The enzymatic activity of inosine 5’-monophosphate dehydrogenase may not be a vulnerable target for Staphylococcus aureus infections

Abstract

Many bacterial pathogens, including Staphylococcus aureus, require inosine 5’-monophosphate dehydrogenase (IMPDH) for infection, making this enzyme a promising new target for antibiotics. Although potent selective inhibitors of bacterial IMPDHs have been reported, relatively few have displayed antibacterial activity. Here we use structure-informed design to obtain inhibitors of S. aureus IMPDH (SaIMPDH) that have potent antibacterial activity (minimal inhibitory concentrations less than 2 μM) and low cytotoxicity in mammalian cells. The physicochemical properties of the most active compounds were within typical Lipinski/Veber space, suggesting that polarity is not a general requirement for achieving antibacterial activity. Five compounds failed to display activity in mouse models of septicemia and abscess infection. Inhibitor-resistant S. aureus strains readily emerged in vitro. Resistance resulted from substitutions in the cofactor/inhibitor binding site of SaIMPDH, confirming on-target antibacterial activity. These mutations decreased the binding of all inhibitors tested, but also decreased catalytic activity. Nonetheless, the resistant strains had comparable virulence to wild-type bacteria. Surprisingly, strains expressing catalytically inactive SaIMPDH displayed only a mild virulence defect. Collectively these observations question the vulnerability of the enzymatic activity of SaIMPDH as a target for the treatment of S. aureus infections, suggesting other functions of this protein may be responsible for its role in infection.

Graphical Abstract

The impending obsolescence of clinical antibiotics demands the development of new therapies¹. Treatment options are already severely limited for many pathogens, including methicillin resistant Staphylococcus aureus (MRSA)². The few new antibiotics developed recently are largely re-tooled β-lactams, tetracyclines, quinolones and aminoglycosides and there is legitimate concern that these too will quickly become compromised. New antimicrobial drugs are urgently needed, ideally with new targets.

Inosine monophosphate dehydrogenase (IMPDH) is a promising potential antibiotic target³. This enzyme catalyzes the rate limiting step in guanine nucleotide biosynthesis, the conversion of IMP to XMP with concomitant oxidation of NAD⁺. The gene encoding IMPDH (guaB) has been found to be essential for infection by many pathogens in genome-wide screens^3–13. IMPDH is a particularly alluring target for S. aureus infections. Both genome-wide transposon and single gene knockout experiments indicate that guaB is essential for growth in whole blood and ocular fluids and for infections in mice^{8, 14}. Nonetheless, these observations are somewhat surprising given that S. aureus possesses salvage pathways that should provide guanine nucleotides in the absence of IMPDH.

The cofactor binding site of S. aureus IMPDH (SaIMPDH) is very different from those found in the human enzymes^{15, 16}. Importantly, selective inhibitors, derived from several scaffolds, have already been reported for IMPDHs with similar cofactor binding sites, including IMPDHs from the intracellular protozoan pathogen Cryptosporidium parvum (Cp)^17–22, Bacillus anthracis (Ba)^{15, 23}, Clostridium perfringens¹⁵, Campylobacter jejuni¹⁵, Francisella tularensis²⁴ and Mycobacterium tuberculosis (Mtb)^25–31. However, relatively few compounds display antibacterial activity. Moreover, when antibacterial activity is observed, the spectrum is usually limited to a few pathogens, most likely reflecting differences in the structure activity relationships (SAR) for both enzyme inhibition and compound uptake.

The urea and triazole scaffolds, designated P and A in our previous reports, respectively, displayed the greatest antibacterial activity against B. anthracis, as exemplified by P150 and A98 (Figure 1A)²³. These compounds were originally designed for uptake into mammalian and protozoan cells in our CpIMPDH inhibitor program, and are consequently more lipophilic than typical for antibiotics^32–34. For example, P32 and P37 are moderately potent inhibitors of BaIMPDH with weak antibacterial activity (Figure 1A)²³, while the more polar P146 displays superior antibacterial activity despite being a considerably weaker enzyme inhibitor. These observations suggested that antibacterial activity might be improved if the compounds were redesigned to increase polarity while maintaining potency of enzyme inhibition.

Figure 1. Bacterial IMPDH inhibitors.

A. Inhibitors with antibacterial activity against B. anthracis²³. B. The residues within 5 Å of P32 in BaIMPDH are shown (PDB 4MYX, A-C’ active site). The subunit listed first contains the IMP site while the subunit denoted with a prime contains the adenosine subsite. Val25’ (light green) is Ile in SaIMPDH. The aromatic rings of P32 (pink) are denoted “R” and “L”. IMP is gray. Hydrogen and halogen bonds are shown in cyan. Water molecules were removed for clarity. C. Inhibitor SAR of BaIMPDH and SaIMPDH are similar. The black line denotes equal values of K_i,app for both enzymes. The dotted lines denote 2-fold differences. D. The P32 complex of BaIMPDH (PDB id 4MYX, chains A-C’) is overlayed with the cofactor complex of MtbIMPDH2 (PDB id 4ZQM, chains D-C’). Residues within 5 Å of P32 and NAD⁺ are shown. Water molecules removed for clarity. BaIMPDH, tan; MtbIMPDH2, sky blue, XMP light green (MtbIMPDH), IMP gray (BaIMPDH) and light green (MtbIMPDH); P32 (pink); NAD, dark cyan. BaIMPDH residues are labeled¹⁵.

Here we report second generation bacterial IMPDH inhibitors that display potent antibacterial activity against S. aureus in vitro and low cytotoxicity in mammalian cells, but do not display in vivo efficacy. Moreover, resistant strains of S. aureus readily emerged that have comparable virulence to wild-type strains. Lastly, bacteria expressing catalytically inactive IMPDH displayed only a mild virulence defect. Collectively these observations question the value of targeting the enzymatic activity of IMPDH for the treatment of S. aureus infections.

RESULTS

Characterization of SaIMPDH.

We previously characterized five bacteria IMPDHs predicted to be sensitive to CpIMPDH inhibitors^{15, 23, 25, 35}. The SAR for enzyme inhibition was surprisingly varied despite the close similarity of the inhibitor binding sites in these enzymes. Therefore, we characterized IMPDH from S. aureus (SaIMPDH) to better enable the development of inhibitors for the treatment of S. aureus infections. We expressed and purified N-terminal His-tagged SaIMPDH (sequence NC_007622.1) in E. coli BL21ΔguaB as described previously for other bacterial IMPDHs¹⁵. The Michaelis-Menten parameters were similar to those of BaIMPDH (Table 1). The value of K_m(NAD⁺) for SaIMPDH was smaller than that previously reported by another laboratory (literature value 2.3 mM; other parameters were not reported³⁶), but these assays were performed at different temperature and pH.

Table 1.

Steady state kinetic parameters for bacterial IMPDHs.

Parameter	BaIMPDHa	SaIMPDHb	Sa2IMPDHb,c	G445Sb,d	Y446Cb,e
K m (IMP), μM	18 ± 2	35 ± 2	25 ± 2	88 ± 3	96 ± 5f
K m (NAD + ), μM	550 ± 50	390 ± 20	228 ± 8	200 ± 10	5800 ± 400
K ii (NAD + ), mM	3.9 ± 0.5	8.0 ± 0.4	7.3 ± 0.6	12.0 ± 0.5	n.a.
k cat , s −1	1.4 ± 0.2	2.3 ± 0.0	2.0 ± 0.0	0.50 ± 0.01	0.90 ± 0.01d

Buffer 50 mM Tris, pH 8.0, 150 mM KCl, 3 mM EDTA and 1 mM DTT.

^a.BaIMPDH, recombinant IMPDH from B. anthracis, values from³⁵.

^b.SaIMPDH, recombinant IMPDH from S. aureus as reported in NC_007622.1. This enzyme was used routinely for inhibitor characterization. Sa2IMPDH, recombinant IMPDH from S. aureus Smith. Sa2IMPDH-G445S and Sa2IMPDH-Y446C, recombinant IMPDH from S. aureus Smith containing the eponymous mutations associated with resistance to 25.

^c.[E] = 20 nM;

^d.[E] = 20 nM;

^e.[E] = 200 nM;

^f.[NAD⁺] = 8 mM. n.a., not applicable.

Many x-ray crystal structures of IMPDH inhibitor complexes have been reported, including ten with BaIMPDH^{15, 35}. We compared the structure-activity relationship (SAR) of enzyme inhibition for SaIMPDH and BaIMPDH in order to determine the relevance of these structures for inhibitor design. SaIMPDH is 73% identical to BaIMPDH overall; only one residue within 5 Å of inhibitor P32 is different in the two enzymes (Val25 is Ile in SaIMPDH; Figure 1B). We surveyed a set of 34 compounds drawn from the A (15 compounds) and P (19 compounds) series, including A98, A110, P146 and P150 (Figure 1C, Table S1). The values of K_i,app for SaIMPDH and BaIMPDH were very similar for all of the A compounds. However, while the P compounds followed the same trend, the values of K_i,app were lower for SaIMPDH by an average factor of 2.5. The structural features that account for the different SAR of the A and P compounds are not readily apparent. Nonetheless, these observations suggested that the structures of BaIMPDH inhibitor complexes would be a useful guide for further medicinal chemistry optimization.

Redesign of the P series to increase t-PSA.

We initially focused on further optimization of the P series because these compounds include the most potent inhibitors of SaIMPDH and are more readily synthesized^{18, 23}. We retained the 4-Cl on the R-ring because this moiety forms a halogen bond with the main chain carbonyl C of Gly444’ (Figure 1B). We first increased the polarity at the 3-position on the R-ring with the addition of sulfonamides and sulfoxides (1, 2 and 3) (Table 2). We also substituted the 3-position with small heterocycles that improved antibacterial activity in topoisomerase inhibitors (4, 5, 6)^{37, 38}. All of these substitutions had a deleterious effect on enzyme inhibition relative to P37, ranging from a factor of ~2 (1 and 5) to a factor of ~20 (2). Nonetheless, several compounds displayed better antibacterial activity than P37, most notably the trifluoromethylpyrazole 4 and the trifluoromethylthiazole 5, where antibacterial activity improved by factors of 40–60.

Table 2.

Optimization of the P series.

ID	Cmpd	t-PSA (Å2)	cLogP	Kiapp (nM)	MIC (μM)
					−Guo	+Guo
P37 a		73.72	3.58	5.5 ± 0.4	60a	n.d.
P146 a		103.26	2.67	26 ± 9	2 a	>30a
P150 a		94.03	3.77	15 ± 2	4 a	>30a
1		128.27	2.55	10 ± 5	13 ± 5 b	>40c
2		139.79	2.10	110 ± 50	n.d.	n.d.
3		136.55	2.19	37 ± 3	13 ± 5	>64
4		91.54	4.47	29 ± 10	2.5 ± 1.5	32
5		97.38	4.47	13 ± 2	1 ± 0	>64
6		117.56	3.66	28 ± 7	8 ± 0	>64
7		112.75	2.29	51 ± 9	>30c	>30c
8		124.27	1.84	830 ± 50c	n.d.	n.d.
9		121.03	1.93	250 ± 90	> 30c	>30c
10		76.02	4.21	118 ± 2	>30c	>30c
11		99.33	4.94	25 ± 1	>30b	>30b
12		119.47	2.32	220 ± 20	n.d.	n.d.
13		108.32	1.97	2800 ± 300c	n.d.	n.d.
14 a		41.13	5.04	9.5 ± 1.0	60	n.d.
15		64.92	3.36	560 ± 150	n.d.	n.d.
16		81.26	3.17	1600 ± 200c	n.d.	n.d.
17		70.23	3.20	500 ± 130c	n.d.	n.d.
18		61.44	3.10	960 ± 160b	n.d.	n.d.
19		99.74	2.77	21.5 ± 0.2	4 ± 2	28 ± 4
20		99.74	3.40	12.9 ± 0.1	1.2 ± 0.1	>32
21		67.15	4.36	12.8 ± 0.7	1.7 ± 0.3	>32
22		61.36	4.68	9 ± 4	3 ± 1	>50
23		90.49	4.60	4 ± 1	>40	>40
24		76.66	5.30	19 ± 1	>64	>64 c
25		93.45	3.95	6 ± 1	1.1 ± 0.1	>32
26		99.69	3.87	6 ± 2	4 ± 0 b	≥64
27		109.25	3.84	16 ± 1	1.7 ± 0.5	>64
28		126.04	2.50	2.4 ± 0.3	1 ± 0	>50
29 d		123.08	3.51	44 ± 11	11 ± 5	>64
30		120.28	3.50	20 ± 7	6 ± 2	>30
31 d		123.08	3.51	20 ± 5	2 ± 2	>64
32		154.29	1.74	13 ± 2	>64	>64 c
33		162.57	1.38	52 ± 19	>64	n.d.
34		140.87	4.39	10 ± 2	0.8 ± 0.3	>40
35		117.56	3.66	26 ± 5	1.5 ± 0.5	>40

The values of K_i,app for SaIMPDH are averages and standard deviations of at least 3 independent determinations unless otherwise noted. Note that a single determination is derived from at least 30 rate measurements, triplicates of at least 10 different inhibitor concentrations. Conditions: 20 nM enzyme, 1000 μM IMP and 1000 μM NAD⁺. Minimum inhibitory concentration (MIC) for S. aureus Smith in RPMI media are the average and range of at least two independent determinations.

^a.Synthesis/data from²³.

^b.Average and range of 2 independent determinations.

^c.Value and error of the fit for 1 determination. n.d., not determined. cLogP and t-PSA calculated on the FAFDrugs4 website³⁹.

^d.Compound is likely a mixture of both pyranoses (major) and both furanoses (minor)^40–42.

In general, replacement of the oxime on the L-ring with a ketone was deleterious for enzyme inhibition, increasing the values of K_iapp by factors of 4–7 (compare 1 vs. 7, 2 vs. 8, 3 vs. 9 and 4 vs. 10). The ketone analog of 5 (11) was also a good inhibitor of SaIMPDH, but displayed little antibacterial activity. Replacement of the oxime with cyano was more deleterious, increasing the value of K_iapp by factors of 22–75 (1 vs. 12 and 3 vs. 13). To further explore the requirements of the L-ring oxime, we compared analogs of P37 where the oxime was replaced with alkene (14), cyano (15), carboxylate (16), methylamide (17), dimethylamide (18) and hydroxyamidine (19) (Table 2). All of these substitutions were deleterious with respect to enzyme inhibition. Most of these substitutions increased K_i,app by factors of 100 or more. However, the alkene and hydroxyamidine substitutions had increased K_i,app by only factors of 2 and 4, respectively. Remarkably, the hydroxyamidine 19 displayed a 10-fold improvement in antibacterial activity despite the loss of potency in enzyme inhibition. We also evaluated the di-Cl derivative 20 since prior work indicated that this substitution improved antibacterial activity²³. This substitution improved both enzyme inhibition relative to 19 by a factor of 18 and antibacterial activity by a factor of 3.3 (20 MIC = 1.2 μM).

Redesign to interact with the adenosine ribose binding site.

Recent x-ray crystal structures of bacterial IMPDHs revealed that the A and P compounds bind in the cofactor binding site^{15, 25}. Overlay of the structures of the BaIMPDH•P32 complex and MtbIMPDH2•cofactor complex indicates that the R-ring of the P compounds binds in a subsite occupied by the adenine base, but does not extend into the ribose binding subsite (Figure 1D). Further inspection suggested that modifications of the 3-position on the R-ring would extend into the ribose binding site, providing another attractive strategy for increasing potency and polarity.

We turned to a starting scaffold that contains an alkene group on the L-ring because it is more easily synthesized than the oximes and hydroxyamidines (e.g. 14, Table 2). Although 14 was a potent inhibitor of SaIMPDH, it displayed little antibacterial activity. We first added 3-NH₂ (21) and 3-OH (22) substituents to enable further elaboration of the 3-position. These compounds were potent inhibitors of SaIMPDH, equivalent to 14 (Table 2). Both 21 and 22 displayed improved antibacterial activity relative to 14 (MIC = 1.7 and 3 μM, respectively).

We next added hydrogen bonding moieties to the 3-position of the R-ring with the intent of introducing an interaction with Ser257 (23, 24, 25 and 26). Addition of an acetate group (23) increased enzyme affinity, but eliminated antibacterial activity. Such substitutions have been similarly debilitating in the antibacterial activity of topoisomerase inhibitors³⁷. The corresponding ethyl ester derivative (24) had a deleterious effect on enzyme affinity, and also failed to display antibacterial activity. In contrast, the amide (25) and hydroxamic acid (26) derivatives remained good enzyme inhibitors and displayed superior antibacterial activity (MIC = 1.1 and 4 μM, respectively). Since the oxime substituted L-ring was more polar and soluble, we synthesized the oxime versions of 24 (e.g. 27) and 25 (e.g. 28). Compound 27 displayed similar enzyme inhibition as 24. Nonetheless, antibacterial activity improved by more than a factor of 30. In contrast, no improvement in antibacterial activity was observed for 28 relative to 25, although enzyme inhibition improved by a factor of approximately 2.5. These observations reveal that the structure-activity relationship (SAR) for antibacterial activity is very context dependent.

Further inspection of the crystal structure of BaIMPDH•IMP•P32 suggested that a sugar attached to the 3-position of the R-ring would bind in the site occupied by the nicotinamide ribose of the cofactor (Figure 1D). Therefore, ribose (29 and 30) and arabinose (31) were appended to this position. These compounds exist as a mixture of anomers. In addition, the reaction of aniline with ribose gives a mixture containing 50% of the β-pyranose, 35% of the α-pyranose, 10% of the β-furanose and 5% of the α-furanose, that can equilibrate readily under neutral conditions by ring opening to the hydroxy imine intermediate. Thus, 29 and 31 are likely mixtures of both pyranoses (major) and both furanoses (minor)^40–42. These substitutions decreased the potency of enzyme inhibition by 2 to 4-fold relative to 21 and 22. The antibacterial activity of 29 was similarly decreased, while 30 and 31 maintained antibacterial activity.

Combination of L- and R-ring substitutions.

We combined the L-ring hydroxyamidine with the successful R-ring substituents to produce 32, 33, 34 and 35, with the goal of further improving antibacterial activity. Hydroxyamidine substitution had little effect on enzyme inhibition relative to the oxime and alkene analogs (e.g., 1 vs 32, 3 vs. 33, 4 vs. 35 and 5 vs. 34) (Table 2). Disappointingly, the hydroxamidine substitution did not improve antibacterial activity relative to the oxime analogs. No antibacterial activity was observed for 32 and 33. The antibacterial activity of 34 and 35 was similar to their oxime analogs with MIC values of ~1 μM. These observations further underscore the seemingly capricious nature of antibacterial SAR.

A-series SaIMPDH inhibition and antibacterial activity.

Two compounds in the A series, A98 and A110, previously demonstrated potent antibacterial activity against S. aureus²³. Given the improved antibacterial activity imparted by the thiazole in the P series, we decided to introduce this heterocycle into the 3-position of the phenyl of the A series (Table 3). In the four compounds evaluated (36-39), which contained various quinoline and quinoline N-oxide ethers, SaIMPDH inhibition was maintained with K_i,app values ≤ 30 nM. Antibacterial activity was also retained with MIC values of 3 – 4 μM, although this was not an improvement relative to A98.

Table 3.

A series compounds.

ID	Cmpd	t-PSAa (Å2)	cLogPa	Ki,appb (nM)	MIC (μM) c
					−Guo	+Guo
A98 d		52.83	4.15	20 ± 4	2a	16a
A110 d		65.4	3.15	50 ± 20	8a	>30a
36		93.96	5.76	7 ± 2	3 ± 1	> 64
37		93.96	5.76	8 ± 2	3 ± 1	>64b
38		106.53	4.76	30 ± 10	4 ± 0	> 64
39		106.53	4.76	16 ± 2	3 ± 1	>64

See legend Table 2 for conditions.

^a.cLogP and t-PSA as calculated by FAFDrugs4 website³⁹.

^b.Average and standard deviation from at least 3 independent determinations.

^d.Average and range from at least two independent determinations.

^d.Synthesis described in²³.

Antibacterial activity is due to inhibition of SaIMPDH.

Figure 2 shows the correlation of antibacterial activity and inhibition of SaIMPDH for the P and A compounds described above. Greater antibacterial activity was associated with more potent enzyme inhibition, as expected if antibacterial activity derived from inhibition of SaIMPDH. Poor enzyme inhibitors displayed at best weak antibacterial activity. Some potent enzyme inhibitors did not display antibacterial activity, perhaps due to poor uptake or transformation into an inactive metabolite by the bacteria. To further confirm on-target antibacterial activity, we determined the values of MIC in the presence of guanosine, which allows cells to grow in the absence of IMPDH activity. None of the inhibitors displayed antibacterial activity in the presence of guanosine (MIC >30 μM), with the exception of 19, where the value of MIC increased by a factor of 7 (Tables 2 and 3). These observations indicate that antibacterial activity derives from inhibition of SaIMPDH as designed.

Figure 2. Physicochemical properties of enzyme inhibition and antibacterial activity.

P compounds are denoted with circles, A compounds with squares. The dotted line denotes MIC = 2 μM. Compounds with no antibacterial activity were arbitrarily assigned MIC = 60 μM. Symbols are colored according to cLogP and t-PSA as calculated by FAFDrugs4 website³⁹.

The physicochemical properties of antibacterial activity.

Although some studies have suggested that antibiotics are smaller and more polar than Lipinski/Veber guidelines^{32–34, 43}, these trends were not observed for the SaIMPDH inhibitors (Figure 2 and Figure S1). The average values of cLogP and t-PSA are 3.7 and 107 Ų, respectively, for the 11 P compounds with values of MIC ≤ 2 μM. The best antibacterial compounds, 34, 20 and 28, vary in cLogP from 2.50 to 4.39, while 32 does not display antibacterial activity despite being a potent inhibitor (K_i,app = 13 nM) with low lipophillicity (cLogP = 1.74). All of the A compounds have comparable antibacterial activity despite large differences in lipophilicity (cLogP varies from 3.15 to 5.76) and polar surface area (t-PSA varies from 52.83 to 106.53 Ų). These observations suggest that increasing polarity may not be a general strategy for antibiotic design.

Structures of enzyme-inhibitor complexes.

To determine how the thiazole, ribose, arabinose and hydroxyamidine groups interacted with the enzyme, we solved the x-ray crystal structures of five inhibitors (5, 19, 29, 30 and 31) with the catalytic domain of BaIMPDH (BaIMPDHΔL). This protein is missing the regulatory CBS subdomain, yet retains catalytic activity equivalent to the wild-type enzyme¹⁵. All complexes included IMP in addition to inhibitor. The crystal, data collection and refinement statistics can be found in Table S2.

The five structures were very similar overall (Figure 3). The unit cell of the complex with inhibitor 5 contained one tetramer, while two tetramers were found in the unit cells of the other complexes. All five inhibitors bound in the same site as P32. The L and R rings occupied the same position in all active sites. As expected, the L-ring interacted with IMP and the R-ring interacted with Tyr445’ in the adjacent subunit while the urea nitrogens formed hydrogen bonds with Glu416¹⁵. All of the inhibitors also interacted with residues Ala253, His254 and Pro27’, and all formed a halogen bond between the 4-Cl on the R-ring and the main chain carbonyl carbon of Gly444’. The 3-substituents of the R ring extended into the adenosine ribose subsite in all cases. In contrast, the structure of the “flap” (residues 390–418) displayed varying amounts of disorder in different complexes as well as in different active sites of the same complex (Figure 3). At least 4 different flap conformations were observed among the five structures, suggesting that the flap was dynamic.

Figure 3. The flap has multiple conformations in inhibitor complexes of BaIMPDHΔL.

Representative active sites from E•IMP•inhibitor complexes are shown. Hydrogen bonds are indicated by solid cyan lines. The subunit containing the IMP binding site is listed first and the subunit containing the adenosine subsite of the cofactor binding site is denoted with a prime. The disordered regions are denoted by dashed lines. The structure of BaIMPDH•IMP•5 contains a single tetramer (PDB id 5UUV). Two flap conformations are exemplified in active sites A-D’ (light green) and D-B’ (lime green). The structure of BaIMPDH•IMP•19 contains two tetramers (PDB id 5UUZ). Two flap conformations were observed in the 8 active sites, exemplified by C-D’ (coral) and E-G’ (orange). The structure of BaIMPDH•IMP•29 complex contains two tetramers (PDB id 7MTU). Three flap conformations were observed in the 8 active sites, exemplified by A-C’ (tan), G-H’ (sienna), and H-F’ (brown). The structure of BaIMPDH•IMP•30 contains two tetramers of BaIMPDH (PDB id 5URS). Two flap conformations were observed in the 8 active sites, exemplified by D-B’ (plum) and G-H’ (medium purple). The structure of BaIMPDH•IMP•31 contains two tetramers (PDB id 7MTX). The flap conformations varied among the active sites, with B-A’ (light blue) and E-G’ (dark sky blue) representing the most divergent positions.

Distinct interactions were observed with the other substituents of the L and R rings in each inhibitor complex (Figures 4 and 5). For 5, L-ring thiazole ring formed a hydrogen bond with the side chain of Ser257 and the R-ring oxime OH formed a hydrogen bonds with the side chains of Thr310, Glu416, and Tyr445’ (Figure 4A). These hydrogen bonds were not observed with the oxime of P32 (Figure 1B). The L-ring hydroxyamidine N2 of 19 also formed a hydrogen bond with the carboxylate of Glu416 while the OH made a water mediated hydrogen bond with the carboxylate of Asp27 (Figure 4B). Thus, the L-ring substituents interact with BaIMPDH in a context-dependent manner that may explain the unpredictable SAR.

Figure 4. Interactions of the oxime and hydroxyamidine substitutions on the A ring.

Hydrogen bonds are shown in cyan, Halogen bond in gold. A. Structure BaIMPDH•IMP•5 (PDB id 5UUV). 5 is magenta. Contact residues are shown for the D-B’ active site. B. Structure of BaIMPDH•IMP•19 (PDB id 5UUZ). 19 is green. Contact residues are shown for the active site formed by chains A-C’).

Figure 5. Interactions of sugar containing inhibitors with BaIMPDH.

Contact residues in representative active sites are shown. Hydrogen bonds are shown in cyan and halogen bonds are gold. A. E•IMP•29 (PDB id 7mtx). D-B’ protein is plum, 29 is purple. B-A’ protein is light green, 29 is dark green. E-G’ protein is sky blue, 29 is dark sky blue. B. E•IMP•30 (PDB id 5urs). G-H’ protein is plum, 30 is purple; H-F’ protein is light green, 30 is dark green. C. E•IMP•31 (PDB id 7mtu). A-C’ protein is plum, 31 is purple. H-F’ protein is light green, 31 is dark green.

The sugar moieties of 29, 30 and 31 bound in the adenosine ribose site as designed (Figure 5A–C). The structures were refined with the sugar conformer that best fit the electron density in each active site. However, the resolution was not sufficient to unambiguously distinguish tautomeric and anomeric configurations and it is possible that each active site contained a dynamic mixture of species. The ribose of 29 was modeled as pyranose, the favored tautomer of aniline-linked pentoses^40–42. The sugar moiety appeared to form hydrogen bonds to the side chain OHs of Thr252 and Ser257 and the main chain carbonyl Os of Val229 and His254 (Figure 5A). These interactions variously involved the ring O, 2’-OH, 3’-OH and/or 4’-OH depending on the active site. The sugar moiety of 30 was modeled as a furanose (Figure 5B). Depending on the active site, the 2’-OH variously formed hydrogen bonds with Thr252 side chain OH and the main chain NH and/or carbonyl O of His254. The 3’-OH also formed hydrogen bonds with the OH of Thr252 while the 5’-OH formed hydrogen bonds with Ser257. The arabinose moiety of 31 was modeled as a pyranose. The sugar appeared to make hydrogen bonds with the side chain hydroxyls of Thr252 and Ser257 and main chain carbonyl of His254. These interactions involved the 2’-OH, 3’-OH and 4’-OH depending on the sugar conformation modeled into a given active site (Figure 5C).

Inhibitor resistant S. aureus strains have mutations in SaIMPDH.

We selected strains of S. aureus Smith resistant to inhibitor 25 to gain further insight into the antibacterial activity of the SaIMPDH inhibitors. We chose this compound because it displayed potent antibacterial activity (MIC = 1 μM) and was easily synthesized. Twenty-six independent resistant clones were isolated that could grow in the presence of 25 (62 μM, 25 μg/mL) after multiple passages on agar plates. All 26 mutants displayed MICs ≥ 320 μM (128 μg/mL). The gene encoding IMPDH (guaB) gene was cloned from each strain and sequenced (note that the promoter was not included in sequencing because it is approximately 2.4 kb upstream from guaB, with two intervening genes). The wild-type guaB gene contained 21 substitutions relative to the sequence of SaIMPDH (NC_007622.1). We will denote the S. aureus Smith enzyme as Sa2IMPDH. Twenty-one resistant strains, exemplified by strain Sa26, contained a G->A mutation at nt1333 of guaB, causing the substitution of Ser for Gly445’. This residue is analogous to Gly444 in BaIMPDH, which interacts with the R-ring of the inhibitors (Figure 1D). Four resistant clones, exemplified by strain Sa12, contained a single A->G mutation at nt1337, causing the substitution of Cys for Tyr446’. The corresponding residue in BaIMPDH, Tyr445, also interacts with the R-ring. Intriguingly, the analogous Cys for Tyr substitution was also observed in an Mtb strain selected for resistance to another bacterial IMPDH inhibitor²⁶. No mutations were discovered in the guaB coding sequence of the remaining resistant clone. The observation that 25/26 resistant clones contain mutations in guaB further confirms that antibacterial activity derives from inhibition of SaIMPDH as designed.

Resistance mutations impair enzyme activity.

Both Gly445’ and Tyr446’ interact with the R-ring of the P inhibitors, so substitutions at these residues can explain resistance to 25. However, these residues are also part of the cofactor binding site, which suggested that the substitutions might also impair cofactor binding and enzyme activity. Sa12 (expressing Y446C) displayed a slow growth phenotype (Figure S1), further suggesting that the enzymatic activity might be compromised in the resistant variants.

We characterized purified recombinant wild-type Sa2IMPDH and mutant enzymes Sa2IMPDH-G445S and Sa2IMPDH-Y446C to assess the functional consequences of the substitutions (Table 1). The kinetic parameters of Sa2IMPDH are very similar to SaIMPDH used in the SAR experiments described above. The catalytic activity of Sa2IMPDH-G445S was slightly compromised relative to the wild-type enzyme, with a 3.5-fold increase in the value of K_m for IMP and a decrease in the value of k_cat by a factor of 4. The catalytic activity of Sa2IMPDH-Y446C was also impaired. The values of K_m for both IMP and NAD⁺ were increased, by 4-fold and 15-fold, respectively, and the value of k_cat was reduced by a factor of 2-fold. Assuming cellular concentrations of 270 μM IMP and 2.5 mM NAD^{+ 44}, the relative activities of Sa2IMPDH, Sa2IMPDH-G445S and Sa2IMPDH-Y446C would be approximately 1, 0.6 and 0.2, respectively.

Resistant Sa2IMPDH mutants are likely resistant to all inhibitors.

As expected, G445S and Y446C displayed lower affinity for 25, with values of K_i,app approximately 200-fold greater than the wild-type enzyme (Table 4). All bacterial IMPDH-specific inhibitors bind in the same site and interact with Gly445’ and Tyr446’^{15, 17, 22, 25–27, 35, 45, 46}, which suggested that Sa2IMPDH-G445S and Sa2IMPDH-Y446C would display decreased affinity to all bacterial IMPDH inhibitors. Indeed, Sa2IMPDH-G445S and Sa2IMPDH-Y446C were resistant to all twelve inhibitors tested, which included examples of 5 different scaffolds (Table 4). The two mutations had similar effects on the binding of most inhibitors, with K_i values typically 100–300-fold higher than those of the wild-type enzyme. Therefore, these mutations will likely provide resistance against any inhibitor targeting the cofactor site.

Table 4.

Inhibition of wild-type and mutant S. aureus IMPDHs.

Compound	Sa2IMPDHa	G445Sb		Y446Cc
	Kiapp (nM)	Kiapp (nM)	Ratio	Kiapp (nM)	Ratio
25	2.9 ± 0.5	570 ± 70	190	680 ± 40	230
29	39 ± 7	6200 ± 700	160	12000 ± 1500	310
30	15 ± 2	3800 ± 100	250	6200 ± 40	410
5	4.2 ± 1.7	1330 ± 30	320	920 ± 30	220
19	12.7 ± 0.4	4200 ± 100	330	3800 ± 400	300
34	4.8 ± 1.5	930 ± 30	190	1170 ± 5	240
35	20.5 ± 0.1	2500 ± 100	125	2800 ± 20	140
31	15 ± 5	3350 ± 150	220	5270 ± 90	350
A110 d	52 ± 5	8600 ± 100	160	>5000	94
36	2.7 ± 0.7	460 ± 10	170	10000 ± 1000	3700
Q21 e	8 ± 1	113 ± 5	14	3700 ± 2000	450
C91 f	77 ± 4	>5000	>65	>5000	>65
D67 g	320 ± 50	>5000	>16	>5000	>16
P32 h	29 ± 4	7100 ± 500	240	5600 ± 200	190

Sa2IMPDH, recombinant IMPDH from S. aureus Smith. G445S and Y446C, recombinant IMPDH from S. aureus Smith containing the eponymous mutations associated with resistance to 25. Buffer 50 mM Tris, pH 8.0, 150 mM KCl, 3 mM EDTA and 1 mM DTT. [IMP] = 1.0 mM.

^a.[E] = 20 nM, [NAD⁺] = 500 μM;

^b.[E] = 20 nM, [NAD⁺] = 500 μM.

^c.[E] = 200 nM, [NAD⁺] = 8 mM. n.a., not applicable.

^d.21;

^e.17;

^f.20;

^g.19;

^h.18.

Resistant mutants are virulent.

The above observations demonstrated that the mutations impaired the activity of Sa2IMPDH, suggesting that the resistant strains might have a reduced virulence. Initial testing of the wild-type strain established that infecting mice with an inoculum of 5×10⁶ CFUs caused mortality reproducibly within 24 to 48 h (Table S3). Survival stabilized after 48 h and no additional mortality was observed over a total of 7 days. Survival increased as the inoculum was decreased by factors of 10. Two wild-type samples were used, one freshly revived from storage (designated “A”) and the second passaged to approximate the isolation of the resistant strains (designated “B”). Comparable survival was observed for mice infected with the two wild type samples, indicating that culture history had no effect on virulence. The mutant strains Sa12 and Sa26 were infected at higher numbers than the wild-type because reduced virulence was anticipated (Figure 6A and Table S3). However, survival was similar for mice infected with wild type and resistant strains, suggesting that there was little fitness cost associated with resistant mutations. These observations suggest that resistance will develop readily if IMPDH inhibitors are used as monotherapy. Moreover, the uncompromised virulence of Sa12, despite expressing an enzyme with only 20% of wild-type activity, suggests that in vivo efficacy will require inhibition exceeding 80%.

Figure 6. Virulence of S. aureus strains.

A. Virulence of 25-resistant strains. WT-A, S. aureus Smith wild type strain freshly revived from storage; WT-B, denotes the wild-type strain passaged to approximate the isolation of the resistant strains. Sa12 carries Y446C substitution in IMPDH; Sa26 carries the G445S mutation. Survival was monitored over 7 days. B. Virulence of S. aureus JE2 expressing catalytically inactive IMPDH containing Ala for the essential catalytic residue Cys307.

SaIMPDH inhibitors are not cytotoxic.

Ten compounds with potent antibacterial activity, 8 from the P series and 2 from the A series, were tested for cytotoxicity in HepG2 cells by monitoring release of lactate dehydrogenase (Figure S3). Significant cytotoxicity (>10% release) was only observed in cells treated with 25 μM 34 and 35. Approximately 6% LDH release was observed when cells were treated with 25 μM 20 and 27. No cytotoxicity (<2% release) was observed in HepG2 cells treated with P146, A98, 5, 21, 31 and 39.

No in vivo efficacy is observed in mouse septicemia.

We tested the efficacy of five compounds, 20, 27, 34, 35 and 39 in a mouse septicemia model to further assess the antibiotic potential of SaIMPDH inhibitors. In the infection control group, only one mouse survived of the ten inoculated with 1.6 × 10⁶ colony forming units (CFUs) of S. aureus Smith (Table S4). Survival increased markedly when mice were infected with fewer bacteria: 40% survival was observed when mice were inoculated with 1.6 × 10⁵ CFUs while 90% survival was observed with 1.6 × 10⁴ CFUs. As expected, all five mice treated with vancomycin (10 mg/kg i.v.) survived, further validating the model. These observations demonstrate that the experiment would be able to reveal in vivo efficacy. Disappointingly, none of the compounds displayed significant in vivo activity (Table S4). Only one mouse survived of the five treated with 20 (75 mg/kg i.v.) or 27 (25 mg/kg i.v.). These results are statistically indistinguishable from the untreated controls. No mice survived when treated with 34 (50 mg/kg i.v.), 35 (50 mg/kg i.v.) or 39 (30 mg/kg i.v.).

Preliminary pharmacokinetic evaluation.

We evaluated the pharmacokinetics of 20 and 39 to determine if sufficient exposure was achieved in the septicemia model. For 20 (75 mg/kg i.v.), C₀ = 195 μM and τ_1/2 = 1.1 h. For 39 (30 mg/kg i.v.), C₀ = 150 μM and τ_1/2 = 6.4 h. In both cases, the values of C₀ greatly exceed MIC (1.2 μM and 2.1 μM for 20 and 39, respectively). However, both compounds display high plasma protein binding, with free fractions of 0.0043 and 0.014 for 20 and 39, respectively. These observations suggest that the concentration of free drug was at most 0.8 μM and 2.1 μM for 20 and 39, respectively, which would be approximately the value of MIC for each compound. Since survival improved when the inoculum decreased by a factor of 10, MIC concentrations of free 20 or 39 for even a few hours would be expected to display in vivo efficacy.

No in vivo efficacy is observed in mouse abscess infection model.

We also tested the efficacy of four compounds, 20, 34, 35 and 39, in S. aureus thigh abscess infection of mice, where SaIMPDH has been shown to be essential^{8, 14}. To avoid potential issues with metabolic stability and plasma protein binding, compounds were injected directly into the abscess, once 2 h after inoculation and again after 12 h. A 2.5 log₁₀ increase in colony forming units (CFUs) was observed after 26 h in the control abscesses treated with vehicle, while only a 1 log₁₀ increase in CFUs was observed in mice treated with the positive control linezolid (Table S3). Despite the high local concentrations of compound, no statistically significant decreases in CFUs were observed relative to the vehicle control (Table S3). These results cast further doubt on the in vivo vulnerability of SaIMPDH.

Bacteria expressing catalytically inactive SaIMPDH have modestly reduced virulence.

Both genome-wide transposon analysis and single gene knockout experiments indicate that guaB, the gene encoding SaIMPDH, is essential for S. aureus infection in murine models of infection^{8, 14}. These experiments suggest that loss of SaIMPDH reduces virulence by a factor of 10⁵-10⁶. However, these experiments did not address whether virulence requires the catalytic activity of IMPDH or some other function. Therefore, we constructed S. aureus JE2 strains expressing catalytically inactive IMPDH containing an Ala substitution at the catalytic Cys307 (guaB::C307A). This strain is a MRSA USA300 isolate comparable to NRS384, the strain used in previous work¹⁴. As expected, guaB::C307A cannot grow in absence of guanine (Figure S4). Growth is restored by plasmids expressing wild-type SaIMPDH.

We tested the virulence of wild-type and two guaB::C307A clones in the mouse septicemia model (Figure 6B and Tables S5 and S6). No survival was observed when mice were inoculated with 2.3×10⁸ CFUs of wild-type bacteria, while 60% survival was observed with a 2.3×10⁷ CFU inoculum. Approximately 20-fold more guaB::C307A bacteria were required to observe comparable mortality. Therefore, virulence appears to be only modestly dependent on the catalytic activity of IMPDH, suggesting that another function of this protein may be responsible for the essentiality associated with ΔguaB strains. Collectively these observations suggest that inhibitors of SaIMPDH may not be effective in treating S. aureus infections.

DISCUSSION

Gene knockout experiments have suggested that SaIMPDH is a promising target for the development of antibiotics to treat S. aureus infections^{8, 14} and structure-guided optimization of SaIMPDH inhibitors reported above yielded potent inhibitors with excellent antibacterial activity in vitro (MIC ≤ 2 μM). Some themes emerge from the SAR for enzyme inhibition that can be rationalized from the crystal structure. For example, the L-ring oxime and hydroxyamidine moieties form hydrogen bonds with the enzyme that may account for their increased affinity relative to the alkene and ketone groups. Nonetheless, different constellations of hydrogen bonds are observed depending on the R-ring, accounting for the varying effect of the oxime/hydroxyamidine substitutions. In contrast, the SAR for antibacterial activity appears almost capricious in comparison. For example, the addition of small heterocycles to the R-ring improves antibacterial activity by 60-fold in the context of the L-ring oxime, but has no effect in an L-ring hydroxyamidine (5 versus P37 and 34). The processes responsible for the unpredictable SAR of antibacterial activity are unclear, but could include inhibitor uptake and metabolism as well as regulation of enzyme function by post translational modification or other interactions. Moreover, it appears that these inhibitors bind to two different enzyme complexes, E•IMP and the covalent intermediate E-XMP*, and the preference for each complex varies between compounds. Therefore, potency can vary depending upon whether substrates are saturating (causing E-XMP* to accumulate) or sub-saturating (causing E•IMP to accumulate). We do not know which complex predominates in cells.

Since antibiotics appear to be characterized by low lipophilicity and high t-PSA^{32, 34, 43}, the optimization strategy involved decreasing lipophilicity and increasing polarity. Nonetheless, the physicochemical properties of the 11 most active compounds were indistinguishable from inactive compounds with comparable potency in enzyme inhibition (MIC ≥ 30 μM): average cLogP = 3.7 and 3.5 while average t-PSA = 103 and 101 Ų for active and inactive compounds, respectively. Thus decreasing lipophilicity/increasing t-PSA may not be a general strategy for antibiotic design. A survey of antibiotic discovery programs at AstraZeneca reached a similar conclusion³³.

While we succeeded in identifying SaIMPDH inhibitors with potent in vitro antibacterial activity and low cytotoxicity, we were unable to demonstrate in vivo antibacterial activity. The pharmacokinetic characteristics of the tested compounds were not ideal, nonetheless our experiments suggest that inhibiting the enzymatic activity of SaIMPDH may not be an effective strategy to treat S. aureus infection in vivo. Moreover, resistance develops easily, and the resulting strains have similar virulence as the wild-type. While the emergence of resistance can be mitigated with combination therapy, this observation is particularly troubling because one strain expressed an enzyme with approximately 20% of wild-type activity. This observation suggests that inhibition must exceed 80% to block infection. Achieving and sustaining such high levels of inhibition is likely to be challenging in vivo. Worryingly, incomplete inhibition could conceivably promote infection: CodY, the master regulator of metabolism and repressor of virulence in S. aureus, is activated by GTP⁴⁷. PurR, the master regulator of purine biosynthesis, also represses toxin expression⁴⁸. Therefore reduction of the guanine nucleotide pools could conceivably increase virulence by relieving repression of toxin production.

The essentiality of SaIMPDH was surprising given that S. aureus also has the pathways required to salvage xanthine, xanthosine and guanosine, which would enable the production of guanine nucleotides in the absence of SaIMPDH. Xanthine is present at ~0.8 μM in mouse plasma⁴⁹, and concentrations as high as 4.9 μM have been reported in human plasma⁵⁰. Xanthosine (5 μM) and guanosine (0.8 μM) are also present in plasma, while guanine is undetectable. Therefore, we became concerned that the essentiality of SaIMPDH might not derive from enzymatic activity and examined the virulence of strains expressing catalytically inactive enzyme. These strains displayed a modest virulence defect, casting further doubt on the in vivo vulnerability of S. aureus to SaIMPDH inhibitors. Salvage pathways vary widely among bacterial pathogens, so IMPDH inhibitors may be effective treatment for other infections. For example, Mtb does not salvage xanthine, xanthosine or guanosine. Mtb does salvage guanine, but requires concentrations at the limit of solubility⁵¹.

Our experiments suggest that the essentiality of SaIMPDH may derive from extra-catalytic functions of IMPDH. SaIMPDH, like most IMPDHs, contains a regulatory domain of uncertain function in addition to the catalytic domain⁵². Deletion of CBS subdomain dysregulates the purine nucleotide pools in E. coli, leading to the toxic accumulation of ATP and other adenine nucleotides^{53, 54}. Intriguingly, the antiproliferative effects of sanglifehrin A in mammalian cells derives from interactions with the CBS subdomain that do not affect catalytic activity⁵⁵. Similar moonlighting functions of the CBS domain may be responsible for the essentiality of SaIMPDH. In addition, SaIMPDH has been reported to interact with 62 proteins, including the ribosome, tRNA synthetases, penicillin binding proteins, uridylate kinase, PRPP synthetase, adenylosuccinate synthetase and DNA helicase⁵⁶. Perhaps the most intriguing SaIMPDH interactor is the transcription repressor PurR. The physiological consequences of these interactions are unknown. Given these myriad interactions, the failure of SaIMPDH inhibitors to phenocopy knockouts should perhaps not be surprising.

METHODS

Materials.

IMP, Tris, and common chemicals were purchased from Sigma-Aldrich (St. Louis, MO). NAD⁺ was purchased from Roche. The synthesis of the inhibitors is described in the Supporting Material. Values of clogP and t-PSA were calculated by FAFDrugs4 website³⁹.

Expression and purification of SaIMPDH.

SaIMPDH was expressed as a N-terminal 6-His tagged protein using pET28b in BL21(DE3)ΔguaB bacteria. The cells were grown in LB media with 50 μg/ml of kanamycin at 37°C to an OD₆₀₀ 0.5–0.8. Protein expression was initiated by 0.25 mM isopropyl β-D-thiogalactoside (IPTG) and the culture was maintained at 30°C overnight. Cells were harvested, resuspended in lysis buffer [50 mM phosphate (pH 8.0), 500 mM KCl, 10 mM imidazole, 0.5 mM (tris(2-carboxyethyl)phosphine) (TCEP) and 10% glycerol], and stored at −80 °C. SaIMPDH was purified according to a standard protocol for His tagged protein. Protease inhibitor cocktail (Roche, Indianapolis, IN; 50 mL/g of wet cells) were added to the thawed cell suspension. The cells lysed on ice by sonication. The lysate was clarified by centrifugation at 18000g for 1 h and filtered through a 0.44 μm membrane. Clarified lysate was applied to a 5 mL Ni-NTA column (McLab, San Francisco CA). The column was washed with 10 column volume (CV) lysis buffer, 10 CV of lysis buffer containing 30 mM imidazole, and the protein was eluted with the same buffer containing 250 mM imidazole. SaIMPDH was dialyzed against 20 mM Hepes (pH 8.0), 150 mM KCl, 3 mM EDTA and 2 mM DTT, concentrated, flash-frozen, and stored in liquid nitrogen.

Enzyme assays.

IMPDH assays were performed in 50 mM Tris-HCl, pH 8.0, 100 mM KCl, 3 mM EDTA and 1 mM DTT. Activity was routinely assayed in the presence of 20–50 nM IMPDH at 25 °C. NADH production was monitored either by following absorbance change at wavelength 340 nm using a Hitachi U-2000 spectrophotometer (ε = 6.2 mM^–1 cm^–1) or by following fluorescence on a Biotek plate reader (excitation wavelength = 340 nm, emission wavelength = 460 nm). Each determination of K_i,app was derived from duplicate measurements of enzyme activity in the presence of twelve different inhibitor concentrations. Reported values are averages of at least two independent determinations. Enzyme was incubated with inhibitor (50 pM - 100 μM) for 10 min at room temperature prior to addition of substrates. The concentrations of IMP and NAD⁺ were fixed at 1.0 mM (saturating) and 1.0 mM (~2.5 × K_m), respectively, for SaIMPDH. K_i,app values were calculated for each inhibitor according to Equation 1 (when K_i,app >> [E]) or Equation 2 (when K_i,app ~ [E]) using the SigmaPlot program (SPSS, Inc.):

$$v_i=v_0/\left(1+\left[\text{I}\right]/K_{i,app}\right)$$ (1)

where v_i is initial velocity in the presence of inhibitor (I) and v₀ is the initial velocity in the absence of inhibitor. The Morrison equation (Equation 2) was used to evaluate tight-binding inhibitors⁵⁷.

$$v_i/v_0=\left\{E_0–I_0–K*+{\left[{\left(E_0–I_0–K_{i,app}\right)}^2+4•E_0•K_{i,app}\right]}^{0.5}\right\}/2•E_0$$ (2)

where v_i is the initial velocity in the presence of inhibitor, v₀ is the initial velocity in the absence of inhibitor, E₀ is the total concentration of enzyme, I₀ is the total concentration of inhibitor, and K_i,app is the apparent inhibition constant. Alternatively Dynafit was used to determine K_i,app⁵⁸.

Antibacterial activity.

Broth micro-dilution MICs were performed following CLSI standard procedures against S. aureus (ATCC 13709, Smith). Stock compound solutions were prepared from powders at a concentration of 0.64 mg/mL (0.32 mg/mL for vancomycin) in 100% DMSO. Two-fold serial dilutions were performed in 96 well plates. The bacterial strain was prepared from an overnight culture plate by re-suspending in sterile saline and adjusting to a 0.5 McFarland standard confirmed by OD. The bacterial suspension was then diluted 1:100 in RPMI without phenol red or RPMI without phenol red containing 1 mM guanosine media. The prepared bacterial inoculum was then added to each corresponding 96 well plate at 190 uL per well, resulting in the top concentration of test article to be 32 μg/mL. The inoculated plates were incubated at 35°C for 18–20 h. The MICs were identified as the first well with no growth by visual inspection of the 96 well plates. Bacterial input inoculum was verified by plating the prepared suspension on TSA plates and incubating overnight resulting in 5.2×10⁵ CFU/well.

Cytotoxicity.

All compounds were dissolved in DMSO and further diluted with culture medium before use in tissue culture assays (final DMSO concentrations were ≤ 0.1%). To determine cytotoxicity, LDH release was measured with the LDH Cytotoxicity Assay Kit (Pierce) according to manufacturer’s protocol. Briefly, 96 well plates were seeded with 13,000 HepG2 cells (all other cell lines seeded at 6,000 to 8,000 cells per well) and the cells were cultured for 24 h prior to drug treatment. The cells were incubated in 110 μL of EMEM containing compound or DMSO (vehicle only, control) for 24 h at 37°C. At least four wells from each plate were used as either ‘spontaneous’ LDH controls or as ‘maximum’ LDH controls per manufacturer’s instructions. Cytotoxicity was determined by measuring absorbance on a microplate reader. Data represent two independent experiments each performed in quadruplicate (n =8).

Plasma protein binding.

Compound (10 μM) was incubated in mouse plasma (BIOIVT, Westbury NY). The free fraction was collected by filtration through a 3K Microcon® centrifugal filter. The concentration of free inhibitor was determined by measuring inhibition of BaIMPDH.

Resistant strains.

Resistant strains were selected for growth on RPMI agar medium in the presence of 25 (25 μg/mL, 25xMIC). The S. aureus guaB gene is the third gene in a four gene operon, so sequencing focused on the coding region and did not include the promoter. Oligonucleotides SaF1 and SaR1 were used to amplify the coding sequence of guaB. DNA sequencing was performed with primers SaF1, SaR1 and internal primers SaF2 and SaR2. Sequences were deposited in GenBank with the following accession numbers: guaBwt, MW436410; guaB-G445S, MW436411 and guaB-Y446C, MW436412.

Name	sequence	purpose
SaF1	AATTGTTTTAGGTGCAATCTCTGC	Outside gene
SaR1	TGGCTACCAAAGTCTAAGACAAGG	Outside gene
SaF2	TTTTAACAAACCGCGACTTACG	Inside gene
SaR2v2	TACCATGTTTGCGTGCTTCA	Inside gene

Construction of guaB::C307A.

The C307A mutation was inserted guaB using overlapping PCR to construct a fragment with ~ 1 kb flanking sequence to facilitate homologous recombination. The PCR fragment was cloned into temperature-sensitive shuttle plasmid pMAD⁵⁹ in IM08B, an E. coli DC10B derivative that methylates DNA according to the CC8 hsdMS methylation complex⁶⁰. Plasmid DNA from this E. coli strain was directly transformed into NRS384. Transformants were selected by erythromycin (Erm, 5 μg/ml) resistance and blue color on X-gal at 30°C. Bacteria were grown in Erm-containing medium at 30°C for the isolation of plasmid to confirm the presence of right construct. Cells with right construct from 30°C cultures was diluted to grow at 43°C for overnight to promote single crossover (Erm^R and light blue), and subsequently diluted to grow at 30°C without erythromycin containing medium to promote homologous recombination to yield white and Erm^S colonies. Mutants were verified by chromosomal PCR followed by DNA sequencing to confirm the right mutation and the junction of the recombination. The guaB::C307A strains could not grow in the absence of a guanosine. This growth defect was complemented by pEPSA5 carrying xylose-inducible wild-type guaB.

In vivo studies.

All experiments were performed in accordance with institutional guidelines as defined by Institutional Animal Care and Use Committees of Brandeis University, Avastus Preclinical Services (Cambridge, MA) and NeoSome Life Sciences, LLC (Lexington, MA), as well as OLAW standards. Pharmacokinetic studies were performed by GVK Biosciences (Hyderbad India). Compounds were administered in a formulation of 20% DMSO/ 80% diluted liquid fill⁶¹. Liquid fill was prepared by combining 375 g of Etocas 35NF, 4.4 g of pluronic acid F68, 50.8 g of polyethylene glycol 400, and 10.8 ml of water and stirring overnight. Aliquots of 15 ml were stored at −80°C. Liquid fill was thawed as needed, and mixed with 33 ml of water to make a diluted liquid fill stock. Compounds were dissolved in DMSO and mixed 1 part to 4 parts diluted liquid fill to give a final dosage volume of 0.2 ml for a 20 g mouse.

Efficacy and virulence studies were performed by Avastus Preclinical Services (Cambridge, MA) and NeoSome Life Sciences, LLC (Lexington, MA). CD-1 female mice (18–20 g) were obtained from Charles River Laboratories and allowed to acclimate for 5 days prior to start of study. Animals were housed 5 per cage with free access to food and water. Bacteria were prepared from an overnight liquid culture. The overnight culture was diluted 1:10 and grown for another 5 h while shaking. The 5 h culture was serially diluted in 5% hog gastric mucin (pH=7.4) to a concentration to achieve at least 90% mortality within 48 hours after infection. Mice were infected with 0.5 ml of bacterial suspension via intraperitoneal injection, 5 mice per test article group (10 mice for the infection control groups). The test articles were delivered 1 h post infection as a slow bolus IV administration at a volume of 10 mL/kg. Infection control mice were dosed with test article vehicle.

Abscess experiments were performed at NeoSome Life Sciences, LLC (Waltham MA). Female CD-1 mice S. aureus ATCC 13709 (Smith), was prepared for infection from an overnight plate culture. A portion of the plate was resuspended in sterile saline and adjusted to an OD of 0.1 at 625 nm. The adjusted bacterial suspension was further diluted to target an infecting inoculum of 1.0×10⁶ CFU/mouse, the actual inoculum size was 1.4×10⁶ CFU/mouse. Plate counts of the inoculum were performed to confirm inoculum concentration. Under anesthesia, mice were injected subcutaneously in the right flank with 0.25 mL of the prepared bacterial inoculum mixed with cyclodextrin beads. Animals were returned to their home cages and allowed to recover from the anesthesia. Beginning at two hours post infection mice were dosed with either test article or positive control antibiotic. Mice receiving test compounds were dosed directly into the subcutaneous abscess in a 100 uL volume, a second dose of test agent were delivered 12 hours later. Linezolid served as the positive control agent and was delivered orally at 20 mg/kg. Five animals were dosed per group. One group of five mice were euthanized at initiation of therapy (T = 0) and CFUs determined. All remaining mice were euthanized at 26 h post infection. Abscesses were aseptically removed, homogenized to a uniform consistency, serially diluted and plated on bacterial growth media. The CFUs were enumerated after overnight incubation. The average and standard deviations for each group were determined.

Supplementary Material

Modi supporting

Table S1. Comparison of inhibition of BaIMPDH and SaIMPDH

Table S2. Crystal and Data Collection and Refinement Statistics.

Table S3. Virulence of Sa12 and Sa26.

Table S4. In vivo activity of SaIMPDH inhibitors.

Table S5. Virulence of guaB::C307A strains.

Table S6. List of strains related to guaB C307A mutations.

Table S7. Compound naming convention.

Figure S1. Physicochemical properties of antibacterial activity.

Figure S2. P226-resistant strain Sa12 displays a growth phenotype.

Figure S3. Cytotoxicity of select compounds.

Figure S4. S. aureus JE2 strains guaB::C307A require guanine or xanthine for growth.

Methods, including enzyme purification and crystallization, data collection and structure solution and refinement and virulence studies

Compound characterization

Abbreviations Used:

Ba

Bacillus anthracis

IMPDH

inosine 5’-monophosphate dehydrogenase

Mtb

Mycobacterium tuberculosis

Sa

Staphylococcus aureus

SAR

structure activity relationship

t-PSA

topological polar surface area