Biochemical and Structural Characterization of a Five-domain GH115 α-Glucuronidase from the Marine Bacterium Saccharophagus degradans 2-40T*

Glucuronic acid (GlcAp) and/or methylglucuronic acid (MeGlcAp) decorate the major forms of xylan in hardwood and coniferous softwoods as well as many cereal grains. Accordingly, the complete utilization of glucuronoxylans or conversion to sugar precursors requires the action of main chain xylanases as well as α-glucuronidases that release the α- (1→2)-linked (Me)GlcAp side groups. Herein, a family GH115 enzymefrom the marine bacterium Saccharophagus degradans 2-40T, SdeAgu115A, demonstrated activity toward glucuronoxylan and oligomers thereof with preference toward MeGlcAp linked to internal xylopyranosyl residues. Unique biochemical characteristics of NaCl activation were also observed. The crystal structure of SdeAgu115A revealed a five-domain architecture, with an additional insertion C+ domain that had significant impact on the domain arrangement of SdeAgu115A monomer and its dimerization. The participation of domain C+ in substrate binding was supported by reduced substrate inhibition upon introducing W773A, W689A, and F696A substitutions within this domain. In addition to Asp-335, the catalytic essentiality of Glu-216 was revealed by site-specific mutagenesis. A primary sequence analysis suggested that the SdeAgu115A architecture is shared by more than half of GH115 members, thus defining a distinct archetype for GH115 enzymes.

Glucuronic acid (GlcAp) and/or methylglucuronic acid (MeGlcAp) decorate the major forms of xylan in hardwood and coniferous softwoods as well as many cereal grains. Accordingly, the complete utilization of glucuronoxylans or conversion to sugar precursors requires the action of main chain xylanases as well as ␣-glucuronidases that release the ␣-(132)-linked (Me)GlcAp side groups. Herein, a family GH115 enzyme from the marine bacterium Saccharophagus degradans 2-40 T , SdeAgu115A, demonstrated activity toward glucuronoxylan and oligomers thereof with preference toward MeGlcAp linked to internal xylopyranosyl residues. Unique biochemical characteristics of NaCl activation were also observed. The crystal structure of SdeAgu115A revealed a five-domain architecture, with an additional insertion C ؉ domain that had significant impact on the domain arrangement of SdeAgu115A monomer and its dimerization. The participation of domain C ؉ in substrate binding was supported by reduced substrate inhibition upon introducing W773A, W689A, and F696A substitutions within this domain. In addition to Asp-335, the catalytic essentiality of Glu-216 was revealed by site-specific mutagenesis. A primary sequence analysis suggested that the SdeAgu115A architecture is shared by more than half of GH115 members, thus defining a distinct archetype for GH115 enzymes.
At present, more than 300 GH115 sequences are recorded in the CAZy database; however, only five representatives have been characterized to date, including the first one from Schizophyllum commune (ScAgu115A) (16,22) as well as PsAgu115A from Pichia stipitis (now Scheffersomyces stipitis) (17), SpAgu115A from Streptomyces pristinaespiralis (25), BoAgu-115A from Bacteroides ovatus (23) and BtGH115A from B. thetaiotaomicron (24). 1 H NMR analysis of two GH115 ␣-glucuronidases from P. stipitis and S. commune confirmed that, like GH67 enzymes, ␣-glucuronidases from GH115 adopt an inverting catalytic mechanism (26). The first crystal structure of a GH115 enzyme was reported in 2013. In their study, Bolam and co-workers (23) showed that BoAgu115A from B. ovatus forms a tight dimer with each monomer composed of four distinct domains (A-D). The dimerization interface comprises interactions between helical bundles of C domains in each protomer and reciprocally between the C-terminal ␤-sandwich (domain D) of protomer 1 and the TIM barrel of protomer 2 (domain BЈ) (23). The BoAgu115A-GlcAp complex structure allowed these authors to localize the enzyme's active site to the deep pocket within the TIM barrel domain (B/BЈ) of each protomer. Mutagenesis analysis demonstrated the important role of several residues, including Asp-206, Asp-332, or Glu-375, as candidates for catalytic residues and Arg-328 involved in substrate recognition. Notably, the overall structures of the first three domains of BoAgu115A bear distant evolutionary similarity with corresponding domains in GH67 enzymes, and although catalytic residues identified in each of these CAZy families do not directly align, they are all located at the C-terminal end of domain B. More recently, the BtGH115A structure revealed a four-domain architecture similar to that of BoAgu115A; however, the position of C-terminal domain D was altered and placed atop of domain C (24). Although representing the first two molecular images of GH115 family enzymes, the structural insights gained to date cannot be directly applied to more than 50% GH115 members due to a substantial sequence insertion, thus necessitating further studies of corresponding GH115 representatives.
To further map the diversity of GH115 family enzymes, we undertook a detailed biochemical and structural characterization of the GH115 ␣-glucuronidase from S. degradans 2-40 T (SdeAgu115A), which shares Ͻ40% sequence identity with previously characterized GH115 representatives (33% with BoAgu115A, 39% with PsAgu115A, 38% with ScAgu115A, and 23% with BtGH115A). The crystal structure of the SdeAgu115A monomer revealed five distinct domains, in contrast to the four-domain composition of BoAgu115A and BtGH115A (23,24), resulting from the insertion of an additional 125-amino acid domain between domains C and D, herein referred to as C ϩ . Detailed structural and mutagenesis studies of SdeAgu-115A demonstrated the role of domain C ϩ in domain arrangement, enzyme dimerization, and substrate binding; Glu-216 in SdeAgu115A was also identified as an essential catalytic residue. Finally, biochemical analyses using defined xylo-oligosaccharides as well as multiple xylan sources confirmed SdeAgu115A action toward ␣-(132)-glycosidic linkages between (Me)GlcAp and singly substituted Xylp residues at both internal and terminal positions of the substrate.

Experimental Procedures
Materials-Pfx polymerase was purchased from Invitrogen. The QIAEX II gel extraction kit and QIAquick PCR purification kit were purchased from Qiagen. Phosphorylase b from rabbit muscle, beechwood xylan, and oat spelt xylan were purchased from Sigma-Aldrich; GH11 endoxylanase from Neocallimastix patriciarum and the D-glucuronic acid assay kit were purchased from Megazyme; GH10 xylanase (Shearzyme) and GH11 xylanase (Pentopan Mono BG) were obtained from Novozymes; pNP-GlcAp was purchased from TCI. Spruce arabinoglucuronoxylan was prepared as reported previously (32). Genomic DNA of S. degradans 2-40 T (ATCC catalog no. 43961) was purchased from American Type Culture Collection. All other chemicals were analytical grade and obtained from Sigma-Aldrich or Fisher.
DNA Manipulation-The gene encoding the mature form of SdeAgu115A (GenBank TM accession number ABD81015) without a signal peptide (aa residues 1-32) was amplified from S. degradans genomic DNA with Pfx DNA polymerase (Invitrogen). The infusion cloning kit from Clontech was used to transfer purified PCR products to the p15Tv-L expression vector (GenBank TM accession number EF456736), generating p15Tv-L_SdeAgu115A. Site-specific mutagenesis was carried out according to a modified QuikChange TM (Stratagene) method (33). All constructs were verified by DNA sequencing at the Center of Applied Genomics at the SickKids Hospital in Toronto.
GH115 Enzyme Sequence Analysis-Enzyme sequences were retrieved from the CAZy database and the JGI fungal genomes database (November 22, 2014). The sequences were curated by removing members that had sequences shorter than 200 amino acid residues or lacked the catalytic domain B. In total, 303 GH115 ␣-glucuronidase sequences were collected and then aligned using MUSCLE in Geneious version 6.0.6. Geneious Tree Builder version 6.0.6 was used to create phylogenetic tree with the neighbor-joining method.
The induced cultures were harvested by centrifugation at 6000 ϫ g for 15 min. Cell pellets (ϳ4.3 g fresh weight) were suspended in 30 ml of binding buffer (300 mM NaCl, 50 mM HEPES (pH 7.0), 5% glycerol, 5 mM imidazole) and disrupted by sonication. Cell debris was removed by centrifugation (17,500 ϫ g, 20 min), and supernatants were passed through a 0.45-m filter before being incubated for 45 min with 5.0 ml of pre-equilibrated nickel-nitrilotriacetic acid resin (Qiagen) at 4°C. Resin was then washed with 300 ml of washing buffer (300 mM NaCl, 50 mM HEPES (pH 7.0), 5% (v/v) glycerol, 50 mM imidazole), and bound protein was eluted with ϳ30 ml of elution buffer (300 mM NaCl, 50 mM HEPES (pH 7.0), 5% (v/v) glycerol, 250 mM imidazole) in a stepwise manner. Active fractions were pooled and adjusted to 1.0 M NaCl before being applied to a phenyl-Sepharose TM hydrophobic interaction chromatography column (1 ϫ 18.5 cm) equilibrated with the same buffer (50 mM HEPES (pH 7.0) containing 1.0 M NaCl). Fractions (5 ml) were collected at a flow rate of 1.5 ml min Ϫ1 and linear gradient of 1.0 to 0 M NaCl in 50 mM HEPES (pH 7.0) over 5 column volumes. All chromatographic steps were performed using the BIOSHOP Duoflow system (Bio-Rad). Fractions containing SdeAgu115A were exchanged to 25 mM HEPES buffer (pH 7.0) containing 300 mM NaCl using a Bio-Gel P10 column. Amicon stirred ultrafiltration cells (Millipore) equipped with a polyethersulfone membrane (nominal molecular weight 100,000) were then used to concentrate purified protein samples before aliquots were flash-frozen in liquid nitrogen and then stored at Ϫ80°C.
Protein concentrations were determined using the Bradford assay and bovine serum albumin as a standard (34). SDS-PAGE was performed and stained with Coomassie Blue according to established procedures (35).
Production of MeGlcAp-substituted xylooligosaccharides (U 4m2 XX and XU 4m2 XX)-Birchwood xylan 1% (w/v) prepared in 300 ml of 50 mM sodium acetate buffer (pH 5.0) was boiled for 5 min and then cooled before the addition of endoxylanases. GH10 xylanase (Shearzyme) (20 microkatals/g xylan) and GH11 xylanase (Pentopan Mono BG) (30 microkatals/g xylan) were separately added to generate U 4m2 XX and XU 4m2 XX, respectively. Enzymatic hydrolyses were performed at 40°C for 72 h with a continuous mixing, and the extent of hydrolysis was regularly checked by TLC. The reaction was stopped by 20-min boiling, and the supernatant was collected after a centrifugation of 15,000 ϫ g for 15 min. Acidic glucuronyl xylooligosaccharides (XOS) were isolated using Dowex 1 ϫ 2 anion exchange resin (75 ml; Sigma), which had been regenerated with 10 volumes of degassed 2 M ammonium formate (pH 6.0) and washed with 2 volumes of water in an Äkta chromatography system (GE Healthcare). Unbound neutral XOS were washed with water (1 ml min Ϫ1 ). Bound acidic glucuronyl XOS were eluted with an 800-ml gradient of 0 -400 mM ammonium formate (pH 6.0) and concentrated to 4 ml in a rotatory evaporator and further purified using BioGel-P2 size exclusion chromatography (8.9 ϫ 135 cm; Bio-Rad) with Milli-Q water (Millipore) as the eluent. Pure sugar factions were concentrated using a rotary evaporator, and the total carbohydrates were determined by the phenol-sulfuric acid method using xylose as the standard (36). The purity was determined by TLC, high performance anion exchange chromatography with pulsed amperometric detection, and mass spectrometry using purified U 4m2 XX and XU 4m2 XX as the standard (37). Mass spectrometric analyses of purified XOS were performed using an Agilent XCT Plus model ion trap mass spectrometer (Agilent Technologies, Waldbronn, Germany) equipped with LC/MSD Trap Software version 5.2, as described by Vuong et al. (38). Hydrolysis products and chromatography fractions were analyzed with TLC on Silica Gel 60 aluminum plates (Merck, Darmstadt, Germany), which were developed in ethyl acetate/acetic acid/isopropyl alcohol/formic acid/water (25:10:5:1: The plate was sprayed with 0.2% orcinol solution in 80% ethanol (v/v) and 10% H 2 SO 4 (v/v) and heated at 100°C for 2-5 min to detect XOS.
Effect of Temperature, pH, Metal Ions, and NaCl-The thermostability of SdeAgu115A was evaluated by incubating the enzyme (300 g/ml) at pH 6.5 in 0.2 ml of 200 mM MES for up to 4 h at 25, 35, 40, and 50°C. Following incubation, residual activity was measured by transferring 6.0 g of the heat-treated enzyme to a final 0.2-ml reaction mixture comprising 200 mM MES (pH 6.5) and 1.0% (w/v) beechwood xylan. After a 10-min incubation at 40°C, reaction products were quantified using the Nelson-Somogyi assay for reducing sugars (39). The pH optimum of SdeAgu115A was determined by measuring activity as described above, this time in buffers with a pH range from 4.0 to 9.5. Buffers used were 200 mM CH 3 CO 2 Na (pH 4 -5.5), 200 mM MES (pH 6.0 -8.0), and 200 mM glycine-NaOH (pH 8.5-9.5). To determine pH stability, residual SdeAgu115A activity was measured after a 24-h incubation at 4°C in buffers with a pH range from 4.0 to 9.5. The effect of metal ions was determined by adding 1.0 mM CaCl 2 , MgCl 2 , CoCl 2 , MnCl 2 , NiCl 2 , CdCl 2, AgNO 3 , ZnCl 2 , CuSO 4 , or HgCl 2 to the activity assay mixture.
To further investigate the impact of NaCl on SdeAgu115A activity, the enzyme was dialyzed against 100 mM MES buffer (pH 6.5) before activity measurements in 100 mM MES buffer (pH 6.5) containing 0 -1.5 M NaCl.
Analytical Ultracentrifugation for Oligomeric State Analysis-Sedimentation equilibrium analytical ultracentrifugation was carried out using a Beckman Optima XL-A analytical ultracentrifuge equipped with an An-60 Ti rotor. Protein absorbance was monitored at 280 nm. Data were collected at speeds of 6000, 7000, and 8000 rpm using SdeAgu115A at varying enzyme concentrations (5.9, 3.0, and 1.5 M) in 20 mM HEPES (pH 7.0, 300 mM NaCl) at 20°C. Data analysis was performed using the Origin MicroCal XL-A/CL-I Data Analysis Software Package version 4.0. The density of buffer and the partial specific volumes of protein were calculated using SedNterp (40).
CD Spectra-CD was recorded with an Aviv model 202 circular dichroism spectrometer from 200 to 300 nm at 25°C in a 0.5-cm path length cell. The spectral bandwidth was 1.0 nm, the step size was 1.0 nm, and the averaging time was 1.0 s. The protein sample was in 0.2 M sodium phosphate buffer (pH 7.5) with an adjusted concentration of 1.0 mg/ml.
Affinity Gel Electrophoresis-The binding of SdeAgu115A and its mutants (E216A and D335A) to beechwood xylan was examined by native affinity gel electrophoresis (41) with minor modifications. Briefly, the native polyacrylamide gels contained 6.0% (w/v) acrylamide in 15 mM Tris, 150 mM glycine, 100 mM sodium phosphate buffer (pH 8.8), and 0.05 and 0.1% (w/v) of the test polysaccharide. Approximately 7.0 g of SdeAgu115A and each mutant were loaded onto the gels and then run at 70 V for 6.0 h. Relative binding affinities were inferred from the migration distance of the wild type and mutant enzymes on gels with and without the test polysaccharides. BSA and phosphorylase b from rabbit muscle (7.0 g) were used as reference proteins.
Protein Crystallization, Data Collection, and Structure Determination-For crystallographic purposes, the selenomethionine-substituted protein was overexpressed using M9 high yield growth medium (Medicilon Inc.) and purified according to the above-described procedure for recombinant SdeAgu115A.
SdeAgu115A was concentrated to 13.0 mg ml Ϫ1 in 10 mM HEPES (pH 7.5) containing 500 mM NaCl. The crystallization screen was performed by mixing 0.5 l of protein with 0.5 l of reservoir buffer using a Mosquito crystal (TTPlab) liquid-handling system. The SdeAgu115A crystals were obtained by a sitting drop vapor diffusion method at 23°C in the presence of 200 mM sodium dihydrogen phosphate, 100 mM sodium acetate (pH 4.6), 200 mM ammonium acetate, and 30% PEG 4000. The crystals were flash-frozen in liquid nitrogen using paratone-N oil as cryoprotectant. Diffraction data for SdeAgu115A crystals were collected at the Argonne National Laboratory, Structural Biology Center at the Advanced Photon Source, using beamline 19-ID with an ADSC Quantum 315R detector according to a previously described experimental set-up (42,43). All diffraction data were processed using the HKL3000 suite of programs (44). The SdeAgu115A structure was determined by the singlewavelength anomalous diffraction method using phasing, density modification, and initial protein model building as implemented in the HKL3000 software package (45)(46)(47)(48)(49)(50)(51). Cycles of manual corrections of the model were carried out in COOT and with REFMAC of CCP4 and finalized with Phenix using the implemented TLS group refinement routine. The final model was refined against all reflections except for 5% randomly selected reflections, which were used for monitoring R free . Final refinement statistics are presented in Table 1. Structure figures were prepared using PyMOL v1.3 (Schrödinger LLC, New York).
Substrate Preference and Enzyme Kinetics-SdeAgu115A activity on various xylans and xylooligosaccharides was measured using the D-glucuronic acid assay kit from Megazyme. Reactions were performed for 10 min at 40°C and were initiated by adding 5 l of enzyme to 20 l of 1.0% (w/v) substrate in 200 mM MES buffer (pH 6.5). Enzyme reactions were stopped by adding 2.5 l of 2.0 M HCl, followed by incubation at 25°C for 10 min. NaOH (2.5 l of a 2.0 M solution) was used to neutralize reaction mixtures before product detection. In all cases, enzyme doses were optimized to obtain initial rates of reaction. D-(ϩ)-Glucuronic acid (0.01-1.2 mg/ml) was used to generate a standard curve.
The activity on pNP-GlcAp was determined by measuring pNP release. The reaction mixture contained substrate in 0.15 ml of 200 mM MES buffer (pH 6.5). SdeAgu115A doses were adjusted to obtain initial rates of reaction at 40°C over 10 min. Reactions were terminated by adding 0.15 ml of 4% (w/v) Na 2 CO 3 . The amount of pNP formed was measured by absorbance at 405 nm, where pNP (50 -400 M) was used to generate a standard curve.
Finally, mass spectrometry was used to confirm the specific release of MeGlcAp from beechwood xylan by SdeAgu115A. Briefly, 160 l of 1% (w/v) beechwood xylan was incubated with 40 l of SdeAgu115A (54 g) in water at 22°C for 12 h. Electrospray ionization mass spectrometric analyses were performed using a Q-Exactive mass spectrometer (Thermo Scientific) in negative mode, and product spectra were compared with D-glucuronic acid.

Results and Discussion
Sequence Analysis and General Biochemical Properties of SdeAgu115A-Multiple-sequence alignment of 303 curated GH115 sequences exhibited two main groups of GH115 enzymes that were distinguished by the presence of an amino acid insertion motif corresponding to amino acid residues 667-792 in SdeAgu115A (Fig. 1A). Based on sequence alignment, nearly all (97%, 94 of 97 sequences) of the fungal GH115s appeared to comprise this conserved sequence insertion motif, whereas roughly 56% (109 of 195 sequences) of bacterial sequences, including previously characterized BoAgu115A and BtGH115A enzymes, lacked the insertion. In addition to these two major forms, 11 GH115 sequences featured additional C-terminal sequences, and six GH115 enzyme sequences corresponded to only the three N-terminal domains identified in BoAgu115A. Phylogenetic analysis of GH115 sequences showed that those with the insertion motif from both fungi and bacteria were grouped together (Fig. 1B), further suggesting a common yet uncharacterized structural arrangement shared by these enzymes. Notably, the N-terminal sequence of SdeAgu115A revealed a secretion motif (signal peptide/signal sequence) in line with the proposed extracellular function of these enzymes.
To characterize an ␣-glucuronidase from the GH115 subgroup containing an insertion motif, SdeAgu115A lacking the native signal peptide sequence (residues 1-32) was expressed in E. coli in fusion with an N-terminal His tag and purified to homogeneity. The weight-average molecular mass of native SdeAgu115A was determined to be 189 Ϯ 3.9 kDa by sedimentation-equilibrium analytical ultracentrifugation (supplemental Fig. S1), indicating that SdeAgu115A forms homodimers under the used experimental conditions (25 mM HEPES buffer (pH 7.0) containing 0.3 M NaCl). By comparison, BoAgu115A was also shown to form a stable dimer in solution (23), whereas BtGH115A appears to remain in monomeric form (24).
Of the 10 ions tested, the presence of 1.0 mM Ni 2ϩ , Cd 2ϩ , Ag ϩ , Zn 2ϩ , Cu 2ϩ , and Hg 2ϩ inhibited SdeAgu115A activity by ϳ40 -100%, whereas the rest did not have any significant affect on the enzyme's activity (Fig. 2C). Moreover, the addition of 10 mM EDTA did not significantly affect SdeAgu115A activity, indicating that this enzyme does not strictly depend on ion supplementation for its activity under the tested experimental conditions (Fig. 2C). Consistently, no metal ions were observed in the crystal structure of the enzyme. The effect of metal ions on the specific activity of other GH115 enzymes has not been tested. Altogether, these data will inform optimization efforts for future enzyme applications.
Crystal Structure of SdeAgu115A Revealed an Additional Domain-The SdeAgu115A crystal structure was determined using the single-wavelength anomalous diffraction approach and selenomethionine-labeled protein. The final model of SdeAgu115A consisted of two polypeptide chains in the asymmetric unit, spanning residues 40 -974 (chain A) and 40 -975 (chain B). Additional density was assigned as two sodium and two phosphate ions, one glycerol, and one acetate molecule. The final model exhibited excellent crystallographic and geometric statistics, which are summarized in Table 1.  In contrast to previously structurally characterized GH115 enzymes(23,24),thecrystalstructureoftheSdeAgu115Amonomer revealed a five-consecutive domain architecture with N-terminal domains A (residues 40 -205), B (residues 206 -494), C (residues 495-670), and the unique C ϩ domain corresponding to the insertion (residues 671-790) conserved in one of two major groups within the GH115 family ( Figs. 1 and 3); in SdeAgu115A, domain C ϩ is then followed by the C-terminal D domain (residues 791-975).
Superimposition of SdeAgu115A with BoAgu115A and BtGH115A demonstrated a root mean square deviation of 2.4 and 2.1 Å, respectively ( Fig. 3 and supplemental Table S1). Like BoAgu115A and BtGH115A, domain A of SdeAgu115A adopted a two-layer ␤-sandwich fold, composed of a central five-stranded ␤-sheet linked to a smaller two-stranded antiparallel ␤-sheet between three ␣-helices (classified as Pfam 03648), whereas domain B displayed a distorted (␤/␣) 8 TIM barrel fold featuring short third and fourth ␤-strands (classified as Pfam 15979). Domain C of SdeAgu115A comprised a fivehelix bundle with a ϳ35-residue long loop disrupted by a short ␤-strand. In the case of SdeAgu115A, the five-helix bundle is wedged between domains B and C ϩ , whereas the loop and short strands are located at the side of this domain (Fig. 3A). In addition to close structural similarity to corresponding domains in BoAgu115A and BtGH115A, domains A-C in SdeAgu115A also share overall structural similarity with a GH67 ␣-glucuronidase from Geobacillus stearothermophilus (PDB code 1K9E) (20), GH84 O-GlcNAcase from Clostridium perfringens (PDB  code 2VUR) (53), and GH20 lacto-N-biosidase from Bifidobacterium bifidum (PDB code 4H04) (54) (supplemental Table S1).
Also similar to BoAgu115A and BtGH115, the C-terminal domain D of SdeAgu115A adopted a "jelly-roll" ␤-sandwich fold consisting of two ␤-sheets, each having five anti-parallel ␤-strands. However, the presence of the C ϩ domain in SdeAgu115A significantly impacted the positioning of this domain D compared with BoAgu115A (Fig. 3) (23).
As mentioned above, domain C ϩ in the SdeAgu115A structure corresponds to the additional insertion sequence, which is present in more than half of GH115 members. The C ϩ domain in SdeAgu115A displayed a ␤-barrel composed of eight antiparallel ␤-strands (Fig. 3IV). A structural similarity search using the Dali server (55) showed that it is most similar to the S pilus periplasmic chaperone protein SfaE (PDB code 1L4I) (56) and the chaperone protein FimC from E. coli (PDB code 3BWU) (57). Both proteins are involved in bacterial pilus biogenesis. More interestingly, domain C ϩ is also structurally similar to the Bacon domain (Bacteriodetes-associated carbohydrate-binding often N-terminal, Pfam 13004) in BoGH5A, a GH5 endoxyloglucanase from B. ovatus (supplemental Table S1) (58). Bacon domains have been identified in several other carbohydrateactive enzymes as well as proteases (2704 sequences deposited in the current Pfam database) (59), suggesting that the C ϩ domain in GH115 enzymes might play a role in carbohydrate binding as well as protein-protein interactions.  (Figs. 3 and 4). This change of domain composition and arrangement in SdeAgu115A prevented the enzyme from forming a butterfly-like dimer, as observed in the BoAgu115A structure (23). In BoAgu115A, the dimer is arranged by the crossing of complementary helical bundles of domain C from each promoter and is further strengthened by interactions of domain D of protomer A with domain BЈ from protomer B and vice versa (Fig. 4II). In contrast, the interface of the SdeAgu115A dimer was contributed primarily through interactions between anti-parallel ␤-sheets and their flanking region (residues 665-670) from C domains of each protomer, which formed on the opposite side of domain C relative to domain D (Fig. 4I). Additional contacts between SdeAgu115A subunits were formed between the loops of the C ϩ domain in protomer A and loop regions from domains CЈ and BЈ in the protomer B. The overall interface between two SdeAgu115A promoters covers ϳ1371 Å 2 according to a PDBsum/PISA server estimate (60), which corresponds to only 3.9% of each promoter surface area and involves 42 residues in each subunit. In comparison, the overall interface between BoAgu115A promoters is ϳ4000 Å 2 and involves 158 residues, accounting for 12.4% of each subunit surface area.

Distinct Domain Composition of SdeAgu115A Significantly Impacted Its Domain Arrangement and Dimerization-The
The recently determined structure of BtGH115A unveiled a second four-domain architecture for the GH115 family (24). This structure closely resembled the domain organization of SdeAgu115A, with its domain D positioned adjacent to domain C and a 15-residue loop to link domains C and D (Fig. 3III); however, BtGH115A was reported to be a monomer (24). Comparison of the BoAgu115A and BtGH115A structures suggests that the placement of domain D atop domain C blocks access to one side of domain C, thereby limiting the interface available for oligomer formation in BtGH115A. In SdeAgu115A, the insertion of domain C ϩ along with the anti-parallel ␤-sheets formed by residues 667-670 in domain C appeared to reconstitute the interface for stable dimerization of the enzyme.
Site-specific Mutagenesis Highlighted the Role of Individual Residues in Catalysis-In line with previously structurally characterized GH115 enzymes, the active site of SdeAgu115A localized in the large cavity at the C-terminal end of the ␤/␣ TIM barrel of the domain B. To probe the essential residues for catalysis, four conserved acidic amino acid residues from domain B (Asp-215, Glu-216, Asp-335, and Glu-381) and one basic arginine residue (Arg-331), were individually replaced by alanine (Fig. 5). In addition to the D335A variant, no activity was detected for the E216A variant of SdeAgu115A (Table 2). CD spectra of SdeAgu115A and its mutants (D335A and E216A) confirmed similar overall structures of the proteins (Fig. 6A). Affinity gel electrophoresis was also used to evaluate the impact of mutating Asp-335 and Glu-216 on substrate binding. The slightly slower migration and band smearing observed for the wild-type enzyme and D335A and E216A variants in gels containing beechwood xylan indicated weak binding to the substrate (Fig. 6B). Nevertheless, replacing Asp-335 or Glu-216 with alanine appeared to reduce this binding (Fig. 6B), suggest-ing that an activity-associated conformational change may enhance substrate binding by this enzyme. By contrast, D215A and E381A variants remained active, albeit with 15 and 330 times lower catalytic efficiencies, respectively, on beechwood xylan ( Table 2). Similar to the R328A mutant of BoAgu115A,  (n ϭ 3). ND, no detectable activity.  the R331A substitution in SdeAgu115A led to a 26,000-fold decrease in catalytic efficiency, further confirming the importance of this residue for GH115 enzyme activity, probably due to its involvement in positioning and stabilizing the Asp-335 side chain and likely interactions with the carboxylate group in (Me)GlcAp as well.

Residue location k cat(app) K m(app) K i(app) k cat(app) /K m(app)
Asp-206, Asp-332, or Glu-375 has been proposed to act as the catalytic acid/base in BoAgu115A (23), whereas these roles were assigned to residues Asp-299 and Glu-346 in BtGH115A (24). However, both E375A and D206A variants of BoAgu115A remained active on beechwood xylan, which is consistent with observations for E381A and D215A variants of SdeAgu115A (Table 2). To date, the catalytic essentiality was biochemically confirmed only for the equivalent residue of Asp-335 in SdeAgu115A (Asp-332 in BoAgu115A). Herein, complete activity loss was also observed for E216A in SdeAgu115A, suggesting its essential role in catalysis. Despite the high conservation of the Glu-216 residue among GH115 enzymes, this residue has not previously been reported as a catalytically important residue (23,24). Whereas Glu-216 and Asp-335 may represent a catalytic acid/base pair in SdeAgu115A, the distance between O⑀1 of Glu-216 and O␦2 of Asp-335 is 22.6 Å (Fig. 5B), which is considerably higher than the typical distance of 8 -12 Å observed between catalytic residues in other glycosidases that use an inverting mechanism (61). Because these residues are localized to a flexible region, it is possible that their position can change upon substrate binding, as was also proposed for BoAgu115A (23). An overlay of B domains from three available GH115 structures revealed a significant positional shift for C␦ in equivalent residues of Asp-335 in SdeAgu115A (12.6 Å relative to C␦ in Asp-299 of BtGH115, 13.6 Å relative to Asp-332 in BoAgu115A). Comparatively, shifts of equivalent residues to Arg-331 in SdeAgu115A are low, with a distance of 2.5 Å for C relative to Arg-295 in BtGH115A and 3.0 Å relative to Arg-328 in BoAgu115A. These shifts can be explained by movement in the loop formed by residues 330 -340 in SdeAgu115A (Fig. 7). Unfortunately, attempts to obtain the SdeAgu115A-GlcAp complex structure to further clarify involvement of specific residues in catalysis and substrate binding were unsuccessful (see "Experimental Procedures" for details).
Domain C ϩ Participated in Xylan Substrate Binding-Further analysis of domain B in SdeAgu115A revealed a long and deep cleft above and flanking the active site cavity, which could accommodate the backbone of xylan substrates. Notably, this cleft further extended into domain D of the same SdeAgu115A monomer (Fig. 5A). By contrast, the predicted xylan-binding cleft of BoAgu115A was formed by domain B and the C-terminal domain DЈ of the neighboring protomer (23).
Two loop regions spanning residues 689 -696 and 770 -775 in the additional C ϩ Ј domain of SdeAgu115A extended the sides of the xylan binding cleft, increasing its depth (Fig. 5A). To assess the potential role of these two loop regions in substrate binding, three aromatic residues (Trp-773, Trp-689, and Phe-696) within these loops were individually mutated to alanine, and their activities were evaluated. Although k cat values of all three mutants (W773A, W689A, and F696A) was decreased by 3-5 times compared with the wild type, these mutants displayed catalytic efficiency on beechwood xylan similar to that of the wild type enzyme due to slightly decreased K m(app) values ( Table 2). All three mutants also displayed reduced substrate inhibition with increased K i(app) values, particularly for W689A, where the K i(app) was increased by 15 times relative to wild type. The reduction in K m(app) values observed for each of the W773A, W689A, and F696A mutants was at first surprising given the presumed role of these aromatic residues in substrate stacking. However, the relative stability of the ␣-(132) linkage between MeGlcAp and Xylp compared with ␤-(134)-D-Xylp backbone linkages is well recognized (62). It is conceivable, then, that the aromatic residues present in loop regions of the additional domain C ϩ increase torsional strain within the bound substrate, thus decreasing the stability of the targeted glycosidic linkage.

Implications of SdeAgu115A Structure on Its Substrate
Profile-The glucurono-xylooligosaccharides were generated with a yield Ͼ10% (w/w); their purity (over 98%) was confirmed by TLC, high performance anion exchange chromatography with pulsed amperometric detection, and electrospray ionization-MS. The activity (k cat ) of SdeAgu115A on XU 4m2 XX and U 4m2 XX was 173.3 Ϯ 3.0 and 65.7 Ϯ 3.0 s Ϫ1 , respectively (Fig.  8), indicating that the enzyme was active on both terminal and internal MeGlcAp substitutions. Furthermore, SdeAgu115A revealed a catalytic efficiency (k cat /K m ) of 16.6 s Ϫ1 mM Ϫ1 for XU 4m2 XX, which is 5.5 times higher than that of U 4m2 XX (3.0 s Ϫ1 mM Ϫ1 ) (Fig. 8). The comparatively high catalytic efficiency measured using XU 4m2 XX was explained by both a 2.6-fold increase in k cat and a 2.1-fold decrease in K m compared with the value obtained using U 4m2 XX. A similar pattern was reported for BoAgu115A (23) and suggests that, as in BoAgu115A, at least one subsite (ϩ2NR) beyond the ϩ1 subsite of SdeAgu115A participates in xylan backbone binding.
Among the xylan substrates, SdeAgu115A showed the highest activity toward beechwood glucuronoxylan, followed by spruce arabinoglucuronoxylan and oat spelt xylan (Table 3). This is the first report directly comparing GH115 specific activity on hardwood, softwood, and cereal xylans. Beechwood glucuronoxylan comprises a ␤-(134)-Xylp backbone substituted with ␣-(132)-MeGlcAp, where the ratio of Xylp to MeGlcAp is ϳ8:1 (3) (Fig. 9). Similarly, the ␤-(134)-Xylp backbone of spruce arabinoglucuronoxylan and oat spelt xylan is decorated by ␣-(132)-MeGlcAp; however, these xylans are further substituted by Araf, where the ratio of MeGlcAp to Araf: Xylp in spruce arabinoglucuronoxylan and oat spelt xylan is ϳ2:1:11 and 1:3:24, respectively (32, 64). The lower activity of SdeAgu115A toward oat spelt xylan could be explained by the comparatively low water solubility of this substrate preparation, which would lower the effective MeGlcAp content available to the enzyme. Lower activity toward spruce arabinoglucuronoxylan was more difficult to explain, given the comparatively high water solubility of this substrate and the 1.5-fold higher MeGlcAp content compared with beechwood xylan. Previous studies of GH115 ␣-glucuronidase using acetylated glucurono-xylooligosaccharides demonstrate that although GH115 enzymes can access (Me)GlcAp substituents from internal regions of xylan, they may be restricted to singly substituted Xylp residues (16). Analogously, it could be that lower activity of SdeAgu115A on spruce arabinoglucuronoxylan is due to partial obstruction by Araf residues despite its position on other Xylp units, which would further support the formation of an extended substrate binding site by this and potentially other GH115 enzymes.

TABLE 3 Activity profile of SdeAgu115A
The final concentration of substrates in activity assay solution was 1% (w/v). No activity was detected toward alginate, rhamnogalacturonan, apple pectin, and gum arabic. Values shown are mean Ϯ S.D. (n ϭ 3). Infusion negative ion spectra confirmed that only MeGlcAp (m/z 207) was released from beechwood xylan by SdeAgu115A (data not shown). SdeAgu115A displayed low but detectable activity toward pNP-GlcAp (Fig. 8), which has not been tested for GH67 and other GH115 enzymes characterized to date. Still, k cat values obtained using pNP-GlcAp were 8200 -36,000 times lower than apparent k cat values obtained using glucurono-xylooligosaccharides and beechwood xylan (Fig. 8), underscoring the critical importance of including non-synthetic substrates in screens for new enzyme activity. Kinetic analyses also revealed substrate inhibition of SdeAgu115A by beechwood xylan (Fig. 8C), suggesting the non-catalytic interactions between SdeAgu115A and this substrate. Substrate inhibition can be caused by the non-productive binding of substrate molecules to different subsites within the enzyme substrate binding site (65). Thus, the substrate inhibition observed for SdeAgu115A implies a multisubsite binding cleft at the surface of this enzyme. Notably, substrate inhibition has not been reported for other GH115 enzymes characterized to date.

Polymeric substrates Activity
Impact of NaCl on the Aggregation and Activity of SdeAgu115A-SdeAgu115A was observed to aggregate in 20 mM HEPES buffer (pH 7.0). The aggregation was reversed by adding NaCl up to 40 mM in 20 mM HEPES buffer (pH 7.5) (Fig.  10A). Interestingly, the presence of NaCl (up to 300 mM) enhanced SdeAgu115A activity toward beechwood xylan up to 6-fold. Moreover, 80% of SdeAgu115A activity was retained in 1.5 M NaCl (Fig. 10B). NaCl tolerance and activation have been reported for several enzymes from other marine bacteria (52,63). Altogether, it was consistent with the marine habitat of the enzyme source organism.
Conclusion-Only a few ␣-glucuronidases from family GH115 have been characterized to date. Two recently published GH115 structures both exhibited four-domain architectures. Herein, we present a GH115 ␣-glucuronidase (SdeAgu115A) from the marine bacterium S. degradans 2-40 T with a novel five-domain architecture, which will be shared by more than  . The effect of NaCl to SdeAgu115A reversible aggregation and its activity on beechwood xylan. A, dialysis of enzyme solution against 20 mM Hepes buffer (pH 7.0) led to protein aggregation; turbidity at A 450 nm decreased through titration with NaCl. B, the Nelson-Somogyi method was used to measure SdeAgu115A activity in 100 mM MES buffer (pH 6.5) containing different NaCl concentrations. Full relative activity (100%) was obtained at 300 mM NaCl. Error bars, S.D.
half of GH115 members. The domain C ϩ insertion in SdeAgu115A significantly impacted domain arrangement and its dimerization, which is the predominant functional oligomeric state of this enzyme. In addition, an essential catalytic residue, Glu-216 in SdeAgu115A, was identified for the first time in a GH115 enzyme. Furthermore, site-specific mutagenesis revealed the participation of the C ϩ domain in xylan substrate binding and the benefit of altering the C ϩ region to modulate substrate inhibition. These results may benefit efforts to minimize detrimental impacts of Araf and acetyl substitutions on GH115 action through protein engineering. Determination of high resolution complex structures of SdeAgu115A in complex with ligands will be key to further evaluating these possibilities.
Author Contributions-E. R. M. and A. S. conceived and coordinated the study; W. W., R. Y., B. P. N., T. V. V., R. D. L., X. X., H. C., M. T., P. G., and G. T. conducted the experiments; P. G. and G. T. prepared the spruce arabinoglucuronoxylan used in the study; and W. W., R. Y., B. P. N., A. S., and E. R. M. wrote the paper. All authors reviewed the results and approved the final version of the manuscript.