Uridine Bisphosphonates Differentiate Phosphoglycosyl Transferase Superfamilies

Abstract

Complex bacterial glycoconjugates drive interactions between pathogens and symbionts and their human hosts. Glycoconjugate biosynthesis is initiated at the membrane interface by phosphoglycosyl transferases (PGTs), which catalyze the transfer of a phosphosugar from a soluble uridine diphospho-sugar (UDP-sugar) substrate to a membrane-bound polyprenol-phosphate (Pren-P). The two distinct superfamilies of PGT enzymes (polytopic and monotopic) show striking differences in structure and mechanism. We designed and synthesized a series of uridine bisphosphonates (UBPs), wherein the diphosphate of the UDP and UDP-sugar is replaced by a substituted methylene bisphosphonate (CXY-BPs; X/Y = F/F, Cl/Cl, (S)-H/F, (R)-H/F, H/H, CH₃/CH₃). UBPs and UBPs incorporating an N-acetylglucosamine (GlcNAc) substituent at the β-phosphonate were evaluated as inhibitors of a polytopic PGT (WecA from Thermotoga maritima) and a monotopic PGT (PglC from Campylobacter jejuni). Although CHF-BP most closely mimics diphosphate with respect to its acid/base properties, the less basic CF₂-BP conjugate more strongly inhibited PglC, whereas the more basic CH₂-BP analogue was the strongest inhibitor of WecA. These surprising differences indicate different modes of ligand binding for the different PGT superfamilies implicating a modified P–O- interaction with the structural Mg²⁺. For the monoPGT enzyme, the two diastereomeric CHF-BP conjugates, which feature a chiral center at the Pα-CHF-Pβ carbon, also exhibited strikingly different binding affinities and the inclusion of GlcNAc with the native α-anomer configuration significantly improved binding affinity. UBP-sugars are thus revealed as informative new mechanistic probes of PGTs that may aid development of novel antibiotic agents for the exclusively prokaryotic monoPGT superfamily.

INTRODUCTION

Nucleoside diphosphate sugar (NDP-sugar) substrates are involved in a variety of essential cellular processes and serve key roles as glycosyl and phosphoglycosyl donor substrates for the biosynthesis of complex glycoconjugates.¹ An important strategy for glycoconjugate assembly involves a membrane-associated initiation step catalyzed by a phosphoglycosyl transferase (PGT). PGTs mediate the transfer of a phosphosugar from an NDP-sugar substrate to a membrane-anchored polyprenol phosphate (Pren-P) acceptor (Figure 1).² The resulting PGT product, a Pren-PP-sugar, is then further elaborated by stepwise addition of sugars from NDP-sugar donors, catalyzed by the sequential action of a set of glycosyl transferases (GTs).³

Figure 1.

A. Ribbon diagram of polyPGT MraY from Aquifex aeolicus (PDB 5CKR). Insert shows close-up view of the MraY active site. PolyPGTs invoke a mechanism involving a ternary complex intermediate and feature a conserved Gly-Xaa-Xaa-Asp-Asp motif. B. Ribbon diagram of monoPGT PglC from Campylobacter concisus (PDB 5W7L), insert shows the close-up view of the PglC active site Asp-Glu catalytic dyad. MonoPGTs follow a ping-pong mechanism, wherein a covalent phosphoglycosyl-Asp intermediate is formed.

There is emerging interest in PGTs due in part to the surprising mechanistic and structural dichotomy between the two known PGT superfamilies.⁴ These superfamilies include either a polytopic or a monotopic functional domain,⁵ and, although the different PGTs catalyze chemically equivalent transformations, thus far, it has been shown that polytopic PGTs (polyPGTs) proceed through a ternary complex mechanism,^6–8 and monotopic PGTs (monoPGTs) invoke a ping-pong mechanism (Figure 1AB).⁹ Although polyPGTs are observed across all domains of life, extensive bioinformatics analyses have revealed that the monoPGT superfamily is exclusive to prokaryotes.^10,11 The biological importance of PGTs in the first step of glycoconjugate assembly in bacterial pathogens and symbionts highlights the significance of structural and mechanistic studies to provide new insight into the determinants of ligand specificity and the key drivers of catalysis as the foundation for inhibitor and chemical probe discovery.

Progress in our understanding of the structures and mechanisms of PGTs is challenged by the fact that these enzymes are integral membrane proteins. To date, structures of two polyPGTs have been reported. These are MraY (Aquifex aeolicus)¹² the essential PGT in the biosynthesis of the bacterial cell wall peptidoglycan and the human GlcNAc-1-P-transferase or GPT (DPAGT1), which catalyzes the initiating step in the dolichol pathway for N-linked protein glycosylation.^8,13,14 The structures are intriguing as several show complexes of the polyPGTs with bisubstrate analogues of natural product origin (e.g. tunicamycin and muraymycin D2).^12,14,15 The polyPGT superfamily features a conserved Gly-Xaa-Xaa-Asp-Asp (GXXDD) motif with the conserved adjacent aspartic acid residues proposed to be required for substrate orientation and essential for activity in these and other polyPGTs.^16,17 Although many of the inhibitor- and substrate-bound complexes lack the Mg²⁺ co-factor, two structures of the human GPT with tunicamycin¹⁴ and the UDP-GlcNAc/Mg²⁺ complex¹³ provide insight into polyPGT-small molecule binding.

For the monoPGT superfamily, there is a single X-ray crystal structure of PglC (from Campylobacter concisus).¹⁸ Structural and mechanistic studies support a mechanism involving a catalytic DE dyad, wherein the aspartic acid carboxyl attacks at the β-phosphate of the NDP-sugar substrate forming a covalent phosphosugar intermediate.⁹ In the structure, the essential Mg²⁺ cofactor is coordinated to the Asp carboxyl group and a phosphate ion.¹⁸ In this case, the two-step ping-pong mechanism of the monoPGTs precludes determination of the structure of the Michaelis complex in the presence of the NDP-sugar substrate bound and Mg²⁺, as the covalent intermediate forms even in the absence of the Pren-P acceptor substrate. At first glance, the GXXDD motif of the polyPGT superfamily appears to bear resemblance to the Asp-Glu (DE) catalytic dyad in the monoPGT family. However, based on the crystal structures of the PGTs, the geometry of these motifs are quite distinct; the conserved adjacent aspartic acid residues of the polyPGTs are presented on a canonical α-helix,¹⁹ whereas the DE catalytic-dyad of the monoPGT superfamily has side chains with co-facial positioning along the 3₁₀ helix that scaffolds the active site (Figure 1, inserts).¹⁸

Despite the biological significance of PGTs, the mechanistic tools available for probing these crucial enzymes in glycoconjugate assembly are limited. Furthermore, many of the existing small molecule probes are derived from complex nucleoside natural products, which are mostly proposed to bind to the polyPGTs as bisubstrate analogues.^15,20 As a complement to the natural product inhibitors and the synthetic analogues that they inspire,²¹ non-hydrolyzable NDP-sugar CXY-bisphosphonate analogues (NBPs) can potentially provide insight into the structure and function of enzymes, such as the PGTs, that act on NDP-sugar substrates. Importantly, NBP-sugars could provide critical information for structure determination,^22,23 as inhibitor scaffolds for identifying structure-activity relationships (SAR),²¹ including inhibitor binding modes²⁴ and, in the case of monoPGTs, as leads for agents with antibiotic activity.

With these opportunities in mind, we describe here the modular syntheses of a series of uridine 5’-bisphosphonate (CXY-UBP) and uridine 5’-bisphosphonate-N-acetylglucosamine (GlcNAc-CXY-UBP) analogues (Figure 2) in which the central diphosphate (P-O-P) oxygen is replaced by a substituted methylene (P-CXY-P), wherein X/Y = H/H, F/F, Cl/Cl, CH₃/CH₃, (S)-H/F and (R)-H/F. The α,β-CXY- and β,γ-bisphosphonate analogues of deoxynucleotides^25–29 have been previously used as probes of nucleic acid polymerase structure and function.^30–33 In addition to introduction of a non-hydrolyzable bisphosphonate mimicking the natural diphosphate, the CXY substitution allows variation in such properties as POH/PO^– acidity, P-O/P-C bond lengths, P-O-P/P-C-P bond angles,³⁴ and the effect of steric perturbation adjacent to the CXY group.³⁵ Finally, for BPs where X ≠ Y, as in CHF, the individual diastereomeric analogues will have nearly equivalent POH/PO^– acidity but different orientations of the C-X and C-Y substituents within the binding site. Herein, the PGT substrate analogues are applied in comparative inhibition studies of representative members of the two PGT superfamilies.

Figure 2.

Non-hydrolyzable uridine 5’-bisphosphonate analogues (H-CXY-UBP) and uridine 5’-bisphosphonate-N-acetylglucosamine analogues (GlcNAc-CXY-UBP) synthesized and analyzed in this study. The diphosphate bridging oxygen of the UDP is replaced by a panel of substituted methylene groups (CXY; X/Y = F/F, Cl/Cl, (R)-H/F, (S)-F/H, H/H, CH₃/CH₃). Uridine is denoted as a circled “U”. The stereochemical assignments of 7 and 8 are (S)-CHF and (R)-CHF, respectively.

EXPERIMENTAL

Synthesis of the CXY-UBP and GlcNAc-CXY-UBP analogues

We developed two synthetic routes to the nucleotide analogues in this study (Figure 2, 1-11). The routes differ in the initial conjugation partner for the bisphosphonate moiety (uridine or GlcNAc). Scheme 1A, exemplified by the synthesis of α-GlcNAc-CH₂ (11) (11% overall yield), benefits from the use of unprotected uridine but requires two demethylation steps due to the use of trimethyl ester bisphosphonate. In contrast, Scheme 1B is more modular, facilitating diversification of the bisphosphonate components, and affords higher overall yields. Scheme 1B was utilized to synthesize five R-CXY-UBPs (R=H, CYX: CCl₂ (1), CF₂ (3), (R/S)-CHF (6), C(CH₃)₃ (9), CH₂ (10) (71–90% overall yields) and five GlcNAc-CXY-UBP derivatives α-GlcNAc-CCl₂ (2), α-GlcNAc-CF₂ (4), β-GlcNAc-CF₂ (5), α-GlcNAc-(S)-CHF (7), α-GlcNAc-(R)-CHF (8) (~25% overall yields). All compounds were characterized by ³¹P NMR, ¹H NMR, ¹³C NMR, ¹⁹F NMR (for CHF and CF₂ analogues), COSY, HSQC, and high-resolution mass spectrometry (HRMS). To assign the CHF stereocenter in α-GlcNAc-(S)-CHF (7) and α-GlcNAc-(R)-CHF (8), a modified synthesis incorporating a sterochemically-defined bisphosphonate intermediate derivatized with a chiral auxiliary was deployed (Scheme 2).

Scheme 1.

General synthesis of α-GlcNAc-UBP derivatives

A. Reagents and conditions: (a) Triflic acid (HOTf), CH₂Cl₂, rt, inert atmosphere, 2 h (quant. yield of 13); (b) 14 (from tetramethyl ester refluxed 5.5 h in MeCN with excess TEA, monoacid generated with DOWEX H⁺(90%)) and 13, dioxane, 55 °C, inert atmosphere, 18 h (95%); (c) NaI, acetone, rt, 18 h (quant. yield of symmetrical disodium salt (16); (d) DOWEX H⁺, MeOH, H₂O, 0 °C (quant. yield of 17) (e) 18 (uridine), dioxane, DMF, PPh₃, DIAD, rt, inert atmosphere, 18 h (yield of 19 not measured); (f) PhSH, DIPEA, NMP, 40 °C, inert atmosphere, 48 h (20% yield of 20 from 17); (g) H₂O, NH₄OH (pH 11), rt, 1 h (70%). B. Reagents and conditions: (h) 21 (2’,3’-O-isopropylidene uridine), p-TsCl, pyridine, rt, inert atmosphere, 18 h (22, 95%); (i) general procedure: 23 [from tetraacid, H₂O, EtOH, 3 equiv. (n-Bu)₄NOH, pH ~7.5, rt, 5 min], MeCN, rt, inert atmosphere (75–95%); (j) aq. HCl pH ~0.5, rt, 2 h (quant.); (k) 1) DOWEX H⁺, H₂O; 2) 13 (GlcNAc oxazoline), DMF, 45 °C, inert atmosphere, 48 h (49–70%); to obtain β-anomer: 75 °C, 3 h (75%)), the mixture of (R/S)-CHF diastereomers was cleanly separated by preparative reverse-phase HPLC; (l) aq. NH₄OH (pH 10.5–11.5), rt, 2.5 h (quant. yield). α-GlcNAc-(S)-CHF (7) and α-GlcNAc-(R)-CHF (8) were established as (S)-CHF and (R)-CHF diastereomers, respectively.

Scheme 2.

Synthesis of stereochemically-defined α-GlcNAc-CHF-UBP

Reagents and conditions: (a) i. 1 M HCl, rt, ii. DOWEX H⁺, H₂O, MeOH (quant. yield of 27) (b) 22, DMF, 50 °C, inert atmosphere (44% yield of 28), (c) H₂O, UV (λ = 365 nm), 48 h (76% yield of 29) (d) 1 M HCl, rt, 1 h (quant. yield of 30) (e) [bisphosphonate preparation: 1:1 H₂O, MeOH, 3 equiv. (n-Bu)₄NOH, pH ~8.3, rt], DMF, GlcNAc oxazoline = 13, 45 °C, inert atmosphere, 5.5 h (57% yield of 31α and 31β) (f) H₂O, NH₄OH pH 10.5, 28 °C, overnight (31% yield of 8α and 8β). Note: The designation of chirality at the -CHF- center changes through the scheme due to the change in Cahn-Ingold-Prelog priority of the phosphonate ester substituents.

Scheme 1A begins with peracetyl-GlcNAc, 12, which is converted to the corresponding oxazoline (13) following known procedures.³⁶ Compound 13, in the presence of trimethyl methylenebis(phosphonate) (14) resulted in the bisphosphonate ester 15. The bisphosphonate ester (15) was symmetrically di-demethylated with sodium iodide to obtain the intermediate 16, which was converted to 17 by passage through DOWEX H⁺. The Mitsunobu coupling of 17 and uridine (18), which favored attack at the less sterically hindered β-P atom provided intermediate 19, which was then demethylated with DIPEA/PhSH to yield 20.^37,38 The O-acetyl groups in 20 were removed with aqueous ammonium hydroxide solution to afford the desired product α-GlcNAc-CH₂ (11). The second route (Scheme 1B) begins with activation of 2’,3’-O-isopropylidene uridine (21) as the 5’-tosyl ester 22,³⁹ followed by reaction with the tris(tetrabutylammonium) salt of the selected methylene-bisphosphonates (23a-e) and then ion exchange to afford 24a-e. The 2’,3’-isopropylidene protecting group was then removed under acidic conditions to obtain five H-CXY-UBPs: H-CCl₂ (1), H-CF₂ (3), H-(R/S)-CHF (6), H-C(CH₃)₂ (9), H-CH₂ (10). Selected H-CXY-UBP products (1, 3, 6) were reacted with 13 to afford O-acetyl protected GlcNAc-CXY-UBPs (25a-c). In summary, the reaction of 1 with 13 at 45 °C, provided 25aα. Under similar conditions, UBP 3 provided 25bα. In addition, when 3 was subjected to modified conditions, including higher temperatures, the glycosidic linkage was epimerized to the β-anomer at C-1 designated as 25bβ.³⁶ Finally, the H-(R/S)-CHF mixture, 6*, was converted to 25cα* by reaction with 13 using the standard 45 °C conditions and individual (R/S)-CHF diastereomers were separated by RP-HPLC. O-acetyl deprotection of the GlcNAc under basic conditions produced α-GlcNAc-CCl₂ (2), α-GlcNAc-CF₂ (4), β-GlcNAc-CF₂ (5), and the individual stereoisomer α-GlcNAc-(S)-CHF (7) and α-GlcNAc-(R)-CHF (8). For 7 and 8, the O-acetylated GlcNAc substituent was crucial to this separation as the precursor H-(R/S)-CHF (6) could not be resolved by RP-HPLC, and thus was studied as an epimeric mixture 6*. Isolation of each O-acetylated GlcNAc containing CHF-epimer (25cα) was carried out by RP-HPLC at the penultimate step in Scheme 1B, by separating 25cα*. This purification yielded individual, but not yet stereochemically-defined, epimers, primed for the final synthetic step.²⁵ Compound 9 was not converted to the corresponding α-GlcNAc-C(CH₃)₂ derivative due to the weak inhibition of the PGTs observed for this UBP precursor (see below).

The CHF-bisphosphonate intermediate (23c) is prochiral, thus, the product mono- and di-phosphonate esters with uridine and GlcNAc, such as 7 and 8, were obtained as a 1:1 mixture of diastereomers. These related isomers could be cleanly separated by preparative HPLC. However, as presented later, the diastereomeric α-GlcNAc-(S/R)-CHF compounds showed different inhibitory activity with the mono- and polyPGTs thus motivating elucidation of the absolute stereochemistry at the CHF chiral center. To accomplish this, we prepared CHF bisphosphonate derivatives modified with an enantiopure chiral auxiliary, (R)-(+)-α-ethylbenzylamine, and a photolabile protecting group (26) that allowed the correlation of chromatographic elution and NMR properties with absolute stereochemistry as defined by X-ray crystallographic analysis (Scheme 2, shown for a selected isomer).²⁵ The chiral auxiliary was removed from 26 under acidic conditions producing 27, which was then coupled to the tosyl-activated uridine isopropylidine (22) to give 28. Photochemical deprotection at 365 nm yielded 29, followed by uridine deprotection (to 30) and coupling to the GlcNAc oxazoline (13) afforded the acetyl protected α- and β-nucleotide analogues 31α and 31β. After deacetylation of 31α, chromatographic analysis was conducted as in the initial synthesis of 7 and 8, and the HPLC retention times correlated α-GlcNAc-(S)-CHF (7) as the (S)-isomer and α-GlcNAc-(R)-CHF (8) as the (R)-isomer. Detailed NMR analysis was also carried out to support the absolute assignment of configuration (S186 and S187).

Enzyme targets and biochemical analysis

The bisphosphonate analogues were assessed for inhibition of representative members of each of the two PGT superfamilies. For the monoPGT analysis we screened two PglC orthologs; one from Campylobacter jejuni (PglC (Cj), which is a human food-borne entero-pathogen that is a significant cause of gastroenteritis worldwide⁴⁰ and the other from C. concisus (72% sequence homology with Cj).⁴¹ The PglC (Cc) was included as it is the target of the only successful monoPGT X-ray structure determination,¹⁸ however as the orthologs showed similar trends and as the majority of the mechanistic studies had been carried out with PglC (Cj),⁹ we pursued detailed analysis of the Cj ortholog. The PglCs from Campylobacter catalyzes the first membrane-committed step in the biosynthesis of N-linked glycoproteins.⁴² The monoPGT data are compared with that obtained for the polyPGT WecA from Thermotoga maritima (WecA (Tm)).⁶ WecA is an integral membrane protein that initiates the O-antigen and enterobacterial common antigen biosynthesis pathways by catalyzing the transfer of GlcNAc-1-phosphate to undecaprenyl phosphate (Und-P) to produce Und-PP-GlcNAc.^17,43 WecA natively uses UDP-GlcNAc as substrate, with a K_m of 0.62 ± 0.13 mM.⁶

Biochemical analysis has shown that the biochemically-preferred substrate for the Campylobacter PglCs is UDP-N,N′-diacetylbacillosamine (UDP-diNAcBac), with a K_m of 24.61 ± 3.30 μM for PglC (Cj), and 22.9 ± 2.5 μM for PglC (Cc).^3,9 However, for these studies, N-acetylglucosamine (GlcNAc) was included in the structures of the UBP analogues rather than diNAcBac to simplify and expedite synthesis. GlcNAc shares structural features with diNAcBac, including similar stereochemistry and the C2-N-acetamido moiety. UDP-GlcNAc is accepted as a substrate by the PglC enzymes included in this analysis although in general the K_m values are significantly higher (~20-fold) than those for UDP-diNAcBac.

Using the Promega UMP/CMP-Glo^™ system, PGT steady-state kinetics were studied under initial linear rate conditions with <10% UDP-diNAcBac turnover. The UMP/CMP-Glo^™ assay quantifies PGT activity through detection of released UMP by enzyme-mediated conversion of the UMP by-product to a luciferase substrate that can be monitored by luminescence. In all cases, a correction to account for any off-target inhibition of the UMP Glo^™ assay reagent enzymes was included in all analyses.⁴⁴ Due to off-target interactions with assay components when analyzing the simple (non-glycosylated) UBP analogues, inhibition studies were also carried out applying an orthogonal radioactivity-based assay which monitors transfer of a radiolabeled-phosphosugar from UDP-[³H]-diNAcBac to the unlabeled Und-P acceptor; substrate conversion is quantified by scintillation counting after liquid/liquid extraction.³

RESULTS

Inhibition of PGTs by selected H-CXY-UBP and GlcNAc-CXY-UBP analogues

The simple UBP analogues including 1 (H-CCl₂), 3 (H-CF₂), 6 (H-(R/S)-CHF), 9 (H-C(CH₃)₂), and 10 (H-CH₂), (Figure 2, R=H), were first screened at 100 μM with the PGT enzymes to identify the bisphosphonate bridging atoms that conferred the best binding (Figure S1). Compound (3) H-CF₂ demonstrated modest inhibition towards each PGT, with 33% ± 6% inhibition of PglC (Cj), 14% ± 11% inhibition of PglC (Cc), and 19% ± 10% inhibition of WecA (Tm). Inhibition by the other UBP analogues was negligible and in some cases the limitations of the UMP-Glo assay precluded conclusive interpretation of inhibition at the R=H UBP stage.

To develop the probes, we then investigated modification of 3 with GlcNAc featuring either α- or β- anomer linkages (α-GlcNAc-CF₂ (4) and β-GlcNAc-CF₂ (5)). The inclusion of both anomers was valuable for determining whether the GlcNAc-UBPs bound in a substrate-like mode, as the PGTs in this analysis are known to act on the α-anomers of bacterial UDP-sugars. Inhibition studies at 100 μM with PglC (Cj) and WecA (Tm), comparing UBP H-CF₂ (3) with α-GlcNAc-CF₂ (4), and β-GlcNAc-CF₂ (5) are illustrated in Figure 3A. With PglC (Cj), comparison of H-CF₂ (3) with α-GlcNAc-CF₂ (4) showed that incorporation of α-GlcNAc into the bisphosphonate analogue enhances binding and increases inhibition from 33% ± 6% to 76% ± 13%. In contrast, β-GlcNAc-CF₂ (5) shows < 5% inhibition under identical conditions, confirming that the inclusion of a sugar with the native stereochemistry promotes more efficient binding. Inhibition of PglC (Cj) by 3 and 4 is concentration dependent (Fig. 3B), and the IC₅₀ of 4 for PglC (Cj) was determined to be 32 μM ± 5.3 μM (Fig. 3C) under assay conditions with competing UDP-diNAcBac at 20 μM. In contrast, α-GlcNAc-CF₂ (4) showed minimal inhibition of the polyPGT WecA (Tm) relative to the parent UBP (3) (Fig. 3A), consistent with the hypothesis that the UBP-CXY moiety supports a different mode of binding for polyPGTs and highlighting the catalytic divergence between PGT superfamilies.^8,9

Figure 3.

Activity of PglC (Cj) and WecA (Tm) measured after incubation with UBP probes followed by a reaction with UDP-diNAcBac (for PglC) and UDP-GlcNAc (for WecA), in reference to a control with no inhibitor. Solid bars represent the inhibition of PglC (Cj) activity. Striped bars represent the inhibition of WecA (Tm) activity. A. Inhibition of UBPs comprising CXY = F/F with R substituent = H (3), α-GlcNAc (4) and β-GlcNAc (5). B. Varied inhibitor concentration with activity of PglC (Cj) measured by incubation with the UBP probes comprising CXY = F/F, followed by a reaction with UDP-sugar substrate in reference to a control with no inhibitor. Black bars represent 200 μM inhibitor, dark grey bars represent 100 μM inhibitor, and light grey bars represent 50 μM inhibitor. Error bars indicate mean ± SD; n = 3. C. The IC₅₀ value of 4 with PglC (Cj) was measured after 10 min incubation with 4 at varied concentrations (0 μM – 700 μM), followed by a reaction with UDP-diNAcBac substrate (20 μM) and quenched after 10% of the UMP product is formed (15 min). The IC₅₀ value was found to be 32 μM ± 5.3 μM. D. Inhibition of UBPs comprising CXY = H/F chiral center with R substituent = α-GlcNAc (compounds 7 and 8, defined as (S)-CHF and (R)-CHF, respectively). Data plotted using GraphPad Prism as percentage remaining activity compared to no inhibitor. Error bars indicate mean ± SD; n = 3.

The micromolar IC₅₀ of α-GlcNAc-CF₂ (4) prompted an investigation into the specific effect of the fluorine atoms in the asymmetric environment of the PGT active sites. The contribution from each fluorine atom in the CXY moiety was parsed out using stereochemically-defined UBP analogues with the R substituent as α-GlcNAc, namely: α-GlcNAc-(S)-CHF (7) and α-GlcNAc-(R)-CHF (8). Figure 3D shows the comparison of the percent inhibition of α-GlcNAc-(S)-CHF (7) and α-GlcNAc-(R)-CHF (8) at 100 μM towards PglC (Cj) and WecA (Tm). The results demonstrate a strong preference for the (S)-CHF stereochemistry of α-GlcNAc-(S)-CHF (7) for PglC (Cj) (52% ± 10% inhibition) vs (R)-CHF of α-GlcNAc-(R)-CHF (8) (<2% inhibition). In contrast, WecA (Tm) demonstrates distinct binding preferences for these UBP probes, with α-GlcNAc-(R)-CHF (8) showing very slightly preferential inhibition of α-GlcNAc-(R)-CHF (8) relative to α-GlcNAc-(S)-CHF (7) (Fig. 3D). Notably, with respect to PglC (Cj), the percent inhibition of α-GlcNAc-(S)-CHF (7) (52% ± 10% inhibition) is not additive with total percent inhibition from α-GlcNAc-CF₂ (4) (76% ± 13%), suggesting additional physicochemical contributions, for example effects on the pK_a values, from the more electronegative CF₂ moiety that may drive binding beyond a specific interaction of the C-F moiety with the enzyme.

Due to the improved binding of GlcNAc-modified UBPs, such as 4, relative to the simple UBP analogues, we investigated PGT inhibition with analogues having a methylene bridge, CXY = H/H. The UBP analogue H-CH₂ (10), with no R substituent, shows very limited inhibition of PglC (Cj) and WecA (Tm). However, comparison of H-CH₂ (10) with the α-GlcNAc modification (11), reinforces that the combination of CXY moiety with an α-linked sugar also supports a substrate-like binding mode, in this case, for a polyPGT. α-GlcNAc-CH₂ (11) is the best inhibitor of WecA (Tm) with 48% ± 7% inhibition, compared to 28% ± 12% inhibition with PglC (Cj) at 100 μM. The IC₅₀ value of α-GlcNAc-CH₂ (11) with WecA (Tm) was found to be 41 μM ± 10 μM (Figure S2).

Overall, the inhibition of PglC and WecA by H-CCl₂ (1), α-GlcNAc-CCl₂ (2), H-CF₂ (3), α-GlcNAc-CF₂ (4), α-GlcNAc-(S)-CHF (7), α-GlcNAc-(R)-CHF (8), H-CH₂ (10), and α-GlcNAc-CH₂ (11) are graphed and compared (Figure S3). The relative affinities of all the UBP-sugar analogues are summarized in Figure 4.

Figure 4.

Comparison of inhibition by GlcNAc-UBP derivatives.

DISCUSSION

Traditionally, nucleoside sugar diphosphates have been synthesized by directly phosphorylating the nucleoside monophosphate or via the well-known phosphoroamidite method.⁴⁵ However, these approaches are unsuitable for synthesis of UBP derivatives. Vaghefi et al. have described a route to nucleoside methylenediphosphonate sugars in which the sugar moiety is attached to diphenyl bisphosphonate prior to coupling to the nucleoside.⁴⁶ The sugar coupling step was performed at 170°C in vacuo, followed by catalytic hydrogenation at 1000 psi to remove the phenyl groups. Our general method (Scheme 1A) begins with the more conveniently accessible tetramethyl bisphosphonate and monodealkylation of this BP ester is carried out under mild conditions to afford 14. To obtain the stereochemically-defined α-GlcNAc-CHF-UBP derivatives, a streamlined approach was implemented which also gave higher yields (Scheme 1B). Here the activated nucleoside is first coupled to the CXY bisphosphonic acid, followed by reaction with activated sugar and simple acid or base deprotection.

The NBP-sugar synthetic chemistry presented here is modular and should be adaptable to prepare a wide range of nucleotide-sugar mimics with tunable P-CXY-P bisphosphonate moieties, different nucleobases and various carbohydrates or carbohydrate bioisosteres.

The relative inhibitory properties of the UBP-sugar analogues with the two PGT superfamilies are illustrated graphically in Figure 4. In general, although the simple uridine 5’-bisphosphonates (CXY-UBP) showed limited inhibition of both PGT superfamilies, the initial analysis with these analogues served as a guide to prioritize further elaboration into GlcNAc-modified nucleoside analogues. Elaboration of the CF₂-UBP with an α-linked GlcNAc provides the best inhibitor of the monoPGT PglC (Cj) (IC₅₀ 32 μM ± 5.3 μM, Fig. 3C). The improvement in inhibition contrasts the complete loss of activity with the β-linked GlcNAc analog, supporting that substrate mimicry is essential for binding. This observation highlights opportunities for improving monoPGT inhibition, for example, through integration of the diNAcBac sugar, which is the biochemically-preferred sugar for the PglCs in this study, instead of GlcNAc. Although UDP-GlcNAc is a far poorer substrate for PglC than UDP-diNAcBac with a K_M value ~20 fold higher, the manipulation of diNAcBac is synthetically more challenging and there are no commercial sources for this rare sugar or any of its derivatives. Alternatively, the modular synthetic approach readily enables synthesis of modified UBP conjugates that include moieties that substitute for the sugar. For example, in glycan-binding proteins recognition of sugar substrates is often promoted by electrostatic interactions wherein an electropositive C−H in carbohydrates interacts with electron-rich aromatic amino acids yielding CH−π interactions.⁴⁷ Thus a carbohydrate substitute could mimic these CH-π interactions while also affording simplified synthetic routes. When occupying the sugar binding site, aryl groups can act as sugar bioisosteres by participating in similarly favorable π - π stacking interactions.

As illustrated in Figure 4, the mono- and polyPGTs show distinct preferences for different UBP-sugar analogues. GlcNAc-CF₂-UBP (4) is the preferred inhibitor for the monoPGT PglC (Cj), whereas GlcNAc-CH₂-UBP (11) is the preferred inhibitor for the polyPGT WecA (Tm). Both GlcNAc-CCl₂-UBP (2) and H-C(CH₃)₂-UBP (9) show very weak inhibition of the PGTs, however, we do not explicitly discuss these analogues in the trends due to the additional steric burden imposed at the bridging site of the bisphosphonate. Analysis of the trends for WecA (11>8>7>4), show that there is a correlation between the increasing basicity of the bisphosphonate moiety with the strength of inhibitor binding.²⁵ This trend becomes evident when referencing the pK_a values of the CXY-BP moieties: CF₂-BP has the lowest pK_a values, corresponding to the lowest relative basicity in its anionic forms, while CH₂-BP and C(CH₃)₂-BP have the highest pK_a values and therefore the highest relative basicity.²⁵ The IC₅₀ of 11 for WecA (Tm) is 41 μM ± 10 μM (Figure S2). The preferential binding to 11 suggests that the primary factor responsible for GlcNAc-UBP analogue binding is the structural Mg²⁺ cofactor. This Mg²⁺ cofactor may principally be responsible for the stability of the ternary complex by playing a role in substrate binding and orientation. Specifically, increasing the basicity of the bisphosphonate moiety will strengthen the ionic interactions between the phosphonate oxygen anions and the Mg²⁺.

A potential complicating factor that might also influence Mg²⁺ binding is structural variation in the differently substituted bisphosphonates (Table 1). Based on our previously published crystal structures for β,γ-CXY nucleotides,⁴⁸ the P-C-P bond angles are more acute than the P-O-P bond angle in diphosphate by 15–21°and the P-C bond is about 0.3 Å longer. As shown in Table 1, with respect to the distance between the O and O’ oxygen anions which interact with Mg^2+, these effects cancel, with the result that the O,O’ distance in diphosphate and in all of the CXY-bisphosphonates is very similar. Quantum mechanical analysis of these geometries is consistent with this observation.³⁴

Table 1.

Geometric properties of some CXY-bisphosphonates

Z	O	CHCl	CHF	CHCH3	CCl2	CH2	CF2
PDB	2FMS	6TCL	6CTK	6CTJ	6CTI	6CTP	6CTO
Length Pβ-Z (Å)	1.51	1.83	1.89	1.83	1.85	1.87	1.85
Length Pγ-Z (Å)	1.51	1.82	1.79	1.82	1.80	1.85	1.85
Angle P-Z-P (°)	128	113	110	110	107	112	107
Distance Pβ-Pγ (Å)	2.72	3.05	3.01	3.00	2.94	3.09	2.98
Distance PβO- PγO (Å)	2.76	2.93	2.89	2.69	2.82	2.94	2.95
Pβ Distance O-Mg (Å)	2.04	2.14	2.18	1.92	2.10	2.20	2.15
Pγ Distance O-Mg (Å)	2.08	2.09	2.15	2.04	2.11	2.18	2.10

Geometric properties of CXY (= Z) bisphosphonate derivatives determined from X-ray crystallographic data for β,γ-CXY nucleotides bound to pol β.⁴⁸ Calculated parameters for the corresponding dimethyl β,γ-CXY BPs are consistent with these results (Table S2).

In Figure 5, β,γ-CF₂ dTTP is overlaid with dUpNHPP, demonstrating three-dimensionally that the Mg²⁺ coordination geometry is not perturbed by the structural differences from the diphosphate to the bisphosphonate ligand.

Figure 5.

Overlay of CF₂ dTTP (dark grey) and 2’-deoxyuridine-5’-[(α,β)-imido]triphosphate (light grey) in the active site of DNA Pol β showing the distance between the phosphonate (blue dotted)/phosphate (black solid) oxygens and the structural Mg²⁺ (green). The image was created with BIOVIA Discovery Studio Visualizer using data from PDB files 2FMS and 6CTO.

Analysis of the structures of GPT (the human polyPGTs) provide some additional insight to the role Mg²⁺ coordination. GPT also features a conserved DD motif (Fig. S190) and uses UDP-GlcNAc as substrate, either in the presence of a UDP-GlcNAc-Mg²⁺ complex,¹³ or tunicamycin, a potent bisubstrate analog inhibitor¹⁴ (Fig. S191A–D). The two structures show the UDP-GlcNAc-Mg²⁺ and tunicamycin coinciding in the bound state. However, there is not a unique structure showing the Mg²⁺ bound directly to the conserved aspartates in GPT. Therefore, this implies that the Mg²⁺ supports a role in substrate orientation, but not a direct role in catalysis for the polyPGTs. This proposal is both consistent with its known function¹⁷ and our previously reported results where a similar correlation was observed between the basicity of bisphosphonate analogues and the binding affinity to Pol β.⁴⁹

In contrast to the polyPGT, PglC (Cj) inhibition follows a different trend (4>7>11>8), with the highest inhibition (IC₅₀ 32 μM ± 5.3 μM, Fig. 3C) observed with the least basic analogue (4). This alludes to the possibility that the role of Mg²⁺ in the monoPGT system is most important for catalysis in the ping-pong mechanism and plays a less significant role in substrate orientation. In proteins, such as Pol β, catalytic Mg²⁺ is known to exhibit a lower affinity to the protein in the active site compared to structural Mg²⁺.⁴⁹ This suggests that the substrate orientations in complex with PglC are nonsuperimposable and ground state conformations of this complex are mainly defined by protein-ligand interactions.

The binding of 4 to PglC supports that the CF₂ group is the best replacement of the oxygen atom in the parent UDP-sugar. However, this does not give information on the contribution of each C-F bond. There may be specific interactions of the C-F bond with the target enzyme or, the effect could principally be one of modulating the electronic and structural properties of the analog relative to the parent UDP-GlcNAc. To address this issue, we investigated the individual effects of the diastereotopic C-F bonds in 7 and 8 by assessing the inhibitory activity with stereochemically-defined α-GlcNAc-CHF-UBPs. The significant difference between (S)-CHF (7) and (R)-CHF (8) (7>>8) is noteworthy. We considered two hypotheses. First, that the C-F in 7 may form a specific interaction upon binding to PglC. In this case, a possible amino acid candidate is PglC residue Lys59 (Fig. 1B). Lys59 is a catalytically-essential residue, which in the current mechanistic proposal and based on the structure (PDB: 5WL7), would be close to the diphosphate oxygen to protonate the UMP leaving group upon formation of the covalent intermediate.^9,18 Alternatively, the (S)-CHF (7) and (R)-CHF (8) isomers might preferentially adopt different solution state conformations. The solution-state conformations of UDP-GlcNAc, (S)-CHF (7), and (R)-CHF (8) in water in the presence of MgCl₂ were assessed by nuclear Overhauser effect (NOE) analysis (SI NMR summary and Figs. S188 and S189). Under these conditions, all three compounds have uracil protons in an observable NOE range (< 5 Å) to many of the GlcNAc protons, indicating a collapsed structure. The key difference between these three structures arises with the interproton distances between uracil and the N-acetyl protons on the GlcNAc. In UDP-GlcNAc and (S)-CHF (7) the N-acetyl protons are within range to observe clear NOEs with the uracil C6 proton. The (R)-CHF (8) also has the acetyl group protons within range, albeit with a much weaker signal. This difference is significant enough to conclude that 7 and 8 have different favored conformations in solution, and that 7 behaves similarly to UDP-GlcNAc in solution. A possible outcome of conformational differences between 7 and 8 may be that Mg²⁺-bound 7, like UDP-GlcNAc, may be preorganized into a favorable state for binding to the monoPGT leading to improved affinity relative to 8. In contrast to the results with PglC, there is no significant difference between the inhibitory activity of 7 and 8 with WecA. Future structural and computational studies will be applied to further investigate the observed variations in binding to the PGT superfamilies.

CONCLUSIONS

Nucleoside diphosphate sugar (NDP-sugar) substrates feature in innumerable cellular processes and there is considerable interest in the development and study of non-hydrolyzable analogues for applications in structural biology, mechanistic enzymology, and as leads for inhibitor development. In this study we have focused on phosphoglycosyl transferases, which catalyze the first membrane-committed step in many glycoconjugate assembly pathways. PGTs play pivotal roles in initiating production of diverse glycans that are essential for bacterial survival and pathogen-host interactions. In bacteria, PGTs and their NDP-sugar substrates are far more varied than in eukaryotes, reflecting the greater diversity of glycoconjugates in unicellular microorganisms and the potential for selective antibiotic and probe development.

We have presented the synthesis and biochemical analysis of a panel of uridine 5’-bisphosphonate (CXY-UBP) and uridine 5’-bisphosphonate-N-acetylglucosamine (GlcNAc-CXY-UBP) analogues of the UDP-sugar substrates of phosphoglycosyl transferases. The analogues feature a central substituted methylene group that can be tuned to modify the steric and electronic properties of the bridging bisphosphonate. The two PGT superfamilies are differentiated by mechanism and involve either - the direct attack by UndP on the β-phosphate of a UDP-sugar for the polyPGTs, or attack by the conserved aspartic acid residue on the UDP-sugar for the monoPGTs. The polyPGT and monoPGT superfamilies are also distinguished by highly divergent 3D structures. Together, the studies presented here underscore the mechanistic dichotomy of the PGT superfamilies and provide a pathway towards selective inhibition of either the prokaryotic monoPGT superfamily or the polyPGT superfamily found across domains of life. As the monoPGTs are exclusively prokaryotic,¹¹ these enzymes represent potential new targets for the development of antibiotic and antivirulence agents due to the pivotal roles played by the complex glycoconjugates that are biosynthesized in pathways that are initiated by monoPGTs in bacteria.

ABBREVIATIONS

PGT

phosphoglycosyl transferase

monoPGT

monotopic PGT

polyPGT

polytopic PGT

NDP

nucleoside diphosphate

UDP

uridine diphosphate

BP

bisphosphonate

UBP

Uridine bisphosphonate

Cj

Campylobacter jejuni

Cc

Campylobacter concisus

Tm

Thermotoga maritime

GlcNAc

N-acetyl glucosamine

Und-P

undecaprenyl phosphate

diNAcBac

di-N-acetylbacillosamine

Pgl

protein glycosylation