Structure-guided inhibitors of hβ4GalT7

Among glycosaminoglycan (GAG) biosynthetic enzymes, the human β1,4galactosyltransferase 7 (hβ4GalT7) is characterized by its unique capacity to take over xyloside derivatives linked to a hydrophobic aglycone as substrates and/or inhibitors. This glycosyltransferase is thus a prime target for the development of regulators of GAG synthesis in therapeutics. Here, we report the structureguided design of hβ4GalT7 inhibitors. By combining molecular modeling, in vitro mutagenesis and kinetic measurements, and in cellulo analysis of GAG anabolism and decorin glycosylation, we mapped the organization of the acceptor binding pocket, in complex with 4methylumbelliferone-xylose (4-MUX) as prototype substrate. We show that its organization is governed, on one side, by three tyrosine residues, Y194, Y196 and Y199, which create a hydrophobic environment and provide stacking interactions with both xylopyranoside and aglycone rings. On the opposite side, a hydrogen-bond network is established between the charged amino acids D228, D229 and R226, and the hydroxyl groups of xylose. We identified two key structural features i.e. the strategic position of Y194 forming stacking interactions with the aglycone, and the hydrogen bond between H195 nitrogen-backbone and the carbonyl group of the coumarinyl molecule to develop a tight binder of hβ4GalT7. This led to the synthesis of 4-deoxy-4-fluoro-xylose linked to 4-MU that inhibited hβ4GalT7 activity in vitro with a Ki ten-times lower than the Km value and efficiently impaired GAG synthesis in a cell assay. This study provides a valuable probe for the investigation of GAG biology and opens avenues towards the development of bioactive compounds to correct GAG synthesis disorders implicated in different types of malignancies. Glycosaminoglycans (GAGs) are linear heteropolysaccharide chains covalently attached to the core protein of a variety of proteoglycans (PGs). Owing to their high structural diversity and their http://www.jbc.org/cgi/doi/10.1074/jbc.M114.628123 The latest version is at JBC Papers in Press. Published on January 7, 2015 as Manuscript M114.628123 Copyright 2015 by The American Society for Biochemistry and Molecular Biology, Inc. by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 2 anionic characteristics, GAGs interact with a network of cellular and extracellular mediators including cytokines and chemokines, enzymes and enzyme inhibitors, matrix proteins and membrane receptors (1). There is currently great emphasis on the crucial roles of GAGs in numerous physiological events including cell differentiation, proliferation and migration (2), and its pathological aspects, such as tumor formation, progression and metastasis (3). Furthermore, since PGs are ubiquitously expressed in extracellular matrices and on cell surfaces of virtually every tissue, they are also involved in the normal and pathological functions of the cardiovascular and osteo-articular system (4), in amyloid disorders (5) and in axonal deand regeneration (6). GAG biosynthesis is initiated by the formation of a tetrasaccharide linkage region (GlcAβ1-3Galβ1-3Galβ1-4Xylβ1-O) covalently linked to serine residues of the PG core protein (7). This tetrasaccharide acts as a primer for the elongation of major GAG chains i.e. chondroitin-/dermatan-sulfate (CS/DS) or heparin/heparan-sulfate (HS), which polymerization involves the coordinated activities of CS-synthases (CSS) and HS-synthases (exostosins, EXT), respectively (8, 9). Mature GAG chains are finally produced by the modifications of their constitutive disaccharide units catalyzed by epimerases and sulfotransferases which considerably increase their structural and functional diversity (10, 11). The human xylosylprotein β1,4-galactosyl transferase, (EC 2.4.1.1337, hβ4GalT7) catalyses the transfer of the first Gal residue of the tetrasaccharide linkage from the activated sugar UDP-galactose (UDP-Gal) onto Xyl residues attached to the PG core protein (12). Since all GAGs share the same stem core tetrasaccharide, β4GalT7 is a central enzyme in GAG biosynthesis. Indeed, hβ4GalT7 mutations have been associated with a rare genetic condition, the progeroid form of EhlersDanlos syndrome (EDS), a group of connective tissue disorders, characterized by a major deficiency in PG synthesis. As a consequence of GAG defect, EDS patients exhibit motor development delay, and musculoskeletal malformations, hypermobile joints and wound healing defaults (13). Patients gene sequencing revealed the presence of missense mutations leading to L206P, A186D (14, 15) and R270C substitutions (16) in the catalytic domain, resulting in a partially or totally inactive enzyme. Recently, we showed that R270C replacement reduced affinity towards xyloside acceptor and strongly affected GAG chains formation in β4GalT7-deficient CHOpgsB-618 cells (17). There is currently no effective therapy for treating EDS patients. Interestingly, the biosynthesis of GAGs can be manipulated by simple xylosides carrying a hydrophobic aglycone, which act as substrates and/or inhibitors of hβ4GalT7. Xyloside analogs have been shown to efficiently induce GAG synthesis bypassing the natural Xyl-substituted core protein of PGs for several decades (18, 19). The xyloside-primed GAG chains are usually excreted and show interesting biological functions such as activation of fibroblast growth factor (FGF)signaling (20, 21), antithrombotic (22), tissue regenerating (23), anti-angiogenic (24) and antiproliferative properties (25, 26). In addition, several groups have synthesized series of xyloside analogs as potential inhibitors of GAG synthesis. Such compounds would represent highly valuable chemical biology tools to probe the functions of GAGs in cell systems and model organisms and as a starting point towards the development of pharmaceuticals, in particular anti-tumor agents. Recently, Garud et al. (27) and Tsuzuki et al. (28) used click chemistry to generate libraries of 4deoxy-4-fluoro-triazole analogs comprising a set of hydrophobic molecules appended to the anomeric carbon of the xyloside. Siegbahn et al. (29, 30) developed a collection of naphthyland benzylxylosides substituted on different positions of the Xyl moiety. These studies led to the discovery of promising xyloside-derived inhibitors of GAG synthesis when screened in cell models. However, until recently, the development of substrates and inhibitors of β4GalT7 has been mostly limited to the synthesis of libraries of analog compounds and their testing in cell assays. Towards the rational design of hβ4GalT7 inhibitors, we have been involved in structure-activity relationship studies of the recombinant human enzyme for several years and identified critical active site amino acids implicated in catalysis and/or substrate binding (17, 31, 32). We previously investigated the importance of conserved DVD and FWGWRGEDDE motifs in the organization of the catalytic domain. Our data have highlighted the crucial role of W224 in substrate recognition and suggested a catalytic role for D228 (31). These findings were in accordance with the structural data from the recently solved crystal structure of the catalytic domain of Drosophila melanogaster dβ4GalT7 (33) and that of the human enzyme (34). In the current study, we developed a structureguided approach for the design of xyloside inhibitors of hβ4GalT7 that were tested on its galactosyltransferase activity in vitro and on GAG biosynthesis in cell assays. We explored the organization of the acceptor binding pocket, specifically probing the functional and structural contribution of a set of residues located in the vicinity of the catalytic center, and highlighted the by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 3 crucial role of three tyrosine residues i.e. Y194, Y196 and Y199 in the architecture of the acceptor substrate binding site and the creation of a hydrophobic environment. Based on these and previous findings, we synthesized compounds that incorporate critical structural elements both on the xylopyranoside and on the aglycone moieties to tightly bind the acceptor site of hβ4GalT7. This work revealed that the 4-deoxy-4-fluoro-Xyl linked to 4-methylumbelliferone (4-MU) strongly inhibited hβ4GalT7 activity in vitro and efficiently impaired GAG synthesis in a cell context. Such compound will be a valuable tool for the exploration of GAG and PG synthesis and opens avenues towards the development of bioactive oligosaccharide structures for GAG biosynthesis regulation in a number of diseases implicating disorders of GAG synthesis. EXPERIMENTAL PROCEDURES Chemicals and Reagents – 4-Methylumbelliferyl-βD-xylopyranoside (4-MUX), UDP-α-D-Gal (UDPGal) and anti-goat IgG (whole molecule)-peroxidase conjugated antibody were provided from Sigma Aldrich. Anti-myc antibodies were from Invitrogen and anti-mouse IgG-peroxidase antibodies were purchased from Cell Signaling whereas anti-decorin antibodies were from R&D systems. Na2[S]SO4 was from PerkinElmer Life Sciences. Cell culture medium was purchased from Life Technologies and restriction enzymes, T4 DNA ligase and peptide Nglycosidase F (PNGase F) from New England Biolabs. The eukaryotic expression vector pcDNA3.1(+) and competent One Shot Top 10 Escherichia coli (E. coli) cells were provided from Invitrogen and the bacterial expression vector pET41a(+) and E. coli BL21(DE3) cells were from Novagen-EMD Chemicals. The QuikChange sitedirected mutagenesis kit was from Stratagene and the transfection agent ExGen 500 from Euromedex. Chemical Synthesis – Naphthyl-4-deoxy-β-Dxylopyranoside (4H-Xyl-NP) (29) was obtained after protection of the 2,3 position of naphthyl-β-Dxylopyranoside by isopropylidene acetal followed by radical deoxygenation and deprotection. 4Methylumbelliferyl-4-deoxy-β-D-xylopyranoside (4H-Xyl-MU) and 4-methylumbelliferyl-4-fluoro-βD-xylopyranoside (4F-Xyl-MU) were synthesized from the reported starting material 4methylumbelliferyl-2,3-di-O-benzoyl-β-Dxylopyranoside (35) by radical deoxygenation or stereocontrolled 4-fluorination followed by final deprotection (data not shown). Molecular modeling of hβ4GalT7 active site in the presence of 4-MUX and UDP-Gal – The crystal structure of hβ4GalT7 bound to UDP and to the manganese ion (PDB ID: 4IRQ) was used as template (34). The crystal structure of dβ4GalT7 (PDB ID: 4M4K), an inactive mutant (D211N) of dβ4GalT7 in complex with UDP-Gal and xylobiose was superposed to the human enzyme structure, which was straightforward considering their strong sequence similarity (58% overall identity). Due to crystallization conditions, a Tris molecule is bound within the active-site of the hβ4GalT7. When retrieved, it frees space within the cavity that can thus accommodate the Gal moiety. The coordinates of the Gal molecule from the dβ4GalT7 complex were merged to the UDP moiety of hβ4GalT7. This did not generate any steric clash within the active site. The resulting complex was then prepared using the Protein Preparation Wizard tool of the Schrödinger Suite (http://www.schrodinger.com, Schrödinger LLC, New York), with default settings (36). All the water molecules were retrieved, except the one that coordinates the manganese ion. The hydrogen atoms were added to the protein and the ligand, ascribing a pH of 7.0. The histidine residues were treated as neutral. The selection of the histidine enantiomers and the orientation of the asparagine and glutamine side-chains were performed so as to maximize the hydrogen bond network. The partial atomic charges derived from the OPLS-2005 force field (37) were assigned to all ligand and protein atoms. Finally, an all-atom energy minimization with a 0.3 Å heavy-atom root-mean-square deviation (RMSD) criteria for termination was performed using the Impref module of Impact and OPLS-2005 (38). The 4-MUX ligand was prepared using the ligprep module (Schrödinger Release 2014-22014). The docking program Glide was used in Standard Precision mode, with OPLS2005, to run rigid-receptor docking calculations (39, 40). The shape and physico-chemical properties of the binding site were mapped onto a cubic grid with dimensions of 20 Å centered on the xylobiose. During the docking calculations, the parameters for van der Waals radii were scaled by 0.80 for receptor atoms with partial charges less than 0.15e. Ring conformational sampling was not allowed to maintain the C1 conformation of the Xyl ring, and no constraint was introduced. A maximum of 100 poses were retained and ranked according to the GlideScore scoring function. The best-docked pose of the 4-MUX ligand showed a RMSD on the Xyl ring heavy atoms of 0.5 Å with the crystallographic xylobiose ligand, thus validating the docking protocol able to recover the position of this moiety. Expression vector construction – The hβ4GalT7 sequence (GenBank nucleotide sequence accession number NM_007255) was cloned by PCR amplification from a placenta cDNA library (Clontech), as previously described (41). For bacterial expression, a truncated form of hβ4GalT7 was expressed as a fusion protein with glutathioneby gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 4 S-transferase (GST). The sequence lacking the codons of the first 60 N-terminal amino acids was amplified from the full-length cDNA and subcloned into NcoI and NotI sites sites of pET-41a(+) to produce the plasmid pET-β4GalT7 (31). For the heterologous expression of hβ4GalT7 in mammalian cell lines, the full-length cDNA sequence was modified by PCR at the 5' end to include a KpnI site and a Kozak consensus sequence, and at the 3' end to include a sequence encoding a myc tag and an XbaI site to be then subcloned into the KpnI-XbaI sites of the eukaryotic expression vector pcDNA3.1(+) to produce pcDNA-β4GalT7 as previously described (31). Mutations were constructed using QuikChange site-directed mutagenesis kit, employing pcDNA-β4GalT7 or pET-β4GalT7 as template. Sense and antisense primers are listed in supplemental Table 1. Mutants were systematically checked by double strand sequencing. The human decorin cDNA sequence (GenBank accession number NM_001920.3) was cloned by PCR amplification from a placenta cDNA library (Clontech). For heterologous expression in eukaryotic cells, the full-length cDNA sequence was modified by PCR to include an AflII site, a Kozak consensus sequence at the 5' end, a sequence encoding a His5 tag, and an XhoI site at the 3' end. This sequence was subcloned into pcDNA3.1(+) to produce pcDNA-decorinHis as previously described (31). Expression and purification of the soluble form of hβ4GalT7 – A single colony of E. coli BL21(DE3) cells transformed with the pET-β4GalT7 plasmid was cultured overnight at 37 °C in a Luria broth (LB) medium containing 50 μg/mL kanamycin. The overnight culture was transferred into fresh LB medium (1:100 dilution), supplemented with 50 μg/mL kanamycin and incubated at 37 °C until the OD600 value reached 0.6-0.8. Expression of hβ4GalT7 was induced by addition of 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) to the cell suspension, that was then incubated overnight at 20 °C under continuous shaking (200 rpm). The bacterial cells were then harvested by centrifugation at 7,000 × g for 10 min at 4 °C. The pellet was resuspended in Lysis buffer (50 mM sodium phosphate, 1 mM phenymethylsulfofluoride, 1 mM EDTA, and 5% (v/v) glycerol, pH 7.4) supplemented with protease inhibitor cocktail tablets (1 tablet /12 mL; Roche Diagnostics) and Benzonase Nuclease (250 units/10 mL, Sigma Aldrich). The suspended cells were then sonicated for 8 cycles of 30 sec, at 30% power (Badelin Sonoplus GM70) with a 20 sec-interval on ice between each cycle. Soluble proteins were collected from the supernatant after centrifugation for 25 min at 12,000 × g and clarification by filtration (0.2 μM Supor Membrane; PALL-Life Science). 10 mL of clarified extracts were applied onto a 1 mL a Glutathione Sepharose High Performance column (GSTrap HP; GE Healthcare) connected to an AKTA prime plus instrument (GE Healthcare). Protein were eluted as 1 mL-fractions using 50 mM Tris-HCl, pH 8.0 containing 10 mM reduced glutathione buffer. Protein purity of the eluted fractions was evaluated by 12% (w/v) SDS-PAGE analysis, followed by staining with Coomassie Brillant Blue. Fractions containing the pure protein were used to determine the kinetic parameters of the enzyme. The same procedure was used for purification of the mutants. Protein concentration was measured using Quant-iT assay kit and Qubit spectrofluorimeter. Determination of the in vitro kinetic parameters of hβ4GalT7 – The kinetic parameters kcat and Km towards 4-MUX and UDP-Gal were determined as described (31). Briefly, 0.2 μg of purified wild-type or mutated GST-hβ4GalT7 were incubated for 30 min at 37 °C in a 100 mM sodium cacodylate buffer pH 7.0, 10 mM MnCl2, with concentrations from 0 to 5 mM 4-MUX in the presence of fixed 1 mM UDP-Gal to determine the apparent Km towards 4MUX, and with concentrations from 0 to 5 mM UDP-Gal in the presence of fixed 5 mM 4-MUX to determine the apparent Km towards UDP-Gal. The incubation mixture was then centrifuged at 10,000 × g for 10 min at 4 °C. The supernatant was analyzed by high performance liquid chromatography (HPLC) with a reverse phase C18 column (xBridge, 4.6 x 150 mm, 5 μm, Waters) using a Waters equipment (Alliance Waters e2695) coupled to a UV detector (Shimadzu SPD-10A). Kinetic parameters were determined by nonlinear least squares regression analysis of the data fitted to MichaelisMenten rate equation using the curve-fitter program of Sigmaplot 9.0 (Erkraft, Germany). In vitro competition assays of hβ4GalT7 activity by C4-modified xylosides – The in vitro inhibition assays of the wild-type GST-hβ4GalT7 were carried out using 0.2 μg of purified protein incubated for 30 min at 37 °C in a 100 mM sodium cacodylate buffer pH 7.0, 10 mM MnCl2, with 0.5 mM 4-MUX and 1 mM UDP-Gal, in the presence of concentrations from 0 to 5 mM of either 4H-Xyl-NP, 4H-Xyl-MU, or 4F-Xyl-MU. Quantification of the reaction product was carried out by HPLC, as described above. The enzyme activities were reported as function of the logarithmic values of inhibitor concentration. IC50 values were determined by fitting the experimental dose-response curves using the curve-fitter program of Sigmaplot 9.0 (Erkraft, Germany). Ki values were calculated from IC50 values according to the Cheng-Prusoff's equation (42, 43). by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 5 In cellulo analysis of GAG chains biosynthesis by Na2[SO4] incorporation – GAG chains biosynthesis using 4-MUX as primer substrate was determined with CHOpgsB-618 cells (American Type Culture Collection). Cells were cultured in Dulbecco’s modified Eagle’s medium-F12 (DMEM/F12) (1:1), supplemented with 10% fetal bovine serum (FBS, Dutscher), penicillin (100 units/ml)/ streptomycin (100 mg/ml), and 1 mM glutamine, then transfected with the wild-type or the mutant pcDNA-β4GalT7-myc plasmid or with the empty pcDNA3.1 vector at 70% cell confluency. Transfected cells were then incubated in a low sulfate medium (Fisher) supplemented with 10 μCi/mL Na2[SO4] in the presence of 0.5 or 10 μM 4-MUX for 16 h. For GAG chains isolation 1 mL culture medium was applied to a G-50 column (GE Healthcare) to separate radiolabeled GAG chains from the non-incorporated Na2[SO4] and radiolabeling was quantified by scintillation counting. In parallel, hβ4GalT7 expression level was checked by western blotting using a primary anti-myc (1/5000) and a secondary anti-mouse antibody (1/10000). To test the inhibitory potency of C4-modified xylosides, the molecules were added at 0 to 100 μM concentration together with 4-MUX (5 μM) for 16 h prior to isolation and quantification of radiolabeled GAGs. To test the cytotoxicity of xyloside inhibitors in CHOpgsB-618 cells expressing the wild-type hβ4GalT7, cells were seeded at 150,000 cells/well in 12-well plates, and incubated for 48 h at 37 oC in the presence of 0 to 400 μM of inhibitor, or 4-MUX as a control. The ratio of viable cells upon the total number of cells was determined using the cell counter TC20 (BioRad) in the presence of a vital marker (Trypan blue). In cellulo analysis of decorin core protein glycosylation – CHOpgs-B618 cells stably transfected with pcDNA-decorinHis encoding the human decorin core protein (31, 44) were transiently transfected with pcDNA3.1 or with recombinant vector encoding either the wild-type or mutated hβ4GalT7-myc as described above. 48 Hours following transfection, the cell medium was collected, concentrated by centrifugation at 4 oC for 15 min at 3000 × g, using the Amicon Ultracell 30 MWCO concentrating system (Merck Millipore, Germany) and submitted to SDS-PAGE (25 μg protein per well). The glycosylation level of the decorin core protein was monitored by immunoblot using a 1/5,000 dilution of primary polyclonal antihuman decorin antibody (VWR) and a 1/10,000 dilution of secondary anti-goat antibody coupled to horseradish peroxidase (Sigma Aldrich), then quantified using ImageJ software. Briefly, the level of decorin glycosylation was expressed as relative band intensity by normalizing the band intensity value for the glycosylated form upon the total intensity value for the bands corresponding to the glycosylated and non-glycosylated forms of decorin core protein. The expression level of the decorin core protein in pcDNA3.1-transfected cells was used as negative control and served as a loading control. On the other hand, the level of glycosylated decorin in cells expressing the wild-type hβ4GalT7 was used as positive control for decorin glycosylation. RESULTS Molecular modeling of the hβ4GalT7 acceptor binding site – In the present study, we aimed to identify amino acids important for the structural organization of the hβ4GalT7 acceptor substrate binding site. We took advantage of the recent crystal structure of hβ4GalT7 in complex with UDP (34) to build a molecular model of this enzyme in complex with both the sugar donor UDP-Gal and the acceptor 4-MUX (Fig. 1A). The modeled structure is in a closed conformation, considered to be the catalytically competent form, and the hydrogen bond network around the UDP moiety is fully conserved. We first examined the position of a series of tyrosine residues i.e. Y194, Y196 and Y199 that were suggested to be involved in the binding of xylobiose in the dβ4GalT7 structure (33). Our computational analysis indicates that Y194 stabilizes both the donor and acceptor substrates location by establishing a hydrogen bond with a βphosphate oxygen of UDP-Gal and a π-stacking interaction with the 4-methylumbelliferyl moiety, respectively (Fig. 1A). Residue Y196 is not hydrogen-bonded to the substrates but to the sidechain of residue D229, allowing its second carboxylic oxygen to be suitably oriented to form a hydrogen bond with the O2 atom of Xyl. The spatial orientation of Y199 inside the substrate binding pocket allows the formation of a hydrogen bond between its side-chain hydroxyl and the O2 atom of the Gal moiety of UDP-Gal. Altogether, residues Y194, Y196 and Y199 form a strongly hydrophobic cluster that is required for correct binding of the substrates. Analysis of the position of H195, a conserved amino acid located between the two active site Y194 and Y196 residues, shows no hydrogen bond involving its side-chain. However, the backbone nitrogen atom of this residue is hydrogen-bonded with the CO group of 4-MUX (Fig. 1A). As illustrated in figure 1B, R226 is located on the surface of the acceptor binding site contributing to an amphipathic entry door with the aromatic residues. In our model, there is no hydrogen bond involving the side-chain of R226. Instead, its by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 6 backbone nitrogen atom is hydrogen-bonded with the O3 atom of the Xyl moiety of 4-MUX (Fig. 1A). The structural impact of R270 on enzyme activity, in the context of EDS was also addressed. The model structure of hβ4GalT7 reveals that R270 belongs to the flexible loop [261-284] that moves upon donor substrate binding, thus creating the acceptor substrate binding site (Fig. 1A). This conformational change leads to the closed and catalytically competent conformation of the active site. However, the crystal structure of the human enzyme (34), as well as our own model in complex with both the donor and acceptor substrates do not highlight specific interactions established by this residue, although its close location to the surface of the active site has to be underlined (Fig. 1B). Kinetic properties of the human recombinant hβ4GalT7 mutants expressed in E. coli – To assess the functional importance of the residues of the acceptor binding site highlighted by our model, we carried out point mutagenesis and analyzed the consequences of conservative and non-conservative mutations on the kinetic parameters of hβ4GalT7 expressed and purified from recombinant E. coli cells. The wild-type enzyme and engineered mutants were produced as truncated fusion proteins lacking the 60 N-terminal amino acids (including the transmembrane domain and part of the stem region) linked to GST and purified by affinity chromatography (data not shown). This led to 1.0 to 2.5 mg of pure protein per liter of culture for wildtype and mutant hβ4GalT7. Kinetic assays were performed using 4-MUX as acceptor substrate, which allowed quantification of the transfer reaction product by UV-detection coupled to HPLC. The kcat and Km values of the wild-type enzyme towards UDP-Gal and 4-MUX shown in table 1 were in agreement with previous work (17, 31). Substitution of Y194 by alanine led to an inactive enzyme, and its conservative substitution by phenylalanine did not restore the galactosyltransferase activity of hβ4GalT7 (Table 1), indicating a critical role of this residue and, importantly, of the hydroxyl group of the tyrosine side-chain. The mutation of Y196 to alanine totally abolished enzyme activity whereas replacement of this residue by phenylalanine led to a slightly active enzyme. The Y196F mutation did not impair enzyme affinity towards the donor substrate to a major extent but this mutant presented a lower affinity towards 4-MUX with a Km value about three-fold that of the wild-type enzyme (Table 1). As observed in the case of Y194 and Y196, the nonconservative mutation Y199A led to a total loss of enzyme activity. However, similarly with what was observed for Y196, substitution of Y199 by phenylalanine led to an active hβ4GalT7 enzyme with Km values towards UDP-Gal and 4-MUX and kcat value only weakly affected compared to the wild-type enzyme. These data suggest that the aromatic ring of phenylalanine at position 199 is sufficient to support xyloside binding and activity. Substitution of H195 by alanine, glutamine or arginine was carried out. The Km values of all three mutants towards UDP-Gal and 4-MUX were mostly comparable to those of the wild-type enzyme, indicating that these mutations had no major effect upon hβ4GalT7 affinity towards its substrates. Moreover, the substitutions at position 195 did not affect the rate of reaction transfer, as indicated by the kcat values that were essentially unchanged (Table 1). Altogether, these results indicate that the side-chain of H195 does not play a critical role in xyloside binding and hβ4GalT7 catalytic activity. Substitution of R226 by alanine or lysine did not impair the affinity towards the substrates with Km values for UDP-Gal and 4-MUX, that were in the same range to that of the wild-type enzyme, and produced a moderate decrease (about two-fold) of the catalytic constant value (Table 1). The R270 residue is mutated to cysteine in the progeroid form of EDS syndrome, and we previously showed that this mutation led to a significant decrease in hβ4GalT7 activity, mainly due to a reduced affinity towards 4-MUX (about ten-fold, see ref. 17). In order to ascertain the contribution of this residue in hβ4GalT7 activity and xyloside binding, we performed kinetic assays following the conservative R270K and nonconservative R270A mutations. The kcat values for both mutants were about two-times lower than that found for the wild-type enzyme and the Km value towards UDP-Gal was almost unaffected (Table 1). Effect of Y194, Y196, Y199, H195, R226 and R270 mutations on the galactosyltransferase activity of hβ4GalT7 towards 4-MUX in cellulo – To check the importance of the selected residues onto hβ4GalT7 function in a cellular context, we designed an experimental procedure involving CHOpgsB-618 cells transfected with hβ4GalT7 cDNA encoding either the wild-type or the mutant enzymes fused to a myc-tag sequence at their C-terminal end, as described in Experimental Procedures. CHOpgsB618 cells expressing the recombinant enzymes were tested for in cellulo galactosyltransferase activity in the absence or in the presence of 4-MUX as exogenous acceptor substrate. The expression level of enzymes was checked by immunoblot analysis using ImageJ software. As shown in figure 2, an additional upper band at ~39kDa was observed (Fig. 2, insert). This ~39kDa band can be attributed to the N-glycosylated form of the enzyme as it disappears upon PNGase F digestion (Fig. 2, insert, left panel). Both bands were taken into account to quantify the by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 7 total enzyme expression level. The results indicate that all mutants considered in these experiments were expressed at a similar level to that of the wildtype protein (Fig. 2, insert, right panel). As shown in figure 2, the GAG synthesis level in cells expressing wild-type hβ4GalT7 was about 7and 9.5-fold higher in the presence of 4-MUX at 5 and 10 μM concentration, respectively, than in the absence of acceptor substrate, indicating that CHOpgsB-618 cells expressing hβ4GalT7 are able to prime efficiently GAG chains synthesis from 4MUX, in agreement with previous studies (17, 31). As expected, cells expressing either Y194A or Y194F mutant in the presence of 4-MUX showed GAG synthesis levels comparable to those obtained with cells transfected with empty vector (Fig. 2). These in cellulo assays confirm that any mutation affecting the Y194 position leads to a total loss of hβ4GalT7 activity. Substitution of Y196 to alanine dramatically reduced the GAG synthesis rate whose level in the presence of 4-MUX was comparable to that obtained with cells transfected with empty vector. This is consistent with the loss of enzymatic activity observed in the in vitro assays (Table 1). By contrast, the conservative mutation Y196F allowed GAG chain priming from 4-MUX. However, the GAG expression level reached with this mutant was about 2 to 3-fold lower than that of the wild-type enzyme, at 4-MUX at 5 and 10 μM concentrations, respectively (Fig. 2). This result is consistent with the drastic decrease of the kcat/Km value towards 4MUX found for the purified Y196F mutant. Comparable results were obtained with cells expressing hβ4GalT7 whose sequence is mutated on the Y199 position. Indeed, cells expressing Y199A mutant were unable to synthesize GAG chains from 4-MUX, whereas cells expressing Y199F showed GAG chains synthesis at a level about half of that observed with cells expressing wild-type enzyme (Fig. 2). These results are in line with the reduced efficiency exhibited in vitro by the enzyme substituted on the Y199 position (Table 1). Altogether, these results demonstrate that both conservative and non-conservative mutations affecting either Y194, Y196 or Y199 position significantly impaired GAG chains biosynthesis in a cellular context, in line with in vitro data (Table 1). In the presence of 4-MUX, the GAG synthesis rate in cells expressing H195A, H195Q and H195R mutants was moderately reduced, i.e. 10 to 15% lower than that of cells expressing the wild-type enzyme (Fig. 2). These results indicate that the sidechain of this residue does not influence the galactosyltransferase activity of hβ4GalT7 in the context of 4-MUX-primed GAG chains in eukaryotic cells, corroborating the findings that none of the mutations of H195 did significantly affect in vitro activity (Table 1). The level of [SO4] incorporation in the presence of 4-MUX in cells expressing R226A was about 2-times lower than that of the wild-type enzyme (Fig. 2). In addition, replacement of R226 by lysine slightly increased the GAG expression level compared to alanine substitution, reaching about 60% that obtained with the wild-type, at 5 and 10 μM 4-MUX concentration. Corroborating in vitro kinetic parameters, these cellular assays indicate that modification of the side-chain of R226 produces minor effects on galactosyltransferase activity. We finally examined the impact of mutations of the R270 residue upon GAG synthesis in eukaryotic cells. We observed that the GAG synthesis rate in cells expressing R270A mutant was about 55% lower than that of the wild-type enzyme, at 5 and 10 μM 4-MUX concentration (Fig. 2). The GAG synthesis level of the conservative mutant R270K was also about two-fold reduced compared to the wild-type (Fig. 2). These results confirm that mutations of R270 significantly affect the capacity of hβ4GalT7 to synthesize GAG chains from 4MUX in a cellular context. Effect of Y194, Y196, Y199, H195, R226 and R270 mutations on the ability of hβ4GalT7 to initiate the glycosylation of the decorin PG in cellulo – We next determined whether the mutations would affect GAG chain formation on the core protein of decorin, used as a model PG (31). To this aim, CHOpgsB-618 cells were engineered to stably express the recombinant human decorin, and were transiently transfected with a pcDNA3.1 vector encoding the wild-type or the mutant forms of hβ4GalT7. This allowed monitoring GAG substitution of the secreted PG by western blot analysis (Fig. 3, insert, left and right panels). The rates of the decorin PG glycosylation in cells expressing the wild-type or mutant hβ4GalT7 were determined as described in Experimental Procedures, then reported onto the histogram (Fig. 3). Results showed that non-conservative mutations of the tyrosine residues at position 194, 196 and 199 as well as mutations of Y194 and Y196 to phenylalanine fully abolished the glycosylation of decorin (Fig. 3). These data were consistent with the drop of GAG chains primed from 4-MUX in cells expressing hβ4GalT7 mutated at these positions (Fig. 2). However, the substitution of Y196 to phenylalanine induced a more dramatic effect on the glycosylation of decorin than on the in vitro or in cellulo activity towards 4-MUX. Furthermore, results shown in figure 3 indicate that the conservative mutation Y199F allowed to recover up to 70% of the decorin glycosylation level compared to cells expressing the wild-type hβ4GalT7. We also assessed the role of the H195 residue in by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 8 the glycosylation process of decorin. The results shown in figure 3 indicate that none of the mutations H195A, H195Q or H195R significantly affected the level of decorin glycosylation. These data are consistent with the results obtained on the in vitro and in cellulo activity of the enzyme towards 4-MUX. Together, mutagenesis experiments indicate that modification of the amino acid side chain at position 195 did not greatly affect xyloside binding and galactosyltransferase activity of hβ4GalT7. Investigation of the effect of R226 mutation upon the ability of CHOpgsB-618 cells to glycosylate decorin showed that the glycosylation level reached with cells expressing R226A or R226K was about 65% of that obtained with cells expressing the wild-type enzyme. These results were consistent with the in vitro and in cellulo GAG chain synthesis assays (Fig. 3). Since we aimed to better understand the molecular basis of the EDS syndrome, it was important to further investigate the impact of mutations affecting the R270 position upon the decorin glycosylation. The decorin glycosylation level reached with cells expressing either R270A or R270K mutant was about half of that obtained with cells expressing the wild-type hβ4GalT7, consistent with in vitro and in cellulo galactosyltransferase assays (Fig. 3). Xyloside inhibitors design and in vitro and in cellulo competition assays – We next took advantage of the knowledge gained from our investigation of the organization of the acceptor substrate binding site to synthesize and test xyloside analogs as potential inhibitors of hβ4GalT7. To this end, in vitro competition assays were performed as described in Experimental Procedures. The specific activity as a function of the logarithm values of the inhibitor concentrations are reported in figure 4B and data fitted to the logistic equation provided IC50 values reported in Table 2. Since the C4-position is critical for both binding and transfer of the Gal residue from UDP-Gal onto the xyloside acceptor, we first synthesized a 4deoxy derivative of 4-MUX (4H-Xyl-MU, Fig. 4A) and tested this compound as inhibitor of hβ4GalT7 in vitro. 4H-Xyl-MU was able to inhibit up to 50% of the initial activity at a 2 mM concentration (Fig. 4B), with an IC50 value of about 1 mM and a Ki value of about 0.5 mM (Table 2). To test whether hydrogen bond formation between 4-MUX and the protein via H195 is important for the inhibitory potency, we synthesized 4H-Xyl-NP which aglycone structure is unable to establish such interaction and compared its inhibitory effect to 4HXyl-MU. This compound produced a decrease of hβ4GalT7 activity towards 4-MUX less than 25% at the highest concentration (Fig. 4B) that did not allow determining IC50 and Ki values. These results clearly indicated that 4H-Xyl-NP is a weak inhibitor of hβ4GalT7. We next substituted the equatorial hydrogen of the C4 atom of the Xyl moiety by a fluorine atom, closer to oxygen in term of electronegativity and predicted to fit the active site in term of steric hindrance. 4F-Xyl-MU led to up to 60% inhibition of the hβ4GalT7 activity at the highest concentration (Fig. 4B), with an IC50 of 0.06 mM and a Ki of 0.03 mM. The inhibition constant for this compound is more than ten-times lower than that reached for the deoxy-analog (Table 2). To complement the in vitro assay, we assessed the ability of the synthesized xyloside derivatives to inhibit GAG chains biosynthesis in cellulo. Addition of 4H-Xyl-NP produced a moderate but significant 20% decrease of GAG chains synthesis in CHOpgsB-618 cells, when used at 50 and 100 μM (Fig. 5A). This correlated with the weak in vitro inhibition level obtained with this compound. The compound 4H-Xyl-MU allowed a larger inhibition of GAG chains synthesis, with up to 30% reduction of the GAG synthesis rate at similar concentrations (Fig. 5B). The best inhibitory effect was observed when performed in the presence of 4F-Xyl-MU leading to up to 50% inhibition of the initial GAG chains synthesis rate at 100 μM concentration (Fig. 5C). Interestingly, preliminary results indicated that 400 μM of 4F-Xyl-MU inhibited the initial decorin glycosylation rate by about 50%, without affecting the viability of CHOpgsB-618 cells (data not shown). Altogether, the latter data confirmed that this fluorinated compound should be considered as a promising non-cytotoxic xyloside-based inhibitor of hβ4GalT7. DISCUSSION β4GalT7 is a unique enzyme in the GAG biosynthetic pathway with regard to its capacity to use exogenous xyloside molecules as substrates and/or inhibitors that can efficiently modulate GAG synthesis in vitro and in vivo (19, 20, 45). This glycosyltransferase is also central in the GAG synthesis process since the formation of the tetrasaccharide linker is a prerequisite to the polymerization of both HS and CS/DS chains. The human enzyme thus represents a prime target for the design of effectors of GAG synthesis as drugs to correct GAG disorders associated with numerous malignant conditions such as genetic diseases and cancer. To meet this challenge, we pioneered structure-function studies of the recombinant hβ4GalT7. We previously carried out structural, thermodynamical and phylogenetic investigations that identified key amino acid residues mainly implicated in the recognition and binding of the by gest on Jne 9, 2017 hp://w w w .jb.org/ D ow nladed from Structure-guided inhibitors of hβ4GalT7 9 donor substrate (31, 46). We also provided insight into the molecular basis of the GAG defects characterizing rare forms of EDS syndrome (17, 47). In the present work, in order to develop xyloside compounds that will specifically target the hβ4GalT7 activity for therapeutic purpose, we explored the architecture of the acceptor substrate binding site. To this aim, we combined functional investigations including site-directed mutagenesis, kinetic analyses, in vitro and in cellulo evaluation of galactosyltransferase activity and GAG synthesis, and a computational approach. This allowed mapping the acceptor binding site and to design and synthesize a potent xyloside-based inhibitor of GAG