Fucosylated inhibitors of recently identified bangle lectin from Photorhabdus asymbiotica

A recently described bangle lectin (PHL) from the bacterium Photorhabdus asymbiotica was identified as a mainly fucose-binding protein that could play an important role in the host-pathogen interaction and in the modulation of host immune response. Structural studies showed that PHL is a homo-dimer that contains up to seven l-fucose-specific binding sites per monomer. For these reasons, potential ligands of the PHL lectin: α-l-fucopyranosyl-containing mono-, di-, tetra-, hexa- and dodecavalent ligands were tested. Two types of polyvalent structures were investigated – calix[4]arenes and dendrimers. The shared feature of all these structures was a C-glycosidic bond instead of the more common but physiologically unstable O-glycosidic bond. The inhibition potential of the tested structures was assessed using different techniques – hemagglutination, surface plasmon resonance, isothermal titration calorimetry, and cell cross-linking. All the ligands proved to be better than free l-fucose. The most active hexavalent dendrimer exhibited affinity three orders of magnitude higher than that of standard l-fucose. To determine the binding mode of some ligands, crystal complex PHL/fucosides 2 – 4 were prepared and studied using X-ray crystallography. The electron density in complexes proved the presence of the compounds in 6 out of 7 fucose-binding sites.

Lewis b tetrasaccharide. Tetravalent constructs 6 and 7 contain a calix [4]arene scaffold in a 1,3-alternate conformation that ensures sufficient solubility in water. Both 6 and 7 24 were prepared by CuAAC click reaction 25,26 from azide 1 and the respective tetrapropargyl calix [4]arene. Finally, a set of dendrimers 8-10 22 with valences ranging from 4 to 12 are more flexible and complex ligands than calix [4]arene derivatives, thus a significant increase in the complexity of the binding event can be expected. The affinity improvements in multivalent carbohydrate−lectin interactions have been attributed to the so-called "cluster glycoside effect" 27-29 . Hemagglutination inhibition assay. PHL as a fucose-specific lectin was shown to agglutinate papain-treated RBCs of blood group O 16 with the surface-exposed terminal trisaccharide α-l-Fucp-(1 → 2)-β-d-Galp-(1 → 3/4)-d-GlcpNAc. The set of potential ligands 1-10 with different levels of fucosylation was tested, and their inhibition potency on hemagglutination by PHL was assessed (Table 1, Fig. 2). DMSO-containing blank did not affect the hemagglutination in any way. To assess the contribution of valency to the affinity increase, an affinity improvement factor β was calculated as the relationship MIC basic unit /(valency × MIC ligand ).
Simple monovalent fucosides 2-4 failed to substantially inhibit the hemagglutination caused by PHL, while azide 1 and difucoside 5 were 16-fold better inhibitors than l-fucose and twice as good as methyl α-l-fucopyranoside. If the structures of simple ligands 1-5 are compared, it seems that the nonpolar substituent at the anomeric position of l-fucose provides an advantage. The activities of tetravalent calixarenes 6 and 7 are identical, giving a potency of 31.3 without regard to the presence or absence of a tert-butyl group on the upper rim of the skeleton. Tetravalent dendrimer 8 was as efficient inhibitor of hemagglutination as calixarenes 6 and 7, although it occupies a different topology. Hexa-and dodecafucosylated dendrimers 9 and 10 were proven to be the best inhibitors. They were more than 100-fold better ligands than natural l-fucose. However, according to the β factor calculated with respect to azide 1 (β Azide1 ), all multivalent compounds were comparable with their building unit.
Surface plasmon resonance. To analyse the competitive inhibition of PHL binding to a multivalent surface, the surface plasmon resonance (SPR) technique was employed. A sensor chip presenting α-l-fucopyranoside residues was treated with a constant concentration of PHL, while an increasing concentration of the ligands was used to determine IC 50 (Table 2, Fig. 3). Only the ligands comparable with or better than α-Me-Fuc in the hemagglutination assay were investigated by SPR. To assess the contribution of valency to the affinity increase, an affinity improvement factor β was calculated as the relationship IC 50,basic unit /(valency × IC 50,ligand ) 30 .
Comparing the IC 50 values and potency, the results from SPR are in good agreement with the finding obtained by hemagglutination. Azide 1 and fucoside 3 exhibit 32-fold and 38-fold, respectively, stronger affinity to PHL than its counterpart l-fucose; in fact, an 86-fold increase in affinity was found for difucoside 5. The affinity improvement factors of tetravalent calixarenes 6 and 7 as well as dendrimer 8 were comparable with β Azide1 value reaching approximately 15. The hexavalent dendrimer 9 was the best ligand of the lectin PHL with IC 50 = 37 nM, β Azide1 = 52 and with potency 9,800-fold and 310-fold higher than l-fucose and azide 1, respectively. Although dodecavalent dendrimer 10 was also an efficient ligand (IC 50 = 80 nM), its affinity improvement factor was substantially lower (β Azide1 = 12) than that of the hexavalent dendrimer 9. This observation could be caused by unwanted side effects such as cross-linking or aggregation induced by the sterically demanding structure of 10. Nevertheless, the data support a strong glycoside clustering effect.
Isothermal titration calorimetry. The binding of PHL to different α-l-fucoside compounds was further characterized by isothermal titration calorimetry (ITC), enabling determination of the complete thermodynamic profile of the molecular interactions (Table 3, Fig. 4). As for calixarenes, both compounds caused a cross-linking/ aggregation of PHL accompanied by the formation of visible precipitates. Because of this, it was not possible to evaluate the data, despite the obvious interaction. The dendrimers also cross-linked/aggregated the protein as with calixarenes, however the curves provided clear outcomes with an affinity in the low micromolar range. The dissociation constants (K D ) of all dendrimers correlate with their valency -the higher number of fucose, the more decreased K D . This implies that dodecafucosylated dendrimer 10 was assessed as the best inhibitor, with K D almost 4 times better than the second-best hexavalent dendrimer 9 and 33 times better than azide 1 (β Azide1 = 3). The stoichiometry value n corresponds to the number of terminal fucoses per ligand molecule. The n value decreases from 4 to 1 as the number of fucose units increases from 1 to 12 (Table 3). As for monovalent (azide 1, fucoside 3) and divalent (difucoside 5) compounds, all of them exhibited a low affinity towards PHL (in the millimolar range), as is usually observed for lectin/saccharide interactions. They also correspond with the same β factor. P. asymbiotica cross-linking -Bacterial aggregation properties. Because of the aggregation of PHL by calixarenes and dendrimers during SPR and ITC experiments, we decided to compare tetravalent compounds at a higher level. We performed a series of in vitro aggregation assays to reveal their aggregation properties towards P. asymbiotica subsp. australis (Table 4, Fig. 5). Prior to each assay, we confirmed the absence of aggregates in two negative controls -bacterial cells in PBS buffer and in the presence of azide 1 as a monomer. With all chosen compounds, we observed bacterial aggregates of a variable size that confirmed that they are capable   www.nature.com/scientificreports www.nature.com/scientificreports/ of interacting with a cell surface. This effect may be caused by interaction of the multivalent inhibitors with PHL or other receptors. The concentration of the first appearance of aggregates varied according to the nature of the ligand used. The highest efficiency (0.625 mM) was observed for the tetravalent dendrimer 8, as we expected. Calixarene 6 proved to be only half as good as dendrimer 8, and aggregated cells at a concentration of 1.25 mM. An unexpected observation was that calixarene 7 had the lowest aggregation potential. In contrast to other techniques (hemagglutination, SPR and ITC), where calixarenes 6 and 7 provided similar results, during cross-linking experiments calixarene 7 was four times worse than calixarene 6. At the same time, both tetrameric calixarenes 6 and 7 are in a 1,3-alternate conformation with the two-faced presentation of the fucopyranosyl epitopes on opposite regions probably occupying a similar spatial orientation. The only difference is the tert-Bu group on the upper rim in calixarene 6, which might contribute to nonpolar interactions with whole P. asymbiotica cells.
X-ray structure of PHL/fucosides complex. The PHL lectin forms a seven-bladed β-propeller assembling into a homo-dimer. Both units contain two types of binding sites situated between the blades and have different amino acid compositions and binding preferences 16 . As mixing PHL with the branched compounds leads to protein precipitation and thus prevents co-crystallization, the direct study of complexes was not possible. Soaking ligand-free PHL crystals with a low concentration of individual branched compounds did not result in complex formation while soaking with their higher concentrations led to a crystal decomposition. Therefore, we soaked the PHL crystals with monomeric fucoside 2, 3 and 4, respectively. The structure of these complexes was determined using X-ray diffraction (Table 5).   Table 3. Thermodynamic profiles for interaction between PHL and ligands determined by isothermal titration calorimetry at 25 °C. Thermodynamic parameters were calculated from two independent measurements. To assess the contribution of valency to the affinity increase, an affinity improvement factor β was calculated as the relationship K D,basic unit /(valency × K D,ligand ). *The stoichiometry value of l-fucose was fixed during the fitting procedure because of the low-affinity interaction that also causes that ΔH and ΔS values may suffer from higher inaccuracy. The value was based on the α-Me-Fuc measurement and on the X-ray crystal structure of the PHL/ BGH complex.  www.nature.com/scientificreports www.nature.com/scientificreports/ fucosides and experimentally confirmed a presence of at least 6 binding sites per monomer. In previous studies, only 2 to 5 binding sites were detected 16,31 . The second saccharide unit was clearly assigned for all three fucosides 2-4 in sites 1 and 3. For fucoside 3 and 4, the second saccharide was also distinguishable in sites 4-7 of at least one protein monomer (Table 6). This allows us to compare the binding mode for C-glycoside-based molecules and published previously O-glycoside structures. The fucose moiety is coordinated in the same way as was shown for α-Me-Fuc and blood group H trisaccharide 16 . Briefly, the O3, O4 and O5 of fucose are coordinated by the backbone atoms of two neighbouring loops and in sites 1, 3, and 5-7 also by the side chain of nearby threonine. The C6 methyl group and nonpolar surface of fucose are stabilized by a CH-π interaction with two tryptophan side chains. Considering fucosides 2-4, the position of the bridging methylene group is close to the α-Me-Fuc O1 atom, which is not directly coordinated by the protein in the complex. Additional direct interactions between the protein and the studied molecules are infrequent. The O4 of the d-arabino-hexopyranosyl unit in fucoside 2 and 4 is coordinated by a glycine backbone nitrogen in sites 1, 3, 5, and 6, while the O6 atom of fucoside 3 is linked to a tryptophan backbone oxygen in sites 4 and 6. When comparing the overall orientation of the second hexopyranosyl unit of fucosides 2-4, it is similar to the galactose unit of blood group H trisaccharide, with minor distortions in individual binding sites influenced e.g. by crystal packing. The remaining parts of fucosides 2 and 3 were not resolved in the electron density, suggesting their negligible importance for ligand binding.

conclusions
In the previous study 16 , we identified an interesting lectin from emerging human pathogen P. asymbiotica harbouring seven potential fucose-binding sites per monomer and able to modulate innate immune system of human. Infection strategies used by pathogens often involve highly specific protein/carbohydrate interactions and therefore design of suitable inhibitors preventing this interaction is needed. Thus, we tested inhibition potency of l-fucose-based compounds with different valences and topologies. All these structures share the same feature of bearing C-glycosidic bond instead of the common but physiologically unstable O-glycosidic bond.
α-l-fucopyranosyl-containing mono-, di-, tetra-, hexa-and dodecavalent ligands were investigated by hemagglutination, ITC, SPR, X-ray crystallography, and cell cross-linking. The binding mode of monovalent ligands was studied via crystal complexes with PHL where the electron density proved the presence of the fucosyl part of the compounds in all accessible fucose-binding sites. The second saccharide unit of studied fucosides was only marginally coordinated by PHL binding sites what, together with affinity comparable to monosaccharides, suggests a low importance of this type of compounds for inhibitor design. On the other hand, the affinity towards azide 1 is thirteen times higher than towards l-fucose and comparable to previously studied propargyl-α-l-fucoside 31 . This suggests, together with data from the previously studied PHL complexes, that introduction of a linker with delocalized π-electrons has not negligible contributory effect to binding affinity and increases also binding site availability/occupancy (4-5 binding sites for azide 1/propargyl-α-l-fucoside vs. 2-3 sites for α-Me-Fuc). This effect was already reported for the PA-IIL lectin and p-nitrophenyl-α-l-fucoside 32 . The affinity of PHL for multivalent ligands reached low micromolar values, which corresponds to an affinity three orders of magnitude higher than that of standard l-fucose and two orders of magnitude higher than that of monovalent azide 1. We have further demonstrated that all of the tested compounds were able to inhibit the PHL binding towards both artificial and natural fucosylated surfaces. Generally, the potency of ligands depended on the valency and types of polyvalent structures (calix [4]arenes and dendrimers). Even though the relation between the increasing number of fucoses per cluster and increasing affinity, the most active compound was a hexavalent dendrimer exhibited an IC 50 of 37 nM and K D of 13 µM. Several tetravalent compounds of different type were also tested for their ability to agglutinate P. asymbiotica bacteria. The cell clumps at different concentrations of tested compounds were observed www.nature.com/scientificreports www.nature.com/scientificreports/ further supporting the previously reported presence of lectins on surface as was also proved for other bacteria (e.g. P. aeruginosa, E. coli) 31,[33][34][35] .
Based on the findings presented in this work, we consider these compounds as a milestone on the way to design efficient inhibitors targeting P. asymbiotica.
Protein production and purification. The PHL lectin was produced in Escherichia coli Tuner(DE3)/ pET25b_phl cells as previously published 16 . In short, the cells were grown in LB broth medium containing 100 μM ampicillin at 37 °C. After reaching an OD 600 of 0.5, gene expression was induced with 0.2 mM isopropyl  www.nature.com/scientificreports www.nature.com/scientificreports/ β-d-1-thiogalactopyranoside (IPTG). Cells were incubated for an additional 20 hours at 18 °C, harvested by centrifugation at 12,000 g for 10 min and resuspended in buffer A (20 mM Tris/HCl, 300 mM NaCl, pH 7.5). Harvested cells were stored at −20 °C prior to protein purification.
For protein purification, cells were disintegrated by sonication (VCX 500, Sonics & Materials, Inc., USA) and the cytosolic fraction containing soluble PHL was collected by centrifugation at 21,000 g at 4 °C for 1 hour and filtrated through a 0.45 μm pore size filter (Carl Roth, Germany). Recombinant protein PHL was purified with isocratic elution on a d-mannose-agarose (Sigma-Aldrich, USA) resin equilibrated with buffer A by affinity chromatography using an ÄKTA FPLC system (GE Healthcare, UK) and used for further studies. For agglutination assays, purified PHL was dialyzed against PBS buffer (137 mM NaCl, 2.7 mM KCl, 8 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , pH 7.4).

Hemagglutination inhibition assay.
Human red blood cells (RBCs) O were processed according to previously published work 16 . In brief, RBCs O treated with sodium citrate were washed four times by PBS buffer (137 mM NaCl, 2.7 mM KCl, 8 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , pH 7.4), diluted to 50% by PBS with 0.005% (w/w) sodium azide and treated by 0.1% papain for the duration of 30 minutes.
Hemagglutination inhibition assay was performed for specificity and semi-quantitative affinity of the PHL interaction with the compounds. All carbohydrate inhibitors in 50 mM starting concentration were serially diluted in the PBS buffer and used for a determination of the lowest inhibiting concentration. As a blank, the PBS buffer supplemented with the appropriate amount of DMSO was used in order to exclude potential interference with the interaction. The individual samples were mixed with the lectin in concentration 2 mg/ml in a 5 μl:5 μl ratio. Thereafter, 10 μl of 10% RBCs O in PBS buffer was added, thoroughly mixed and incubated for 10 minutes at room temperature 36 . Following incubation, the mixture was again mixed and the reaction was observed on microscope slides using Levenhuk 2L NG microscope with Levenhuk D2L digital camera (Levenhuk, USA). Imagines were obtained via software ToupView for Windows (Levenhuk). The positive (PHL without inhibitor) and negative control (reaction without PHL) were prepared and processed in the same way using the appropriate volume of PBS buffer instead of the not included components. The lowest concentration of inhibitor able to inhibit agglutination was determined and compared with the standard (l-fucose). Minimal inhibitory concentrations (MIC) of synthesised inhibitors were determined from two independent measurements. Surface plasmon resonance. Surface plasmon resonance (SPR) experiments were performed on a BIAcore T200 instrument (GE Healthcare, UK) at 25 °C, using buffer A supplemented with 0.05% Tween 20 as a running buffer. α-l-fucoside was immobilized onto CM5 sensor chip (GE Healthcare, UK) covered with a carboxymethylated dextran matrix. The sensor chip surface was activated with N-ethyl-N-(3-dimethylaminopropyl)carbodiimide/N-hydroxysuccinimide solution and then coated with streptavidin using the manufacturer's standard protocol using HBS buffer (10 mM HEPES, 150 mM NaCl, 0.05% Tween 20, pH 7. 5). Unreacted groups were blocked with 1 M ethanolamine-HCl, pH 8.5. Biotinylated carbohydrate (biotinilated monomeric probes, Lectinity, Russia) was injected onto particular measuring channel and pure biotin on blank channel at a flow rate of 5 μl/min. SPR inhibition measurements were carried out on measuring channel with l-fucoside at a flow rate of 5 µl/ min. PHL was diluted to a concentration of 20 μg/ml by the buffer A supplemented with 0.05% Tween 20, mixed 1:1 (v/v) with various concentrations of inhibitors (500-5 μM) in the same buffer and injected onto the sensor chip. The response of lectin bound to the sugar surface at equilibrium was plotted against the concentration of inhibitor in order to determine IC 50 (concentration of inhibitor resulting in 50% inhibition of binding). As IC 50 is not a constant and depends on the experimental set-up, a parameter called potency was used for characterization. The potency of a tested inhibitor is the ratio of IC 50 of a chosen standard inhibitor (in this case l-fucose) and the inhibitor in question. Pure PHL lectin was used as a control (0% inhibition) and channel with pure biotin served as a blank. IC 50 was determined from triplicates.
Isothermal titration calorimetry. PHL protein in buffer A was equilibrated at room temperature at least for 30 min before ITC measurement. All ITC experiments were performed using ITC200 calorimeter (Malvern Panalytical, UK) at 25 °C. Carbohydrate ligands dissolved in buffer A were used at different concentrations. Protein in the calorimeter cell (50 μM) was titrated by consecutive additions (2 μl) of the ligand (1.5-5 mM) in the syringe while stirring at 1000 rpm. Integrated heat effects were analysed by nonlinear regression using a single-site binding model and global fitting in Origin 7 software (Microcal Instruments) 37 . The thermodynamic parameters of synthesised inhibitors were determined from two independent measurements.  www.nature.com/scientificreports www.nature.com/scientificreports/ Crystallization and data collection. PHL was concentrated to 13.6 mg/ml using an ultrafiltration unit with a 10-kDa cut-off membrane (Vivaspin 20, Sartorius, Germany) and crystallized under previously described conditions 16 . The crystals were obtained under the following conditions: 4 μl sitting drop, protein solution mixed with precipitant (3.7 M NaCl, 100 mM Hepes, pH 7.5) in ratios 1:1, 3:5, 1:3 and 1:7. The drops were set against