The agar-specific hydrolase ZgAgaC from the marine bacterium Zobellia galactanivorans defines a new GH16 protein subfamily

Agars are sulfated galactans from red macroalgae and are composed of a d-galactose (G unit) and l-galactose (L unit) alternatively linked by α-1,3 and β-1,4 glycosidic bonds. These polysaccharides display high complexity, with numerous modifications of their backbone (e.g. presence of a 3,6-anhydro-bridge (LA unit) and sulfations and methylation). Currently, bacterial polysaccharidases that hydrolyze agars (β-agarases and β-porphyranases) have been characterized on simple agarose and more rarely on porphyran, a polymer containing both agarobiose (G-LA) and porphyranobiose (GL6S) motifs. How bacteria can degrade complex agars remains therefore an open question. Here, we studied an enzyme from the marine bacterium Zobellia galactanivorans (ZgAgaC) that is distantly related to the glycoside hydrolase 16 (GH16) family β-agarases and β-porphyranases. Using a large red algae collection, we demonstrate that ZgAgaC hydrolyzes not only agarose but also complex agars from Ceramiales species. Using tandem MS analysis, we elucidated the structure of a purified hexasaccharide product, L6S-G-LA2Me-G(2Pentose)-LA2S-G, released by the activity of ZgAgaC on agar extracted from Osmundea pinnatifida. By resolving the crystal structure of ZgAgaC at high resolution (1.3 Å) and comparison with the structures of ZgAgaB and ZgPorA in complex with their respective substrates, we determined that ZgAgaC recognizes agarose via a mechanism different from that of classical β-agarases. Moreover, we identified conserved residues involved in the binding of complex oligoagars and demonstrate a probable influence of the acidic polysaccharide's pH microenvironment on hydrolase activity. Finally, a phylogenetic analysis supported the notion that ZgAgaC homologs define a new GH16 subfamily distinct from β-porphyranases and classical β-agarases.

The main cell wall polysaccharides of marine red macroalgae are unique sulfated galactans, carrageenans or agars (1). These polysaccharides consist of a backbone of galactopyranose units linked by alternating ␣-1,3 and ␤-1,4 linkages. Whereas all of the 3-linked residues of these galactans are in the D-configuration (G unit), the 4-linked galactose units are in the D-configuration in carrageenans (D unit) and in the L-configuration in agars (L unit). A further layer of complexity is introduced by the systematic occurrence of either a 3,6-anhydro bridge or a sulfate group at C6 in the 4-linked galactose residues. Galactose 6-sulfate (referred to as D6S in carrageenans and L6S in agars) is considered as the biogenic precursor of 3,6-anhydro-galactose (referred to as DA in carrageenans and LA in agars). Indeed, the conversion of galactose 6-sulfate into 3,6-anhydro-galactose is catalyzed by galactose-6-sulfurylase (2, 3) enzymes, which have been identified only in genomes of red macroalgae (4,5). The regular structure of the backbone of red algal galactans is often masked by additional chemical modifications, such as ester sulfate groups, methyl groups, or pyruvic acid acetal groups (6 -8). This complexity has been taken into account in carrageenan nomenclature, and traditionally carrageenans are identified by a Greek prefix, indicating the major component of the sample (7). This Greek prefix nomenclature is widely used in the literature, in industry, and even in legislation. However, this system is not sufficient to describe more complex carrageenans, and a letter code-based nomenclature, inspired by the IUPAC nomenclature, was proposed by Knutsen et al. (9) for systematically describing these complex galactans. A Greek prefix nomenclature has never been introduced for agars, probably because the academic and private sectors have essentially focused on agarose, a high-gelling agar essentially devoid of modifications that is used worldwide in food industries, in molecular biology and for chromatography matrices. However, natural agars are largely as complex as carrageenans (6,8, these polymers (as we do here). To investigate the structure of natural agars, the combination of biophysical methods and specific enzymes has become a powerful strategy. The main enzymatic tools are the ␤-agarases and the ␤-porphyranases, which specifically cleave the ␤-1,4 linkages in agars and release oligosaccharides of the neo-agarobiose (LA-G) and neo-porphyranobiose (L6S-G) series, respectively (13,14). For instance, the use of bacterial ␤-porphyranases with 1 H NMR (14,15) demonstrated that the agar from Porphyra umbilicalis, commonly referred to as porphyran, is composed of one-third agarobiose motifs (G-LA) and twothirds porphyranobiose motifs (G-L6S). Tetrasaccharides with a C6-methylated D-galactose (L6S-G6Me-L6S-G) were also identified. Recent reinvestigation of these olio-porphyrans using tandem mass spectrometry (MS/MS) has revealed an even greater complexity with different degrees of methylation, substitution by uronic acids, and branching with a pentose unit (16,17). Zobellia galactanivorans Dsij T is a model alga-associated bacterium (18), which has already provided several enzymatic tools to study agar structure. Its agarolytic system includes four potential ␤-agarases and five potential ␤-porphyranases (all belonging to family 16 of the glycoside hydrolases; GH16; http://www.cazy.org 6 (19)) and at least two 1,3-␣-3,6anhydro-L-galactosidases (GH117 family) (20,21). Three ␤-agarases (ZgAgaA, ZgAgaB, and ZgAgaD) and two ␤-porphyranases (ZgPorA and ZgPorB) have already been characterized at the biochemical and structural level (13,14,(22)(23)(24). These studies notably revealed a gradient of tolerance for modifications in the agar chain, from ZgAgaD, which is strictly specific for long stretches of unsubstituted agarobiose motifs, to ZgAgaB and ZgPorB, which can tolerate substituents at some subsites (24).
The last putative ␤-agarase, which has not been characterized in Zobellia galactanivorans, is ZgAgaC. The first mention in the literature of this enzyme was the purification of the WT enzyme from the culture medium of Z. galactanivorans (13). The authors demonstrated that the extracellular agarolytic activity of this marine flavobacterium is encompassed by two enzymes: ZgAgaA, which was retained on a Sepharose CL6B affinity chromatography column, and ZgAgaC, which was found in the flowthrough of the agarose-containing column. By Edman degradation, an oligopeptide of ZgAgaC was sequenced (ATYDFTGNTP), but the agaC gene was not successfully cloned (13). Later, the sequencing of the Z. galactanivorans genome revealed that the ATYDFTGNTP peptide was only found in the ORF ZGAL_4267, which was thus named agaC (18). Here, we present the characterization of the recombinant ZgAgaC on agarose and on a natural complex agar, revealing a previously undescribed substrate specificity. The crystal structure of this enzyme gives insight into the molecular bases of its recognition of agars. Finally, an updated phylogeny of the GH16 galactanases indicates that ZgAgaC and its close homologues constitute a new subfamily within the GH16 family.

The agaC gene and its genomic context
The agaC gene encodes a protein of 328 amino acids with a theoretical molecular mass of 37.6 kDa. The protein sequence contains a lipoprotein signal peptide (residues 1-17), with a lysine at the ϩ2 position (Lys 19 ) suggesting that ZgAgaC is anchored to the outer membrane (25), a region rich in glutamate and lysine (Cys 18 -Asn 67 ) and a catalytic module of the GH16 family (Ala 68 -Glu 328 ) (Fig. 1A). The low-complexity region is predicted by the DisEMBL server (26) as being disordered (Hot-loops definition: Leu 26 -Glu 61 ; Remark-465 definition: Pro 32 -Glu 61 ), suggesting that this region is a flexible linker. Interestingly, this protein does not feature a C-terminal type IX secretion system (T9SS) domain, which is unique to Bacteroidetes (27), whereas ZgAgaC was shown to be an extracellular enzyme (13). This suggests that ZgAgaC is indeed anchored through a lipid to the outer membrane and likely oriented toward the external medium. Its secretion might be the result of a fortuitous proteolytic cleavage of the linker region.
The position of agaC in the genome is noteworthy. Although this gene is not included in a polysaccharide utilization locus (28,29), it is nonetheless found in a carbohydrate-related genomic context. Indeed, agaC (locus ID: ZGAL_4267) is located next to the mannitol utilization operon (ZGAL_4259 -ZGAL_4264) (30), the -carrageenase gene cgiA1 (locus ID: ZGAL_4265) (31,32), and a gene predicted to encode a ␤-helix fold protein (locus ID: ZGAL_4268). In a recent characterization of the carrageenolytic regulon in Z. galactanivorans (33), the genes cgiA1 and ZGAL_4268 were found to be strongly induced in the presence of carrageenans (-and -carrageenans for cgiA1; only -carrageenan for ZGAL_4268). But this was not the case for agaC despite its neighboring location (33). In another transcriptomic study on Z. galactanivorans (34), agaC was found to be induced by agar but not by porphyran. More surprisingly, this gene was also induced in the presence of laminarin, the storage polysaccharide of brown algae.

Phylogenetic analysis of GH16 family galactanases
Currently, the GH16 family includes three different types of enzymes specific for sulfated galactans from red algae: -carrageenases (35), ␤-agarases (13), and ␤-porphyranases (14). Based on pairwise sequence comparisons with characterized galactanases from Z. galactanivorans, the GH16 module of ZgAgaC appears highly divergent from these different enzymes. The closest homologue is ZgPorA (31% identity), distantly followed by ZgAgaA (23%), ZgAgaB (21%), ZgPorB (21%), and ZgCgkA (19%). To have a better idea on the relationship between these enzyme groups, we searched GenBank TM for homologues of these GH16 galactanases from Z. galactanivorans. We thus identified 154 sequences, essentially originating from diverse phyla of marine heterotrophic bacteria. Using the GH16 family laminarinases ZgLamA (36) and ZgLamB (37) as outgroups, we calculated a phylogenetic tree of these sequences ( Fig. 2 and Fig.  S1). This tree is divided into five very solid clades, with bootstrap values ranging from 86 to 99%. As expected, -carrageenases and ␤-agarases form two different clades. This analysis Characterization and structure of ZgAgaC also indicates that ZgAgaC and its closest homologues (25 sequences) constitute a monophyletic group distinct from "classical" ␤-agarases. The most surprising result is that ␤-porphyranases form two distinct monophyletic groups, the ZgPorA-like group and the ZgPorB-like group. Some relative positions of these clades are also solid. The -carrageenase clade is the closest to the root and is a sister group of all of the agar-specific enzymes (bootstrap value: 99%). The ZgPorAlike clade is the most ancestral group of agar-specific enzymes, solidly rooting a cluster composed of the three other clades (bootstrap value: 97%). Within this cluster, the relative positions between the clade of the classical ␤-agarases, the ZgAgaC-like clade and ZgPorB-like clade are unclear, because all of the nodes connecting these clades have bootstrap values below 50%.

ZgAgaC is an agar-specific enzyme degrading highly modified agars
The recombinant ZgAgaC corresponds to the catalytic GH16 module encoded by agaC (Thr 69 -Glu 328 ) without the lipoprotein signal peptide and the N-terminal low-complexity region (Fig. 1B). This His-tagged protein (31.3 kDa, pI ϭ 5.81) was produced in soluble form in Escherichia coli BL21(DE3) with a high yield of ϳ185 mg/liter of culture. A two-step purification, immobilized metal ion affinity and size-exclusion chromatography (SEC), 7 was necessary to purify ZgAgaC to electrophoretic homogeneity. The SEC analysis suggested that ZgAgaC is a monomer in solution, and this result was confirmed by dynamic light scattering.
To understand the specificity of ZgAgaC, we compared the degradation pattern of this GH16 enzyme with the already characterized ␤-agarase ZgAgaB and ␤-porphyranase ZgPorB (13, 14, 24) on diverse natural agars. Thirteen red agarophyte algae were tested with ZgAgaC, ZgPorB, ZgAgaB, and the combination of ZgPorB and ZgAgaB. The oligosaccharide degradation patterns of these GH16 enzymes were analyzed by fluorophore-assisted carbohydrate electrophoresis (FACE). For most algae (e.g. Vertebrata lanosa, Rhodomella sp., Polysiphonia simulans, and Polysiphonia brodei), the degradation pattern of ZgAgaC was very close to that of ZgAgaB and was different from that of ZgPorB. This suggests that ZgAgaC essentially behaves as a classical ␤-agarase on the agars from these algal species. Nonetheless, the degradation pattern of ZgAgaC was unique for Polysiphonia elongata and Osmundea pinnatifida (formerly named Laurencia pinnatifida) (Fig. 3). On P. elongata, most of the bands of the ZgAgaC profile are com-mon with the ZgAgaB profile; however, several bands of the ZgAgaB profile are missing in the ZgAgaC profile. Conversely, a band corresponding to a low-molecular mass oligosaccharide is only released by ZgAgaC (Fig. 3A). The most differences between the degradation patterns were observed for O. pinnatifida (Fig.  3B). ZgPorB degraded the agar of this species but did not produce a large amount of oligosaccharides. The degradation of this polysaccharide by ZgAgaB appears even less efficient, with only very few bands. The combination of ZgPorB and ZgAgaB resulted in a larger release of oligosaccharides, suggesting a synergistic effect on this substrate. In contrast, ZgAgaC was able alone to produce numerous oligosaccharides with a profile, which partially resembles that of ZgPorB but also features bands unique to ZgAgaC (Fig.  3B).

ZgAgaC enzymatic characterization
The enzymatic activity of ZgAgaC was further tested on purified polysaccharides: agarose and laminarin (Sigma), "pure porphyran" (a native porphyran pretreated with ZgAgaB to remove agarobiose motifs), and the native agar extracted from O. pinnatifida. Preliminary tests demonstrated that ZgAgaC is active on agarose and on the O. pinnatifida agar but has no activity on pure porphyran or on laminarin. To determine the mode of action of ZgAgaC, the enzymatic hydrolysis of agarose was monitored by FACE for 1 h at 40°C (Fig. 4). After 1 min, a large rangeofoligosaccharideswasreleasedbyZgAgaC,fromoligosaccharideswithhighdegreesofpolymerizationtosmalleroligosaccharides. All along the kinetic experiment, the population of larger oligosaccharides decreased with a concomitant increase of the apparent quantity of small oligosaccharides. This evolution of the degradation pattern indicates that ZgAgaC proceeds with an endolytic mode of action.
The pH dependence of ZgAgaC was studied on both agarose and the agar extracted from O. pinnatifida. Due to a problem of precipitation of ZgAgaC with the universal buffer Teorell-Stenhagen (38), several buffers were used separately to measure the pH dependence. Although, this strategy generally results in gaps between the different curves, this was not the case here, likely due to the use of 150 mM of NaCl in each buffer to maintain a constant ionic strength. Interestingly, the pH optimum of ZgAgaC is strongly dependent on the substrate used, pH 6.5 with agarose and pH 9 with the O. pinnatifida agar (Fig. 5). The temperature dependence of ZgAgaC was evaluated only with the O. pinnatifida agar, because neutral agarose forms gels below 40°C, whereas the O. pinnatifida agar remained soluble at all tested temperatures. With the O. pinnatifida agar, the activity of ZgAgaC was optimal at 50°C. The kinetic parameters of ZgAgaC have been thus evaluated at 50°C but at the pH

Purification and structure of the O. pinnatifida oligosaccharides
The oligosaccharides released by the action of ZgAgaC on complex agar from O. pinnatifida were purified by SEC as described previously (22). 48 fractions were collected and were The phylogenetic tree was generated using the maximum likelihood approach with the program MEGA version 6 (57). Bootstrap numbers are indicated. The tree was rooted by the laminarinases ZgLamA and ZgLamB from Z. galactanivorans. The clades of the -carrageenases, the ZgPorA homologues, the ZgPorB homologues, and the classical ␤-agarases have been collapsed. The sequence of ZgAgaC from Z. galactanivorans is marked with a black diamond.

Characterization and structure of ZgAgaC
analyzed by FACE gels. The three fractions (named OP30, OP36, and OP44) that contained the smallest oligosaccharides identified by FACE gels were characterized by electrospray ionization-MS (ESI-MS; Fig. 6). The major species in the OP44 fraction was measured at m/z 723.17 as [M Ϫ H] Ϫ (Fig. 6A). This oligosaccharide is attributed to a neoagarotetraose that contains two LA units, two G units, one sulfate group, and one methyl group (exact mass of the [M Ϫ H] Ϫ calculated at 723.165). This tetrasaccharide seems to be the terminal form of the products released from O. pinnatifida by the action of ZgAgaC. The OP30 fraction contained a larger species. This oligosaccharide was measured as a [M Ϫ 2H] 2Ϫ at m/z 629.14 ( Fig. 6C) and attributed to a species containing two LA units, four G (or L) units, two sulfate groups, and one methyl group (exact mass of the [M Ϫ 2H] 2Ϫ calculated at 629.131). To decipher the complete structure of this species, this ion was further studied using extreme UV dissociative photoionization (XUV-DPI) tandem mass spectrometry (MS/MS) at the synchrotron SOLEIL facility on the DISCO beamline (39). In contrast to the classical tandem MS approach, XUV-DPI MS-MS allows one to obtain a definitive structural characterization of oligosaccharides and, especially, a complete description of the methylation and sulfate patterns (16). The XUV-DPI MS-MS spectrum (Fig. 7) allows the attribution of the species appearing at m/z 629.14 to a L6S-G-(2-O-Me)-LA-G(2Pentose)-LA2S-G unit. This structure highlights the diversity of the subunits and linkages that can be found in O. pinnatifida agar and the originality of the oligosaccharide structures that can be released by ZgAgaC.

Structural comparison of ZgAgaC with ␤-agarases and ␤-porphyranases
The crystal structure of ZgAgaC was solved at high resolution (1.3 Å) using the automatic molecular replacement pipeline MoRDa (40). This program used a combination of several structures (PDB entries 4ATE, 2YCB, and 5FD3) to create the initial model. The crystal belonged to the orthorhombic space group P2 1 2 1 2, and its unit cell dimensions were as follows: a ϭ 60.6 Å, b ϭ 101.5 Å, c ϭ 46.7 Å. The asymmetric unit contains one protein chain, one magnesium ion, two glycerol molecules, two ethylene glycol molecules, and 331 water molecules. The electron density map was of high quality, allowing the modeling of the complete recombinant ZgAgaC (Thr 69 -Glu 328 ) and even two of the residues corresponding to the BamHI cloning site (Gly-Ser) as well as the N-terminal His 6 tag. Interestingly, the His 6 tag of the ZgAgaC chain of one asymmetric unit interacts with residues of the active site of a symmetric ZgAgaC chain (Fig. 8A). This unusual interaction likely favored the crystallization of ZgAgaC and strengthened the crystal packing.

Characterization and structure of ZgAgaC
which is strictly conserved in the ZgAgaC-like homologues (Fig.  S2), cannot play a similar function. Nonetheless Gln 213 , which is also located in strand ␤13 (Fig. 1D), is hydrogen-bonded to Asp 190 (Gln 213 NE2-Asp 190 OD2: 2.97 Å). This glutamine is invariant in the ZgAgaC-like subgroup (Fig. S2) and may functionally replace the usually conserved histidine.
The representation of the molecular surface of ZgAgaC indicates that its active site is an open groove (Fig. 9C). Such an active site topology is consistent with the endolytic mode of action of this enzyme (Fig. 4). With the exception of the acidic catalytic residues, the active groove displays a strong basic character (Fig. 9C). The two major positively charged patches are due to Arg 186 (negative subsites) and to Arg 224 and Lys 226 (positive subsites). These basic amino acids are invariant among the ZgAgaC-like homologues (Fig. S2) and are strong candidates for interacting with the negatively charged substituents of complex agars (e.g. sulfate groups, uronic acids).
To have a more precise idea about the molecular bases of agar recognition by ZgAgaC, we compared this structure to the structure of ZgAgaB (24) and of ZgPorA (14) in complex with their respective substrates (Figs. 10 and 11). Among the residues involved in substrate recognition, only two residues are strictly conserved with either ZgAgaB or ZgPorA: Trp 110 and Trp 177 (ZgAgaC numbering). The tryptophans equivalent to Trp 177 stack against the pyranose ring of the D-galactose moiety
Trp 110 is more significant for the specificity of ZgAgaC. Indeed, the equivalent tryptophans in ZgAgaB (Trp 109 ) and ZgPorA (Trp 56 ) do not superimpose with Trp 110 , despite their invariance in term of sequence (Fig. 12A). The distances between the C␣ of Trp 110 (ZgAgaC) and those of Trp 109 (ZgAgaB) and Trp 56 (ZgPorA) are 2.65 and 2.31 Å, respectively, and the distance differences are even greater for the side chains. This is due to the dissimilarities in length and composition of the loops between the strands ␤2 and ␤3 (8, 10, and 11 residues in ZgAgaC, ZgPorA, and ZgAgaB, respectively; Fig. 1D), which result in a different positioning of the conserved tryptophan in each type of enzyme. In the ZgPorA complex, Trp 56 is nearly perpendicular to the ring of the L-galactose-6-sulfate at the Ϫ2 subsite, whereas its NE1 atom establishes a hydrogen bond with the O6 of the D-galactose at the Ϫ3 subsite (Fig. 11). Trp 109 forms a parallel hydrophobic platform for the D-galactose moiety at the Ϫ3 subsite in the ZgAgaB complex (Fig. 10). Trp 110 (ZgAgaC) overlays neither with Trp 109 (ZgAgaB) nor with Trp56 (ZgPorA) (Fig. 12A), suggesting that this residue likely interacts with a sugar moiety in subsite Ϫ2 or Ϫ3 of ZgAgaC, but in a fashion differing from both ␤-agarases and ␤-porphyranases. Some residues of ZgAgaC occupy similar three-dimensional positions as nonequivalent residues from ZgAgAB or ZgPorA. This is the case for the guanidine group of Arg 186 (located in the strand ␤11 of ZgAgaC) and of Arg 133 (located in the strand ␤10 of ZgPorA) that nearly overlap (Fig. 11). Arg 133 is a key residue in ZgPorA, which recognizes the sulfate group of L6S at the Ϫ2 subsite (14), but it is not conserved in the ZgAgaC sequence (Fig. 1D). The functional group of Arg 219 (ZgAgaB) is also apparently close to that of Arg 186 (ZgAgaC; Fig. 10). However, a global view of the superimposition of these enzymes reveals that the side chain of Arg 219 (ZgAgaB) would clash sterically with the loop Asp 181 -Arg 186 of ZgAgaC and thus cannot play a similar role. The side chains of Trp 297 (ZgAgaC) and Trp 312 (ZgAgaB) are also roughly at the same position (Fig. 10), although these tryptophans originate from loops differing in length and sequence (Fig. 1D). Trp 312 forms a hydrophobic platform for the D-galactose bound in the ϩ2 subsite of ZgAgaB (24).
There are also some striking substitutions between ZgAgaC and the other galactanases. Notably, Glu 308 (ZgAgaB numbering) is strictly conserved not only in ␤-agarases but also in the ZgPorA-like and ZgPorB-like ␤-porphyranases (Fig. 1D). This glutamate is hydrogen-bonded to the axial hydroxyl group OH4 of the D-galactose at the cleavage subsite Ϫ1 (Fig. 10) and is

Characterization and structure of ZgAgaC
considered as an essential difference between ␤-galactanases and ␤-glucanases in the GH16 family (23,24). In ZgAgaC, this key glutamate is replaced by a tryptophan (Trp 291 ; Figs. 1D, 10, and Fig. 12B), which is strictly conserved in the ZgAgaC homologues (Fig. S2). Similarly, Gly 111 (ZgAgaB) is strictly conserved in ␤-agarases and ␤-porphyranases (Fig. 1D), likely to make room for the neighboring residues Trp 312 and Glu 308 (ZgAgaB). This glycine is replaced by Tyr 112 in ZgAgaC, which stacks against Trp 291 and points toward the Ϫ1 subsite (Fig. 12B). Tyr 112 is also invariant in the ZgAgaC subgroup (Fig. S2). Such a drastic substitution is rendered possible by the above-mentioned shift of Trp 110 (ZgAgaC). Considering their position, it is most likely that Tyr 112 and Trp 291 are involved in the recognition of the D-galactose moiety at the Ϫ1 subsite, possibly through their functional groups, -OH and -NE1, respectively.
Interestingly, the identification of unexpected ligands bound to ZgAgaC strengthens some of our hypotheses. A glycerol is bound to the Ϫ1 subsite of ZgAgaC. Its hydroxyl groups OH3 and OH2 are hydrogen-bonded to Trp 291 -NE1 and to His 63 -NE2 of the symmetrical His tag, respectively (Fig. 8B). The  (72). The labels of the ZgAgaC residues conserved with ZgAgaB are shown in blue. In both panels, the catalytic residues are underlined with a red line.

Characterization and structure of ZgAgaC
superposition of ZgAgaC and ZgAgaB confirmed that this glycerol partially overlaps the D-galactose bound at the Ϫ1 subsite of ZgAgaB (Fig. 12B). This is consistent with a crucial role of Trp 291 in the recognition of D-galactose at subsite Ϫ1. Moreover, His 61 of the symmetrical His tag is stacked against Trp 110 (Fig. 12C). This histidine likely mimics a monosaccharide, confirming the assumption of an involvement of Trp 110 in substrate recognition.
We have further attempted to co-crystallize the initial ZgAgaC construct with agar oligosaccharides. Crystals were obtained, but the determined structure was identical to the apo-ZgAgaC structure (data not shown). As an attempt to solve this problem, we have overproduced another construct of ZgAgaC with a cleavable N-terminal GST tag (Fig. 1C). This recombinant enzyme was also inactivated by mutating the acid/ base catalyst Glu 193 into a threonine. We developed a protocol for cleaving the GST tag and purifying the cleaved enzyme ZgAgaC E193T . Unfortunately, we were not yet able to crystallize this recombinant protein.

Discussion
Z. galactanivorans Dsij T was isolated from Delesseria sanguinea (50), a red alga, which produces a highly substituted agar (51,52). In its natural environment, this agarolytic bacterium is expected to feed on such complex agars, notably originating from red algae of the orders Bangiales, Corallinales, Gracilariales, and Ceramiales (which includes D. sanguinea and O. pinnatifida) (8). To date, three ␤-agarases and two ␤-porphyranases from Z. galactanivorans have been characterized (13,14,(22)(23)(24). Notably, they were shown to be more or less tolerant to some substitutions, such as the sulfation at C6 of the L unit and the methylation of the C6 of the G unit (24). However, it is yet unclear whether Z. galactanivorans can cope with other modifications found in complex agars. Our present work on ZgAgaC gives some insight into this question.
Although ZgAgaC was annotated as a putative ␤-agarase in the Z. galactanivorans genome (18), this enzyme is highly divergent from characterized GH16 ␤-agarases (23 and 21% with ZgAgaA and ZgAgaB, respectively). The transcription of the agaC gene is induced by agarose and laminarin but not by pure porphyran (34). We have demonstrated here that a recombinant ZgAgaC can degrade agarose (Fig. 4) but is inactive on pure porphyran and on laminarin. Interestingly, the induction of agaC by laminarin (whereas ZgAgaC is inactive on this brown algal compound) suggests that laminarin could act as a broad signal for inducing the catabolism of various algal polysaccharides in Z. galactanivorans. This assumption is consistent with the partial overlap of the laminarin-and alginate-induced transcriptomes of Z. galactanivorans (34). This potential signaling function of laminarin may be due to the high abundance of this storage compound in coastal ecosystems (53).  (72). The labels of the ZgAgaC residues conserved with ZgPorA are shown in blue. In both panels, the catalytic residues are underlined with a red line.

Characterization and structure of ZgAgaC
Based on these first results, ZgAgaC seems to be a classical ␤-agarase. However, an activity screening on a collection of 13 agarophytes has revealed that the patterns of released oligosaccharides by ZgAgaB and ZgAgaC differed significantly for several algal species (e.g. P. elongata and O. pinnatifida), suggesting that the substrate specificity of ZgAgaC is broader than expected. Before discussing in more detail the exact specificity of ZgAgaC, it is noteworthy that the pH optimum of ZgAgaC strongly varies, depending on the nature of the substrate (pH 6.5 and pH 9.1 with agarose and the O. pinnatifida agar, respectively; Fig. 5). The presence of a microenvironment around the polymer may explain this phenomenon. There is little documentation on the influence of a microenvironment on carbohydrate-active enzymes; however, a previous study by Li et al.
(54), demonstrates a significant positive influence of charged (nonsubstrate) polysaccharides on enzyme stability over a wide pH range. For highly sulfated agars, protons are likely required as counterions, creating an acidic environment around the surface of the polymer. In contrast, agarose does not display charged modifications, and thus there is probably no such pH microenvironment. If this is to be placed in a biological context, the pH microenvironment around the acidic polysaccharide will be different from that observed farther away from the algal surface. Thus, an apparent pH optimum of 9.1 is likely overestimated on the molecular level near the active site and does not reflect the reality of the local environment. This also raises the interesting question of whether the pH of the local environment plays a molecular role in the regulation of the activities of the different GH16 enzymes from Z. galactanivorans.
By a combination of UHPLC and XUVPDI-MS/MS, we have succeeded in characterizing purified oligosaccharides released by the action of ZgAgaC on O. pinnatifida agar. The smallest characterized product is monosulfated and monomethylated neoagarotetraose (LA-G-LA-G), although we were unable to determine the exact position of the substituents. Nonetheless, the structure of a hexasaccharide was completely elucidated: L6S-G-LA2Me-G(2Pentose)-LA2S-G. This result is consistent with the chemical composition of the O. pinnatifida agar, which was shown to contain D-xylose branches (10,11), suggesting that the pentose on the hexasaccharide is likely a D-xylose moiety. The structure of the hexasaccharide is reminiscent of that proposed for the tetrasaccharide product. This suggests that the sulfate and methyl groups of this tetrasaccharide could also be carried by the C2 of the LA units, at the reducing and nonreducing end, respectively. In any case, this MS analysis demonstrates that ZgAgaC can act on a highly complex agar motif. Taking together these oligosaccharide structures, the activity of ZgAgaC on agarose, and its inactivity on pure porphyran, we can deduce the following characteristics of the active site of ZgAgaC. (i) The subsites Ϫ2 and Ϫ1 are specific for a neoagarobiose unit, either unmodified (agarose) or sulfated at C2 on the LA moiety, and (ii) the subsites ϩ1 and ϩ2 can bind either a neoagarobiose unit (as in agarose) or neoporphyranobiose (L6S-G; as found on the nonreducing of the hexasaccharide). The fact that ZgAgaC cannot hydrolyze pure porphyran, whereas it is able to bind neoporphyranobiose at subsites ϩ1 and ϩ2, strongly indicates that the subsites Ϫ2 and Ϫ1 cannot recognize a neoporphyranobiose unit. (iii) The sub-sites Ϫ3 and Ϫ4 can bind either a neoagarobiose unit (as in agarose) or LA2Me-G(2Pentose).
The crystal structure of ZgAgaC is not particularly different from ␤-agarases and ␤-porphyranases at the fold level, almost all secondary structures being conserved (Fig. 1D). In contrast, the ZgAgaC active site is strongly modified compared with that of the other GH16 galactanases. Based on the structural comparison with the complex structures of ZgAgaB (24) and ZgPorA (14) (Figs. 10 and 11) and on the sequence alignment of the ZgAgaC homologues (Fig. S2), we can predict that the following conserved residues of ZgAgaC are involved in substrate recognition: Trp 110 (subsite Ϫ2 or Ϫ3); Arg 186 (subsite Ϫ2), Tyr 112 , Trp 177 , and Trp 291 (subsite Ϫ1); Lys 226 and Arg 244 (subsite ϩ1); and Trp 297 (subsite ϩ2). Among these residues, only Trp 110 and Trp 177 are conserved with ␤-agarases and ␤-porphyranases, but the position of Trp 110 is shifted compared with the equivalent tryptophan in the other enzymes (Fig. 12A). The most spectacular change is the substitution of a tryptophan (Trp 291 ) for the conserved glutamate (Glu 308 , ZgAgaB numbering) that is important in recognition of the D-galactose unit at the Ϫ1 cleavage subsite in both ␤-agarases and ␤-porphyranases (Fig. 12B). We propose that both Trp 291 and Tyr 112 functionally replace this crucial glutamate. These assumptions on the crucial role in substrate recognition of both Trp 110 and Trp 291 are strengthened by the observed interactions of ZgAgaC with a glycerol molecule at the Ϫ1 subsite (Fig. 8B) and with the His tag of the symmetrical ZgAgaC chain (Fig. 8C). Considering their spatial location, Arg 186 and Lys 226 /Arg 224 most likely interact with the sulfate groups of the here-characterized complex hexasaccharide (LA2S (Ϫ2 subsite) and L6S (ϩ1 subsite), respectively). Altogether, ZgAgaC recognizes agarose in a way completely different from that of classical ␤-agarases and is also adapted to bind and cleave highly substituted hybrid agars (i.e. sulfated, methylated, and/or branched oligoagars).
The updated phylogenetic tree of the GH16 galactanases ( Fig. 2 and Fig. S1) unambiguously suggests that the ZgAgaC homologues constitute a monophyletic clade distinct from the classical ␤-agarases and the ␤-porphyranases. Considering their structural and activity differences, we thus propose that the ZgAgaC-like enzymes form a new subfamily within the GH16 family. Another important result of this phylogenetic analysis is that ZgPorA-like and ZgPorB-like enzymes now constitute two solid, distinct clades. This is not a complete surprise for two reasons: (i) ZgPorA and ZgPorB display only 24% sequence identity; (ii) in the initial phylogenetic tree of the GH16 galactanases in 2010 (14), the node connecting the ZgPorA-like and the ZgPorB-like enzymes was supported by a low bootstrap value (55%). The superimposition of ZgPorA and ZgPorB confirms that these GH16 enzymes recognize porphyran in a partially different way (e.g. the L6S at the Ϫ2 subsite is not recognized by the same arginine) (14). More generally, this updated phylogenetic analysis suggests that the ZgPorA homologues are the most early diverging type of agar-specific enzymes, solidly rooting a group comprising the ZgPorB-like clade, the ZgAgaC-like clade, and the classical ␤-agarases. Moreover, the -carrageenases constitute a sister clade of all of the agar-specific enzymes. Therefore, the common ancestor of Characterization and structure of ZgAgaC GH16 galactanases was most likely an enzyme acting on complex, sulfated galactans rather than on neutral galactan. The ZgAgaC and ZgPorB homologues also act on sulfated galactans, suggesting that the classical ␤-agarases, which are more specific for neutral agarose, have diverged more recently.
To summarize, this study highlights the diversity of GH16 agar-specific enzymes, which appeared as a bacterial response to red macroalgal cell walls, to cope with the complexity of natural agars. However, this creates a practical difficulty in terms of enzyme nomenclature. How can we simply distinguish ZgAgaC-like enzymes from classical ␤-agarases, or ZgPorA-like from ZgPorB-like ␤-porphyranases? The Greek letter nomenclature of carrageenans is particularly useful for naming the different types of carrageenases (e.g. -, -, -, and ␤-carrageenases). A similar nomenclature for agars and agarases could be a solution. However, until a satisfying nomenclature is created, we recommend mentioning the type of ␤-agarases (classical or ZgAgaA-like; ZgAgaC-like) and ␤-porphyranases (ZgPorAlike; ZgPorB-like), or referring to the corresponding GH16 subfamily. 8

Experimental procedures
Except where mentioned, all chemicals were purchased from Sigma (France).

Cloning and site-directed mutagenesis of the agaC gene
The agaC gene from Z. galactanivorans encoding the ZgAgaC protein (locus identifier: ZGAL_4267, GenBank TM accession number FQ073843.1) was cloned as by Groisillier et al. (58). Briefly primers were designed to amplify the coding region corresponding to the catalytic module of ZgAgaC (forward primer, AAAA-AAGGATCCACCTATGATTTTACCGGAAACACCC; reverse primer, TTTTTTCTGCAGTTATTCCTCTACCAATTGATA-GGTATG) by PCR from Z. galactanivorans genomic DNA. After digestion with the restriction enzymes BamHI and PstI, the purified PCR product was ligated using the T4 DNA ligase into the expression vector pFO4 (58) predigested by BamHI and NsiI, resulting in a recombinant protein with an N-terminal hexahistidine tag. This plasmid, named pZG234, was transformed into E. coli DH5␣ strain for storage and into E. coli BL21(DE3) strain for protein expression. The putative nucleophile Glu 193 was replaced by a threonine by site-directed mutagenesis performed using the QuikChange II XL site-directed mutagenesis kit (Stratagene), the pZG234 plasmid, and the primers CAGGAGGGGT-TGGAGTTATTCGTTATTCGTTATAACGTCAATTTCGT-TACG and GTCCTCCCCAACCTCAATAAGCAATATTG-CAGTTAAAGCAATGC. The resulting plasmid is named pZG234 E193T . For expressing a variant of ZgAgaC E193T with an N-terminal GST tag, the cloning was performed using the In-Fusion HD Cloning Kit (Clontech), and the manufacturer's protocol was followed. Briefly, the gene was amplified from pZG234 E193T with the primers 5Ј-GGGGCCCCTGGGATCCGGATCCAC-CTATGATTTTACCGGAAA-3Ј and 5Ј-AGTCACGATGCGG-CCGCTTATTCCTCTACCAATTGATAGGTATGTATCCA-GTCTATTTTCATGGTGT-3Ј, these primers bearing the 15-bp homology necessary for InFusion cloning in pGex-6p3. pGex6p3 was digested by BamHI and NotI. All PCR amplifications were done with the high-fidelity polymerase CloneAmp (Clontech). The resulting plasmid is named pGEX_ZG234 E193T. Plasmid amplifications were performed in E. coli XL10-Gold ultracompetent cells (Stratagene).

Overexpression and purification of ZgAgaC and ZgAgaC E193T
E. coli BL21(DE3) cells harboring the plasmid pZG234 were cultivated at 20°C in a 1-liter autoinduction ZYP 5052 medium (59) 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 and 3,000 ϫ g. The cells were resuspended in 20 ml of buffer A (50 mM Tris, pH 8, 500 mM NaCl, 20 mM imidazole) and chemically lysed as described previously (21). Afterward, the lysate was clarified at 12,000 g for 30 min at 4°C, and the supernatant was filtered at 0.22 m. The supernatant was loaded onto a HyperCell PAL column charged with NiCl 2 (0.1 M) and pre-equilibrated with buffer A. The column was washed with buffer A, and the protein was eluted with a linear imidazole gradient produced by the mixing of buffer A and buffer B (50 mM Tris, pH 8, 500 mM NaCl, 500 mM imidazole) at a flow rate of 1 ml⅐min Ϫ1 . The different fractions were concentrated on an Amicon Ultra 15 unit (10 kDa; Merck Millipore) to reach a volume of 2 ml. Finally, the protein was injected onto a Sephacryl S-200 size-exclusion column (GE Healthcare) pre-equilibrated with buffer C (50 mM Tris, pH 8, 300 mM NaCl).
The fusion protein GST-ZgAgaC E193T, which has an N-terminal GST tag, was produced from E. coli BL21 cells harboring the plasmid pGEX_ZG234 E193T with the same protocol used for ZgAgaC. The cells were resuspended in 20 ml of buffer D (50 mM Tris, pH 8, 200 mM NaCl, 1 mM DTT) and chemically lysed. After clarification as described above, the supernatant was loaded onto a 5-ml GST trap 4B column (GE Healthcare) equilibrated with buffer D. The column was washed extensively with buffer D, and the elution was performed with buffer E (50 mM Tris, pH 8, 200 mM NaCl, 1 mM DTT, 50 mM GSH, 7 g⅐liter Ϫ1 agarose oligosaccharides). The agarose oligosaccharides were produced as described previously (22). For removal of the GST tag, GST-ZgAgaC E193T was incubated with the GST-tagged human rhinovirus 3C protease (GST-PreScission) (1 M) for 16 h. The cleaved protein, referred to as ZgAgaC E193T , was separated from the free GST tag and the GST-PreScission protease by injection onto a 5-ml GST trap 4B column. The column was washed with buffer F (50 mM Tris, pH 8, 400 mM NaCl, 1 mM DTT, 50 mM GSH), and ZgAgaC E193T , which has affinity for the Sepharose matrix, was eluted with buffer G (50 mM Tris, pH 8, 200 mM NaCl, 1 mM DTT, 7 g⅐liter Ϫ1 agarose oligosaccharides). A final size-exclusion chromatography was undertaken with a Sephacryl S-200 column (GE Healthcare) pre-equilibrated with buffer C.

Comparison of the pattern of oligosaccharides released by ZgAgaC, ZgAgaB, and ZgPorB
Thirteen species of agarophyte red algae were harvested in June 2016 in Roscoff (Brittany, France): O. pinnatifida, Dumontia contorta, P. simulans, P. elongata, Polysiphonia brodiei, Rhodomella sp., Chondria dasyphylla, Cryptopleura ramosa, Gracillaria sp., V. lanosa, Ceramium rubrum, Chylocladia verticillata, and D. contorta. These algae were cryo-ground with CryoMill (Retsch). For comparison purposes, the ␤-agarase ZgAgaB and the ␤-porphyranases ZgPorB from Z. galactanivorans were produced and purified as described previously (13,14). The ground algae were resuspended in the buffer of the respective tested enzymes. Each reaction mixture contained 1 M enzyme and 0.1 g⅐ml Ϫ1 of ground algae and was incubated at 35°C under agitation for 24 h. These reaction mixtures were centrifuged at 11,000 ϫ g over 20 min, and the supernatants were conserved at Ϫ20°C for subsequent fluorophore-assisted carbohydrate electrophoresis analyses.

Fluorophore-assisted carbohydrate electrophoresis
The different degradation reactions and the oligosaccharide fractions were analyzed by FACE (60). 100 l of the enzymatically degraded algae or 1 ml of the purified oligosaccharide fractions were dried in a speed-vacuum centrifuge. Oligosaccharides were labeled with 2-aminoacridone or 8-aminonaphthalene-1,3,6-trisulfonate (ANTS) as described previously (61). Briefly, for fluorophore labeling, the dried oligosaccharide pellet was dissolved with 2 l of 2-aminoacridone or ANTS solution, and 2 l of 1 M sodium cyanobohydride in DMSO was added. The mixture was incubated at 37°C for 16 h in the dark. Glycerol 20% (20 l) was added to the samples before loading onto a 31% polyacrylamide gel. The electrophoresis was carried out at 4°C in the dark for 2 h at 175 V.

Extraction and preparation of natural agars
Natural agars from Porphyra sp. and O. pinnatifida were extracted as follows. Algae were treated to obtain alcohol-insoluble residues, as described by Hervé et al. (62). Briefly, the dried algae were successively washed with 70% ethanol, 96% ethanol, methanol/chloroform (1:1, v/v), and 100% acetone. Each step was repeated three times. After this treatment, agars from Porphyra sp., referred to as porphyran, and from O. pinnatifida were extracted using an autoclave at 100°C at 1 bar over 1 h. The polysaccharides were precipitated by the addition of 4 vol-umes of ethanol and were retrieved by centrifugation at 6,000 ϫ g over 30 min.
Porphyran is usually constituted of one-third agarobiose motifs and two-thirds porphyranobiose motifs (14). It was necessary to undertake enzymatic assays on a substrate containing only porphyranobiose motifs. Thus, the native porphyran was solubilized in water at 1% and digested by ZgAgaB (13) at a final enzyme concentration of 1.5 M. The reaction mixture was filtered on Amicon Ultra 15 (3-kDa cutoff). The retentate was recovered and freeze-dried for use in the enzymatic assays. This polysaccharidic fraction is referred to as pure porphyran (i.e. without agarobiose motifs).

Enzymatic assays
ZgAgaC activity was initially tested by FACE to determine potential substrates. The amount of reducing ends produced by enzymatic digestion was followed using a method adapted from Kidby and Davidson (63). Aliquots of reaction medium (20 l) were mixed with 180 l of ferricyanide solution (300 mg of potassium hexocyanoferrate III, 29 g of Na 2 CO 3 , 1 ml of 5 M NaOH, completed to 1 liter with water). The mixture was heated to 95°C over 5 min and cooled down to 4°C. Its absorbance was measured at 420 nm. The pH optimum of ZgAgaC was determined by monitoring enzymatic activity at 25°C and in a pH range of 3.85-10.8 for each polysaccharide. Several buffers were separately used to measure the pH optimum: (i) for agarose, 100 mM phosphate was used between pH 6 and 8, 100 mM sodium acetate between pH 3.85 and 5, and 100 mM sodium borate for pH 9; (ii) for the O. pinnatifida agar, 100 mM Tris was used between pH 7 and 9 and glycine NaOH between pH 9 and 10.5. For all of these buffers, 150 mM NaCl was added. The temperature optimum of ZgAgaC was evaluated by activity measurement on the O. pinatifida agar at temperatures ranging from 25 to 65°C.
To determine the kinetic parameters of ZgAgaC, purified ZgAgaC was used at a concentration of 0.3 M for agarose and 1 M for O. pinnatifida agar. The substrate concentrations ranged from 0.0075 to 0.2% (w/v) for agarose and from 0.1 to 0.8% (w/v) for the O. pinnatifida agar. Reducing end equivalents were determined by the ferricyanide assay with galactose standard curves. The reaction buffers were 100 mM phosphate, pH 6.5, 150 mM NaCl for agarose and 100 mM glycine, pH 9, 150 mM NaCl for the O. pinnatifida agar. All of the enzymatic assays were performed at 50°C. The reactions were monitored over 80 s with a point every 10 s, and all values were determined in triplicate. The molar concentration of agarose was determined using the molecular mass of the repeating unit neoagarobiose. The K m and the k cat were determined by a nonlinear regression analysis using the program R.

Purification of O. pinnatifida oligoagars
The agar from O. pinnatifida was dissolved at 1% (w/v) and incubated at 37°C for 48 h with 1 M of purified ZgAgaC. Degradation products were ultrafiltrated with Amicon Ultra 15 (3 kDa; Merck Millipore) and the oligosaccharides contained in the filtrate were further purified as described previously (22). Briefly, the oligoagars were purified by preparative size-exclusion chromatography with three columns of Superdex 30 (26/ Characterization and structure of ZgAgaC 60) (GE Healthcare) in series, which were operated on an HPLC system (Gilson). Detection was performed using a refractive index detector (Spectra System RI-50). The purification was monitored by Unipoint Software (Gilson). The elution was performed using 50 mM ammonium carbonate ((NH 4 ) 2 CO 3 ) at a flow rate of 1 ml⅐min Ϫ1 . The fractions were collected and freeze-dried before MS analyses.

ESI MS measurements
The mass measurements were performed on a Synapt G2Si high-definition mass spectrometer (Waters Corp., Manchester, UK) equipped with an ESI source. The instrument was operated in negative polarity, as well as in "sensitivity" mode. Samples were diluted at 10 g⅐ml Ϫ1 in a solution of H 2 O/MeOH (50:50) and infused with a flow rate of 3 l⅐min Ϫ1 .

XUV-DPI tandem MS measurements
The experimental setup of the extreme XUV-DPI was developed at the SOLEIL synchrotron radiation facility at the end station of the DISCO beamline (64). A bending magnet-based synchrotron beamline was coupled to a linear ion trap (LTQ XL, Thermo Fisher Scientific). An automatic shutter was used to synchronize the photon beam (tuned to 18 eV, 68.9 nm) with the trapped precursor ions. Precursor ions were isolated with a window of 2 Da and exposed to XUV photons for 500 ms and were ejected for their measurement after a delay of 50 ms. Samples were diluted to a concentration of 50 g⅐ml Ϫ1 and infused at a flow rate of 5 l⅐min Ϫ1 . Measurements were performed in negative ion mode on the doubly charged species observed at 629.1 m/z. The nomenclature used for annotations is according to that defined by Domon and Costello (65). Raw data were processed with mMass 5.3.0 (66).

Crystallization and structure determination of ZgAgaC
ZgAgaC and ZgAgaC E193T were concentrated at 50 mg⅐ml Ϫ1 and 70 mg⅐ml Ϫ1 , respectively, and stored at 4°C in buffer C (50 mM Tris, pH 8, 300 mM NaCl). Crystallization screening was undertaken with the nanodrop-robot Crystal Gryphon (Art Robbins instruments) with four sparse-matrix-sampling kits (Qiagen and Molecular Dimensions). For ZgAgaC E193T, 0.5 mg of agar oligosaccharides (neoagarotetraose and neoporphyranotetraose, gifts from Dr. F. Le Sourd; terminal products (released by the action of ZgAgaC on the O. pinnatifida agar)) were added to the protein solution prior to the crystallization screening. The initial crystallization conditions were manually optimized, and single crystals were obtained using the hanging drop vapor diffusion method as follows. For ZgAgaC, 2 l of enzyme (50 mg⅐ml Ϫ1 ) were mixed with 1 l of reservoir solution containing 2.1 M sodium malonate and 1% glycerol. Crystallization screening of ZgAgaC E193T was also attempted with four sparse-matrix-sampling kits. Unfortunately, no crystals were obtained.
Diffraction data for a ZgAgaC crystal (Table 1) were collected at 1.3 Å resolution on the ID29 beamline (ESRF, Grenoble, France). X-ray diffraction data were integrated using XDS (67) and scaled with Scala (68). The structure was solved by molecular replacement, using the automatic pipeline MoRDa (40). The model provided by MoRDa was further manually modified and corrected using COOT (69) and refined with REFMAC5 (70).

PDB code 6HY3
Characterization and structure of ZgAgaC