NMR structure and dynamics of the C-terminal domain of R-type lectin from the earthworm Lumbricus terrestris

Authors

  • Hikaru Hemmi,

    Corresponding author
    • National Food Research Institute, National Agriculture and Food Research Organization (NARO), Tsukuba, Japan
    Search for more papers by this author
  • Atsushi Kuno,

    1. Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
    Search for more papers by this author
  • Jun Hirabayashi

    1. Research Center for Medical Glycoscience, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
    Current affiliation:
    1. Research Center for Stem Cell Engineering, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
    Search for more papers by this author

Correspondence

H. Hemmi, National Food Research Institute, National Agriculture and Food Research Organization (NARO), 2-1-12 Kannondai, Tsukuba, 305-8642 Ibaraki, Japan

Fax: +81 298 387996

Tel: +81 298 388033

E-mail: hemmi@affrc.go.jp

Abstract

The C-terminal domain (Ch; C-half) of the R-type earthworm 29-kDa lectin (EW29), isolated from the earthworm Lumbricus terrestris, has two sugar-binding sites, in subdomains α and γ, and the protein uses the two sugar-binding sites for its function as a single domain-type haemagglutinin. Our previous NMR titration experiments showed that the α sugar-binding site is a high-affinity site and the γ sugar-binding site is a low-affinity site. However, it remains unclear why the α sugar-binding site of EW29Ch binds to lactose much more strongly because the crystal structure of lactose-bound EW29Ch showed that the interaction between the α sugar-binding site and lactose was almost same as that between the γ sugar-binding site and lactose. In the present study, we have determined the NMR structure of EW29Ch in the sugar-free state and performed 15N relaxation experiments for EW29Ch in both the sugar-free state and the lactose-bound states. The conformation of EW29Ch in the sugar-free state was similar to that of EW29Ch in complex with lactose. Conformational changes upon binding of lactose were observed only for the α sugar-binding site. By contrast, the 15N relaxation experiments revealed a conformational exchange at the α sugar-binding site in the sugar-free state, which was suppressed in the lactose-bound state. The conformational exchange phenomenon observed for the α sugar-binding site was not observed for the γ sugar-binding site. Differences in the conformational change and the backbone dynamics between subdomains α and γ may be associated with the difference of the sugar-binding modes between the two sugar-binding sites.

Database

Structural data for the NMR structure of EW29Ch in the sugar-free state have been deposited in the Protein Data Bank database under accession number 2RST.

Abbreviations
Ch (C-half)

the C-terminal domain

DSS

4,4-dimethyl-4-silapentane-1-sulfate

EW29

earthworm 29-kDa lectin

HSQC

heteronuclear single quantum coherence

PDB

Protein Data Bank

RDC

residual dipolar coupling

SRC

Sia-recognition EW29Ch

Introduction

Sugar-binding proteins, known as lectins, exist ubiquitously in both animals and plants. In the lectin family, carbohydrate-recognition proteins containing short conserved triple-repeated QXW motifs form the R-type lectin family [1, 2]. The R-type lectin family is classified into two groups (i.e. the lectin group and the enzyme group) and the 3D structures of several proteins in each of the two groups of the R-type lectin family have been determined [3, 4]. These proteins possess the common β-trefoil fold structures displaying the characteristic pseudo-three-fold symmetry by the three subdomains designated α, β and γ, although their sugar binding differs according to their ligand specificities. Furthermore, the tandem repeat-type proteins in the R-type lectin family generally contain one sugar-binding site per domain, suggesting that truncated mutants comprising a single domain may have no haemagglutinating activity [1, 4, 5].

An R-type earthworm 29-kDa lectin (EW29), isolated from the earthworm Lumbricus terrestris, consists of two homologous domains (14 500 Da) (i.e. N- and C-terminal domains). The C-terminal domain (Ch; C-half) of EW29 (EW29Ch) bound to asialofetuin-agarose as strongly as the whole protein and retained its haemagglutinating activity [3]. The crystal structures of the complex between EW29Ch and lactose or N-acetylgalactosamine [Protein Data Bank (PDB) code: 2ZQN or 2ZQO] indicate that the overall structure of EW29Ch resembles β-trefoil fold, and that the protein has two sugar-binding sites in the subdomains α and γ [4, 6]. Furthermore, our previous NMR titration experiments showed that the α sugar-binding site of EW29Ch (Kd = 0.01–0.07 mm for lactose) has a much tighter sugar-binding mode than the γ sugar-binding site (Kd = 2.66 mm for lactose) [7]. However, the structural basis of the lectin-sugar interactions of EW29 does not explain why the α sugar-binding site of EW29Ch binds to lactose much more strongly because the crystal structure of lactose-bound EW29Ch shows that the interaction between the α sugar-binding site and lactose is almost same as that between the γ sugar-binding site and lactose [4]. Therefore, additional studies are required to analyze the interaction between the protein and lactose in the solution state and to determine the structural basis of the carbohydrate cross-linking properties of the lectin.

In the present study, we have determined the refined NMR structure of EW29Ch in the sugar-free state using residual dipolar coupling (RDC) data, and we have performed a comparison of the refined sugar-free NMR structure with the crystal structure of the complex between EW29Ch and lactose. Furthermore, we have examined the dynamics of EW29Ch in the sugar-free state or in the sugar-bound state by 15N relaxation methods. We report that differences in the conformational changes and the relaxation data between the two states were observed, particularly for the α sugar-binding site, even though the sugar-free NMR structure of EW29Ch was similar to the crystal structure of lactose-liganded EW29Ch. On the basis of the structural data, we discuss the relationship between the motions of the backbone of EW29Ch and the sugar-binding mode.

Results

High-resolution NMR structure of EW29Ch in the sugar-free state and its structural comparison with lactose-liganded EW29Ch in the crystal state

First, we determined the solution structure of EW29Ch in the sugar-free state by NMR and compared the NMR structure with the crystal structure of EW29Ch in complex with lactose to determine the conformational changes upon the binding of lactose to EW29Ch.

To obtain the high-resolution NMR structure of EW29Ch in the sugar-free state, we refined the NMR structure of the protein calculated with distance and dihedral angle restraints using residual 1DNH couplings. A well-defined ensemble of the final 20 structures was obtained (Fig. 1A and Table 1) with an rmsd from the mean structure for backbone atoms (residues 5–130) of 0.35 Å and for all heavy atoms of 0.80 Å. The high-resolution NMR structure of EW29Ch in the sugar-free state is characterized by a β-trefoil fold, as expected from the crystal structure of lactose-bound EW29Ch (Fig. 1B). The similarity of the lowest total-energy NMR structure of EW29Ch with the crystal structure of EW29Ch is reflected by a pairwise rmsd of 1.10 Å for the backbone atoms of most of the residues, excluding the first four amino acids in the N-terminus and the last two amino acids in the C-terminus (Fig. 1C). Specifically, a large part of the secondary structure regions (residues 6–10, 16–19, 28–33, 45–48, 54–56, 61–64, 71–74, 81–87, 90–93, 96–105, 114–118 and 123–125), except for the three 310-helices (residues 20–22 and 40–44) and the C-terminus β-strand (residues 127–131), of the NMR structure of EW29Ch were very similar to the corresponding region of the crystal structure of EW29Ch, with a pairwise rmsd of 0.74 Å.

Table 1. Statistics for 20 NMR structures of EW29Ch in the sugar-free state
 Number of restraints
  1. a

    procheck-nmr [46] was used for the Ramachandran plot analysis.

Total distance restraints1943
Intraresidue351
Sequential664
Medium (1 < |i − j| < 5)178
Long (|i − j| ≥ 5)634
Hydrogen bond58 × 2
Total dihedral angle restraints148
φ64
ψ59
χ125
Residual dipole couplings, 1DNH109
rmsd from experimental restraints
NOE distance restraints (Å)0.0285 ± 0.0006
Dihedral angle restraints (°)0.3412 ± 0.0284
1DNH restraints (Hz)0.554 ± 0.0124
rmsd from ideal covalent geometry
Bonds (Å)0.0022 ± 0.00004
Angles (°)0.3799 ± 0.0054
Impropers0.2789 ± 0.0098
Ramachandran plot analysis (%) (residues 5–130)a
Most favoured regions82.4
Additional allowed regions16.2
Generously allowed regions1.0
Disallowed regions0.4
rmsd relative to the mean structure (Å)
Backbone atoms (residues 5–130)0.35 ± 0.07
Heavy atoms (residues 5–130)0.80 ± 0.08
Figure 1.

High-resolution NMR structure of EW29Ch in the sugar-free state. (A) Superposition of the 20 best structures of EW29Ch. (B) Ribbon diagram of the lowest-energy structure of EW29Ch. (C) Schematic representation of the lowest-energy NMR structure of EW29Ch (green) overlaid with the crystal structure of the lactose-liganded EW29Ch (red). Key residues for sugar binding are labelled.

The similarity of the backbone conformations of the NMR structure and the crystal structure of EW29Ch was also confirmed by the Q-factor of 28% between 109 measured residual 1DNH couplings of EW29Ch and those calculated from the crystal structure (Fig. S1) because the Q-factor is very sensitive to small changes in structure, and values of this factor that are smaller than 30% still indicate a good agreement between the calculated coordinates (the RDC calculated from the coordinates) and the real structure (the measured RDC) [8]. Although the refined NMR structure of EW29Ch in the sugar-free state and the crystal structure of EW29Ch in complex with lactose are closely similar, there are still significant differences between them. Conformational differences were observed for some loop regions (residues 36–39, 57–60 and 76–79) with a local rmsd > 1.5 Å. In the [1H,15N] heteronuclear single quantum coherence (HSQC) spectrum of EW29Ch in the sugar-free state, some NMR signals (Asn59, Asp60 and Asp76) in the loop regions were missing. Furthermore, the 15N{1H}-NOE value of Gly38 in the sugar-free state is < 0.70. Thus, the conformational differences of the loop regions between the two structures may be a result of the flexibility of the loop regions.

By contrast, the local rmsd for the main chain atoms of a large portion of the sugar-binding key residues (Asp18, Ile31, Trp33, Asp102, Lys105, Cys115, Trp117, His120 and Asn124) between the two proteins were < 1.0 Å, except for some residues identified from the EW29Ch-lactose complex structure (Gly21, Gln22, Lys36 and Asn43) (Fig. 2A,B). These results indicate that the main chain conformation of the key sugar-binding residues (Asp18, Ile31 and Trp33) in the α sugar-binding site and that of all of the key sugar-binding residues in the γ sugar-binding site of the NMR structure of EW29Ch in the sugar-free state and the crystal structure of EW29Ch in complex with lactose are similar. The structural difference in the main chain of the key residues, Gly21 and Gln22, of the NMR structures and the crystal structure of EW29Ch may be a result of the ill-defined region from Gly21 to Pro24 among the 20 NMR structures with an average local rmsd of the main chain atoms > 1.0 Å because the amide resonances of Gly21 and Ans23 were not observed in the sugar-free state [7]. The main chain atoms of the key residues, Lys36 and Asn43, are well-defined among the 20 NMR structures, with an average local rmsd of 0.58 and 0.27 Å, respectively. Residue Lys36 is located in the loop region from Lys36 to Pro39, as described above. Residue Asn43, as well as Gly21 and Gln22, is located in 310-helix regions. The three residues showed high R2 values in the sugar-free state and their high R2 values were suppressed in the lactose-bound state, as described below. Thus, the results suggest that the main chain of the γ sugar-binding site of EW29Ch in the sugar-free state does not undergo major conformational changes upon binding of lactose to EW29Ch, although the conformation of the main chain of the α sugar-binding site partly changes by binding with lactose.

Figure 2.

Interactions between the two sugar-binding sites (α and γ) of EW29Ch and lactose. A list is shown of possible interactions at the α binding site (A) and the γ binding site (B) from the crystal structure of EW29Ch in complex with lactose (PDB code: 2ZQN). These cartoons were generated by moe, 2011.10 (CCG Inc., Montreal, QC, Canada). A comparison of the structures of the side-chains of the key residues for the α binding site (C) and the γ binding site (D) of the 20 NMR structures (green) with those of the crystal structure of EW29Ch in complex with lactose (PDB code: 2ZQN) (red) is illustrated using lines. The bound lactose in the complex structure with EW29Ch is shown as a ball-and-stick.

We also compared the side-chains of the key residues in the sugar-binding sites (α or γ) of EW29Ch in the sugar-free state with the crystal structure of EW29Ch in complex with lactose (Fig. 2C,D). In the crystal structure of lactose-bound EW29Ch, the lactose-binding manner of the α sugar-binding site was similar to that in the γ sugar-binding site [4]. The Asp18 Oδ2, Asn43 Nδ2 and Lys36 Nζ atoms in the α sugar-binding site and the Asp102 Oδ2, Asn124 Nδ2 and His120 Nε2 in the γ sugar-binding site form hydrogen bonds to the O3 atom of the galactose residue of lactose. The Lys36 Nζ atom also interacts with the O2 atom of the galactose moiety. Furthermore, Asp18 Oδ1 and Gly49 main chain N atoms in the α sugar-binding site and Asp102 Oδ1 and Lys105 main chain N atoms in the γ sugar-binding site form hydrogen bonds to the O4 atom of the galactose residue of lactose. Other important interactions are made by Trp33 in the α sugar-binding site and Trp117 in the γ sugar-binding site because noncovalent interactions mediated by aromatic rings of protein are pivotal to sugar binding [9]. The indole plane of each tryptophan residue makes partial stacking interactions with C3, C4, C5 and C6 atoms of the galactose moiety [4]. In addition to those interactions, van der Waals interactions also occur between the sugar and protein molecule.

In the 20 NMR structures in the sugar-free state, the side-chain atoms of Asp18 and Asp102 were well-defined, with an average local rmsd of 0.95 and 0.75 Å, respectively. The distances between the side-chain atoms of these residues in the 20 NMR structures and galactose residue of the lactose are slightly longer than the distance between the corresponding residues in the crystal structure of lactose-bound EW29Ch and the galactose residue. The side-chain atoms of Ile31 were also well-defined, with an average local rmsd of 0.25 Å, and the positions of the side-chains of Ile31 in NMR structures are also slightly farther from the galactose residue than those of the corresponding residue in the crystal structure of lactose-bound EW29Ch. Thus, these results showed that the side-chain atoms of Asp18, Ile31 and Asp102 in the sugar-free state moved to a suitable position to bind with lactose.

The side-chain atoms of Trp33 and Trp117 are well-defined or less well-defined, with an average local rmsd of 0.58 and 1.23 Å, respectively. The indole planes of both tryptophan residues in the NMR structures are slightly farther away from the galactose residue compared to that of the corresponding residues in the crystal structure of lactose-bound EW29Ch (Fig. 2C,D). Thus, these results showed that both tryptophan residues of EW29Ch in the sugar-free state also moved to a suitable position to bind with lactose.

The side-chain atoms of Cys115, His120 and Asn124 in the 20 NMR structures are well-defined or somewhat less well-defined, with an average local rmsd of 0.83, 1.12 and 1.01 Å, respectively, and the side-chains of these residues in some of the 20 NMR structures are located spatially near the side-chains of the corresponding residues in the crystal structure of lactose-bound EW29Ch. The side-chain atoms of Asn43 in the 20 NMR structures are poorly defined and the side-chains of Gln22, Lys36 and Lys105 in the 20 NMR structures are disordered, with an average local rmsd of 1.47, 3.61, 2.39 and 2.07 Å, respectively. The side-chains of these residues in some of the 20 NMR structures are also located spatially near the side-chains of the corresponding residues in the crystal structure of lactose-bound EW29Ch. Thus, the side-chains of these residues may be arranged at suitable positions by binding with lactose. In a previous study [7], we showed that both sugar-binding sites of EW29Ch interact only with the galactose residue of the lactose molecule, even though, in the crystal structure of lactose-bound EW29Ch, the glucose residue of the lactose molecule interacted with the side-chain of Lys105 (Fig. 2B). We suggested that the interaction between the glucose residue and the side-chain of Lys105 might be an artefact caused by the crystallization of lactose-bound EW29Ch. In practice, the side-chains of Lys105 among the 20 NMR structures were disordered, indicating that the side-chains of Lys105 are flexible in the sugar-free state.

Solution dynamics of EW29Ch in the sugar-free state and in the lactose-bound state by 15N relaxation experiments

Figure 3 shows the superposition of the 1H-15N HSQC spectra of 15N-labelled EW29Ch in the sugar-free state with those in the lactose-bound state ([EW29Ch]/[lactose] = 1/0.5, 1/1 and 1/8) and in the fully lactose-bound state ([EW29Ch]/[lactose] = 1/80). In a previous study [7], we found four patterns of chemical-shift change for the amide resonances of EW29Ch upon the addition of lactose: (a) amide resonances showing slow chemical exchange; (b) fast chemical exchange; (c) no or slight chemical-shift change; and (d) four amide resonances (Gly21, Asn23, Asp60 and Asp107) not observed in the sugar-free state that appeared in the lactose-bound state (Fig. 3). The amide resonance of Asp107 was still very weak in the fully lactose-bound state. There were two amide resonances (Asn59 and Asp76) that were not observed in either the sugar-free state or the lactose-bound state.

Figure 3.

Chemical-shift changes in NMR titration of EW29Ch with lactose. Five 1H-15N HSQC spectra of 15N-labelled EW29Ch in the presence of protein/lactose molar ratios of 0 (red), 0.5 (green), 1.0 (purple), 8.0 (orange) and 80.0 (cyan) are overlaid. Overlay spectra show that free and bound forms of some residues in the protein are in slow exchange on the NMR timescale and those of other residues are in fast exchange. Representative resonance signals for slow and fast exchanges are labelled. Four resonance signals (Gly21, Asn23, Asp60 and Asp107) disappearing in the sugar-free state and appearing in the lactose-bound state are also labelled.

To characterize the dynamics behaviour of EW29Ch upon the binding of lactose, 15N relaxation measurements were performed in the sugar-free state or in lactose-bound states at [EW29Ch]/[lactose] = 1/1, 1/8 and 1/80 (Fig. 4; see also Fig. S2). The lactose occupancy levels of the two sugar-binding sites of EW29Ch at each of the [EW29Ch]/[lactose] molar ratios used for the 15N relaxation measurements were estimated by the Kd values measured previously [7] (Table S1). The average R1 values were 1.52 s−1 for the sugar-free state and 1.50, 1.46 and 1.44 s−1 for the lactose-bound states at [EW29Ch]/[lactose] = 1/1, 1/8 and 1/80, respectively. Decreased average R1 values upon the addition of lactose are assumed to be a result of the effect of viscosity in solution. However, the patterns of R1 values for each residue were similar among the sugar-free state and the three lactose-bound states (Fig. 4A). The average 15N{1H}-NOE values (∼0.81 for the sugar-free state and the three sugar-bound states) and the pattern of 15N{1H}-NOE values for each residue were similar among the sugar-free state and the three lactose-bound states (Fig. 4C). The average R2 values were also similar among the sugar-free state and the three lactose-bound states; the average R2 values were 12.11 s−1 for the sugar-free state and 11.94, 11.91 and 12.11 s−1 for the lactose-bound states at [EW29Ch]/[lactose] = 1/1, 1/8, and 1/80, respectively (Fig. 4B). However, the patterns of the R2 values for each residue were different between the sugar-free state and the sugar-bound states (Figs 4B and 5; see also Figs S2, S3 and S4). Five residues (Glu20, Gln22, Gly27, Asn43 and Gln44) in subdomain α and two residues (Ser56 and Lys57) near subdomain α exhibited significantly higher R2 values in the sugar-free state compared to the same residues in the lactose-bound states. The R2 values of six residues (Glu20, Gln22, Gly27, Asn43, Gln44 and Ser56) in the lactose-bound state were similar to the average R2 values in the sugar-free state or in the lactose-bound states. The R2 values of Lys57 in the three lactose-bound states (14.91–16.85 s−1) were higher than the average R2 values but were near the average R2 values compared to the R2 value of Lys57 in the sugar-free state (26.17 s−1) (Fig. 4B).

Figure 4.

Plots of the relaxation times (A) R1(15N), (B) R2(15N) and (C) 15N{1H}-NOE for the backbone of EW29Ch in the sugar-free state (black) and in the lactose-bound states at [EW29Ch]/[lactose] = 1/1 (blue), 1/8 (red) and 1/80 (purple). The secondary structures are indicated above the plots.

Figure 5.

Mapping of the transverse relaxation rate (R2 s−1) differences (ΔR2) between R2 in the sugar-free state (R2-free s−1) and that in the lactose-bound state (R2-bound s−1) at [EW29Ch]/[lactose] = 1/1 (A), 1/8 (B) or 1/80 (C) (ΔR2 = R2-free − R2-bound) on a ribbon diagram of the lowest-energy structure of EW29Ch in the sugar-free state determined by NMR. The spheres represent the 15N atom of the main chain of each residue in the protein. Residues showing ΔR2 > 5.0 are indicated in green and those showing 1.5 ≤ ΔR2 ≤ 5.0 are indicated in light green. Residues showing ΔR2 < −5.0 are indicated in red and those showing −5.0 ≤ ΔR2 ≤ −1.5 are indicated in light red. Residues showing R2-free > 15 s−1 and R2-bound > 15 s−1 or showing R2-free > 15 s−1 and very broad signals in the lactose-bound state are indicated in blue. The NMR signals of the [1H,15N]-HSQC spectrum of residues in the sugar-free state that were missing or very broad are indicated in yellow. Others are shown in grey. Residues shown in red, light red, green, light green, blue and yellow are labelled.

The global tumbling parameters of EW29Ch in the sugar-free state and in the lactose-bound states were estimated from the R2/R1 values to perform the model-free analysis. After the data had been filtered to exclude residues showing a high degree of internal mobility or 15N line broadening [10], a fit of the averaged R2/R1 values yielded a correlation time (τC) of 7.67 ns for the sugar-free state and 7.89, 8.10 and 8.23 ns for the lactose-bound states at [EW29Ch]/[lactose] = 1/1, 1/8, and 1/80, respectively. The increase in correlation time τC upon the addition of lactose could be caused by the effect of viscosity in solution [11]. The model-free analysis was performed using tensor2 [12] under conditions of anisotropic tumbling. On the basis of the model selection criteria implemented in tensor2, the averaged order parameters (S2) were 0.90 for the sugar-free state and 0.91, 0.91 and 0.91 for the lactose-bound states at [EW29Ch]/[lactose] = 1/1, 1/8, and 1/80, respectively. The patterns of the S2 and conformational exchange broadening factors (Rex) values for each residue were very similar to those of the 15N{1H}-NOE and R2 values, respectively (Fig. S5).

Effect of lactose binding on backbone dynamics

In a previous study [7], we reported that the resonance signals of residues Gly21 and Asn23 in subdomain α were missing in the sugar-free state and appeared in the sugar-bound state. These results suggested that certain key residues for sugar-binding (Glu20, Gln22, Asn43 and Gln44) in subdomain α identified from the crystal structure of EW29Ch–lactose complex, the loop region (residues 23–27) in subdomain α, and two residues (Ser56 and Lys57) located near subdomain α exhibit a conformational exchange in the sugar-free state but do not exhibit a conformational exchange as a result of sugar-binding in the lactose-bound states. Three residues (Gln50, Ile70 and Glu71) in subdomain β and two residues (Ser106 and Lys108) in subdomain γ exhibited high R2 values in both sugar-free and lactose-bound states. The R2 values of Ile70 and Glu71 in the lactose-bound states at [EW29Ch]/[lactose] = 1/8 and [EW29Ch]/[lactose] = 1/80 could not be determined because the resonance signals of the two residues were very broad, which indicates that the R2 values of Ile70 and Glu71 in the two lactose-bound states are also high. The R2 values of the three residues (Gln50, Ile70 and Glu71) in subdomain β upon the addition of lactose remained high because there is no sugar-binding site in subdomain β of EW29Ch. The R2 values upon the addition of lactose of the three residues, Lys105, Ser106 and Lys108 in the γ sugar-binding site of EW29Ch were also high. The 15N{1H}-NOE value of Lys105 and Ser106 could not be determined in the sugar-free state or in the lactose-bound state at [EW29Ch]/[lactose] = 1/1 because the resonance signals of the two residues were very weak. The resonance signal of Lys105 in the lactose-bound state at [EW29Ch]/[lactose] = 1/8 was partly overlapped. Thus, S2 and Rex for Lys105 in the sugar-free state and in the two lactose-bound states at [EW29Ch]/[lactose] = 1/1 and 1/8 and for Ser106 in the sugar-free state and in the lactose-bound state at [EW29Ch]/[lactose] = 1/1 could not be determined either. Furthermore, the resonance signal of Asp107 was missing in both the sugar-free state and the two lactose-bound states at [EW29Ch]/[lactose] = 1/1 and 1/8. These results suggest that the region from Lys105 to Lys108, a part of the large loop region (residues 104–113) in subdomain γ, may be flexible in both the sugar-free state and the lactose-bound state.

It is worth noting that a novel sialic acid-binding lectin was created from EW29Ch by natural evolution-mimicry (designated Sia-recognition EW29Ch; SRC), and a significant conformational change was observed in a large loop region of the γ subdomain of the SRC, including the region corresponding to EW29Ch (residues 105–108), even though the overall structure of SRC was identical to that of EW29Ch [6]. SRC bears a single-binding site formed by subdomain α and the extended loop of subdomain γ, which has dual binding ability to α2-6Sia and, to a lesser extent, galactose (Gal).

Finally, eleven residues (Leu17, Ile30, Asp49, Gly52, Gln69, Thr72, Ala112, Ile114, Cys115, Ala116 and Gly122) displayed a unique pattern of R2 values because the R2 values of those residues increased upon the addition of lactose. In particular, the R2 values of three residues, Leu17 in subdomain α and Ile114 and Cys115 in subdomain γ, were remarkably increased upon the addition of lactose, from 13.61 s−1 in the sugar-free state to 28.21 s−1 in the lactose-bound state at [EW29Ch]/[lactose] = 1/80 (Leu17), from 11.32 s−1 to 17.04 s−1 (Ile114), and from 11.34 s−1 to 19.57 s−1 (Cys115), even though the R1 values of the three residues were similar between the sugar-free state and the lactose-bound state at [EW29Ch]/[lactose] = 1/80 and the 15N{1H}-NOE values of the three residues were > 0.80 in both the sugar-free state and the lactose-bound state (Fig. 4). We cannot explain the cause of the increased R2 values at present. The residues Leu17, Ile114 and Cys115 are located on β-strands that include the key residues (Asp18 and Ile19: β2, Trp117: β11) for sugar binding. Thus, the increased R2 values upon addition of lactose may be related to sugar binding. Furthermore, most of the residues in which R2 values increased upon addition of lactose are located at the centre of a β-trefoil fold of EW29Ch (Fig. 5). Unlike other R-type lectins, EW29Ch contains no disulfide bridges, and the cylindrical hydrophobic core located at the centre of the β-trefoil fold of EW29Ch effectively acts as a stabilizer of the peptide fold. Thus, the increased R2 values upon addition of lactose may be also related to the stability of the peptide fold.

Discussion

The R-type lectin family exists ubiquitously in nature and mainly binds to the galactose unit of sugar chains. The R-type lectins possess common β-trefoil fold structures, although their sugar-binding affinities, as well as their ligand specificities, differ significantly. For example, the Kd values determined for two sugar-binding sites of the ricin B-chain and EW29Ch differ [7, 13-23], even though the sugar-binding properties of the two relevant sugar-binding sites of these R-type lectins are quite similar to each other in their crystal structures [1, 4]. To clarify the differences of the sugar-binding properties between the two sugar-binding sites of the same molecule (EW29Ch) from the viewpoint of structural biology, we performed high-resolution NMR structure determination using RDC data, and examined the backbone dynamics by 15N relaxation experiments. Comparison of the NMR structure of sugar-free EW29Ch with the crystal structure of EW29Ch in complex with lactose revealed that the NMR structure of sugar-free EW29Ch was closely similar to the crystal structure of lactose-bound EW29Ch, except for some loop regions. Furthermore, we found that backbone conformational changes upon addition of lactose were observed for the α sugar-binding site, whereas these were not observed for the γ sugar-binding site. Thus, differences in the conformational changes between subdomains α and γ may be associated with the difference of sugar binding modes between the two sugar-binding sites. By contrast, we also determined that there are differences in the backbone dynamics between the sugar-free state and the lactose-bound state in the α sugar-binding site of EW29Ch.

Chemical exchange phenomena in NMR spectra can reveal protein motions that are associated with biological functions on a micro- to millisecond timescale, including catalysis, ligand binding and protein folding. Chemical exchange processes include conformational exchange, side-chain reorientation and other types of internal motion, as well as real proton exchange [24, 25]. Conformational exchange on the micro- to millisecond timescale is frequently identified in intermolecular interfaces [26]. These motions may reflect the ability to sample ensembles of conformers, thus allowing their adaptation to multiple binding sites.

Recently, the NMR structure determination of a human RegIV mutant (hRegIV-P91S) and the backbone dynamics for free and mannan-bound hRegIV-P91S by 15N relaxation parameters and simplified Carr–Purcell–Meiboom–Gill relaxation dispersion experiments were performed to determine the carbohydrate recognition mechanism of the hRegIV protein. hRegIV is known to contain a sequence motif homologous to the calcium-dependent (C-type) lectin-like domain, whereas it can bind to polysaccharides, such as mannan and heparin, even in the absence of calcium ions [27]. The structural properties and carbohydrate-binding ability of hRegIV-P91S are almost identical to those of the wild-type protein. The NMR structure of hRegIV-P91S adopts a typical fold of C-type lectin, which contains two α-helices and eight β-strands consisting of two antiparallel β-sheets. However, on the basis of chemical shift perturbation of amide resonances, the protein was shown to bind to mannan with two calcium-independent sites (sites I and II). Furthermore, the backbone dynamics for free and mannan-bound hRegIV-P91S by NMR indicated that the fast time-scale motions of site II residues were decreased in the complex form, whereas both the fast and slow time-scale motions for site I residues were restricted with mannan binding: mannan binding significantly reduced the micro- to millisecond backbone motions of several residues in site I, including two loop regions. The micro- to millisecond backbone motions identified in site I residues implied that this site may adapt different conformers for different carbohydrate recognitions.

The backbone dynamics of the lectin domain of malectin, a novel endoplasmic reticulum-resident lectin, in free and maltose-bound forms, was investigated by 15N-relaxation experiments because line broadening, as observed for the ligand-binding loops of the malectin domain, was often caused by motion on the micro- or millisecond timescale, and the aromatic ring resonances of Tyr67, Tyr89, Tyr116 and Phe117 residues in the binding loops were also exchange-broadened. The entire binding site and the surrounding residues on the four loops in the malectin domain showed a high mobility in the free form that freezes upon ligand binding. These findings suggest that, among the many possible interconverting loop conformations, each ligand selects for the one that best accommodates the respective carbohydrate [28].

In the present study, we found that conformational exchange existed within and near the α sugar-binding site, including the binding loop in EW29Ch in the sugar-free state, and that the conformational exchange was largely reduced in the lactose-bound state by 15N relaxation experiments. These results indicated that lactose binding to subdomain α suppresses the conformational exchange. By contrast, this phenomenon was not observed in subdomain γ. For all sugars used in our previous study (lactose, melibiose, d-galactose, methyl α-d-galactopyranoside and β-d-galactopyranoside), the sugar-binding ability of the α sugar-binding site of EW29Ch was ∼10- to 100-fold higher than that of the γ sugar-binding site [7]. Thus, we hypothesized that the α sugar-binding site of EW29Ch may adapt more suitable conformers for the respective sugars by micro- to millisecond backbone motions occurring the conformational exchange. Therefore, the difference of the backbone dynamics between subdomains α and γ, as well as the difference of the conformational change between them, may be associated with the difference of sugar binding modes between the α sugar-binding site (high-affinity site) and the γ sugar-binding site (low-affinity site) in the same molecule.

To support the hypothesis, we need to address the balance of enthalpic/entropic contributions for sugar binding modes of the α sugar-binding site or the γ sugar-binding site. In both sugar-binding sites, the side-chains of the key residues of EW29Ch moved to or arranged to the suitable positions to bind with lactose, indicating that the enthalpic contributions resulting from the formation of favourable protein-lactose hydrogen-bonding and van der Waals interactions would compensate for the loss of conformational entropy of the protein side-chains. By contrast, the balance of the enthalpic/entropic contributions for sugar-binding may differ between the main chains of the key residues in the α sugar-binding site and those in the γ sugar-binding site, even though the crystal structure of lactose-bound EW29Ch showed that the interaction between the α sugar-binding site and lactose was almost same as that between the γ sugar-binding site and lactose. The balance of the enthalpic/entropic contributions for sugar-binding may change for the main chain of the key residues in the α sugar-binding site because the conformation of the main chain of the α sugar-binding site partly changed by binding with lactose, and the change in the conformational exchange induced by lactose-binding was also observed, whereas the balance of the enthalpic/entropic contributions for sugar-binding may not change for the main chain of the key residues in the γ sugar-binding site because the structural changes induced by lactose-binding are very subtle, and the change in the conformational exchange induced by lactose-binding was not observed. We anticipate that the relationship between the difference of sugar binding modes and the changes of the balance of the enthalpic/entropic contributions for sugar-binding will be addressed in more detail in further studies employing calorimetric analysis for mutants with a single sugar-binding site.

As described above, a novel sialic acid-binding lectin (SRC) was created from EW29Ch by natural evolution-mimicry. The overall structure of SRC was identical to that of EW29Ch, and a significant conformational change was observed in a large loop region of the γ subdomain of SRC, including the region corresponding to EW29Ch (residues 105–108) [6]. SRC bears a single-binding site formed by subdomain α and the extended loop of subdomain γ, which has dual binding ability to α2-6Sia and, to a lesser extent, Gal. It is tempting to speculate that the single binding site of SRC has dual binding ability to α2-6Sia and, to a lesser extent, Gal because the α binding site of EW29Ch may adapt different conformers for different types of carbohydrate recognition by conformational exchange on a micro- to millisecond timescale, and the γ binding site has a flexible loop. Thus, characterization of the sugar-binding mode of the sugar-binding site by backbone dynamics of the lectin using NMR in combination with the structural basis of lectin-sugar interactions should be useful in the design of other lectins with respective carbohydrate recognition that form critical biomarkers.

Experimental procedures

Expression and purification of EW29Ch

The expression and purification of the 15N-labelled or 13C, 15N-labelled EW29Ch were performed using 15N-labelled or doubly-labelled C.H.L. medium (Chlorella Co., Tokyo, Japan) as described previously [29]. The final product contains the full-length 131 amino acid EW29Ch sequence of L. terrestris (residues Lys130−Glu260 in EW29) [3], with an N-terminal methionine residue (total length of 132 amino acids). Residues are numbered from the N-terminal methionine residue (Met1-Lys2-Pro3 …). Thus, the residue Lys2 of EW29Ch corresponds to the residue Lys130 of EW29.

For the NMR studies in the sugar-free state, purified EW29Ch was dissolved in 50 mm potassium phosphate buffer (pH 6.1) and a protease inhibitor cocktail (Sigma Chemical Co., St Louis, MO, USA) in either a 90% H2O/10% 2H2O mixture or 99.96% 2H2O. The final concentration of the protein was 0.9 mm. For the amide proton exchange studies, protein was lyophilized and dissolved in D2O. For the purpose of collecting residual dipolar coupling restraints, non-ionic liquid crystalline medium was used, consisting of 50 mm potassium phosphate buffer (pH 6.1), 10% 2H2O and a 5% C12E5PEG [n-dodecyl penta(ethylene glycol)]/hexanol mixture with a surfactant-to-alcohol ratio of 0.96 [30]. In addition, five mixtures (1 : 0.5, 1 : 1, 1 : 8, or 1 : 80) of 15N-labelled EW29Ch and lactose were also prepared for NMR titration studies or 15N relaxation studies as a lactose-bound state.

NMR spectroscopy and structural determination

NMR spectra were acquired at 25 °C on Bruker DRX 600 and Avance 800 NMR spectrometers (Bruker Instruments, Inc., Bellerica, MA, USA). All 1H dimensions were referenced to internal 4,4-dimethyl-4-silapentane-1-sulfate (DSS), and 13C and 15N were indirectly referenced to DSS [31]. All multidimensional NMR spectra were acquired in a phase-sensitive mode employing the states–time proportional phase increment method [32] or the echo–anti-echo mode [33]. Shifted sine-bell window functions were applied to the NMR data before zero-filling and Fourier transformation. NMR data were processed using felix2000 (Accelrys, San Diego, CA, USA) or nmrpipe [34], and analyzed using sparky (T. D. Goddard and D. G. Kneller; University of California, San Francisco, CA, USA). 1H, 13C and 15N assignments were obtained from standard multidimensional NMR methods described previously [29].

Distance restraints were constructed from intensities of NOE cross peaks in 2D NOESY, 3D 15N-edited NOESY and 13C-edited NOESY spectra with mixing times of 150 and 200 ms, respectively. Pseudo-atom corrections were made for nonstereospecifically assigned methylene and methyl resonance [35]. An additional 0.5 Å was added to the upper bounds for methyl protons [36]. Backbone Φ and ψ dihedral angle restraints were evaluated from 3JHNα values obtained from an HNHA experiment and the talos+ [37]. χ1 angles were estimated from HNHB and HN(CO)HB experiments [38, 39]. Hydrogen-bond restraints were derived from h3JNC' couplings observed in the h3JNC'HNCO experiments [40, 41] and were also identified from the pattern of sequential and interstrand NOEs involving HN and Hα protons, chemical shift index, analysis of the slowly and very slowly exchanging amide protons, and the first preliminary structures. Residual 1DNH couplings were extracted from the difference in J splitting measured for isotropic and anisotropic samples. The J splittings for 15N-1HN were obtained from 2D 1H-15N inphase–antiphase experiments performed in an interleaved manner [42].

Initial structures were calculated by simulated annealing using torsion angle dynamics with cns, version 1.1 [43], using NOE-derived distance restraints, hydrogen-bond restraints, and dihedral angle restraints. Finally, an ensemble of 100 EW29Ch structures was calculated using cns with residual 1DNH couplings by a simulated annealing protocol. Initial estimation for the axial component of the molecular alignment tensor (Da) and the rhombicity (R) were obtained on the basis of the preliminary structure calculated without residual 1DNH coupling using pales [44]. These values were optimized in an iterative manner, using the structures calculated by cns. The final values of Da and R were −8.0 Hz and 0.567, respectively. The final 20 lowest-energy ensemble structures were analyzed by molmol [45] and procheck-nmr [46] and images were created using molmol. The atomic coordinate data have been deposited in the Protein Data Bank under accession number 2RST.

15N NMR relaxation measurements

15N longitudinal and transverse relaxation times, as well as 15N{1H} NOE, were measured at 600 MHz at 25 °C. 15N longitudinal relaxation times (T1) were measured with relaxation delays of 10, 20, 40, 80, 160, 240, 320, 640, 960, 1280 and 1920 ms. 15N transverse relaxation times (T2) were obtained with 15N 180° Carr–Purcell–Meiboom–Gill pulses at total relaxation delays of 8, 16, 32, 48, 64, 80, 96, 112, 128, 160 and 192 ms. 15N{1H}-NOEs were determined by comparison of the peak intensities between the HSQC spectra with and without 1H saturation of 3 s recorded in an interleaved manner. The model-free analysis was carried out using tensor2 [12].

Acknowledgements

We thank Dr Masaki Mishima (Tokyo Metropolitan University) and Dr Chojiro Kojima (Osaka University) for help with the analysis of the RDC measurements. We also thank Ms Sachiko Unno (AIST) for the preparation of the proteins. This work was supported by a Grant-in-Aid for Scientific Research (KAKENHI) (C) (20580373 and 24580500) from the Japan Society for the Promotion of Science (JSPS) (to H.H., A.K. and J.H.).

Ancillary