Tailoring a Novel Sialic Acid-Binding Lectin from a Ricin-B Chain-like Galactose-Binding Protein by Natural Evolution-Mimicry

Complex glycans are synthesized in various biological contexts, especially in those of higher animals including humans. Sialic acids (Sia) represent a group of non-reducing terminal sugars in mammalian glycoconjugates, possessing great structural diversity and biological significance. Variation in Sia derivatives, e.g. in linkage and positional isomers, greatly increases the range of glycan structural complexity and functional diversity. The latter includes regulatory roles in cell–cell interaction, cell-signalling and viral infections (1). Striking changes in sialylation features on tumour cells have been shown to serve as useful ‘tumour markers’ (2). In inflammation, the first step of leucocytic transmigration into endothelial cells is a cell-adhesion event mediated by the specific recognition of sialyl Lewis X by selections (3). In differentiation and development, SSEA-4 (Siaα2-3Galβ1-3GalNAcβ1-3Galα1-4R) is a well-known biomarker (4). Recently, structural glycomics using mass spectrometry (MS) has been used to elucidate functional aspects of glycans (5). However, direct analysis of sialoglycans by MS is technically difficult, because of the low ionization efficiency of this 9-carbon acidic sugar, and substantial decomposition of Sia linkages.

Lectin-based glycotechnology is expected to be an alternative means to profile complex features of glycoproteins in a rapid and comprehensive manner (6–8). In this regard, various types of lectin microarray have now become available for analysing glycan profiles expressed on individual glycoproteins and even cell and tissue extracts (9–15). However, these systems are incomplete in terms of the repertoire of Sia-binding lectins: this is partly because stable production in bacteria of mammalian (and thus, endogenous) Sia-binding lectins is difficult. In addition, few Sia-binding lectins of plant origin are available. Strong support for a biotechnological approach appears to be necessary in order to develop a satisfactory repertoire of Sia-binding lectins.

In this study, we describe a novel lectin generated by a rational strategy mimicking the natural process of lectin evolution (Fig. 1). In essence, most lectins in nature are thought to have developed from a more primitive molecule in the order Glc/Man- to Gal-specific lectins. Applying this hypothesis, Sia-specific lectins should have emerged in the last step of evolution to recognize ‘Sia-modified Gal’ residue (16). In fact, certain trace features of Gal-recognition remain in the naturally occurring Sia-binding lectins, e.g. Sambucus nigra agglutinin (SNA), Sambucus sieboldiana agglutinin (SSA), Polyporus squamosus lectin (PSL), Maackia amurensis leukoagglutinin (MAL) and galectins (17–24). Therefore, valuable Sia-recognition lectins should be easily acquired from the ancestral Gal-binding lectins by molecular evolution methodology.

This nature-mimicking strategy has been developed into a high-throughput technology by combination of error-prone PCR with a recently reinforced ribosome display system [Fig. 2; originally developed by Hanes et al. (25)]. From a practical viewpoint, the choice of an initial scaffold protein is crucial to the success of the strategy, as it must be sufficiently stable and productive in a bacterial expression system. In this study, we chose earthworm 29-kDa Gal-binding lectin (EW29) to fulfill the requirements (26). After enrichment with fetuin (a representative serum sialoglycoprotein) as a selection probe, one of the derived fetuin-binding lectins (designated SRC) was found to show substantial affinity for α2,6-sialylated glycans, which its parental molecule completely lacked. Notably, SRC retained lowered but detectable affinity for Gal, reminiscent of molecular evolution of Gal-binding lectins. X-ray crystallographic analysis of SRC strongly supported our hypothesis: i.e. modification of the original lactose-binding pocket resulted in the formation of an extended sub-recognition site for α2-6Sia. As far as we know, this is the first example of the application of molecular evolution to produce novel Sia-binding lectins. These findings have both academic and practical implications for future approaches to molecular design of carbohydrate-recognition molecules.

MATERIALS AND METHODS

Construction of the Plasmid, pRARE/ew29ch

For in vitro selection, Gene III gene containing an additional region utilized as a spacer (0.4 bp) was inserted into pET-27b vector (Qiagen) via XhoI and NheI sites (designated pRARE vector). EW29Ch gene was ligated into pRARE vector via NdeI and XhoI sites. His-SoCBM13β′ mutant gene from Streptomyces xylanase was inserted into pRARE vector via XbaI and XhoI sites (pRARE/his-socbm13β′) (27).

Preparation of a DNA library

Open reading frame of EW29Ch into pRARE/ew29ch was randomly mutated by error-prone PCR with 100 ng/ml pRARE/ew29ch, Master Mix (Promega), 5 mM MgCl₂, 0.5 mM MnCl₂, 1 μM T7P-for primer (5′-TAATACGACTCACTATAGGGGAATTGTG-3′) and 1 μM gene3-rev primer (5′-CCACCTCCCTCACCGCCGGTAGAAGATATC-3′) (2 min at 98°C, followed by 30 cycles of 30 s at 98°C, 30 s at 65°C and 45 s at 72°C, with a final 5 min at 72°C) (28). A spacer fragment was amplified by PCR with 200 ng/ml of pRARE/his-socbm13β′, 1 μM of the forward primer, ‘gene3-for’ (5′-CTCGAGGGGATATCTTCTACCGGCGGAGGA-3′), and 1 μM of the reverse primer, ‘RARE-rev’ (5′-GCTAGCAACCCGAACCCGTCCCCGTGAATT-3′). The synthesized DNA fragments were treated with DpnI and then were blunt-ended by T4 DNA polymerase. Using an overlap PCR method (29), the library fragment was fused to the spacer fragment. The overlap PCR product was thoroughly amplified using T7P-for primer and RARE-rev primer (5′-GCTAGCAACCCGAACCCGTCCCCGTGAATT-3′). The synthesized fragments were purified with QIAquick PCR purification Kit (Qiagen) and were transcribed with RiboMAX™ Express Large Scale RNA production Systems-T7 (Promega). The resulting mRNA was purified with RNeasy (Qiagen) and was used immediately for selection.

Selection with Fetuin–agarose

In vitro translation was performed with the mRNA library using Escherichia coli S30 Extract System for Linear Templates (Promega) for 10 min at 37°C. The translation reaction was stopped by placing on ice for 5 min and by diluting with ice-cold binding buffer (200 mM sodium phosphate buffer, 150 mM NaCl, 50 mM MgCl₂ and 1% Tween20, pH 5.7). Immediately after centrifugation at 10,000 × g for 5 min at 4°C, the supernatant was added to the pre-mixed solution containing 0.2 mg/ml fetuin-immobilized agarose, 1% BSA, 4 U/μl RNasin (Promega) in the binding buffer, and the reaction tube was rotated at 4°C for 10 min. After centrifugation, the slurry was washed with the binding buffer. The mRNA coupled with both ribosome and synthesized protein was dissociated with elution buffer (200 mM sodium phosphate buffer, 150 mM NaCl and 50 mM EDTA, pH 5.7) and the resulting mRNA fragment was purified with RNeasy. The purified RNA was reverse-transcribed and amplified with a one-step RNA PCR kit (AMV) (TaKaRa) using the ‘T7P-for’ and ‘RARE-rev’ primer set. The derived DNA fragment was used in the next selection. After the fourth selection cycle, the RT-PCR product was re-amplified by PCR using the ‘T7P-for’ and ‘gene3-rev’ primer set, and the resulting DNA fragments were inserted into pET-27b vector (Novagen) via NdeI and XhoI sites and were subjected to sequence analysis.

Protein Expression and Purification

Expression and purification of selected mutants were performed essentially as described previously (30). The selected mutants were expressed into E. coli BL21-CodonPlus (DE3)-RIL strain (Stratagene), and were purified by lactose–agarose (2 ml column volume) (Sigma). The purified mutant proteins were desalted and concentrated with an Amicon Ultra–15 (5000 MWCO, Millipore).

Lectin Microarray Analysis

Lectin microarray was performed as described previously (9). Cy3-labeled asialofetuin was prepared from Cy3-labeled fetuin by Sialidase A (Prozyme) digestion. Fifty μl of Cy3-labeled glycoprotein solution in the probing buffer (1% Triton X, 1 mM Ca²⁺, 1 mM Mn²⁺ and 0.5 M glycine in TBS) was applied to the lectin array on the glass slide, which was incubated at 20°C for 8 h. The reaction solution was discarded and the glass slide was rinsed twice with the probing buffer. A fluorescence image of the microarray was acquired by an evanescent-field fluorescence scanner, GTMASScan III (Nippon Laser & Electronics lab. Nagoya, Japan).

Frontal Affinity Chromatography (FAC)

FAC was performed using an automated machine, FAC-1 (Shimadzu) as described previously (31). Flow rate and column temperature were kept at 0.125 ml/min and 25°C, respectively. Elution of pyridylamine (PA)- and 2-aminobenzoic acid (AA)-labelled glycans and fetuin was monitored with a fluorescence detector (excitation/emission wave lengths: 310/380, 335/410 and 285/350 nm for PA, AA and fetuin, respectively). Elution of pNP-glycans was monitored with a UV detector (280 nm). Dissociation constants (K_d) and effective ligand contents (B_t) in FAC were determined as described previously (31, 32). FAC competition assay was performed using α2,6-sialylated bi-antennary N-glycan (503, see Fig. 3) and fetuin in the presence of 10–200 mM lactose (Wako).

Crystallization

pET27/SRCstop vector, in which a stop codon was inserted in the 3′ terminus of the SRC gene was constructed. Ligand-free SRC, SRC/lactose and SRC/6′-sialyllactose complexes were crystallized by mixing 5 μl of 15 mg/ml protein solution with/without 15 mg/ml lactose and 6′-sialyllactose (Funakoshi), and an equal volume of precipitant solution (conditions for ligand-free SRC and SRC/lactose complex; 0.2 M ammonium sulphate, 0.1 M sodium acetate trihydrate pH4.6, 30% PEG MME 2000; those for SRC/6′-sialyllactose complex; 0.2 M NaCl, 0.1 M phosphate–citrate pH4.2, 20% PEG 8000) equilibrated against 500 μl of the reservoir solution by the sitting- or hanging-drop vapour-diffusion methods. Crystals of SRC and two complexes were obtained at 4°C within 5 days.

Data Collection

X-ray diffraction experiments were conducted using a R-AXIS VII imaging-plate X-ray detector with CuK_α X-ray diffractometer (Rigaku). Data collections were processed and scaled using DENZO and SCALEPACK from HKL2000 package (33). All crystals belonged to monoclinic space group P2₁ and diffracted beyond ∼2.0 Å. Data collection and refinement statistics are summarized in Table 1. Matthews coefficient V_M values for these crystals were calculated to be 2.2 ų/Da and these crystals corresponded to a solvent content of ∼47.1, 47.0 and 45.5%, respectively (34). An initial model of ligand-free SRC was built by the molecular replacement method using MolRep in CCP4 suite with structure of EW29Ch as a search model (PDB entry 2D12) (35, 36). All crystallographic refinements were performed using CNS and the parameter set of Engh and Huber (37, 38). Manual model rebuilding was conducted with XtalView (39). Model building of SRC/sugar complexes was conducted as described earlier with the structure of SRC (PDB entry 2DRY) as a search model. The quality of all models was analysed with PROCHECK (40). The coordinates for both free and complex forms of SRC have been deposited in the RCSB Protein Data Bank (accession numbers 2DRY, 2DRZ and 2DS0).

RESULTS

Selection for Sia-Binding Lectins by Natural Evolution-Mimicry

To create Sia-binding lectins using the proposed strategy—natural evolution-mimicry, EW29 was chosen as a scaffold protein (26). EW29, is a Gal-binding lectin belonging to the ricin-B chain (R-type lectin) family. The molecule comprises two homologous domains, i.e. N-terminal (N-half, EW29Nh) and C-terminal domains (C-half, EW29Ch), each divided into three subdomains designated α, β and γ. Recent X-ray crystallographic analysis of EW29Ch in complex with lactose demonstrated that it had two Gal-binding pockets in subdomains α and γ. On the other hand, EW29Nh has a much weaker sugar-binding affinity, though it is required for high-affinity binding. From a practical viewpoint, EW29Ch has a considerable advantage over other proteins in terms of productivity (>10 mg soluble protein is obtained from a 1 l culture). In addition, R-type lectins have a wide range of sugar-binding specificities. Therefore, EW29Ch is expected to be well tailored for the development of Sia-binding lectins.

As a selection probe, we chose fetuin having α2,6-sialylated tri-antennary N-glycan as a major component, as well as short fragments of O-glycans known as sialyl Tn and T antigens. For selection, fetuin was immobilized to agarose at a concentration of 10 mg/ml. For construction of a library of randomly point-mutated EW29Ch genes, error-prone PCR was performed under the optimized conditions to attain both high error-incorporation and substantial amplification: i.e. 5 mM Mg²⁺, 0.5 mM Mn²⁺ and low-fidelity DNA polymerase were used for amplification of 1 μg of the DNA fragment (28). Theoretical diversity and average mutation rate were estimated to be >10¹² and 7.5 bps/gene, respectively.

After enrichment cycles with fetuin–agarose (Fig. 2), the resultant libraries were subjected to PCR under the normal conditions using a specific primer set (T7P-for and RARE-rev), which encompassed an entire region of the EW29Ch gene (∼1 kbp). The DNA fragments thus amplified were found to be substantially enriched as selection cycles proceeded (Fig. 4), when compared with a control EW29Ch. After the fourth cycle, the amplified fragments were inserted into an expression vector (pET-27b) via restriction sites (NdeI/XhoI), and the derived clones were extensively characterized. Sequencing of 40 such clones revealed considerable diversity at both nucleotide (8.8 mutations/gene) and amino acid (5.8) levels (Fig. 5). No overlap clone was found between the third (20) and the fourth (40) cycles (data not shown), while two (4F20 and 4F61, designated SRC later) of the latter were identical and had six amino acid changes; i.e. E148G, I227N, D230G, I231V, E237G and G239S (Fig. 6A). Of the 40 mutants, 31 were successfully expressed in E. coli as a major component of soluble extract, and 10 of them were bound to a lactose–agarose column and eluted with 0.2 M lactose. The selected 10 mutant lectins were then analysed by FAC using 12 representative glycans (003, 103, 105, 307, 313, 323, 503, 504, 602, 724, 728 and 736 in Fig. 3) (31). The above clones, 4F20 and 4F61, showed substantial affinity for α2,6-sialylated glycans, while the others and the parental EW29Ch did not (data not shown). We focused further analysis on this α2,6Sia-binding EW29Ch mutant (designated SRC, Sia-recognition EW29Ch).

Sugar-binding Specificity of SRC

As described earlier, SRC was successfully expressed in E. coli as a soluble form (>20 mg/l culture), and was purified on lactose–agarose. To investigate its binding features, we performed lectin microarray analysis (9). SRC, as well as SNA (α2,6Sia-binding lectin) belonging to the R-type lectin family, were spotted on the same surface-activated glass slide at appropriate concentrations (2 and 1 mg/ml, respectively). As probes, Cy3-labeled fetuin, sialidase-treated fetuin (asialofetuin) and bovine serum albumin (BSA, negative control) were used at a concentration of 1 μg/ml. Intense signals were observed on SRC when Cy3-fetuin (Sia-terminated) was used as a probe, whereas less intense signals were observed for Cy3-asialofetuin (Gal-terminated; Fig. 6B). On the other hand, EW29Ch showed specific binding for asialofetuin, but not at all for fetuin. This indicated that SRC had acquired substantial affinity for Sia, which the parental EW29Ch completely lacked. Notably, SRC retains weakened but detectable affinity for Gal. The results for SNA were consistent with the previous reports (17, 18).

To characterize the Sia-binding properties of SRC in greater detail, we further carried out FAC analysis using 54 glycans (Fig. 3) as described previously (31). The effective ligand content (B_t) of the SRC-column used, and the dissociation constant (K_d) for pNP-β-6′-sialyllactose were determined to be 1.1 nmol and 57 μM, respectively (Table 2 and Supplementary Figure 1). Overall sugar-binding features of SRC and EW29Ch are summarized in Fig. 7A and B, respectively, in terms of affinity constant (K_a) (K_a = 1/K_d; for K_d values, see Supplementary Table 1). SRC showed high affinity for a series of α2,6-sialylated N-glycans (501–506). The binding affinity increased in the order of the branching number; i.e. α2,6-sialylated mono- (501, K_d = 69 μM; 502, 70 μM), bi- (503, 40 μM) and tri-antennary (504, 27 μM) N-glycans. As observed in lectin microarray experiments, SRC showed weakened but detectable affinity for N-glycans having exposed Gal units (302–418; 99–400 μM). Affinity for α2,6-sialylated bi-antennary glycan (503, 40 μM) was 7.2-times stronger than its galactosylated form (307, 286 μM). On the other hand, parental EW29Ch showed no affinity for α2,6-sialylated N-glycans (503–506), whereas it showed high affinity for a series of galactosylated N-glycans (302–418; 1.9–13 μM). Neither SRC nor EW29Ch had detectable affinity for fully α2,3- (602, 705, 706, 710 and 712) and α2,8-sialylated glycans (707 and 713). The essential aspects of the investigated features are expressed in a radar chart (Fig. 7C and D).

Though showing affinity for both α2-6Sia and Gal, it is still possible that SRC could have a single sugar-binding site. To examine this point, we carried out a competition assay using FAC. When α2,6-sialylated bi-antennary N-glycan (503) was eluted as a fluorescent probe in the presence of a competitor saccharide, lactose, the retardation was completely abolished (Fig. 8A). When fetuin was eluted, the retardation (monitored by fluorescence due to Trp) was also canceled by the presence of lactose (Fig. 8B). This result clearly demonstrated that SRC had a single binding pocket responsible for both lactose and α2,6-sialylated glycans. This conclusion is consistent with the observation that SRC had no hemagglutinating activity (data not shown). When compared with carbohydrate-binding properties of naturally occurring Sia-binding lectins, SNA and SSA, belonging to the R-type lectin family, the obtained SRC showed rather moderate binding affinity for α2,6-sialylated glycans and lowered but significant affinity for their asialoforms (unpublished data). In this context, SRC is considered to be an intermediate between original Gal-binding and evolved Sia-binding R-type lectins.

Structural Determination of SRC

To elucidate a molecular-evolutionary mechanism for SRC, we performed X-ray crystallographic analysis for both the free and complex forms of SRC. For the latter, lactose and 6′-sialyllactose were used as a ligand (for crystal parameters and refinement statistics, see Table 1). The overall structure of SRC was identical to that of EW29Ch, i.e. β-trefoil fold (Fig. 9A and B) (Suzuki, R. et al., unpublished observations). However, a significant conformational change was observed in a large loop region of the subdomain γ (Fig. 9C). Since this area contains three mutations (D230G, I231V and E237G), these may have induced a dramatic flip of the side chains of W245 and I232, which in turn caused a dynamic shift (i.e. sharper hairpin formation) of the loop region containing E237G. This conformational change, together with an additional mutation, G239S, evolved a novel function, the recognition of α2-6Sia (described later). On the other hand, an electron density map corresponding to lactose was observed only in subdomain α of SRC (Fig. 9A). In this subdomain, the binding mechanism was essentially the same as that of EW29Ch (for detailed interactions, see Table 3).

The structure of SRC in complex with 6′-sialyllactose is shown in Fig. 9D. As in the case of SRC/lactose complex, an electron density map corresponding to the bound 6′-sialyllactose was observed only in the subdomain α of SRC (for detailed interactions, see Table 4). Notably, the manner of binding to the lactose unit of 6′-sialyllactose was identical to the binding to lactose (Fig. 9E). Importantly, additional interactions with α2-6Sia are depicted in SRC: they consist of two hydrogen bonds via G148 in subdomain α and S239 in the extended loop of subdomain γ (described earlier). In conclusion, 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.

DISCUSSION

A recent trend in molecular evolutionary engineering has been to focus mutation on local regions, such as complementarity-determining regions (CDR) in antibodies, substrate-binding sites in enzymes and loop regions of carbohydrate-recognition proteins (41–43). This type of approach, i.e. targeted mutagenesis, is represented by saturation mutagenesis and point mutagenesis based on modelling studies. In this report, we describe an alternative approach, i.e. random mutagenesis on an entire region, under the concept natural evolution-mimicry. Starting from a Gal-binding scaffold protein (EW29Ch), a novel Sia-binding lectin was successfully obtained by error-prone PCR and the reinforced ribosome display system. The resulting SRC preserved the practical advantages of high productivity in E. coli, and ease in purification using a conventional lactose-column. Both features apparently derived from the parental EW29Ch. Both lectin microarray and FAC analyses revealed specific SRC affinity for Sia. X-ray crystallographic analysis of SRC/sugar complexes demonstrated that SRC recognized 6′-sialyllactose with the conserved lactose-binding pocket (subdomain α) together with additional contacts in a distant region (γ). This cannot be easily attained by targeted mutagenesis.

As far as we know, there has been no report about three-dimensional structure of a Sia-binding lectin belonging to the R-type lectin family. Therefore, our present data will also be helpful to understand molecular recognition mechanisms of natural Sia-binding lectins, e.g. SNA, SSA and PSL, belonging to the same family (17–21).

Novel lectins, including SRC, produced by the above strategies should contribute, in various ways, to structural profiling of sialoglycoproteins, i.e. sialoglycomics. These include (i) lectin microarray analysis, (ii) FAC profiling, (iii) flow cytometry analysis and (iv) enrichment of sialoglycoproteins (6–9, 31). The present method will also be applied to create other Sia-binding lectins, e.g. those specific for α2-3/α2-8Sia glycosides, which form critical biomarkers.

Supplementary data are available at JB online.

The authors thank S. Nakamura-Tsuruta, N. Uchiyama, J. Kominami and Y. Itakura, AIST, for critical discussion and technical supports about FAC and lectin microarray analyses. We also thank J. Iwaki, AIST, for a helpful comment about the X-ray crystallographic study. This work was supported in part by New Energy and Industrial Technology Development Organization (NEDO) in Japan.

REFERENCES

Abbreviations:

Abbreviations:
 
• EW29

  earthworm 29-kDa lectin

 
• EW29Ch

  C-terminal domain of EW29

 
• EW29Nh

  N-terminal domain of EW29

 
• FAC

  frontal affinity chromatography

 
• PA

  pyridylaminated

 
• pNP

  p-nitrophenyl

 
• SRC

  Sia-recognition EW29Ch