LysM Domains from Pteris ryukyuensis Chitinase-A

The LysM domain probably binds peptidoglycans, but how it does so has yet to be described. For this report, we measured the thermal stabilities of recombinant LysM domains derived from Pteris ryukyuensis chitinase-A (PrChi-A) and monitored their binding to N-acetylglucosamine oligomers ((GlcNAc)n) using differential scanning calorimetry, isothermal titration calorimetry, and NMR spectroscopy. We thereby characterized certain of the domains' functional and structural features. We observed that the domains are very resistant to thermal denaturation and that this resistance depends on the presence of disulfide bonds. We also show that the stoichiometry of (GlcNAc)n/LysM domain binding is 1:1. (GlcNAc)5 titration experiments, monitored by NMR spectroscopy, allowed us to identify the domain residues that are critical for (GlcNAc)5 binding. The binding site is a shallow groove formed by the N-terminal part of helix 1, the loop between strand 1 and helix 1, the C-terminal part of helix 2, and the loop between helix 2 and strand 2. Furthermore, mutagenesis experiments reiterate the critical involvement of Tyr72 in (GlcNAc)n/LysM domain binding. Ours is the first report describing the physical structure of a LysM oligosaccharide-binding site based on experimental data.

Carbohydrates are often the signals that initiate biochemical processes. For example, as a response to pathogenic attacks, plant defense proteins first recognize and then attack the carbohydrate components of the pathogens' cell walls (1). As these defense proteins bind carbohydrates (2), via their noncatalytic carbohydrate-binding modules, their binding mechanisms are physiologically interesting. Many carbohydrate-binding modules have been structurally and/or functionally characterized (3) and, based on these characteristics, are currently grouped into 49 families in the Carbohydrate-Active Enzyme (CAZy) data base (available on the World Wide Web).
The LysM domain, which probably binds chitin, a ␤-(134)linked N-acetylglucosamine oligosaccharide ((GlcNAc) n ), was originally identified as a component of bacterial lysins. This domain is found in many of the enzymes involved in cell wall degradation and is also present in other proteins that are asso-ciated with bacterial cell walls (4 -8). Basal level resistance by plants against certain pathogens also appears to involve the recognition of chitin oligosaccharides and related compounds. Recently, a chitin oligosaccharide elicitor receptor was purified, and it contains a LysM domain within its extracellular domain (9). Additionally, certain symbiotic relationships between leguminous plants and rhizobial bacteria appear to be mediated by LysM domain/chitin oligosaccharide interactions. Recent studies have identified a class of proteins that recognize Nod factors, which are the lipochitin oligosaccharide signal molecules secreted by symbiotic bacteria (10 -12). These plant proteins are members of a serine/threonine receptor kinase family and contain extracellular LysM domains. After reviewing the available literature on LysM domains, it appears to us that its function is to bind peptidoglycans, although presently, it is not officially recognized as a carbohydrate-binding module, since it is not included in the CAZy data base.
Although the LysM domain is found in many proteins, as described above, only three tertiary structures for it have been reported. Bateman and Bycroft (13) determined the three-dimensional solution structure of the LysM domain found in the Escherichia coli membrane-bound lytic murein transglycosylase D (MltD; Protein Data Bank code 1EOG). Recently, the crystal structure of Bacillus subtilis YkuD, a protein containing one LysM domain was reported (Protein Data Bank code 1Y7M) (14). The structure of a LysM domain from the human hypothetical protein SB145 has been deposited in the Protein Data Bank (Protein Data Bank code 2DJP). These domains have a ␤␣␣␤ fold with the two helices packing against one side of the two-stranded antiparallel ␤-sheet.
Although the canonical three-dimensional LysM domain structure is now known, how a LysM domain physically interacts with an oligosaccharide is not. Bateman and Bycroft (13) reported that the loop between strand 1 and helix 1 lies at the end of a shallow groove on the surface of the domain and suggested that this region might be the ligand binding site. They based their suggestion both on an examination of the MltD solution structure and on an alignment of its sequence with other LysM domains (13). Recently, Mulder et al. (15) performed surface analyses and docking calculations using models (built by homology from the MltD structure) of the three LysM domains of the Medicago truncatula Nod factor perception protein with the goal of predicting the most favorable binding modes between chitin oligosaccharides and Nod factors. Their consensus interaction model primarily involves van der Waals interactions between a bound tetrasaccharide and the residues of the long loop between helix 2 and strand 2. Neither of the aforementioned reports is based on an experimentally defined oligosaccharide-LysM complex, so it is not necessarily surprising that the two suggested binding sites differ significantly. Obviously, the oligosaccharide binding sites for LysM domains need to be experimentally defined.
Recently, we cloned the gene (prchiA) for a novel chitinase from Pteris ryukyuensis. 2 It codes for two N-terminal LysM domains and a family 18-chitinase catalytic domain at its C terminus. For the study reported herein, we expressed and purified two recombinant LysM domains (LysM single and LysM tandem) derived from this chitinase. Using calorimetric and NMR techniques, we determined their thermostabilities and their binding affinities to various GlcNAc oligomers and characterized the carbohydrate binding site.
All clones were transformed into E. coli BL21(DE3) cells. The bacteria were grown at 37°C in Luria-Bertani broth for nonlabeled protein expression and in M9 minimal medium for uniformly 15 Nand 15 N/ 13 C-labeled protein expression. Ampicillin (50 g/ml) was included in all broths. Protein expression was induced by the addition of isopropyl-␤-D-thiogalactopyranoside at a final concentration of 1 mM. Cells were then grown for 24 h at 18°C, centrifuged at 5,000 ϫ g for 15 min, and frozen.
Frozen pellets were thawed on ice, suspended in 20 mM sodium phosphate, pH 7.0, and lysed by sonication. Insoluble material was then removed by centrifugation at 27,000 ϫ g for 30 min. The supernatants were dialyzed against 10 mM sodium acetate, pH 4.0, and precipitates were discarded. For LysM tandem, the sample was then concentrated and purified over Hi-Load Superdex 30 pg 26/60 (GE Healthcare Ltd.) in 20 mM sodium phosphate, pH 7.0, 100 mM NaCl. For LysM single, the supernatant was loaded onto a column of ToyoScreen Hexyl-650C (Tosoh Corp., Tokyo, Japan) resin. After washing the protein-loaded resin with 50 mM sodium acetate, pH 5.0, 2.5 M ammonium sulfate, protein was eluted using a linear 2.5 to 0 M ammonium sulfate gradient. The fraction containing LysM single was concentrated and purified over HiLoad Superdex 30 pg 26/60 in 20 mM sodium phosphate buffer, pH 7.0, 100 mM NaCl.
Calorimetric Studies-Thermal denaturations of both isolated LysM domains and LysM tandem in the presence of the reducing reagent ␤-mercaptoethanol (␤-ME) 3 or tris(2-carboxylethyl)phosphine (TCEP), each at a concentration of 5 mM, or in the presence of a 20-fold molar excess of (GlcNAc) n were monitored by differential scanning calorimetry (DSC) using a MicroCal VP-DSC calorimeter (MicroCal, Northampton, MA). DSC thermograms were obtained from 20 to 130°C at scan rates of 1°C/min with protein concentrations of 1 mg/ml in 20 mM sodium phosphate, pH 7.0. Base-line corrections were performed by subtracting a buffer thermogram, obtained under identical conditions, from experimental thermograms. The data were analyzed using MicroCal Origin 7.0 (Origin-Lab Corp., Northampton, MA), assuming a two-state unfolding model.
Isothermal titration calorimetry (ITC) experiments were performed with a MicroCal VP-ITC calorimeter. The instrument's design and its operation have been described in detail elsewhere (16). The LysM samples used in these experiments were prepared by first dialyzing them for about 16 h against 20 mM sodium phosphate, pH 7.0. (GlcNAc) n s were dissolved in the same buffer preparations.
For a given experiment, 7-l aliquots of 9.65-31.9 mM (Glc-NAc) n in 20 mM sodium phosphate, pH 7.0, were added via a 250-l syringe to a sample cell that contained 1.4482 ml of a stirred (310 rpm) 178 -364 M LysM solution equilibrated at 30°C. After each injection, the heat released was measured. Prior to experimentation, the instrument containing water in the sample and reference cells was equilibrated overnight. Stable base lines were defined as those with root mean square noise levels of less than 5 ncal s Ϫ1 . The heat of dilution caused by an injection of (GlcNAc) n was measured under identical buffer, injection, and temperature conditions but by adding ligand to a sample solution that lacked protein. The heat of dilution was subtracted from the heat change that occurred when protein was present. Nonlinear fitting of the data was performed using MicroCal Origin 7.0 (Origin-Lab Corp.). The parameters that were varied to minimize the S.D. of the fit to the experimental data were the binding constant (K b ), the enthalpy change (⌬H), and the number of binding sites per protein molecule (stoichiometry; n). The derived values for K b, ⌬H, and n at 30°C were then used to calculate the changes in free energy (⌬G) and entropy (⌬S).
Two-dimensional 1 H-15 N HSQC spectra were recorded for 0.25 mM 15 N-labeled LysM single in 20 mM sodium phosphate, pH 7.0, 100 mM NaCl, with and without 0.3 mM (GlcNAc) 5 . The amide 1 H, 13 N cross-peaks of the complex's HSQC spectrum were identified using assignments obtained from the threedimensional triple resonance experiments, three-dimensional CBCA(CO)HN and three-dimensional HNCACB, of a solution containing 0.6 mM 15 N/ 13 C-labeled LysM and 0.72 mM (GlcNAc) 5 .
Homology Modeling-Multiple alignments of LysM domains were performed both manually and with the ClustalX program (24). The homology modeling program COMPOSER (25), which is part of the Sybyl 7.1 software package (SYBYL Tripos Associates, St. Louis, MO), was used to build the model LysM single structure. The structurally conserved regions (defined by COMPOSER) were built by homology using the NMR solution structure of the E. coli MltD LysM domain (13). Those peptide fragments, found in the Sybyl library of tertiary structures, that are most sequentially similar to the LysM domain loops, were used to model the loops. The loop geometries were optimized using the Tripos force field (26). The stereochemistries of the resulting models were checked using the PROCHECK program (27). Connolly surfaces were calculated using the MOLCAD program (28).
Oligosaccharide Docking-Oligosaccharide docking was performed using Surflex-Dock 2.1 (29). Atomic coordinates for (GlcNAc) 2 and (GlcNAc) 4 were derived from the coordinates found for GlcNAc in the cyclooxygenase-2/indomethacin/Glc-NAc complex (PDB code 4COX) (30) and were then used for docking. The binding cavity was centered within a region defined by those residues whose NMR resonances were perturbed by (GlcNAc) 5 binding. Surflex docking was performed using the "whole molecule" approach, with a maximum of 100 simultaneously considered posed fragments at each stage of the incremental construction process and default settings for all other parameters.
Chitin-binding Assay-The ability of LysM single and its Y72A mutant to bind chitin was examined using a column, packed with powdered chitin (0.5 ϫ 10 cm), as described by Yamagami and Funatsu (31). Fifty g of each domain were dissolved in 10 mM sodium phosphate, pH 7.0, and loaded onto the chitin column equilibrated with the same buffer. After washing the chitin with the same buffer to remove unadsorbed protein, the chitin was washed 1) with 1 M NaCl in the same buffer, 2) with 0.1 M acetic acid, 3) with 1 M acetic acid, and 4) with 5 M acetic acid.
Antifungal Activity-The antifungal activities of PrChi-A mutants were tested by measuring their abilities to inhibit Trichoderma viride growth. A fungus inoculum was placed at the center of a Petri dish containing potato dextrose agar. Wells were subsequently punched into the agar at distances of 15 mm from the center of the Petri dish. Ten-l samples, containing 200 or 400 pmol of protein in distilled water, were added to the wells. The plates were incubated for 24 h at room temperature and then photographed.
Characterizing the Thermostability of the LysM Domain by DSC-The energetics of a macromolecular conformational transition (e.g. protein denaturation) can be derived from the transition's partial heat capacity's dependence on temperature, which is obtainable from DSC measurements (32,33). Fig. 2A shows the dependence of the C p of unfolding on temperature for 1.0 mg/ml solutions of LysM tandem (black line) and LysM single (red line) in 20 mM sodium phosphate, pH 7.0. The LysM tandem and single domains denature at high temperatures, with thermal transition (T m ) values of 90.0 and 96.5°C, respectively. Attempts to refold the domains by reducing the temperature failed, indicating that the thermal denaturations were completely irreversible for both domains. When heat denaturation causes protein aggregation, the higher temperature side of the corresponding DSC curve sharply decreases as a consequence of exothermic effects. At least in the case of the LysM tandem domain, irreversible denaturation is possibly caused by aggregation.
It is likely that the four cysteines found per domain form two disulfides. To determine if cysteine thiols are present in LysM tandem, LysM single, and/or LysM single Y72A, the proteins were assayed with the sulfhydryl-modifying reagent, 5,5Ј-dithiobis(2-nitrobenzoic acid) (34). No free thiols were detected in the proteins, suggesting that all cysteines participate in disul-fide bridges (data not shown). (Other indirect evidence for the presence of disulfide bonds is discussed below.) To determine whether such likely disulfide bonds contribute significantly to domain stability, the DSC experiment using LysM tandem was repeated, but this time in the presence of 5 mM ␤-ME or 5 mM TCEP (Fig. 2B). Since TCEP is more stable and a more effective reducing reagent than is ␤-ME (35) and since reducing reagents sometimes produce spurious calorimetric data, using two different reducing reagents seemed prudent. For LysM tandem, in the presence of ␤-ME, its T m value is 77.4°C, and in the presence of TCEP, its T m value is 68.4°C. Apparently, there are disulfides present in LysM tandem that remarkably stabilize its structure.
The effects of GlcNAc oligomers on the thermal stability of LysM tandem were also studied. Fig. 2C shows the temperature dependence of C p for uncomplexed LysM tandem (black line) and LysM tandem in the presence of (GlcNAc) 3 (green line), (GlcNAc) 4 (blue line), and (GlcNAc) 5 (red line) after base-line subtraction. The T m values are somewhat affected by the presence of the oligosaccharides: 89.9°C for the isolated domain and for the complexes, 90.2°C in the presence of (GlcNAc) 3 , 90.9°C in the presence of (GlcNAc) 4 , and 91.6°C in the presence of (GlcNAc) 5 . The T m values for LysM tandem unfolding increased slightly as the length of the GlcNAc oligomer increased.  Thermodynamics of (GlcNAc) n /LysM Binding Monitored by ITC-ITC is the most direct method for measuring the heat of binding at constant temperature. A ligand is titrated into a solution containing its binding partner with the result that the complex binding constant (K b ), the enthalpy change for binding (⌬H), and the stoichiometry of the complex (n) can be derived (32,36). ITC measurements of (GlcNAc) n (n ϭ 3, 4, and 5) binding to the LysM constructs were measured at 30°C. Fig. 3 shows the results for an ITC experiment when 7-l portions of 9.65 mM (GlcNAc) 5 were titrated into a 1.4482-ml solution containing 129.8 M LysM single in 20 mM phosphate, pH 7.0. In Fig. 3A, the area within each trough is a measure of the heat released upon the addition of (GlcNAc) 5 . Fig. 3B is a plot of the heat released, normalized to 1 mol of oligosaccharide, versus the molar ratio of LysM single to a GlcNAc oligomer. The theoretical "best fit" binding curves are also shown in Fig. 3B. No matter which GlcNAc oligomer was used, when n was free to vary during fitting, the data were best fit with a stoichiometry of approximately 1 for the LysM single complex data and a stoichiometry of approximately 2 for the LysM tandem complex data (data not shown).
The Overall Modeled Structure for a PrChi-A LysM Domain-Although the sequences for more than 1,500 proteins that con-tain a LysM domain are listed in the PFAM data base (37), the three-dimensional structures for only three of those domains have been solved. The NMR solution structure of an isolated LysM domain from the E. coli membrane-bound lytic murein transglycosidase D (MltD) has been reported (13). A second structure is part of the B. subtilis YkuD crystal structure (14). A third structure, which has been deposited in the Protein Data Bank, is part of the human hypothetical protein SB145. These

The Chitin-binding Site for a LysM Domain
would be Cys 60 with Cys 104 and Cys 71 with Cys 94 . These disulfide bonds were added to the model before refinement, and their geometries were optimized using the Tripos force field. The overall quality of the resulting model was verified by the PROCHECK program. Fig. 4A displays the model of the PrChi-A LysM single domain, and Fig. 4B displays the template structure, which is the MltD LysM domain. The modeled PrChi-A LysM single has a ␤␣␣␤ fold, with a geometry that is almost same as that of the MltD LysM domain.
Identification of the Residues of LysM Single That Interact with (GlcNAc) 5 -To identify which residues of LysM single interact with (GlcNAc) 5 , we recorded 1 H-15 N HSQC spectra at 30°C for LysM single in the presence and absence of (Glc-NAc) 5 . These experiments identified the protein amide 15 N and 1 H chemical shifts that are affected by (GlcNAc) 5 and thereby identify the residues that interact directly or indirectly with (GlcNAc) 5 (Fig. 5A). Whether (GlcNAc) 5 was present or not, the LysM single spectra contain well dispersed, intense signals, indicating well defined molecular structures. Certain LysM sin-gle resonances shift dramatically when (GlcNAc) 5 is present, presumably because their corresponding protein residues interact with (GlcNAc) 5 in a LysM-(GlcNAc) 5 complex.
The backbone amide resonances of the 1 H-15 N HSQC spectra of LysM single in the presence and absence of (GlcNAc) 5 were assigned using the correlations obtained from the CBCA-(CO)NH and HNCACB spectra. The cross-peaks of a CBCA-(CO)NH spectrum can be used to correlate a residue's 15 N, HN amide resonances to the C␣ and C␤ resonances of the preceding residue; a HNCACB spectrum provides the means to correlate intraresidue 15 N, HN, C␣ and 15 N, HN, C␤ resonances. The residue assignments are given in Fig. 5A.
The cross-peaks of five residues, Ser 81 , Glu 86 , Ala 90 , Cys 94 , and Asn 95 , displayed significant chemical shift perturbations when (GlcNAc) 5 was present (Fig. 5, A and B). The amide signals of five other residues, Gly 68 , Thr 70 , Cys 71 , Tyr 72 , and Ile 74 , were undetectable when the oligosaccharide was present, probably because of extreme peak broadening, which would be indicative of an intermediate binding exchange rate compared with the NMR time scale. The residues with chemical shifts most affected by (GlcNAc) 5 are colorcoded in the model of LysM single (Fig. 6A). Gly 68 is located at the loop that resides between strand 1 and helix 1, and residues Thr 70 -Ile 74 are positioned near the N-terminal of helix 1. Glu 86 is located at the C-terminal of helix 2, and Ala 90 , Asn 94 , and Ala 95 are part of the loop between helix 2 and strand 2. These residues are all found on one side of the molecule and form a long shallow groove at the molecular surface.
(Evidence for the groove is given below.) Docking of (GlcNAc) n onto the PrChi-A LysM Single Model-Accessible surface calculations on PrChi-A LysM single were performed with the MOLCAD program. The results of these calculations allowed us to identify a few cavities that might be GlcNAc binding sites. Most notably, 1) there is a large cavity that is formed by residues whose NMR resonances are not affected by oligosaccharide binding, but 2) a cavity associated with those residues whose resonance are affected by oligosaccharide binding could not be readily discerned. We therefore performed a second accessible surface area calculation within the region defined by the residues whose resonances are affected by (GlcNAc) 5 binding. A shallow groove delineated by those residues could be discerned, and it was used as the LysM docking surface. Docking simulations were then performed with Surflex-Dock. Docking was initially performed using (Glc-NAc) 2 and the model of LysM single. The results of this simulation identified many possible docking sites (data not shown). We then constructed a model of (GlcNAc) 4 using the (Glc-NAc) 2 coordinates and docked (GlcNAc) 4 to LysM single. The binding modes of (GlcNAc) 4 for all docking runs are very similar to each other for the portion of the oligosaccharide found to dock at the N-terminal region of helix 1 and the loop between strand 2 and helix 2 (lower region of the molecule in Fig. 6B). However, the binding modes for the oligosaccharide were much more diverse at the top of the groove (Fig. 6B).
Mutational Analysis-Because the binding of carbohydrates to proteins is known to often be mediated by aromatic residues (3), we postulated that mutation of Tyr 72 , which is one of the residues whose NMR amide signals became undetectable in the presence of (GlcNAc) 5 , might disrupt LysM chitin binding and enzymatic antifungal activity of PrChi-A. To establish if the conformations of the wild-type and mutant forms of LysM single are the same, we compared their 1 H-15 N HSQC spectra. The amide signals of the mutant are as well dispersed as those of the wild type spectrum, and in many cases the cross-peaks of the two spectra overlap (Fig.  7A). Furthermore, the thermostability of LysM single Y72A, as measured by DSC, is the same as that of the wild type. Both the shapes of the thermograms and the T m values of the two proteins coincide (data not shown). Therefore, LysM single Y72A and the wild type domain have virtually the same fold. Having eliminated the possibility that loss of activity upon mutation could be attributed to a conformational change, we then applied LysM single and LysM single Y72A to the chitin column to evaluate the chitin binding ability of these proteins by comparing their relative elution positions. As shown in Fig. 7B, LysM single was retained on the column, demonstrating its chitin binding ability. Although it was eluted as separate peaks, LysM single was mainly eluted from the column by 1 M acetic acid. On the other hand, approximately half of the applied Y72A mutant passed through the column, and almost all of the rest of it eluted with 0.1 M acetic acid, indicating that Tyr 72 in LysM single is responsible for chitin binding of the LysM domain. We then postulated that since mutation of Tyr 72 to Ala disrupts LysM chitin-binding, it might also disrupt the enzymatic antifungal activity of PrChi-A, which is the case (Fig. 7C). The antifungal activity of the Y72A site-directed mutant of ⌬1LysM, which consists of the second LysM domain and the catalytic domain of PrChi-A, was significantly reduced in comparison with ⌬1LysM (Fig. 7C). The binding of (GlcNAc) 5 to LysM single Y72A was monitored using ITC under the same conditions as used for the wild-type domain (Fig. 7D). The binding affinity for (GlcNAc) 5 by LysM single Y72A is less than that of LysM single. Therefore, Tyr 72 plays important roles in chitin binding and fungal cell wall binding.

DISCUSSION
There is intense interest in LysM domain proteins because of their important roles in chitin oligosaccharide recognition and their widespread occurrence. In higher plants, numerous genes for LysM domain-containing proteins have been identified, although there is only limited information concerning their three-dimensional structures. Therefore, little is known about their structure-function relationships, specifically chitin oligosaccharide binding modes.
Disulfide Bonds Structurally Stabilize the PrChi-A Domain-We have shown, using DSC experiments (Fig. 2), that LysM single and LysM tandem are thermostable (i.e. have high T m values). DSC measurements were also made on LysM tandem with the reducing reagents, ␤-ME and TCEP (35), present (Fig.  2B). The T m value of LysM tandem decreased in the presence of FIGURE 7. A, two-dimensional 1 H-15 N HSQC spectra of LysM single (red-colored cross-peaks) and LysM single Y72A (green-colored cross-peaks). The residues whose resonances shift significantly due to the mutation are labeled. The side-chain amide signals of Asn and Gln residues are connected by horizontal lines. B, elution profiles of LysM single (top) and LysM single Y72A (bottom) from a chitin affinity column chromatography. Protein was loaded onto a chitin column equilibrated with 20 mM phosphate, pH 7.0. After unabsorbed protein was removed by washing with the same buffer, the adsorbed protein was eluted successively with 1 M NaCl, 0.1 M acetate, 1 M acetate, and 5 M acetate. C, antifungal activity of ⌬1LysM-Y72A, ⌬1LysM, and CatD. (The constructs are shown in the figure.) The test samples (200 or 400 pmol of a purified chitinase in 10 l of distilled water) were individually placed into wells positioned around a field of T. viride. D, an isothermal microcalorimetric profile of a titration of a 9.65 mM (GlcNAc) 5 solution into a 129.8 M LysM single Y72A solution. The area of each trough corresponds to the heat generated after the addition of 7 l of (GlcNAc) 5 to 1.4482 ml of protein solution.