Cell division inhibitory activity of the tested compounds was performed as described previously (Stokes et al., 2005). Overnight B. subtilis culture was grown in starvation medium supplemented with 1% hydrolyzed casein and then diluted in starvation medium supplemented with 3% hydrolyzed casein and grown at 37°C. The culture was diluted to A₆₀₀ of ∼0.06, and 10 μL aliquots were added to transparent 96-well microtiter plates containing dilutions of the screening compounds in 100 μL volumes of medium. After incubation for approximately 5 h (4–5 generations) at 37°C, 20 μL culture samples were transferred to poly-L-lysine-coated slides for microscopy. Cell morphology was assessed by phase-contrast light microscopy.

B. subtilis, S. aureus USA300 #417 and S. aureus USA300 #2690 tested in this assay were already available in our in-house collection, S. pyogenes (ID:0716100032) and S. agalactiae (ID:0216100214) are clinical isolated strains in the First Affiliated Hospital of Guangzhou Medical University, other strains were purchased from American Type Culture Collection (ATCC, USA). Antimicrobial susceptibility tests were conducted in 96-well microplates using the broth microdilution procedure described in the Clinical and Laboratory Standards Institute (CLSI) guidelines (Wikler et al., 2009). Cation-adjusted Mueller Hinton broth for all the S. aureus strains, or Brain Heart Infusion broth for antibiotic-susceptible E. faecium strain ATCC 49624, vancomycin-resistant E. faecium strain ATCC 700221, S. agalactiae ATCC 13813, S. pyogenes ATCC 12344, S. pyogenes ATCC 19615 and M. catarrhalis ATCC 25240, or Mueller Hinton broth for the other strains were used in the assays. After incubation for 18 h at 37°C, the absorbance at 600 nm (A₆₀₀) was recorded using a microplate reader (Bio-Rad laboratory Ltd., UK) and the percentage of bacterial cell inhibition with respect to vehicles (1% DMSO) was calculated. The MIC was defined as the lowest compound concentration at which the growth of bacteria was inhibited by ≥90%. Three independent assays were performed for each test.

A growing culture of S. aureus ATCC 29213 or E. coli ATCC 25922 were diluted to approximately 10⁵ CFU.mL^-1 in volumes of Cation-adjusted Mueller Hinton broth or Mueller Hinton broth, respectively, containing various concentrations of compound 1. Cultures were incubated at 37°C, shaking. At the appropriate time intervals, 100 μL samples were removed for serial dilution in 900 μL volumes of Mueller Hinton broth and 100 μL volumes from three dilutions were spread on to MH agar. Cell counts (CFU.ml^-1) were enumerated after incubating the plates at 37°C for 18 h.

Staphylococcus aureus and E. coli FtsZ was cloned, overexpressed, and purified as described previously (Sun et al., 2014; Li et al., 2015). The light scattering assay was performed using a protocol adapted from the literature (Andreu et al., 2010). The polymerization of recombinant SaFtsZ or EcFtsZ was measured using 90° light scattering in a thermostatically (37°C) controlled fluorescence spectrometer (Agilent Cary Eclipse). Both excitation and emission wavelengths were set to 600 nm with a slit width of 2.5 nm. FtsZ (6 μM) in 20 mM of Tris buffer (pH 7.4, containing 0.01% Triton X-100 to avoid compound aggregation) was placed in a 10 × 2 mm (excitation path) cell, the reaction was started by consecutive additions of 20 mM KCl, 5 mM MgCl₂, 1 mM GTP and different concentrations of the test compound. One percentage DMSO and 20 μg/mL methicillin were tested as vehicle and negative in this assay.

Staphylococcus aureus FtsZ (12 μM) was incubated in the absence and in the presence of 3 μg/mL of 1 in 50 mM MOPS buffer (pH 6.5) at 25°C. After 10 min, 5 mM MgCl₂, 50 mM KCl, and 1 mM GTP were added to the reaction mixtures and incubated at 37°C for 15 min. Then, 10 μL of the sample mixtures were placed on a glow-discharged Formvar carbon-coated copper grid (400 mesh) for 10 min. The grids were subsequently subjected to negative staining using 10 μL of 0.5% phosphotungstic acid (PTA) for 30 s, air-dried and digital images of the specimen were observed with a transmission electron microscope (JEOL model JEM 2010) operated at 200 kV and equipped with a Gatan MSC 794 CCD camera.

The Effect of 1 on the GTPase activity of recombinant FtsZ was measured in 96-well microplates using a CytoPhos phosphate assay Biochem Kit (Cytoskeleton, USA) according to an optimized protocol and the manufacturer’s instructions. FtsZ (6 μM) was preincubated with vehicle (1% DMSO) or different concentrations of each test compound in 20 mM Tris buffer (pH 7.4, containing 0.01% Triton X-100 to avoid compound aggregation) for 10 min at 25°C. Then 5 mM of MgCl₂ and 200 mM of KCl were added. Reactions were started with the addition of 500 mM GTP and incubated at 37°C. After 30 min, the reactions were quenched by adding 100 mL of Cytophos reagent for 10 min. Inorganic phosphate was quantified by measuring the absorbance at 650 nm with a microplate reader (Bio-Rad laboratory Ltd., UK).

SaFtsZ (10 μM) was incubated without or with 1 at different concentrations (3 and 6 μg/mL) in 20 mM Tris buffer (pH 7.4, containing 0.01% Triton X-100 to avoid compound aggregation) for 30 min at 25°C. The far-UV CD spectrum was monitored over a wavelength range of 200–250 nm using a JASCO J-810 spectropolarimeter equipped with a temperature controller and a 0.1 cm path length quartz cuvette. An average of five scans was taken for each spectrum. Deconvolution and statistical analysis of the CD spectra were performed using Jasco software and Origin 6.0 software, respectively.

A culture of B. subtilis containing the IPTG-inducible plasmid for the overexpression of green fluorescence protein (GFP)-tagged FtsZ was grown in LB medium supplemented with 30 μg/mL of chloramphenicol. After overnight incubation, a sample of the culture was diluted to 1% in the LB medium containing 1 μg/mL of 1 and 40 μM of IPTG. After 4 h incubation at 37°C, the B. subtilis cells were fixed, harvested and resuspended in PBS buffer containing 0.25% of agarose. 10 μL of sample mixture were added to a pretreated microscopic slide with 0.1% (w/v) poly-L-lysine and visualized using a fluorescence microscope at 60× oil immersion magnification with a standard FITC filter set. The images were captured using an Olympus Bio Imaging Navigator FSX 100 microscope.

The molecular modeling procedures were performed using the MolSoft ICM 3.8.4 software (Neves et al., 2012). The X-ray crystal structure of S. aureus FtsZ in complex with a cell division inhibitor (PC190723) and GDP was downloaded from the PDB database (PDB entry: 4DXD; resolution: 2.0 Å) (Tan et al., 2012). Water molecules and co-crystal ligands were removed from the structure and the protein was prepared for docking using an automated procedure of ICM. Briefly, the hydrogen atoms were added to the structure, MMFF atom types and charges were assigned, and His, Asn and Gln residues were optimized to the best hydrogen bonding network. Energy maps accounting for hydrogen bonding, van der Waals, hydrophobic and electrostatic interactions were pre-calculated for two possible binding sites centered at either the co-crystal ligand PC190723, or at the GDP molecule. Each putative binding site was defined as a cubic grid including all FtsZ residues within a 10 Å cutoff from the co-crystal structure. The grid spacing was set to 0.5 Å. The structure of 1 was sketched in 2D and converted into 3D using the ICM molecule editor. The molecules was flexibly docked against the pre-calculated grid potential maps and scored using the ICM scoring function. One hundred independent docking runs were performed and the results were clustered in order to filter repeated binding modes. The top-scoring poses were visually inspected.

The effects of 1 on eukaryotic tubulin polymerization were monitored by fluorescence microscopy via using a tubulin polymerization assay kit (BK011P, Cytoskeleton, Inc). A known tubulin polymerization enhancer (paclitaxel) and a known inhibitor of tubulin-dependent GTP hydrolysis (vinblastine) were tested as reference compounds in the same assay conditions. The concentration of 1 (20 μM) in this assay is much higher than that in the FtsZ polymerization assay, to make sure whether this compound has an effect on tubulin or not.

L929 mouse fibroblasts cells and HK-2 renal epithelial cells were removed from the sterile cell culture flasks with trypsin and neutralized with fetal bovine serum. After washing with phosphate buffered saline and centrifugation, cells were re-suspended in complete cell culture medium and the concentration was adjusted at approximately 1 × 10⁵ cells mL^-1. Cells seeded in the 96 wells microtiter plates for 24 h were used for the evaluation of the tested compounds. 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) was purchased from Promega (Madison, WI, USA). For the MTS assay, cells (5,000 per well) were seeded into 96-well plates. After treatment with compounds of different concentrations for 24 h, cells were added with MTS at a final concentration of 0.3 mg/mL, followed by incubation for another 2 h. The optical density (OD) of each well was determined at 490 nm (background subtraction at 690 nm) by a SpectraMax 340 microplate reader (Molecular Devices, Sunnyvale, CA, USA). The growth inhibitory ratio was calculated as follows: Growth inhibitory ratio = (A_control - A_sample)/A_control (where A is the OD value per well).

To determine the growth curves of this assay, a single colony of B. subtilis was inoculated in 10 mL LB or 10^-2 LB at 37°C for 24 h with shaking. The culture was harvested by centrifuging at 13,000 rpm for 5 min at 4°C, and the cells were diluted with LB or 10^-2 LB to 0.5 McFarland standard (approximate cell density is ∼10⁸ CFU.mL^-1). Aliquot of 10 μL 0.5 McFarland standard cultures were diluted into 10 mL LB or 10^-2 LB containing various concentrations of compound 1, respectively. Cultures were incubated at 37°C with shaking. At appropriate time intervals, 100 μL samples were taken from the culture solution for serial dilution in 900 μL of the cultured medium broth and 100 μL from three dilutions were spread over MH agar. Cell counts (CFU.ml^-1) were enumerated after the plates incubated at 37°C for 18 h. For the morphology observation, B. subtilis cells were cultured in LB or 10^-2 LB for 24 h without or with 0.75 μg/mL of 1. Then 20 μL culture samples were transferred to poly-L-lysine-coated slides for microscopy. Cell morphology was assessed by phase-contrast light microscopy.

Refer to the Scheme S1 and Figures S3–S6 in the Electronic Supplementary Material.

FtsZ targeting compounds can disrupt the cell division function of FtsZ protein, leading rod-shaped bacteria into cell elongation. Because most of the reported FtsZ inhibitors do not have or only have weak inhibitory effect against the Gram-negative strains (Haranahalli et al., 2016; Hurley et al., 2016), we therefore initiated the screen with B. subtilis 168 in the presence of a small compound library including 40 natural products, 20 natural product derivatives and 140 general compounds which are either commercially available or synthesized in our laboratory. The screen (Stokes et al., 2005; Li et al., 2015) generated one potential hit, 2-((E)-4-hydroxystyryl)-1-methyl-4-((Z)-(3-methylbenzo[d]thiazol-2(3H)-ylidene)methyl)quinolin-1-ium iodide (compound 1). As shown in Figure 2B, 1 is able to increase the cell length of B. subtilis significantly at low concentration (1 μg/mL), as compared to untreated cells (Figure 2A), suggesting a mechanism of antibacterial-induced cell filamentation.

We further tested the antibacterial activity of 1 against an extended panel of clinically relevant bacterial strains, including antibiotic-resistant strains. Methicillin and ceftazidime were tested under the same assay conditions as a reference compound. The results showed that 1 has potent and broad spectrum antibacterial activity against most of the Gram-positive and -negative strains which were tested, as well as all strains and species of Staphylococci that were tested [minimal inhibitory concentration (MIC) in the range of 0.75–3.0 μg/mL], including nine MRSA strains (Table 1). 1 can inhibit the growth of antibiotic-susceptible and antibiotic-resistant S. aureus strains with MIC values of 1.5–3 μg/mL, revealed that 1 have a better anti-S. aureus effect than methicillin and ceftazidime (Table 1). The potency of 1 was more than 100 times greater than that of methicillin against most of the MRSA strains. The growth of vancomycin-susceptible and vancomycin-resistant E. faecium were inhibited with MIC values of 0.75–1.5 μg/mL. For Gram-negative strains, 1 strongly inhibited the growth of E. coli and its drug resistant strain, an NDM-1 carbapenemase isolate that is resistant to many of the major classes of antibiotics, with MIC values of 3 μg/mL. The growth of clinical isolated strains, Streptococcus pyogenes and S. agalactiae were inhibited with MIC values of 1.5 μg/mL. M. catarrhalis ATCC 25240 and P. aeruginosa ATCC BAA-2108 were inhibited with MIC values of 3 μg/mL and 6 μg/mL, respectively. In addition, 1 also showed a moderate antibacterial activity against Klebsiella pneumoniae and Acinetobacter baumannii strains with MIC values of 48 μg/mL, which are close to that of ceftazidime (Table 1).

Viable counts for the determination of killing curves were performed as previously described to investigate whether the antibacterial activity of 1 is bactericidal or not (Wikler et al., 2009). Sample killing curves resulting from 1 against S. aureus ATCC 29213 and E. coli ATCC 25922 are presented in Figure 3. The control showed no reduction in the counts of CFU from control inoculum. Figure 3A showed that 1 at 1× MIC concentration caused a reduction of 1 × 10² CFU mL^-1 for S. aureus in 4 h and to below the lowest detectable limit (10³ CFU mL^-1) in 24 h. In the E. coli bacterial survival assay, 4× MIC of 1 can rapidly reduce the viable counts below the lowest detectable limit after 4 h incubation, and the counts at MIC concentration were maintained under the lowest detectable limit for over 24 h (Figure 3B). The results revealed that the antibacterial activity was consistent with a bactericidal mode of action.

It is interesting to note that similar cell filamentation phenomena were also found with FtsZ inhibitors from different chemotypes, such as 9-phenoxyalkyl substituted berberine derivatives and PC190723 (Haydon et al., 2008; Sun et al., 2014), suggesting that 1-induced cell elongation may be attributed to the 1-FtsZ interaction. To confirm whether this compound is able to disturb FtsZ activity in vitro, S. aureus FtsZ and E. coli FtsZ were cloned, overexpressed, and purified, and some biochemical assays were employed. As a first step to validate that FtsZ is the target of 1, we assessed the impact of 1 on the polymerization dynamics of FtsZ using 90° light scattering in a thermostatically (37°C) controlled fluorescence spectrometer in which changes in FtsZ polymerization are reflected by corresponding changes in absorbance at 600 nm (Chan et al., 2015). Figure 4A shows the time-dependent polymerization profiles of SaFtsZ in the absence and presence of 1 at a concentration range from 1 to 6 μg/mL. The similar phenomena can be observed in the time-dependent polymerization profiles of EcFtsZ with 1 (Supplementary Figure S1). These results prove that 1 stimulates FtsZ polymerization in a concentration-dependent manner, which is similar to that of some FtsZ-targeting compounds (Andreu et al., 2010; Kaul et al., 2012, 2013b, 2015; Kelley et al., 2012). Methicillin is also included in the assay as a non-FtsZ-targeting control antibiotic and, as expected, it does not affect FtsZ polymerization. The effect of 1 on FtsZ polymer formation was also visualized using transmission electron microscopy (TEM) (Kaul et al., 2012). It was found that the size and thickness of the FtsZ polymers and the bundling of the FtsZ protofilaments were drastically increased with 1 (Figures 4B,C). At 3 μg/mL of 1, the thickness of FtsZ protofilaments was found to have increased from ∼10 nm to ∼90 nm (Figure 4C). We further tested the inhibitory effect of 1 on the GTPase activity of FtsZ. The effects of various concentrations of 1 on the GTPase activity of SaFtsZ or EcFtsZ were estimated after 30 min of FtsZ assembly. The results shown that 1 decreased the rate of GTP hydrolysis of FtsZ in a concentration-dependent fashion (Figure 4D and Supplementary Figure S2). For example, 1, 4, and 8 μg/mL of 1 inhibited GTPase activity of SaFtsZ by 14, 38, and 58%, respectively (Figure 4D). The results indicate that 1 inhibits the rate of GTP hydrolysis of FtsZ may due to modulating the assembly of FtsZ. Apart from polymerization and GTPase assays, the effects of 1 on the secondary structures of FtsZ were determined by monitoring far-UV CD spectra of FtsZ (Figure 4E). The results revealed that 1 can significantly perturb the secondary structures of FtsZ. By analysis of the CD spectra using Yang’s reference, it indicated that the secondary structure of FtsZ contained ∼31.5% α-helix, ∼21.1% β-sheet, and ∼47.4% other structures. When using 6 μg/mL of 1, α-helix increased to ∼37.6% and β-sheet decreased to 14.2%. These confirmation changes may also be the cause of disruption of FtsZ function.

In addition to in vitro study, on-target effect of 1 in live bacteria is also important. A green fluorescent protein (GFP)-FtsZ constructed in B. subtilis 168 was used to monitor the impact of the compound on the formation of FtsZ Z-rings (Stokes et al., 2005). We treated the bacteria with DMSO (solvent control), or 1 and examined the bacteria using fluorescence microscopy. In the absence of compound, fluorescent foci corresponding to Z-rings are evident at midcell (Figure 5A). In contrast, the 1-treated bacteria lack these midcell foci, and FtsZ was found to distribute as discrete and punctate foci throughout the elongated cell, indicating mislocalization of the FtsZ protein (Figure 5B). As 1 enhances FtsZ polymerization, it is likely that those discrete and punctate fluorescent foci in 1-treated bacteria are multiple non-functional FtsZ polymeric structures. Observation of such disorganized FtsZ in elongated bacterial cells is typical of agents that disrupt FtsZ polymerization (Hurley et al., 2016).

A 2.01 Å crystal structure of S. aureus FtsZ apo-form (Tan et al., 2012) was used to identify a potential binding site for 1. The highest docking score positioned the ligands bind near the T7-loop and H7-helix. Since the binding site is a relatively narrow cleft delimited by the H7-helix, the T7-loop, and a four-stranded β-sheet, the substrate requires some degree of planarity in its structure to fit in. The docking result suggested a large number of favorable hydrophobic interactions between the molecule and the side chains of Leu200, Val203, Leu209, Met226, Ile228, Leu261, Val297, Leu302, Val307, and Ile311. Moreover, the hydroxyl group of 1 was predicted to participate in two hydrogen bonds with the backbone carbonyl of Val203 and the backbone amide of Leu209 (Figure 6).

Tubulin is the closest mammalian functional homolog to bacterial FtsZ. We therefore sought to determine whether 1 would exert a similar effect on mammalian tubulin because our results (Figure 4) indicated that the compound is a potent stimulator of FtsZ polymerization. In the presence of vinblastine (30 μM), a tubulin inhibitor, the polymerization of mammalian tubulin was completely inhibited. On the opposite, the fluorescence intensity was significantly increased in the presence of paclitaxel (20 μM), confirming that the rate of polymerization was significantly enhanced. 1 (20 μM) showed similar results to a control experiment with 2% DMSO-treated mammalian tubulin, indicating that 1 is not a stimulator/inhibitor of tubulin polymerization (Figure 7).

To probe for any potential mammalian cytotoxicity, we used a MTS tetrazolium assay to assess the cytotoxicity of 1 against two mammalian cell lines (L929, HK-2). 1 was found to be minimally toxic to these cell types, with 50% inhibitory concentrations (IC₅₀s) higher than 50 μg/mL (Table 2), which are much higher than the MICs (0.75–3.0 μg/mL) of 1 against bacterial strains, indicating no significant toxicity toward normal mammalian cells.

Bacteria change their size and growth rate when cultivated in nutrient poor medium (Chien et al., 2012; Mandal et al., 2013). It was reported that bacteria were more resistant to beta-lactam antibiotics in a rich medium than in a relatively poor medium (Wang et al., 2015). To further investigated the antimicrobial activity of Compound 1 under different nutrient conditions, we investigated the growth curves of B. subtilis 168 cells in LB or 10^-2 LB medium with different concentrations of 1. And also observed the morphology of B. subtilis growing in the two different nutrient conditions. The results indicated that B. subtilis cells cultured in a nutrient rich medium have a faster growth rate and longer morphology than that cultured in the nutrient poor medium (Figures 8, 9). Figure 8A showed that 1 at 1.5 μg/mL decreased the cell numbers of B. subtilis below the lowest detectable limit (10³ CFU mL^-1) after incubation for 6 h; however, using 0.75 μg/mL of 1 cannot inhibit the growth of B. subtilis in LB. On the other hand, when B. subtilis cells were cultured in 10^-2 LB, using 1.5 μg/mL of 1 can decreased the cell numbers to the lowest detectable limit after incubation for 4 h, and also using 0.75 μg/mL of 1 showed inhibitory effect on the growth of the cells (Figure 8B). These results indicated that 1 has a better antibacterial activity against bacteria cultured in the nutrient poor condition than that in the nutrient rich condition. These results may due to rapid growth cells with bigger size having much more FtsZ proteins than the slow growing cells in the nutrient poor medium (Chien et al., 2012). In addition, using 0.75 μg/mL of 1 can inhibit the cell division of B. subtilis in both of the culture media. And the phenomena of cell elongation were found sharper in LB than that of in 10^-2 LB (Figures 9C,D).