Extending Janus lectins architecture: Characterization and application to protocells

Graphical abstract


Introduction
Due to their ability to crosslink various cells, such as red blood cells, lectins are originally known as agglutinins. They are generally multivalent and their selective interaction with glycoconjugates found many applications in the field of biotechnology and biomedicine as glycan-profiling tools [27]. Therefore, fine-tuning the specificity or the valency of lectins is a promising approach for obtaining novel tools. Synthetic biology strategies, such as the engineering of protein architecture, bring novel possibilities for lectin applications [8,13,23,24,26]. Lectin engineering is a novel domain of the synthetic glycobiology [30,32] and it can be performed at different levels, from glycan specificity to supramolecular architecture.
The trimeric oligomerization of Ralstonia solanacearum lectin (RSL) in a b-propeller shape (Fig. 1A) was recently used as a scaffold for building Janus lectins [22,25] with two faces and different specificities. The two Janus lectins, RSL-CBM40 and RSL-CBM77 Rf , were obtained by fusion of the monomeric RSL with two carbohydrate-binding modules (CBMs) with different specificity but rather similar shape, consisting of a single b-sandwich domain. This resulted in the creation of synthetic bispecific chimeras able to establish the interaction with fucose on one side and sialic acid (RSL-CBM40) or homogalacturonan (RSL-CBM77 Rf ), respectively, on the other side.
Other protein domains with different specificity, but also topology diverging from the b-sandwich CBM, could be considered for building novel Janus lectins from the trimeric RSL scaffold.
b-trefoils are robust domains that contain an internal tandem forming a three-lobed architecture with almost no secondary structures [19]. They display a large variety of functions, and many show carbohydrate-binding activity. They are considered as lectins due to their multivalence, but also as CBMs (e.g., CBM13) since they sometimes occur as domains associated with glyco-active enzymes or toxins [20,21]. While their shape is well-conserved, their sequences do not show strong similarity, except for the presence of hydrophobic residues in the core and a QXW repeat found in most sequences [11]. A large number of crystal structures of btrefoil lectins are available (212 from 63 different proteins), and based on their sequences they have been classified into 12 classes in the UniLectin3D database [2]. Datamining of genomes based on the lobe sequence signature of each class resulted in the prediction of thousands of putative b-trefoil lectins spanning all kingdoms of life [21]. The threefold symmetry results in the presence of three carbohydrate-binding sites referred to as a, b, and c although vari-ations in amino acid sequences sometimes result in only one or two active binding sites.
Among the known structures of b-trefoil lectin, the agglutinin from the fairy ring mushroom Marasmius oreades (MOA) presents an interesting strict specificity for aGal-containing oligosaccharides. The lectin rose attention already in the second half of the 20th century due to its high selectivity and affinity to blood group B oligosaccharide whereas very poor binding was detected to blood group A or H oligosaccharides [7,33]. MOA is specific for oligosaccharides with terminal non-reducing a1-3 linked galactose, while galactosides linked in a1-2, a1-4, and a1-6 showed no binding [33]. Due to its preference for the aGal1-3Gal disaccharide, MOA binds efficiently to blood group B (Gala1-3[Fuca1-2]Gal) but also to linear oligosaccharides such as Gala1-3Galb1-4Glc on glycosphingolipid isoGb3 and Gala1-3Galb1-4GlcNAc (Galili epitope) on glycoproteins [18]. This latter epitope is present on the cell surface of most mammals, but not in humans, apes, and Old World mon- keys due to the deactivation of the a3-galactosyltranferase during evolution [9]. Humans possess preformed antibodies directed against Gala1-3Gal, which are responsible for hyperacute rejection of animal (mainly porcine) organs in xenotransplantation attempts [4]. They are also responsible for a strong immune response against biodrugs such as therapeutical antibodies if produced in rodent cell cultures with active a1-3galactosyltranferase [3]. Highly selective lectins are therefore needed for the identification of such epitopes [17,29,33].
For extending the concept of Janus lectin, a fucose-specific bpropeller domain, and an a-galactose specific b-trefoil were selected ( Fig. 1), first to demonstrate the possibility of utilizing a b-trefoil lectin instead of a small CBM, and second to create the dual specificity fucose/aGal that does not exist in natural or artificial glycan-binding proteins. This can be of interest in biotechnology for identifying epitopes and in synthetic biology for crosslinking cells in artificial tissues. The crystal structure of MOA revealed that the lectin assembles as a homodimer with the monomer composed of two distinct domains, adopting b-trefoil fold at the N-terminus and a/b fold at the C-terminus (Fig. 1B). The C-terminal domain serves as a dimerization interface and retains a proteolytic function [5,15]. The lectin b-trefoil domain has a threefold symmetry, and the conserved motif (Gln-X-Trp) 3 is involved in the hydrophobic core of the structure. Co-crystals of MOA with Gala1-3Galb1-4GlcNAc revealed that each binding site has different ligand occupancy, emphasizing the fact that slight differences in amino acids might affect the binding (Fig. 1B) [10].
In the present study, we designed a Janus lectin as a fusion protein of monomeric RSL at the N-terminus and the MOA b-trefoil domain at the C-terminus (Fig. 1C). We produced and characterized the Janus lectin RSL-MOA with double specificity toward fucosylated and a-galactosylated glycans. The ability to bind these epitopes was tested with H1299 lung epithelial cancer cells and giant unilamellar vesicles. We also compared the biophysical behavior of this novel Janus lectin with RSL-CBM40 [25] and RSL-CBM77 Rf [22]. Additionally, we have engineered the b-trefoil domain of MOA (MOAbT) and compared its activity with RSL-MOA.

MOAbT
The gene for the b-trefoil domain of MOA was obtained by polymerase chain reaction where the plasmid pET-25b+RSL-MOA was used as a template. The primers, TTACATATGAGCTTACGTCGTGGC (Forward) and ATTACTCGAGTTACATGCGGTTGAAGTACC (Reverse) were designed to align to the full sequence of the b-trefoil domain of MOA and were ordered from Eurofins Genomics (Ebersberg, Germany). The restriction enzyme sites of NdeI and XhoI were added at 5 0 and 3 0 ends, respectively. Subsequently, the gene moabt and plasmid pET-TEV vector [12] were digested by RE NdeI and XhoI and ligated resulting in the pET-TEV-MOAbT vector. After transformation by heat shock in the E. coli DH5a strain, a colony screening was performed, and the positive plasmids were amplified and controlled by sequencing.

RSL-MOA
The original amino acid sequence of the lectin MOA from Marasmius oreades was obtained from the PDB database. The gene rsl-moa was designed as a fusion chimera with RSL at the N-terminus and the b-trefoil domain of MOA at the C-terminus via the linker PNGELLSS ( Supplementary Fig. 1). The gene was ordered from Eurofins Genomics (Ebersberg, Germany) after codon optimization for the expression in the bacteria Escherichia coli. The restriction enzyme sites of NdeI and XhoI were added at 5 0 and 3 0 ends, respectively. The synthesized gene was delivered in plasmid pEX-A2-RSL-MOA. Subsequently, plasmid pEX-A2-RSL-MOA and the pET-25b+ were digested by NdeI and XhoI restriction enzymes to ligate rsl-moa in pET-25b+. After transformation by heat shock in the E. coli DH5a strain, a colony screening was performed, and the positive plasmids were amplified and controlled by sequencing. E. coli KRX (Promega) cells were transformed by heat shock with the pET-25b+RSL-MOA plasmid and pre-cultured in LB media substituted with 50 lg/mL ampicillin at 37°C under agitation at 180 rpm overnight. The following day, 10 mL of preculture was used to inoculate 1 L LB medium with 50 lg/mL ampicillin at 37°C and agitation at 180 rpm. When reached an OD 600nm of 0.6-0.8, the protein expression was induced by adding 1 % Lrhamnose, and the cells were cultured at 16°C for 20 h.
The cells were harvested by centrifugation at 14000 Â g for 20 min at 4°C and the cell paste was resuspended in 20 mM Tris/HCl pH 7.5, 100 mM NaCl (Buffer A), and lysed by a pressure cell disruptor (Constant Cell Disruption System) with a pressure of 1.9 kBar. The lysate was centrifuged at 24 000 Â g for 30 min at 4°C and filtered on a 0.45 lm syringe filter prior to loading on an affinity column.

MOAbT
The cell lysate was loaded on 1 mL HisTrap column (Cytiva) preequilibrated with Buffer A. The column was washed with Buffer A to remove all contaminants and unbound proteins. The MOA was eluted by Buffer A in steps during which the concentration of imidazole was increased from 25 mM to 500 mM. The fractions were analyzed by 12 % SDS PAGE and those containing MOAbT were collected and deprived of imidazole by dialysis in Buffer A. The Nterminal His-tag was removed by TEV cleavage with the ratio 1:50 mg of TEV:protein in the presence of 0.5 mM EDTA and 1 mM TCEP over night at 19°C. After, the protein mixture was repurified on 1 mL HisTrap column (Cytiva) and the pure protein was concentrated by Pall centrifugal device with MWCO 3 kDa and stored at 4°C.

RSL-MOA
The cell lysate was loaded on 10 mL D-mannose-agarose resin (Merck) pre-equilibrated with Buffer A. The column was washed with Buffer A to remove all contaminants and unbound proteins and the flow-through was collected. RSL-MOA was eluted by Buffer A with the addition of 100 mM D-mannose or 100 mM L-fucose in one step. Due to not sufficient binding capacity of the column, the flow-through was reloaded on the column several times and the protein was eluted as described previously. The fractions were analyzed by 12 % SDS PAGE and those containing RSL-MOA were collected and dialyzed against Buffer A. The protein was concentrated by Pall centrifugal device with MWCO 30 kDa and the pure protein fractions were pooled, concentrated, and stored at 4°C.

Isothermal titration calorimetry (ITC)
ITC experiments were performed with MicroCaliTC200 (Malvern Panalytical). Experiments were carried out at 25°C ± 0.1°C. Protein and ligand samples were prepared in Buffer A. The ITC cell contained proteins in a concentration range from 0.05 mM to 0.2 mM. The syringe contained the ligand solutions in a concentration from 50 lM to 10 mM. 2 lL of ligands solutions were injected into the sample cell at intervals of 120 s while stirring at 750 rpm. Integrated heat effects were analyzed by nonlinear regression using one site binding model (MicroCal PEAQ-ITC Analysis software). The experimental data were fitted to a theoretical curve, which gave the dissociation constant (K d ) and the enthalpy of binding (DH).

Surface plasmon resonance (SPR)
The SPR experiments were performed using a Biacore X100 biosensor instrument (GE Healthcare) at 25°C. Biotinylated polyacrylamide-attached (PAA) sugars, such as PAA-a-fucose, PAA-a-galactose, and PAA-b-galactose (Lectinity) were immobilized on CM5 chips (GE Healthcare) pre-coated with streptavidin, as previously described [25]. In the sample cell, the PAA-sugars were immobilized either as a mixture for low-density chips (1:

Protein labeling
RSL-MOA and MOAbT were dissolved at 1 mg/mL in Dulbecco's phosphate-buffered saline (PBS) and stored at 4°C prior to usage. For fluorescent labeling, NHS-ester conjugated Atto488 (Thermo Fisher) or Cy5 (GE Healthcare) were used. Fluorescent dyes were dissolved at a final concentration of 10 mg/mL in water-free DMSO (Carl RothGmbH & Co), aliquoted, and stored at À20°C before usage according to the manufacturers protocol. For the labeling reaction, 100 lL of lectin (1 mg/mL) was supplemented with 10 lL of a 1 M NaHCO 3 (pH 9.0) solution. Hereby, the molar ratio between dye and lectin was 5:1 for RSL-MOA (cell assays) and 2:1 for RSL-MOA and MOAbT for GUV assays. The labeling mixture was incubated at 4°C for 90 min, and uncoupled dyes were separated using Zeba Spin TM desalting columns (7 kDa MWCO, 0.5 mL, Thermo Fischer). Labeled lectins were stored at 4°C, protected from light.
GUVs were prepared by the electroformation method as earlier described [34]. Briefly, lipids dissolved in chloroform of a total concentration of 0.5 mg/mL were spread on indium tin oxid-covered (ITO) glass slides and dried in a vacuum for at least one hour or overnight. Two ITO slides were assembled to create a chamber filled with sucrose solution adapted to the osmolarity of the imaging buffer of choice, either HBSS (for live-cell imaging) or PBS (for GUV-only imaging). Then, an alternating electrical field with a field strength of 1 V/mm was implemented for 2.5 h at RT. Later we observed the GUVs in chambers manually built as described [34].

Imaging of RSL-MOA and MOAbT binding to GUVs
Samples containing GUVs and lectins were imaged using a confocal laser scanning microscope (Nikon Eclipse Ti-E inverted microscope equipped with Nikon A1R confocal laser scanning system, 60x oil immersion objective, NA = 1.49, and four laser lines: 405 nm, 488 nm, 561 nm, and 640 nm). Image acquisition and processing were made using the software NIS-Elements (version 4.5, Nikon) and open-source Fiji software (https://imagej.net/software/fiji/).

Cell culture
The human lung epithelial cell line H1299 (American Type Culture Collection, CRL-5803) was cultured in Roswell Park Memorial Institute (RPMI) medium supplemented with 10 % fetal calf serum (FCS) and 4 mM L-glutamine at 37°C and 5 % CO2, under sterile conditions. Cells were cultivated in standard TC-dishes 100 (Sarstedt AG & Co. KG) until 90 % confluence, detached with trypsin (0.05 % trypsin-EDTA (1x) solution; Gibco, Thermo Fischer Scientific) and re-seeded for a subculture or for experiments. For experiments, cells were incubated with different concentrations of RSL-MOA for indicated time points.

Cell proliferation (MTT) assay
Cell viability of H1299 cells upon treatment with RSL-MOA was investigated in a standard MTT assay. Cells were treated with increasing concentrations of RSL-MOA for 24 h to monitor cell proliferation or cytotoxicity. 1 Â 10 4 cells per well were transferred to a 96-well plate with a U-bottom. The cells were centrifuged at 1600 Â g for 3 min at RT. The cell pellet was re-suspended in 100 lL of variously concentrated protein solutions (9, 18, 36, 90, 180, 360, 720, 1000 nM) and transferred to a 96-well flatbottomed plate. The cells were incubated for 24 h at 37°C. Subsequently, 10 lL of MTT labelling solution (MTT Cell Proliferation Kit, Roche) was added to each well, and the cells were incubated for 4 h at 37°C. At the end of incubation, 100 lL of the solubilisation reagent was added to each well, and the plate was further incubated at 37°C overnight. The next day, the absorbance of the samples was measured at 550 nm using a BioTek microplate reader. The reference wavelength was set at 690 nm. The data was further analyzed using Microsoft Excel and GraphPad Prism software.

Statistical analysis
All data in graphs are presented as mean ± standard deviation (SD) and were calculated from the results of independent experiments. Statistical testing was performed with GraphPad Prism software and Microsoft Excel using data of ! 3 biological replicates. Statistical differences in independent, identical samples were determined with a two-tailed, unpaired t-test. Non-significant results are not highlighted.

Differential scanning calorimetry (DSC)
The DSC experiments were carried out in Micro-Cal PEAQ DSC instrument (Malvern Panalytical). The protein samples were prepared in Buffer A, which could be substituted with the addition of a ligand. The sample cell contained the protein solutions in the concentration range of 9-29 mM. The reference cell contained the same buffer as present in the sample but without protein. The increase in temperature was measured from 20 to 130°C with a scan rate of 200°C/min. The data were analyzed by Micro-Cal PEAQ DSC software with a non-two-state and progress baseline method fitting model.

Design and production of the b-trefoil domain of MOA (MOAbT)
MOA occurs naturally as a dimer, assembled by a strong association between the two C-terminal domains. Monomeric MOA composed only of the b-trefoil domain would be of interest for biotechnology applications. The gene of the MOA b-trefoil domain was defined as the 156 N-terminal AAs of MOA full sequence and was amplified by PCR and subcloned into the vector pET TEV as a fusion with an N-terminal His-tag sequence. The resulting protein was named MOAbT and was recombinantly produced in soluble form in the bacterium E. coli BL21(DE3). The protein was purified by immobilized metal ion chromatography followed by His-tag cleavage by TEV (tobacco etch virus) protease. The resulting protein has an estimated molecular weight of 17.2 kDa. TEV cleavage followed by additional immobilized metal ion chromatography reduced the sample contaminants and the protein purity was verified by 12 % SDS PAGE electrophoresis (Supplementary Fig. 2A).

Design and production of Janus lectin RSL-MOA
Janus lectin RSL-MOA was designed as a gene fusion of monomeric RSL (N-terminus) and b-trefoil domain of lectin MOA (Cterminus). The fusion of the two protein domains by the eight AAs linker PNGELLSS results in the protein sequence of 255 amino acids (Supplementary Fig. 1). The gene sequence was optimized for bacterial expression, and the synthesized gene was subcloned into the plasmid pET25b+. The protein was produced in soluble form in the bacterium E. coli KRX. The protein was subsequently purified by affinity chromatography on an agarose-mannose column due to the interaction between the RSL domain and mannose residues. Protein analysis by 12 % SDS PAGE electrophoresis showed that Janus lectin RSL-MOA has an apparent size of 83 kDa, which corresponds to a trimer. RSL oligomerization appears to be resistant to denaturation conditions. A smaller amount of dimers and monomers were also visible, probably because of partial oligomer denaturation ( Supplementary Fig. 2B).

Biophysical characterization of MOAbT and RSL-MOA by isothermal titration calorimetry (ITC)
MOAbT affinity towards Gala1-3Gal disaccharide, the terminal disaccharide of Galili epitope, was assayed by titration microcalorimetry. Large exothermic peaks were obtained at the beginning of the titration (Fig. 2). The affinity was not strong enough to obtain a sigmoidal curve, and therefore, as recommended in such cases [31], the stoichiometry (N) was fixed, using a value of N = 2 since, as previously shown by Grahn et al., [10], MOA binding sites have different affinities, and we expected at least two of them to be active. A dissociation constant (K d ) of 150 lM was obtained, which confirms the functionality of the isolated b-trefoil domain, and is in good agreement with the K d measured for the whole MOA protein, i.e., 182 lM [33] (Fig. 2A).
RSL-MOA was designed to possess six binding sites for fucose and up to nine binding sites for a-galactose on the opposite faces.
The functionality of both binding interfaces was tested by various biophysical approaches. To this end, we designed an ITC experiment with a consecutive injection of both ligands. First, Gala1-3Gal was titrated into the cell containing RSL-MOA. Subsequently, the cell content from the first experiment (complex RSL-MOA/ Gala1-3Gal) was titrated by a-methyl fucoside (a-MeFuc) (Fig. 2B).
The thermogram obtained by titrating Gala1-3Gal in RSL-MOA solution was very similar to the one obtained for MOAbT. The dissociation constant (K d = 226 lM) was comparable to the K d for the isolated b-trefoil domain (as described above) and for the whole MOA [33]. The subsequent titration by a-MeFuc resulted in a sigmoid shape, due to stronger affinity, with a measured stoichiometry of N = 2, corresponding to the presence of two fucose binding sites per RSL monomer. The K d was measured to be 0.4 lM, in excellent agreement with the previously measured affinity (0.7 lM) for RSL [16]. This experiment proved that both parts of RSL-MOA are functional and able to bind their ligands at the same time.

Biophysical characterization of RSL-MOA by surface plasmon resonance
To evaluate the effect of multivalency and therefore measure avidity instead of affinity, surface plasmon resonance (SPR) was used. Streptavidin CM5 chips were functionalized by different ligands, i.e., biotinylated PAA-a-fucose, PAA-a-galactose, and PAAb-galactose.
No binding of RSL-MOA could be observed with a-galactose chip, even with high density surface (200 lg/mL biotinylated PAA-a-Gal), probably because of its very low affinity for the monosaccharide a-galactose (K D = 8 mM) [33]. On the opposite, RSL-MOA bound efficiently to fucosylated chips. In order to avoid mass transfer [6], low density (LD) CM5 chip was prepared with mixture of PAA-a-fucose/PAA-b-galactose in the ratio 1:9 (200 lg/mL) and RSL-MOA was injected in increasing concentrations (5 to 1000 nM). The regeneration step was performed between each injection using 1 M fucose solution. As shown in Fig. 3A, a dose-dependent response was obtained, with a steep association phase and very weak dissociation event. The steadystate analysis showed that even at the highest protein concentration (1000 nM) the chip surface was not saturated by RSL-MOA, i.e., the plateau phase was not reached (Fig. 3A). Nonetheless, the fitting of the kinetics constant could be performed and values for affinity (K D = 4.9*10 -10 M) and kinetics (k on and k off ) were estimated ( Table 1).
As an alternative, a single-cycle kinetic experiment was performed. RSL-MOA was injected at four concentrations (10 to 500 nM) (Fig. 3B). The advantage of this method is that no regeneration step is needed. The estimated K D of 5.9*10 -12 M was two orders lower if compared to the titration experiment; however, such a variation is acceptable in binding events characterized by avidity. In order to evaluate the capacity of monosaccharides to inhibit the multivalent binding and compete with protein binding to the chip surface, 50 nM RSL-MOA was pre-incubated with various concentrations of fucose (5 to 5000 lM). As shown in Fig. 3C, the complete inhibition was achieved in the presence of high concentrations of fucose (2500 and 5000 lM), and a IC 50 value of 7.8 lM was obtained.
The ability of RSL-MOA to bind the glycan-decorated surface of SPR chips was confirmed by several experimental setups. The variations in the affinity constant just illustrate the difficulty to quantify binding when avidity is the major event. These results are in agreement with previous ones evaluating RSL avidity for the fucose chip [1]. This avidity constant is at least 1000-fold higher than the affinity measured for fucose monosaccharide in solution, confirming the very strong cluster effect resulting from the topology of the RSL b-propeller with the presentation of six binding sites on the same face, binding very efficiently to fucose presented in a multivalent manner on a surface.
The Janus lectin RSL-MOA-Atto488 (green) binds efficiently to the surface of FSL-iGb3-GUVs (red), even at a concentration of 200 nM (Fig. 4C). The difference is striking when compared to the absence of binding of MOAbT-Atto488 (500 nM) (Fig. 4A). The super-multivalency resulting from the presence of three btrefoil domains, displaying a total of nine possible aGal binding sites, makes a very strong difference in terms of ability to bind to the glycosylated surfaces.
The capacity of RSL-MOA to crosslink GUVs carrying different oligosaccharides was tested using two populations of fluorescently labeled vesicles, FSL-iGb3-GUVs (labelled with Atto647N-DOPE, red) and FSL-A-GUVs (labelled with Atto488-DOPE, green). When the unlabeled RSL-MOA was incubated with the liposomes, crosslinking was observed between red and green GUVs, as well as, cross-linking between GUVs of the same color. The cross-linking between the same population GUVs is possibly due to lectin topology and multivalency, b-trefoil of MOA, and b-propeller of RSL. Herein, due to the double specificity of lectin RSL-MOA, we confirmed that multivalent and two-site-oriented topology has  Table 1 SPR statistics from titration and single-cycle kinetics experiments. The experiments were carried out on LD PAA-a-Fuc CM5 chips with a flow of 10 lL/min. clear potential in the cross-linking of glyco-decorated liposomes and proto-tissues (Fig. 4D).

RSL-MOA binds to human epithelial cells
The ability of RSL-MOA to recognize and bind to glycans on the surface of human cells was investigated by flow cytometry. RSL-MOA was fluorescently labeled with a Cy5 dye (RSL-MOA-Cy5) and incubated with the non-small cell lung cancer cell line H1299, for 30 min at 4°C. As depicted in the histograms of fluorescence intensity (Fig. 5A), cells treated with increasing concentra-tions of RSL-MOA-Cy5 (0.07 -0.7 lM) showed a dose-dependent binding of the lectin to the cell surface. In order to confirm lectin specificity and glycan-driven binding to receptors at the plasma membrane of H1299 cells, a number of inhibition assays was performed (Fig. 5B, C). RSL-MOA-Cy5 was pre-incubated with 100 mM fucose or 100 mM synthetic analogue of a-Gal epitope, pnitrophenyl-a-D-galactopyranoside (PNPG), respectively. The high concentrations of ligands were chosen due to previous observations with lectin RSL, for which a complete inhibition of binding to fucosylated receptors on the surface of H1299 cells was achieved only in the presence of 100 mM fucose [28]. The presence of 100 mM fucose significantly decreased RSL-MOA-Cy5 binding to treated cells by approx. 50 % compared to unblocked RSL-MOA-Cy5 (Fig. 5B), but did not abolish it entirely. Increasing the concentration up to 500 mM of fucose did not result in a further decrease in protein binding. The interaction of the engineered lectin with the cell surface was therefore not completely inhibited upon saturation of the fucose-binding sites, suggesting that the MOA domain partially compensates for the binding of Janus lectin to H1299 cells. On the other hand, 100 mM PNPG had almost no effect on RSL-MOA-Cy5 binding to treated cells, as indicated by the histograms in Fig. 5C. Fluorescence of cells incubated with RSL-MOA-Cy5 or RSL-MOA-Cy5 pre-treated with PNPG exhibited a similar intensity, as illustrated by the overlapping histograms (Fig. 5C). Reasonably, inhibition of protein binding was not achieved due to the low affinity of MOA toward monosaccharides [33].
These results suggest that both RSL and MOA domains of Janus lectin are able to bind to glycans present on H1299 cancer cells. Furthermore, we evaluated the potential cytotoxicity induced by RSL-MOA in treated cells. As illustrated in Supplementary Fig. 3, cell viability and proliferation were preserved at 24 h posttreatment with increasing concentrations of RSL-MOA (in the range 0 -1 lM), confirming the potential of this tool in biomedical research and applications.

Comparison of thermal stability of Janus lectins and their distinct domains
Janus lectins are created by assembling protein domains with different biophysical properties. It is therefore of interest to evaluate the effect of the fusion of these modules, and also to compare the biophysical properties of Janus lectins.
The thermal stability of RSL-MOA and its constituting domains was measured with differential scanning calorimetry (DSC). Thermograms of MOAbT and RSL are displayed in Fig. 6. These two modules have very different thermal stability with a denaturation/midpoint temperature (Tm) of 42°C for the MOAbT and 92°C (with two Tm at 90.5°C and 93.8°C) for the b-propeller RSL. It could be hypothesized that MOAbT is easily denatured as it represents only a single domain of the whole original protein.
The very strong thermal resistance of RSL is related to the compact and robust architecture of the b-propeller, and the two slightly different Tms extracted from the curve may correspond to the dissociation of the trimer followed by the unfolding of each monomer.
As expected, adding a-MeFuc as a ligand to the protein solution increased the Tm of RSL to values of approximately 100°C and 103°C, for sugar concentrations of 125 and 250 lM, respectively.
The thermal unfolding of RSL-MOA and other Janus lectins was carried out in the presence of a high concentration of fucose since the protein was tested after the purification procedure. RSL-MOA displayed two events of denaturation at very different temperatures, i.e., 40°C and 115°C. From the results obtained on the separated domains, it is clear that the lower Tm corresponds to the unfolding of the MOA b-trefoil and the second event, with higher Tm, corresponds to the denaturation of RSL. The Tm of the MOA moiety is not different when isolated or linked to RSL. The Tm of the RSL moiety is higher than for the b-propeller alone, which is probably due to the high concentration of fucose in the medium.
Other Janus lectins were also assayed by DSC for comparison. The proteins RSL-CBM40 and RSL-CBM77 Rf were obtained as described previously [22,25] and the buffer was supplemented by fucose as in the case of RSL-MOA. As already experienced with RSL-MOA, two separate events of denaturation were observed (Fig. 6) ( Table 2). The high-temperature Tm, corresponding to the denaturation of RSL, is observed, with a value as high as 121°C for RSL-CBM40. CBM77 Rf and CBM40 have a similar b-sandwich fold, and they present rather similar melting temperatures of 67°C and 65°C, respectively. This value is more than 25°C higher than the one observed for MOA, indicating a clear difference in thermal stability between b-sandwich and b-trefoil.

Discussion and conclusion
In this work, we extended the concept of Janus lectin and therefore developed a universal strategy for increasing lectin valency and introducing an additional specificity [22,25]. Janus lectins are engineered as fusion chimeras and due to the presence of the lectin RSL, these synthetic proteins assemble as trimers, resulting in the multiplication of lectin binding sites and thus expected higher affinity toward ligands. The first two Janus lectins, RSL-CBM40 [25,28] and RSL-CBM77 Rf [22] were designed as fusion chimeras of a lectin and CBM domain with a possible application as drug carriers or protein crosslinker in plant cell wall engineering, respectively. However, here we propose an alternative where the Janus lectin RSL-MOA is composed of two individual lectin domains, btrefoil from lectin MOA and b-propeller of RSL.
The biophysical properties of RSL-MOA were compared with MOAbT, the engineered b-trefoil domain of MOA lectin. Both proteins showed the same affinity toward ligands in solution (ITC experiments) confirming their activity. However, their behavior resulted differently when tested with glyco-decorated liposomes (GUV experiments). We observed that MOAbT showed almost no binding to FSL-iGb3-GUVs while RSL-MOA interacts with such vesicles and even mediates their crosslinking. These observations are probably conditioned by the super-multivalency of engineered RSL-MOA that displays nine sugar-binding sites, instead of only three in MOAbT. On the other hand, MOAbT displayed active binding with blood group B GUVs (FSL-B-containing GUVs) confirming that the lectin domain is able to bind glycoconjugates on the membrane surface. RSL-MOA also retains the ability to bind two differently glyco-decorated GUVs and crosslink them. This confirms that the engineered topology does not affect protein binding and such a lectin is of interest for the development of a highly selective tool for the construction of proto-tissues.
The SPR analysis of RSL-MOA revealed the strong dosedependent binding of the b-propeller RSL toward low-density PAA-a-Fuc CM5 chips. Even though almost no dissociation event was observed, the fitting procedure allowed for the estimation of binding constants. Depending of the experimental design, some variations are observed in affinity and kinetics, due to the strong avidity of the system, generating complexity. Nonetheless,  RSL-MOA showed at least 1000 higher affinity for surfacepresented oligosaccharides compared to the solution state, supporting the fact that the addition of MOA does not affect its binding properties.
The ability of RSL-MOA to recognize glycans exposed at the surface of H1299 cells was analyzed by flow cytometry, demonstrating a strong dose-dependent binding. H1299 cells are characterized by the presence of highly fucosylated glycoconjugates, among others, on their surface [14]. Furthermore, the effect of saturation of RSL-MOA sugar-binding sites with 100 mM fucose or 100 mM PNPG, was investigated and resulted in a decrease in the interaction between lectin and cells. We believe that the presence of the galactose-specific MOA domain partially compensates for the inhibition of fucose-dependent interaction between the lectin and the cell surface-exposed glycans. Our hypothesis is supported by previous observations with the lectin RSL, which showed similar binding properties to H1299 cells that could be fully inhibited by 100 mM fucose. On the other hand, 100 mM PNPG did not affect RSL-MOA binding, suggesting that the synthetic analogue of a-galactose (PNPG) did not bind MOA with sufficient affinity, as already observed previously with a-galactose monosaccharide [33] and confirmed by SPR observations in the present work. Additionally, RSL-MOA does not show any cytotoxicity toward H1299 cells making it a suitable candidate for the further development in biomedical applications. However, the probable ability of this lectin to agglutinate erythrocytes is a limitation for in vivo use, and the main application would be as a research tool, before further developments are carried out.
The thermal stabilities of three Janus lectins, i.e., RSL-MOA, RSL-CBM40, and RSL-CBM77 Rf , were compared by DSC. During protein denaturation, two events of unfolding are observed for all Janus lectins implying the fact that each protein domain has a different stability. Structurally similar CBMs (b-sandwich fold) share almost the same Tm, while b-trefoil is much more thermally unstable. On the other hand, the b-propeller RSL showed extremely high Tm, whereas the addition of the ligand has a tremendous effect on its stability. Our findings, therefore, suggest that the structural fold of the individual domains should be taken into account when designing novel Janus lectins if stability is to be considered.
Until now, three Janus lectin with different specificities and architecture have been engineered. Therefore, we are confident that with this strategy, the novel bispecific lectins with improved valency can be obtained and find their applications in numerous fields of biotechnology and biomedicine.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.