required for production of the P. aeruginosa Pel polysaccharide (20 -22). In addition to its degrading A. fumigatus biofilms, the recombinant hydrolase domain of PelA (PelA h ) also disrupts Pel-dependent biofilms of P. aeruginosa (21). Although, the structure and anomeric configuration of the Pel polysaccharide remains to be determined, it has been found to contain partially deacetylated 1,4-linked GalNAc and N-acetylglucosamine (GlcNAc) in a ratio of 5:1 (20,23). The presence of GalNAc-rich regions in both GAG and Pel polysaccharides suggests that PelA h may also be specific for GalNAc.
Structural and functional characterization of the glycoside hydrolase domain of Sph3 (Sph3 h ) revealed that this enzyme is essential for GAG biosynthesis and belongs to the glycoside hydrolase (GH) family 135 (24). Sph3 h has an (␤/␣) 8 -fold with a shallow conserved active site groove. Co-crystallization of Sph3 h with GalNAc and mutagenesis of residues within the active site groove suggest that this enzyme hydrolyzes GAG within GalNAc-rich regions of the polymer (24).
Less is known about structure and function of the glycoside hydrolase domain of PelA, which is predicted by Phyre 2 to contain a (␤/␣)-barrel fold (20) and by the conserved domain database to belong to the GH114 superfamily (www.ncbi.nlm. nih.gov/Structure/cdd/wrpsb.cgi). However, a BLAST search against all CAZy, GH114 members, and a search of the sequence against all Hidden Markov Models of CAZy families, failed to find any significant sequence identity with classified glycoside hydrolases, suggesting that PelA h belongs to a new CAZy family.
Herein, we characterize the molecular mechanisms underlying the cross-kingdom GAG activity of the glycoside hydrolases, Sph3 h and PelA h . Mass spectrometry enzymatic fingerprinting and NMR studies revealed that both enzymes share retaining endo-␣-1,4-N-acetylgalactosaminidase activity and require a minimal substrate length of seven GalNAc residues. The substrate interacts with the binding cleft differently in the two enzymes leading Sph3 h to cleave substrates closer to the nonreducing end, whereas PelA h cleaves proximal to the reducing end of oligosaccharides. The enzymes also differ in their ability to cleave partially deacetylated substrates. Although neither enzyme is active against fully deacetylated oligosaccharides, PelA h has a higher propensity to cleave GalNAc linkages within regions of partially deacetylated GAG. The structure of PelA h was determined, and comparison with Sph3 h revealed a high degree of structural similarity within the catalytic site of these enzymes. The presence of a deeper, more electronegative groove in PelA h likely underlies its ability to bind and cleave cationic partially deacetylated substrates more effectively than Sph3 h .

Sph3 h and PelA h are ␣-1,4-N-acetylgalactosaminidases
To elucidate the molecular mechanisms by which Sph3 h and PelA h mediate biofilm disruption, these enzymes were incubated with pre-grown A. fumigatus biofilms and the profile of oligosaccharides that were released was analyzed by matrixassisted laser desorption and ionization time of flight mass spectrometry (MALDI-TOF MS) enzymatic fingerprinting. Consistent with our previous studies, nanomolar concentrations of Sph3 h or PelA h disrupted A. fumigatus biofilms (Fig. 1,  A and B) (19). Analysis of the oligosaccharides released by treatment with either enzyme revealed similar MS fingerprints, with spectra displaying ions of m/z ratio corresponding to N-acetylhexosamine (HexNAc) oligomers. Treatment of biofilms with 100 nM Sph3 h produced ions with a m/z ratio from 1056.5684 to 2478.2696 with a repetitive occurrence of 203.1002 Ϯ 0.0059 matching with the HexNAc m/z ratio (25). This profile is consistent with the release of HexNAc oligomers ranging in size from pentamers to dodecamers (12-mers) (Fig. 1A). Treatment of fungal biofilms with 100 nM PelA h generated a similar spectrum of ions ranging from m/z 1056.6345 to 3290.8078, with an interval m/z difference of 203.1107 Ϯ 0.0053 between ions, suggesting that PelA h treatment released HexNAc oligosaccharides ranging in size from pentamers to hexadecamer (16-mers) (Fig. 1B). MS-MS fragmentation analysis confirmed these ions were composed uniquely of m/z ratio 203.10 units, suggesting that these oligosaccharides are HexNAc homopolymers (Fig.  1C). GC coupled to MS (GC-MS) analysis of the monosaccharide composition of these oligosaccharides confirmed they are composed solely of GalNAc, supporting the hypothesis that these enzymes cleave regions of homo-GalNAc within GAG (Fig. 1D).
To further confirm the specificity of Sph3 h and PelA h for the ␣-1,4-GalNAc linkages within GAG, the ability of these enzymes to cleave synthetic oligo-␣-1,4-galactose, and oligo-␣-1,4-galactosamine were tested. Neither Sph3 h nor PelA h exhibited activity against these components of GAG, even at concentrations as high as 5 M (Fig. 2, A and B). Similarly, neither enzyme was able to degrade chitin (␤-1,4 -linked GlcNAc), the only other major hexosamine-containing polysaccharide Molecular mechanism of glycoside hydrolases Sph3 h and PelA h within the fungal cell wall (Fig. 2C). These data suggest that biofilm disruption by Sph3 h and PelA h is a consequence of cleavage of ␣-1,4-GalNAc homopolymeric regions within GAG.
To determine the minimum substrate size that can be cleaved by Sph3 h and PelA h , ␣-1,4-GalNAc hexamers and heptamers purified from partial Sph3 h digestion of A. fumigatus biofilms were treated with each enzyme and the degradation products were analyzed by MALDI-TOF MS fingerprinting.

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h
GalNAc heptasaccharides but not hexasaccharides were rapidly hydrolyzed by both enzymes (Fig. 3, A-D). Hydrolysis of GalNAc heptasaccharides by Sph3 h resulted in the accumulation of pentasaccharides (Fig. 3B), whereas PelA h hydrolysis produced predominantly both tetra-and pentasaccharides (Fig. 3D), suggesting that these enzymes function as endoacting glycoside hydrolases. The ability of PelA h to degrade the heptamers into two sets of products suggests some flexibility in the positioning of the oligosaccharide in the PelA hbinding site.

Sph3 h and PelA h are retaining endoglycoside hydrolases
To investigate the molecular mechanism of the two glycoside hydrolases, the stereochemical outcome of oligosaccharide cleavage by these enzymes was evaluated by 1 H NMR spectroscopy using synthetic GalNAc octamers as a substrate. Spectral analysis of the pre-reaction substrate demonstrated the pres-ence of four doublets with coupling constants between 3.5 and 4.0 Hz in the anomeric region of the spectra, characteristic of ␣-glycosidic linkages. The addition of Sph3 h or PelA h led to the appearance of a new peak at 5.29 ppm with a coupling constant of 4.0 Hz characteristic of an ␣-anomer (Fig. 4, A and B). Acquisition of 1D spectra at 25 and 37°C (Fig. 4, A and B), and of 2D COSY 1 H-1 H spectra (Fig. S1) after 24 h of reaction revealed the appearance of a ␤-anomer signal at 4.73 ppm on the 1D spectra consistent with secondary mutarotation at the new reducing end. The addition of purified monomeric GalNAc to samples at the end of the experiment resulted in the appearance of anomeric signals at 4.66 and 5.25 ppm. These signals were distinct from the experimental products, demonstrating that neither Sph3 h nor PelA h released monosaccharides from GalNAc octamers (Fig. S2). Collectively, these data suggest that Sph3 h and PelA h are retaining endoglycosidases.

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h PelA h has a (␤/␣)-barrel with a deep substrate-binding groove
To shed light on the mechanisms by which PelA h and Sph3 h hydrolyze ␣-1,4-GalNAc polymers, and to complement the available structure of Sph3 h , we pursued the structure determination of PelA h . Crystals of PelA h formed readily in a variety of different crystallization conditions and the structure was subsequently determined using selenomethionine incorporation and the single-wavelength anomalous diffraction method to 1.54 Å and refined to a final R work and R free of 16.9 and 19.2%, respectively. Unambiguous electron density allowed the modeling of residues 48 to 300. No interpretable density was observed for residues 47, 301-303, or the hexa-histidine purification tag. The structure of PelA h revealed a ␤ 8 /␣ 7 -barrelfold with a C-terminal region lacking the secondary structure ( Fig. 5A). This region, residues 290 -300, packs against the core ␤-barrel and makes multiple contacts to neighboring helices suggesting that the lack of helical structure is not an artifact of the construct design. Furthermore, secondary structure prediction using Jpred4 (26) predicts that residues 297-310 form a ␤-strand supporting the lack of an 8 helix.
In addition to the core (␤/␣)-fold, PelA h contains two structural insertions. The first is a 38-amino acid loop between ␤3 and ␣3 containing a two strand anti-parallel ␤-sheet and two ␣-helices (Loop3 in Fig. 5A). The second is a ␤-hairpin loop (␤HL) between ␤6 and ␣6. These insertions contribute to the creation of a 15-Å deep, ϳ36 Å long, active site groove on the "top" face of the barrel at the C termini of the ␤-strands (Fig.  5A). This groove is composed of several aromatics that could potentially participate in polysaccharide-protein interactions (Fig. 5B).
When the structure of PelA h was submitted to the tertiary structure similarity server, DALI (27), the most similar structures found were: (i) a hypothetical protein TM1410 from Ther-matoga maritima, which has sequence similarity to GH114 family members (PDB 2AAM, 2.5 Å root mean square deviation (RMSD) over 203 residues); (ii) a cycloisomaltooligosaccharide glucotransferase from GH66 (PDB 3WNK, 2.9 Å RMSD over 165 residues); (iii) Cwp19, a peptidoglycan hydrolase reported as a GH-like 10 family member (PDB 5OQ2, 2.9 Å RMSD over 193 residues), and (iv) a dextranase from GH66 (PDB 3VMN, 2.9 RMSD over 155 residues). TM1410 has the highest similarity in structure and highest sequence identity to PelA h (15.7%). TM1410 also contains an insertion after ␤3, consisting of a three-stranded ␤-sheet and a small ␣-helix (Loop3 TM1410 ). Loop3 TM1410 folds further over the putative active site than the equivalent Loop3 in PelA h (Fig. 5C). Electron density corresponding to a ring containing ligand, as well as multiple glycerol molecules, was found in the deep groove of the TM1410 structure. It is possible that Loop3 TM1410 has some flexibility and the presence of a ligand in TM1410 causes the loop to cap the groove. Loop3 PelA and the ␤HL have the highest B-factors of the PelA h structure, reaching 134 Å 2 in Loop3 as compared with an average protein B-factor of 31.6 Å 2 , suggesting conformational heterogeneity in the crystal, further supporting the flexibility of these regions (Fig. 5D).
PelA h is most closely related to GH114 family as previously noted. According to the automated carbohydrate active enzymes annotation server dbCAN2, PelA h has very low overlap with the GH114 HMM profile and has insufficient sequence identity to be assigned to this family (28,29). These results suggest that PelA h represents a new GH family related to GH114. Our experimental evidence that PelA h exhibits glycoside hydrolase activity allows for the creation of the GH166 family. Although no GH114 structure has been solved to date, the similarity in the predicted structure suggests that GH166 and GH114 may create a new GH clan (30).

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h Structural comparison of GAG hydrolyzing PelA h and Sph3 h reveals architectural differences but conserved active-site residues
A superimposition of PelA h and Sph3 h (RMSD of 3.05 Å over 167 C␣ atoms) aligned the active site of Sph3 h (24) with that of PelA h (19) and revealed that the insertions in the PelA h structure are unique to this protein (Fig. 6A). Sph3 h has a much shallower binding cleft compared with PelA h (Fig. 6B). Active sites situated in clefts and tunnels suggest endo-activity and correlates with increased processivity (31)(32)(33). Shallow substrate grooves correlate with low processivity but have been found in enzymes that cleave crystalline, recalcitrant substrates (31)(32)(33). The deeper cleft found in PelA h is indicative of a more processive activity as compared with Sph3 h .
Despite differences in cleft architecture, examination of the putative active sites (RMSD 0.25 Å over active site motif backbone atoms) revealed a high degree of conservation, with residues Asp-166, Asn-202, and Glu-222 of Sph3 h that interact with GalNAc in our co-crystal structure (24), superimposing with residues Asp-160, Asn-199, and Glu-218 of PelA h (Fig. 5, C  and D). Sph3 h activity was previously shown to be dependent on Asp-166 as conservative mutation of this residue abolished Sph3 h degradation of GAG (24). Mutation of Glu-222 to alanine also abolished Sph3 h activity but replacement with glutamine had a lesser affect. Biofilm disruption by PelA h was shown to involve Glu-218, which aligns to Glu-222 of Sph3 h , suggesting this residue is involved in catalysis (19). The similarity of the residues within the catalytic site of Sph3 h and PelA h suggests

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h
that they share a common catalytic mechanism. The distance between carboxyl side chains of the catalytic residues Asp-166 and Glu-222 in Sph3 h and the homologous residues Asp-160 and Glu-218 in PelA h are 5.2-6.7 and 4.8 -6.8 Å, respectively. This distance is consistent with a retaining (ϳ5.5 Å) rather than inverting mechanism of cleavage (ϳ10 Å) (32).

Sph3 h and PelA h have different distribution of substratebinding subsites
To determine whether Sph3 h and PelA h cleave GalNAc oligosaccharides closer to the reducing or nonreducing end, an oligosaccharide preparation enriched in (␣-1,4-GalNAc) 9 was reduced by sodium borohydride treatment, conferring an additional m/z ratio of 2 to the reducing end. These reduced oligosaccharides were then incubated with each enzyme and the resulting products analyzed by MALDI-TOF MS. Both enzymes produced multiple oligosaccharide products ranging in length from tetramers to heptamers. However, oligosaccharides produced by Sph3 h treatment were consistent with cleavage of (␣-1,4-GalNAc) 9 near the reducing end of the substrate, whereas PelA h treatment released reduced GalNAc 4 -6 consistent with cleavage near the nonreducing end (Fig. 7).
Mapping of sequence conservation based on alignments of Sph3 h and PelA h to respective homologous proteins revealed high degrees of sequence conservation in the active-site groove of each enzymes (Fig. 7, C and D). However, closer examination of the patterns of conservation shows differences between the two enzymes (Fig. 7, C and D). For Sph3 h , surface conservation extends further on the nonreducing side of ϩ1/Ϫ1 cleavage site suggesting approximately five conserved substrate-binding subsites (Ϫ5 to Ϫ1, Fig. 7C). This observation is consistent with our data that shows that this enzyme produces a minimum length pentasaccharide from the nonreducing end (Fig. 7A). In contrast, PelA h has surface conservation that extends on the reducing end side of cleavage and at least five subsites could be mapped (ϩ1 to ϩ5, Fig. 7B). Surface residue conservation thus correlates well with the results of the GalNAc 9 hydrolysis experiments.

PelA h cleaves deacetylated-rich regions within GAG
Comparison of the structures of Sph3 h and PelA h revealed that the PelA h active site groove is more electronegative than that of Sph3 h (Fig. 8A), suggesting that PelA h may be able to bind and potentially cleave cationic oligosaccharides, such as partially deacetylated oligo-GalNAc. Although neither enzyme was able to cleave ␣-1,4 GalN homopolymers, it is possible that these enzymes may be able to cleave ␣-1,4 GalNAc linkages within GalN-rich regions of GAG. Although GalN-containing oligosaccharides were not detected during the biofilm disruption assay (Fig. 1), these cationic degradation products could have remained adherent to the negatively charged hyphal cell wall. To test this hypothesis, a cell-free GAG degradation assay was performed using purified secreted GAG. Sph3 h treatment of secreted GAG resulted in the release of predominately homo-GalNAc oligosaccharides (86%) with a smaller amount of mono-deacetylated GalNAc oligosaccharides (14%, Fig. 8B). In contrast, treatment of secreted GAG with PelA h produced almost exclusively partially deacetylated GalNAc oligosaccharides (98%, Fig. 8B) suggesting that PelA h preferentially binds and degrades GalN-containing regions of GAG. Consistent with our prior results, no galactose homo-or heteropolymers, In orange are the three residues (denoted with single letter abbreviations) identified as highly conserved in Sph3 homologues. C, C␣ alignment of PelA h (yellow and blue) to hypothetical protein TM1410 (PDB 2AAM, gray and black) shows similarity in tertiary structure topology. The insertions of PelA h are colored as in A showing that TM1410 also contains these additions including a loop after ␤3 (Loop3 TM1410 , black). An unknown ligand (red) and glycerol (gray) were found in the groove of the TM1410 structure. D, visualization of PelA h structure B-factors, colored blue to red for relative low to high values.

Discussion
We previously described the structure of Sph3 h and found that this enzyme constituted the first member of a new glycoside hydrolase family, GH135, with activity against GAG and Aspergillus biofilms (24). More recently, the glycoside hydrolase domain of PelA from P. aeruginosa has also been reported to disrupt A. fumigatus biofilms (19). However, the molecular mechanism and specificity these enzymes had not been elucidated.
Herein, the mechanism by which these enzymes disrupt A. fumigatus biofilms was studied using a combination of structural biology, and mass and NMR spectroscopy. These studies revealed that Sph3 h and PelA h share a conserved active site and cleave ␣-1,4-GalNAc glycosidic linkages using a retaining enzyme mechanism. Despite these similarities of specificity and mechanism, the interaction between the enzymes and the substrate differ. The Sph3 h has a neutral electrostatic cleft in contrast to the deep electronegative groove of the PelA h . These observations correlate with the nature of the products observed after cleavage of secreted GAG. The electronegative PelA h was able to bind and release cationic GalN-containing oligosaccharide more efficiently, whereas the neutral Sph3 h released mostly neutral GalNAc oligosaccharides. Furthermore, despite interacting with the same substrate, the two glycoside hydrolases have extended conserved substrate-binding sites on opposites sides of the catalytic Ϫ1/ϩ1 site. This difference leads to Sph3 h cleaving closer to the reducing end, whereas PelA h cleaves nearer to the nonreducing end of oligosaccharides.
Sph3 h is structurally similar to GH27 family members, which rely on two acidic residues to cleave the glycosidic bond (24). GH27 catalytic residues align structurally with Asp-166 and Glu-222 in Sph3 h . Previously, mutagenesis studies on Sph3 h identified Asp-166 as essential for catalysis (24). Glu-222 was not essential in vitro as mutation to glutamine retained some activity (24). Replacement of the catalytic acid/base residue with glutamine does not abolish activity in some retaining hydrolases but slows the rate of reaction (34,35). Mutation of the acid/base glutamate in the Sulfolobus solfataricus ␤-glyco-

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h
sidase reduced activity 10 -60 -fold depending on the substrate (36). In Sph3 h , Glu-222 is coordinated by Tyr-88 and Asn-202, and these residues may aid in activation of the glutamine in the E222Q mutant. PelA h has equivalent catalytic residues that align structurally to those of Sph3 h . NMR analysis found that both enzymes utilize a retaining mechanism. Analysis of the structure of Sph3 h and PelA h revealed that the distances between the carboxyl side chains of the catalytic residues Glu-222 and Asp-166 in Sph3 h and Glu-218 and Asp-160 in PelA h correspond to the average distance between the catalytic residues of retaining hydrolases (32). Thus, the structures of Sph3 h and PelA h and the NMR results are consistent with these enzymes utilizing a retaining mechanism (Fig. 4).
Our findings demonstrate that PelA h and Sph3 h share a similar (␤/␣)-barrel-fold with central grooves containing a highly conserved active sites, underlying their endo-␣-1,4-N-acetylgalactosaminidase activity. The high identity between the catalytic residues of the active site suggests that the difference in the ability of these enzymes to hydrolyze oligosaccharides containing deacetylated residues is a reflection of differences in polysaccharide binding affinity. This hypothesis is supported by the fact that, unlike Sph3 h , the binding groove of PelA h is highly electronegative, suggesting an affinity for cationic substrates as seen in the soluble GAG-degradation assay. PelA is a multidomain protein with both hydrolase and carbohydrate deacetylase activity. The deacetylase activity has been shown to be required for in vivo Pel-dependent biofilm formation (20). Our results suggest that the PelA deacetylase domain likely acts on Pel first, rendering the polysaccharide cationic prior to hydrolysis of the polymer. Similar findings have been reported for the poly-␤1-6-N-acetyl-D-glucosamine modifying enzyme, PgaB, which contains both a deacetylase domain and a GH153 hydrolase domain (37). Studies of PgaB demonstrated that the GH153 domain requires a specific pattern of partially deacetylated poly-␤1-6-N-acetyl-D-glucosamine as a substrate (37). PelA h appears to be more promiscuous in its substrate specificity, as the enzyme can hydrolyze pure acetylated oligomers.
Another difference between Sph3 h and PelA h was revealed by their activity on reduced GalNAc nonamers (Fig. 7). Sph3 h hydrolyzed oligosaccharides proximal to the reducing end, whereas PelA h cleaved near the nonreducing end. This observation suggests differences in the roles of the substrate-binding residues on either side of the catalytic site in each enzyme. This hypothesis is substantiated by the differences in surface residue conservation in the putative binding grooves of Sph3 h and PelA h . Sph3 h contains a longer stretch of conserved residues on the nonreducing side of the Ϫ1/ϩ1 cleavage site. The conserved residues in Sph3 h span the Ϫ5 to ϩ2 binding subsites. In contrast, the conserved surface of PelA h aligns with Ϫ2 to ϩ5 subsites.
Previously, we reported that PelA h was able to disrupt both GAG-and Pel-dependent biofilms. In contrast, Sph3 h was active only against GAG-dependent fungal biofilms, and was unable to cleave Pel-dependent biofilms despite being able to bind Pel polysaccharide (19,24). The results of our structurefunction studies reveal important differences in the substrate specificity and structure of these enzymes that may explain these observations. PelA h was found to have a deeper, more

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h
electronegative substrate-binding groove and exhibit preferential activity against partially deacetylated substrates compared with Sph3 h . Although Sph3 h treatment of soluble GAG releases predominately acetylated oligomers, treatment with PelA h releases 98% partially deacetylated products. Although the detailed structure of Pel has not been determined, compositional studies suggest it contains both GlcNAc and GalNAc (23). The degree of deacetylation, and the identity of the sugars deacetylated in the Pel polysaccharide have not been determined. Our results suggest the inactivity of Sph3h against Pel may reflect the fact that Pel is more extensively deacetylated than GAG. Alternately, whereas the activity of PelA against ␣-linked GalNAc suggests that these ␣-linkages are present within Pel, it is possible that the presence of GlcNAc sugars within the Pel polymer may interfere with Sph3 h ability to cleave the Pel polymer. Distinguishing among these possibilities, and confirming the presence of ␣-linked GalNAc in Pel will require a complete structural analysis of this polysaccharide. The results of these studies assign enzymatic function to two new glycoside hydrolases. Sph3 h is the only GH135 family member that has been functionally characterized. Although the structure of Sph3 h had been previously determined (24), the current study expands our understanding of the activity of this enzyme and demonstrates that Sph3 h functions as a retaining endo-␣-1,4-N-acetylgalactosaminidase with specificity for fungal GAG. We also found that the bacterial enzyme PelA h is a retaining endo-␣-1,4-N-acetylgalactosaminidase. As the sequence identity between PelA h and current glycoside hydrolase families is low, our structural and functional characterization of this protein enables us to classify PelA h as the first member of a new family, GH166.

Protein expression and purification of PelA h for structural studies
PelA h encompassing residues 47-303 of the mature PelA h protein was expressed and purified as previously described (21). Selenomethionine (SeMet)-labeled protein was produced as previously described (38) with B834 Met Escherichia coli cells (Novagen) and purified as described for the native protein. 10 4 A. fumigatus conidia were grown in Brian media in polystyrene, nontissue culture treated 96-well-plates for 21 h at 37°C and then treated with the indicated concentration of glycoside hydrolase in 1ϫ PBS for 1 h at room temperature under gentle agitation. Biofilms were then washed, stained with 0.1% (w/v) crystal violet and destained with 100% ethanol for 10 min. The optical density of the destain solution was measured at 600 nm.

MALDI-TOF MS enzymatic fingerprint
Products of enzymatic digestions were diluted in 0.2% trifluoroacetic acid (TFA) before being spotted on the MALDI-TOF plate in a ratio 1:1 (v/v) with 5 mg/ml of dihydroxybenzoic acid matrix reconstituted in acetonitrile (ACN), 0.2% TFA (70: 30, v/v). Spectra were recorded on a Bruker UltrafleXtreme in positive reflector mode and represent an accumulation of 5000 laser shots. MALDI-TOF MS/MS experiments were performed using the same mass spectrometer.

GC-mass spectrometry monosaccharide composition
Oligosaccharides enzymatically extracted from biofilm were hydrolyzed with 6 M hydrochloric acid (HCl) for 4 h at 100°C. After drying, samples were derivatized and analyzed as previously described (39). Briefly, samples were then converted in methyl glycosides by heating in 1 M methanol-HCl (Supelco) for 16 h at 80°C. Samples were dried and washed twice with methanol prior re-N-acetylating hexosamine residues. Re-N-acetylation was performed by incubation with a mix of methanol: pyridine:anhydride acetic (10:2:3) for 1 h at room temperature. Samples were then treated with hexamethyldisilazane:trimethylchlorosilane:pyridine solution (3:1:9, Supelco) for 20 min at 80°C. The resulting TMS methyl glycosides were dried, resuspended in 1 ml of cyclohexane, and injected in the Trace1300 GC-MS system equipped with a CP-Sil5-CB capillary column (Agilent Technologies). Elution was performed with the following temperature gradient: 120 to 160°C at a rate of 10°C/min, 160 to 220°C at a rate of 1.5°C/min, 220 to 280°C at a rate of 20°C/min. Identification and quantification of each monosaccharide were carried out using standards and response factors determined for each monosaccharide.

Production of oligosaccharide of chitin, ␣-1,4-Gal, and ␣-1,4-GalN and specificity study
Oligosaccharides of chitin/chitosan were produced by acidic partial hydrolysis of chitin from shrimp shell (Sigma). Briefly, chitin was incubated in 0.1 M HCl for 2 h at 100°C. Solubilized oligosaccharides were then purified on a Hypersep Hypercarb SPE cartridge (ThermoFisher) conditioned as per the manufacturer's instructions. After loading the sample, the cartridge was washed with water, 5% (v/v) ACN, and oligosaccharides were eluted with 50% (v/v) ACN.
The ␣-1,4-Gal and ␣-1,4-GalN oligosaccharides were chemically synthetized based on the use of di-tert-butylsilylene group-protected building blocks. These extremely powerful cis-galactosylating agents only produced ␣-isomer products. After glycosylation reactions, the di-tert-butylsilylene group was removed with HF-pyridine. The resulting free 6-hydroxyl group was protected with the benzoyl group selectively to afford the acceptor. The final compounds were obtained after deprotection by saponification, debenzylation, azide reduction, and acetylation of amine groups.
All oligosaccharides were incubated with 5 M Sph3 h or PelA h for 1 h at room temperature and analyzed by MALDI-TOF MS. Controls of enzyme functionality were performed using as substrate ␣-1,4-GalNAc oligosaccharides produced as described below.

Enzymatic mechanism determination by 1 H NMR
Proton NMR spectra were recorded on an AVANCE III HD 600 NMR spectrometer (Ascend TM 600 magnet, Bruker Biospin Ltd.) operating at a frequency of 600.17 MHz for 1 h and equipped with a quadruple resonance CryoProbe (CPQCI 1 H-31 P/ 13 C/ 15 N) and a SampleJET TM autosampler. For each Molecular mechanism of glycoside hydrolases Sph3 h and PelA h sample, 0.45 mg of ␣-1,4-GalNAc octamer was resuspended in 160 l of 0.1ϫ PBS in D2O containing 0.5 mM trimethylsilylpropanoic acid and transferred in a 3-mm NMR tube. The 1 H NMR spectra were continuously acquired at 25°C for 4 h with lock and shim performed on every 10 experiments. A new spectrum at 25°C was recorded after 24 h to observe the mutarotation and confirmation of this event was validated with the acquisition of 1 H NMR spectra at 37°C and a 2D 1 H-1 H correlation spectroscopy (COSY) spectra at 25°C. The 1 H spectra were acquired using the pulse sequence noesypr1d (Bruker Biospin Ltd.) to achieve good suppression of the water signal. Each 1 H spectrum was acquired with 32 scans, a 1 H 90°pulse length of 7.8 s, a mixing time of 10 ms, a spectral width of 12 kHz, and a recycle delay of 4 s for a total of 66,000 data points. The 2D 1 H-1 H COSY spectra were acquired using 16 scans with a 1 H 90°pulse length of 8 s, a spectral width of 3 kHz in both dimensions, and a repetition delay of 1.8 s for a total of 2048 data points in F2 and 128 increments in F1. All spectra were processed using TOPSPIN software (version 3.5 pl 7, Bruker Biospin Ltd.).

Production of pure ␣-1,4-GalNAc oligosaccharides
A. fumigatus biofilm were incubated with 5 nM Sph3 h for 1 h at room temperature, solubilized oligosaccharides were then further purified on a Sep-Pak C 18 cartridge. In brief, cartridges were conditioned using absolute ethanol followed by water. Samples were then loaded onto the cartridge before washing and eluting using a 0.25% (v/v) step gradient of methanol from 0 to 4% (v/v) followed by a 1% (v/v) step gradient of ACN from 1 to 4%.

Reduction of oligosaccharides
Reduction of the ␣-1,4-GalNAc 9-mer was performed resuspending the oligosaccharides in 1 M ammonium hydroxide containing 10 mg/ml of sodium borohydride and incubating overnight at room temperature. Reaction was quenched adding dropwise 30% (v/v) acetic acid and samples were purified using a Hypersep Hypercarb SPE cartridge (ThermoFisher) as per chitin oligosaccharide purification.

Crystallization, data collection, and structure solution
Purified PelA h was concentrated to ϳ20 mg/ml and crystallization trials were performed using MCSG1-4 sparse-matrix screens (Microlytic) in 48-well hanging-drop VDX plates (Hampton Research) using a 2-l drop with a 1:1 protein:precipitant ratio at 20°C. Initial crystallization hits were obtained in several conditions. Diffraction quality native PelA h crystals were grown using 0.1 M BisTris, pH 7.5, 25% (w/v) PEG MME 5000 at a 1:2 ratio of protein to crystallization solution at 20°C. SeMet-labeled PelA h was crystallized in a similar condition with 26% (w/v) PEG MME 5000. Both crystals were cryoprotected for 10 s in mother liquor supplemented with 15% (v/v) ethylene glycol prior to vitrification in liquid nitrogen.
Diffraction data were collected at Ϫ173°C with wavelengths of 0.9791 and 1.075 Å on beamline X29, National Synchrotron Light Source (NSLS) ( Table 1) for the SeMet labeled and native crystal, respectively. A high redundancy dataset was generated for the SeMet-labeled PelA h by collecting 90 images with 2°o scillation at 90% beam attenuation and with an exposure time of 0.3 s/image and 360 images with 1°oscillation with 50% beam attenuation with an exposure time of 0.4 s/image on an ADSC Quantum-315 detector with a 260-mm crystal-to-detector distance. The native PelA h dataset was collected using the same strategy as described above for the SeMet crystal but with a 180-mm crystal-to-detector distance. Autosol (40) was used to determine initial phases and generate a density-modified map. The resulting electron density map was of high quality and enabled PHENIX AutoBuild to build Ͼ95% of the protein. The remaining residues were built manually in COOT (41,42) and the structure refined using PHENIX.REFINE (43). Translation/ Libration/Screw (TLS) groups were added to the refinement in PHENIX through the use of the TLSMD server (44,45).
All structure figures were generated using the PyMOL molecular graphics system (DeLano Scientific) (46), or Chimera (47) for electrostatics using APBS (48). Structural similarity to deposited structures in the Protein Data Bank (PDB) was determined using DALI and structure alignment was performed in COOT (27,41). Amino acid conservation was calculated using the Consurf server (49) aligned to 113 proteins using the default settings. Programs used for crystallographic data processing and analysis were accessed through SBGrid (50).

Secreted GAG purification and digest
Secreted purified GAG was prepared as previously reported (13). Briefly, culture supernatant of a 3-day-old Af293 culture a R merge ϭ ΑΑ͉I (k) Ϫ ͗I͉͘/Α I (k), where I (k) and ͗I͘ represent the diffraction intensity values of the individual measurements and the corresponding mean values. The summation is over all unique measurements. b R work ϭ ΑʈF obs ͉ Ϫ k͉F calc ʈ/͉F obs ͉, where F obs and F calc are the observed and calculated structure factors, respectively. c R free is the sum extended over a subset of reflections (5%) excluded from all stages of the refinement. d As calculated using MolProbity (51). e Maximum-Likelihood Based Coordinate Error, as determined by PHENIX (42).

Molecular mechanism of glycoside hydrolases Sph3 h and PelA h
was filtered on Miracloth prior to being ethanol precipitated. Precipitate was then successively washed with 70% (v/v) ethanol twice, 150 mM NaCl, and water. The remaining gel was then freeze dried. The dried purified GAG was incubated with the 1 M glycoside hydrolase for 1 h in 0.1ϫ PBS. The released soluble oligosaccharides were then analyzed using the MALDI-TOF MS enzymatic fingerprint technique.