Myxobacterium-Produced Antibiotic TA (Myxovirescin) Inhibits Type II Signal Peptidase

Abstract

Antibiotic TA is a macrocyclic secondary metabolite produced by myxobacteria that has broad-spectrum bactericidal activity. The structure of TA is unique, and its molecular target is unknown. Here, we sought to elucidate TA's mode of action (MOA) through two parallel genetic approaches. First, chromosomal Escherichia coli TA-resistant mutants were isolated. One mutant that showed specific resistance toward TA was mapped and resulted from an IS4 insertion in the lpp gene, which encodes an abundant outer membrane (Braun's) lipoprotein. In a second approach, the comprehensive E. coli ASKA plasmid library was screened for overexpressing clones that conferred TA^r. This effort resulted in the isolation of the lspA gene, which encodes the type II signal peptidase that cleaves signal sequences from prolipoproteins. In whole cells, TA was shown to inhibit Lpp prolipoprotein processing, similar to the known LspA inhibitor globomycin. Based on genetic evidence and prior globomycin studies, a block in Lpp expression or prevention of Lpp covalent cell wall attachment confers TA^r by alleviating a toxic buildup of mislocalized pro-Lpp. Taken together, these data argue that LspA is the molecular target of TA. Strikingly, the giant ta biosynthetic gene cluster encodes two lspA paralogs that we hypothesize play a role in producer strain resistance.

INTRODUCTION

Apersistent challenge in clinical antibiotic use is resistance development. This, combined with the abandonment of antibiotic drug discovery by most pharmaceutical companies, has led to a public health concern about the availability of efficacious antibiotics (36). Historically, microbial natural products (NPs) have been the primary source of clinical antibiotics. The advent of mass genome sequencing and new discovery technologies has reinvigorated NP drug discovery. Genomic analysis has highlighted the point that certain microbes are well endowed for NP production. One such group comprises the myxobacteria, where up to 10% of their genomes can encode secondary metabolite pathways (7, 13, 43, 44). Importantly, myxobacterial NPs often have novel structures and activities (25). For instance, the common lab strain DK1622 encodes 18 secondary metabolite gene clusters. Therefore, myxobacteria are a rich source for new drugs, particularly antibiotics, which are worthy of pursuit.

Antibiotic TA (producer strain isolated from Tel Aviv), also known as myxovirescin, megovalicin, or M-230B (6, 17, 21, 28), is a promising lead compound. TA has a novel structure consisting of 28-membered macrolactam lactone. The biosynthetic pathway for TA production has been partly elucidated and consists of hybrid polyketide and nonribosomal peptide synthetases (1, 22, 33–35). TA is a rapid bactericidal agent and has activity against many Gram-negative and some Gram-positive bacteria. Antibacterial activity is specific, as TA shows no toxicity toward fungi, protozoa, eukaryotic cells, rodents, or even humans (26). TA also exhibits unusually high adhesive properties toward biological and abiotic materials (27). For these reasons, TA has been proposed for the treatment or prevention of biofilm infections, such as periodontal diseases or infections derived from indwelling medical devices (4, 15, 31, 32).

The attractive properties of TA have led to synthetic synthesis and optimization efforts (3, 45). However, TA synthesis is arduous and not knowing the molecular target has blinded optimization efforts. Prior metabolic labeling experiments suggest that TA may inhibit cell wall biosynthesis (6, 51). However, the specific molecular target remains unknown. We thus pursued two complementary genetic approaches to elucidate TA's mode of action (MOA). As described below, genetic and biochemical results show that TA targets type II signal peptidase encoded by the lspA gene.

MATERIALS AND METHODS

Strains, plasmids, and growth conditions.

Bacterial strains and plasmids used in this study are listed in Table 1. The ASKA Escherichia coli clone library was generously provided by the National Institute of Genetics, Japan (12). Myxococcus xanthus was routinely grown at 33°C in CTT (1% Casitone, 10 mM Tris-HCl, pH 7.6, 8 mM MgSO₄, 1 mM KH₂PO₄, pH 7.6). E. coli was typically grown in Luria-Bertani (LB) medium. Antibiotics were used at the following concentrations: kanamycin (KAN), 40 μg/ml; ampicillin (AMP), 100 μg/ml; tetracycline (TET), 12.5 μg/ml; and chloramphenicol (CM), 25 μg/ml. Isopropyl-β-d-1-thiogalactopyranoside (IPTG) was added at 10 μM when screening the ASKA library and at 20 μM for cells carrying plasmid p-lspA (pCA24N-lspA). When needed, arabinose or glucose was supplemented at 0.2% (wt/vol).

Table 1.

Strains and plasmids

Strain or plasmid	Genotype or relevant properties	Phenotype	Construction	Source or reference
M. xanthus
DK1622	Wild type	TA+		Lab collection
DW1034	Δta1	TA−		47
ER5014B	ΩTn5::lac	TAOP Kanr		E. Rosenberg
E. coli
MG1655	Wild type			ATCC
DW37	imp4213	Kanr	BAS849 × P1:CAG18620→carB3092::Tn10(kan) imp4213	This study
E609L	lpp::Tn10	Tetr		K. Miller; 19
YX23	imp4213 lpp::IS4	Kanr	DW37 NTG mutagenesis→select TAr	This study
YX104	pCA24N-lspA	Cmr	MG1655 × pCA24N-lspA→select Cmr	This study
YX127	lpp::Tn10 pJY811	Tetr Cmr	E609L × pJY811→select Cmr	This study
YX130	lpp::Tn10 pJY851	Tetr Cmr	E609L × pJY851→select Cmr	This study
YX135	imp4213 pJY811	Kanr Cmr	YX23 × pJY811→select Cmr	This study
XL1-Blue	Cloning strain	Tetr		Stratagene
ASKA library	AG1 host	Cmr		12
CGSC8171	F−, purR106::Tn10	Tetr		CGSCa
CGSC6920	F−, zdh-299::Tn10	Tetr		CGSC
CGSC7385	F−, zdi-925::Tn10	Tetr		CGSC
CGSC6756	Hfr (PO118) zdi-57::Tn10	Tetr		CGSC
CGSC6728	Hfr (PO2A) pfkB20::Tn10	Tetr		CGSC
CGSC7387	F−, zdj-276::Tn10	Tetr		CGSC
CGSC7390	F−, zea-3068::Tn10	Tetr		CGSC
CGSC7234	F−, ruvA60 (polar)::Tn10	Tetr		CGSC
CGSC6758	Hfr (PO45) zed-977::Tn10	Tetr		CGSC
CGSC7395	F−, zed-3069::Tn10	Tetr		CGSC
CGSC7397	F−, zef-3129::Tn10	Tetr		CGSC
Plasmids
pJY811	pLspA-SK (WT), PBAD promoter	Cmr		49
pJY851	pLspA-SR, PBAD promoter	Cmr		49
pCA24N-lspA	pLspA+, PT5-lac promoter	Cmr		12

^aCGSC, Coli Genetic Stock Center.

Translation-dependent killing.

E. coli strain MG1655 was grown in LB broth to mid-log phase and diluted 1:10,000 in fresh prewarmed LB. Prior to the addition of test antibiotics, cultures were incubated with CM (20 μg/ml) for 5 min at 37°C. TA and polymyxin B were added at 1× the MIC (4 μg/ml and 0.5 μg/ml, respectively). At various times, aliquots from each treatment were removed, serially diluted, and then plated on LB agar to determine CFU.

Zone of inhibition (ZOI).

Overnight myxobacterial cultures were collected by centrifugation and resuspended in TPM buffer (10 mM Tris-HCl, pH 7.6, 8 mM MgSO₄, 1 mM KH₂PO₄, pH 7.6) to a calculated absorbance of 100 Klett units (∼3 × 10⁸ CFU/ml). A 5-μl aliquot of cells was then spotted on assay plates with or without arabinose or IPTG. Assay plates consisted of 1.5% agar in 1/2 CTT (CTT broth containing 0.5% Casitone). Plates with M. xanthus spots were incubated at 33°C for 12 or 48 h prior to indicator strain overlay (100 μl of OD₆₀₀ [optical density at 600 nm] = 1 cell culture mixed in 3 ml of molten 1/2 CTT–0.7% agar). Plate overlays were incubated overnight at 33°C before observation.

MIC.

To determine MIC values, 2-fold serial broth dilutions of purified antibiotics were done in 96-well microtiter dishes in LB medium. When indicated, IPTG was added at 20 μM. Inocula contained 5 × 10⁵ CFU/ml (160 μl of culture in final volume of 180 μl). Both antibiotic-only and strain-only controls were included. Microtiter dishes were scored after incubations at 33°C for 18 h. The MIC was scored as the lowest concentration of antibiotic that resulted in no visual detection of growth.

TA preparation.

Crude TA extract was prepared by following established protocols except that chloroform was used as the extraction solvent (35). A 100-ml culture of DK1622 yielded 100 μl of crude TA extract dissolved in chloroform. TA was also purified from strain Mx v48 to >95% purity as determined by high-performance liquid chromatography-mass spectrometry-diode array detector (HPLC-MS-DAD) analysis as previously described (6). As the availability of purified TA was limited, purified compound was used only for biochemical and MIC tests. Genetic screens were conducted against an extract or M. xanthus TA producing lawns.

Mutagenesis and isolation of resistance mutants.

N-Methyl-N′-nitro-N-nitrosoguanidine (NG) mutagenesis was conducted on a hyperpermeable E. coli strain, DW37 (imp4213) (Table 1) (30). Freshly dissolved NG was added (100 μg/ml) in TM buffer (10 mM Tris-HCl, pH 7.6, and 8 mM MgSO₄), and the culture was shaken at 37°C. At various times, cells were collected and washed twice in TM buffer and then were overlaid with soft agar on LB agar plates. A filter disk containing 2 μl of TA extract was placed on top of the cell layer. After overnight incubation at 37°C, a ZOI formed around the filter disk, within which resistant colonies subsequently arose. Resistant colonies from early to late time points (24 to 72 h) were picked and tested for TA^r (filter disk ZOI) by comparing sensitivity to the parent strain.

Mapping the TA^r locus.

To identify the mutation that caused TA^r in YX23, Hfr crosses and phage P1 transductions were conducted by following standard procedures (16). For Hfr crosses, a series of donor strains were mated with strain YX23. Exconjugants were then cross-streaked over a TA extract line dispensed with a capillary tube on LB agar (supplemented with TET and KAN). The DW37 and YX23 strains served as negative and positive controls, respectively. The TA sensitivity of exconjugants was scored, and the percentage of Tet^r TA^s cells was calculated for mapping. Subsequently, bacteriophage P1 transductions from a panel of known Tn10 insertions were used to more precisely map the TA^r locus using the described scoring method (5, 20). Flanking PCR primers were then designed to amplify and sequence the putative TA^r locus. Primers used for amplification were Lpp-F (5′-AAGTGCCTTCCCATCAAAAA-3′) and Lpp-R (5′-AGTAGCGGTAAACGGCAGAC-3′).

Genetic screen against ASKA library.

An overnight culture of strain ER5014B was collected by centrifugation and resuspended in TPM buffer to a calculated absorbance of 200 Klett units (∼6 × 10⁸ CFU/ml). Then, 600 μl of cells was mixed with 1 ml 1/2 CTT broth and 6 ml of molten 1/2 CTT–0.7% agar and carefully overlaid on 1/2 CTT–1.5% agar with 10 μM IPTG in an OmniTray (catalog no. 62409-600; Thermo Scientific). This inducer concentration was empirically determined because it allowed >95% of the ASKA clones to grow. These plates (56 in total) were incubated at 33°C for 26 h. Thereafter, plates were overlaid with 6 ml molten LB–0.7% agar containing 10 μM IPTG. After brief solidification and drying, the ASKA library, which was in a 96-well format (LB broth with CM and 10 μM IPTG), was then transferred onto OmniTrays with a 96-pin replicator. YX23 and XL1-Blue strains were included in empty wells of each 96-well microtiter plate as controls. The plates were incubated at 37°C for 18 h before visual inspection of resistant colonies.

IPTG dose response.

A lawn of ER5014B was prepared as described above on 1/2 CTT–1.5% agar OmniTrays containing IPTG ranging from 0 to 200 μM. After incubation at 33°C for 26 h, a thin layer of LB agar containing the corresponding concentration of IPTG was overlaid. Overnight cultures of MG1655 and YX104 (MG1655 p-lspA) were 10-fold serially diluted in a 96-well microtiter dish and then transferred with a replicator on top of the agar surface and incubated for 18 h at 37°C. The plating efficiency of both strains under different IPTG concentrations was determined and plotted.

Lpp processing and Western blot analysis.

An overnight culture of E. coli strain YX127 (lpp::Tn10 pJY811) was diluted 1:100 in LB medium supplemented with CM and 0.2% glucose and grown to early log phase (∼2 × 10⁸ CFU/ml). The culture was made into 500-μl aliquots which were treated with purified TA or globomycin at various concentrations for 5 min at 37°C prior to the addition of inducer (0.2% arabinose). Control cells were treated only with 0.2% glucose or 0.2% arabinose. After shaking at 37°C for 30 min, cells were harvested, washed, and resuspended at an adjusted optical density (OD₆₀₀) of 2.5 in 2× Laemmli sample buffer (Bio-Rad). These samples were then diluted 1:10, and 5 μl of each sample was loaded for SDS-PAGE separation. Immunoblottings were done according to standard protocols with a polyvinylidene difluoride (PVDF) membrane and blocked with 3% skim milk in TBST buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20) for 1 h (29). Both mature and unprocessed forms of Lpp were probed with polyclonal rabbit Lpp antiserum (kindly provided by S. Inouye) at a 1:60,000 dilution, and goat anti-rabbit horseradish peroxidate (HRP)-conjugated secondary antibody was used at a 1:25,000 dilution (Pierce).

Fluorescence microscopy.

An overnight culture of E. coli MG1655 was diluted in LB 1:100 and cultured to an OD₆₀₀ of 0.3. Antibiotics were added at 1× the MIC, and cultures were shaken at 37°C for 35 min. Cells were pelleted by centrifugation and resuspended in TPM buffer. Prior to microscopic observation, dye FM4-64 (Invitrogen) was added to the sample to a concentration of 1 μg/ml. Microscopic analyses were performed on a Nikon E800 microscope coupled to a digital imaging system (42).

RESULTS

Bactericidal activity of TA depends on de novo protein synthesis.

The bactericidal activity of most clinically used antibiotics depends on cell metabolism, as they target essential cellular enzymes or processes. To gain insight whether TA kills in a target-specific manner, we tested whether a bacteriostatic antibiotic (CM) could block TA's bactericidal activity (52). Polymyxin B, a bactericidal antibiotic that disrupts membrane integrity in a non-cell-growth-dependent manner, was used as a control. As shown in Fig. 1, we confirmed that TA has rapid bactericidal activity. Importantly, this bactericidal activity was blocked when cells were pretreated with the translation inhibitor CM (Fig. 1). In contrast, polymyxin B retained bactericidal activity when E. coli was pretreated with CM. Thus, unlike polymyxin B, the bactericidal activity of TA requires new protein synthesis and thus cell metabolism.

Fig 1.

Bactericidal activity of TA depends on cell growth. Comparisons of E. coli MG1655 kill kinetics against polymyxin B (PMB), TA, and chloramphenicol (CM). CM was static at 20 μg/ml. PMB (0.5 μg/ml) kill kinetics were identical with or without CM pretreatment. The bactericidal activity of TA (4 μg/ml) was blocked by CM pretreatment.

Antibiotic TA is the major diffusible factor that blocks E. coli growth.

Myxobacteria produce a wide range of antibiotics, and one of the best studied is TA. Recently, we showed that M. xanthus antibiotics play a role in predation and that TA is the major diffusible factor that blocks E. coli growth (47). Strain DW1034 (Δta1) cannot produce TA and failed to form a ZOI against wild-type (WT) E. coli compared to the parental DK1622 (TA⁺) strain (Fig. 2). This ZOI defect of the TA⁻ strain persisted even after prolonged preincubation (48 h) (Fig. 2). In contrast, a highly permeable E. coli strain (DW37; imp4213) elicited a large ZOI after a short 12-h M. xanthus preincubation and a significantly reduced but clearly detectable inhibition halo against the TA⁻ strain. This narrow inhibition halo shows that DW37 was susceptible to another unknown diffusible factor(s) produced by the TA⁻ mutant. By extension, a TA overproducer (TA^OP) strain (ER5014B) created a broader inhibition halo against both E. coli strains (Fig. 2, center and right middle panels) than did DK1622. We note that the TA^OP strain poorly swarms, which enhanced the apparent size of the inhibition halo (Fig. 2). These observations served as a foundation for conducting the genetic screens described below in lieu of using purified TA.

Fig 2.

Antibiotic TA is the major diffusible factor that blocks E. coli growth. A comparison of DK1622 (TA⁺), DW1034 (TA⁻; Δta1) and ER5014B (TA^OP) ZOI from growth on 1/2 CTT agar is shown against E. coli MG1655 and DW37 (imp4213) strains. M. xanthus preincubation times are shown.

Isolation and mapping of an lpp chromosomal mutation that confers TA^r.

Classic genetic methods are a proven approach to isolate antibiotic-resistant mutants that can lead to the elucidation of antibiotic MOA. To initiate such a pursuit, we chose to use E. coli, as facile genetic tools are available. To ensure screening sensitivity, the DW37 (imp4213) strain was selected for mutagenesis (Table 1). Fourteen TA^r mutants were initially isolated, and four were selected for further characterization. To determine whether resistance was specific toward TA, these isolates were tested against a panel of antibiotics. One mutant (YX23) was shown to exhibit selective resistance toward TA and was not cross-resistant toward other antibiotics (Tables 2 and 3). Consequently, YX23 was selected for genetic mapping. The TA^r locus was initially mapped by Hfr crosses to 38 to 43 min on the E. coli chromosome. Subsequent P1 transductions with a set of known Tn10 insertions in this region more precisely mapped the mutation to ∼37.5 min. As the lpp gene mapped to this region and was a plausible candidate, we tested and found that a plasmid containing lpp could complement and thus restore TA sensitivity to YX23 (Fig. 3B). Western blotting analysis with Lpp antibodies found that the Lpp protein was absent in YX23 (Fig. 3A). The lpp locus was then PCR amplified with a pair of flanking primers. Interestingly, the size of the PCR product was ∼1.3 kb larger than the size of the expected parental control PCR product. This large ∼1.7-kb PCR product was sequenced and found to contain an IS4 insertion in the lpp locus (nucleotide position 29) (the open reading frame [ORF] encodes Braun's lipoprotein). This result suggests that the loss of Lpp function confers TA^r.

Table 2.

Susceptibility of YX23 to antibiotic panel

Antibiotic	MIC (μg/ml)
	DW37 (imp4213)	YX23
Bacitracin	16	8
Ampicillin	0.5	0.5
Polymyxin B	0.5	0.25
Rifampin	0.063	0.031
Chloramphenicol	4	4

Table 3.

Effect of lpp and p-lspA genes on antibiotic susceptibility

Strain/gene	MIC (μg/ml)
	TA	Globomycin
DW37 (imp4213)	0.063	0.125
YX23 (imp4213 lpp::IS4)	1	1
MG1655	4	8
YX104 (MG1655 pCA24N-lspA)a	16	64
YX127 (lpp::Tn10 pJY811)b	4	32

^aSupplemented with 0.02 mM IPTG.

^bSupplemented with 0.2% arabinose.

Fig 3.

lpp null mutations confer TA resistance. (A) Western blot analysis with anti-Lpp antibody. A 6-kDa band was detected in YX127 (lpp⁺) and absent in YX23 (lpp::IS4). (B) TA sensitivity of indicated strains against a TA⁺ (DK1622) strain. lpp disruptions (YX23 and E609L) confer TA resistance, while a complementing p-lpp⁺ clone restores sensitivity (YX135 and YX127, respectively). Note that strain YX135 was hypersensitive to TA and exhibits a high frequency of mucoid colony formation. In contrast, an E609L strain ectopically expressing an lpp78K→R allele (YX130), which has a C-terminal substitution that blocks peptidoglycan covalent attachment (49), results in TA^r.

To test the above conclusion, we obtained the E. coli strain E609L, which has a known lpp mutation, resulting from Tn10 insertion/excision events (19). By Western blot analysis we confirmed that E609L does not make Lpp (data not shown) and further found that this strain was TA resistant compared to its parental strain (lpp⁺). When E609L was transformed with a plasmid expressing Lpp, the susceptibility of E609L toward TA was restored (Fig. 3B). In contrast, a missense mutation, which replaced the C-terminal Lys with an Arg in Lpp (lpp78K→R) and consequently cannot covalently attach to the cell wall, resulted in TA^r when expressed from a plasmid (Fig. 3B) (49). We therefore conclude that the absence or the failure of Lpp to attach to the cell wall confers TA^r. This result coupled with the fact that Lpp is not essential for cell growth led us to conclude that Lpp was not the direct target of TA.

Screen of ASKA library identifies lspA clone that confers TA^r.

The overexpression of a target protein against a cognate antibiotic can confer resistance toward that antibiotic and thus provides a genetic means to help elucidate compound MOA (48). Consequently, in parallel to the above studies, the comprehensive ASKA library was screened for a plasmid(s) that confers TA^r (12). This library contains ∼4,300 E. coli ORFs in an expression vector under P_Tn5-lac control. As the availability of purified TA was limited, we devised a screen using a TA^OP strain (Fig. 4). Here, M. xanthus strain ER5014B was grown as a lawn and then overlaid with soft agar upon which the ASKA library was transferred, the idea being that the M. xanthus lawn produces TA which diffuses through the agar and blocks E. coli growth. Conditions were optimized such that only resistant colonies could grow. From this systematic screen, nine candidate clones were found that conferred TA^r. To test whether these clones breed true and to change the genetic background from a cloning strain (AG1) to a laboratory WT strain, these plasmids were purified and transformed into MG1655. Importantly, only 1 of the initial 9 clones was found to breed true in MG1655. This clone carries lspA, and as shown in Fig. 5A, its overexpression confers TA^r, as indicated by the lack of a ZOI. Figure 5B further shows an IPTG dose response, illustrating that increased lspA expression increases levels of TA^r (Table 3). We note that at high IPTG levels (≥200 μM), the strain's plating efficiency drops due to toxic LspA overexpression (data not shown). Since lspA is an essential gene, these results suggest that LspA is the target of TA (46, 50).

Fig 4.

Genetic selection strategy used to identify E. coli overexpression clones that confer TA^r. The comprehensive ASKA library was grown in 96-well plates and transferred with a 96-pin replicator on an M. xanthus lawn that overproduces TA (ER5014B). (Bottom right) Screening result with a single TA^r colony (arrow). (Bottom left) E. coli growth in the absence of M. xanthus lawn.

Fig 5.

LspA overexpression confers TA resistance. (A) Comparison of ZOI of DK1622 against MG1655 and MG1655 overexpressing LspA. (B) IPTG dose response of LspA expression on efficiency of plating on an ER5014B lawn. At ≥0.2 mM IPTG, the efficiency of plating dropped due to toxic LspA overexpression. The parent E. coli strain without the lspA plasmid was used as a control.

Purified TA blocks Lpp lipoprotein processing.

The above genetic results indicate that TA may act as an LspA inhibitor. LspA encodes the type II signal peptidase that cleaves signal sequences from prolipoproteins. Prior studies with a known LspA inhibitor, globomycin, have similarly found that LspA overexpression confers globomycin resistance (8, 40) and that globomycin treatment blocks Lpp processing by inhibiting LspA (9). To biochemically investigate whether LspA was the target of TA, we tested whether TA blocks Lpp lipoprotein processing. For these studies a Δlpp E. coli strain was transformed with a plasmid that ectopically expressed lpp from the P_BAD promoter. In this strain under glucose repression, Lpp protein was not detected by Western blot analysis, while in the presence of the arabinose inducer Lpp was readily detected (Fig. 6). To test for inhibition of Lpp processing, early-log-phase E. coli cells were treated with TA or globomycin at the same time as Lpp expression was induced. As shown, mature (processed) Lpp migrates at a position that corresponds to 6 kDa (Fig. 6). Unprocessed Lpp migrates at a higher mobility, of 8 kDa. Importantly, inhibition of Lpp processing was observed with both TA and globomycin at sub-MICs (Fig. 6). Globomycin at 2 μg/ml and TA at 0.25 μg/ml inhibited approximately 50% of Lpp processing, indicating their respective 50% effective concentrations (EC₅₀s). These results show that TA inhibits Lpp processing in whole cells and support the genetic findings that LspA is the target of TA.

Fig 6.

TA and globomycin inhibit Lpp processing. A Western blot was probed with polyclonal sera to Lpp. The E. coli strain YX127 (lpp::Tn10, p-P_BAD-lpp) was grown and treated as indicated. The MICs of globomycin and TA against YX127 were 32 μg/ml and 4 μg/ml, respectively (Table 3). Glucose (0.2%, wt/vol) blocks Lpp expression, and arabinose (0.2%, wt/vol) induces Lpp expression. As indicated, antibiotics were added to cultures 5 min prior to arabinose induction. Mature Lpp was 5.9 kDa, while unprocessed Lpp was 8.3 kDa.

TA and globomycin cause similar morphological changes.

Prior studies showed that globomycin treatment causes morphological changes in bacterial cells that included localized cell lysis (11). To investigate the effect that TA has on cell morphology, we treated E. coli with these antibiotics at their respective 1× MICs (Table 3). Time course experiments found that a 35-min treatment resulted in noticeable morphological changes to many cells, yet only a few cells had actually lysed. Specifically, TA treatment resulted in localized cell lysis or plasmolysis that morphologically looked identical to globomycin treatment (Fig. 7) (11). Fluorescence microscopy of identical fields with a membrane stain showed that plasmolyzed cells had a membrane border, suggesting that the E. coli inner membrane had detached and was pulled back from the cell wall. We observed similar morphological changes to E. coli expressing an inner membrane-localized lipo-mCherry fluorescent reporter (42) following TA and globomycin treatment, again suggesting that local cell lysis originated from membrane detachment (data not shown).

Fig 7.

TA and globomycin cause similar morphological changes to E. coli (MG1655). Cells were treated with the indicated antibiotics at 1× MIC for 35 min prior to harvesting and washing in TPM buffer. Cells were observed with a 100× phase-contrast objective lens (left panels), and membranes were stained with FM4-64 (right panels).

DISCUSSION

The bactericidal activity of TA requires de novo protein synthesis (Fig. 1), suggesting that synthesis of new proteins may be required for killing. Parallel genetic approaches further led to the conclusion that type II signal peptidase (LspA) was the target of TA. First, LspA overexpression specifically conferred TA^r to E. coli. Second, inactivation of lpp also conferred TA^r. These findings were identical to previous reports concerning the known LspA inhibitor globomycin (14, 40, 46). Mechanistically, the expression of Lpp following globomycin (or TA) exposure appears to result in a toxic buildup of Lpp in the inner membrane, leading to a lethal cross-linking event between the cell wall and inner membrane (49). In support of this, a C-terminal Lys-to-Arg change at the Lpp murein attachment residue conferred TA^r by presumably relieving Lpp toxicity (Fig. 3B). Biochemically, we show that TA inhibits pro-Lpp processing in whole cells, similarly to globomycin (Fig. 6). In combination, these results provide compelling evidence that TA targets LspA. Killing likely occurs by two mechanisms. One mechanism involves the aforementioned mislocalization and toxic buildup of Lpp. The second mechanism likely prevents essential lipoproteins, e.g., LptE (2), from being properly localized to the outer membrane.

Previous metabolic labeling studies found that TA and globomycin cause a delayed inhibition in cell wall biosynthesis (6, 11, 51). In light of our discovery, we reinterpret those results: namely, the inhibition of LspA blocks the maturation of key lipoproteins required for murein biosynthesis and thus indirectly blocks cell wall biosynthesis. Consistent with this, two independent reports recently showed that the LpoA and LpoB lipoproteins bind to and are essential for PBP1A and PBP1B function in cell wall biosynthesis (24, 41). Thus, we hypothesize that TA and globomycin block the maturation of LpoA and LpoB, thus preventing their outer membrane localization and blocking their essential interactions with PBP1A and PBP1B. In addition, microscopic analysis showed that TA and globomycin cause similar cell morphology changes that result in plasmolysis (Fig. 7) (11). This deformation involves membrane detachment from the cell wall and likely correlates with a block in cell wall synthesis.

Visual comparison of the two-dimensional structures of TA and globomycin reveals structural similarities. For instance, the left portions of these molecules are lipophilic (Fig. 8, arrows), perhaps facilitating membrane insertion where the LspA target resides as a small integral membrane protein (18, 38). Second, the structures highlighted with dashed lines (Fig. 8) are similar and thus could serve as the active pharmacophore. Consistent with this, these regions contain amide bonds, the substrate of the LspA peptidase, and could thus bind the active site (38).

Fig 8.

Chemical structures of TA and globomycin. Arrows indicate lipophilic regions (left portion of TA) that may facilitate membrane penetration. Dashed lines highlight similar substructures that might represent pharmacophores.

It is intriguing that the M. xanthus genome includes four lspA genes, two of which are carried within the TA biosynthetic gene cluster (lspA1 and lspA2) (23). Most bacteria carry only one lspA gene (10). Based on operon structure, lspA3 and/or lspA4 (MXAN_0368 and MXAN_0369) appears to constitute a housekeeping lipoprotein processing function as these genes reside next to lgt, which encodes prolipoprotein diacylglyceryl transferase, also involved in lipoprotein processing (39). Bioinformatic analysis suggests that all four LspA proteins are functional, as they contain nearly all the described conserved/essential 15 residues (38). The only exception is that the LspA1 gene encodes an Ala in place of a Gly within region II (amino acid position 58). Importantly, this substitution was shown to result in a functional LspA peptidase in Bacillus subtilis (38). We hypothesize that the LspA1 and/or LspA2 protein confers TA^r. Thus, when the ta gene cluster is expressed, host resistance would be coexpressed (lspA1), an idea we are currently testing.

For a number of reasons, LspA is an attractive target for antibiotic drug discovery. First, this signal peptidase is, to the best of our knowledge, universally found in bacteria and is broadly essential in Gram-negative bacteria. In Gram-positive organisms, LspA appears to be conditionally essential or nonessential and plays a key role in pathogenesis, as many virulence factors are lipoproteins or require lipoprotein function (10, 37). Second, LspA is absent in eukaryotic cells, which eliminates any concerns about target-based toxicity in animals. Third, from a clinical perspective, LspA represents a novel target. Thus, current clinical resistance mechanisms are unlikely to show cross-resistance toward a novel antibiotic structure that acts on a novel target. Ideal antibiotics, which have a very low frequency of resistance development, typically inhibit multiple essential bacterial targets. Although TA appears to inhibit only LspA, its rapid bactericidal activity likely reduces resistance development. In addition, against E. coli, spontaneously TA-resistant colonies were detected at low levels: <10⁻⁸ per cell (unpublished data).

The identification of the molecular target for TA opens new opportunities for lead optimization. In particular, solving the three-dimensional structure of LspA alone and bound to TA would facilitate a rational lead optimization approach to improve potency and possibly the spectrum of activity. The availability of globomycin as an alternate scaffold for an LspA inhibitor would synergize these efforts and also open an opportunity to design a new scaffold that might be more amendable to synthetic chemistry. As TA exhibits significantly better whole-cell potency (2- to 10-fold) over globomycin, as determined by MIC and EC₅₀ (Table 3 and Fig. 6), and its structure has fewer amines (Fig. 8), suggesting a superior toxicity profile (26), we argue that TA represents a more desirable lead than globomycin.