The β-Glucanase ZgLamA from Zobellia galactanivorans Evolved a Bent Active Site Adapted for Efficient Degradation of Algal Laminarin*

Background: The marine bacterium Zobellia galactanivorans consumes laminarin, a main algal storage polysaccharide, as a carbon source. Results: The β-glucanase ZgLamAGH16 was structurally and biochemically characterized. Conclusion: ZgLamAGH16 evolved a unique bent active site, making the enzyme highly efficient for laminarin degradation. Significance: Within family GH16, highly specific laminarinase evolved from ancestral β-glucanases with a broader specificity. Laminarinase is commonly used to describe β-1,3-glucanases widespread throughout Archaea, bacteria, and several eukaryotic lineages. Some β-1,3-glucanases have already been structurally and biochemically characterized, but very few from organisms that are in contact with genuine laminarin, the storage polysaccharide of brown algae. Here we report the heterologous expression and subsequent biochemical and structural characterization of ZgLamAGH16 from Zobellia galactanivorans, the first GH16 laminarinase from a marine bacterium associated with seaweeds. ZgLamAGH16 contains a unique additional loop, compared with other GH16 laminarinases, which is composed of 17 amino acids and gives a bent shape to the active site cleft of the enzyme. This particular topology is perfectly adapted to the U-shaped conformation of laminarin chains in solution and thus explains the predominant specificity of ZgLamAGH16 for this substrate. The three-dimensional structure of the enzyme and two enzyme-substrate complexes, one with laminaritetraose and the other with a trisaccharide of 1,3–1,4-β-d-glucan, have been determined at 1.5, 1.35, and 1.13 Å resolution, respectively. The structural comparison of substrate recognition pattern between these complexes allows the proposition that ZgLamAGH16 likely diverged from an ancestral broad specificity GH16 β-glucanase and evolved toward a bent active site topology adapted to efficient degradation of algal laminarin.

Laminarinase is commonly used to describe ␤-1,3-glucanases widespread throughout Archaea, bacteria, and several eukaryotic lineages. Some ␤-1,3-glucanases have already been structurally and biochemically characterized, but very few from organisms that are in contact with genuine laminarin, the storage polysaccharide of brown algae. Here we report the heterologous expression and subsequent biochemical and structural characterization of ZgLamA GH16 from Zobellia galactanivorans, the first GH16 laminarinase from a marine bacterium associated with seaweeds. ZgLamA GH16 contains a unique additional loop, compared with other GH16 laminarinases, which is composed of 17 amino acids and gives a bent shape to the active site cleft of the enzyme. This particular topology is perfectly adapted to the U-shaped conformation of laminarin chains in solution and thus explains the predominant specificity of ZgLamA GH16 for this substrate. The three-dimensional structure of the enzyme and two enzyme-substrate complexes, one with laminaritetraose and the other with a trisaccharide of 1,3-1,4-␤-D-glucan, have been determined at 1.5, 1.35, and 1.13 Å resolution, respectively. The structural comparison of substrate recognition pattern between these complexes allows the proposition that ZgLamA GH16 likely diverged from an ancestral broad specificity GH16 ␤-glucanase and evolved toward a bent active site topology adapted to efficient degradation of algal laminarin.
1,3-␤-Glucans constitute one of the main types of storage compounds in eukaryotes (1). Their biosynthesis is a very ancient pathway and likely coexisted with the glycogen metabolism in the last common ancestor of eukaryotes (2). Storage 1,3-␤-glucans are particularly well distributed in marine algae, and these algal polysaccharides are structurally diverse. Euglenoid microalgae (Excavata phylum) store carbon as paramylon, a linear 1,3-␤-glucan that is deposited as semicrystalline granules in their cytosol (3). The storage polysaccharides of haptophyte phytoplanktons consist of ␤-1,3and ␤-1,6-linked glucose polymers, but they are mainly composed of ␤-1,3-linkages in Phaeocystales (4), whereas the ratio ␤-1,6/␤-1,3-linkages is ϳ1.5 in Coccolithales (5). Soluble 1,3-␤-glucans with occasional ␤-1,6-linked branches are the hallmark storage compounds of the Stramenopile phylum, which includes diatoms (phytoplanktons), the multicellular brown algae, and the nonphotosynthetic phytopathogens Oomycetes. Diatoms and Oomycetes produce 1,3-1,6-␤-glucans only constituted by glucose moieties, which are referred to as chrysolaminarin (6) and mycolaminarin (7), respectively. The storage polysaccharide of brown algae called laminarin, in reference to the Laminaria genus (kelp), has a slightly different structure. Like chrysolaminarin and mycolaminarin, it is a small vacuolar polymer that contains on average 25 glucosyl residues and some occasional ␤-1,6-linked branches (8). However, laminarin has the particularity to be composed of two series, the minor G series, which only contains glucose residues, and the more abundant M series, which displays a D-mannitol residue at the reducing end (9). The unique presence of mannitol in laminarin is explained by a horizontal gene transfer event between the common ancestor of brown algae and an ancestral actinobacterium. This horizontal gene transfer event, which occurred after the divergence from Oomycetes and diatoms, had a major impact on the evolution of brown algae, resulting in the acquisition of the mannitol metabolism (2) and the biosynthetic route for alginate, the main cell wall polysaccharide of extant brown seaweeds (10).
Brown seaweeds and 1,3-␤-glucan-producing microalgae represent a huge biomass in marine ecosystems. Brown algae dominate the intertidal and the upper sublittoral zones of rocky shores in temperate and polar regions and represent, with other marine macrophytes, a global carbon sink (11,12). Annually recurring phytoplankton blooms are large enough to be observed from space by satellites (13), and diatoms and haptophytes are frequently the dominant microalgal groups in such blooms (14,15). Therefore, 1,3-␤-glucans are abundant nutrients in marine trophic networks, and particularly, they constitute a crucial carbon source for specific marine heterotrophic bacteria (16,17). Moreover, a recent study of bacterioplankton responding to a spring diatom bloom in the North Sea revealed a dynamic succession of distinct bacterial populations specialized for successive decomposition of algal biomass. The first bacterial peak, which was dominated by Flavobacteria, has been accompanied by a high abundance of family GH16 laminarinases (18), underlying the environmental importance of this class of glycoside hydrolases. Flavobacteria are also found to be associated to macroalgae, and some species are specialists in the degradation of high molecular weight organic matter (19). This is the case with Zobellia galactanivorans, which is a model microorganism for the bioconversion of algal polysaccharides. This marine flavobacterium isolated from the red alga Delesseria sanguinea in Roscoff (20) has been extensively studied for its capacity to degrade agars (21)(22)(23) and carrageenans (24,25), which are sulfated galactans from red algae. However, Z. galactanivorans is also able to metabolize some polysaccharides from brown algae, such as alginate. Indeed, this microorganism possesses two alginolytic operons induced by the presence of alginate (26), and the two first alginate lyases of this complex system (AlyA1 and AlyA5) have been recently characterized at the biochemical and structural level (27). Z. galactanivorans grows with brown algal laminarin as its sole carbon source, and its genome contains five putative laminarinases: four of the family GH16 and one of the family GH64. This enzymatic diversity may seem surprising at first sight, considering that laminarin is a small, soluble polysaccharide. It is difficult to predict whether these enzymes have redundant or complementary activities or whether they match the biological diversity of 1,3-␤-glucans present in the sea. Moreover, some laminarinases are also known to be active on 1,3-1,4-␤-glucans, adding another degree of complexity to the characterization of this group of enzymes. As a first step to an in depth understanding of the laminarin utilization system of Z. galactanovorans, we report here the biochemical and structural analysis of its first GH16 laminarinase, ZgLamA.

MATERIALS AND METHODS
Except when mentioned, all chemicals were purchased from Sigma.
Cloning and Site-directed Mutagenesis of ZgLamA GH16 -The gene encoding the putative laminarinase ZgLamA (locus identifier: Zobellia_2431, GenBank TM accession number CAZ96583) was cloned as by Groisillier et al. (28). Briefly, primers were designed to amplify the coding region corresponding to the catalytic module of LamA (forward primer ggggggggatccgcctttaataccttagtgttttcaga, reverse primer cccccccaattgttatt-gatagatccttacatagtctatttc) by PCR from Z. galactanivorans genomic DNA. After digestion with the restriction enzymes BamHI and MfeI, the purified PCR product was ligated using the T4 DNA ligase into the expression vector pFO4 predigested by BamHI and EcoRI, resulting in a recombinant protein with a N-terminal hexahistidine tag (plasmid pLamA cat ). The plasmid was transformed into Escherichia coli DH5␣ strain for storage and in E. coli BL21(DE3) strain for protein expression. Sitedirected mutagenesis was performed using the QuikChange II site-directed mutagenesis kit (Stratagene) and the previous plasmid. The two putative catalytic residues Glu-269 and Glu-274 were replaced either by a serine or an alanine (mutant E269A: forward primer tggcccgcatgcggcgcaattgacattctggaa, reverse primer ttccagaatgtcaattgcgccgcatgcgggcca; mutant E269S: forward primer tggcccgcatgcggctcaattgacattctggaa, reverse primer ttccagaatgtcaattgagccgcatgcgggcca; mutant E274A: forward primer gaaattgacattctggcacaaaacggctgggac, reverse primer gtcccagccgttttgtgccagaatgtcaatttc; and mutant E274S: forward primer gaaattgacattctgtcacaaaacggctgggac, reverse primer gtcccagccgttttgtgacagaatgtcaatttc). Mutant plasmids were sequenced to confirm that the mutation occurred at the correct position. These variant plasmids were also transformed into E. coli DH5␣ strain for storage and in E. coli BL21(DE3) strains for protein expression.
Overexpression and Purification of ZgLamA GH16 and ZgLamA GH16-E269S -E. coli BL21(DE3) cells harboring the plasmid pLamA cat were cultivated at 20°C in a 1-liter autoinduction ZYP 5052 medium (29) supplemented with 100 g⅐ml Ϫ1 ampicillin. Cultures were stopped when the cell growth reached the stationary phase and were centrifuged for 35 min at 4°C at 3,000 ϫ g. The cells were resuspended in a 20 ml of buffer A (50 mM HEPES, pH 7.5, 500 mM NaCl, 50 mM imidazole). An anti-proteases mixture (Complete EDTA-free; Roche Applied Science) and 0.1 mg/ml of DNase were added. The cells were disrupted in a French press. After centrifugation at 12,500 ϫ g for 2 h at 4°C, the supernatant was loaded onto a 10-ml Sepharose column (GE Healthcare) previously charged with 100 mM NiSO 4 and equilibrated with buffer A. The column was washed with buffer A (90 ml), and the protein was eluted with 60 ml of linear gradient between buffer A and buffer B (50 mM HEPES, pH 7.5, 500 mM NaCl, 500 mM imidazole) with a flow rate at 1 ml⅐min Ϫ1 . The different fractions (1 ml each) were analyzed by SDS-PAGE. The fractions corresponding to a single band at the expected size (28 kDa) were pooled (47 ml) and were concentrated by ultrafiltration on an Amicon membrane (10-kDa cutoff) (30 ml at 9 mg⅐ml Ϫ1 ). An aliquot of 5 ml (9 mg⅐ml Ϫ1 ) was loaded onto a 120-ml Superdex 75 (GE Healthcare) column previously equilibrated with buffer C (20 mM Tris, pH 7.5, 200 mM NaCl). The protein was eluted using between 70 and 80 ml of buffer C, and the purity of the fractions was checked by SDS-PAGE. A calibration curve was also used to determine the oligomerization state of ZgLamA GH16 . The mutant protein ZgLamA GH16-E269S was produced by the same procedure and was purified by a single step of metal affinity chromatography as described above. Buffer A was composed of 20 mM Tris, pH 7.5, 300 mM NaCl, and 10 mM imidazole. The protein was eluted with a linear gradient of imidazole (10 -600 mM) with a flow rate at 1 ml⅐min Ϫ1 . ZgLamA GH16-E269S was dialyzed (molecular weight cutoff, 6 -8,000; Spectrum Laboratories) to eliminate imidazole. Finally, ZgLamA GH16 and ZgLamA GH16-E269S were concentrated by ultrafiltration on an Amicon membrane (10-kDa cutoff) to 10 and 14.6 mg⅐ml Ϫ1 , respectively. In addition, the proteins were filtrated on an Ultra free Durapore PVDF 0.1-m membrane before crystallization screening.
Thermostability Analysis-The thermostability of ZgLamA GH16 was studied by dynamic light scattering. A solution of 50 l of ZgLamA GH16 at 10 mg⅐ml Ϫ1 was filtered on a 0.2-m membrane. Using a Zetasizer Nano instrument (Malvern), the protein solution was heated from 10 to 70°C in steps of 1°C during a total period of 12 h, and the hydrodynamic gyration radius (R g ) was measured at each degree. The denaturation temperature was determined as the point of sharp change in gyration radius.
Enzymatic Activity Assays on ␤-Glucans-The hydrolytic activities of the purified ZgLamA GH16 and ZgLamA GH16-E269S were measured by the ferricyanide reducing sugar assay (30) on different ␤-glucans: laminarin from Laminaria digitata (0.1% w/v), mixed linked glucan (MLG) 4 from barley, curdlan from Alcaligenes faecalis, and paramylon from Euglena gracilis (all at 0.2% w/v). Laminarin is a small polysaccharide and has a large amount of reducing ends; thus this substrate was reduced prior to use. 70 mg of NaBH 4 were added to 10 ml of laminarin (20 mg⅐ml Ϫ1 ), and the mixture was incubated at 20°C for 3 days. The solution was acidified by adding concentrated acetic acid, drop by drop, until H 2 release stopped. 40 ml of absolute ethanol were added, and after centrifugation, the pellet was resuspended in 50 ml of absolute ethanol. This washing step was repeated twice, and the pellet was finally dried in vacuum (SpeedVac).
Reduced laminarin was hydrolyzed by 0.7 M of purified enzyme in a 1 ml of buffer C at 40°C for 30 min. Aliquots of the reaction mixture (40 l) were taken at T 0 , 10 min, and 30 min and added to 200 l of 5ϫ ferricyanide reagent. The samples were boiled at 95°C for 15 min and cooled to 20°C before absorbance measurement at 420 nm. All experiments were undertaken in triplicate. A calibration curve with 0 -3.33 mM (0, 0.278, 0.556, 1.11, 1.67, 2.22, 2.78, 3.33) glucose was used to calculate the amount of released reducing ends as glucose-reducing end equivalents. The activity of ZgLamA GH16 on MLG, curdlan, and paramylon was measured similarly, except that the reactions were monitored for 15 h. Aliquots were taken at T 0 , 10 min, 1 h, and 15 h.
The pH optimum for laminarin hydrolysis was determined as follows: 0.06% (w/v) of laminarin were hydrolyzed by 10 nM of ZgLamA GH16 in a 1-ml reaction mixture at 40°C for 10 min. The following different buffers (at 100 mM) were tested with the pH varying from 4 to 9 by 0.5 increments of pH units: phosphate citrate (pH 4 -6), MOPS (pH 6 -7.5), Tris-HCl (pH 7.5-8.5), and glycine-NaOH (pH 8.5-9). Released reducing ends were measured as described above, except that aliquots of reaction mixture (40 l) were taken every 2 min.
The kinetic parameters of ZgLamA GH16 on reduced laminarin and MLG were determined using 10 nM of enzyme in a reaction mixture of 500 l at optimal temperature and pH. The amount of released reducing ends was measured as above. For each substrate, five concentrations were used: 0.06% (w/v), 0.12%, 0.24%, 0.48%, and 0.96% for laminarin and 0.05% (w/v), 0.1%, 0.15%, 0.2%, and 0.25% for MLG. Aliquots of the reaction mixture (40 l) were taken every 2 min for 10 min for laminarin and every 15 min for 1 h for MLG. For each substrate, the K m and k cat were determined from a Lineweaver-Burk plot.
Fluorophore-assisted Carbohydrate Electrophoresis (FACE) Analysis-0.5% (w/v) of laminarin was hydrolyzed using 100 nM of ZgLamA GH16 in a reaction mixture of 500 l of glycine buffer, pH 8.5, at 20°C. An aliquot of 20 l (100 g of oligosaccharides) was taken at 30 s, 1 min, 2 min, 5 min, 15 min, 30 min, and 1 h. The samples were boiled to inactivate the enzyme and then dried in vacuum (SpeedVac). The FACE experiment was undertaken as previously described (31). Briefly, the oligosaccharides were mixed to 2 l of 0.15 M 8-aminonaphthalene-1,3,6-trisulfonic acid (ANTS) and 5 l of 1 M NaBH 3 CN. The reaction mixtures were incubated at 37°C for at least 3 h and dried in vacuum (SpeedVac). The oligosaccharides were resuspended in 20 l of 25% glycerol, and 10 l (50 g) were loaded into a 36% acrylamide gel. The migration was undertaken at 200 V at 4°C and with a 1ϫ migration buffer (192 mM glycine, 25 mM Tris, pH 8.5). The experiment was repeated using 0.5% (w/v) of MLG and 4.5 M ZgLamA GH16 . The reaction mixture was incubated at 40°C, and an aliquot of 20 l was taken at 2, 5, 15, and 30 min.
A glucan tetrasaccharide containing two ␤-1,4-linkages separated by one ␤-1,3-linkage (G4G3G4G) was also purchased from Megazyme. Three samples of this substrate at 50 g were labeled with ANTS as previously described. One of them was used as control. The second sample was resuspended in a reaction mixture of 50 l containing 4.5 M of ZgLamA GH16 and glycine buffer, pH 8.5, at 37°C for 30 min. The same experiment was undertaken on the third sample, except that ZgLamA GH16 was first inactivated by heating. In parallel, 100 g of nonlabeled G4G3G4G was hydrolyzed by ZgLamA GH16 at 37°C for 30 min. After the enzymatic reaction, an aliquot containing 50 g of oligosaccharides (reaction products) was labeled as mentioned above. A sample of 50 g of glucose was also labeled with ANTS and was used as control. 10 l (25 g) of each samples were loaded onto a 36% acrylamide gel.
Crystallization of ZgLamA cat and ZgLamA GH16-E269S , Structure Determination, and Refinement-Crystallization screening was undertaken with the nanodrop robot Honeybee (Cartesian) using the commercial screens PACT and JCSGϩ (Qiagen). The initial crystallization conditions of ZgLamA GH16 were manually optimized, and single crystals were obtained as follows: 2 l of enzyme at 10 mg⅐ml Ϫ1 were mixed with 2 l of reservoir solution containing 24% PEG 3350 and 100 mM sodium citrate, pH 5.2, in hanging drops at 4°C. Single crystals of ZgLamA GH16-E269S in complex with laminarin oligosaccharides were obtained from 2 l of the mixture enzyme/oligosaccharides (13.3 mg⅐ml Ϫ1 of enzyme, 5 mM of purified hexasaccharides) that were added to 1 l of reservoir solution containing 100 mM MIB buffer, pH 4.0 (sodium malonate, imidazole, and boric acid), and 19% of PEG 1500 in hanging drops at 20°C. Single crystals of ZgLamA GH16-E269S in complex with MLG trisaccharides were obtained from 2 l of the mixture enzyme/oligosaccharides (11.7 mg⅐ml Ϫ1 of enzyme, 0.04% (w/v) of MLG degradation products) that were added to 1 l of reservoir solution containing 100 mM MIB buffer, pH 4.0, 17% of PEG 1500, and 10% of glycerol in hanging drops at 12°C. Prior to flash-freezing in a nitrogen stream at 100 K, single crystals were quickly soaked in the same crystallization solution supplemented with 10, 20, or 12% glycerol, respectively. Diffraction data for ZgLamA GH16 were collected on Beamline ID23-1 at the European Synchrotron Radiation Facility (Grenoble, France). The diffraction data for the two complexes, ZgLamA GH16-E269S ⅐laminarin and ZgLamA GH16-E269S ⅐MLG, were collected on Beamline ID29 (European Synchrotron Radiation Facility). X-ray diffraction data were integrated using Mosflm (32) and scaled with Scala (33). The structure of ZgLamA GH16 was determined by molecular replacement with MolRep (34) using the chain A of RmLamR from Rhodothermus marinus (PDB code 3ILN (35) as a starting model. The structure of ZgLamA GH16 was manually built using COOT (36). For both complexes of ZgLamA GH16-E269S , the structures were also determined by molecular replacement, but using the coordinates of the chain A of ZgLamA GH16 . For all the structures, the initial molecular replacement solutions were further refined with the program REFMAC5 (37), alternating with cycles of manual rebuilding using COOT. A subset of 5% randomly selected reflections was excluded from computational refinement to calculate the R free factors throughout the refinement. The addition of the ligand sugar units for the complexed structures was performed manually using COOT. Water molecules were added automatically with REFMAC-ARP/wARP and visually verified. The final refinement was carried out using REFMAC5 with TLS, anisotropic B factors, and NCS restraints and Babinet scaling. Data collection and refinement parameters are presented in Table 1. The atomic coordinates and structure factors of ZgLamA GH16 , ZgLamA GH16-E269S ⅐laminarin and ZgLamA GH16-E269S ⅐MLG (codes 4BQ1, 4BOW, and 4BPZ, respectively) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics (Rutgers University, New Brunswick, NJ).
Sequence Analysis and Structure Comparison-The family GH16 laminarinases with known three-dimensional crystal structures were selected in the CAZY database (38), and their amino acid sequences were recovered in the PDB. These sequences and the ZgLamA GH16 sequence were aligned using MAFFT (39). The multiple sequence alignment was manually refined using Bioedit, based on the superimposition of the structures of the different laminarinases. The crystal structure of ZgLamA GH16 was compared with those of the laminarinase PcLam16A from Phanerochaete chrysosporium (PDB code 2W39) (40) and of the lichenase from Fibrobacter succinogenes (PDB code 1ZM1) (41) using COOT and PyMOL. For phylogenetic analysis, close homologues of ZgLamA GH16 and RmlamR were selected on the basis of a BlastP search on GenBank TM NR.  a The values in parentheses concern the high resolution shell. b R sym ϭ ⌺͉I Ϫ I av ͉/⌺͉I͉, where the summation is over all symmetry-equivalent reflections. c R pim ϭ corresponds to the multiplicity weighted R sym .
The selected proteins were subjected to multiple sequence alignment using MAFFT (42), with the iterative refinement method and the scoring matrix Blosum62. These alignments were manually edited using Bioedit (© Tom Hall). The phylogenetic tree was derived from the refined alignment using the maximum likelihood method with the program MEGA 5 (43). The reliability of the trees was tested by bootstrap analysis using 100 resamplings of the data set.

RESULTS
ZgLamA GH16 Is a ␤-Glucanase Highly Specific for Laminarin-In the genome of Z. galactanivorans, the gene zg2431 is predicted to produce a putative laminarinase referred to as ZgLamA. This protein is predicted to be anchored in the outer membrane because of the presence of a lipoprotein signal peptide and displays a modular architecture (Fig. 1A). In the N-terminal region, the signal peptide is followed by a polycysitic kidney disease-like module composed of 102 residues. Such ␤-sandwich domains are often involved in protein-protein interactions (44). The C-terminal region is composed of a catalytic module of glycoside hydrolase family 16 (GH16). Within the GH16 family, two conserved glutamates in the pattern EXDX(X)E play the role of the catalytic residues (45). In ZgLamA, the equivalent to the nucleophile is Glu-269, whereas the general acid/base is Glu-274 (Fig. 1B). To study the enzymatic properties of ZgLamA without potential interference from the polycysitic kidney disease-like module and also to facilitate crystallization assays, we have decided to only clone the nucleotide sequence corresponding to the GH16 catalytic module into the vector pFO4. The recombinant protein, referred to as ZgLamA GH16 , was produced in soluble form in E. coli BL21(DE3) strain with a yield of ϳ300 mg/liter of culture. A step of immobilized metal ion affinity chromatography followed by size exclusion chromatography was necessary to purify ZgLamA GH16 to electrophoretic homogeneity. The size exclusion chromatography analysis suggested that ZgLamA GH16 is a monomer in solution, and this result was confirmed by dynamic light scattering experiments. The dynamic light scattering was also used to study the protein thermostability. Above 40°C, a sharp increase of the hydrodynamic radius of gyration (R g ) is observed, indicating the beginning of the denaturation of ZgLamA GH16 .
The enzymatic activity of the purified ZgLamA GH16 was screened by the ferricyanide reducing sugar assay on two soluble ␤-glucans, laminarin from L. digitata and 1,3-1,4-␤-D-glucan from barley (mixed linked glucan, MLG), and two crystalline 1,3-␤-glucans, curdlan from A. faecalis and paramylon from E. gracilis. Significant activity was detected in the presence of laminarin and MLG, but not on the crystalline glucans. Prior to determining the kinetic parameters of ZgLamA GH16 , the effect of pH on the enzyme activity was determined using laminarin. The maximum of activity is obtained at pH 8.5 in 0.1 M glycine buffer ( Fig. 2A). ZgLamA GH16 displays no activity at pH 4.0 in 0.1 M phosphate citrate buffer. The activity is detectable from pH 4.5 and increases with the pH until 7.0 (in 0.1 M MOPS), reaching 92% of the maximum activity. In contrast, the Tris buffer clearly has an inhibitory effect. In 0.1 M Tris-HCl, pH 8.5, ZgLamA GH16 only reaches 30% of the optimal activity detected in glycine buffer at the same pH. Altogether, the highest activity was obtained in glycine-NaOH pH 8.5, but the exact pH optimum remains uncertain. The kinetic parameters of ZgLamA GH16 were then measured at 40°C and in glycine-NaOH, pH 8.5, on reduced laminarin and MLG. The turnover of ZgLamA GH16 on laminarin is much higher than on MLG (410 and 32 s Ϫ1 , respectively), whereas its K m values for both substrates are relatively similar (K m 5.0 mM (0.9 mg⅐ml Ϫ1 ) and 8.7 mM (1.57 mg⅐ml Ϫ1 ), respectively) (Fig. 2B). Taking into account these two parameters, ZgLamA GH16 has a catalytic efficiency (k cat /K m ) 22-fold higher on laminarin than on MLG.
The nucleophile candidate Glu-269 was successfully replaced by a serine, by site-directed mutagenesis. The mutant variant ZgLamA GH16-E269S was also produced in E. coli BL21(DE3) in a soluble form with high yield (110 mg/liter of culture) and purified by nickel affinity chromatography. ZgLamA GH16-E269S (10 nM) was assayed on laminarin and MLG with reducing sugar assay as for ZgLamA GH16 , but no enzymatic activity was detected even after 24 h of incubation (data not shown), confirming the involvement of Glu-269 in the catalytic machinery of ZgLamA.
Hydrolysis Pattern of Laminarin and MLG Oligosaccharides by ZgLamA GH16 -The hydrolysis of laminarin and MLG by ZgLamA GH16 was monitored by FACE for 1 h and 30 min, respectively (Fig. 3, A and B). For both substrates, oligosaccharides with a relatively high degree of polymerization were initially released, progressively followed by oligosaccharides of smaller sizes. These patterns of action indicate that ZgLamA GH16 proceeds according to an endolytic mode of action. The degradation products of ZgLamA GH16 were further analyzed by two different approaches. For laminarin, four standard ␤-1,3-glucan oligosaccharides purchased from Megazyme (DP range from 3 to 6) were digested by ZgLamA GH16 to completion. The reducing ends of the reaction products were labeled with ANTS and analyzed by FACE (Fig. 3C). After the hydrolysis of the trisaccharide, a reaction product corresponding to a disaccharide is observed. The released monosaccharide is only partially visible, because it is almost masked by the migration front of the fluorescent marker. For the other oligosaccharides, the same degradation pattern is observed (Fig. 3C). This experiment indicates that the smallest oligosaccharide that can be degraded by ZgLamA GH16 is laminaritriose, and the terminal products are glucose and laminaribiose.
To determine which linkage in MLG is hydrolyzed by ZgLamA GH16 , a glucan tetrasaccharide containing two ␤-1,4linkages separated by one ␤-1,3-linkage (G4G3G4G) was purchased from Megazyme. If the enzyme cleaved the ␤-1,3-linkage, two disaccharides G4G would be released, migrating as a single band. In the case of the cleavage of a ␤-1,4-linkage, a glucose and a trisaccharide G4G3G would be released. When the active ZgLamA GH16 is added to the previously labeled oligosaccharide, no reaction product is observed compared with the controls (Fig. 3D), suggesting that the ANTS labeling precludes the action of ZgLamA GH16 . When the labeling is undertaken after the enzymatic reaction, the reaction products migrate as two bands, corresponding to a monosaccharide and a trisaccharide (Fig. 3D). Therefore, these experiments indicate that ZgLamA GH16 specifically cleaves ␤-1,4-linkages next to Complex Structures of ZgLamA from Z. galactanivorans ␤-1,3-linkages. The trisaccharide G4G3G was not further degraded, indicating that glucose and the MLG trisaccharide are the terminal products.
Crystal Structure of ZgLamA GH16 -The crystal structure of ZgLamA GH16 was solved at a resolution of 1.5 Å by molecular replacement using the laminarinase RmLamR from R. marinus (35). The crystal is orthorhombic (P2 1 2 1 2 1 ), and its asymmetric unit contains two chains of ZgLamA GH16 , two glycerol molecules, two calcium ions (one of each in each protein monomer), and a total of 526 water molecules. Chains A and B are composed of 251 amino acids (from His-133 to Gln-383), respectively. The overall structure of ZgLamA GH16 confirms that this protein adopts the jelly roll fold typical of GH16 enzymes. The core enzyme is composed of 15 ␤-strands forming two twisted ␤-sheets. Some loops extend from the binding cleft in the regions connecting the two ␤-sheets. Par-ticularly, ZgLamA GH16 displays a long loop that is not present in other known structures of family GH16 laminarinases (Fig. 1). This additional loop is composed of 17 amino acids (from Ala-246 to Ala-262) and appears to block the negative subsites ( Fig.  4A) (subsite naming according to Davies and co-workers (46)) in comparison with the straight groove observed in RmLamR (Fig. 4B). The entire loop can be perfectly superimposed between the ZgLamA GH16 enzyme structure and that of its complexes (see below). Moreover, the B factors of the atoms composing this additional loop are not significantly different from the B factors of other protein atoms. Some strong hydrogen bonds are involved in the stability of the loop (distances are given for chain A). Glu-250 O ⑀2 is stabilized by Gln-160 N ⑀2 Taking into account all these elements, this additional loop seems to be stable and provides a bend in the catalytic cleft (Fig. 4C). A calcium ion is found on the convex side of each protein chain and is bound to the carbonyl and carboxylate O atoms of Asp-377, the carbonyl O atom of Gly-189, the carboxylate O atoms of Glu-147 and Glu-145, and one water molecule in an octahedral geometry. With the exception of plant xyloglucan endotransglycosylases/hydrolases (47,48), such a calcium binding site is well conserved in the GH16 enzymes (49), and this cation is known to increase their thermostability (50). A glycerol molecule is bound to subsite Ϫ1 of the catalytic cleft of each ZgLamA GH16 monomer mimicking a bound glucose moiety. The carbon backbone of the glycerol is stacked against the aromatic rings of Trp-238 and Trp-242; the hydroxyl group O3 is hydrogen-bonded to Trp-238 N ⑀1 (3.11Å) and Glu-274 O ⑀1 (2.82 Å), whereas the hydroxyl group O1 makes a strong hydrogen bond to Glu-269 O ⑀2 (2.61 Å).

Structure of ZgLamA GH16-E269S Complexes: Molecular Basis of Laminarin and MLG Recognition-The inactive mutant
ZgLamA GH16-E269S was cocrystallized with either purified laminarin hexasaccharides or MLG terminal degradation products, both types of oligosaccharides being produced by ZgLamA GH16 . The two complex structures were determined at resolutions of 1.35 and 1.13 Å, respectively, by molecular replacement using ZgLamA GH16 as a model. The unit cells of these complex crystals and their symmetry were similar to that of the native crystal. In both complex structures, the conserved  GH16 laminarinases (B). A, LP, lipoprotein signal peptide suggesting that ZgLamA is anchored in the outer membrane. In the N-terminal region, the signal peptide is followed by a polycysitic kidney disease-like module (PKD) composed of 102 residues. The C-terminal region is composed of a catalytic module of glycoside hydrolase family 16 (GH16). B, ZgLamA GH16 is compared with all the laminarinases structurally characterized so far. The ␣ helices and ␤ strands are represented as helices and arrows, respectively, and ␤ turns are marked with TT. This sequence alignment was created using the following sequences from the Protein Data Bank: laminarinase 16A from P. chrysosporium (2W39, residues 1-298), LamR from R. marinus (3ILN, residues 2-249), BglF from Nocardiopsis sp. F96 (2HYK, residues 4 -244), endo-1,3-␤-glucanase from Cellulosimicrobium cellulans (3ATG, residues 1-255), TM_0024 from Thermotoga maritima (3AZX, residues 7-256), Tpet_0899 from  cation binding site on the convex side of the protein is occupied by a calcium ion, as observed in the structure of ZgLamA GH16 .
The asymmetric unit of the "laminarin complex" crystal contains two ZgLamA GH16-E269S molecules (from Ala-136 to Gln-383), a sodium ion, two calcium ions, and 545 water molecules. Each protein chain binds a laminarin oligosaccharide, but only three glucose units are visible in chain A (subsites Ϫ1 to Ϫ3), whereas four glucose moieties were modeled in chain B (subsites Ϫ1 to Ϫ4). In both cases, the missing units of the laminarihexaoses are localized in the solvent region without contact to the protein and thus likely disordered. Except in a few cases (23), an oligosaccharide spanning the negative and positive subsites is rarely observed in complexed structures of enzymes from Clan GH-B (GH7 and GH16). Indeed, the sugar in subsite Ϫ1 is expected to bind with a distorted skew boat conformation to give rise to the preferred axial orientation for the leaving group (51). Such a distorted conformation is unstable, explaining the usual difficulty in obtaining a simultaneous occupation of negative and positive subsites in Clan GH-B enzymes. Because of the high resolution of the enzyme-complex structure determination, we see alternate conformations for these oligosaccharides. In chain B, the D-glucose residue in subsite Ϫ1 displays both ␣ and ␤ conformation for the hydroxyl group of the anomeric carbon C1, confirming that the sugar at subsite Ϫ1 is the reducing end of the oligosaccharide. Because of the replacement of the nucleophile Glu-269 by a serine, the glucose in ␣ conformation mimics the expected glycosyl-enzyme intermediate conformation. For both laminarin oligosaccharides, the electron density of the glucose residues in subsites Ϫ1, Ϫ2, and Ϫ3 are perfectly defined. The fourth glucose moiety is more disordered. In chain B, the four visible linked glucose units adopt a helical conformation, confirming the tendency of ␤-1,3-glucans to form helices (Fig. 4E). In both independent ZgLamA GH16 chains, the three first negative subsites are similar (Fig. 5, A and B). In subsite Ϫ1, the glucose unit in the ␣ configuration binds like the glycerol molecule in the native structure of ZgLamA GH16 and forms similar hydrogen bonds with Glu-274 and Trp-238, as well as with three additional residues Asp-271, Asn-171, and Ser-269. This glucose unit forms a hydrophobic interaction with Trp-238 and Trp-242, the latter being conserved in all family GH16 enzymes, in which it constitutes a hydrophobic platform correctly orienting the sugar ring at the Ϫ1 subsite (49). The subsite Ϫ2 is characterized by hydrogen bonds between the glucose unit and three polar amino acids (Asn-171, Glu-250, and Arg-213), with Glu-250 belonging to the additional loop unique to ZgLamA. As men-  1, 2, 4, 6, and 8). 100 g at 0.1% of the oligosaccharides from DP3 to DP6 were incubated with 4.5 M of ZgLamA GH16 at 37°C for 12 h (lanes 3, 5, 7, and 9). D, the reaction mixtures contain 0.1% (w/v) of the tetrasaccharide (G4G3G4G) and 4.5 M of active (lanes 3 and 5) or inactive (lane 4) ZgLamA GH16 in glycine buffer, pH 8,.5 at 37°C for 30 min. The asterisk indicates that the G4G3G4G oligosaccharides were labeled before the enzymatic reaction, whereas the absence of an asterisk indicates that the oligosaccharides were labeled after the reaction.
tioned before, Arg-213 plays an important role in the stability of this loop by forming a hydrogen bond with Glu-250. Moreover, this arginine likely orients Glu-250 to properly interact with the substrate. The glucose unit at subsite Ϫ2 is sandwiched by Trp-264 and His-170. The ϳ90°orientation between Trp-242, situ-ated below the sugar in subsite Ϫ1, and His-170 and Trp-264, which sandwich the sugar in subsite Ϫ2, precisely orient the laminarin chain toward the catalytic center. Finally, the glucose in subsite Ϫ3 is stabilized by a hydrogen bond between its hydroxyl group Glc3 O6 with the carbonyl of Trp-264, whereas The asymmetric unit of the "MLG complex" crystal contains two ZgLamA GH16-E269S molecules (from His-133 to Gln-383 for chain A and from His-132 to Gln-383 for chain B), two calcium ions, and 585 water molecules. Three glucose units of a MLG oligosaccharide are clearly visible in the negative subsites of each enzyme. A ␤-1,3-linkage is observed between subsites Ϫ1 and Ϫ2, whereas a ␤-1,4-linkage is positioned between subsites Ϫ2 and Ϫ3 (Fig. 6, A and B). Even decreasing the sigma level of the difference electron density map to 0.5 does not reveal any additional glucose moiety, indicating that the oligosaccharide bound to ZgLamA GH16-E269S is a trisaccharide, consistent with the result obtained with the FACE experiment (Fig. 3D). The glucose units in the subsites Ϫ1 and Ϫ2 bound similar to the laminarin oligosaccharide, except for the loss of the hydrogen bond between Glc2 O6 and Arg-213 N 2 (Fig. 6, A and B). In subsite Ϫ3, the presence of a ␤-1,4-linkage in the MLG trisaccharide produces a 180°rotation of the third glucose in comparison with the laminarin hexasaccharide, resulting in a totally different recognition of this moiety. The interaction of the hydroxyl group Glc3 O6 with the carbonyl group of Trp-264, observed in the laminarin complex (Fig. 5, A and B), is replaced by the formation of a hydrogen bond between the hydroxyl group Glc3 O3 and Glu-263 O ⑀2 in the MLG complex (Fig. 7). A structural water molecule (HOH139 in chain A and HOH166 in chain B) is also involved in the stability of the MLG complex.
Comparison of ZgLamA GH16-E269S Substrate Complexes with Other ␤-Glucanase Complexes-The ZgLamA GH16-E269S complexes have been compared with other GH16 ␤-glucanases for which complex structures are available: the laminarinase PcLam16A from the terrestrial fungus P. chrysosporium (PDB code 2W39, EC 3.2.1.6) (40) and the lichenase TFs␤-glucanase from the bovine rumen bacterium F. succinogenes (PDB code 1ZM1, EC 3.2.1.73) (41). PcLam16A was initially cocrystallized with a MLG trisaccharide G4G3G, but only a G3G disaccharide spanning the subsites Ϫ1 and Ϫ2 was discernible in the electron density map (40). The subsites Ϫ1 of ZgLamA GH16-E269S and PcLam16A are well conserved. The tryptophan residues Trp-99 and Trp-103 of PcLam16A are oriented in the same way as Trp-238 and Trp-242, respectively (Fig. 5C). PcLam16A displays an additional residue, Asn-162, showing two conformers. The "swung in" conformer makes a hydrogen bond with the hydroxyl group O2 of the glucose residue bound to subsite Ϫ1. In subsite Ϫ2 two residues are also conserved, Arg-70 and Trp-110 in PcLam16A corresponding to Arg-213 and Trp-264 in ZgLamA GH16-E269S . The conserved relative orientation of the tryptophan side chains in subsites Ϫ1 and Ϫ2 is a critical feature explaining why both ZgLamA GH16-E269S and PcLam16A require a ␤-1,3-linkage between subsites Ϫ1 and Ϫ2. In ZgLamA GH16-E269S , three additional residues strengthen the binding of the glucose moiety in subsite Ϫ2; they are Glu-250, which belongs to the loop unique to ZgLamA GH16-E269S ; His-170; and Asn-171. Subsite Ϫ3 of ZgLamA GH16-E269S is essentially constituted by Trp-264, but this key tryptophan is not conserved in PcLam16A. Moreover, the third glucose of the trisaccharide cocrystallized with PcLam16A was disordered, suggesting that this GH16 enzyme does not possess a third negative subsite (Fig. 5C).
The lichenase TFs␤-glucanase is strictly specific for MLG and was cocrystallized with a MLG trisaccharide (G4G3G) (41). With the exception of the conserved catalytic residues, ZgLamA GH16-E269S and TFs␤-glucanase have very few features in common (Fig. 6C). Both have three negative subsites, but only the binding mode in subsite Ϫ1 is partially similar with the conservation of an aromatic residue (Trp-242 in ZgLamA GH16-E269S and Phe-40 in TFs␤-glucanase) providing a hydrophobic platform for the inferior face of the glucose moiety. The conserved tryptophan Trp-141 (Trp-238 in ZgLamA GH16-E269S ) is hydrogenbonded through its intracyclic 2-amino group to the glucose hydroxyl group O6. All other residues involved in MLG recognition are not conserved between ZgLamA GH16-E269S and TFs␤-glucanase. The orientation of the MLG trisaccharides are also completely different, parallel to the inner ␤-sheet in TFs␤glucanase (Fig. 6C), whereas this oligosaccharide diverges from the ZgLamA GH16-E269S ␤-sheet by an angle of ϳ60° (Fig. 6B).
In this context, we have characterized the structure and specificity of ZgLamA, one of the five putative laminarinases found in the genome of Z. galactanivorans. ZgLamA GH16 is an endolytic ␤-glucanase highly efficient for the degradation of algal laminarin but has no significant activity on semicrystalline 1,3-␤-glucans such as curdlan and paramylon. Its minimal substrate is laminaritriose, releasing glucose and laminaribiose. ZgLamA GH16 has also a residual activity on MLG from barley, but its catalytic efficiency on this polysaccharide is 22-fold inferior in comparison with laminarin. The structure of ZgLamA GH16-E269S in complex with MLG trisaccharide (G4G3G) reveals that the ␤-1,3 glycosidic bond is localized between subsites Ϫ1 and Ϫ2. Consistently, our FACE analysis of the hydrolysis of the commercial tetrasaccharide G4G3G4G has demonstrated that ZgLamA GH16 specifically cleaves ␤-1,4linkages next to ␤-1,3-linkages (Fig. 3D). Because of its residual activity on MLG, ZgLamA GH16 is thus formally assigned the EC number 3.2.1.6, although its efficiency is much higher on laminarin. This highlights the limits of the EC nomenclature. Interestingly, the laminarinase RmLamR from R. marinus, which is relatively similar to ZgLamA GH16 (41% identity), is in fact ϳ5-fold more active on MLG than on laminarin (specific activities, 3,111 and 656 units/mg) (52).
How did such a remarkable difference in substrate specificity develop for two closely related enzymes? The comparison of the structure of ZgLamA GH16 to RmLamR (35) gives rise to an obvious explanation. The succession of ␤-1,3 bonds gives to laminarin a helical conformation (Fig. 4E). In contrast, MLG has a linear shape (Fig. 4F), because of its fine structure dominated by the presence of ␤-1,4-linkages (ϳ70%) and irregular interruptions of single ␤-1,3 bonds (53). The external rim of the RmLamR sugar-binding cleft is parallel to the inner ␤-sheet, resulting in a straight groove, which is well suited for binding linear MLG (Fig. 4D). In contrast, ZgLamA GH16 displays an additional loop that precludes the binding of a polysaccharide parallel to the ␤-sheets. Instead, this additional loop provides a bend in the active site of the enzyme (Fig. 4C), which is complimentary to the helical conformation of laminarin (Fig. 4E), explaining the high efficiency of ZgLamA GH16 on 1,3-␤-glucans and its limited activity on MLG.
The comparison of the complexes of ZgLamA GH16-E269S , PcLam16A, and TFs␤-glucanase highlights other crucial differences in the molecular recognition of laminarin and MLG by these various GH16 ␤-glucanases. Like RmLamR, the fungal laminarinase PcLam16A displays a quite straight cleft, consistent with its broad specificity for curdlan, laminarin, and MLG (40,54). The mode of laminarin recognition is partially shared between ZgLamA GH16-E269S and PcLam16A, but it is limited to the subsites Ϫ1 and Ϫ2 (Fig. 5). Moreover, ZgLamA GH16-E269S binds the glucose unit in subsite Ϫ2 more tightly than PcLam16A, with three additional amino acids involved in the recognition. ZgLamA GH16 displays a larger interaction surface because of a third negative subsite (absent in PcLam16A), which thus increases the affinity of this enzyme for laminarin.
TFs␤-glucanase and ZgLamA GH16 have been both crystallized with a trisaccharide G4G3G, but they only share a partially conserved subsite Ϫ1 in common (Fig. 6). The groove of TFs␤glucanase is with a straight topology, well adapted to bind linear MLG. A similar binding in ZgLamA GH16 would provoke a steric clash between the oligosaccharide and the additional loop of this laminarinase. Thus, although these GH16 enzymes bind the same substrate, they essentially differ in their interaction with MLG.

CONCLUSION
Phylogenetic and structural evidence supports that the common ancestor of the GH16 family featured a ␤-bulge between the conserved catalytic residues and was likely a ␤-1,3-glucanase (49,55). This hypothesis is consistent with the fact that 1,3-␤-glucan is an ancestral storage polysaccharide in eukaryotes (2) and thus a very ancient source of carbon for marine bacteria. Nonetheless, numerous laminarinases in the family GH16 hydrolyze both 1,3-␤-glucans and MLG. This is the case of the other GH16 laminarinases, which have been structurally characterized (Fig. 1B). These enzymes display a straight catalytic groove, which easily fits a linear polysaccharide such as MLG. In contrast, ZgLamA is highly efficient for the degradation of laminarin and has only a residual activity on MLG. The presence of a unique loop in ZgLamA (Fig. 1) results in a bend in the active groove more adapted to the helical shape of laminarin (Fig. 4, C and E). This bent topology is a simple alteration of the straight groove and appears as a relatively low cost evolution toward a greater specificity for laminarin. Altogether, we propose that the ancestral laminarinases in family GH16 had a broad specificity for both laminarin and MLG. The emergence of GH16 enzymes that are more specific and efficient for the degradation of laminarin is likely a more recent evolutionary event. A phylogenetic analysis of the close homologues of ZgLamA GH16 and RmLamR and of characterized laminarinases strengthens this hypothesis. Indeed, ZgLamA GH16 forms a solid clade with 12 laminarinases (Fig. 7). The additional loop is conserved in these 12 enzymes (data not shown). The clade comprising ZgLamA GH16 is rooted by several clades of laminarinases that do not possess with this specific loop (Fig. 7). These latter GH16 enzymes likely have a straight groove as observed for the structurally characterized laminarinases (PDB codes 2VY0, 3AZX, 3ATG, 2HYK, 3ILN, and 2CL2). Therefore, the ZgLamA GH16 clade emerged more recently than the clades of laminarinases with straight grooves. Such a hypothesis is also already accepted for the evolution of GH16 enzymes more specific for MLG. Indeed, bacterial lichenases (EC 3.2.1.73) have lost the ␤-bulge of the ancestral catalytic center (conserved in most clan GH-B enzymes: all GH7 and most GH16 enzymes) and thus diverge later from ␤-1,3-glucanases (49). Interestingly, a new GH16 enzyme from black cottonwood, PtEG16, has been recently characterized as a broad specificity ␤-glucanase with a nearly equal capacity to hydrolyze MLG and xyloglucans. This new subfamily has been proposed as a key evolutionary intermediate between lichenases and XET/xyloglucanases (56). For us, the discovery of PtEG16 also indicates that the MLG degradation activity has emerged at least three times within the family GH16 (broad specificity laminarinases, lichenases, and EG16). Altogether, our characterization of ZgLamA and the recent results from Brumer and co-workers (56) emphasize the complex and bumpy history of this fascinating GH family and that the evolution between broad and narrow substrate specificity can be back and forth.