LpxC inhibitors as new antibacterial agents and tools for studying regulation of lipid A biosynthesis in Gram-negative pathogens.

UNLABELLED: The problem of multidrug resistance in serious Gram-negative bacterial pathogens has escalated so severely that new cellular targets and pathways need to be exploited to avoid many of the preexisting antibiotic resistance mechanisms that are rapidly disseminating to new strains. The discovery of small-molecule inhibitors of LpxC, the enzyme responsible for the first committed step in the biosynthesis of lipid A, represents a clinically unprecedented strategy to specifically act against Gram-negative organisms such as Pseudomonas aeruginosa and members of the Enterobacteriaceae. In this report, we describe the microbiological characterization of LpxC-4, a recently disclosed inhibitor of this bacterial target, and demonstrate that its spectrum of activity extends to several of the pathogenic species that are most threatening to human health today. We also show that spontaneous generation of LpxC-4 resistance occurs at frequencies comparable to those seen with marketed antibiotics, and we provide an in-depth analysis of the mechanisms of resistance utilized by target pathogens. Interestingly, these isolates also served as tools to further our understanding of the regulation of lipid A biosynthesis and enabled the discovery that this process occurs very distinctly between P. aeruginosa and members of the Enterobacteriaceae. Finally, we demonstrate that LpxC-4 is efficacious in vivo against multiple strains in different models of bacterial infection and that the major first-step resistance mechanisms employed by the intended target organisms can still be effectively treated with this new inhibitor. IMPORTANCE: New antibiotics are needed for the effective treatment of serious infections caused by Gram-negative pathogens, and the responsibility of identifying new drug candidates rests squarely on the shoulders of the infectious disease community. The limited number of validated cellular targets and approaches, along with the increasing amount of antibiotic resistance that is spreading throughout the clinical environment, has prompted us to explore the utility of inhibitors of novel targets and pathways in these resistant organisms, since preexisting target-based resistance should be negligible. Lipid A biosynthesis is an essential process for the formation of lipopolysaccharide, which is a critical component of the Gram-negative outer membrane. In this report, we describe the in vitro and in vivo characterization of novel inhibitors of LpxC, an enzyme whose activity is required for proper lipid A biosynthesis, and demonstrate that our lead compound has the requisite attributes to warrant further consideration as a novel antibiotic.

INTRODUCTION

The war against antibiotic resistance rages on for the anti-infective community, as the emergence and spread of mechanisms that effectively subvert the activity of marketed antibacterial agents continue at a terrifying rate. While efforts to fight this battle have been limited in number, there have been valiant attempts to develop new analogs of existing antibiotic classes, with several of these upgraded molecules advancing to clinical trials recently (1–3). And while each of these agents will undoubtedly prove efficacious against many target species, the potential gaps in strain coverage due to the expression of preexisting resistance mechanisms will likely limit their widespread utility, leaving many patients with very few, if any, viable treatment options. As we continue in our quest to identify emerging pathogens and develop new anti-infective agents to combat multidrug-resistant (MDR) strains, antibacterial discovery efforts must be broadened to include the exploration of new cellular pathways, especially since target-based resistance should not exist against clinically unprecedented cellular targets. Although there are multiple examples of this approach, one of the most intriguing and promising novel pathways for the treatment of Gram-negative bacteria is lipid A biosynthesis.

The outer membrane of Gram-negative pathogens, one of the most important features distinguishing them from Gram-positive organisms, has presented a significant challenge to antibacterial drug discoverers due to its remarkable ability to restrict access of small molecules to the periplasmic space (4, 5). In response, novel and innovative approaches to circumvent this impermeability are currently being explored and developed (6, 7); however, their ultimate potential clinical utility remains unknown. As an alternative strategy, many groups have elected to exploit outer membrane biogenesis pathways to find new antibiotic targets. Among the various components that are responsible for outer membrane assembly, the synthesis of lipid A molecules is among the most critical, since these moieties serve as the anchor on the outer membrane for lipopolysaccharide (LPS) attachment. For most Gram-negative organisms, the inability to decorate the outer membrane with LPS has a bactericidal effect, and thus the interference of lipid A biosynthesis by a small-molecule inhibitor would prevent LPS assembly and result in the death of the target bacterial cell. The UDP-3-O-(R-3-hydroxymyristoyl)-N-acetylglucosamine deacetylase, encoded by lpxC, represents the first committed step in the lipid A biosynthetic pathway and has emerged as a novel antibiotic target for the treatment of Gram-negative infections (8–10).

Here we report the advancement of novel classes of LpxC inhibitors and provide an in-depth characterization of a lead compound, LpxC-4, in terms of its microbiological spectrum, resistance potential, and in vivo efficacy. Through the course of our investigation, using spontaneously resistant isolates generated during these profiling efforts, we identified several unexpected physiological responses that differed among the various Gram-negative pathogens we are targeting. In addition, we show that LpxC-4 still retains efficacy against mutants expressing these different first-step resistance mechanisms, demonstrating the potential clinical utility of this inhibitor class.

RESULTS

LpxC inhibitors are potent and rapidly bactericidal against multiple Gram-negative species.

Our efforts to identify a potent, broad-spectrum inhibitor of LpxC have focused on a Zn2+ binding class of hydroxamic acids. The structures of the lead molecules from two different series of compounds are shown in Fig. 1. LpxC-2, one of our leads from the biphenyl methylsulfone-containing series, has been described previously (11), as have the pyridone-substituted compounds LpxC-3 and LpxC-4 (12). While the 50% inhibitory concentrations (IC50s) for each of these compounds against the Pseudomonas aeruginosa LpxC enzyme are not substantially different, the pyridone analog LpxC-4 demonstrates a clear MIC90 advantage over the biphenyl analog LpxC-2 when tested against a panel of 106 recent clinical isolates (Table 1). By comparison, despite it having enzyme inhibitory activity roughly equivalent to that of LpxC-4, the MIC90 of CHIR-090, a highly potent LpxC inhibitor that is active against a wide variety of MDR Gram-negative bacteria (13, 14), was found to be 4-fold higher against P. aeruginosa. In the case of Klebsiella pneumoniae, an ~10-fold decrease in the enzyme IC50 between LpxC-3 and LpxC-4 resulted in a 16-fold decrease in MIC90 to 1 µg/ml (Table 1). In Acinetobacter baumannii, however, all LpxC inhibitors tested had considerably higher enzyme IC50s and MIC90 values that were 32 µg/ml or greater. It should be noted, however, that a structurally related LpxC inhibitor from the pyridone class was recently demonstrated to have significant in vivo efficacy against an MDR clinical isolate of A. baumannii despite having no whole-cell activity in vitro (15), a disconnect which is likely due to the unusual ability of A. baumannii to survive in vitro without LPS (16). Many of our compounds, most notably LpxC-4, also have strong activity against other important Gram-negative species, including Escherichia coli, Enterobacter spp., Burkholderia cepacia, and Stenotrophomonas maltophilia (Table 1). Since it proved to be a potent compound with broad-spectrum activity, we chose to further characterize LpxC-4 both in vitro and in vivo.

FIG 1 

Structures of LpxC inhibitors used in these studies.

TABLE 1 

In vitro microbiological assessment of novel LpxC inhibitors shows superior activities against a broad spectrum of Gram-negative pathogens compared to CHIR-090 and meropenem

Bacterium (n[a])	Activity of:
	LpxC-2[b]		LpxC-3[c]		LpxC-4		CHIR-090		MPM[d]
	IC50 (nM)	MIC90 (μg/ml)	IC50 (nM)	MIC90 (μg/ml)	IC50 (nM)	MIC90 (μg/ml)	IC50 (nM)	MIC90 (μg/ml)	MIC90 (μg/ml)
P. aeruginosa (138)	1.4	4	3.6	2	1.1	1	<2.1	4	>64
P. aeruginosa PAO1 WT[e]	1.4	0.5	3.6	0.5	1.1	0.5	<2.1	1	0.25
P. aeruginosa PAO1 M62R	7.9	4	2.8	0.5	2.1	0.5	NT	NT	NT
K. pneumoniae (98)	1.0	16	0.68	16	0.069	1	NT	NT	32
E. coli (79)	NT[f]	2	NT	8	NT	0.25	NT	0.25	NT
Enterobacter spp. (52)[g]	NT	4	NT	16	NT	0.5	NT	0.5	0.25
A. baumannii (31)	>41	32	110	32	183	>64	NT	>64	32
B. cepacia (30)	NT	NT	NT	NT	NT	0.5	NT	>64	8
S. maltophilia (30)	NT	NT	NT	NT	NT	2	NT	>64	>64

n, no. of isolates.

MIC data from reference 11.

MIC data from reference 12.

MPM, meropenem.

WT, wild type.

NT, not tested.

Includes 21 E. aerogenes and 31 E. cloacae isolates.

We first conducted standard static time-kill (STK) studies using LpxC-4 concentrations which targeted multiples above and below the MIC against representative strains of P. aeruginosa and K. pneumoniae. Figure 2A shows the rapid bactericidal activity of LpxC-4 against P. aeruginosa UC12120, with regrowth limited to drug concentrations of ≤2× the MIC (0.25 µg/ml). The kill rate is not as dramatic against the cystic fibrosis (CF) clinical isolate PA-1955, which constitutively expresses the mexEF-oprN efflux pump (Fig. 2B). While regrowth is still restricted to drug concentrations of ≤2× the MIC (4 µg/ml) and stasis is seen at 4× the MIC, the kinetics of bacterial killing is considerably slower than that observed with UC12120, which suggests that efflux upregulation may play a role in this phenotype, as has been described previously (11). Testing of a recent, KPC-producing clinical isolate of K. pneumoniae, however, demonstrated the same pattern of rapid bactericidal activity as was seen with UC12120, with no regrowth seen at concentrations of 4× the MIC or higher (Fig. 2C).

FIG 2 

Static-time-kill assays demonstrate sustained bactericidal activities of LpxC-4 against P. aeruginosa UC12120 (A), PA-1955 (B), and K. pneumoniae KP-1487 (C).

LpxC-4 resistance in P. aeruginosa, although infrequent, arises from efflux pump and/or target upregulation.

In order to assess the resistance emergence potential for our lead LpxC inhibitors, we used standard frequency-of-resistance (FOR) methods with representative strains of P. aeruginosa and K. pneumoniae. P. aeruginosa resistance frequencies, which were calculated based on LpxC-4 concentrations above each strain’s respective MIC and are shown in Table 2, demonstrate that each of the strains tested has an extremely low probability of developing spontaneous resistance to LpxC-4. In addition, several resistant isolates were recovered, and their MICs relative to that for the parent strain were determined. For P. aeruginosa PAO1 and UC12120, a 2- to 4-fold increase in the LpxC-4 MIC relative to that of each respective parent strain was observed (Table 2). Spontaneously resistant P. aeruginosa PA-1955 isolates, however, showed a much more pronounced MIC shift relative to findings for the parent strain (8- to 32-fold). Cross-resistance MIC profiling of the P. aeruginosa LpxC-4R isolates demonstrated that quinolone resistance was increased as well (data not shown), which prompted our evaluation of LpxC-4 MICs against these resistant isolates in both the presence and absence of the efflux pump inhibitor PAβN. Addition of 50 µg/ml PAβN fully restored the LpxC-4 MIC to parental levels in all PAO1 and UC12120 resistant isolates tested while only partially reducing the MICs of PA-1955 isolates (Table 2).

TABLE 2 

LpxC-4 resistance frequencies are similar in P. aeruginosa and K. pneumoniae, but resulting first-step mechanisms of resistance are functionally distinct

Strain	FOR (fold MIC)	Resistant isolate no. (genetic alteration)	MIC (μg/ml)		Western blot analysis result			Genotype
			LpxC-4[e]	LpxC-4 + PAβN	OprM	OprN	LpxC
P. aeruginosa
PAO1[a] (OprM+)	3.4 × 10−8 (2×)	1	4	0.5	+++	−	+	MexR frameshift at bp 106
	<5.0 × 10−10 (4×)	2	4	0.5	+++	−	+	MexR premature stop codon at Q55
		3	2	0.5	+++	−	+	MexR R82P
		4	2	0.5	+++	−	+	MexR L131Q
UC12120[b] (OprN-)	2.5 × 10−8 (2×)	1	4	1	+	+	+	MexT G258D
	4.2 × 10−9 (4×)	2	4	1	+	+	+	MexS L46F
		3	4	1	+	+	+	MexT G258D
		4	4	1	+	+	+	MexS R322W
PA-1955[c] (OprN+)	1.5 × 10−8 (4×)	1	16	2	+	+	+	ND
	2.5 × 10−10 (8×)	1 (ΔoprN)	2	ND[f]	+	−	+	ND
		2	16	1	+	+	+	ND
		2 (ΔoprN)	2	ND	+	−	+	ND
		3	16	1	+	+	+	ND
		3 (ΔoprN)	2	ND	+	−	+	ND
		4	16	2	+	+	+	ND
		4 (ΔoprN)	2	ND	+	−	+	ND
		5	64	8	+	+	+++	C-to-A mutation 11 bp upstream of lpxC
		5 (ΔoprN)	16	ND	+	−	+++	ND
		6	64	8	+	+	+++	C-to-A mutation 11 bp upstream of lpxC
		6 (ΔoprN)	16	ND	+	−	+++	ND
K. pneumoniae								Genotype
KP-1487[d]	9.6 × 10−8 (8×)	1	8	4			+	FabZ A78V
	2.2 × 10−9 (16×)	2	16	8			+	FabZ R121L
		3	8	8			+	FabZ P22L
		4	8	4			+	FabZ A78V
		5	8	8			+	FabZ P22S
		6	8	8			+	FabZ P22L

PAO1 MIC = 0.5 μg/ml.

UC12120 MIC = 1 μg/ml.

PA-1955 MIC = 2 μg/ml.

KP-1487 MIC = 1 μg/ml.

For KP-1487, a skipped-well phenomenon was seen at 0.25, 0.5, and 1 μg/ml, MIC reported per CLSI guidelines.

ND, not determined.

Our preliminary MIC results led us to probe total cell extracts of both wild-type and LpxC-4R mutant strains using Western blot analysis with OprM-, OprN-, and OprJ-specific monoclonal antibodies. Interestingly, while PAO1 mutants showed increased production of OprM relative to results for the parental strain, all resistant UC12120 isolates produced OprN, whereas the parent strain did not (Table 2). Additionally, we noted that the PAO1 resistance frequency was approximately 10-fold lower than that of UC12120. One likely explanation for the differential frequency and mechanistic response seen between these two strains is the existence of a P195T point mutation in the PAO1 mexT gene, which encodes the positive regulator of the mexEF-oprN operon. While this particular mutation has not been described previously, an 8-bp insertion within this coding region which results in a frame shift that prevents the activation of this efflux system has been described (17). We have speculated that this point mutation similarly prevents the activation of mexEF-oprN, as evidenced by the lack of FOR-generated PAO1 mutants that express this efflux pump. In the context of LpxC-4 resistance, this would therefore create a requirement for an alternative efflux system to be activated or upregulated. No changes in efflux expression were detected in the LpxC-4R PA-1955 isolates relative to that of the parent strain, which constitutively expresses mexEF-oprN by virtue of a G78S point mutation encoded in mexS, which has previously been described to be involved with activation of this efflux pump (18). OprJ expression was not detected in any P. aeruginosa isolate tested.

Additionally, we noted that no mutants with mutations within the lpxC coding region were recovered. Previous resistance frequency experiments performed with LpxC-2 yielded mutants that harbored an M62R active site point mutation, resulting in an 8-fold increase in the LpxC-2 MIC (Table 1). The MIC shift associated with this particular mutation was explained by the ~5-fold increase in the IC50 for the M62R mutant enzyme versus that for the wild-type version, and backcrossing of this mutation into a wild-type, susceptible P. aeruginosa strain recapitulated the level of LpxC-2 resistance seen in the spontaneously generated mutant (data not shown). Encouragingly, the LpxC-4 IC50 of the M62R mutant enzyme was not significantly elevated, and correspondingly, no MIC increase against LpxC-4 was seen with this mutant strain (Table 1), which would explain our inability to recover such a mutant in FOR studies. The mutation replaces a relatively nonpolar residue (methionine) with a polar residue (arginine), and based on the known binding mode of the pyridone-containing inhibitors for wild-type LpxC (12), the amino acid side chain of both wild-type and M62R mutant enzymes is likely located near the pyridone carbonyl of LpxC-3 and LpxC-4. This may enable a productive H-bonding interaction of these inhibitors with the mutant arginine, thereby maintaining enzyme inhibitory activity. Since LpxC-2 contains a nonpolar phenyl group at this position, a similar productive interaction is not feasible, which may explain the minor loss of activity against the mutant enzyme. Taken together, these FOR results suggest that activation of mexEF-oprN is the preferred first-step mechanism employed by P. aeruginosa to resist the activity of LpxC-4, although utilization of MexAB-OprM can compensate if the former cannot be activated, albeit at a lower frequency.

To further confirm these preliminary mechanistic findings, we conducted whole-genome sequencing of four PAO1 LpxC-4R isolates. Given the Western blot results described above, it was not surprising to find that all of these mutants had mutations in mexR (Table 2), the repressor of the mexAB-oprM operon (19). Sequencing of mexST from four resistant UC12120 isolates revealed that two harbored the exact same point mutation in mexT, while the other two each had unique point mutations in mexS (Table 2). We then generated oprN deletion mutations in multiple resistant UC12120 isolates and found that the LpxC-4 MICs of these isogenic mutants had returned to wild-type UC12120 levels. We also made oprN deletion mutants in six PA-1955 LpxC-4R isolates, but unlike the case with UC12120 mutants, the MICs from only four of the six fully reverted to wild-type levels (Table 2). To help understand the additional mechanism(s) of LpxC-4 resistance potentially being employed in the other two PA-1955 mutants, we sequenced the lpxC gene and also performed Western blot analyses using a P. aeruginosa LpxC-specific monoclonal antibody. The results of these assays, shown in Table 2, demonstrated that while neither of these isolates had any detectable mutations within the lpxC coding region, they both demonstrated increased levels of LpxC production relative to those of both the parental strain and the other four LpxC-resistant mutants. Since increased LpxC production has been described previously as a P. aeruginosa resistance mechanism against other LpxC inhibitors (10), we sequenced the upstream noncoding region of the lpxC gene and identified the same C-to-A mutation 11 bp upstream of the translation initiation site in both of these isolates (Table 2).

A P. aeruginosa-specific sRNA may contribute to LpxC production and resistance to LpxC inhibitors.

The repeated isolation of spontaneous LpxC inhibitor-resistant mutants harboring this upstream, noncoding region mutation, both by our group and by others (10), prompted us to further probe this mechanism of LpxC-4 resistance. We started by backcrossing this mutated allele into PAO397, an engineered PAO1 strain that lacks 5 different RND efflux pumps, so that the contributions of efflux to LpxC-4 resistance would not factor into our analyses. As expected, backcrossing of the C-to-A mutation 11 bp upstream of the lpxC translational start resulted in a 16-fold increase in the LpxC-4 MIC relative to that of the parent strain (data not shown). Examination of the DNA sequence upstream of the ribosome binding site (RBS) suggested that a structural element may exist within this region. Computational analysis of this sequence using the mfold server (20) predicted the existence of a small RNA (sRNA) that spans nearly the entire intergenic region separating lpxC from ftsZ, the gene immediately upstream. Interestingly, this sRNA was recently identified computationally, and its existence was confirmed experimentally by reverse transcription-PCR (RT-PCR) (21, 22), which has led to the naming of this region as PA4406.1 in the PAO1 genome annotation on the Pseudomonas Genome Database (23).

Figure 3A shows the predicted structure of the sRNA encoded by wild-type PAO1. In this model, the cytosine 11 bp upstream of the lpxC translational start (filled arrow) is part of a hairpin structure that pairs with the guanine 18 bp upstream of lpxC (open arrow). Changes to this structure when the C-to-A mutation is introduced are shown in Fig. 3B. Given the relatively simplistic nature of this hairpin, we were skeptical about its contribution to regulation of LpxC translation. Unfortunately, our attempts to further characterize this sRNA by restoring the hairpin structure in the mutant through the introduction of a G-to-T mutation 18 bp upstream of lpxC were unsuccessful. Likewise, the cloning of this sRNA into E. coli also did not prove successful, at least not without identifying additional mutations elsewhere within the structure. We were able to demonstrate one key difference between the wild-type and C-to-A mutant strains, however. Although this mutation has historically been regarded as one that impacts translation of LpxC both by us and by others (10), we speculated that the position of this mutation within the sRNA coding region could actually be affecting the transcription or mRNA stability of lpxC. While quantitative RT-PCR (qRT-PCR) experiments using RNA from wild-type PA-1955, PAO397, and their respective C-to-A mutants did not demonstrate any significant differences in mRNA turnover (data not shown), we were encouraged to see a 3-fold increase in lpxC expression in both mutant strains relative to that in their respective parent strains (Fig. 4). While future work is needed to identify any additional components of this regulatory circuit, to our knowledge this is the first report describing a molecular mechanism controlling LpxC production in P. aeruginosa.

FIG 3 

The sRNA predicted to be encoded in the upstream region of lpxC in P. aeruginosa adopts an altered conformation upon mutation, conferring LpxC-4 resistance. Sequences from the wild-type (A) or C-to-A-mutated (B) variants were analyzed using mfold. The closed arrow indicates the −11 position, while the open arrow indicates the −18 position.

FIG 4 

Mutation of the sRNA upstream of lpxC, which confers high-level resistance to LpxC-4, results in an increased level of lpxC transcription as assessed by qRT-PCR.

Curiously, we have yet to identify an LpxC-overproducing mutant from any P. aeruginosa isolate that does not already have an efflux pump upregulated, whereas in strains like PA-1955, which constitutively expresses MexEF-OprN, this genotype was found among the few stably resistant spontaneous mutants recovered in FOR experiments. Further supporting this preliminary finding, we were also able to recover mutants with increased levels of LpxC production in a P. aeruginosa clinical isolate that overexpresses mexAB-oprM (data not shown). As with PA-1955 LpxC-overexpressing mutants, these mutants also contained the C-to-A mutation 11 bp upstream of the lpxC translational start. This interesting finding prompted us to perform larger-scale studies by examining 96 FOR-generated mutants from strains expressing efflux pumps at a basal level (such as PAO1 and UC12120). The resulting isolates have demonstrated that efflux upregulation is their primary mechanism of resistance to LpxC inhibitors by virtue of the full restoration of the MIC to wild-type levels when PAβN is included in the assay (data not shown). This suggests that strains that do not have upregulated efflux pumps would require two independent mutational events to become resistant to our LpxC inhibitors to levels that are not predicted to be clinically achievable. To evaluate the resistance propensity from hypermutable P. aeruginosa strains, we constructed ΔmutS deletion mutants of both PAO1 and UC12120, which resulted in the expected phenotype as assessed by a sharp increase in the rate of spontaneous rifampin resistance. A corresponding increase in the frequency of LpxC-4 resistance was not detected in either of these hypermutable strains, however, and the mechanisms behind the resistance in recovered isolates were no different from those found in wild-type PAO1 and UC12120 (data not shown).

Mutation of fabZ is the major first-step LpxC-4 resistance mechanism in the Enterobacteriaceae.

The LpxC-4 FOR for KP-1487, a KPC-expressing, MDR K. pneumoniae clinical isolate, is shown in Table 2. Resistant isolates from this strain, which were recovered at frequencies similar to those with P. aeruginosa, showed an 8-fold increase in the MIC relative to that of the parent strain. In contrast to P. aeruginosa, however, we noticed a highly reproducible “skipped-well phenomenon” (e.g., bacterial growth at drug concentrations below the parental strain MIC, followed by a gap in growth surrounding concentrations equivalent to the parental strain MIC, followed by growth at higher concentrations) when many of these resistant isolates were profiled by MIC testing. This phenotype has been reported previously (24), although in that particular report it was described in the context of colistin and polymyxin B MIC testing of Acinetobacter clinical isolates, as opposed to isogenic resistant mutants whose parent strain does not display this pattern of growth. The skipped-well MIC phenomenon has also been described for P. aeruginosa (25), where it was demonstrated that isogenic polymyxin B-resistant mutants can adapt to particular concentrations of this antibacterial agent. To the best of our knowledge, however, this phenomenon has not been described in the context of LpxC inhibitors. While we have reported the MICs of these FOR-generated mutants as 8 µg/ml (Table 2), we also noted that the concentration required to inhibit the growth of wild-type KP-1487 (1 µg/ml) was within the “skipped-well” range when its isogenic mutants were tested, meaning that this concentration was sufficient to kill both wild-type and mutant strains.

The MICs of these resistant isolates were not reduced to parental levels when PAβN was included, nor were these isolates cross-resistant to quinolone antibiotics (data not shown). Both of these phenotypes suggested that efflux pump upregulation/activation was not mediating LpxC-4 resistance in K. pneumoniae. Each of these mutants did display cross-resistance to LpxC-2 and LpxC-3, however, which prompted us to explore other potential resistance mechanisms related to the target. First, we sequenced the entire lpxC gene from 6 LpxC-4R KP-1487 isolates, which did not reveal any mutations in this region relative to sequence of the parent strain. The lack of lpxC upregulation was confirmed by preparing whole-cell extracts of these strains and performing Western blot analyses using a K. pneumoniae LpxC-specific polyclonal antibody (Table 2). Given the previous discovery of fabZ mutations correlating with resistance to LpxC inhibitors (10, 26), we next decided to sequence this gene from our resistant mutants. Sequencing of the 6 mutants described above revealed that each harbored a single point mutation in fabZ (Table 2), at least one of which has been described previously (26). We subsequently performed whole-genome sequencing and confirmed that these point mutations were the only ones evident in these resistant strains (data not shown).

In vivo efficacy against wild-type P. aeruginosa and K. pneumoniae strains is translatable from in vitro MIC values when evaluated in both acute septicemia and neutropenic thigh infection models.

Fifty percent effective dose (ED50) studies were conducted with LpxC-4 against P. aeruginosa UC12120, PA-1950, and PA-1955 and K. pneumoniae KP-1487 in an acute septicemia model. Efficacy, as assessed in this acute infection model, was generally consistent with MIC results, with ED50s ranging from 7.4 mg/kg of body weight for PA-1950 (MIC = 0.25 μg/ml) to 55.9 mg/kg for PA-1955 (MIC = 2 μg/ml). We also established a strong correlation between the expression of P. aeruginosa LpxC-4 resistance determinants and in vivo efficacy by demonstrating that PA-1955 ΔoprN (MIC = 1 µg/ml) had an ED50 that was reduced to 19.8 mg/kg and that the PA-1955 sRNA mutant (MIC = 64 µg/ml) had an ED50 that was >200 mg/kg. To further define the exposure-effect relationship of these data, the ED50 data were normalized to an EC50 accounting for free drug concentrations in these studies. Pharmacokinetics of LpxC-4 following single subcutaneous doses of 18.75, 75, and 300 mg/kg revealed that exposure was generally linear across the dose range, with area under the concentration-time curve (AUC) and maximum concentration of drug in serum (Cmax) increasing with a proportional increase in dose (Table 3). Cmax values ranged from 5.02 to 75.4 µg/ml for doses from 18.75 to 300 mg/kg, respectively. Accounting for a free fraction of 0.31 (12), unbound LpxC-4 AUCs ranged from 1.58 to 23.7 µg·h/ml. In vivo pharmacokinetic/pharmacodynamic (PK/PD) driver and exposure magnitude determinations were characterized in a neutropenic thigh infection model with K. pneumoniae KP-1487 and P. aeruginosa UC12120. Dose fractionation of LpxC-4 suggested that the free AUC/MIC ratio (free AUC/MIC) was the variable most dynamically linked to efficacy against UC12120, with r2 values of 0.79, 0.75, and 0.53 for free AUC/MIC, free Cmax/MIC, and the percentage of the dosage interval in which the level of drug in serum exceeds the MIC (% time > MIC), respectively. Although data were more variable for activity against KP-1487, free AUC/MIC still showed the strongest correlation to activity, with r2 values of 0.58, 0.51, and 0.21 for free AUC/MIC, free Cmax/MIC, and % time > MIC, respectively. Endpoint studies of activity against PA-1950 and KP-1487 in the neutropenic thigh model suggested free AUC/MICs of 6.48 and 5.36, respectively, to reach the stasis (no net change in bacterial burden over 24 h) target. Free AUC/MICs of 7.25 and 13.5 were associated with a 1-log kill versus PA-1950 and KP-1487, respectively (Table 4). While EC50s and targets for stasis and 1-log kill were broadly consistent across neutropenic thigh studies with P. aeruginosa and K. pneumoniae, the KP-1487 P22S FabZ mutant, which showed an 8-fold increase in the MIC versus that for wild-type KP-1487, demonstrated much lower AUC/MICs for PK/PD endpoints of stasis and 1-log kill, suggesting that these mutants are more susceptible in vivo than their in vitro MICs would suggest.

TABLE 3 

Pharmacokinetics of LpxC-4 in CD-1 mice[]

Parameter[b]	Value for dose (mg/kg) of:
	18.75	75	300
Cmax (mg/liter)	5.02 ± 0.61	15.50 ± 4.26	75.40 ± 5.65
Tmax (h)	0.25 ± 0.00	0.33 ± 0.13	0.33 ± 0.13
AUC (mg ⋅ h/liter)	5.09 ± 1.04	17.60 ± 2.49	76.30 ± 0.96
Free AUC (mg ⋅ h/liter)	1.58 ± 0.32	5.46 ± 0.77	23.70 ± 0.30
Half-life (h)	0.60 ± 0.03	0.69 ± 0.03	0.68 ± 0.03
CL (liters/h/kg)	3.79 ± 0.78	4.32 ± 0.67	3.92 ± 0.04
Vss (liters/kg)	2.20 ± 0.19	3.30 ± 0.55	2.53 ± 0.39

Values are means ± SD.

CL, clearance; Vss, volume of distribution at steady state.

TABLE 4 

In vivo efficacies (24-h AUC/MIC) of LpxC-4 in CD-1 mice against sentinel strains of P. aeruginosa and K. pneumoniae

Strain	MIC (µg/ml)	Immunocompetent septicemia EC50[a]	Neutropenic lung EC50	Neutropenic thigh stasis[a]	Neutropenic thigh 1-log kill[a]
P. aeruginosa
UC12120	1	1.23 ± 2.34	NT[b]	NT	NT
PA-1950	0.25	6.31 ± 2.51	<8.40	6.48 ± 2.68	7.25 ± 20.32
PA-1955	2	5.00 ± 5.36	NT	NT	NT
PA-1955 ΔoprN	1	1.02 ± 0.68	NT	NT	NT
PA-1955 sRNA mutant	64	NC[c]	NT	NT	NT
K. pneumoniae
KP-1487[d]	1	NT	NT	5.36 ± 0.49	13.50 ± 2.20
KP-1487[e]	1	NT	NT	14.50 ± 6.20	NC
KP-1487 FabZ P22S[d]	8	NT	NT	1.83 ± 2.69	3.45 ± 2.11

Mean AUC/MIC ± SE.

NT, not tested.

NC, not calculable (ED50 > 200 mg/kg).

Inoculum = 1 × 106 CFU.

Inoculum = 1 × 107 CFU.

Pulmonary infection model of activity against P. aeruginosa PA-1950.

Efficacy of LpxC-4 in an acute (48-h) mouse pneumonia model was achieved against PA-1950 with an ED50 of <25 mg/kg. This result was consistent with an estimated LpxC-4 ED50 of 16.8 mg/kg against PA-1950 in the neutropenic thigh model. Normalizing the dose to free drug exposure and considering an MIC of 0.25 µg/ml, the ED50s in the thigh and lung endpoint studies correspond to free AUC/MICs of 7.2 and 10.7, respectively, suggesting similar exposure requirements for both models.

DISCUSSION

Here we have reported the in vitro and in vivo characterization of a recently disclosed inhibitor of the Gram-negative LpxC enzyme, which has demonstrated the potential of this drug class to provide a new option in the clinical setting. During the course of our experimentation, we were fascinated to find that our lead compounds also served as reagents to better understand the basic physiology of the organisms that they target. Our assessment of LpxC-4 resistance in multiple Gram-negative pathogens has demonstrated how differently many of the basic cellular processes are carried out in these organisms. For example, the fact that efflux is the preferred first-step resistance mechanism in P. aeruginosa but not K. pneumoniae is reflective of the functional distinction between the various RND efflux pumps produced by these bacterial species. Despite the fact that members of the Enterobacteriaceae produce AcrAB-TolC, an efflux pump whose substrate specificity includes several different classes of antibiotics (27) and therefore contributes to multidrug resistance, we have never established that LpxC-4 resistance can be achieved by overexpression of this system. This conclusion was based both on our inability to recover any spontaneously resistant mutants with increased production of AcrAB-TolC and the fact that clinical isolates known to have upregulated efflux expression (included in the MIC90 panels described in Table 1) did not demonstrate an increase in the LpxC-4 MIC. As a result, organisms such as K. pneumoniae require mutations in fabZ to resist the action of LpxC-4, much as E. coli does to resist other LpxC inhibitors described previously (26).

Since FabZ acts on R-3-hydroxymyristoyl acyl carrier protein, a substrate common to both lipid A and fatty acid biosynthesis, we and others (26) have hypothesized that these various fabZ point mutations result in a redistribution of the availability of this precursor, shunting more of it toward lipid A biosynthesis. More recently, however, it was demonstrated that CHIR-090-resistant E. coli mutants, each of which harbored a single point mutation within fabZ, showed a marked decrease in lpxC expression relative to that of their parent strain (28). These results support the contention that E. coli carefully regulates the coordinate expression of both lipid A and fatty acid biosynthetic pathways, since homeostasis between these two cellular components is vital to the survival of the organism. Our discovery that fabZ-mutated K. pneumoniae isolates display a skipped-well phenotype in MIC profiling studies reinforces this point, since reduced LpxC levels as a consequence of fabZ mutation result in fewer target molecules for LpxC-4 to inhibit. Intriguingly, our initial in vivo efficacy data with KP-1487 suggest that the apparent 8-fold MIC shift demonstrated by fabZ mutants in vitro does not translate to reduced susceptibility in an in vivo infection model. FOR and subsequent MIC profiling using similar LpxC compounds in our laboratories against E. coli revealed fabZ mutation as the primary resistance mechanism employed by this pathogen. Similarly to K. pneumoniae, these mutants also displayed the skipped-well pattern of growth. Whether these E. coli FabZ mutant strains retain in vivo susceptibility, in spite of an increase in MIC, is unknown at this time but warrants further investigation.

Another key difference between P. aeruginosa and members of the Enterobacteriaceae is the effect of LpxC overexpression and the regulatory mechanism employed by the former to control the levels of this enzyme. It has long been established that overproduction of LpxC has deleterious effects on bacterial cell survival, a consequence that is carefully avoided in the Enterobacteriaceae by the expression and activity of FtsH, an AAA protease that recognizes a specific C-terminal sequence in LpxC (29). The C terminus of P. aeruginosa LpxC does not include this recognition sequence, however, and this protein was recently shown not to be subjected to FtsH-mediated cleavage (30), a finding that has left the mechanism(s) of LpxC regulation a mystery in this organism. The recent identification of an sRNA upstream of lpxC in P. aeruginosa (21, 22), together with the repeated identification of spontaneous LpxC inhibitor-resistant strains harboring the exact same mutation within this RNA coding region, has provided some preliminary clues to better understand this novel mechanism of LpxC control. Since sRNA-mediated regulation can occur by direct binding via sequence complementarity, we attempted studies to demonstrate that C-to-A mutant sRNAs showed differential binding capabilities relative to wild-type sRNAs. Likewise, since sRNAs can regulate their targets through the recruitment of proteins, we also tested wild-type and mutant sRNAs for their ability to differentially bind to P. aeruginosa cellular extracts. In both cases we were unable to conclusively identify further mechanisms by which this sRNA controls lpxC expression, although optimization of experimental conditions could provide different results in the future. However, we were able to show that this upstream mutation leads to an increase in transcription of lpxC, rather than just its translation as has been proposed previously. The existence of this sRNA-mediated regulation in P. aeruginosa distinguishes it from members of the Enterobacteriaceae, and this difference is further supported by the lack of LpxC-overexpressed mutants recovered from FOR experiments with K. pneumoniae. The latter point may be explained by the significant sequence divergence upstream of lpxC between these pathogens. Whether LpxC overproduction in P. aeruginosa is better tolerated than it is in the Enterobacteriaceae remains to be understood and warrants further investigation.

In relevant in vivo models of infection, LpxC-4 was shown to be efficacious against sentinel strains of P. aeruginosa and K. pneumoniae. Activity in these models was commensurate with MIC values of the wild-type strains evaluated. PK/PD driver analysis suggested free AUC/MIC to be the parameter most closely linked to efficacy, and magnitude requirements for stasis and 1-log kill were determined. Magnitude requirements for a 1-log kill were nearly equivalent for PA-1950 and KP-1487, suggesting similar exposure targets to treat strains of Pseudomonas and Klebsiella. Certainly a greater number of strains will need to be studied to confirm exposure requirements, but based upon preliminary allometric scaling of preclinical PK (data not shown), a clinical dose of approximately 1,200 mg every 8 h (q8h) would be anticipated to sufficiently treat strains with MICs of 1 µg/ml or less. Additionally, if PA-1955 susceptibility (MIC = 2 µg/ml) is representative of first-step efflux mutants, it is likely that compounds like LpxC-4 could be used clinically against these P. aeruginosa strains as well. Unfortunately, our data do not suggest that LpxC-4 can effectively treat second-step P. aeruginosa sRNA mutants with increased LpxC expression, although the frequency of these mutations is quite low and thus far they seem to arise only in strains with preexisting upregulated efflux expression. Therefore, we anticipate that the incidence at which these double mutants would spontaneously arise would be quite rare, although careful monitoring for the emergence of second-step-resistant mutants would be particularly critical for strains with efflux upregulation. Our data also support the conclusion that effective treatment against K. pneumoniae strains and likely other members of the Enterobacteriaceae, such as E. coli, that resist LpxC-4 through fabZ point mutation is achievable in vivo, despite the significant MIC shifts seen in vitro. This serves as yet another example in which the constraints of standardized in vitro susceptibility testing can hinder our ability to truly appreciate the potential for our candidate compounds by not accounting for the substantial differences in cell physiology in the harsh, nutrient-restricted in vivo environment versus that under nutrient-replete conditions used for classical MIC testing.

The need for new antibiotics, particularly those with novel mechanisms of action to minimize the prevalence of existing resistance, has prompted us to identify and evaluate a novel series of LpxC inhibitors for the treatment of serious Gram-negative infections. The results described in this report demonstrate that LpxC-4 is a potent, broad-spectrum inhibitor against which spontaneous resistance arises fairly infrequently. In vivo efficacy studies confirm our in vitro activity findings for multiple Gram-negative pathogens, and PK assessments suggest that the increases in the MIC associated with the major first-step resistance mechanism in P. aeruginosa may still be covered by realistic LpxC-4 dosing regimens. In addition, our identification and characterization of the skipped-well phenomenon displayed by fabZ-mutated K. pneumoniae first-step mutants also support the notion that major resistance mechanisms employed by Enterobacteriaceae may also be successfully treated by LpxC-4 in vivo. This unusual susceptibility phenotype, along with the preliminary work to describe the novel sRNA-mediated mechanism of lpxC regulation by P. aeruginosa, were discoveries that stemmed from the use of our LpxC inhibitors. The utilization and application of antibacterial agents as tools to further our understanding of the fundamental physiological processes used by target bacterial species is an important advancement that has the potential to provide valuable insight into novel antibiotic targets and pathways in the future.

MATERIALS AND METHODS

Bacterial strains and media used.

Bacterial strains used in this study were from the Pfizer internal collection, consisting primarily of strains recovered from patients worldwide since 2000. P. aeruginosa PAO1 (provided by Mike Vasil, University of Colorado, Denver), UC12120, and PA-1950 are all antibiotic-susceptible strains, while strain PA-1955 is a MexEF-OprN-producing isolate recovered from a cystic fibrosis patient in 2000. P. aeruginosa PAO397 is a derivative of PAO1 that has deletion mutations in 5 major RND-type efflux pumps (31). K. pneumoniae strain KP-1487, an MDR clinical isolate that expresses KPC-2, SHV-12, and TEM-1, was obtained from New York-Presbyterian Hospital (Columbia University Medical Center). All strains were routinely grown and maintained on lysogeny broth (LB) and agar. All MIC, static time-kill (STK), and frequency-of-resistance (FOR) experiments were conducted in cation-adjusted Mueller-Hinton broth (caMHB), which was purchased from Difco. Where appropriate, medium was supplemented with 15 µg/ml (for E. coli) or 75 µg/ml (for P. aeruginosa) gentamicin.

LpxC inhibitor IC50, MIC, and STK testing.

Enzyme inhibition of each LpxC compound (shown in Fig. 1) was conducted using a functional LpxC enzyme assay, described previously (12). LpxC-4, which can be purchased commercially (catalog number PZ0194; Sigma-Aldrich), is also known as PF-5081090. MIC assays were conducted according to the guidelines established by the Clinical and Laboratory Standards Institute (CLSI) (32, 33). Where indicated, the efflux pump inhibitor phenyl-arginyl-β-naphthylamide (PAβN) (Sigma) was added at a final concentration of 50 µg/ml. STK assays were also conducted in accordance with CLSI guidelines. Briefly, a 0.5 McFarland suspension from an overnight LB agar plate was diluted to 5 × 105 in caMHB and incubated at 35°C with ambient atmosphere at 220 rpm for 90 min. STK assessments were carried out in 10 ml of preincubated culture with a range of concentrations of LpxC inhibitors. Samples (200 µl/sample) were collected at 0, 2, 4, 6, 8, 10, 24, and 48 h, serially diluted in phosphate-buffered saline, and spotted on blood agar plates before overnight incubation at 37°C for CFU/ml determinations.

FOR determination and resistant isolate characterization.

Frequencies of resistance (FORs) were calculated using standard population analysis as described previously (6). Specifically, bacterial strains were grown in caMHB to mid-log phase (optical density at 600 nm [OD600] = 0.5) prior to being concentrated 40× in fresh medium. One hundred microliters of this suspension (~1 × 109 CFU) was plated on agar plates containing increasing concentrations of LpxC inhibitors, ranging from 0.5 to 64 µg/ml (in 2-fold increments). Plates were incubated at 37°C for 40 h prior to colony counting. Resistance frequencies were calculated by dividing the number of colonies by the density of cells plated from the initial inoculum. Resistance stability of the recovered colonies was assessed by streaking on drug-free medium, freezing at −80°C, restreaking on drug-free medium, and using those grown colonies to conduct standard MIC testing. Isolates that demonstrated elevated MICs above those for the parent strain were further characterized to determine the mechanism of resistance.

FOR-generated isolates were initially characterized by conducting MICs in the presence and absence of PAβN. Total cell lysates were prepared by harvesting log-phase-grown LB cultures (OD600 = 0.5) and centrifuging an equivalent number of cells for each strain. Cell pellets were resuspended in 100 µl of phosphate-buffered saline (PBS) plus 100 µl 2× SDS-PAGE sample buffer (containing 5% β-mercaptoethanol), and samples were sonicated briefly prior to incubation at 100°C for 5 min. Lysate samples were separated by SDS-PAGE and subjected to Western blot analyses using OprM-, OprN-, and OprJ-specific monoclonal antibodies. Lysates were also probed with either a P. aeruginosa or K. pneumoniae anti-LpxC-specific antiserum. The full-length lpxC gene (including the upstream region) was PCR amplified and sequenced from these isolates using the primers PA LpxC.F (5′ ATCAGGAGTAGAGATGTG 3′) and PA LpxC.R (5′ AGAGGCCACCCTGAGGTG 3′) for P. aeruginosa and KP LpxC.F (5′ CATATGATCAAACAAAGGACTCTTAAA 3′) and LpxC.R (5′ GGATCCTTACGCCAGAACCATCGAC 3′) for K. pneumoniae. K. pneumoniae strains were also PCR amplified and sequenced with the fabZ-specific primers FabZ.F (5′ TTGACTACTGACACTCATACTCTGCAC 3′) and FabZ.R (5′ TCAGGCCTCCCGGCTACG 3′).

Mutagenesis of oprN.

In order to generate OprN deletion mutants in P. aeruginosa, an internal portion of the oprN gene was PCR amplified from P. aeruginosa PAO1 genomic DNA using the primers OprN.F (5′ GCGCAGTCGATCCGGAGC 3′) and OprN.R (5′ GCGCTCCTGGCGCTTGCC 3′), and the resulting amplicon was cloned into pCR2.1-TOPO (Life Technologies). Digestion with HincII removed an internal 401-bp fragment of oprN, and the SmaI-digested GmR cassette from pPS856 (34) was ligated in its place. The region containing oprN::GmR was excised from pCR2.1-TOPO using PvuII and ligated into SmaI-digested pEX100T (35), and the resulting plasmid was transformed into E. coli Top10 (Life Technologies). This strain was used as the donor in a triparental mating with various recipient P. aeruginosa strains and E. coli HB101 harboring pRK2013 (36), which served as the helper plasmid. Gentamicin-containing pseudomonas isolation agar (PIA) (Difco) was used to initially select for single-crossover mutants. Isolated colonies from this plate were collected and restruck for isolation on LB agar plates containing gentamicin and 5% sucrose, the latter of which was included to select for cells that had undergone a double-crossover recombination event. Finally, Gmr and sucrose-resistant mutants were confirmed to be cured of pEX100T by plating on LB agar plates containing 500 µg/ml carbenicillin, and mutations were confirmed by DNA sequencing using target gene-specific primers. All restriction enzymes were purchased from New England Biolabs.

Analysis of lpxC upstream region and backcrossing into P. aeruginosa.

The intergenic region separating ftsZ and lpxC was analyzed for the presence of a small RNA (sRNA) using the mfold server (http://mfold.rna.albany.edu). The contributions of mutations in this region to LpxC compound resistance were confirmed by backcrossing individual base changes into PAO397, a PAO1 isogenic mutant that cannot produce the MexAB-OprM, MexEF-OprN, MexCD-OprJ, MexXY, and MexJK efflux pumps (11). The primers LpxC.IG.For (5′ ATGGCGATGATGGGTACC 3′) and LpxC.IG.Rev (5′ GAACTCGTCCTCGTAACG 3′) were used to PCR amplify a 1,333-bp fragment from genomic DNA from both wild-type PA-1955 and a FOR-generated LpxC-4R mutant of PA-1955, which harbors a C-to-A base change 11 bp upstream of the translational start site of lpxC. Each amplicon was cloned into pEX18ApGW (37), and the C-to-A-mutated allele was introduced into PAO397 by conjugation. After primary selection on LB agar plates containing 100 µg/ml carbenicillin, exconjugates were counterselected on LB agar containing 5% sucrose and 0.5 µg/ml LpxC-4. SucroseR and LpxC-4R colonies were confirmed to be cured of pEX18ApGW by replica plating on LB agar containing 100 µg/ml carbenicillin. Successful C-to-A replacement at 11 bp upstream of lpxC was confirmed in carbenicillin-susceptible clones by PCR and sequencing. Backcrossed PAO397 strains were characterized by MIC testing and Western blot analysis as described above.

LpxC and sRNA stability and expressional analyses.

To measure changes in RNA stability as a result of the mutated lpxC gene upstream noncoding region, P. aeruginosa strain PA-1955 and its isogenic C-to-A mutant, along with PAO397 and its backcrossed C-to-A mutant (described above), were grown to mid-logarithmic phase (OD600 = 0.45) in LB with shaking. Transcription was halted through the addition of 200 µg/ml rifampin, and total RNA was extracted at both 0 and 5 min posttreatment. qRT-PCR was conducted according to the 2−ΔΔ method (38) using the primers lpxC.F (5′ AAGAAGTTCATCCGCATCAAGCGC 3′) and lpxC.R (5′ CCTCCTTGACGAAGGAAGTACTGG 3′) and rpsL.F (5′ GCAAGCGCATGGTCGACAAGA 3′) and rpsL.R (5′ CGCTGTGCTCTTGCAGGTTGTGA 3′), the latter of which served as an internal control. Triplicate samples were analyzed, and changes in relative expression levels found in C-to-A mutants were expressed as fold changes over their respective parent strains.

Pharmacokinetic studies and analysis.

LpxC-4 pharmacokinetic properties were determined in neutropenic female CF-1 mice with active thigh infections. Plasma exposure was determined following subcutaneous dosing at dose levels of 18.75, 75, and 300 mg/kg. Animals (n = 3/time point/dose level) were administered doses in 40% cyclodextrin at a dose volume of 0.2 ml. Terminal blood samples were taken via cardiac puncture at 0.25, 0.5, 1, 2, 4, 6, and 8 h postdose, collected into disodium EDTA Microtainers (BD Biosciences, Franklin Lakes, NJ), and processed for plasma. Concentrations of LpxC were determined by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Briefly, plasma samples (50 µl) were subjected to protein precipitation with acetonitrile (200 µl) containing an internal standard (diclofenac; Sigma-Aldrich St. Louis, MO), and 5 µl of the resulting supernatant was injected onto a Luna C18 column (2.7 µm, 30 × 3 mm; Phenomenex, Torrance, CA). The column was equilibrated with solution A (95:5 water-acetonitrile, 0.1% acetic acid) at a flow rate of 1.0 ml/min. The gradient was started at 5% solution B (95:5 acetonitrile-water, 0.1% acetic acid) and was increased to 95% solution B from 0.5 to 1.5 min. Conditions were held at 95% solution B until 2.0 min and then returned to starting conditions by 2.3 min and held for an additional 0.5 min for a total run time of 2.8 min. The effluent was analyzed with an API-4000 mass spectrometer detector (AB Sciex, Foster City, CA), fitted with a turbo ion spray interface and operated in the negative-ion mode. LpxC-4 and diclofenac were detected by multiple reactions monitoring at transitions 410.9/330.8 and 293.9/249.9, respectively. The dynamic range of the assay was 2.5 ng/ml to 10 µg/ml for LpxC-4.

LpxC-4 concentrations (mean data for three mice per time point were used to generate a single concentration-time curve) were fitted to a 1-compartment model using the software program WinNonLin, version 5.2 (Pharsight, St. Louis, MO). The maximum concentration (Cmax) was the highest observed concentration; Tmax was the earliest time at which Cmax occurred; and the elimination half-life was estimated as ln2/kel, where kel is the elimination rate constant derived from the slope of the log concentration-versus-time profile. The area under the concentration-time curve from 0 h to the last time point (AUC0–tlast) was calculated by linear trapezoidal approximation.

General in vivo procedures.

All animal experiments were conducted in accordance with regulations and established guidelines and were reviewed and approved by the Pfizer Institutional Animal Care and Use Committee. Comparative efficacy and dose response were determined in an immunocompetent model of septicemia. Exposure-effect and PK/PD driver assessments were determined using PA-1950 and KP-1487 as representative clinical strains in a neutropenic thigh and lung pneumonia model. Animals were housed in groups of 5 to 10 mice per cage and given access to food and water ad libitum for all studies. For immunosuppression studies, animals were rendered transiently neutropenic by administration of two oral doses of cyclophosphamide monohydrate (Sigma-Aldrich, St. Louis, MO). The first dose (150 mg/kg in 0.2 ml sterile water) was administered 4 days prior to infection, and the second dose (100 mg/kg in 0.2 ml sterile water) was administered 1 day prior to infection.

Acute septicemia model.

Groups of five CF-1 female mice (Charles River Laboratories) were infected intraperitoneally with P. aeruginosa and K. pneumoniae strains diluted in 3% brewer’s yeast (Sigma-Aldrich, St. Louis, MO). Inocula were prepared from frozen stock with 100 µl spread on sheep blood agar plates and incubated overnight at 37°C. The resulting colonies were harvested, washed with brain heart infusion broth (BHIB), and diluted to a target inoculum of ~1 × 104 CFU/mouse delivered in 0.5 ml. Mice were treated with LpxC-4 at 0.5 h and 4 h postinfection. The compound was administered subcutaneously (s.c.) in a dose volume of 0.2 ml. Doses ranged from 0 to 100 mg/kg, and the compound was reconstituted in 40% cyclodextrin in sterile water prior to administration. Animals were euthanized at the onset of therapy and 24 h post-bacterial challenge, spleens were surgically excised, and bacterial colonies were enumerated by standard plate counting methods (CFU/spleen) on MacConkey’s agar for cultivation. Efficacy was established as the difference in bacterial burden over 24 h and applied to fractional dose-response model with the following characteristics: 24-h maximum effect (Emax) = lower limit of detection − 0-h log CFU; 24-h minimum effect (Emin) = 0-h CFU − 24-h upper limit of detection; 0-h mean = average log CFU at onset of therapy. The fractional response (%) = (24-h delta log CFU − Emin)/(Emax − Emin). ED50 values were determined from nonlinear regression analysis of the fractional dose-response using the software program GraphPad Prism, version 3.02 (GraphPad Software, La Jolla, CA).

Neutropenic thigh infection model.

CF-1 female mice were rendered transiently neutropenic as described above and challenged with 0.1 ml of a 1:1 ratio of Cytodex beads (lot 124H0068; Sigma): 10−4 dilution of plate-scraped PA-1950 or KP-1487. The inocula were delivered intramuscularly into the left femoral bicep of CF-1 mice. Subcutaneous therapy was initiated 1 h following bacterial challenge using q24h, q12h, q6h, and q3h dosing intervals that covered a 64-fold dose range from 3.69 to 300 mg/kg/course of therapy. Efficacy was determined by assessing the bacterial burden at 0 and 24 h post-initiation of therapy (1 and 25 h post-bacterial challenge). Infected thighs were aseptically excised and processed for bacterial burden enumeration. The change in CFU over 24 h was determined for each dose level, and the data were analyzed by nonlinear regressional analysis.

Pulmonary infection model.

Groups consisting of 8 isoflurane-anesthetized C3H/HeN female mice (Charles River Laboratories) were infected with P. aeruginosa PA-1950 by placing 40 µl of bacterial suspension in BHIB onto the external nares. Each mouse was held in a vertical position until the droplet was completely inhaled. The targeted inoculum was 1 × 106 CFU/animal. Therapy with LpxC-4 was initiated on day 1 of the study at 0.5 h post-bacterial challenge, and animals received an additional dose at 4 h post-challenge. This twice-daily dosing was repeated on day 2 of the study, animals were euthanized at 48 h post-challenge, and the lungs were aseptically removed and processed for bacterial burden enumeration. Efficacy was assessed as the difference in bacterial burden from the initiation of therapy to 48 h post-challenge.