Novel Cephalosporins Selectively Active on Nonreplicating Mycobacterium tuberculosis.

We report two series of novel cephalosporins that are bactericidal to Mycobacterium tuberculosis alone of the pathogens tested, which only kill M. tuberculosis when its replication is halted by conditions resembling those believed to pertain in the host, and whose bactericidal activity is not dependent upon or enhanced by clavulanate, a β-lactamase inhibitor. The two classes of cephalosporins bear an ester or alternatively an oxadiazole isostere at C-2 of the cephalosporin ring system, a position that is almost exclusively a carboxylic acid in clinically used agents in the class. Representatives of the series kill M. tuberculosis within macrophages without toxicity to the macrophages or other mammalian cells.

Introduction

Antibiotics that rapidly kill Mycobacterium tuberculosis in axenic culture require months to years to produce the same result in tuberculosis (TB) patients. Accordingly, TB treatment lags far behind that of other bacterial diseases in terms of treatment duration, number of antibiotics required, toxicity to the host, and cure rates. Although new drugs such as oxazolidinones (linezolid, Pfizer),[1] diarylquinolines (bedaquiline, Janssen),[2] and nitroimidazoles (e.g., delamanid)[3] offer hope of shortening TB therapy and reducing mortality in patients whose TB is resistant to the standard regimen, there is an urgent need to discover additional anti-TB drugs.

Even when M. tuberculosis is genetically sensitive to existing drugs, a small fraction of a replicating population survives exposure to each such drug in vitro. Such bacteria are termed “persisters”. They display class I phenotypic tolerance[4] in that, when the drug is removed and the persisters are allowed to replicate, application of the drug at the same concentration again kills the vast majority. Mechanisms of class I persistence range from temporary nonreplication of a small subpopulation[5] to heterogeneous expression of proteins that activate prodrugs and misincorporation of amino acids into proteins.[6−9] Genetically susceptible M. tuberculosis can also display class II phenotypic tolerance when external stresses, such as those imposed by the host immune system, prevent most of the population from replicating.[4,10−12] Ideally, TB should be treated with a combination of drugs such that bacteria displaying class I phenotypic tolerance to any one of them are killed by at least one of the others, and at least one of the drugs can kill nonreplicating M. tuberculosis that display class II phenotypic tolerance.

Some of the conditions in the host that can drive M. tuberculosis into replication arrest and class II phenotypic tolerance include residence in interferon γ (IFNγ)-activated macrophages that traffic the bacilli to acidified phagosomes and expose them to nitrosative and oxidative stress[13−15] or release of M. tuberculosis into the hypoxic milieu of necrotic granulomas.[16] The search for compounds active against nonreplicating M. tuberculosis has been pursued in a variety of in vitro nonreplicating models.[17−30] We recently developed a high throughput screening platform to identify small molecules that kill class II persistent M. tuberculosis that are rendered nonreplicating by a combination of four host-relevant conditions:[13−15,21,31−35] low pH (5.0), a flux of nitric oxide (generated from 0.5 mM nitrite at that pH), hypoxia (1% O2), and low concentrations of a fatty acid (0.05% butyrate) as the carbon source instead of the conventional carbon sources dextrose and glycerol.[21,36] Here, we report that this screening protocol has led to the discovery of the first cephalosporins, to our knowledge, that are selectively active against bacteria in a nonreplicating state. We describe an initial analysis of their structure–activity relationship.

Results

Identification of Cephalosporins Active on Nonreplicating M. tuberculosis

A high-throughput screening campaign against replicating and nonreplicating M. tuberculosis was carried out using a library of compounds from an in-house screening collection assembled at the University of Kansas. This screen led to the identification of three cephalosporin esters[37] (1–3, Figure a) whose activity was unique to nonreplicating M. tuberculosis. For comparison, cephalexin 4, a broad-spectrum antibiotic in clinical use, was also tested in the multistress nonreplicating model but found to be inactive. The three hit molecules were resynthesized to >96% purity and displayed MIC90s against nonreplicating M. tuberculosis of 1.8–2.7 μg/mL and 0.5–1.1 μg/mL during exposures lasting 3 or 6 days, respectively (Table ). Initial results were obtained with a strain of M. tuberculosis whose dual auxotrophy for pantothenate and lysine increases its safety for laboratory personnel,[38,39] and key results for select molecules were verified using virulent, wild-type M. tuberculosis H37Rv. Given the propensity of β-lactams for inoculum effects,[40] we also determined the MIC90s against nonreplicating M. tuberculosis using a 10-fold lower inoculum of A580 of 0.01. The results at 3 days (0.7–3.1 μg/mL) were similar to those found for the higher inoculum cultures exposed for 6 days. Activity against replicating M. tuberculosis and against human HepG2 hepatoma cells was not seen up to the highest concentration tested (100 μg/mL), and the best selectivity index was ≥250. Thus, the activity of these compounds against M. tuberculosis was directly dependent on time of exposure and the state of nonreplication, inversely dependent on concentration of the bacteria, and selective for M. tuberculosis over human cells.

Figure 1

Structures of (a) cephalosporins 1–3 selectively active on nonreplicating M. tuberculosis, (b) for an inactive analogue, the clinically used antibiotic cephalexin, and (c) the C-2 oxadiazole cephalosporin 5.

Table 1

Primary Screening Actives that Target Non-Replicating M. tuberculosis

compound	NR-MIC90 3 days OD580 = 0.1 (μg/mL)	NR-MIC90 6 days OD580 = 0.1 (μg/mL)	NR-MIC90 3 days OD580 = 0.01 (μg/mL)	NR-MIC90 6 days OD580 = 0.01 (μg/mL)	R-MIC90 3 days OD580 = 0.01 (μg/mL)	HepG2 LD50 (μg/mL)	d6: % remaining in PBS	d6: % remaining NR medium	d6: % remaining NR medium + 0.5 mM NaNO2
1	1.76	1.05	0.90	0.64	>100	>75.69	80	100	100
2	2.61	0.52	3.11	n.t.a	55.44	>100	80	100	100
3	2.69	0.45	0.7	n.t.a	>100	>100	100	100	40
cephalexin 4	61.35	70.42	97.93	73.04	33.99	>100	50	100	100

n.t. = not tested.

Stability in Cell-Free PBS and Nonreplicating Medium

Because some molecules are chemically unstable in the multistress model of nonreplication,[21,35,36] compounds 1 and 2 were tested and found to be stable for up to 6 days in cell-free PBS and nonreplicating medium containing or omitting NaNO2 (Figure a and b). However, 3 was unstable in cell-free nonreplicating medium containing NaNO2 (Figure c; summarized in Table ). For comparison, cephalexin was partially unstable in cell-free PBS and stable in cell-free nonreplicating medium either containing or lacking NaNO2 (Figure d).

Figure 2

Cell-free stability of primary screening hits. Molecules were incubated at 37 °C in PBS (blue) or nonreplicating medium without (orange) or with (red) NaNO2. Data are averages of replicate samples ± standard deviation.

Structure–Activity Relationship (SAR) Studies

These promising results prompted us to undertake an initial structure–activity relationship survey. For each new analogue, we determined the activity against Mtb under both nonreplicating (NR) and replicating (R) conditions. In addition, each compound was assayed for cytotoxicity against HepG2 cells. For the present discussion, analogues are presented in Tables –5 according to the chemical class investigated.

Table 2

Survey of C-2-Substituted Cephalosporinsb

These data were from a 3-day exposure to compound.

n.d. = not determined. n.t. = not tested.

Table 5

SAR of C-2 Cephalosporin Oxadiazolesb

These data were from a 3-day exposure to compound.

n.d. = not determined. n.t. = not tested.

To test if the ester moiety in compounds 1–3 was essential for activity, we synthesized five free acid analogues of 1 (9) or related molecules and tested three commercially available cephalosporins bearing C-2 carboxylic acids (cephalexin, cefdinir, and cephalothin). The activity of all the cephalosporin carboxylates tested against nonreplicating M. tuberculosis ranged from ∼50 to >100 μg/mL (Table ), signifying the importance of an ester moiety at C-2. To determine if additional functional groups could be tolerated at this position, we prepared and tested analogues containing various amides, alcohols, and ethers at C-2 (Table , compounds 5–18). All were found to be inactive. However, replacement of the ester moiety by the isosteric 1,2,4-oxadiazole in analogue 5 resulted in a compound that was close in activity to 1 (Figure c and Table ). Moreover, compound 5 was stable in cell-free nonreplicating medium containing NaNO2 (Figure ).

Figure 3

Cell-free stability of 5. Compound 5 was incubated at 37 °C in PBS (blue) or nonreplicating medium without (orange) or with (red) NaNO2. Data are averages of replicate samples ± standard deviation.

To determine if simply adding ester groups or an oxadiazole to C-2 of a classical cephalosporin would confer activity against nonreplicating M. tuberculosis, we made cephalexin analogues 4a–c bearing such modifications. Testing determined that analogues 4a and 4c were poorly active against nonreplicating M. tuberculosis and completely inactive against replicating M. tuberculosis (Table ). n-Propyl ester 4b was slightly more active with an NR-MIC90 of 15.5 μg/mL.

Table 3

C-2 Ester and Oxadiazole Analogues of Cephalexin

Given these results, we chose to concentrate on exploring the effect of C-2 and C-7 amino substitution in two series of analogues: the esters (Table ) and the oxadiazoles (Table ). All compounds were synthesized as summarized in Scheme . Esters were generated from commercially available 7-aminodeacetoxycephalosporanic acid (7-ADCA) by amidation, followed by esterification. The oxadiazole series was prepared by installing the desired oxadiazole onto the C-2 of carboxylic acid A[41] as shown in Scheme below. Compounds were typically purified by mass-directed HPLC fractionation and rendered in purities of ≥95% for biological evaluation.

Table 4

SAR of C-2 Cephalosporin Estersb

These data were from a 3-day exposure to compound.

n.d. = not determined.

Scheme 1

Synthetic Route for the Preparation of Ester and Oxadiazole Analogues

Numerous analogues in both series of compounds were active against nonreplicating M. tuberculosis. The activity profiles of these molecules were responsive to these basic chemical changes, which suggested that the compounds were exerting their activity through action at a discrete cellular target. For example, a preference for longish and unbranched esters at C-2 was generally observed with ethyl, propyl, and butyl esters being preferred. Activity was sharply diminished for analogues containing propargyl groups at this position (e.g., compounds 4, 6, and 22), although benzyl esters were tolerated (compounds 14, 24, and 29). Recalling that cephalexin analogues 4a and 4b containing a side chain bearing a primary amine were poorly active, we concentrated our initial SAR on neutral C-7 amide moieties (although the single C-7 propyl amine examined, compound 3, did have significant activity). Most of the compounds examined bore a substituted 3-phenylpropamide side chain or its ethereal analogue (e.g., compound 13). We also observed that moving the double bond from the Δ2,3 to the Δ3,4 had only a modest effect (cf. compounds 34 and 35).

Among the C-2 oxadiazoles (Table ), similar trends were observed with a few addenda. Here, a wider range of carbamates, including a single carbothioate, were prepared and found to be active (compounds 14–17). In this series, we noted that although compounds bearing para electron-withdrawing substituents retained excellent potency, in many cases measurable levels of cellular toxicity were also observed. Particularly notable examples included some p-Cl and 3,4-dichloro analogues (compounds 5, 12, 21, and 29).

Physicochemical Properties and Metabolic Stability

Compounds 1 and 5 were chosen as representative molecules of the alkyl ester and oxadiazole classes of cephalosporins that are active against nonreplicating M. tuberculosis, whereas cephalexin, cefdinir, and cephalothin were chosen as representatives of cephalosporins lacking such activity. The active cephalosporins shared higher values for clogP and pKa, whereas other properties such as H-bond donors, H-bond acceptors, molecular weight, heavy atom count, and rotatable bonds were similar (Table ).

Table 6

Predicted Properties of Representative C-2 Ester and Oxadiazole Cephalosporinsa

MW, molecular weight; HDB, H-bond donor; HBA, H-bond acceptor; PSA, polar surface area.

We next determined the hydrolytic stability of these compounds under strongly acidic conditions, such as would be encountered in the stomach. Both 1 and 5 were more stable at pH 2 (100% remaining after 4 h) than cephalexin (∼74% remaining) (Table ), whereas all three compounds were stable at pH 7 and degraded in base (pH 12). Compound 5 and cephalexin were soluble at 84 and 76 μM at pH 7.4, respectively, whereas 1 was less soluble at 23 μM (Table ).

Table 7

Stability and Solubility of 4, 1, and 5

	cephalexin 4	1	5
pH 2 stability (%)a	73.7	110	102
pH 7 stability (%)a	102	133	101
pH 12 stability (%)a	0.000	0.000	0.000
solubility (μM/pH 7.4)b	75.8	22.7	83.7

Percent remaining after a 4 h incubation at 37 °C.

Determined after shaking at room temperature for 4 h.

Parallel artificial membrane permeability assays (PAMPA) predicted that both 1 and 5 would be membrane permeable (Table ). However, unlike cephalexin, both 1 and 5 were rapidly metabolized by mouse liver microsomes (Table ). Compounds 1 and 5 were less susceptible to metabolism by human liver microsomes with half-lives of ∼80 min and CLint values suggestive of slow metabolism (Table ).

Table 8

Preliminary Pharmacokinetic Properties of 4, 1, and 5

	cephalexin 4	1	5
PAMPA (mean Pe(10–6) cm/s)	<0.0001a	7.28	13.4
PAMPA (log Pe)	<−9.99	–5.15	–4.87
mouse liver microsomes (t1/2, minutes)b	stablec	<5	<5
mouse liver microsomes, CLint (μL/min/mg protein)b	stable	unable to calculateb	unable to calculateb
human liver microsomes (t1/2, minutes)	stable	86.1	76.3
human liver microsomes, CLint (μL/min/mg protein)	stable	8.07	9.13

No compound detected.

Both compounds 1 and 5 unstable in assay conditions.

No metabolism observed at ≤60 min.

Next, we assessed the stability of compounds 1 and 5 in mouse plasma to determine the feasibility of testing these compounds for activity in a mouse model of tuberculosis. Both compounds 1 and 5 were completely transformed in mouse plasma in <5 min (Figure a). In human plasma, compounds 1 and 5 had half-lives of approximately 2–3 h (Figure b). Cephalexin 4 was stable in plasma from both species (Figure a and b).

Figure 4

Stability of compounds 1 and 5 in plasma. Compounds 1, 5, and cephalexin (4) were tested for stability in mouse (a) and human (b) plasma at the indicated time points. Stability was inferred by monitoring the parent ion. One of two similar experiments. Compound 1 was tested once in human plasma.

Selective Bactericidal Action of Cephalosporins on Nonreplicating M. tuberculosis

Narrow spectrum bactericidal activity is preferred for TB drugs for two reasons. First, TB treatment is protracted, and long-term exposure to broad-spectrum antibacterial agents can precipitate severe and sometimes fatal intestinal dysfunction, such as that caused by overgrowth of Clostridium difficile. Second, efficacy of a given drug against other bacterial infections can prompt its use in the community, including in people who have undiagnosed TB. Monotherapy of TB often selects for emergence of genetically resistant strains. The spread of such strains in the community would render the new drug progressively less useful for the treatment of TB. Hence, it was important to test the antimicrobial spectrum of the new cephalosporins against other bacteria. Compounds 1 and 5 had MIC90s > 100 μg/mL against replicating Pseudomonas aeruginosa, Escherichia coli, Staphylococcus aureus, Mycobacterium smegmatis, and Mycobacterium bovis BCG as well as against the fungus Candida albicans (Figure S1).

To determine the extent of bacterial kill, we exposed nonreplicating wild-type M. tuberculosis at an OD580 of 0.01 (low inoculum) to 1 and 5 for 7 days (Figure ). At ∼ 0.7–0.8 μg/mL, both compounds reduced colony-forming units (CFUs) by 2 log10. Compounds 1 and 5, at 3 and 10 μg/mL, respectively, reduced CFU to the extent that there were no recoverable colonies when 10 μL of undiluted sample was plated (≥3.4 log10 kill). Thus, no class I phenotypic tolerance was observed. Bactericidal activity of 1 was not enhanced by the addition of a β-lactamase inhibitor, clavulanate, and addition of clavulanate did not lead to activity of 3, 2, or 1 against replicating M. tuberculosis (data not shown). In contrast, clavulanate enhanced the replicating MIC90 of Meropenem 4-fold.

Figure 5

Bactericidal activity of compounds 1 and 5 for nonreplicating M. tuberculosis. Nonreplicating wild-type M. tuberculosis at an OD580 of 0.01 was exposed to compounds for 7 days, and surviving bacilli were enumerated on 7H11-OADC agar plates. The inoculum is shown in yellow. The limit of detection was 1 colony arising from 10 μL of undiluted sample. Error bars represent standard deviations of triplicates. One of two similar experiments.

Representative cephalosporins (compounds 5, 12e, 12i, 18d, 19d, 21b, 22c, and 23) from the preliminary SAR campaign (Tables –5) were tested for activity against nonreplicating wild-type M. tuberculosis. Analogues derived from compounds 1 and 5 were bactericidal in this assay (Figure S2).

Reactive Nitrogen Species Enhance Bactericidal Activity of 1 and 5 against Nonreplicating M. tuberculosis

The activity of 1 against nonreplicating M. tuberculosis increased in relation to the concentration of NaNO2 (Figure a), whereas that of rifampicin did not at ≤0.5 mM NaNO2 (Figure b). At 1 mM NaNO2, double the concentration used in the nonreplicating screening, we observed nitrite-dependent killing with rifampicin as well. We then tested both 1 and 5 for nitrite-dependence by coupling the outgrowth to a CFU-surrogate assay (charcoal agar resazurin assay; CARA)[42] that determines the approximate concentration of compound leading to ≥2–3 log10 CFU reduction as reflected by the ability of survivors to convert resazurin to a fluorescent product. Both 1 and 5 decreased fluorescence in a dose-dependent manner that was strongly enhanced by the addition of NaNO2 (Figure c and d). As observed for many of the β-lactams in this study, the activities of both 1 and 5 were more potent at a 10-fold lower inoculum of 0.01 and 7-day exposure (Figure d). Both compounds displayed nitrite-independent activity at the lower inoculum (Figure c and Figure d). Thus, nitrite contributed to a 32- to 64-fold enhancement of 1’s activity, but activity was not strictly dependent on an exogenous source of nitrite (Figure d).

Figure 6

Potentiation of activity of cephalosporins against nonreplicating M. tuberculosis by reactive nitrogen species. Wild-type M. tuberculosis was resuspended at an OD580 of 0.1 in nonreplicating medium containing indicated concentrations of NaNO2 (0–1 mM) and dispensed into separate microtiter plates for each NaNO2 concentration. Cells were then exposed to (a) 1 or (b) rifampicin for 7 days, after which a standard outgrowth assay was initiated to estimate the number of surviving cells. In a separate experiment, nonreplicating M. tuberculosis at a standard OD580 of (c) 0.1 or lower inoculum of OD580 of (d) 0.01 were treated with either 1 (red) or 5 (blue) in the presence or absence of 0.5 mM NaNO2 for 7 days. CARA fluorescence provides an estimate of mycobacterial viability; complete loss of fluorescence is associated with ≥2–3 log10 CFU reduction.

Nonreplicating-Active Cephalosporins Kill M. tuberculosis in Macrophages

Wild-type M. tuberculosis is typically growth-arrested, or replicates slowly, in activated macrophages, due in part to phagosomal acidification and macrophage production of reactive nitrogen species (RNS).[13,14] The multistress nonreplicating assay conditions were designed in part to mimic this phagosomal microenvironment.[21,35] We hypothesized that cephalosporins active in the nonreplicating model might be bactericidal against intracellular M. tuberculosis. To test this, we stimulated mouse bone marrow-derived macrophages with IFNγ, or left them unstimulated, infected them with wild-type M. tuberculosis, and treated them with 1, 5, or diluent alone. We observed approximately 1–2 log10 CFU reduction of intracellular M. tuberculosis in activated macrophages treated with 1 or 5 with no apparent toxicity to the macrophages. Compound 5’s bactericidal activity against intracellular M. tuberculosis was strictly IFNγ dependent (Figure a and b).

Figure 7

Bactericidal activity of (a) 1 and (b) 5 against intracellular M. tuberculosis. Mouse bone marrow-derived macrophages activated or not with 50 ng/mL of IFNγ were infected with wild-type M. tuberculosis. After a 4 h period for bacterial uptake, macrophages were washed and treated with 100 μg/mL of 1 or 5 for (a) 4 or (b) 3 days. Morphology of the macrophages was not affected by addition of 1 or 5 at the concentrations shown. One of five similar experiments.

Discussion

To our knowledge, this is the first report of β-lactams that only kill a given bacterium when it is nonreplicating, and the first report of β-lactams with activity against any one bacterial species that lacks broad-spectrum antibacterial activity. Early studies by Tuomanen et al. demonstrated that although many β-lactams lack activity against nongrowing cells, a minority killed starved, nonreplicating Escherichia coli and Streptococcus pneumonia.(43,44) Similar findings were recently observed in M. tuberculosis, as the combination of Meropenem and the β-lactamase inhibitor clavulanate killed both replicating and hypoxic, nonreplicating M. tuberculosis.[45] Meropenem-clavulanate lacked activity against nonreplicating M. tuberculosis in the conditions studied here. In addition to hypoxia, our conditions included a low pH, a flux of reactive nitrogen species, and a fatty acid carbon source. Faropenem was also reported to kill both replicating and nongrowing M. tuberculosis.(46) Like Meropenem, faropenem was inactive in our multistress model of nonreplication. The novel cephalosporins described here did not acquire activity against replicating M. tuberculosis when we included clavulanate in the assays.

Structurally, the two main classes of compounds explored herein differ from clinically used cephalosporins by the lack of a carboxylic acid moiety at C-2 (a notable exception being the prodrug cefuroxime axetil), and indeed, we showed that carboxylic acid 9 is inactive against nonreplicating Mtb. Early in this project, we considered whether the screening hit propyl ester was functioning as a prodrug, but the successful replacement of this moiety with the corresponding oxadiazole renders this possibility unlikely. A hydrophobic moiety at this position is tolerated, as seen with the n-propyl ester and oxadiazole, and possibly preferred, given the fact that the C-2 hydroxymethyl analogue 10 is inactive. Polarity at this position also plays a role given the inactivity of the amide analogous to the active esters. In the preliminary SAR pursued to date, we have also ascertained that biological activity is affected by the amide moiety attached to the central cephalosporin nucleus with chains ending in electron-poor aromatic rings being preferred and a moderate dependence of activity on the length of the chain leading to this point. The role of the β-lactam itself is currently ambiguous; although we know that hydrolytic cleavage of this ring results in an inactive compound, it is not clear whether this is because the β-lactam is essential per se, i.e., in analogy to the generally accepted mechanism of most β-lactams, which involve covalent binding of this group to the target protein, or because of a structural alignment resulting from the cephalosporin ring system.

A question of pressing interest is the molecular target that renders these cephalosporins profoundly active against M. tuberculosis in a nonreplicating state. Experiments to address this are underway by a variety of approaches but have not yet yielded an answer, although we note that our observation of structure–activity relationships is consistent with action through one or more specific targets. The canonical role of β-lactams in killing replicating bacteria has been widely accepted as the arrest of peptidoglycan biosynthesis. Disruption of the balance between new peptidoglycan synthesis and peptidoglycan cleavage by hydrolases leads to cellular lysis[47,48] due to a futile cycle in the synthetic pathway.[49] The sensitivity of hypoxic M. tuberculosis to Meropenem and clavulanate, albeit significantly less than the sensitivity of replicating M. tuberculosis, suggests that M. tuberculosis may require peptidoglycan biosynthesis to survive the hypoxic state.[45] In addition to classical D,D-transpeptidases that catalyze the formation of 4′ → 3′ peptidoglycan cross-links, M. tuberculosis may use at least five L,D-transpeptidases (LdtMT1–LdtMT5; LDTs) for peptidoglycan 3′ → 3′ cross-linking. The 3′ → 3′ cross-links account for ∼80% of peptidoglycan extracted from M. tuberculosis in stationary phase, a form of nonreplication.[50] One of the nonclassical L,D-transpeptidases, LdtMT2, plays a role in M. tuberculosis virulence in a mouse model of infection.[51] Meropenem and other carbapenems bind recombinant LdtMT1,2,4,5.[52] However, there are additional covalent targets of β-lactams, such as signal peptidases and proteases.[53−55] Thus, we anticipate that β-lactams that specifically target nonreplicating populations may have either single or multiple canonical or noncanonical targets.

Two of the cephalosporins with activity against nonreplicating M. tuberculosis were stable in PBS and in nonreplicating medium, whether or not the medium contained NaNO2. An equipotent compound, 3, was unstable. Thus, 3 may have entered the bacilli or otherwise exerted its bactericidal effect before its structure was transformed in the extracellular medium.

In animal and human tuberculosis, M. tuberculosis often resides in macrophages,[56] and the ability to kill intracellular bacilli is an important feature of antimycobacterial compounds. Two of the nonreplicating active cephalosporins, compounds 1 and 5, killed intracellular M. tuberculosis but with differential dependences on immune activation. Immune activation of M. tuberculosis-infected macrophages leads to profound changes of the phagosomal microenvironment that are anticipated to lead to growth arrest. These changes include phagosomal acidification to approximately pH 4.5 and induction of iNOS, which produces nitric oxide.[14,15]M. tuberculosis exhibits variable behavior in mouse bone marrow-derived macrophages, ranging from subexponential replication in nonactivated macrophages to slower replication, no net change in CFU, or a modest decline in CFU in activated macrophages. An inhibitor of dihydrolipoamide acyltransferase (DlaT) selectively kills M. tuberculosis and M. bovis BCG in vitro when they are nonreplicating, and this compound is effective against M. bovis BCG in activated macrophages.[17] Likewise, compound 5 killed intracellular M. tuberculosis when the macrophages were immune stimulated, consistent with our hypothesis that compound 5 exerts activity in an acidic, nitrosative phagosome. However, compound 1 killed intracellular M. tuberculosis both in the absence and presence of IFNγ activation and thus may possess some activity against replicating M. tuberculosis at the concentrations tested.

Summary

The potential ability of β-lactams to treat TB has been suggested for many years but has only recently gained substantial notice with the report of promising results using Meropenem in human trials.[45,46] It would be of considerable interest to test the role of β-lactams that target nonreplicating M. tuberculosis in TB therapy in combination with agents active against replicating M. tuberculosis. Cephalosporins with activity against nonreplicating M. tuberculosis identified in this study, 5 and 1, were nontoxic, stable in cell-free medium, stable at pH 2 and 7, soluble at pH 7.4, predicted to be membrane-permeable, active in macrophages, and inactive against the other bacterial and yeast species tested. Compounds 1 and 5 were relatively stable when incubated with human liver microsomes. Although compounds 1 and 5 were highly labile in mouse plasma, they were more stable in human plasma with half-lives of 2–3 h. Some analogues of compound 5 were active in the ng/mL range. We are continuing to study the SAR of 5 while seeking its targets.

Experimental Section

General Procedure for Synthesis of Esters: Propyl (6R,7R)-3-Methyl-8-oxo-7-(2-phenoxyacetamido)-5-thia-1-azabicyclo[4.2.0]oct-2-ene-2-carboxylate (1)[37]

To (6R,7R)-3-methyl-8-oxo-7-(2-phenoxyacetamido)-5-thia-1-azabicyclo[4.2.0]oct-2-ene-2-carboxylic acid (197.0 mg, 0.57 mmol) was added sodium bicarbonate (48 mg, 0.57 mmol) and a minimal amount of water (6.0 mL) to dissolve the starting material. The mixture was stirred at rt for 30 min until all solids were dissolved and then frozen and lyophilized. DMF (10.0 mL) was slowly added followed by 1-iodopropane (0.55 mL, 5.7 mmol). The reaction was stirred at rt for 16 h, then quenched with water, and extracted twice with Et2O. The Et2O layers were combined and then washed 3 times with water and once with brine solution. The organic layer was dried (Na2SO4), filtered, and concentrated. The residue was purified via MPLC (silica, 30% EtOAc/hexanes → 70% EtOAc/hexanes) to afford the title compound (130.0 mg, 59% yield). [α]D25 +71.2 (c 1.45, CH2Cl2); IR (film) νmax 1780, 1721, 1687, 1524, 1494, 1228, 755 cm–1; 1H NMR (500 MHz, CDCl3) δ 7.34 (m, 2H), 7.30 (s, 1H), 7.05 (tt, J = 7.7, 1.0 Hz, 1H), 6.94 (m, 2H), 5.88 (dd, J = 9.2, 4.7 Hz, 1H), 5.04 (d, J = 4.8 Hz, 1H), 4.58 (s, 2H), 4.24 (m, 2H), 3.53 (dd, J = 18.3, 1.0 Hz, 1H), 3.22 (d, J = 18.3 Hz, 1H), 2.15 (s, 3H), 1.75 (h, J = 7.4 Hz, 2H), 0.99 (t, J = 7.4 Hz, 3H); 13C NMR (126 MHz, CDCl3) δ 168.80, 164.03, 162.35, 157.03, 130.88, 129.96, 129.80, 123.01, 122.52, 115.05, 114.90, 67.51, 67.25, 58.44, 56.94, 30.27, 22.03, 20.18, 10.61; HRMS (ESI-TOF) calcd for C19H22N2O5SNH4+ [M + NH4]+ 408.1588, found 408.1605.

General Procedure for Synthesis of Oxadiazoles: N-((6R,7R)-3-Methyl-2-(3-methyl-1,2,4-oxadiazol-5-yl)-8-oxo-5-thia-1-azabicyclo[4.2.0]oct-2-en-7-yl)-2-phenoxyacetamide (5)

To a solution of 2,4-dinitrophenol (1.03 g, 5.61 mmol) in CH2Cl2 (10 mL) was sequentially added (6R,7R)-3-methyl-8-oxo-7-(2-phenoxyacetamido)-5-thia-1-azabicyclo[4.2.0]oct-2-ene-2-carboxylic acid (1.92 g, 5.5 mmol) in a minimal amount of 1,4-dioxane (8 mL) and DCC (1.15 g, 5.6 mmol) in 8 mL of CH2Cl2. The mixture was stirred at rt for 30 min, after which the mixture was filtered through a plug of cotton to remove the urea. To the filtrate was then added ethylamidoxime (411.0 mg, 5.6 mmol) in CH2Cl2 (7 mL), and the mixture was stirred at rt for 4 h. The mixture was then washed twice with sat. aq NaHCO3, filtered, and concentrated. The residue was then placed in a vacuum oven at 110 °C for 16 h, and the resulting residue was purified via MPLC (silica, 100% hexanes → 60% EtOAc/hexanes) to afford the title compound as an orange solid (703.2 mg, 52% yield). [α]D24 +79.6 (c 0.72, CH2Cl2); IR (film) νmax 1775, 1493, 1331, 1216, 754, 732, 690 cm–1; 1H NMR (400 MHz, CDCl3) δ 7.35 (m, 3H), 7.06 (tt, J = 7.4, 1.0 Hz, 1H), 6.96 (m, 2H), 5.95 (dd, J = 9.1, 4.8 Hz, 1H), 5.16 (d, J = 4.7 Hz, 1H), 4.60 (s, 2H), 3.61 (d, J = 18.4 Hz, 1H), 3.35 (d, J = 18.3 Hz, 1H), 2.49 (s, 3H), 2.25 (s, 3H); 13C NMR (126 MHz, CDCl3) δ 168.79, 167.63, 164.56, 156.99, 130.93, 129.98, 122.56, 117.45, 114.90, 67.25, 58.73, 57.25, 30.02, 20.40, 11.92. HRMS (ESI-TOF) calcd for C18H19N4O4S+ [M + H]+ 387.1122, found 387.1088.

Strains and Growth Conditions

Mycobacterial strains and media were prepared as described.[21,35] Briefly, wild-type M. tuberculosis H37Rv was cultivated at 20% O2 and 5% CO2 in Middlebrook 7H9 bacteriologic medium containing 0.2% glycerol, tyloxapol (0.02%), and 10% OADC supplement, and the M. tuberculosis strain mc26220 (ΔpanCDΔlysA)[38,57] was grown in similar medium with minor modifications: additional glycerol (final: 0.5%), OADC supplement, casamino acids (0.05%), l-lysine (240 μg/mL), and pantothenate (24 μg/mL). Cells were rendered nonreplicating at 1% O2 and 5% CO2 in a Sauton’s-based medium (per liter: 0.5 g of KH2P04, 0.5 g of MgS04, 0.05 g of ferric ammonium citrate, BSA (0.5%), NaCl (0.085%), tyloxapol (0.02%), l-lysine (240 μg/mL), pantothenate (24 μg/mL), butyrate (0.05%), and 0.5 mM NaNO2).

High Throughput Screen

Molecules from the University of Kansas in-house library were screened using a reported protocol[21,35] with minor modifications. For the replicating screen, 500 nL test agent was added to 50 μL of replicating M. tuberculosis mc26220 at an OD580 = 0.01, giving a final concentration of 20 μg/mL and 1% DMSO. After 7 days of incubation at 20% O2 and 5% CO2, the OD580 was determined. For the nonreplicating screen, M. tuberculosis mc26220 was washed 2× in PBS containing tyloxapol (0.02%; PBS-Tyl) and resuspended in nonreplicating medium containing 0.5 mM NaNO2, and 15 μL of cells were dispensed into 384-well tissue culture plates (Greiner, reference 781091). Cells were exposed to 150 nL of test compounds in DMSO, and plates were incubated for 7 days at 1% O2 and 5% CO2. After 3 days of exposure to test agents, M. tuberculosis in each well was diluted 5-fold by the addition of 60 μL of fresh replicating medium using a reagent dispenser (ThermoScientific), which also served to mix cells. After 7 days of outgrowth at 20% O2 and 5% CO2, the OD580 was determined. Primary screening hits and downstream assay data were managed using the CDD Vault from Collaborative Drug Discovery (Burlingame, CA. www.collaborativedrug.com)[58] and JChem for Excel and MarvinView (ChemAxon).

Activity against Replicating and Nonreplicating M. tuberculosis

For minimal inhibitory concentration (MIC) assays, compounds were serially diluted 2-fold in DMSO from 10 to 0.04 mM using a PerkinElmer Janus robot with a P30 row/column MDT head to make 100× compound source stocks in Greiner compound plates (384-well small volume conical well, reference number 784201). Compounds were then distributed into 384-well replicating and nonreplicating assays with M. tuberculosis mc26220 in 384-well microplates as described above. For colony forming unit assays, experiments were set up using wild-type M. tuberculosis single cell suspensions in 96-well tissue culture-treated plates (Corning). At select time points, aliquots of cells were serially diluted in PBS-Tyl and spread on Middlebrook 7H11 agar plates containing a 10% OADC supplement. Colonies were enumerated ∼3 weeks postplating. The minimal bacteriocidal concentration leading to 99% reduction in colony forming units (MBC99) was extrapolated from CFU data.

HepG2 Toxicity Assays

Toxicity assays using the human hepatoma cell line HepG2 were as described.[59] Briefly, HepG2 cells were propagated in Dulbecco’s modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS), pyruvate, glutamine, and nonessential amino acids. HepG2 cells were incubated for 2 days with DMSO vehicle control or test compounds (≤1% DMSO final) at 3000 cells/well in 384-well tissue culture plates (Greiner reference 781091). Cellular viability was determined after 2 days by measuring ATP content with a CellTiter-Glo kit (Promega).

Microbial Spectrum

Select compounds were tested for activity against a panel of replicating Gram positive and Gram negative bacteria (Mycobacterium smegmatis, Staphylococcus aureus, Escherichia coli, and Pseudomonas aeruginosa) and yeast (Candida albicans). Bacteriologic medium and assay conditions were as described.[21] In brief, 200 μL of cells at an OD580 of 0.01 in a sterile, clear tissue culture-treated Corning 96-well plate were exposed to DMSO or drug, and growth was determined by optical density.

Stability Assay

Compounds were dissolved at 50 μg/mL in cell-free PBS (pH 7.4) or cell-free nonreplicating medium (pH 5.0) containing or not 0.5 mM NaNO2. Poorly soluble compounds were dissolved at 5 μg/mL and in a 50:50 (vol:vol) solution of acetonitrile and PBS, or acetonitrile and nonreplicating medium containing or not 0.5 mM NaNO2. The nonreplicating medium was as described above except that BSA, tyloxapol, lysine, and pantothenate were omitted. Solutions containing acetonitrile had their pH adjusted to 5.0 (the additional acetonitrile increased the pH from 5.0 to 5.8). Samples were incubated at 37 °C, and aliquots were removed every 12 h for analysis by LC-MS. Data represent percent remaining of the parent compound compared to that at the start of the experiment.

Cheminformatics

Tanimoto similarity between molecules and cheminformatic analysis of chemical properties (including ClogP values) were determined in Collaborative Drug Discovery (CCD) (Burlingame, CA; www.collaborativedrug.com)[58] using ChemAxon software.

Physicochemical, Permeability, and Metabolism Studies

These assays were performed by BioDuro (Shanghai, China). Mouse and human liver microsomal stability was assayed in triplicate at 0, 15, 30, 45, and 60 min.

Plasma Stability

Cephalosporins 1, 5, and cephalexin were spiked into lithium heparin-treated human and CD-1 mouse plasma (bioreclammation) at 1 μg/mL. Spiked samples were incubated at 37 °C, and extractions were performed at 5, 15, 30, 60, and 180 min. The reactions were quenched and proteins precipitated at each time point by adding 20 μL of spiked plasma to 200 μL of extraction solvent containing methanol/acetonitrile (1:1; vol/vol) and 10 ng/mL of verapamil (Toronto Research Chemicals, Inc.) as an internal standard (IS). In addition, 20 μL of 1:1 acetonitrile/water (ACN/H2O; vol/vol) were added. A reference sample was created by adding 20 μL of unspiked plasma to 200 μL of the extraction solvent. After the plasma enzymes were denatured by the extraction solvent, 20 μL of a 1 μg/mL solution in 1:1 ACN/H2O was added to the reference sample. Extracted samples were vortexed for 5 min and then centrifuged at 3000 rpm for 5 min. Then, 100 μL of extract was transferred to 100 μL of ddH20 for LC-MS analysis. LC-MS analysis was performed with an Agilent 1260 liquid chromatography system coupled to a 4000 Qtrap mass spectrometer (AB Sciex) in multiple reaction monitoring (MRM) mode with positive electrospray ionization (ESI) and an Agilent column, SB-C8, 2.1 × 30 mm, 3.5 μm. Mobile phase A was 0.1% formic acid in 100% H2O, and mobile phase B was 0.1% formic acid in 100% acetonitrile. Injection volumes were routinely 2 μL. The ions monitored were compound 1 (m/z 387.1/195.9), compound 5 (391.1/199.9), cephalexin (348.1/158.1), and verapamil (455.4/165.2). The percentage remaining was determined at each time point by dividing the sample analyte/IS peak area ratio by the reference sample analyte/IS peak area ratio.

Charcoal Agar Resazurin Assay (CARA)

The CARA was used as described.[42] In brief, 10 μL aliquots from replicating or nonreplicating MIC90 assay plates were removed and spotted onto microplates containing 200 μL of 7H11-OADC-charcoal agar in each well. The microplates were then incubated for 7–10 days at 37 °C at 20% O2 and 5% CO2. The film of bacterial growth (microcolonies) on the microplates was semiquantitated by the addition of 40 μL of a 1:1 (v/v) mixture of Alamar blue (AB) and Tween80 (TW80) and 1 h of further incubation at 37 °C at 20% O2 and 5% CO2. In some cases, if the CARA microplate appeared dry, all wells were prewetted with 40 μL of PBS prior to the addition of the AB:TW80 developing solution. Fluorescence was determined by top-read with excitation at 530 nm and emission at 590 nm. The CARA-minimal bactericidal concentration leading to ≥99% loss in CFUs (CARA-MBC≥99) was estimated as the lowest concentration of drug leading to complete loss of Alamar blue fluorescence.

Macrophage Infections

Primary bone marrow-derived macrophage infections were performed as described.[17,60−62] In brief, ∼1 × 105 macrophages isolated from 8-week old female C57Bl6 mice were grown in 48-well plates in DMEM supplemented with 4.5 g/L of glucose, 0.584 g/L of l-glutamine, 1 mM pyruvate, 10% FBS, and 10% L-cell conditioned medium containing or not 50 ng/mL of recombinant mouse IFNγ and infected with wild-type M. tuberculosis H37Rv at a multiplicity of infection of 1–5. Log phase wild-type M. tuberculosis was allowed to infect macrophages for 4 h, after which medium and extracellular M. tuberculosis were removed by two washes with PBS and replaced with fresh medium containing compounds or not at 1% DMSO. At the times indicated, macrophages were washed and lysed with PBS supplemented with 0.5% Triton X-100. Surviving bacilli were enumerated on 7H11-OADC agar plates. Macrophage supernatants were assayed for nitrite with the Greiss assay.