A Parsimonious Mechanism of Sugar Dehydration by Human GDP-Mannose-4,6-dehydratase.

Biosynthesis of 6-deoxy sugars, including l-fucose, involves a mechanistically complex, enzymatic 4,6-dehydration of hexose nucleotide precursors as the first committed step. Here, we determined pre- and postcatalytic complex structures of the human GDP-mannose 4,6-dehydratase at atomic resolution. These structures together with results of molecular dynamics simulation and biochemical characterization of wildtype and mutant enzymes reveal elusive mechanistic details of water elimination from GDP-mannose C5″ and C6″, coupled to NADP-mediated hydride transfer from C4″ to C6″. We show that concerted acid-base catalysis from only two active-site groups, Tyr179 and Glu157, promotes a syn 1,4-elimination from an enol (not an enolate) intermediate. We also show that the overall multistep catalytic reaction involves the fewest position changes of enzyme and substrate groups and that it proceeds under conserved exploitation of the basic (minimal) catalytic machinery of short-chain dehydrogenase/reductases.

6-Deoxysugars, prominently represented by the ubiquitous l-fucose,[1] are functionally important constituents of complex glycans and glycosylated natural products. Their biosynthetic pathways have in common that 4,6-dehydration of a hexose nucleotide precursor constitutes the first committed step.[2,3] The l-fucose (as GDP-l-fucose) is derived from GDP-d-mannose through an evolutionary conserved route via GDP-6-deoxy-α-d-lyxo-hexopyranos-4-ulose (GDP-4″-keto-6″-deoxy-mannose).[4−6] GDP-mannose 4,6-dehydratase (GMD) catalyzes the conversion of GDP-mannose.[5] The basic mechanism of GMD appears to be universally used by sugar 4,6-dehydratases.[2,3] It involves three catalytic steps (Figure a). The GDP-mannose is initially oxidized at C4″ by a NADP+ cofactor tightly bound to the enzyme. Water is eliminated from GDP-α-d-lyxo-hexopyranos-4-ulose (GDP-4″-keto-mannose) at C5″ and C6″ to form a GDP-6-deoxy-β-l-erythro-hex-5-enopyranos-4-ulose (GDP-4″-keto-mannos-5″,6″-ene) intermediate. Reduction of this intermediate by NADPH at C6″ gives the product and regenerates NADP+.[7,8]

Figure 1

Mechanistic basis for 4,6-dehydration of GDP-mannose by human GDP-mannose 4,6-dehydratase (hGMD). (a) The proposed enzymatic mechanism in three catalytic steps. (b) Stepwise mechanism of β-elimination of water from a ketone.[31]

The dehydratase reaction has drawn much interest in structural[2,7,9−20] and biochemical studies[7,8,10,19,21−27] spanning several decades to elucidate the enzymatic mechanism. Sugar 4,6-dehydratases are members of the short-chain dehydrogenase/reductase protein superfamily.[28] Their active sites are equipped with the basic SDR (short-chain dehydrogenase and reductase) catalytic apparatus for oxidation–reduction by NAD(P), which has been well characterized from other enzymes.[22,29,30] Accordingly, central problem in our mechanistic understanding of the dehydratase reaction is to elucidate how enzymes integrate a unique β-elimination of water with an apparently “classical” SDR cycle of catalytic hydride transfer to and from NAD(P).[10,13,24,23,31] However, the switch from alcohol oxidation in the first step of the catalytic reaction to C–C double-bond reduction in the last step (Figure a) is another unique feature of the sugar 1,4-dehydratases that is not well understood mechanistically.[10]

Chemically, β-elimination of a ketone by enzyme catalysis is likely to proceed stepwise, with Cα-H bond cleavage before the Cβ-O bond cleavage. Considering thermodynamic requirements for the Cα-H bond cleavage, Gerlt and Gassman[31] refuted enzymatic reaction via base-catalyzed abstraction of the α-proton to form a carbanion (enolate). They proposed a concerted general acid–general base-catalyzed formation of an enol intermediate from which 1,4-elimination of the β-substituent could occur (Figure b). They also considered that, given suitable geometry of the enzyme–substrate complex as shown in Figure b, the conjugate acid of the base catalyzing the enol formation could also catalyze expulsion of the β-substituent. The proposed mechanism implies a syn stereochemical course for the β-elimination reaction and suggests two as the minimal number of functional groups required for efficient catalysis. Herein, we sought to clarify through study of the human GMD (UniProt accession ID: O60547), whether—and if so to what extent—Gerlt and Gassman’s minimum catalytic principle for β-elimination[31] was in fact incorporated by an actual sugar 1,4-dehydratase that has emerged from evolution through natural selection. We show, based on high-resolution pre- and postcatalytic complex structures of the enzyme, that human GDP-mannose 4,6-dehydratase (hGMD) represents a perfect realization of that principle in its most parsimonious form. We suggest that other sugar 1,4-dehydratases like dTDP-glucose 4,6-dehydratase[13,24] employ the same principle, but in an expanded version.

Our mechanistic analysis builds on four high-resolution crystal structures of hGMD determined in this study (Table S1, Figure ): the wildtype enzyme in complex with the inactive substrate analogue GDP-4″-deoxy-4″-fluoro-mannose (PDB: 6GPJ, 1.94 Å); the inactive E157Q variant in complex with GDP-mannose (PDB: 6GPK, 1.47 Å); the wildtype enzyme in complex with the product GDP-4″-keto-6″-deoxy-mannose (PDB: 6GPL, 1.76 Å); and the S156D variant in complex with GDP-mannose and ADP-ribose (PDB: 6Q94, 2.8 Å). In capturing at atomic resolution the start and end point of the enzymatic reaction, these structures together with biochemical data and evidence from molecular dynamics simulation make detailed suggestion for the catalytic path from substrate to product. It is exactly this important fundamental insight which has been difficult to obtain from previous structural studies on sugar 1,4-dehydratases that could reveal the enzyme–substrate complex[13,15,16] or enzyme complexes with substrate/product analogue.[13,14,17]

Figure 2

High-resolution crystal structures of hGMD. (a) Overall fold of the hGMD dimer (E157Q variant); each monomer has bound GDP-mannose (purple) and NADP+ (yellow). The NADP+ binding loop (cyan), the substrate binding loop (red) and allosteric inhibitor (GDP-l-fucose) binding loop (dark blue) are highlighted. (b–g) Close-up structures of (b,c) wildtype hGMD bound with GDP-4″-deoxy-4″-fluoro-mannose (yellow), (d,e) wildtype hGMD bound with the product GDP-4″-keto-6″-deoxy-mannose (salmon), (f) E157Q variant bound with GDP-mannose (cyan) and (g) S156D variant bound with GDP-mannose (cyan) and ADP-ribose, a cleavage product of NADP+ (white). Hydrogen bonds are shown as dashed black lines, with distances indicated in Å. The 2FO–FC electron density maps of the final structure (gray) are contoured at 2σ and are clipped around the ligands.

Crystals of hGMD (Figure a) contain two to four homodimers in the asymmetric unit, with subunits arranged side to side in an opposite up-and-down orientation. Each subunit adopts the characteristic SDR fold, composed of a prominent Rossman-fold domain for NADP+ binding to which a smaller GDP-Man binding domain is appended (for a detailed structural description, see Figures S3 and S4). The active site is in a cleft at the interface of the two domains. The two actives sites are separate one from another in the dimer structure (Figure a), apparently functioning independently in catalysis. However, there is cooperativity between the subunits for tight binding of NADP+. The loop of residues 55–63 extends into the neighboring subunit and locks down on the NADP+ bound (Figure S3), as seen similarly in prokaryotic and plant GMDs.[14,18] Movement away by this loop and dimer disruption would be necessary for NADP+ to dissociate. The E157Q structure captures the loop (residues 70–78) for binding of the allosteric inhibitor GDP-l-fucose in the same conformation as shown in the GDP-l-fucose complex of wildtype hGMD[32] (Figure a; for details, see Figure S4), despite the fact that no GDP-l-fucose is present. The inhibitor loop is disordered in the two wildtype structures reported here. Although not the main focus of this study, the result bears immediate significance for hGMD inhibition. Conformational sampling, rather than induced fit, by GDP-l-fucose is suggested as the structural principle of inhibitor binding in hGMD. The inhibition entails cooperativity between the protein subunits, as shown in Figure S4. Its mechanistic basis is that GDP-l-fucose binding blocks the entrance to the GDP-mannose binding pocket of the opposing chain.

For structural characterization of the hGMD Michaelis complex (Figure b,c), we first targeted the native enzyme. In search for an unreactive substrate analog, we considered substituting the 4-hydroxy group of α-mannosyl with fluorine. While rendering the substrate incompetent for catalytic turnover, the substitution arguably causes only weak perturbation of the enzyme–substrate interactions originally present.[33−35] We therefore developed a synthesis for GDP-4″-deoxy-4″-fluoro-mannose (Figure S5) and show with isothermal titration calorimetry (ITC; Figure S6) that the fluorinated ligand binds to hGMD with a dissociation constant (Kd) of 1.6 (±0.1) μM, comparable to the 8 (±1) μM Km (∼Kd, vide infra) for GDP-mannose. In a second approach, we targeted the native substrate, necessitating the construction of an inactive hGMD variant (E157Q). The atomic maps of catalytic center interactions shown in the wildtype complex with GDP-4″-deoxy-4″-fluoro-mannose (Figure b,c) and in the E157Q complex with GDP-mannose (Figure f) are consistent and are mutually supportive in suggesting a positioning of the substrate for catalysis. This demonstrates important complementarity in the approaches used.

The hGMD active site is composed of a canonical SDR catalytic triad (Tyr179, Thr155, Lys183) for alcohol oxidation by NADP+, extended by three residues (Glu157, Ser156, Asn208) promoting β-elimination. Tyr179 and Thr155 each form a hydrogen bond with the substituent at α-mannosyl C4″ (fluorine, Figure b,c; hydroxy, Figure f; overlay in Figure S7a). The nicotinamide C4 is above the substrate C4″, with distance (3.5–3.7 Å) and angle (77°–83°; relative to the nicotinamide ring C4–N1 axis) well set for hydride transfer.[9,10] Lys183 establishes a highly conserved SDR proton relay:[36] a chain of hydrogen bonds connects Tyr179 (the catalytic acid/base) via the ribosyl hydroxy group and the ε-amino group of lysine to water (Figure S8). The hGMD is special in that its proton relay does not connect to bulk water, but ends in a reservoir of water molecules buried inside the protein (Figure S8). Glu157, Ser156, and Asn208 each forms a hydrogen bond with the C6″ hydroxy group. The orientation of the hydroxy group is in accordance with a β-elimination having syn stereochemical course.

The structure of hGMD bound with the GDP-4″-keto-6″-deoxy-mannose (Figure d,e) suggests a true postcatalytic complex captured in the crystal. A water molecule, likely the one eliminated from the C6″, is held in place by Glu157, Ser156, and Asn208. The nicotinamide C4 is positioned above the C6″, with distance (3.2 Å) and angle (116°; relative to the C4–N1 axis) both proper for hydride transfer. Tyr179 and Thr155 form a hydrogen bond (2.7 Å each) with the C4″ keto group of product. Comparing the product (Figure d,e) to the substrate complex structure (Figure b,c), the product C6″ has made a 1.1 Å upward movement and is now favorably positioned for hydride transfer to the coenzyme (Figure S9). Apart from this subtle change, all atomic positions defining the catalytic center interactions are effectively the same in substrate and product complex (Figure S7; Figure S9). Therefore, hGMD seems to accommodate the different catalytic steps of its overall reaction (Figure a) without the need for repositioning of the enzyme and substrate groups, thus conforming to the “principle of least nuclear motion” in enzyme catalysis. The central, nearly parallel orientation of the nicotinamide ring to the sugar ring plane determines the strict 1,4-reductive regioselectivity of hGMD. In contrast, as pointed out in study of the hexosamine nucleotide 5,6-dehydratase TunA, a 1,2-selective hydride addition (to reduce the 4″-keto moiety) would necessitate the ring planes to lie in certain angle (observed: 22°) to each other.[19]

Gerlt and Gassman’s mechanism (Figure b) built into hGMD implies a 4,5-enolization of GDP-4″-keto-mannose under concerted general acid–general base catalysis from Tyr179 and Glu157, respectively.[37] In both enzyme structures reporting on the Michaelis complex (Figure b,f), however, the Glu157 is hydrogen bonded to the C6″ hydroxy group. Molecular dynamics simulations of enzyme complex with NADPH and the enol (GDP-mannos-4″,5″-ene) intermediate show that in 21% of 150 structure snapshots analyzed over a total runtime of 15 ns, the Glu157 approaches the C5″ at a distance (∼3.5 Å) plausible for catalytic proton transfer at this position (Figure a, for details, see Figures S12 and S13 as well as Table S2).[38] In the remainder time of the simulation, the Glu157 is in contact with the C6″ hydroxy group. The side-chain conformational flexibility thus revealed is essential for Glu157 to function as catalytic base during the enol formation and, in conjugate acid form, as catalytic acid during the expulsion of water, as proposed in Figure a,c.

Figure 3

Proposed catalytic mechanism of hGMD. (a) Proton abstraction from the C5″ by Glu157 is suggested by results of molecular dynamics simulations. A structure snapshot (3.1 ns) of the complex of wildtype hGMD bound with NADPH and GDP-mannos-4″,5″-ene shows Glu157 in a position suitable for proton transfer. (b) Proton relay or proton uptake from bulk water in the final ketone-forming step of the reaction is shown. (c) Detailed proposal of the catalytic mechanism.

With β-elimination complete, the conversion of the GDP-4″-keto-mannos-5″,6″-ene intermediate to the GDP-4″-keto-6″-deoxy-mannose product likely proceeds in two steps, representing in opposite order the reversal of the previous catalytic steps of oxidation and enolization (Figure c). Accordingly, hydride reduction at C6″ proceeds under catalytic facilitation by Tyr179 as the general acid and gives the 4,5-enol (Figure c), consistent with computational analysis of a small-molecule model of the dehydratase reduction step.[10] The ketone formation involves concerted catalysis by Tyr179 and Glu157 and results in proton transfer from water via Glu157 to the C5″. Our structural analysis of hGMD reveals the path of proton uptake from bulk water (Figure b), and it shows this previously unrecognized path to be largely conserved in other sugar dehydratases (Figure S11).

Mutagenesis combined with measurement of C5″ deuterium exchange, using method adopted from Gross et al.,[23] provides biochemical support to the mechanistic claim that Tyr179 and Glu157 provide concerted catalysis to enol formation. We show with in situ proton NMR that upon incubation in D2O in the presence of the GDP-4″-keto-6″-deoxy-mannose product, the wildtype enzyme catalyzes rapid “wash-in” of solvent deuterium at C5″ (Figure a,c, Table ) while E157Q (Figure b, Table ) and Y179F are inactive (Figure S27, Table ).

Figure 4

In situ 1H NMR monitoring of deuterium incorporation from solvent into C5″ of GDP-4″-keto-6″-deoxy-mannose (3 mM) on incubation with (a) 1.4 μM wildtype hGMD or (b) 71 μM E157Q variant. (a) The H6″ doublets of GDP-4″-keto-6″-deoxy-mannose (keto-H-6″) and the corresponding hydrate (C4″-diol) are gradually transformed to singlets, indicating deuterium incorporation at C5″ catalyzed by wildtype hGMD. Note: signal change from doublet to singlet for H6″ was more conveniently analyzed than signal decrease for H5″. The signal for H5″ was partly overlapped in the 1H-NMR spectra of GDP-4″-keto-6″-deoxy-mannose. (b) The H6″ doublets remain unchanged during incubation with the E157Q variant, indicating the absence of deuterium incorporation at C5″. (c) Deuterium incorporation at C5″ catalyzed by hGMD and spontaneous formation of the 4″-diol form of GDP-4″-keto-6″-deoxy-mannose in aqueous solution (here D2O). At equilibrium, the 4″-keto and 4″-diol forms are present at a ratio of about 3:1.

Table 1

Kinetic Parameters and NADPH Content of hGMD and Variants Thereof

enzyme	kcat [s–1]a	KM [μM]a	NADPH content (%)	deuterium incorporation [s–1]d
WT	0.42 ± 0.08	8 ± 1	3b/n.d.c	0.28 (fast)
Y179F	n.d.	n.d.	70/n.d.	n.d
E157Q	n.d.	n.d.	49/n.d.	n.d
S156D	n.d.	n.d.	15/n.d.	0.13 (medium)
S156A	0.05 ± 0.01	127 ± 5	8/n.d.	0.01 (slow)

From initial rate measurements at 37 °C.

Mol NADPH/mol enzyme subunit, expressed in percent, in the enzyme as isolated.

and in the enzyme at steady state during reaction with GDP-Man.

. For the methods used, see the Supporting Information.

Deuterium exchange rates measured with in situ 1H NMR spectroscopy in D2O at 30 °C and determined from the integrated signal of 6H″ in GDP-4″-keto-6″-deoxy-mannose, GDP-4″-diol-6″-deoxy-mannose, GDP-4″-keto-5″-deutero-6″-deoxy-mannose and GDP-4″-diol-5″-deutero-6″-deoxy-mannose. n.d., not detectable.

Having shown that hGMD represents, in a most elementary form, Gerlt and Gassman’s chemical principle of catalytic β-elimination of a ketone (cf. Figure b and Figure c), we investigated a conspicuous structural variation in sugar 1,4-dehydratase active sites: dTDP-glucose dehydratase has an aspartic acid residue[13,23,24] in place of Ser156 in hGMD (Figure S10). We find that the S156D variant of hGMD is inactive toward GDP-mannose but promotes C5″ deuterium exchange in GDP-4″-keto-6″-deoxy-mannose at half the rate of the wildtype enzyme (Figure S28, Table ). The structure of S156D in complex with GDP-mannose (2.8 Å; Figure e, Table S1) shows electron density for the substrate and for 2′-phospho-ADP-ribose (a cleavage product of NADP+). Lacking coordination with Asp156, the C6″ hydroxy group of GDP-mannose adopts a pseudoaxial position which conflicts with the presence of the nicotinamide ring. Although redox chemistry is thus rendered impossible for the S156D variant, its Tyr179 and Glu157 are in plausible positions to catalyze enol formation in the product. We additionally show that a S156A variant retains activity with GDP-mannose (12% of wildtype, Figure S26) but is much slower (13-fold) in C5″ deuterium exchange (Figure S29, Table ) than the S156D variant. It is also significant that hGMD turns over at a ∼10-fold slower rate (kcat = 0.42 s–1; Table , Figure S25) than dTDP-glucose dehydratase does.

In hGMD, the reduced NADPH form of the enzyme is not detectably present in the reaction at steady state (Figure S18). The kcat is therefore limited by a substrate oxidation that either is slow intrinsically or occurs kinetically coupled to a subsequent step, likely the enolization, that leads to an internal equilibrium far on the side of enzyme-bound GDP-mannose. Although beyond the scope of the current investigation, kinetic isotope effects could be useful to distinguish between these kinetic scenarios for hGMD. In contrast to hGMD, the 4.9 s–1kcat of dTDP-glucose dehydratase is limited partly by dTDP-4″,5″-glucosene reduction[25] and the reaction involves enzyme-NADH (45% of total)[27] at steady state. On the basis of these considerations, we think that nature’s mechanistic rationale for having Asp instead of Ser in the 4,6-dehydratase catalytic apparatus might have been to drive the oxidation-enolization by making proton abstraction from C5″ to the enzyme thermodynamically more favorable, as shown in Figure S30. With the pKa of the catalytic Glu increased in the presence of the neighboring Asp, as analogously observed in other enzymes,[39−41] the energetic barrier to the enol formation might be reduced effectively. A speeding up of the catalytic reaction thus achievable might benefit the enzyme function in particular physiological contexts.

In summary, therefore, this mechanistic account of hGMD advances the detailed understanding of hexose dehydration by a class of sugar 1,4-dehydratases. This is broadly relevant regarding the enzymology of 6-deoxy-sugar biosynthesis. It also provides important insight into the evolution of enzyme structure, function, and mechanism in a superfamily wide context of short-chain dehydrogenases/reductases.