Pyridodiazepine Amines Are Selective Therapeutic Agents for Helicobacter pylori by Suppressing Growth through Inhibition of Glutamate Racemase but Are Predicted To Require Continuous Elevated Levels in Plasma To Achieve Clinical Efficacy

Abstract

A pyridodiazepine amine inhibitor of Helicobacter pylori glutamate racemase (MurI) was characterized. The compound was selectively active against H. pylori, and growth suppression was shown to be mediated through the inhibition of MurI by several methods. In killing kinetics experiments, the compound showed concentration-independent activity, with about a 2-log loss of viability in 24 h. A demonstration of efficacy in a mouse infection model was attempted but not achieved, and this was attributed to the failure to attain extended exposure levels above the MIC for >95% of the time. This index and magnitude were derived from pharmacokinetic-pharmacodynamic (PK-PD) studies with amoxicillin, another inhibitor of peptidoglycan biosynthesis that showed slow killing kinetics similar to those of the pyridodiazepine amines. These studies indicate that MurI and other enzymes involved in peptidoglycan biosynthesis may be less desirable targets for monotherapy directed against H. pylori if once-a-day dosing is required.

INTRODUCTION

Helicobacter pylori is the causative pathogen of gastric ulcers and is implicated in gastric cancer (1–3). It is recommended that treatment to eradicate the organism be initiated upon receiving a diagnosis (4). Current treatments consist of a combination of proton pump inhibitors with up to three nonselective antibiotics (4–6). The complicated regimens of multiple pills combined with the nonselectivity of the antibiotics can lead to side effects, result in poor patient compliance, and therefore result in an unsatisfactory clinical outcome (7). In addition, drug resistance is emerging, rendering some therapies less effective (8). Consequently, there is a need for novel improved therapies. Ideally, such novel therapies would be selective for H. pylori, require a simpler regimen (such as one pill once a day), and be orally bioavailable.

Genomic, biochemical, and structural biology studies identified glutamate racemase (MurI) as a target for which selective inhibitors against H. pylori could be developed (9). Glutamate racemase converts l-glutamate to d-glutamate, which is ligated to UDP-Mur-N-acetyl-Ala (UDP-MurNAc-Ala) by MurD to form UDP-MurNAc-Ala-Glu (10); UDP-MurNAc-Ala-Glu is an intermediate in the synthesis of the essential building block UDP-MurNAc-pentapeptide that is used in the formation of the peptidoglycan layer that protects the cell against osmotic rupture (Fig. 1; 11). Efforts to identify inhibitors against this essential enzyme were undertaken, and an initial series using pyrazolopyrimidinediones was identified using high-throughput screening (HTS) (9). Medicinal chemistry efforts on this series led to selective inhibitors for H. pylori that suppressed growth through the inhibition of MurI (12). This series was also amenable to improvements in oral bioavailability (13), but high systemic exposure of unbound compound, which was needed to demonstrate efficacy in an animal model, was not achieved with potent analogues due to high protein binding. As a result, a demonstration of efficacy in a suitable animal model was not achieved with the pyrazolopyrimidinediones, and the progress of the series was halted.

FIG 1.

Cytoplasmic steps of the peptidoglycan biosynthetic pathway. Amino acids are sequentially added to UDP-N-acetyl muramic acid (UDP-Mur) using MurC to MurF to form the UDP-MurNac-pentapeptide, the final cytoplasmic peptidoglycan precursor (11). MurI provides d-glutamate. The inhibition of MurI results in the inhibition of peptidoglycan biosynthesis at the d-glutamate addition stage (MurD). As a result, the inhibition of MurI will result in the depletion of MurD, MurE, and MurF products and the accumulation of UDP-N-acetyl muramic acid-alanine (MurD substrate). Ala, alanine; Glu, glutamate, Dap, diaminopimelic acid.

Another round of HTS presented a second scaffold, pyridodiazepine amine, and a medicinal chemistry program was initiated with this series to lower protein binding while maintaining or improving potency (14). In this communication, the microbiological and pharmacokinetic properties of an advanced analog and the impacts these have on the pharmacological outcome in animal efficacy experiments are reported.

(Part of these studies were presented at the 46th Interscience Conference for Antimicrobial Agents and Chemotherapy [ICAAC], 17 to 21 September 2006, San Francisco, CA, as abstracts F2-1174 [15] and A2-1106 [16].)

MATERIALS AND METHODS

MIC determination.

Broth microdilution MICs were determined in 96-well plates using 2-fold serial dilutions of compound, as previously described (17).

Killing kinetics.

Bactericidal studies were performed with H. pylori strain SS1 (18) at 37°C in Brucella broth containing 5% fetal bovine serum under a 5% O₂, 10% CO₂, and 85% N₂ atmosphere (Queue Cellstar incubator). The cells were transferred onto blood agar plates, incubated overnight under a 5% O₂, 10% CO₂, and 85% N₂ atmosphere, and then diluted into assay medium to a starting concentration of ∼10⁵ CFU/ml. The compounds were added at multiples of the MIC, and samples taken at different time points were serially diluted 10-fold in assay medium and the dilutions plated onto blood agar. After 5 days of incubation under a 5% O₂, 10% CO₂, and 85% N₂ atmosphere at 37°C, the number of colonies was counted, and the CFU/ml was plotted against time.

Mode of action studies.

The frequency of spontaneous resistance development was determined and genetic manipulations to map the resistance locus and peptidoglycan precursor pool analyses were performed as described before (12).

Measurement of physical chemical and pharmacokinetic properties.

Plasma protein binding, equilibrium solubility, and clearance in rats were measured as described before (19). ClogP, the predicted octanol-water partition coefficient, was calculated using the BioByte algorithm (Daylight Chemical Information Systems, Laguna Niguel, CA).

Gastric fistula experiments.

Female Sprague-Dawley rats with fistula implanted in their stomachs were obtained from Charles River Laboratories. The rats were fasted 24 h before the start of the experiments. At the start, the stomachs were rinsed with distilled water at 37°C until clean, upon which cannula were connected to the fistula. Compound A was formulated in a polyethylene glycol (PEG) 400-to-saline ratio of 1:3, and a 5-mg/kg of body weight bolus dose was injected in the tail vein. Gastric juice was collected for 2 h at four 30-min intervals. Plasma samples were also collected during this time period from a satellite group of rats that was dosed the same way with compound A as were the cannulated rats. The concentrations of compound A in plasma and gastric juice were determined by liquid chromatography-tandem mass spectrometry (LC-MS/MS) from three rats, and the average ± standard deviation (SD) was plotted against time.

Mouse infection model.

Mice were infected with a mouse-adapted strain of H. pylori SS1, as described previously (18). Briefly, each mouse was inoculated with a bacterial suspension of H. pylori SS1 obtained from 2-day liquid cultures (approximately 10⁹ organisms/ml). The animals were dosed a total of three times in a five-day period with 0.1 ml using direct gastric instillation. The infection was allowed to establish itself for 7 days following the third dose of bacteria prior to initiating drug treatment. When dosed with compound A, the mice were also dosed orally with aminobenzotriazole (ABT) (20), either at 100 mg/kg once or 50 mg/kg twice daily, 2 h before compound A was administered to inhibit compound A metabolism via CYP450. Compound A and amoxicillin were administered orally for 4 days on a dosing regimen of every 6 h (q6h). The H. pylori counts per stomach were enumerated 6 h after the last dose to assess infection suppression (10 mice per group). Compound A was formulated in 11.2% Captisol (pH 4.2) and amoxicillin in 0.75% hydroxypropyl methylcellulose (HPMC). For the pharmacokinetic-pharmacodynamic (PK-PD) experiments, amoxicillin was fractionated as one, two, or four doses administered each day for 4 days. Plasma exposure was assessed in a satellite group of three animals.

RESULTS

Activity and mode of action of HTS hit.

HTS with H. pylori MurI against another compound collection identified a new scaffold (Fig. 2) that showed potent and selective activity against the enzyme (14). This potent and selective enzyme inhibition extended to growth inhibition as well, with low MICs for H. pylori, while the MICs for other bacterial and eukaryotic species were very elevated (Table 1). Three lines of evidence showed that this compound suppressed the growth of H. pylori through the inhibition of MurI. First, a mutant overexpressing MurI (12) showed an 8-fold elevation in MIC compared to that of the wild-type parental strain (Table 1). In addition, resistant mutants that were generated with frequencies around ∼10⁻⁶ at 2× to 8× the MIC (Table 2) showed large shifts in the MICs (>16-fold) against the HTS hit, but this was not the case for antibiotics from other classes, with the exception of the previously identified inhibitors of MurI (pyrazolopyrimidinediones; data not shown). MurI was sequenced for some of the isolated mutants, and several mutations (A35T, A75T, and C162Y) were found in separate mutants. The transformation of these mutations into a susceptible isolate led to reduced susceptibilities for the MurI inhibitors, confirming that these mutations were the cause of resistance. Finally, peptidoglycan precursor pool analyses also showed growth suppression of H. pylori by the HTS hit through the inhibition of MurI. The accumulation of UDP-MurNAc-Ala, with a concomitant full disappearance of UDP-MurNAc-pentapeptide, was seen when H. pylori was grown in the presence of the HTS hit (Fig. 3). This profile is expected for a MurI inhibitor, since the inhibition of MurI will block the formation of d-Glu, leaving MurD without d-Glu to ligate to UDP-MurNAc-Ala, resulting in the accumulation of this substrate and the complete disappearance of the final UDP-MurNAc-pentapeptide product.

FIG 2.

Structures of compounds described in this study. Left, HTS hit; right, lead compound A.

TABLE 1.

MICs of HTS hit, compound A, and amoxicillin against several bacterial and eukaryotic species^a

Organism (n)	Description	MIC (μg/ml) of:
		HTS hit	Compound A	Amoxicillin
Helicobacter pylori (4)	Clinical isolates	0.5 to 2	0.13 to 0.5	0.008 to 0.25
H. pylori SS1	Strain used in efficacy studies	0.5	0.25	0.25
H. pylori ARHp80	Clinical isolate	2	0.25	0.008
H. pylori ARHp80.2 (MurI+)	Isogenic MurI-overexpressing strain of ARHp80 (12)	16	64	0.008
H. pylori ARHp80 HefC−	Isogenic efflux compromised strain of ARHp80 (17)	0.5	0.13	0.0009
Streptococcus pneumoniae (4)	Clinical isolates, including PRSP	>64	>64	0.016 to 8
Staphylococcus aureus (2)	Clinical isolates, including MRSA	>64	>64	0.25 to 64
Haemophilus influenzae (3)	Clinical isolates, including β-lactamase positive	>64	>64	0.125 to >64
Escherichia coli	Clinical isolate	≥64	>64	>64
Moraxella catarrhalis	Clinical isolate	8	>64	NDb
Fusobacterium necrophorum	Clinical isolate	ND	>64	ND
Bacteroides thetaiotaomicron	Clinical isolate	ND	>64	ND
Bacteroides fragilis	Clinical isolate	ND	>64	ND
Clostridium difficile	Clinical isolate	ND	32	ND
Fusobacterium mortiferum	Clinical isolate	ND	>64	ND
Candida albicans	Yeast	>64	>64	>64
A549	Human cell line	ND	>64	>64

^aPRSP, penicillin-resistant S. pneumoniae; MRSA, methicillin-resistant S. aureus.

^bND, not determined.

TABLE 2.

Resistance frequencies of HTS hit and compound A against three strains of H. pylori

Compound	Selection (× MIC)	Frequency with strain:
		ARHp55	ARHp80	SS1
HTS hit	2	1.2 × 10−6	1.3 × 10−6	2.7 × 10−6
	4	9.5 × 10−7	9.8 × 10−7	2.2 × 10−6
	8	9.2 × 10−7	8.3 × 10−7	1.3 × 10−6
Compound A	2	1.8 × 10−7	7.7 × 10−8	8.9 × 10−8
	4	1.2 × 10−7	5.1 × 10−8	8.3 × 10−8
	8	7.3 × 10−8	2.6 × 10−8	8.6 × 10−8

FIG 3.

Peptidoglycan precursor pools of H. pylori in the presence (bottom) and absence (top) of pyridodiazepine amine HTS hit. Exponentially grown H. pylori was subjected to compound for one generation at 2× the MIC. Peptidoglycan precursor pools were analyzed using reversed-phase high-performance liquid chromatography (HPLC) from trichloroacetic acid (TCA) extracts, as described before (12). The accumulation of UDP-MurNAc and the decrease in UDP-MurNAc-pentapeptide when H. pylori cells are exposed to the HTS hit are indicated with arrows.

Activity of compound A.

Since the pyridodiazepine amine HTS hit already showed potent activity with the desired mode of action, optimization was primarily focused on improving physical chemical and pharmacokinetic properties while maintaining or improving microbiological activity. These improvements were needed to allow testing in an animal efficacy model, something that eluded the previously studied pyrazolopyrimidinedione series. A medicinal chemistry program that was undertaken to achieve these goals (14) yielded compound A (Fig. 2) as a promising candidate. Compound A showed improved microbiological activity over that of the HTS hit and retained selectivity, and the desired mode of action was confirmed by the elevated MICs observed with the MurI-overexpressing mutant (Table 1) and pyridodiazepine amine-resistant isolates (data not shown). The MIC of compound A for the parental wild-type strain was similar to that of the isogenic hefC gene knockout (Table 1), showing that the activity of compound A was not affected by the main efflux system present in H. pylori (17). In addition, the frequencies of spontaneous resistance development were reduced compared to those with the HTS hit (Table 2).

The compound was not bactericidal, showing only slow concentration-independent killing kinetics that resulted in about a 2-log loss of viability in 24 h of compound exposure (Fig. 4). Killing continued over time, with full sterilization achieved after 48 h. These relative slow killing kinetics were also observed with the β-lactam amoxicillin (Fig. 4), another inhibitor of peptidoglycan biosynthesis, and relative slow killing kinetics have been reported for other β-lactams as well (21, 22). In contrast, clarithromycin, an inhibitor of protein biosynthesis, showed rapid killing kinetics (Fig. 4), as did a fluoroquinolone (data not shown).

FIG 4.

Killing kinetics of H. pylori SS1 with compound A at 0× (cross), 5× (triangle, solid line), 16× (circle, solid line), and 160× (square, solid line) the MIC, amoxicillin at 10× (triangle, dashed line) and 100× (square, dashed line) the MIC, and clarithromycin at 10× (triangle, dotted line) the MIC.

Pharmacokinetics of compound A.

Compound A had greatly improved solubility and reduced protein binding compared to those of the HTS hit (Table 3), and these were much better compared to the potent analogs of the pyrazolopyrimidinedione series, which had solubilities only in the single-digit micromolar range and were >95% bound to protein. The compound also showed promising oral bioavailability (Table 3). The clearance of compound A in rats and mice was moderate but could be vastly improved by pretreating animals with 100 mg/kg of the cytochrome P450 metabolism inhibitor aminobenzotriazole (ABT) (20). Single-dose pharmacokinetics with compound A in the presence of ABT in mice showed linear exposure and plasma blood levels of unbound compound exceeding the MIC for a significant amount of time (see Table S1 in the supplemental material), something that was never accomplished with the pyrazolopyrimidinediones due to high protein binding. Because of the ability to achieve high free plasma exposures, efficacy experiments with compound A in the presence of ABT were initiated in the H. pylori infection mouse model (18).

TABLE 3.

Physical, chemical, and pharmacokinetic properties of HTS hit, compound A, and amoxicillin

Property	HTS hit	Compound A	Amoxicillin
Solubility (μM)	<0.6	1,365	>1,000
ClogP	5.35	3.13	−1.87
Protein binding (% bound)
Mouse	NDa	86	ND
Human	99.7	82	<30
Microsomal clearance (μl/min/mg)
Mouse	>100	30	ND
Human	43	13	<4
Rat pharmacokinetics
Clearance (ml/min/kg)	99	25	16b
t1/2 (h)c	1.5	0.5	0.9b
Bioavailability (%)	ND	39	51b

^aND, not determined.

^bData taken from reference 33.

^ct_1/2, elimination half-life.

Efficacy studies.

Compound A was dosed every 6 h for 4 days at 15, 30, 60, and 120 mg/kg, while ABT was administered once a day at 100 mg/kg, starting 2 h before the first dose of the day. No reduction in H. pylori was observed in the stomachs of the treated animals, even at the highest dose administered (see Fig. S1 in the supplemental material). Based on the single-dose pharmacokinetics (see Table S1 in the supplemental material), the maximum dose administered in the efficacy study (480 mg/kg/day) was expected to result in exposure levels of compound A that were above the MIC for 100% of the time of the efficacy experiment. However, pharmacokinetic studies involving sampling multiple times a day over an extended period revealed that while high exposure of compound A was achieved on the first day, exposure progressively reduced over time, resulting in compound A being above the MIC for <50% of the time of the efficacy experiment.

Amoxicillin, a β-lactam that inhibits peptidoglycan biosynthesis by binding to penicillin-binding proteins (23), was used as a positive control in the efficacy studies to confirm that the inhibition of targets in peptidoglycan biosynthesis in general, and therefore MurI, resulted in a reduction in the H. pylori burden in this model. Amoxicillin was indeed efficacious, although high doses were needed to achieve this efficacy (Fig. 5). To understand what level of exposure was needed for inhibitors of peptidoglycan biosynthesis to result in a reduction in H. pylori burden, a full PK-PD dose fractionation study was undertaken with amoxicillin to determine the plasma PK-PD index and magnitude correlated with efficacy. The time above the MIC was identified as the index, with a need for amoxicillin to be above this level >95% of the time to achieve a 2-log reduction in H. pylori burden in the stomach (r² = 0.9).

FIG 5.

H. pylori counts in stomachs of H. pylori-infected mice upon treatment with vehicle (white), 50 mg/kg/day (light gray), 200 mg/kg/day (dark gray), 800 mg/kg/day (black), and 1,600 mg/kg/day (* indicates that the CFU/stomach was below the limit of detection [<100]) of amoxicillin. The compound was dosed every 6 h for 4 days. Data are presented as means and standard errors.

Compound penetration through gastric mucosa.

Compound A was dosed intravenously in rats, and gastric juice was collected over time. Significant amounts of compound A were collected from the gastric juice (see Fig. S2 in the supplemental material), indicating that compound A can pass the gastric mucosa and thus reach the site of infection.

DISCUSSION

A pyridodiazepine amine was identified through HTS with H. pylori MurI (14). This study showed the HTS hit to be a selective inhibitor for H. pylori, and the suppression of growth was confirmed to be mediated by the inhibition of MurI. The selectivity is most likely the result of the highly specific enzyme inhibition observed with this compound (14), although reduced compound penetration in other organisms cannot be excluded as a contribution to the reduced activity against other species. The suppression of growth was confirmed to be mediated through the inhibition of MurI, with an H. pylori construct that overexpressed MurI showing reduced susceptibility, and the isolation of pyridodiazepine amine-resistant H. pylori mutants with mutations mapping to MurI. Although these types of experiments might point to a resistance mechanism rather than to the target of a compound, the fact that the compound inhibits MurI and that the biochemical potency of the inhibitor was decreased against the purified mutant proteins (24) shows that MurI must be the target. This was backed up by peptidoglycan precursor analyses of H. pylori treated with the pyridodiazepine amine that showed a profile expected for the inhibition of MurI. Thus, straight out of HTS, without the need for optimization, an inhibitor of MurI was identified that suppressed the growth of H. pylori through the inhibition of that enzyme. This was in contrast to a previously identified pyrazolopyrimidinedione HTS hit that required further optimization to achieve growth suppression through the inhibition of MurI (12).

Similar to the pyrazolopyrimidinediones (12), resistant mutants were easily generated, and mutations were found across MurI (A35T, A75T, and C162Y) outside residues that interact with the compound (9, 14). Structural biology and biochemical studies have provided a rationale as to why these mutations confer resistance against both series of MurI inhibitors (9, 24), and it is likely that future inhibitors of H. pylori MurI will suffer from the same issue. While the frequencies of spontaneous resistance were improved for compound A compared to those of the HTS hit, possibly as a result of the potency improvement, they remained relatively high enough to warrant clinical development.

Compound A showed slow killing kinetics with H. pylori, and similar results were seen with amoxicillin. Thus, regardless of whether peptidoglycan biosynthesis was inhibited at the cytoplasmic level through the inhibition of MurI by pyridodiazepine amines, or at the periplasmic level through the inhibition of penicillin-binding proteins by β-lactams, slow killing kinetics seems to be a hallmark for peptidoglycan inhibitors in general for H. pylori, and it is likely the result of the relatively slow growth of the organism (∼3 h generation time; 25). This study found rapid killing kinetics for clarithromycin, and this is in contrast to one study that reported clarithromycin as an agent with slow killing kinetics (21). The reason for these discrepant results may be either the use of early stationary cells in that study, rather than early exponential growing cells, or the choice of strain, as killing kinetics with clarithromycin has been shown to be strain dependent (22).

Efficacy in a mouse stomach model was attempted with compound A, but no reduction in H. pylori burden was observed. Poor outcomes with monotherapy are not uncommon for treatments directed against H. pylori (26). There are several possible explanations for the observed failure with compound A. While the contributions of the topical and systemic routes of compound delivery in the treatment of H. pylori are still unclear (27), it is likely that the topical delivery of compound A to H. pylori is limited due to the ionization of compound A (pK_a, 5.2) at low pH, affecting uptake in H. pylori. It is also possible that the agent did not reach the site of infection through the systemic route, although this seems to be an unlikely explanation, as gastric fistula experiments showed that at least in rats, compound A can penetrate the gastric mucosa where H. pylori resides. Although one cannot rule out that MurI is not a good target in vivo, the fact that amoxicillin treatment led to a reduction in H. pylori burden shows that the inhibitors of peptidoglycan biosynthesis can be agents for therapy directed against this organism, albeit at high concentrations. The most likely reason for the failure to achieve efficacy was that plasma levels of compound A were above the MIC only about 50% of the time in the efficacy experiment and not the >95% level presumed necessary to achieve efficacy. This index and magnitude were based on the PK-PD analysis of amoxicillin. Since both amoxicillin and compound A are inhibitors of peptidoglycan biosynthesis with equal slow killing kinetics, the time above the MIC with an equally high magnitude can be expected to be the index and magnitude for the compound A to achieve efficacy.

The identification of time above the MIC as the PK-PD index needed for amoxicillin to achieve efficacy in this model is in agreement with other studies using other animal efficacy models and pathogens (28–31). However, the magnitude of >95% is much higher than the 30 to 60% reported in those studies. The high magnitude needed to achieve efficacy for amoxicillin in the H. pylori mouse model compared to that in other models can likely be attributed to the much slower killing kinetics seen with H. pylori compared to those with other species, and the inability to maintain such a high magnitude may be another important factor in the poor clinical outcomes seen with amoxicillin monotherapy (32) beyond the proposed antrum-body transitional zones that were identified as sanctuary sites for H. pylori (26). Alternatively, the high magnitude might have been the result of the fact that the MIC determinations were not performed according to CLSI methodologies, which possibly would have led to a lower MIC and increased magnitude of time above the MIC; however, the MIC value of amoxicillin is in line with what was reported when the CLSI guidelines were used (26).

The low levels of compound A were unexpected based on single-dose pharmacokinetics experiments; the reasons for this are unclear but might be due to the upregulation of CYP450 activity, rendering ABT less effective. Efforts to increase the exposure of compound A by dosing ABT twice a day at 50 mg/kg, instead of 100 mg/kg once a day, resulted in higher exposure on the first day but not in subsequent days (data not shown), so the only way forward to demonstrate efficacy in this animal model would be to identify analogs with much lower clearance and/or vastly improved potency.

In order to improve current therapies directed against H. pylori, a single agent with high selectivity, oral bioavailability, and once-a-day dosing is preferred. Two different inhibitor series of H. pylori MurI have been identified that satisfy the first two criteria (9, 12, 14 and this study), validating MurI as a selective target for therapy directed against H. pylori. However, PK-PD studies suggest that agents designed to inhibit this target need to remain above the MIC for almost 100% of the time in order to achieve efficacy. In order to accomplish this with once-a-day dosing, compounds need to be either cleared slowly and/or have very low MICs. While not unfeasible, it is more likely that inhibitors of MurI would require multiple doses a day to achieve the high PK-PD target. For that reason, inhibitors of H. pylori MurI are less desirable candidates for monotherapy directed against H. pylori, and it is expected that this would apply to other inhibitors of peptidoglycan biosynthesis as well, regardless of the enzyme they target in the pathway. It may be more sensible to explore targets that show rapid killing kinetics for therapies directed against H. pylori.