Insights into the Target Interaction of Naturally Occurring Muraymycin Nucleoside Antibiotics

Abstract

Muraymycins are a subclass of antimicrobially active uridine-derived natural products. Biological data on several muraymycin analogues have already been reported, including some inhibitory in vitro activities towards their target protein, the bacterial membrane enzyme MraY. However, a structure-activity relationship (SAR) study on naturally occurring muraymycins based on such in vitro data has been missing so far. In this work, we report a detailed SAR investigation on representatives of the four muraymycin subgroups A–D using a fluorescence-based in vitro MraY assay. For some muraymycins, inhibition of MraY with IC₅₀ values in the low pM range was observed. These inhibitory potencies were compared with antibacterial activities and were correlated to modelling data derived from a previously reported X-ray crystal structure of MraY in complex with a muraymycin inhibitor. Overall, these results will pave the way for the development of muraymycin analogues with optimized properties as antibacterial drug candidates.

Natural products revisited

Naturally occurring muraymycin nucleoside antibiotics were studied for inhibitory activities towards their target protein, the bacterial membrane enzyme MraY. An in vitro MraY assay furnished SAR data on several representative muraymycins. These inhibitory potencies were correlated to antibacterial activities and to modelling data, thus providing detailed insights into the target interaction of muraymycins.

Resistance against clinically used antibiotics continues to emerge and represents a significant challenge for human healthcare.^[1] Therefore, research on novel antimicrobial agents with new or as of yet unexploited modes of action is highly relevant. The bacterial membrane protein MraY (translocase I) represents such a potential new target.^[2] MraY is an enzyme involved in the intracellular membrane-associated stages of cell wall (peptidoglycan) biosynthesis. It catalyzes the transformation of the cytosolic precursor UDP-MurNAc-pentapeptide (‘Park’s nucleotide’) 1 with the membrane anchor undecaprenyl phosphate, yielding the membrane-bound biosynthetic intermediate lipid I 2 (Scheme 1).^[3,4]

Scheme 1.

MraY-catalyzed reaction of Park’s nucleotide 1 towards lipid I 2 (undecaprenyl phosphate represented schematically; UDP = uridine diphosphate; UMP = uridine monophosphate). The exact composition of the pentapeptide moiety can vary in different bacteria.^[4]

Methods for the overexpression of MraY, including cell-free approaches, have been established.^[5] Following a previously described topology model,^[6a] the first ligand-free X-ray crystal structure of MraY (from the extremophile Aquifex aeolicus) confirmed the enzyme to be an integral membrane protein with ten transmembrane helices and five cytoplasmic loops forming the active site.^[6b] Mechanistic studies on MraY including mutagenesis have not been fully conclusive.^[7]

Naturally occurring uridine derivatives (‘nucleoside antibiotics’, e.g., muraymycins, caprazamycins, liposidomycins, capuramycins and mureidomycins) and their analogues are potent inhibitors of MraY and therefore attractive candidate structures for antimicrobial drug development.^[8] Muraymycins (e.g., compounds 3–8, Table 1) were discovered as a collection of 19 structurally related, secondary metabolites from Streptomyces sp. in 2002.^[9] The muraymycin scaffold consists of a (5′S,6′S)-glycyluridine (GlyU) core structure, which is connected to a peptide moiety by an alkyl linker. In most muraymycins, this core is 5′-O-aminoribosylated to generate an unusual disaccharide. Muraymycins have been categorized into four subgroups A–D, based on the chemical modification of the central L-leucine, with groups A–C containing (3S)-3-hydroxy-L-leucine at this position. A-type muraymycins (e.g., A1 3) are 3-O-acylated with ω-functionalized fatty acids terminating with a guanidinium functionality. B-series congeners (e.g., B2 4 and B6 5) are 3-O-acylated with unfunctionalized, branched fatty acids. C-type muraymycins (e.g., C1 6) completely lack the 3-O-acyl unit. Finally, members of the D-series (e.g., D1 7 and D2 8) contain an unmodified L-leucine residue at the central position (Table 1). First insights into the biosynthesis of these antibacterial natural products have already been obtained and are summarized elsewhere.^[8c]

Table 1.

Inhibition data (from in vitro MraY assays) and antibacterial activities of naturally occurring muraymycins 3–8 as well as of synthetic analogue 9.

muraymycin	R1	R2	IC50 (MraY S. aureus) [nM][a]	IC50 (MraY Aquifex aeolicus) [nM][b]	IC50 (MraY C. bolteae solubil./crude) [nM][c]	MIC (S. aureus) [μg/mL][d]	MIC (E. coli ΔtolC/DH5α) [μg/mL][e]
A1 3			0.027 ± 0.003	0.11 ± 0.02	0.16 ± 0.04 0.017 ± 0.004	18	2 > 32
B2 4			0.010 ± 0.002	0.11 ± 0.01	0.14 ± 0.01 0.0083 ± 0.0006	> 32	1 > 32
B6 5			0.021 ± 0.002	0.19 ± 0.02	0.38 ± 0.03 0.036 ± 0.004	> 32	1 > 32
C1 6	OH		0.016 ± 0.002	0.35 ± 0.04	0.45 ± 0.17 0.093 ± 0.020	> 32	1 > 32
D1 7	H		0.48 ± 0.13	99 ± 30	380 ± 110 40 ± 6	> 80	1 > 32
D2 8	H		0.39 ± 0.11	46 ± 13	200 ± 30 25 ± 4	> 80	1 > 32
analogue 9	OH	H	95 ± 19	(50% @100 μM)[f]	58% @100 μM)[f] 860 ± 300	> 50	50 15

^[a]Fluorescence-based in vitro assay with MraY from S. aureus in crude membranes, IC₅₀ ± SD.

^[b]Fluorescence-based in vitro assay with MraY from Aquifex aeolicus in solubilized and purified form, IC₅₀ ± SD (except where indicated).

^[c]Fluorescence-based in vitro assay with MraY from C. bolteae, first value: IC₅₀ ± SD (except where indicated) for protein in solubilized and purified form, second value: IC₅₀ ± SD for protein in crude membranes.

^[d]Minimal inhibitory concentration against growth of S. aureus subsp. aureus (Newman strain).

^[e]Minimal inhibitory concentration against growth of E. coli, first value: efflux-deficient ΔtolC mutant, second value: DH5α strain.

^[f]Percentage of inhibition at the given concentration of the inhibitor.

Several semi-synthetic and synthetic analogues of muraymycins as well as their biological activities have already been reported, thus furnishing some structure-activity relationship (SAR) data mostly for simplified muraymycin variants.^[10] Further insights into the biological action of muraymycins were recently obtained from the X-ray co-crystal structure of MraY from Aquifex aeolicus in complex with the natural product muraymycin D2.^[11] In many of the aforementioned SAR investigations on muraymycin analogues, biological potencies were mainly evaluated on the basis of antibacterial activities (i.e., minimal inhibitory concentrations (MIC)). However, antibacterial activity is the overall result of a compound’s influx into the bacterial cell, its target interaction, its intracellular stability and potential efflux phenomena.

A significantly more informative approach to study the SAR of MraY-inhibiting antibiotics would be to include an investigation into their inhibitory potencies towards MraY using an in vitro enzyme assay. This way, MIC values would provide additional information on other factors influencing antibacterial activity besides target interaction. Such an approach has only been pursued for some of the previously reported muraymycin analogues.^[10] Remarkably, even information on the MraY-inhibiting properties of the parent natural products is scarce. The originally isolated, naturally occurring muraymycins were solely assessed by their inhibition of the formation of lipid II (i.e., the biosynthetic intermediate following lipid I 2) and peptidoglycan in coupled assays, thus only indirectly supporting MraY inhibition.^[9a,b] These assays revealed inhibition of lipid II and peptidoglycan formation at concentrations as low as 0.27 μg/mL, being equivalent to the low nM range. However, quantitative information on direct MraY inhibition remains unknown.

We therefore aimed to isolate and test the MraY inhibition of the naturally produced muraymycins 3–8 as selected representatives of the four subclasses A–D (see Table 1). Using modified fermentation conditions based partly on previously reported methods,^[9a,b] Streptomyces sp. (NRRL30473 for isolation of muraymycins A1 3, B2 4 and B6 5; NRRL30475 for isolation of muraymycins D1 7 and D2 8; NRRL30477 for isolation of muraymycin C1 6) were cultivated for production. Methanol extracts of the mycelium and the water phase were subjected to size exclusion chromatography and semi-preparative HPLC to furnish muraymycins 3–8. For the in vitro evaluation of MraY inhibition, we employed a fluorescence-based MraY assay originally reported by Bugg et al.^[12] In this assay, a fluorescence-labelled derivative of Park’s nucleotide 1 (dansylated in the L-lysine side chain, Scheme 1) is employed and furnishes, upon incubation with MraY and undecaprenyl phosphate, a dansylated derivative of lipid I 2. This product shows stronger fluorescence intensity than the dansylated substrate, thus leading to an increase of fluorescence over time as a measure of MraY activity. We have recently reported the preparation of the dansylated substrate by total synthesis and a modified version of this assay with lower substrate concentration.^[12d]

All isolated natural products 3–8 were initially tested for their inhibitory potency towards two different preparations of MraY: (i) MraY from S. aureus, overexpressed in E. coli and used as crude membranes from overexpressing cells;^[5a,10f,12d] and (ii) MraY from Aquifex aeolicus, overexpressed in E. coli and purified to homogeneity in the presence of a non-denaturing detergent (modified protocol from ref.^[6b,11a]). Attempts to extract and purify the MraY homologue from S. aureus from the crude membranes led to denaturation and therefore afforded inactive protein. Activity assays were performed in triplicates, providing (after data fitting) IC₅₀ values for inhibition with standard deviations (Table 1). We had previously reported the synthetic 5′-defunctionalized (i.e., 5′-deoxy) muraymycin analogue 9 (Table 1),^[10f] which was also included in this study.

Assays with MraY from S. aureus in crude membranes revealed all tested naturally occurring muraymycins 3–8 to be picomolar inhibitors of the enzyme (Table 1). For this source of MraY activity, muraymycins A1 3, B2 4, B6 5 and C1 6 were most active with IC₅₀ values in the range of 10–27 pM. It was therefore concluded that the fatty acyl motif found in muraymycins of the A- and B-series apparently does not contribute to MraY inhibition as it is absent in congener C1 6. This finding supports a previous hypothesis that the fatty acyl unit might instead contribute primarily to the cellular uptake rather than target interaction.^[10d,13] On the other hand, muraymycin D1 7 was a ca. 30-fold less active MraY inhibitor than C1 6, although it only differs in the missing β-hydroxy group in the leucine moiety. This indicated a general beneficial role of such a β-substituent in the leucine unit for MraY inhibition, irrespective of O-acylation in this position. Inhibitory activities of muraymycins D1 7 and D2 8 against MraY from S. aureus were nearly identical, thus demonstrating that the 2-O-methyl motif found in the aminoribose unit of several naturally occurring muraymycins does not contribute to inhibition. The validity of these inhibitory activities was confirmed using the commercially available nucleoside antibiotic tunicamycin as a reference MraY inhibitor (see Supporting Information).

In contrast to the activities of natural products 3–8 in the pM range, synthetic 5′-defunctionalized analogue 9 showed an IC₅₀ value of 95 ± 19 nM against MraY from S. aureus (Table 1). This value was slightly lower than previously reported^[10f] due to differences in the assay protocol, i.e., lower substrate concentration. A comparison of 9 with the most closely related natural product C1 6 revealed a ca. 6000-fold loss in activity due to the absence of the aminoribose unit, thus further establishing that the aminoribose motif mediates a key interaction with MraY. This has also been derived from the X-ray co-crystal structure of MraY from Aquifex aeolicus in complex with muraymycin D2 8.^[11] Nonetheless, 9 was still a fairly strong MraY inhibitor, thus making it a suitable, structurally simplified model system for SAR studies, for instance, on variations in the peptide unit.

We then tested for inhibition of solubilized and purified MraY from Aquifex aeolicus (Table 1). Unexpectedly, inhibitory potencies were generally worse than with the S. aureus protein, with IC₅₀ values of muraymycins 3–6 being up to ca. 22-fold higher. For muraymycins D1 7 and D2 8, IC₅₀ values were in the nM range, demonstrating that inhibitory activities were ca. 200-fold and 120-fold, respectively, lower than for the protein preparation from S. aureus. Synthetic analogue 9 was only active at μM concentrations. Overall, the order of activities within the compound series 3–9 was similar irrespective of the source of MraY activity. However, structural variations which were disadvantageous for MraY inhibition had a much more pronounced effect with the solubilized Aquifex aeolicus protein. For instance, the inhibitory activity of muraymycin D1 7 was ca. 280-fold lower than the one of C1 6 (instead of 30-fold for the S. aureus protein, vide supra), and the formal removal of the aminoribose unit in 9 (relative to 6) even furnished more than five orders of magnitude loss in potency.

It was unclear if this pronounced decrease in inhibitory activities was a result from the solubilization of the Aquifex aeolicus protein, or if it was related to the intrinsic properties of this MraY homologue. As Aquifex aeolicus is an extremophile, the latter could not be ruled out, even if sequence alignment of the S. aureus and Aquifex aeolicus proteins suggested otherwise (Figure S1, Supporting Information). We therefore decided to overexpress a third MraY homologue, i.e., MraY from Clostridium bolteae. This MraY homologue was recently used to obtain a co-crystal structure with tunicamycin^[14] and has three relevant features: (i) it does not originate from an extremophile, but rather from a bacterial class with human pathogenicity; (ii) it can be solubilized and purified (like the Aquifex aeolicus protein); (iii) in contrast to the MraY homologue from Aquifex aeolicus, it is not modified with a maltose-binding protein (MBP) purification tag for overexpression. The MBP tag solubilizes MraY and therefore precludes its membrane insertion. Hence, the activity of the C. bolteae protein can alternatively also be applied in the form of non-purified crude membranes from overexpressing cells, as it is the case for the MraY homologue from S. aureus.

Inhibition data of 3–9 with solubilized and purified MraY from C. bolteae were very similar to activities observed with the solubilized Aquifex aeolicus protein (Table 1). In contrast, IC₅₀ values of muraymycins 3–5 with MraY from C. bolteae in crude membranes were nearly identical to values obtained with the S. aureus protein in crude membranes. The IC₅₀ value of muraymycin C1 6 with MraY from C. bolteae in crude membranes was at least closer to the S. aureus than to the Aquifex aeolicus inhibitory activity. For all other inhibitors 7–9, IC₅₀ values with MraY from C. bolteae in crude membranes was in between the S. aureus and the Aquifex aeolicus values. We therefore concluded that inhibitory potencies may differ for MraY homologues from different bacteria, in a way that structural variations of the inhibitors can have more or less pronounced effects. This can obviously even occur in the case of high sequence homology, as it is displayed by the MraY homologues selected for this study (Figure S1, Supporting Information). However, relative tendencies within a series of inhibitors are notably retained. On the other hand, the preparation of the protein for in vitro MraY assays appears to play a significant role. While solubilization with non-denaturing detergents in principle retains MraY activity, it leads to differences in MraY inhibition. The latter apparently depends on the native lipid environment provided by cellular membranes. The comparison of the X-ray co-crystal structure of MraY from Aquifex aeolicus in complex with muraymycin D2 8^[11] with the previously reported structure of the apo-enzyme^[6b] reveals that MraY undergoes a large conformational change upon inhibitor binding. Such conformational transitions may be different in a lipid membrane environment than in detergent micelles, which could potentially explain the observed differences in inhibition data.

In order to correlate in vitro activities for MraY inhibition with antibacterial properties, we then studied inhibition of bacterial growth by muraymycins 3–9 (Table 1). Antibacterial activities of muraymycins against a range of pathogens have been reported before,^[8c,9a,b] so we mainly aimed to confirm some of the previously reported data and to comparatively study activities against an efflux-deficient bacterial strain. In our hands, muraymycin A1 3 was the only congener with notable activity against S. aureus, which was in good agreement with previously published data.^[9a,b] This result was remarkable though as muraymycins 3–6 all showed very similar inhibitory potencies towards MraY, thus indicating that the unique activity of 3 is correlated to cellular access to the target. We have reported a model system to study the membrane-interacting properties of the ω-functionalized fatty acid moiety of A-series muraymycins such as 3, indicating both an efficient membrane accumulation and penetration mediated by the guanidinium-containing structural motif.^[13] This present study encourages further investigations on the membrane-penetrating properties of O-acylated muraymycins of the A- and B-groups.

Natural products 3–8 were not active against the efflux-competent E. coli strain DH5α, but showed strong activities against the efflux-deficient E. coli ΔtolC mutant (MIC 1–2 μg/mL, Table 1). This demonstrated that efflux also plays a significant role in the antibacterial activity of muraymycins. Remarkably, differences of 3–8 in MraY inhibition did not furnish differences in their inhibition of the growth of E. coli ΔtolC. For instance, muraymycins C1 6 and D1 7 showed a ca. 30- to 430-fold difference in MraY inhibition (crude membranes only), but identical MIC values (1 μg/mL). As both compounds are very similar in size and polarity, it is highly unlikely that they differ in cellular uptake, and efflux is hampered in the E. coli ΔtolC mutant. We therefore conclude that an increase in inhibitory activity towards MraY beyond a certain threshold does not necessarily lead to increased antibacterial activity. Contrastingly, synthetic analogue 9 differed from natural products 3–8 as it inhibited the growth of both the E. coli strain DH5α and the ΔtolC mutant, mostly likely because it was less prone to cellular efflux. Taken together, the obtained antibacterial data suggest that antibacterial activities of muraymycins are largely influenced by cellular uptake and efflux phenomena and are less dependent on pharmacodynamic properties.

The recently reported X-ray co-crystal structure of MraY from Aquifex aeolicus in complex with muraymycin D2 8^[11] provided the first structural insights into the target interactions of muraymycins. With exact inhibition data of compounds 3–9 in hand, we were interested if the relative inhibitory potencies of muraymycins could be correlated to structural aspects by in silico modelling, based on the co-crystal structure. Attempts to dock muraymycin ligands into this structure of MraY (after deletion of 8 from the complex) failed due to the pronounced conformational flexibility of muraymycins (which hampered their energy optimization in solution) and of the enzyme. We therefore proceeded by changing the structure of ligand 8 within the complex into the structure of other muraymycins, and the energy of the respective ligand-protein complex was then minimized in silico. This procedure furnished calculated structures of other muraymycin-MraY complexes (Figure S2, Supporting Information). The structure of muraymycin A1 3 in complex with MraY is a representative example (Figure 1). An overlay of the calculated structure of 3 and the reported structure of muraymycin D2 8 in complex with MraY revealed a very similar orientation of the ligands. The ω-functionalized fatty acid moiety of muraymycin A1 3, which is absent in 8, is oriented towards a hydrophobic cleft of the protein (Figure 1A). Within this hydrophobic cleft, there are no apparent key interactions of the fatty acyl unit with MraY though (Figure 1B).

Figure 1.

A: Overlay of muraymycins A1 3 (hydrocarbon scaffold in orange) and D2 8 (hydrocarbon scaffold in light blue) in complex with MraY from Aquifex aeolicus (grey), with the position of 8 derived from the previously reported X-ray co-crystal structure^[11] and the position of 3 derived from in silico modelling. The orange arrow indicates the position of the fatty acyl side chain of 3, which points into a hydrophobic cleft of the protein. B: Overlay of muraymycins A1 3 (orange) and D2 8 (light blue) from the aforementioned complexes with MraY, showing parts of MraY (thin stick representations) which mediate key interactions (light grey for 3, dark grey for 8). The orange arrow indicates the position of the fatty acyl side chain of 3, which does not show specific key interactions with the protein.

In conclusion, we report a detailed study on the inhibition of the bacterial membrane enzyme MraY by naturally occurring muraymycin nucleoside antibiotics. Several muraymycins proved to be extremely potent MraY inhibitors with IC₅₀ values in the low pM range. The tested compounds provided SAR data on the roles of the aminoribose unit (mediates a key interaction), its 2-O-methylation (not significant for MraY inhibition) and the fatty acid unit (probably mainly required for uptake-related effects). Different sources of MraY activity revealed (i) that inhibitory potencies may differ for MraY homologues from different bacteria, but that relative SAR trends seem to be retained, and (ii) that the preparation of the protein (crude membranes vs. solubilization with detergents) significantly influences inhibitory data from in vitro MraY assays. Antibacterial activities do not fully correlate to SAR data for MraY inhibition, indicating that cellular uptake as well as efflux play important roles. Overall, our results have strong implications for future work. First, the obtained SAR data will contribute to the design of novel muraymycin analogues. Second, MraY assays should be performed with protein preparations providing a lipid environment similar to the native one for MraY, i.e., crude membranes or lipid bilayer nanodiscs. Work along this line is ongoing in our laboratories.

Experimental Section

Full experimental details are disclosed in the Supporting Information.