Structural Investigation of Park’s Nucleotide on Bacterial Translocase MraY: Discovery of Unexpected MraY Inhibitors

Systematic structural modifications of the muramic acid, peptide, and nucleotide moieties of Park’s nucleotide were performed to investigate the substrate specificity of B. subtilis MraY (MraYBS). It was found that the simplest analogue of Park’s nucleotide only bearing the first two amino acids, l-alanine-iso-d-glutamic acid, could function as a MraYBS substrate. Also, the acid group attached to the Cα of iso-d-glutamic acid was found to play an important role for substrate activity. Epimerization of the C4-hydroxyl group of muramic acid and modification at the 5-position of the uracil in Park’s nucleotide were both found to dramatically impair their substrate activity. Unexpectedly, structural modifications on the uracil moiety changed the parent molecule from a substrate to an inhibitor, blocking the MraYBS translocation. One unoptimized inhibitor was found to have a Ki value of 4 ± 1 μM against MraYBS, more potent than tunicamycins.

Peptidoglycan is a polymer consisting of sugars and amino acids that forms the bacterial cell wall. Interrupting the biosynthesis of peptidoglycan can devastate bacterial growth and survival due to the critical role it plays in maintaining cell shape and protecting bacteria from internal osmotic pressure 1,2 . One of the enzymes involved in bacterial cell wall biosynthesis, MraY is an integral membrane protein that catalyzes the transfer of the monophospho-MurNAc-pentapeptide moiety from Park's nucleotide (UDP-MurNAc-pentapeptide) onto the undecaprenyl phosphate, to give Lipid I with concomitant release of UMP (Fig. 1). MraY is an attractive antibacterial target being essential for bacterial growth; highly conserved across many bacterial species; and without a eukaryotic counterpart [3][4][5][6] .
One major class of MraY inhibitors, known as nucleoside antibiotics, shares a uridine nucleoside as a common moiety with Park's nucleotide [6][7][8][9][10] . Accordingly, an understanding of the interactions between Park's nucleotide and MraY might be useful for the design of new MraY inhibitors. Recent disclosure of an apo crystal structure of MraY from Aquifex aeolicus (MraY AA ) shows the overall architecture of this interesting enzyme 11 . However, due to the lack of available complex crystal structure, detailed mechanisms or interactions between substrates or inhibitors toward MraY remain to be explored. Although some brief substrate studies of Park's nucleotide toward MraY have been reported, their scope is limited to the structural diversity accessible by biocatalysis 12 . Obviously, the substrate study of MraY is hampered by difficulties to acquire the structurally complex substrates. Chemical synthesis seems to be the most straightforward approach towards the generation of pure and systematically modified samples of various desired molecules for testing against MraY.
To more thoroughly investigate how structural modification of Park's nucleotide affects MraY substrate recognition, we first sought to identify a proper polyprenyl phosphate substrate that would be conserved for all the Park's nucleotide analogues tested. In our preliminary HPLC-based MraY activity study, NBD-Park's nucleotide 6 was completely consumed in 1 h when undecaprenyl phosphate (C55P) was applied as a polyprenyl phosphate substrate in our hands (Supplementary Figure 1) 13 . In contrast, other polyprenyl phosphates with a shorter length or different configurations still can be recognized as a MraY substrate but their substrate activity is much weaker than undecaprenyl phosphate (C55P) (Supplementary Table 1). Our observation of this broad substrate specificity of MraY is consistent with previous studies in the combined MraY-MurG system or membrane fractions containing both MraY and MurG [14][15][16] . According to our results, C55P was chosen as the substrate coupling partner for all the Park's nucleotide analogues studies, and the substrate activity was measured after 1 h reaction for convenient purposes. Moreover, it was decided not to modify the pyrophosphate group as it is at this position that translocation occurs.
Herein, we describe the systematic preparation of Park's nucleotides with varying three parts including the peptide, N-substituted muramic acid, and uridine moieties for evaluation as MraY BS substrates (Fig. 2). This information will provide us with the essential moieties and the specificity requirements of the MraY for Park's nucleotide analogues, as an effort toward development of new inhibitors.

Results and Discussions
Preparation of Park's nucleotide analogues and evaluation of their substrate activity. As shown in Fig. 3, O-debenzylation of 1 followed by a phosphorylation and phosphitylation/oxidation sequence gave the phosphate 2 in 71% yield over three steps 17 . Compound 4 was obtained via the debenzylation of 2. Finally, conjugation of 4 with activated UMP-morpholine-N,N'-dicyclohexyl carboxamidine salt and global deprotection under basic conditions gave Park's nucleotide 9 in 69% yield. For the preparation of 5, selective deprotection of the trimethylsilyl ethyl ester (TMSE) in 2 by treatment with TBAF in THF, followed by  coupling with H-d-iso-Glu(OMe)-l-Lys-(TFA)-d-Ala-d-Ala(OMe) and debenzylation gave the corresponding 3 in 67% yield over three steps. Compound 3 was then coupled with activated UMP-morpholine-N,N'dicyclohexylcarboxamidine salt, followed by global deprotection under basic conditions gave Park's nucleotide 5 in 35% yield over two steps. A fluorescent probe 6 was prepared from 5 by conjugating a nitronbenzoxadiazole (NBD) fluorophore at the terminal amine site of lysine on the peptide stem in 88% yield. Compounds 7 and 8 were similarly prepared (Fig. 3).
The substrate activity study of 5-10 toward MraY BS was performed using the HPLC-based MraY functional assay. Substrate consumption curves of 5-10 were shown in Fig. 4A. Compounds 5-8 were recognized as a MraY BS substrate, but 9 and 10 were not. The similar curves of 5 and 6 suggest that the NBD-fluorophore attaching to the side chain of Lys on the pentapeptide stem of Park's nucleotide does not cause any significant effect on its substrate activity ( Supplementary Figures 1 and 2). Compound 7, lacking the terminal two amino acids (d-Ala-d-Ala), was only slightly less active than 5 (17% activity reduced after 1 h reaction, Fig. 4B), showing that the d-Ala-d-Ala moiety is not essential for MraY BS recognition. The previous study reported by Hammes and Neuhaus pointed out that 7 is a much weaker substrate than 5 when intrinsic membrane fractions are used as a source of lipidphosphate and enzyme 12 . In our conditions, only the purified enzyme and two pure substrates were utilized, and Park's nucleotide analogue was the limiting reagent compared to the other substrate C55P. Both individual studies show different degrees of the substrate activity loss that might be attributed to several factors such as enzyme activity, substrate ratio and assay platform. Compound 8, similar to 5 but lacking the terminal three amino acids, was a weak substrate (40% activity remained after 1 h reaction, Fig. 4B). Moreover, 9 (bearing only one amino acid (l-Ala)) and UDP-GlcNAc (10) were not substrates under these assay conditions, showing that this 3-O-lactyl-tripeptide (d-Lac-l-Ala-γ -d-Gln-l-Lys) moiety in Park's nucleotide is important for the MraY BS catalyzing process.
Next, more subtle structural changes of Park's nucleotide 5 were proposed, and the resultant molecules conjugated with a NBD fluorophore on the peptide stem for easy monitoring (Fig. 5A). All analogues except 17 were synthesized in a manner similar to that for 5. Initial attempts to prepare 17 by coupling of 3 and morpholine-activated 5-amino-uridine-5′ -monophosphate in the presence of 1H-tetrazole were not successful. Most of the morpholine-activated 5′ -NH 2 -UMP was found to degrade into 5′ -NH 2 -UMP, and only trace among of product was detected in the reaction mixure 18 . To overcome this problem, the synthetic strategy was re-designed to entail activation of the sugar moiety with the carbonyl diimidazole (CDI) instead of activation of 5-amino-uridine-5′ -monophosphate, followed by global deprotection and the NBD labeling 19 . In this way, 17 was obtained in a yield of 31% over four steps (see also Supplementary Methods).
As illustrated in Fig. 5B, both N-glycolyl 12, the natural substrate for mycobacterial MraY (also called MurX), and unnatural N-glycinyl 13 had similar substrate activity to 6, indicating that there are no extra interactions, such as additional hydrogen bonds, to increase the activity between the N-substituent moiety on muramic acid of Park's nucleotide analogues and MraY BS 20 . Analogue 14 (R 4 = H) had similar activity to 6, suggesting the methyl group on the lactate moiety to be unessential 21 . Likewise, 15 (R 5 = H) was slightly less active than 6 (about 80% relative activity after 1 h reaction, Fig. 5C) 12 . Surprisely, 16 (R 6 = H) was found to be a very poor substrate compared to 6 (< 10% relative activity after 1 h reaction, Fig. 5C), showing the acid group attached to the Cα of iso-d-Glu moiety in Park's nucleotide to be critical.
In order to evaluate the effects of peptide moiety of Park's nucleotide, investigation of the binding affinity of 5, 6, 9 and 16 was performed using a biolayer interferometry-based binding (BLI) assay. Initial attempts to perform the MraY BS binding assay in the presence of both substrates (C55P and Park's nucleotide analog) didn't work properly because a strong non-specific binding signal was observed; presumably, the hydrophobic part of C55 might mainly contribute this non-specific interaction 22 . To simplify the assay conditions, only Park's nucleotides were utilized to measure the binding affinity with MraY BS. As shown in Fig. 6, compounds 5 and 6 exhibited similar binding affinity with K D values of 120 and 127 μ M, respectively. This suggests the NBD tag in 6 does not affect the binding affinity with MraY BS , and this observation is consistent with the substrate activity result in Fig. 4. Structurally, 9 is the truncated form of 5 (lacking the outermost four amino acids, including iso-d-Glu); and 16 is very similar to 6 -the only difference being removal of the acid group attached to the Cα of iso-d-glutamic acid (R 6 = H). However, 9 and 16 showed no proper binding affinity with MraY BS -only a very low binding signal was detected, even at concentrations up to 500 μ M. Our results indicate that the acid moiety (R 6 = COOH) on iso-d-Glu of Park's nucleotide plays an important role for both binding affinity and substrate activity. In addition, 11 (R 1 = OH/R 2 = H) and 17 (R 7 = NH 2 ) did not function as substrates, even under extreme reaction conditions, showing the equatorial hydroxyl group at R 2 position to be critical and the modifications at R 7 position not tolerated. Construction of the molecular model of the Park's nucleotide-MraY BS complex. Based on these preceding results, as well as our mutagenesis (Table 1) and computational modeling studies, a putative Park's nucleotide binding site on MraY BS is proposed (Fig. 7) 23,24 . As illustrated in Fig. 7A, Park's nucleotide 5 could specifically interact with the W297, K102, and Q271 of MraY, and the phosphate group of C55P. As shown in Table 1, the enzyme activities of four MraY BS mutants, including T53A, K102A, Q271A, and W297A, were significantly decreased, suggesting these residues to be important for enzyme activity. All four mutants had higher K M values compared with the wild-type MraY BS (K M = 18 μ M and NBD-Park's nucleotide 6 applied as a substrate), showing that the mutations caused a loss of binding affinity. The highly conserved threonine (T53) located on loop A is close to the proposed catalytic pocket (Fig. 7B), and may participate in the enzyme process. In addition, the uracil moiety is embedded in a deep grove, which may interact with W297 on loop E of MraY BS . Our substrate specificity and site-directed mutagenesis study strongly suggest that Q271 on MraY BS might interact with the iso-d-glutamic acid of Park's nucleotide through a hydrogen bond to stabilize the peptide chain.

Discovery of Park's nucleotide analogues bearing modifications at the uracil 5-position as
MraY BS inhibitors. We were curious whether analogues 11 (modified at R 1 /R 2 ) and 17 (modified at R 7 )neither of which were active substrates -could inhibit the function of MraY BS . To further evaluate the role of positions at R 1 , R 2 and R 7 , we re-designed and synthesized 18-21 with a truncated peptide (Fig. 8). The inhibitory activity of each of these was determined using a fluorescent enhancement assay against MraY BS , with tunicamycins as reference (Supplementary Figure 5). As shown in Table 2, 18 (R 7 = NH 2 ) had no inhibition activity, but 19 (R 7 = NHAc) showed weak inhibitory activity (K i = 764 μ M) against MraY BS ; and 20, bearing a p-tolylacetamide moiety at R 7 , became a more potent inhibitor (K i = 11 μ M) -strongly indicating that appropriate N-substitution (R 7 ) can enhance inhibitory activity. However, 21, the C4-hydroxyl epimer of 20, was a very poor inhibitor (30% inhibition at 1 mM). This finding implied the C4-hydroxyl epimerization of Park's nucleotide dramatically impairs both the substrate and inhibitory activities. To improve the inhibitory activity, we reinstalled the tetrapeptide      orientation of the C4-hydroxyl group of muramic acid plays an essential role in molecule-enzyme recognition.
Notably, the binding affinity of 22 (K D = 86 μ M) was approximately 2.3 fold stronger than that of 20, which can be attributed to the contribution of the tetrapeptide moiety.
The antibacterial activities of 20 and 22 were also investigated and the minimal inhibitory concentrations (MIC) against S. aureus and B. subtilis were determined using standard-broth dilution methods 25 . Unfortunately, both 20 and 22 showed no antibacterial activity, even at a high concentration of 200 μ M. It may be because the compounds containing the highly charged pyrophosphate moiety were difficult to penetrate the bacterial cytoplasmic membrane 26 . In order to improve the antibacterial activity, finding a surrogate to replace the pyrophosphate moiety on 22 remains to be explored.

Conclusions
A series of Park's nucleotide analogues with modifications at the peptide, muramic acid, and nucleotide moieties has been designed and synthesized, and their MraY BS substrate activity and specificity were evaluated. Our results led to several important findings: (1) the first two amino acids (l-alanine-iso-d-glutamic acid) of the oligopeptide chain are essential for MraY BS recognition; (2) the configuration of the C4-OH on muramic acid is important for MraY BS substrate specificity; and (3) modifications at the 5-position of the uracil dramatically impair the substrate activity (Fig. 9). Also, the substrate specificity data together with mutagenesis and computational modeling studies allowed us to infer a putative Park's nucleotide binding site on MraY BS .
Unexpectedly, analogues bearing modifications at the 5-position of the uracil were found to be MraY BS inhibitors though these molecules contain a pyrophosphate moiety. Of these, an unoptimized inhibitor 22 (K i = 4 μ M) was found to be roughly twice as potent as tunicamycins (K i = 9 μ M) against MraY BS , the first example of Park's nucleotide-based inhibitors. However, our results not only allow us to infer the minimal structure requirements of Park's nucleotide as a MraY substrate, but also illuminate a new direction for MraY inhibitor design. And more generally, the concrete nature of these conclusions validate our strategy of systematic substrate structure modifications for the elucidation of enzyme binding site mapping, for membrane proteins without available co-crystal structures. Investigation of the MraY mechanisms, and development of more potent inhibitors with in vivo antibacterial activity are currently ongoing in our laboratory.

Methods
General. All chemicals were obtained from commercial suppliers and used without further purification. All solvents were anhydrous grade unless indicated otherwise. All non-aqueous reactions were performed in ovendried glassware under a slight positive pressure of argon unless otherwise noted. Reactions were magnetically stirred and monitored by thin-layer chromatography on silica gel. Flash chromatography was performed on silica gel of 40-63 μ m particle size. Concentration refers to rotary evaporation. Yields are reported for spectroscopically pure compounds. NMR spectra were recorded on dilute solutions in D 2 O, CDCl 3 and CD 3 OD on Bruker AVANCE 600 at ambient temperature. Chemical shifts are given in δ values and coupling constants J are given in Hz. The splitting patterns are reported as s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), and dd (double of doublets). High resolution ESI mass spectra were recorded on a Bruker Daltonics spectrometer. The reaction was allowed to warm to RT over a period of 1 h. The mixture was diluted with CH 2 Cl 2 (50 mL), and extract with water (20 mL × 2). The organic layers were collected, dried over MgSO 4 , concentrated, and purified     After stirring for 4 h, the reaction was neutralized by 1.0 N HCl (aq) , concentrated, and purified by cc (iPrOH/NH 4 OH (aq) = 2/1, silica gel) to give 5 as white solid (32 mg, 0.027 mmol, 35% over two steps). 1