Galectin Binding to Neo-Glycoproteins: LacDiNAc Conjugated BSA as Ligand for Human Galectin-3

Carbohydrate-lectin interactions are relatively weak. As they play an important role in biological recognition processes, multivalent glycan ligands are designed to enhance binding affinity and inhibitory potency. We here report on novel neo-glycoproteins based on bovine serum albumin as scaffold for multivalent presentation of ligands for galectins. We prepared two kinds of tetrasaccharides (N-acetyllactosamine and N,N-diacetyllactosamine terminated) by multi-step chemo-enzymatic synthesis utilizing recombinant glycosyltransferases. Subsequent conjugation of these glycans to lysine groups of bovine serum albumin via squaric acid diethyl ester yielded a set of 22 different neo-glycoproteins with tuned ligand density. The neo-glycoproteins were analyzed by biochemical and chromatographic methods proving various modification degrees. The neo-glycoproteins were used for binding and inhibition studies with human galectin-3 showing high affinity. Binding strength and inhibition potency are closely related to modification density and show binding enhancement by multivalent ligand presentation. At galectin-3 concentrations comparable to serum levels of cancer patients, we detect the highest avidities. Selectivity of N,N-diacetyllactosamine terminated structures towards galectin-3 in comparison to galectin-1 is demonstrated. Moreover, we also see strong inhibitory potency of our scaffolds towards galectin-3 binding. These novel neo-glycoproteins may therefore serve as selective and strong galectin-3 ligands in cancer related biomedical research.

Here, we use serum albumin as scaffold for multivalent glycan presentation for lectin binding. Albumin is a highly abundant non-glycosylated serum protein and the protein of choice because it contains multiple lysine residues which are easily addressable for chemical conjugation methods [32][33][34]. Neo-glycoproteins are in general synthesized by chemical conjugation of glycans to native proteins [35,36]. They can be specifically designed for lectin binding studies by attachment of various types of glycan epitopes and by their glycosylation density [37][38][39].
The LacDiNAc (GalNAcȕ1,4GlcNAc) epitope is overexpressed in certain parasites, but has a rather low abundancy in mammalian cells [60]. Gal-3 was found to bind to LacDiNAc and mediate immune recognition [60]. In human cancer progression, cancer type dependent up-or downregulation of LacDiNAc expression was observed [61,62]. Additionally, it was shown that the LacDiNAc epitope is present on the gastric mucin and recognized by an adhesin from Helicobacter pylori [63,64] which may play a role for bacterial colonization in the gastric mucosa.
Recently, we could confirm LacDiNAc as a selective ligand for Gal-3 compared to galectin-1 (Gal-1) [65]. The inhibitory effect of divalent LacDiNAc for Gal-3 binding to asialofetuin glycoprotein, a model ligand for galectins, was weakly dependent on the linker length. However, a multivalent effect was not observed. In our previous studies, we found that the poly-LacNAc ([3Galȕ1,4GlcNAcȕ1]n = 2 -4) oligomers are preferentially bound by Gal-3 [66]. We concluded that multivalent presentation of a tetrasaccharide glycan structure with a LacDiNAc epitope on a suitable scaffold should lead to increased inhibition of Gal-3 binding to glycoproteins.
We here report on the synthesis of two types of bovine serum albumin (BSA) based neo-glycoproteins carrying the tetrasaccharide structures LacNAc-LacNAc and LacDiNAc-LacNAc, respectively, with various degrees of multivalency. First, the glycans carrying an amino terminated linker at their reducing end were synthesized by multi-step chemo-enzymatic synthesis as previously reported [65,67]. Chemical conjugation to lysine residues of BSA was accomplished by homobifunctional amino-reactive linker squaric acid diethyl ester enabling crosslinking via primary amino groups [68][69][70][71][72][73][74][75]. Variation of the molar ratios of glycan with respect to lysine residues resulted in the synthesis of 11 neo-glycoproteins of both types with different degrees of glycan modification. The multivalent neo-glycoproteins were finally evaluated in binding studies with human Gal-3 and Gal-1 to determine their binding and affinity properties as well as inhibitory potential.

Results and Discussion
Here, we present neo-glycoproteins with varying glycosylation density based on BSA and their application in galectin binding studies. For our purposes, the oligomers LacNAc-LacNAc and LacDiNAc-LacNAc were synthesized de novo chemo-enzymatically. Decoration of BSA was accomplished by a two-step conjugation reaction using squaric acid diethyl ester as a linker. Irrespective of the accessibility, BSA can be decorated with up to 60 glycans per molecule due to the presence of 60 lysine residues. The synthesized neo-glycoproteins are tested as ligands for human Gal-3 and Gal-1.
In conclusion, multi-step chemo-enzymatic synthesis and deprotection provided two pure tetrasaccharides (LacNAc-LacNAc and LacDiNAc-LacNAc) in good yields. Scheme I. Chemo-enzymatic synthesis of tetrasaccharides 4 and 5 using recombinant glycosyltransferases for complete conversions.

Synthesis and Purification of Squaric Acid Monoamide Esters
Consecutive attachment of two different ligands with the same functional group is the most attractive feature of squaric acid diethyl ester (8) [72]. Thus, 8 offered ideal chemical properties for BSA decoration with LacNAc-LacNAc or LacDiNAc-LacNAc. Amino group providing oligomers 6 and 7 were conjugated to 8 forming squaric acid monoamide esters 9 and 10 (Scheme II). Formation of squaric acid bisamide was suppressed by constant pH 7 and the presence of triethylamine (Et3N) as external base. Complete conversion of 6 and 7 could only be achieved by applying a four-fold molar excess of 8 and equal amounts of Et3N. The purity of the products 9 and 10 had highest priority as trace impurities of 8 would lead to protein cross-linking reactions during the second conjugation step. Thus, removal of residual 8 was achieved using preparative HPLC. Additionally, LC-MS results ( Figure S1c,d) confirmed integrity and purity of compounds 9 and 10. Molecular masses of 972.3 m/z and 1013.6 m/z, respectively, were found for 9 and for 10.

Synthesis and Analysis of Neo-Glycoproteins
Aqueous solution with increased pH was utilized as reaction buffer for squaric acid bisamide production, as recommended [70,72,79]. In order to generate neo-glycoproteins presenting variable numbers of sugar moieties, different concentrations of 9 or 10 were applied for the coupling process (Scheme II). Active charcoal delipidated BSA was dissolved in borate buffer (pH 9.0) and ratios of 9 or 10 ranging between 0.025 and 1.5 with respect to BSA lysine residues were adjusted. Thus, neo-glycoproteins 11a-k and 12a-k with variable numbers of LacNAc-LacNAc and LacDiNAc-LacNAc, respectively, were obtained after an incubation period of six days.
Trinitrobenzene sulfonic acid (TNBSA) was utilized for quantification of free lysine residues of BSA. Nucleophilic character of TNBSA as nitrobenzene derivative is well known for precise labeling of primary amines. The photometric measurement of the resulting N-trinitrophenylamine at 420 nm was demonstrated as the suitable wavelength for colorimetric quantification of free lysine residues [80]. In order to minimize consumption of valuable neo-glycoproteins experimental setup of the TNBSA assay was optimized. Lysine hydrochloride and BSA were utilized for highly accurate calibration ( Figure S2) of the TNBSA assay. Finally, the number of unmodified (free) lysine residues of the neo-glycoproteins 11a-k and 12a-k was determined by the optimized TNBSA assay as shown in Table 1. Molecular weights of the glycoproteins were then calculated according to the number of attached LacNAc-LacNAc or LacDiNAc-LacNAc glycans (amidated lysine residues). Increased ratios of 9 or 10 with respect to lysine residues of BSA led to an increasing modification density as more lysine residues were amidated.
For 11a-d and 12a-d the coupling efficiencies of squaric acid monoamide esters 9 and 10 were between 73.0% and 94.4%. However, at higher molar ratios coupling efficiency was only 32.2% for 9 and 30.6% for 10 yielding BSA based neo-glycoproteins 11i and 12i, respectively. Possible reasons for this observed limitation may be low accessibility of some of the lysine residues and hydrolysis of the squaric acid monoamide esters [72,79]. In summary, different amounts of 9 or 10 were attached to BSA reaching numbers for modified lysine residues between 2 and 29 per BSA molecule. Our results are in accordance with published data, where lactose conjugated BSA with maximal 30 glycation sites was obtained at lactose-BSA ratios of 20:1 [75]. In contrast, we reach this grade of modification already with a 1.5-fold molar excess of squaric acid monoamide esters (9 or 10) despite their approximately doubled molecular masses with respect to lactose. Ultrafiltration was used for removal of residual 9 or 10 to exclude effects on galectin binding assays. Pure neo-glycoproteins 11a-k (yields of 45.6%-68.2%) and 12a-k (yields of 61.7%-87.7%) were finally obtained ( Figure 1). To gain higher modification densities, a fed-batch instead of batch mode for chemical conjugation might be an alternative to overcome squaric acid monoamide ester hydrolysis. Neo-glycoproteins 11j-k and 12j-k featuring almost identical modification degrees were synthesized for subsequent comparison of LacNAc-LacNAc conjugated BSA (11j-k) and LacDiNAc-LacNAc conjugated BSA (12j-k) in galectin inhibition assay.
TNBSA assay results (Table 1) indicated consistent increase of theoretical molecular weights of neo-glycoproteins 11a-k and 12a-k. For further evaluation, apparent molecular weights of neo-glycoproteins were determined by reducing polyacrylamide gel electrophoresis (SDS-PAGE) shown in Figure 1. As molecular weight of BSA (66.4 kDa) could be confirmed via SDS-PAGE, gentle active charcoal treatment for removal of lipids does obviously not influence protein integrity. Moreover, Figure 1 provides direct evidence for molecular weight shift of neo-glycoproteins towards higher values. As expected, the mass differences of 11a-k or 12a-k compared to unmodified control (lane C) increase with higher initial concentrations of substances 9 or 10. Neo-glycoproteins 11a-k and 12a-k do not appear as distinct bands while smearing effect increases with size. Smearing appearance of glycosylated BSA can be due to the statistical character of the coupling reaction and thus presence of BSA molecules with different degrees of modification. Moreover, hydrophilic sugar chains are known to lack interaction with detergents like SDS [81]. Consequently, detergent induced development of linearized proteins is prevented proportionally to the amount of glycan moieties. Nevertheless, molecular weights of neo-glycoproteins are evaluated. For this purpose, retardation factors (Rf) are determined for each sample (11a-k and 12a-k). Molecular weight differences with regard to unmodified BSA were calculated and are summarized in Table 2.  Table 2. Molecular weight differences (¨MW) of neo-glycoproteins 11a-k and 12a-k compared to unmodified BSA. Values for retardation factor (Rf) were calculated based on electrophoretic mobility of samples resulting from SDS-PAGE analysis ( Figure 1). Measured Rf values were transformed to molecular weights by linear regression (y = mx + b; y-log MW; m-slope; x-Rf; b-y-intercept). Differences between the calculated molecular weights of neo-glycoproteins and those of unmodified BSA are compared with data determined by TNBSA assay. Coupling reactions of squaric acid monoamide esters 9 or 10 lead to mass increase of 0.943 kDa or of 0.984 kDa per amidated lysine residue. Results from biochemical TNBSA assay and SDS-PAGE provide direct evidence for molecular weight shift of neo-glycoproteins 11a-k and 12a-k. Both approaches are suitable analysis methods as approximately equal differences of molecular weight are identified. Most molecular weights indicated by SDS-PAGE are slightly smaller compared to those resulted from TNBSA assay. The analysis of diffuse bands leads to certain inaccuracy of Rf values.
In conclusion, BSA based neo-glycoproteins 11a-k and 12a-k were analyzed using different experimental setups. Maximal 29 of 60 lysine residues were demonstrated to be accessible by squaric acid monoamide esters by a tuned molar excess (1.5-fold) of those. Neo-glycoproteins appeared as smeary bands on reducing gel due to glycan proportions. However, shifts of molecular weights already indicated by TNBSA assay could be confirmed by SDS-PAGE. Clean BSA based neo-glycoproteins varying in the degree of presented oligosaccharide moieties were synthesized. Consequently, samples 11a-k and 12a-k were used for ELISA-type galectin binding tests.

Comparison of Galectin-3 and Galectin-1 Binding to Neo-Glycoproteins
The selective binding of human Gal-1 and Gal-3 to the synthesized neo-glycoproteins was investigated by ELISA-type assay using galectin concentration of 1 μM (Figure 2). Both galectins show no binding to unmodified BSA. Whereas binding signals of both galectins to the standard glycoprotein asialofetuin (ASF) are similar, Gal-3 shows higher binding to neo-glycoproteins compared to Gal-1. At low modification degrees of LacNAc-LacNAc conjugated BSA (11a-d), Gal-3 binds up to 10 times higher, whereas at higher glycan numbers (11e-i), Gal-3 reaches two-fold higher binding signals than Gal-1 ( Figure 2 and Table S1). The binding differences between Gal-1 and Gal-3 are proven to be significant with p < 0.002. Our data confirm previous studies that galactose terminated tetrasaccharides and oligosaccharides have higher selectivity for binding of Gal-3 compared to Gal-1 [52,66,82,83]. Most importantly, the difference between Gal-3 and Gal-1 binding is more distinct regarding LacDiNAc-LacNAc conjugated BSA ( Figure 2 and Table S1). The binding of Gal-3 to neo-glycoproteins 12a-d is up to 60-fold higher, and to BSA with higher glycan densities (12e-i) seven-fold higher when compared to Gal-1. The smaller difference between the binding potencies of both galectins to highly modified neo-glycoproteins is probably caused by reaching the maximal binding density of Gal-3 to neo-glycoproteins as well as increased binding of Gal-1 if multiple ligands are presented. Gal-1 is known to recognize terminal and not internal galactose [84,85], but in the present and earlier studies, weak binding to internal galactose occurred [66,86,87]. Due to the fact that Gal-1 does not bind to LacDiNAc [60,65,83,84], weak binding of Gal-1 to LacDiNAc-LacNAc conjugated BSA is based on recognizing the internal LacNAc unit since multiple ligands are presented.

Galectin-3 Binding to Neo-Glycoproteins at Different Galectin Concentrations
LacNAc and LacDiNAc epitopes are recognized by human Gal-3 with preferential binding to LacDiNAc [65]. Moreover, we identified in previous studies the LacNAc-LacNAc tetrasaccharide as the preferable Gal-3 ligand [66] pointing out that the glycans 4 and 5 are suitable candidates for developing multivalent neo-glycoproteins. BSA with varying numbers of 4 and 5 were analyzed for their binding characteristics as ligands for human Gal-3.
In Figure 3, binding signals of Gal-3 on immobilized neo-glycoproteins increase with higher modification densities. The results confirm Gal-3 binding to both types of BSA based neo-glycoproteins and no binding to unmodified BSA. However, significant increases of saturated binding signals are detected for BSA based neo-glycoproteins up to a modification degree of approximately 20 glycans per BSA, neo-glycoproteins 11a-g and 12a-f, respectively. Binding of Gal-3 hardly improves with increasing numbers of glycans above these modification degrees which may be due to approaching the maximal Gal-3 steric occupancy per molecule. The maximal binding signal of Gal-3 is for 11a-c and 12a-c in the same range but is slightly higher for 12d-i compared to 11d-i which may hint at better binding of Gal-3 to the LacDiNAc epitope.
The binding curves were further analyzed and used for calculation of the Kd values for Gal-3 binding (Figure 4 and Table S2). In general, LacDiNAc-LacNAc conjugated BSA shows better binding of Gal-3 compared to LacNAc-LacNAc conjugated BSA resulting in general lower Kd values for all modification degrees 12a-i. Significant differences between both types of neo-glycoproteins are obvious reflecting different selectivity of Gal-3 for the glycans. For binding of Gal-3 to 11a-i Kd values decrease more gradually reaching the lowest Kd value of 0.1 μM for 11i (Table S2). A significant drop of the Kd value is observed for the neo-glycoproteins 11f and 11i with about 19 and 29 LacNAc-LacNAc glycans, respectively. In comparison, Kd values are already significantly lowered for 12d-f with about nine, 14 and 21 LacDiNAc-LacNAc glycans, respectively, reaching a Kd value for Gal-3 in the nanomolar range (3 × 10 í8 M for 12f). These data emphasize high selectivity of Gal-3 for LacDiNAc-LacNAc glycans. In order to evaluate multivalent effects, the binding signal per glycan ligand was calculated for all applied Gal-3 concentrations and was related to 11d and 12d, respectively ( Figure 5 and Table S3). For 11a-h and 12a-c, binding signals are only detectable for Gal-3 concentrations from 0.05 μM ( Figure S3). In contrast, for LacDiNAc-LacNAc conjugated BSA modification densities of at least nine glycans per BSA (12d) are sufficient to reach binding signals of Gal-3 at concentrations even below 0.05 μM. Thus, binding signals per glycan for 11d and 12d are used as the benchmark for calculating binding potencies. Enhancement of binding by multivalent presentation is not observed for LacNAc-LacNAc conjugated BSA. Binding signals of Gal-3 to 11a-i in relation to one binding site reach nearly the same level, in the used concentration range. In contrast, Gal-3 shows enhanced avidity for binding to neo-glycoproteins 12e-i at concentrations below 0.05 μM with highest multivalent effects for 0.005 μM. Up to 100 times higher binding in relation to one binding site is observed for modification densities above 20 glycans per BSA. However, at Gal-3 concentrations above 0.1 μM, the binding signals in relation to one glycan are for higher modified BSA even lower than for weak glycosylated BSA. This is possibly due to hindered accessibility of all binding sites in the presence of Gal-3 excess. Consequently, multivalent ligand presentation on BSA can induce multivalent effects depending on the ligand, the valency and the galectin concentration. Here, only LacDiNAc-LacNAc conjugated BSA with modification degrees above eight glycans per BSA enhances the avidity of Gal-3 at concentrations below 0.05 μM. Multivalent effects were also found for Gal-3 using N-glycan modified human serum albumin synthesized by click-chemistry [88]. The multivalent neo-glycoprotein carrying eight biantennary N-glycans showed a 30-fold increase in affinity per glycan unit in comparison to the monovalent reference. For Gal-1 it was reported that BSA based neo-glycoproteins with up to eight conjugated glycans per BSA did not show enhanced avidity [84]. The monovalent recognition of the glycans by Gal-1 was explained by the high molecular weight of BSA and therefore the long distance between the glycans.
Mostly, Gal-3 binding was investigated for multivalent ligands based on chemical scaffolds like calixarenes, dendrimers or quantum dots [58,59,[89][90][91]. Comparison of our results with findings for these synthetic scaffolds is difficult but can be drawn with regard to the glycan type and number. With multivalent LacNAc modified quantum-dots presenting 108 glycans, a 92-fold higher affinity towards Gal-3 in comparison to soluble LacNAc was shown using surface plasmon resonance spectroscopy [58]. With respect to one ligand, we gain higher enhancement of Gal-3 binding to BSA based neo-glycoproteins modified with a much smaller number of glycans, namely at least 14 LacDiNAc-LacNAc glycans (12f). Neo-glycoproteins enhancing the binding potency of low concentrated Gal-3 by a factor of 100 may be promising inhibitors for anti-cancer therapy. The expression of Gal-3 is upregulated in different tumor cells and induces tumor progression and metastasis [42][43][44][45][46][47][48]92]. The serum level of Gal-3 in patients with cancer is likewise significantly increased compared with healthy individuals [93][94][95]. Therefore, interfering with Gal-3 by applying effective inhibitors can reduce the angiogenic and metastatic potential of tumor cells [96][97][98][99]. Multivalent systems, e.g., based on dendrimer scaffolds, are of high interest as they induce cluster formation and exhibit enhanced inhibitory potencies [53,91,100]. In the present study, we have generated neo-glycoproteins which are effective for multivalent binding at high modification degrees (12f-i) with the Gal-3 selective epitope LacDiNAc-LacNAc. The highest multivalent effects occur at Gal-3 concentrations of 5-10 nM (130-260 ng/mL) which is in the range of Gal-3 level in human serum of cancer patients (mean value = 320 ng/mL). This is about five times higher than in healthy individuals (mean value = 62 ng/mL) [95]. Hence, in roughly that concentration range, Gal-3 inhibition should reduce tumor aggressiveness making these LacDiNAc-LacNAc conjugated BSA promising candidates for anti-cancer therapy. In vivo applications are possible, because BSA based neo-glycoproteins were already applied to mice, rats and rabbits [101,102].
In conclusion, BSA-based neo-glycoproteins are high-affinity ligands for Gal-3 with higher binding for increasing modification densities. At Gal-3 concentrations below 0.05 μM, up to 100-fold enhanced avidity is shown for neo-glycoproteins presenting more than 14 LacDiNAc-LacNAc glycans.

Inhibitory Potency of Neo-Glycoproteins
For possible biomedical application of these neo-glycoproteins, their ability to inhibit Gal-3 binding has to be proven. Thus, competitive inhibition assay was performed to investigate Gal-3 binding to immobilized ASF in the presence of neo-glycoproteins 11j, 11k, 11g and 12j, 12k, 12g, respectively. To compare inhibitory potencies with soluble oligosaccharides, 4 and 5 were also used in this assay. Plotting residual Gal-3 binding signals against inhibitor concentration provides the sigmoidal curves depicted in Figure 6 and are used for IC50 calculation (Figure 7 and Table 3).
Inhibition of Gal-3 binding by neo-glycoproteins is shown by a shift to lower inhibitor concentrations indicating higher inhibition potencies of ligands presented multivalently on a protein scaffold. IC50 values of 4 and 5 are significantly higher than those reached by the neo-glycoproteins (p < 0.002). With increasing glycosylation densities of the neo-glycoproteins, stronger inhibition of Gal-3 binding to ASF is observed. BSA modified with 18 and 24 binding sites (11k, 11g and 12k, 12g) are obviously stronger inhibitors than BSA presenting 7.5 glycans (11j and 12j) though only small differences exist between both higher modification degrees. Increased modification of BSA results in higher inhibition, but does not strengthen the inhibition proportionally. When comparing the tetrasaccharides LacDiNAc-LacNAc (5) and LacNAc-LAcNAc (4), 5 reveals a 1.5-fold higher inhibition of Gal-3 binding. This is based on better binding to LacDiNAc terminated glycan as already demonstrated [65].  , 12k and 12g) is shown. Unmodified BSA has no influence on the galectin-3 binding. Inhibition curves were fitted using Sigma Plot software.  In this context, the evaluation of multivalent effects is relevant. Therefore, we chose neo-glycoproteins of low (11j and 12j), medium (11k and 12k) and high (11g and 12g) modification degree used in this competitive inhibition assay and calculated relative potencies per glycan ( Table 3). The improvement of the inhibitory potency by multivalent presentation is overall more pronounced for LacNAc-LacNAc conjugated BSA than for LacDiNAc-LacNAc conjugated BSA since the IC50 value for 4 is higher than for 5. Thereby, the inhibitory potency can be more enhanced. The general presentation of glycans on BSA has little influence on the inhibitory potency indicated by only up to two times improved relative potency in relation to one binding site compared to 4 and 5. Here, we can conclude that multivalent effects are not relevant for low glycosylation degrees of about 7.5 glycans per BSA (11j, 12j). However, compounds 11j and 12j improve the inhibition of Gal-3 binding to ASF by a factor of 15 and 6.5, respectively. For medium and high modification densities (11k, 11g and 12k, 12g), the inhibitory potency is drastically enhanced by factors of 130-370 related to the corresponding tetrasaccharide. In relation to one binding site, inhibitory potency is improved by 11k four times and by 11g 7.5 times in comparison to 11j, and almost twice by 11g compared to 11k. Regarding LacDiNAc-LacNAc conjugation, inhibition potency is enhanced by 12k by a factor of 7 per glycan compared to 12j, whereas with 12g scarcely any further improvement is gained. Higher inhibitory potency of LacDiNAc-LacNAc conjugated BSA over LacNAc-LacNAc conjugated BSA is not observed. In contrast, we see binding enhancement of Gal-3 to LacDiNAc-LacNAc presenting neo-glycoproteins with at least 14 glycans (12e-i) at Gal-3 concentrations below 0.05 μM (Figure 5b). The inhibition assay requires higher Gal-3 concentration (5 μM) due to weaker binding to ASF than to neo-glycoproteins. Therefore, equal inhibition potencies of both types of modifications are obtained.
However, we determined IC50 values below 100 nM for BSA based neo-glycoproteins modified with at least 18 glycans. Reaching such high inhibitory potencies is caused by binding enhancement due to protein chelation in the presence of multivalent ligands [11]. As Gal-3 forms oligomers involving multivalent ligands [54,55] it is prone to be a candidate for binding enhancement by multivalent effects. In contrast, Gal-3 binding to ASF was characterized by high affinity to the first LacNAc epitope of the triantennary N-glycan with increasing negative cooperativity for subsequent binding events [56]. However, lactose based dendrimers showed strong multivalent effects in a solid phase inhibition assay with ASF [103]. Gal-3 binding to monovalent lactose dendrimer was already about 10 times higher than to free lactose indicating interactions with the dendrimer backbone [103,104], contrary to the neo-glycoproteins in the present study. Moreover, calixarenes multivalently presenting lactose increased the inhibitory potency towards Gal-3 binding to ASF by a factor of 12.5 reaching an IC50 of 200 μM [89]. Multivalent LacNAc bearing calix [4/6]arenes with 2/3' substitutions in the LacNAc core showed IC50 values up to 0.15 μM for Gal-3 binding to ASF, which were about 1500 times lower compared to monovalent species [90]. Hence, multivalent presented LacNAc increases Gal-3 binding more than multivalent lactose. In the present study, multivalent presentation of tetrasaccharides on BSA yields lower IC50 values up to 0.04 μM indicating a large benefit using tetrasaccharides instead of disaccharides.
In conclusion, we could prove inhibitory potency for Gal-3 binding to ASF applying the presented neo-glycoproteins. For modification densities of at least 18 glycans per BSA, up to 15 times higher potencies in relation to one binding site are recorded. We suggest that IC50 values can even be reduced in assays using lower Gal-3 concentration, e.g., in vivo assays, as we could show up to 100 times increased Gal-3 binding at concentrations below 0.05 μM ( Figure 5).

Production of Recombinant Enzymes
Enzymes were expressed and purified as described previously [65,67,76,105]. Briefly, the fusion protein of human ȕ1,4-galactosyltransferase-1 (His6-Propeptide-catȕ4GalT-1, ȕ4GalT) and the Y284L mutant (ȕ4GalTY284L) were expressed in E. coli Shuffle T7 Express (DE3) (NEB, Frankfurt/Main, Germany). The ȕ1,3-N-acetylglucosaminyltransferase from Helicobacter pylori (ȕ3GlcNAcT) and UDP-Glc 4ƍ-epimerase from Campylobacter jejuni, both fused to maltose binding protein, were expressed in E. coli BL21 (DE3) (Novagen/Merck, Darmstadt, Germany). TB medium with 0.5 mM IPTG was used for induction of expression. After 24 h, cells were harvested by centrifugation and appropriate cell mass was sonicated two times for 30 s for follow-up affinity chromatography. Recombinant MBP-tagged proteins were purified by amylose affinity chromatography using MBPTrap TM HP 5 mL column (GE Healthcare, Munich, Germany) and His6-tagged proteins by immobilized metal-ion affinity chromatography (IMAC) using HisTrap TM HP 5 mL column (GE Healthcare). Purification was carried out according to manufacturer's instructions. After column equilibration bacterial crude extracts were loaded onto the column at a flow rate of 1 mL/min and washed until baseline signal was reached. Target proteins were eluted by application of elution buffer containing 10 mM maltose or 500 mM imidazole. The buffer for ȕ4GalTY284L was subsequently changed to 0.1 M MOPS buffer (pH 6.8) including 20% (v/v) glycerol.

Synthesis of Neo-Glycoproteins Based on BSA
Prior to BSA modification, active charcoal delipidation was carried out for removal of lipid contaminations as described [106], because commercial preparations of BSA may contain fatty compounds. Coupling reaction to BSA (Carl Roth, Karlsruhe, Germany) was carried out under slightly basic conditions using sodium tetraborate buffer (50 mM, pH 9.0). Compounds 9 and 10 dissolved in defined volume of water were applied in varying molar ratios compared to lysine residues of BSA (from 1:60 to 1.5:1). Reaction mixtures were incubated for up to 136 h at room temperature while gently shaking. The products 11a-k and 12a-k were purified via filtration to remove unbound glycans and analyzed regarding BSA lysine residue modification as described below.

Analysis of Neo-Glycoproteins
Determination of free lysine residues (İ-amino groups) in proteins was carried out using 2,4,6-trinitrobenzene sulfonic acid (TNBSA, Sigma-Aldrich) as detecting agent. The TNBSA assay was performed as described [80] but optimized for our application. Protein samples were diluted in 50 mM sodium tetraborate buffer (pH 9.0) to yield a concentration of 12.5 μM and 50 μL were mixed with equal volume of 7.5 mM TNBSA. After incubation at room temperature for 15 min, absorbance at 420 nm was measured in microplate reader (SPECTRAmax Plus, Molecular Devices, Ismaning, Germany). Standard curves were generated using lysine hydrochloride and crude BSA as standard. Protein amounts were measured by Bradford assay using calibration with unmodified BSA.
The modification of BSA with glycans resulted in increasing molecular weights that were verified by SDS-PAGE. SDS-PAGE was done using 8% gels and constant current of 25 mA.

Expression and Purification of Recombinant Galectins
Human His6Gal-1C2S and human His6Gal-3 were expressed and purified as described elsewhere [66,107] except that we used E. coli Rosetta (DE3) pLysS (Novagen/Merck, Darmstadt, Germany) for expression. The purification via IMAC was performed as mentioned above. His6Gal-1C2S and His6Gal-3 were stored at 4 °C in phosphate buffered saline containing 2 mM EDTA (EPBS, pH 7.5) including 10% (v/v) glycerol for His6Gal-1C2S. Protein amounts were measured by Bradford assay using calibration with BSA.

Galectin Binding Assays on Neo-Glycoproteins
The potential of neo-glycoproteins acting as ligands for galectins was evaluated in an ELISA-type assay [66]. The neo-glycoproteins 11a-i and 12a-i as well as unmodified BSA (50 μL of 0.1 μM protein in PBS pH 7.5) were immobilized in microtiter plates (MaxiSorp, Nunc, Wiesbaden, Germany) overnight. Residual binding sites were blocked with 2% delipidated BSA in PBS followed by 1 h incubation of galectin in EPBS. Anti-His6-peroxidase (Roche, Mannheim, Germany, 1:2000 in PBS) was used to detect bound galectin that subsequently converted OPD substrate (o-phenylenediamine, Dako, Hamburg, Germany) with read-out at 492 nm. Between the different incubation steps wells were washed three times with PBS-Tween (0.05% (v/v)). Measured data were fitted and Kd values calculated using Sigma Plot (Systat software GmbH, Erkrath, Germany). Significance of the data was investigated by independent two sample t-test with p < 0.002.

Inhibition of Galectin Binding with Neo-Glycoproteins
Neo-glycoproteins were used as inhibitors of Gal-3 to ASF binding in competitive inhibition assays performed as described [66]. The standard glycoprotein ASF (Sigma-Aldrich) was immobilized in microtiter plates as described above for neo-glycoproteins. After blocking, 5 μM Gal-3 and varying concentration of 11j, 11k, 11g and 12j, 12k, 12g were simultaneously incubated for 1 h followed by detection of bound Gal-3 as described above. Measured data were fitted and IC50 values calculated using Sigma Plot software (Systat software GmbH).

Conclusions
We here demonstrate the efficient synthesis of high-affinity ligands for human Gal-3. We show that LacNAc-LacNAc or LacDiNAc-LacNAc conjugation to BSA is tunable using squaric acid diethyl ester as linker. We report here for the first time on selective Gal-3 binding to multivalent neo-glycoproteins revealing and quantifying the influence of defined multivalency. The highest binding enhancement by a factor of about 100 was observed with LacDiNAc-LacNAc conjugated BSA modified with at least 21 glycans at a Gal-3 concentration of 5 nM. The efficient and selective binding of such neo-glycoproteins at serum level concentrations of Gal-3 may have high impact in anti-cancer therapy. The here presented multivalent neo-glycoproteins can be further loaded with cytotoxic compounds and labeled with fluorescent dyes enabling tumor diagnostic by molecular imaging and tumor therapy. In summary, our tailor-made neo-glycoproteins are suitable candidates for targeting Gal-3 in cancer related biomedical research.