Inhibition of Guanosine Monophosphate Synthetase by the Substrate Enantiomer l-XMP

Studies with mirror-image l-enantiomer nucleosides and nucleotides have revealed relaxed enantioselectivities of several cellular kinases and viral polymerases.^[1,2] This feature of enzyme-ligand molecular recognition has been exploited in the design of efficacious antiviral l-nucleoside drugs, which have lowered host cell toxicity.^[3-6] For example, lamivudine (2′,3′-dideoxy-3′-thiacytidine, 3TC), an l-nucleoside drug, exploits the relaxed enantioselectivity of HIV reverse transcriptase to inhibit viral replication.^[7] Conversely, the enantioselectivities of the majority of nucleotide biosynthesis enzymes have not been characterized. The depletion of cellular nucleotide pools has been shown to result in antiproliferative, antibacterial, and immunosuppressive effects.^[8-11]

GMP Synthetase (GMPS), an enzyme in de novo nucleotide biosynthesis, catalyzes the amination of xanthosine 5′-monophosphate (XMP) to guanosine 5′-monophosphate (GMP) in the presence of glutamine (the amine source) and ATP.^[10,12] GMPS possesses two active sites that are separated by approximately 30 Å, suggesting that GMPS undergoes a significant conformational change during catalysis.^[13,14] In the amidotransferase active site, a glutamine residue is hydrolyzed to liberate ammonia, which subsequently functions as the nucleophile in the amination of XMP (Equation 1).^[15] In the synthetase active site, the 2-carbonyl of XMP is adenylated with ATP (Equation 2) to activate the aromatic ring for subsequent aminolysis (Equation 3).^{[10,12,13,16]} Formation of adenyl-XMP is believed to trigger glutamine hydrolysis in the amidotransferase active site.^[12]

$$\text{Glutamine}+{\mathrm{H}}_2\mathrm{O}\stackrel{\mathrm{GMPS}}{\to }\text{Glutamate}+{\mathrm{NH}}_3$$ (1)

$$\mathrm{XMP}+\mathrm{ATP}\stackrel{\mathrm{GMPS}}{\rightleftarrows }\mathrm{adenyl}-\mathrm{XMP}+\mathrm{PPi}$$ (2)

$$\mathrm{adenyl}-\mathrm{XMP}+{\mathrm{NH}}_3\stackrel{\mathrm{GMPS}}{\to }\mathrm{GMP}+\mathrm{AMP}$$ (3)

A crystal structure of E. coli GMPS has been solved that reveals a large solvent-accessible synthetase pocket with considerable surface area.^[13] Several base-modified d-XMP analogues have been shown to function as substrates for GMPS and be converted to their amine derivatives,^[17] and non-hydrolyzable adenyl-XMP analogues have been synthesized.^[18] Based on this structural information and our interest in characterizing for the first time the enantioselectivity of GMPS, we hypothesized that l-XMP, the enantiomer of native ligand d-XMP, could target GMPS and modulate enzymatic activity. We hypothesized that l-XMP could incorporate into the synthetase active site and inhibit enzyme function, or less likely, l-XMP could function as a substrate for GMPS and undergo aminolysis to yield l-GMP. In either case, enzymatic synthesis of d-GMP would be affected, either by direct enzyme inhibition or by the activity of a suicide substrate. Given the central importance of GMPS in eukaryote and prokaryote biochemistry, we examined the enantioselectivity of the enzyme.

Preceding this work, a synthesis of l-XMP (6), the enantiomer of natural ligand d-XMP, had not been reported. Our synthesis of l-XMP (6) started from l-arabinose, which was elaborated to 1-O-acetyl-2,3,5-tri-O-benzoyl-β-l-ribofuranoside (1) by reported methods (Scheme 1).^[19] A Vorbrüggen coupling with trimethylsilyl protected xanthine 2 gave a separable mixture protected l-xanthosine isomers 3 (N⁹ isomer) and 4 (N⁷ isomer, not shown).^{[20, 21]} Deprotection of the benzoyl protecting groups of 3 using ammonia afforded l-xanthosine (5). Selective phosphorylation of the 5′-OH of 5 utilizing phosphorous oxychloride gave l-XMP (6).^[22]

Scheme 1.

Synthesis of L-XMP: Reagents and conditions: (i) TMSOTf, DCE, reflux, 70% (for 3), 21% (for N⁷-isomer 4, not shown); (ii) NH₃, MeOH, 55 °C (sealed tube), 93%; (iii) POCl₃, PO(OMe)₃, Proton-Sponge; aq. TEAB, 19%.

E. coli GMPS was overexpressed and purified (Figure S1) and an HPLC-based assay was developed to quantitate enzymatic reaction products. GMPS was incubated with test substrates and NH₄OAc (ammonia source), and the reaction was terminated at various time points by addition of EDTA. GMPS protein was then removed by a molecular weight spin-column (30 kDa), and enzymatic reaction products were analyzed by reverse-phase HPLC. Surprisingly, we found that incubation of l-XMP and GMPS yielded a new peak of identical retention time as d-GMP (Figure 1). Characterization of this new peak (MS analysis, Figure S3) confirmed that l-XMP was converted to l-GMP by GMPS, demonstrating that turn over of the opposite enantiomer substrate was possible.

Figure 1.

HPLC analysis of L-XMP conversion to L-GMP by GMPS.

Biochemical characterization of the kinetics of l-XMP conversion to l-GMP by GMPS, as well as d-XMP conversion to D-GMP, was measured by fitting the individual GMP/CMP (cytosine 5′-monophosphate, an external standard) ratios from each sample into the slope-intercept equation from the calibration plot. Initial velocity measurements of GMP production as a function of time were measured at a variety of substrate (XMP) concentrations, and at fixed saturating concentrations of the non-varied substrates ATP and NH₄OAc (Figure 2A). Fitting these data to the Michaelis-Menten equation and analysis by non-linear regression allowed measurement of kinetic parameters (Table 1).

Figure 2.

Initial velocities versus substrate plots for D-XMP and L-XMP as measured by A) HPLC analysis and B) UV-Vis analysis.

Table 1.

Kinetic parameters of GMPS enzymatic activity in the presence of enantiomeric substrates d-XMPand l-XMP.^[a]

Substrate	kcat (s−1)	Km (μm)	kcat/Km (μm−1 s−1)
d-XMP[b]	4.8 (± 0.3) × 10−2	35.3 ± 8.5	1.4 × 10−3
d-XMP[c]	4.3 (± 0.7) × 10−2	24.9 ± 6.6	1.7 × 10−3
l-XMP[b]	3.2 (± 0.3) × 10−6	316.7 ± 55.6	1.0 × 10−8
l-XMP[c]	3.1 (± 0.7) × 10−6	329.9 ± 104.9	9.4 × 10−9

^[a]Performed in triplicate. Values shown are the average +/- S.D.

^[b]HPLC analysis.

^[c]UV-Vis analysis.

Analysis of d-XMP revealed an apparent K_m of 35.3 μm, which was similar to previously reported K_m values for E. coli GMPS (29 μm and 166 μm).^{[23, 24]} The turnover number (k_cat) was found to be 4.8×10^–2 s^–1, which was comparable to a previous report of 9.4×10^–2 s^–1.^[24] Analysis of l-XMP revealed an apparent K_m of 316.7 μM, which is ~10-fold higher than the natural enantiomer. Surprisingly, the k_cat was measured at 3.2×10^–6 s^–1, which is a 15,000-fold difference in turnover number compared to d-XMP. The specific activity (k_cat/K_m) of l-XMP decreased 140,000-fold from the natural enantiomer (1.4×10^–3 μm^–1 s^–1 for d-XMP versus 1.0×10^–8 μm^–1 s^–1 for l-XMP). These results suggest that l-XMP may also inhibit GMPS.

To confirm the values derived from the HPLC assay, a known continuous UV spectrophotometric assay was also employed.^[23,25] This assay monitors a reduction in 290 nm absorbance resulting from conversion of XMP (ε₂₉₀ = 4800 m^–1 cm^–1) to GMP (ε₂₉₀ = 3300 m^–1 cm^–1). UV-based kinetic data was calculated analogously to the HPLC-derived data (Figure 2B). Analysis of both d-XMP and l-XMP revealed nearly identical results to the HPLC assay (for l-XMP: k_cat/K_m = 9.4×10^–9 μm^–1 s^–1 (UV) versus 1.0×10^–8 μm^–1 s^–1 (HPLC); Table 1).

Although l-XMP conversion to l-GMP by GMPS was demonstrated, the weak affinity of l-XMP for GMPS, coupled with its slow turnover number, suggested possible enzyme inhibition by this ligand. To probe for GMPS inhibition, we performed enzymatic activity experiments with our xanthosine-based molecules to demonstrate reduction of GMPS-mediated amination of d-XMP. Addition of a fixed concentration of inhibitor to varying d-XMP concentrations, followed by analysis by UV-Vis and fitting to the competitive inhibition equation (or uncompetitive for decoyinine towards XMP),^[10] yielded the K_i data shown in Table 2. Evaluation of the known GMPS uncompetitive inhibitor decoyinine revealed a K_i = 54.1 μm, which was similar to a previous report (26 μm).^[26] Mizoribine (bredinin trade name), a known immunosuppresive drug and competitive GMPS inhibitor, was found to be more potent in our hands (K_i = 1.8 μM), and this activity is similar to reports of the same compound against GMPS isolated from rat Walker sarcoma cells (K_i = 10 μM).^[11] l-XMP (6) inhibition results were quite interesting, revealing that l-XMP is almost 7-fold more potent than decoyinine inhibition against E. coli GMPS (K_i = 7.5 μM). Both d-xanthosine and l-xanthosine nucleosides were also tested and neither molecule inhibited GMPS, suggesting that 5′-monophosphorylation is required for inhibition. Mizoribine does not require phosphorylation for GMPS inhibition. Our results suggest that l-XMP can inhibit GMPS enzymatic activity with potency similar to or slightly better than other known inhibitors.^{[9, 26-28]}

Table 2.

GMPS inhibition data of known inhibitors and xanthosine analogues.^[a]

Inhibitor	Ki (μm)
Decoyinine	54.1 ± 14.5
Mizoribine	1.8 ± 0.7
d-Xanthosine	> 1500
l-Xanthosine	> 500
l-XMP	7.5 ± 1.8

^[a]Performed in triplicate. Values shown are the average +/- S.D.

To understand the molecular interactions of d-XMP and l-XMP within the GMPS active site, energy minimized three-dimensional conformations of the biochemical reaction intermediates (adenyl-d-XMP and adenyl-l-XMP; Figure S4B) were docked (Surflex-dock in the SYBYL software suite) into the crystal structure of E. coli GMPS (PDB 1GPM).^[13] The molecule of AMP observed in the x-ray crystal structure was extracted and re-docked into GMPS with a calculated similarity of 0.908 (1.0 is the theoretical maximum), showing reliability in docking accuracy (Figure S4A). Several stabilizing molecular interactions were observed between adenyl-d-XMP and GMPS, such as hydrogen bonds between Lys856 and the xanthine nucleobase; Arg875, Arg765, and Glu768 to ribose alcohols; and Asn761 to the phosphate of d-XMP (Figure 3A). The considerable size of the GMPS synthetase pocket readily accommodated the docking of adenyl-l-XMP; however, the l-ribose sugar occupied a substantially different position within the synthetase domain (Figure 3B). No longer present were many of the key molecular interactions between the nucleobase and ribose alcohols as evident by a decreased in consensus score, which is an estimate of the overall ligand binding affinity (CScore = 7.68 for d-XMP versus 6.16 for l-XMP). One compensating molecular interaction was observed for adenyl-LXMP, which was a hydrogen bond between Asn761 to a ribose alcohol. The conformation of the l-ribose sugar in adenyl-l-XMP also forces C² of the nucleobase to be positioned approximately 2.0 Å away from the region in space occupied by natural adenyl-d-XMP ligand. This perturbation to nucleobase conformation may deter aminolysis of the adenylated unnatural monophosphate, thereby slowing enzyme turnover. Additionally, the loss of key hydrogen-bonding interactions may also contribute to the loss in enzyme efficiency. Nonetheless, our observation of the synthesis of l-GMP from l-XMP implies the large size of the synthetase pocket must allow some movement of adenyl-l-XMP to obtain the correct conformation for amination.

Figure 3.

Docking of A) adenyl-D-XMP and B) adenyl-L-XMP intermediates into the E. coli GMPS synthetase binding pocket. Key hydrogen-bonding interactions between the adenylated ligands and GMPS are marked with dashed lines. The 2-position of the xanthine nucleobase is marked (2.0 Å conformational shift).

In conclusion, the synthesis of l-GMP from l-XMP provides new insight into the substrate promiscuity of GMPS. GMPS was also inhibited by l-XMP at low micromolar levels, which is comparable to other known inhibitors. These results provide new insight into GMPS-ligand interactions that will be useful for future inhibitor designs.