Binding of the AVR4 Elicitor of Cladosporium fulvum to Chitotriose Units Is Facilitated by Positive Allosteric Protein-Protein Interactions THE CHITIN-BINDING SITE OF AVR4 REPRESENTS A NOVEL BINDING SITE ON THE FOLDING SCAFFOLD SHARED BETWEEN THE INVERTEBRATE AND THE PLANT CHITIN-BINDING DOMAIN*

The attack of fungal cell walls by plant chitinases is an important plant defense response to fungal infection. Anti-fungal activity of plant chitinases is largely re-stricted to chitinases that contain a noncatalytic, plant-specific chitin-binding domain (ChBD) (also called Hevein domain). Current data confirm that the race-specific elicitor AVR4 of the tomato pathogen Cladosporium fulvum can protect fungi against plant chitinases, which is based on the presence of a novel type of ChBD in AVR4 that was first identified in invertebrates. Although these two classes of ChBDs (Hevein and invertebrate) are se-quentially unrelated, they share structural homology. Here, we show that the chitin-binding sites of these two classes of ChBDs have different topologies and charac-teristics. The K D , (cid:1) 4.6, and 50 m M sodium chloride. Isotopic labeling and purification of AVR4 was performed as described (37). All of the NMR samples were prepared in a mixture of 95% H 2 O, 5% D 2 O (v/v) and contained trace amounts of sodium azide as preservative. All of the NMR spectra were acquired at 298 K on Varian Inova 500, 600, or 800 MHz, and Bruker AMX500 spectrometers. Triple and double resonance heteronuclear NMR experiments performed to obtain backbone and side chain assign- ments of AVR4 included three-dimensional HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, (H)CCH-TOCSY, and HC(C)H-TOCSY (Pro- tein Pack; Varian Inc.). The assignment was performed using the standard assignment procedures based on triple and double resonance NMR spectra. First, 15 N HSQC spectra were used to obtain a set of 1 H- 15 N resonance frequencies. Sequential assignment was then performed us- ing these shift pairs in combination with HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH spectra. Assignment of the nonaromatic side chain resonances was obtained by means of (H)CCH-TOCSY and HC(C)H-TOCSY spectra. A 15 N NOE spectroscopy-HSQC (43) spectrum was used for NOE assignments of the backbone 1 HN and tryptophan side chain HE1 protons. All of the data processing and analysis were done using the programs NMRPipe (44) and XEASY (45), respectively. The chemical shifts of Tachycitin were retrieved from BioMagResBank (www.bmrb.wisc.edu). NMR Titration Experiments— Binding of chito-oligomers to AVR4 was followed by recording 1 H- 15 N HSQC spectra at five different tem-peratures. Thereto, temperature in the NMR tube was carefully cali- brated by referencing the water resonance to sodium 2,2-dimethyl-2-silapentane-5-sulfonate. The protein concentration All were performed in m M potassium phosphate, m M sodium chloride, pH 7.0. The average result of at least three independent experiments is shown Thermodynamics of AVR4 binding to chito-oligomers using isothermal titration calorimetry.

Binding and conversion of carbohydrates by proteins is of fundamental importance in numerous biological processes, including (self and nonself) cell-cell recognition, cell adhesion, and carbohydrate turnover. Recently, protein domains responsible for this interaction have been reclassified into distinct carbohydrate-binding modules (CBMs) 1 (1). CBMs are often present in carbohydrate-degrading enzymes, where they appear to mediate a prolonged and more intimate contact between the catalytic domain and insoluble carbohydrate polymers (2,3). Lectins, on the other hand, are carbohydratebinding proteins that lack enzymatic activity but often contain tandem repeats of CBMs.
Chitin, a polymer consisting of ␤-1,4-linked GlcNAc residues, is a major component of crustacean shells, insect exoskeletons, and fungal cell walls but is absent in plants. In higher organisms, two CBMs predominantly confer binding of proteins to chitin, i.e. the Hevein domain (hereafter denoted as CBM18) (4) and the invertebrate chitin-binding domain (CBM14) (5). CBM18 is nearly exclusively found in plants (to date one additional member of CBM18 has been identified in Streptomyces griseus; Ref. 6), whereas CBM14 is commonly found in the genomes of baculoviridae, invertebrates, and mammals but absent in plants (5). Both ChBDs are typical CBMs, i.e. lectins with tandem repeats are known for both (e.g. wheat germ agglutinin (WGA) (7) and peritrophin-44 (8)), and both domains can be found in chitinases. However, CBM18 is only fused to the plant-specific family 19 catalytic domain, whereas chitinases of mammals and invertebrates utilize CBM14 in combination with the family 18 catalytic domain. Sequence homology is missing between the two motifs, but the three-dimensional structure of Tachycitin revealed that CBM14 and CBM18 partially share their tertiary structure (9).
The race-specific elicitor AVR4 of the tomato pathogen Cladosporium fulvum is hitherto the only fungal protein containing a CBM14 (10). 2 AVR4 binds specifically to chitin (10) 2 and appears to have a high affinity for crude fungal components that resist harsh treatments such as heating and treatment with proteinase K (12). Originally, AVR4 was identified as an extracellular race-specific elicitor of C. fulvum that induces plant defense responses in tomato plants carrying the comple-mentary Cf-4 resistance gene. Recognition of AVR4 is sufficient for induction of complete resistance in tomato against isolates of the fungus C. fulvum that carry the AVR4 encoding gene (13,14). The natural isolates of C. fulvum that were found to evade Cf-4-mediated resistance were reported to secrete proteasesensitive isoforms of AVR4, whereas native mature AVR4 (86 amino acids) is insensitive to these proteases (10,14). The corresponding avr4 alleles in these isolates all contain single nucleotide polymorphisms causing in all but one case single amino acid substitutions (13). These mutations appear to have no direct effect on the chitin binding properties of the isoforms as was shown for a set of Cys-to-Tyr mutations (10).
C. fulvum is reported to be insensitive to a combination of tomato chitinases and ␤-1,3-glucanases, at least under in vitro conditions (28). Studies using two other fungi, i.e. Trichoderma viride and Fusarium solani f.sp. phaseoli, showed that AVR4 can protect these two fungi against anti-fungal activity of PR-3 chitinases. 2 The protective effect was further substantiated by the observation that AVR4 binds to chitin present in the cell walls of these two fungi.
To better understand the role of AVR4 during infection of tomato, we here examined the binding properties of AVR4 to chitin using soluble chito-oligomers. This system allows for a detailed comparison between AVR4 and CBM18 lectins (e.g. Hevein, Prohevein, UDA, and WGA). For the CBM18 lectins, the use of chito-oligomers has provided a detailed description of the chitin-binding site. In CBM18 lectins, the binding site consists of three binding subsites (a subsite is defined as all amino acids that interact with one sugar residue). Subsite ϩ1 is formed by the residues Ser 19 , Trp 23 , and Tyr 30 , whereas Trp 21 is involved in subsites ϩ2 and ϩ3 (29 -36). A hallmark of the CBM18 lectins is that they already interact with one GlcNAc residue. Here we show that binding of AVR4 requires at least a stretch of three GlcNAc residues. Using NMR, we identified several residues in AVR4 that are important for ligand binding. These residues are indeed positioned in the structural motif shared by CBM14 and CBM18, but they appear to highlight different binding sites rather than overlapping binding sites as compared with CBM18 (9).
Isothermal Titration Calorimetry-ITC measurements were performed at 298 K following standard procedures using a Microcal MCS titration calorimeter (40). The reaction cell (with a volume of ϳ1.35 ml) containing the AVR4 protein sample was continuously stirred while successive aliquots of ligand solution were added (final volume of the additions was 250 l). Ligand and protein were dissolved in the same buffer. The AVR4 concentration in the cell was in the range of 90 -360 M depending on degree of polymerization (DP) of the chito-oligomer added (see legend of Fig. 1). The chito-oligomer concentrations used were 23, 20, 3.2, and 2.0 mM for DP ϭ 3, 4, 5, and 6, respectively. The integrated heat effects after correction for heat of dilution were analyzed using standard software provided by Microcal Inc. The cumulative heat effect (Q) during the titration process for a simple set of binding sites is given by the following equations, where M t is the macromolecule concentration in the calorimetric cell, characterized by the volume (V 0 ), n is the number of binding sites with a binding enthalpy of ⌬H, and is the fractional saturation of the binding sites, which can be related to the apparent association constant (K A ) and to the total ligand concentration (L T ), where L f is the concentration of free ligand. Other thermodynamic parameters were calculated using the following standard thermodynamic equation.
Tryptophan Fluorescence Quenching-The fluorescence measurements were performed with a Varian Cary Eclipse thermostatted at 293 K. The excitation wavelength was 295 nm with an excitation slit of 2.5 nm. Emission intensities were collected over the wavelength range of 315-400 nm with an emission slit of 5 nm. The spectra were the averages of three scans and corrected for the effect of dilution, buffer, and chito-oligomer additions. Quantitative binding experiments were performed in a volume of 3 ml to which aliquots of a ligand solution (5-30 l) were added under continuous stirring. AVR4 was dissolved at a protein concentration of 3.6 M in 20 mM potassium phosphate buffer, pH 7.0, containing 50 mM sodium chloride. Chito-oligomers were dissolved in the same buffer at a concentration of 38, 30, 16, and 2.0 mM for a DP ϭ 3, 4, 5, and 6, respectively. The maximum change in volume caused by the ligand additions was less than 5%. The fluorescence quenching at full saturation of binding (F ϱ ) was estimated by plotting 1/(F 0 Ϫ F) versus 1/[S], and extrapolating to the y axis, where F 0 is the fluorescence intensity of AVR4 without ligand, and F is the fluorescence intensity of AVR4 at the chito-oligomer concentration [S]. Association constants (K A ) were estimated using two methods: the fluorescence quenching titration equivalent of the Hill Plot (i.e. log (F 0 Ϫ F/F Ϫ F ϱ ) versus log [S]) (41) and Scatchard plot analysis (i.e. /L f versus ) (42).
Size Exclusion Chromatography-pH-dependent size exclusion chromatography was performed at 293 K using a Superdex-75 (HR 10/30; Amersham Biosciences) column operated at a flow rate of 0.5 ml/min. The apparent molecular mass of the oligomeric/complexed state of AVR4 (25 M in 50 l of injection volume) was estimated from a standard curve produced at different pH values (5.0, 7.0, and 8.6) in buffer containing 50 mM potassium phosphate and 150 mM potassium chloride. The standard curves were obtained by plotting the log molecular mass of protein standards (aprotinin, insulin, ubiquitin, ribonuclease 5A, serum albumin (all bovine), horse myoglobin, chicken albumin, and blue dextran) versus K av . The K av is defined as follows, where V E is the elution volume, V V is the void volume, and V B is the bed volume of the column matrix.
Mass Spectrometry-ESI-MS was performed with a Q-Tof Ultima Global mass spectrometer (Waters Corporation, MS Technologies Centre, UK). AVR4 and the chito-oligomers were dissolved in 10 mM ammonium acetate/acetic acid (pH range, 5.0 -8.6). The sample infusion flow rate was 10 l/min. The instrument settings were: capillary potential, 3 kV; cone voltage, 100 V; desolvation gas flow rate, 150 liters/h; source temperature, 90°C; and radio frequency 1, 225 kHz; and the matrix-assisted laser desorption ionization strip was positioned at 3600 arbitrary units resulting in an elevated intermediate pressure of 4.35 mbar. The instrument was operated under standard ESI conditions. Calibration of the TOF analyzer was performed with a CsI solution of 2 mg/ml in isopropanol/water (50:50, v/v) over the mass range of 800 -7100 Da.
Nuclear Magnetic Resonance Spectroscopy-The NMR samples contained typically 1.5 mM 13 C/ 15 N AVR4 dissolved in 20 mM acetate-d 4 , pH 4.6, and 50 mM sodium chloride. Isotopic labeling and purification of AVR4 was performed as described (37). All of the NMR samples were prepared in a mixture of 95% H 2 O, 5% D 2 O (v/v) and contained trace amounts of sodium azide as preservative. All of the NMR spectra were acquired at 298 K on Varian Inova 500, 600, or 800 MHz, and Bruker AMX500 spectrometers. Triple and double resonance heteronuclear NMR experiments performed to obtain backbone and side chain assignments of AVR4 included three-dimensional HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, (H)CCH-TOCSY, and HC(C)H-TOCSY (Protein Pack; Varian Inc.). The assignment was performed using the standard assignment procedures based on triple and double resonance NMR spectra. First, 15 N HSQC spectra were used to obtain a set of 1 H-15 N resonance frequencies. Sequential assignment was then performed using these shift pairs in combination with HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH spectra. Assignment of the nonaromatic side chain resonances was obtained by means of (H)CCH-TOCSY and HC(C)H-TOCSY spectra. A 15 N NOE spectroscopy-HSQC (43) spectrum was used for NOE assignments of the backbone 1 HN and tryptophan side chain HE1 protons. All of the data processing and analysis were done using the programs NMRPipe (44) and XEASY (45), respectively. The chemical shifts of Tachycitin were retrieved from BioMagResBank (www.bmrb.wisc.edu).
NMR Titration Experiments-Binding of chito-oligomers to AVR4 was followed by recording 1 H-15 N HSQC spectra at five different temperatures. Thereto, temperature in the NMR tube was carefully calibrated by referencing the water resonance to sodium 2,2-dimethyl-2silapentane-5-sulfonate. The protein concentration was kept constant during the titration (1 mM), whereas the ligand concentration was increased in successive steps (30). As buffer we used 20 mM acetate-d 4 , pH 6.0, and 150 mM NaCl. Final concentrations of the ligands were 50 mM GlcNAc, 35 mM (GlcNAc) 2 , 27.2 mM (GlcNAc) 3 , and 3.5 mM (Glc-NAc) 6 . The association constants (K A ) were estimated using a NMR derivative of the Scatchard plot, where ⌬ ϭ ␦ observed Ϫ ␦ free and ⌬ 0 ϭ ␦ saturated Ϫ ␦ free . Thermodynamic parameters (⌬H and ⌬S) were estimated from a van't Hoff plot based on a set of K A values obtained from a set of backbone resonances.
Calorimetric Titrations-ITC was used to determine the affinity of AVR4 for linear soluble chitin fragments. No heat of binding was detected when 50 mM GlcNAc or 50 mM (GlcNAc) 2 (chitiobiose) was added to AVR4. The addition of longer chitooligomers resulted, however, in substantial heat of binding (Table I). The binding curves obtained for (GlcNAc) 3 , (GlcNAc) 4 , and (GlcNAc) 5 could be fitted, assuming a onebinding site model (Fig. 1, A-C). The binding curve obtained for (GlcNAc) 6 deviated from a one-binding site model. However, a model with two AVR4 binding sites per one (GlcNAc) 6 molecule described accurately the binding event (Fig. 1D). This second AVR4-binding site was not observed for (GlcNAc) 5 (Fig. 1C).
The dissociation constant (K D ) found for the binding of (GlcNAc) 6 to AVR4 is ϳ45 times lower than the K D observed for (GlcNAc) 5 and ϳ200 times lower than the K D observed for (GlcNAc) 4 (Table I). This decreased K D for (GlcNAc) 6 originates from a steep decrease in ⌬H as compared with ⌬H for (GlcNAc) 5 (concomitantly, a small negative contribution to binding comes from a slight decrease in the ⌬S). The substantially decreased ⌬H could indicate that positive allosteric interactions occur between the two AVR4 molecules that bind to one (GlcNAc) 6 . Additionally, a substantial part of the decreased ⌬H could originate from a decrease in the solvent-exposed area of the two bound AVR4 molecules. The existence of such a protein-protein interaction is supported by the fact that the number of binding sites (n) appears to be pH-dependent with an apparent pK a of ϳ4 (Table II). This implies protonation of an acidic residue (i.e. Asp or Glu), which could effectively disrupt an interaction at the protein-protein interface.
The slight decrease of ⌬S found for the binding of (GlcNAc) 6 to AVR4 as compared with (GlcNAc) 5 and (GlcNAc) 4 suggests that an increased number of translations and rotations in the sugar chain is restrained upon binding of AVR4 to (GlcNAc) 6 . Potentially each of the GlcNAc residues is restrained because of the interaction with the two AVR4 molecules leading to an overall reduced flexibility of (GlcNAc) 6 . Correspondingly, a certain degree of freedom should remain in the chito-oligomer chain for (GlcNAc) 4 and (GlcNAc) 5 when bound to AVR4. Indeed, ⌬S is increased for both (GlcNAc) 4 and (GlcNAc) 5 as compared with (GlcNAc) 3 . Hevein differs from the foregoing situation in that two Hevein molecules already bind to (GlcNAc) 5 rather than (GlcNAc) 6 (33). The corresponding K D is ϳ45 times lower than the K D for (GlcNAc) 4 . In this case, ⌬S is the main contributor to the decreased K D (Fig. 2), which originates from the existence of several Hevein-(GlcNAc) 5 complexes with different stoichiometries (1:1 and 2:1 protein ligand complexes) (30,33).
This cost of restraining translations and rotations of the GlcNAc chains has been given as one explanation for the "enthalpy-entropy" compensation generally observed for lectinsugar interactions, which results in K D values in the micromolar to millimolar range (48) (see also supplementary data). The correspondingly small and negative ⌬H and ⌬S, as obtained here for AVR4 and as reported for other lectins such as Hevein (30,33), Prohevein (32), and UDA (42) (Fig. 2), points out that these interactions are enthalpically driven. Moreover, this sign and this order of magnitude of ⌬H indicates that hydrogen bonds, CHinteractions, and Van der Waals' forces are the principal forces stabilizing the complex (49 -52). However, this conclusion does not extend beyond small soluble chito-oli-   (GlcNAc) 3 showed that both have a surface-exposed binding site, whereas the binding site of WGA (29) is completely solvent-buried at the interface of the WGA dimer. This more solvent-buried binding site is reflected in a 3-fold increased ⌬H for WGA, but it is also compensated by a more negative ⌬S, so that for WGA ⌬G does not differ from the ⌬G obtained for the other plant chitin-binding lectins (Fig. 2). In overall conclusion, the ITC data support a model where the binding site of AVR4 is solvent-exposed and AVR4 exclusively interacts with (Glc-NAc) 3 repeats. Tryptophan Fluorescence Quenching-Surface-exposed tryptophans are often involved in protein-carbohydrate interactions forming CH-interactions (52). We used Trp fluorescence quenching (41) to study the role of the two Trp residues in AVR4 (Trp 63 and Trp 71 ). The addition of GlcNAc or (GlcNAc) 2 (up to 50 mM) did not result in quenching or a blue shift of the Trp fluorescence (Table III). However, when longer chito-oligomers were added to AVR4, the Trp fluorescence was significantly quenched, which was accompanied by a small blue shift from 354 to 348 nm (Fig. 3). This blue shift indicates that one of the Trp residues becomes more solvent-buried upon complexation, which is most likely Trp 71 because the corresponding residue in Tachycitin is solvent-exposed (9). In contrast, Trp 63 would remain solvent-buried because Trp 63 is involved in a hydrophobic interaction in the core of the protein with Tyr 38 (again based on the structure of Tachycitin) (9). This hydrophobic interaction is strictly conserved based on the high degree of conservation of both aromatic residues in the CBM14 family.
Irrespective of the length of the ligand, Trp fluorescence quenching was always ϳ50% at full saturation of binding (F ϱ ) ( Table III). The fact that we observed no difference in quenching between (GlcNAc) 6 an the smaller ligands suggests that both Trp residues are distant from the protein-protein interface. Second, these data give no further indication for additional interactions between AVR4 and the ligand as the length of the ligand increases. These data corroborate, therefore, that AVR4 exclusively interacts with (GlcNAc) 3 repeats. In contrast, it is known for CBM18 lectins (e.g. Hevein) that additional interactions occur for longer ligands such as (GlcNAc) 4 and (GlcNAc) 5 (33).
The fluorescence quenching experiments were used as a second method (besides ITC) to determine K D values for the interaction between AVR4 and chito-oligomers (Table III). Estimates for the K D were obtained from both Scatchard (Fig. 3B) and Hill plot analyses (supplementary data). These K D values are in good agreement with the K D values obtained by ITC. The slopes obtained for the corresponding Hill plots approached unity for all chito-oligomers, including (GlcNAc) 6 . However, in the case of (GlcNAc) 6 the Scatchard plot was clearly curved (Fig. 3B), whereas smaller chito-oligomers showed a perfect linear regression (as expected for a single binding event). This confirms the presence of two AVR4-binding sites at (GlcNAc) 6 . Moreover, the curved Scatchard plot provides a second indication that positive cooperativity contributes to binding of AVR4 to chitin. A re-examination of the ITC data obtained for (GlcNAc) 6 using now a model with two dependent binding sites gave, however, no statistically significant improvement of the data fit. Therefore, we only present one K D for (GlcNAc) 6 , but this K D ITC is only an apparent value. Trp fluorescence quenching experiments have also been reported for peritrophin-44, a CBM14 lectin containing four ChBD repeats (8). In this case, Trp fluorescence quenching was ϳ16% at full saturation for (GlcNAc) 3 , whereas peritrophin-44 only contains one Trp residue in one of the four CBM14 repeats. Strikingly, similar experiments with UDA and WGA (both CBM18) resulted in enhanced fluorescence in the presence of the chito-oligomers, i.e. up by 27% for UDA (42) and ϳ36% for WGA (41). The increased fluorescence for CBM18 as compared with fluorescence quenching for CBM14 indicates that the Trp residues have a different topology in regard to the ligand in the two types of ChBDs.
Analytical Size Exclusion Chromatography-To exclude the possibility that nonspecific aggregation of AVR4 occurred under any of the tested conditions, we performed analytical size exclusion chromatography. AVR4 eluted from the column as a monomer at acidic to neutral pH (Table IV), whereas at pH 8.6 higher order complexes were observed (the estimated pI of AVR4 is 8.6). The formation of higher order complexes (dimer, trimer, etc.) proved to be reversible because the equilibrium shifted to monomer when the pH was decreased again. The monomer of AVR4 eluted at an apparent molecular mass of 10.3 kDa at pH 7.0 (Table IV), which is 8% higher than the mass determined by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (37). This supports that AVR4 behaves like a globular protein on the column. Subsequently, the column was equilibrated with either 35 mM (GlcNAc) 5 or 25 mM (GlcNAc) 6 . Under these conditions, the apparent molecular mass of AVR4 increased significantly. When the column was equilibrated with (GlcNAc) 5 , a protein complex eluted at an apparent molecular mass corresponding to the molecular mass of AVR4 plus one (GlcNAc) 5 molecule. However, in the case of (GlcNAc) 6, a protein complex eluted at an apparent molecular mass that was 60% higher than expected for AVR4 alone. Apparently, we observed an continuous equilibrium between one and two AVR4 molecules that bind to (GlcNAc) 6  more increased concentrations of the chito-oligomers were found to be impossible because the column pressure increased above the column operation conditions recommended by the manufacturer.
Mass Spectrometry-ESI-MS is increasingly used for the detection of noncovalent complexes over an extended range of K D values from 10 Ϫ6 to 10 Ϫ14 M (54, 55). Despite the K D values in the order of 10 Ϫ2 -10 Ϫ6 M for the chito-oligomers, we investigated whether we could specifically detect noncovalent complexes between AVR4 and chito-oligomers using ESI-MS. Care was taken to optimize the instrument settings, such as source temperature, cone potential, and the desolvation gas flow rate, all of which have been reported to influence the detection of noncovalent complexes (56,57). Instrument settings were optimized using the mass peak of the tetramer of yeast alcohol dehydrogenase in 10 mM ammonium acetate at pH 7.0. Subsequently, ESI-MS was performed on a sample of 20 M AVR4 in the presence of 20 M (GlcNAc) 6 (both dissolved in 10 mM ammonium acetate at pH 7) (Fig. 4). Under these conditions, three distinct entities were observed in the mass spectrum, i.e. AVR4, (GlcNAc) 6 , and a noncovalent complex between AVR4 and (GlcNAc) 6 with a 1:1 stoichiometry (with a charge state distribution ranging from ϩ3 to ϩ8 with the dominance of the ϩ6 charge state for AVR4).
A primary concern for the detection of complexes with ESI-MS is the specificity of the observed complex. First of all, the observed 1:1 stoichiometry for the complex in the gas phase is regarded as a good indicator for specificity (54). A surplus of AVR4 (100 M) gave a relative reduction in the intensity of the  3. Tryptophan fluorescence quenching of AVR4 in the presence of (GlcNAc) 5 and (GlcNAc) 6 . A, emission spectrum of AVR4 (3.6 M) without ligand (solidline) and in the presence of 77 M (GlcNAc) 6 (full saturation; dashed line; and the difference spectrum, dotted line). B, the corresponding Scatchard plots for (GlcNAc) 5 (DP ϭ 5) and (GlcNAc) 6 (DP ϭ 6) indicate that only (GlcNAc) 5 binds to AVR4 (i.e. linear regression with an intercept at y axis at ϭ 1.0), whereas two AVR4 molecules bind to (GlcNAc) 6 (DP ϭ 6). The second binding events shows positive cooperativity. Binding of smaller ligands (i.e. GlcNAc) 3 and (GlcNAc) 4 ) gave also a linear regression like for (GlcNAc) 5 (not shown). L f , concentration of free ligand; , fraction of occupied binding sites.  5 11.7 1.10 AVR4 ϩ (GlcNAc) 6 16.7 1.55 (GlcNAc) 5 b 2. 5 1.050 (GlcNAc) 6 2. mass peaks corresponding to the 1:1 complex (as expected). On the other hand, a five times surplus of the ligand (100 M), as tested for (GlcNAc) 5 and (GlcNAc) 6 , gave nonspecific aggregates that contained one AVR4 molecule and two, three, or even four chito-oligomers. Next, we performed a survey over the pH range 3.5-8.5 (with 10 mM ammonium acetate as buffer). Higher order "hybrid" assemblies were not observed over this pH range, but the mass peak corresponding to the 1:1 complex was best observed at neutral pH. This agreed with the ITC data, which already had shown that the interaction weakens toward acidic pH (Table II). Finally, when we compared the relative peak intensities for the different AVR4-GlcNAc complexes (all ϩ5 charge state), we noted that the relative peak intensity of the noncovalent complex displayed a positive correlation with increasing length of the chito-oligomer (Fig. 4B). No complex could be observed for GlcNAc or (GlcNAc) 2 . Control experiments under identical conditions with the proteins AVR9 of C. fulvum (3.3 kDa) and bovine ribonuclease A (13.5 kDa) showed that these proteins interacted in a none-specific manner with the chito-oligomers, i.e. the peak intensity of the complex was always less than 5%, and it appeared to be independent of the length of the chito-oligomer (Fig. 4B). These two proteins were chosen because they have a relatively small size, a basic pI, and no known affinities for carbohydrates. Altogether, the MS data appear to reflect our previous K D values, as obtained with ITC and fluorescence quenching, and therefore, the detected complexes would in fact be specific complexes. Our initial idea was, however, that we would be able to detect complexes consisting of two AVR4 molecules and one (GlcNAc) 6 molecule. None of our conditions (including variations of the instrument settings) resulted in mass peaks that would correspond with such a complex. A possible explanation could be the gas phase itself, which is known to affect the stability of complexes.
Nuclear Magnetic Resonance-The 1 H, 13 C, 15 N backbone and side chain resonances of AVR4 were assigned using common NMR protocols. In general, the backbone amides were well resolved and dispersed in the 15 N HSQC spectrum, indicative of a folded protein (37). Of the 72 expected amide cross-peaks in the 15 N HSQC, the peaks Cys 57 , Gly 68 , and Cys 72 could not be assigned, presumably because of unfavorable chemical exchange processes. Determination of the solution structure of AVR4 was impaired by the presence of 14 prolines and the overlap of their side chain 1 H and 13 C resonances. We obtained,  4. Detection of the AVR4/chito-oligomer complex using ESI mass spectrometry. A, ESI mass spectrum of AVR4 in the presence of (GlcNAc) 6 . The inset shows an enlargement of m/z 1850 -2150 Da showing two mass peaks that correspond to the noncomplexed AVR4 (A 5ϩ) and AVR4 bound to (GlcNAc) 6 (AH 5ϩ). B, the detected complexes are not an artifact caused by ESI-MS as the relative mass peak intensities obtained for the different AVR4-chito-oligomer complexes (black bars) correlate with the binding affinities (Table I). No complex is observed for GlcNAc and (GlcNAc) 2 . Similar measurements with AVR9 (white bars) and ribonuclease A (hatched bars) are shown as control experiments. The peak intensities were normalized using the mass peaks that corresponded to free protein with the same charge state. Both protein and ligand were used at a concentration of 20 M.
however, information about the secondary structure of AVR4. First, the NOE patterns in the 15 N NOE spectroscopy-HSQC spectrum clearly indicated an ␣-helix (58) for residues 14 -22. 13 CA chemical shift index analysis (59) confirmed the ␣-helical character of these residues (Fig. 5A). Notably, these residues form a sequence insertion in AVR4 that is connected to the core of the protein by an additional disulfide bond, Cus 21 -Cys 27 (10). Second, the 1 HA chemical shift of the residues 25-46 and 58 -80 follows closely the 1 HA shift of the corresponding residues in Tachycitin (Fig. 6). These two stretches of residues form the consensus of CBM14 motif (5) and are in the core of the protein fold of Tachycitin, i.e. the two anti-parallel ␤-sheets in Tachycitin excluding the loop regions. The 13 CA chemical shift index plot of AVR4 shows also two long stretches with ␤-sheet propensity, which overlap with the ␤-sheets in Tachycitin. Correspondingly, we noted strong d␣N(iϩ1) NOE contacts for residues in these two stretches indicative for ␤-sheet (data not shown). Six Cys residues are conserved in the CBM14 motif, and we showed recently that the corresponding Cys residues in AVR4 are indeed involved in a disulfide bond pattern similar to that found in Tachycitin (10). Based on these facts, we conclude that the protein fold of AVR4 is similar to the fold of Tachycitin with the exception that we found one additional ␣-helix that comprises residues 14 -22. NMR Studies of the AVR4-Chito-oligomer Complex-Residues in AVR4 that interact with chitin were identified from changes in chemical shifts of the 1 HN and 15 N resonances of AVR4 induced by adding chito-oligomers. When aliquots of (GlcNAc) 3 were added to AVR4, a ligand concentration-dependent change in chemical shift was noted for a set of 1 HN resonances without substantial line broadening (see supplementary data). The addition of GlcNAc or (GlcNAc) 2 did not induce such changes in the NMR spectrum. The continuous change in chemical shift is characteristic for fast exchange on the NMR time scale (60). Residues Asp 73 and Tyr 74 showed the largest concentration-dependent changes in chemical shift upon binding (Fig. 5B) and were used to derive binding constants using the NMR-derivative of the Scatchard plot. For Asp 73 and Tyr 74 , respectively, K D values of 5.3 mM and 5.4 mM (at 298 K) were obtained in close agreement with the ITC and fluorescence quenching data. Using the van't Hoff analysis (i.e. ϪR ln(1/K D ) versus 1/T), we were able to estimate ⌬H and ⌬S using the average K D values of Asp 73 and Tyr 74 (supplementary data). This gave a slope that corresponds with ⌬H o vH (van't Hoff enthalpy) ϭ Ϫ7.78 Kcal/mol yielding ⌬S ϭ Ϫ15.7 cal mol Ϫ1 K Ϫ1 . These numbers are in the same order as our ITC data. Nevertheless, these numbers should be regarded as qualitative rather than quantitative because the derivation of thermodynamic parameters from a van't Hoff plot assumes that ⌬H o is independent of the heat capacity ⌬C p (ϭ ␦⌬H/␦T). However, a small but negative contribution of ⌬C p to ⌬H o cal (calorimetric enthalpy) is generally observed for lectin-sugar interactions but is not included in ⌬H o vH (11,51,61). The titration of AVR4 with (GlcNAc) 6 3 . Conclusively, for the initial additions of ligand, we observed binding of AVR4 to (GlcNAc) 6 , but the exchange between bound and free AVR4 had changed to the intermediate regime as a consequence of the decreased dissociation constant. As a result, we were not able to estimate the K D for (GlcNAc) 6 from our NMR data. The fact that at increased concentrations of (GlcNAc) 6 the entire spectrum was affected by line broadening points to an increased rotational correlation time c . Because the c reflects the apparent size of AVR4, higher order complexes must have been present, most likely complexes with a 2:1 protein-ligand stoichiometry (as seen with the ITC and fluorescence experiments).
Residues Involved in Chitin Binding-Studies with CBM18 lectins have shown that both the 1 HN and HA resonances can be used as indicators for residues involved in ligand binding, i.e. the resonances important for binding show shifts exceeding 0.1 ppm at full saturation, whereas shifts Ͻ0.1 ppm are apparently caused by a reorientation of the aromatic side chains influencing other residues as well (30,31,33,34). In addition, large conformational changes were never found for these types of protein-lectin interactions. Performing similar experiments for AVR4, we found that the 1 HN resonances of Asn 64 , Asp 65 , Asn 66 , Asp 73 , and Tyr 74 experienced large shifts in the presence of (GlcNAc) 3 (Fig. 5B). These five residues are located near the predicted chitin-binding site of Tachycitin (Fig. 7), i.e. the second ␤-sheet that shows structural similarities with Hevein (9). Fig. 7B shows a ribbon structure of both Tachycitin and Hevein with the residues involved in binding shown in red. The first conclusion is that the residues involved in binding in CBM14 and CBM18 have only a limited overlap. Asn 64 and Asn 66 would align with Ser 19 (subsite ϩ1 in Hevein) and Trp 21 (subsite ϩ2) (Fig. 7A). Asn 64 is highly conserved in the CBM14 FIG. 8. Proposed binding model for the chitin-binding domains CBM14 and CBM18 (i.e. AVR4 and Hevein). Panel I, AVR4 interacts with a ligand with a DP of 3 or more, whereas Hevein already interacts with N-acetyl-D-glucosamine. Panel II, a second binding site becomes only available for AVR4 when the ligand is six sugar residues long, whereas for Hevein a second binding site is available at (GlcNAc) 5 . For AVR4 the second binding event is accompanied by positive cooperativity. In the case of Hevein several complexes with a 1:1 and a 1:2 stoichiometry were noted, which does not support positive cooperativity for Hevein. family, which adds to a role in chitin binding (90% similarity: Asn, Asp, and less often Ser). However, Asn 66 is not conserved in the CBM14 family. The residues Asp 73 and Tyr 74 , which experienced the largest shift upon binding of AVR4 to (Glc-NAc) 3 , are highly conserved in the CBM14 family (based on the 233 annotated sequences in the Pfam protein data base) but not in the CBM18 family (Fig. 7A). Based on the structural similarities between Tachycitin and Hevein, it was previously proposed that the chitin-binding site in Tachycitin would overlap with the binding site in Hevein. 2 However, Fig. 7B shows that Asp 73 and Tyr 74 are not situated near the putative binding site. Moreover, our data indicate a novel binding site on the folding scaffold shared between Tachycitin and Hevein. This binding site is solvent-exposed as suggested by the ITC data, but perhaps more interestingly these residues appear to form a stretch of residues at the surface of Tachycitin (Fig. 7B). This extended binding site could explain why AVR4 exclusively interacts with (GlcNAc) 3 repeats. In contrast, for Hevein a small binding pocket is seen in the form of subsite ϩ1, which seems to provide enough contacts to sustain an interaction with GlcNAc alone (Fig. 7B). Remarkably, Hemmi et al. (46) reported recently that the three-dimensional structure of the antifungal peptide scarabaecin from the coconut beetle Orycetes rhinoceros also shares a significant structural similarity with Hevein and Tachycitin. Again, this peptide has no overall sequence similarity with either one of the two ChBDs, but a structural comparison of the region of the putative chitin binding site indicates that binding of chitin by scarabaecin is likely to occur in a fashion similar to that of Hevein, i.e. all three residues that form subsite ϩ1 in Hevein are conserved in scarabaecin (i.e. Asn 25 , Phe 27 , and Phe 35 ) (46), whereas no residues appear to correspond with Asp 73 and Tyr 74 (Fig. 7A). Nevertheless, those results provide additional evidence for the idea of convergent evolution between the CBM14 and CBM18.
The role of Trp 71 in the binding site of AVR4 is more elusive. Trp 71 would structurally align with Trp 23 in Hevein, which is an important residue in subsite ϩ1 of the CBM18 lectins. Trp 71 appears not to be required for binding, because a large set of CBM14 members does not contain an aromatic residue at this position. Nevertheless, our NMR titration data showed effectively that the side chain of Trp 71 experiences a shift upon binding of (GlcNAc) 3 to AVR4 (supplementary data). Likewise, low concentrations of (GlcNAc) 6 caused significant line broadening of the side chain of Trp 71 (data not shown). Thus, Trp 71 is affected by the interaction but is not necessarily required for binding for all CBM14 members. The side chain of Trp 71 becomes solvent buried upon binding, confirming our initial conclusions based on the fluorescence quenching experiments. In addition, the NMR data confirm that Trp 63 is not directly involved in binding, because both the backbone amide and the side chain of Trp 63 experienced only a subtle effect in the presence of (GlcNAc) 3 and low concentrations of (GlcNAc) 6 (supplementary data).
Conclusion-Binding of AVR4 to chitin appears to be limited to an interaction with repeats of three GlcNAc residues. Our experiments did not indicate that additional interactions occur with GlcNAc residues situated outside this repeat. However, we detected positive allosteric interactions between AVR4 molecules that bind to (GlcNAc) 6 . Positive cooperativity has not been reported for any of the CBM18 lectins. This raises an interesting point. The "interlocking" process of the AVR4 molecules during binding suggests that binding of chitin by AVR4 will be very effective and that it tends to reach saturation of binding (Fig. 8). This would explain why AVR4 effectively protects the cell wall of the fungi T. viride and F. solani f.sp. phaseoli against anti-fungal activity by basic PR-3 chitinases (10). 2 Westerink et al. (12) reported recently that AVR4 binds to crude fungal components with a binding affinity in the order of nanomolar. This could indicate that the interlocking process would be even more effective than suspected on the basis of the data presented here. Thermodynamically, the affinity of AVR4 for the substrate will only be increased in the case of chitin as compared with the chito-oligomers because of the inherent reduced flexibility of chitin. In addition, a further decrease of ⌬H is expected when additional AVR4 molecules interact with chitin, because this will further reduce the solvent-exposed area of the bound AVR4 molecules. Conclusively, our data support a model where AVR4 effectively protects the chitin in the cell wall from degradation in favor of fungal growth and sustaining cell wall formation at the hyphal tip, potentially even in a hostile environment containing increased concentrations of plant chitinases caused by host responses.