Screening and Development of New Inhibitors of FtsZ from M. Tuberculosis

A variety of commercial analogs and a newer series of Sulindac derivatives were screened for inhibition of M. tuberculosis (Mtb) in vitro and specifically as inhibitors of the essential mycobacterial tubulin homolog, FtsZ. Due to the ease of preparing diverse analogs and a favorable in vivo pharmacokinetic and toxicity profile of a representative analog, the Sulindac scaffold may be useful for further development against Mtb with respect to in vitro bacterial growth inhibition and selective activity for Mtb FtsZ versus mammalian tubulin. Further discovery efforts will require separating reported mammalian cell activity from both antibacterial activity and inhibition of Mtb FtsZ. Modeling studies suggest that these analogs bind in a specific region of the Mtb FtsZ polymer that differs from human tubulin and, in combination with a pharmacophore model presented herein, future hybrid analogs of the reported active molecules that more efficiently bind in this pocket may improve antibacterial activity while improving other drug characteristics.


Introduction
Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), is one of the most prevalent infectious diseases worldwide. A recent report by the World Health Organization (WHO) reveals that in 2014 there were an estimated 9.6 million new cases of TB (12% co-infected with HIV) and 1.5 million people died from the disease, including 1.1 million deaths among HIVnegative individuals and 400,000 among people who were HIV-positive. [1] Moreover, the increasing development of multi-drug resistant TB (MDR-TB) strains, forms of TB that do not respond to standard front line regimens, is a serious burden on public health systems throughout the world. More recently, extensively drug resistant tuberculosis (XDR-TB) that is also resistant to second line drugs has emerged in numerous areas. These stark realities emphasize the acute need for constant reinvention of the anti-TB drug pipeline targeting novel, alternative drug pathways. [2][3] New targets for tuberculosis are regularly being identified from existing genomic data and new screening techniques. [3] The bacterial tubulin homolog, FtsZ (Filamenting temperaturesensitive protein Z), is essential for bacterial cell division and is an ideal target for novel antimicrobials. In recent years, FtsZ has received considerable attention for the discovery of novel antibacterial and anti-TB drugs. [4][5][6][7][8][9][10][11][12][13][14][15][16][17] In our search for new active scaffolds against Mtb FtsZ, a variety of small molecules reported to have antibacterial activity, and, furthermore, some of which were considered to inhibit Mtb or other bacterial FtsZs were acquired and screened for both antitubercular activity and Mtb FtsZ inhibition. A consistent set of antibacterial activity data in parallel with FtsZ screening results should be useful to prioritize active scaffolds for new analog optimization. Furthermore, the potent combination of a new crystal structure and these activity data will allow advancement of robust consensus binding models that should help medicinal chemists enhance selective activity against the bacterial protein target and whole bacteria while potentially minimizing off-target effects against the direct mammalian homolog, tubulin, as well as reducing mammalian toxicity through other off-target activities. Beyond the aforementioned antibacterial/FtsZ actives or related compounds, we were particularly intrigued by the reported similarities of certain non-steroidal anti-inflammatory drugs (NSAIDs), e.g. Indomethacin and Sulindac analogs, to the known tubulin polymerization inhibitor Colchicine. [18,19] Colchicine has been reported to be one of the few known tubulin inhibitors that demonstrates activity against Mtb FtsZ. [15] Sulindac belongs to this chemically diverse group and, importantly, is not overtly toxic in vivo but shows clinical efficacy for longer term treatment regimens in cancer chemoprevention. [20][21][22][23] The NSAIDs are excellent pharmacophores showing good activity through animal models and in the clinic for numerous indications. As part of an ongoing program to study the chemical biology of interesting NSAID scaffolds such as Sulindac, we have investigated a variety of analogs and their on-target (COX-1 and 2) and off-target (e.g. cell cytotoxicity, PDE5, PDE10A) activities. [24][25] Among the interesting and atypical activities of the NSAIDs, certain known drugs (e.g. Ibuprofen, Aspirin) have been reported to show antibacterial activity. [26][27][28][29][30] An indomethacine analog closely related to sulindac sulfide amide (SSA) has been reported to inhibit tubulin polymerization in vitro in a dose response manner. [18] Hence, we added this early lead Sulindac analog to our initial anti-TB/FtsZ assays and confirmed that it is a modest potency inhibitor of Mtb FtsZ while showing no inhibition of human tubulin at 100 μM concentration. The activity of this initial lead warrants the exploration of new sulindac analog series against FtsZ. Herein, we report the screening of a number of acquired and synthesized samples and a lead Sulindac analog available in our labs against FtsZ from M. tuberculosis. These compounds advanced in a series of assays depending on activity that include Mtb FtsZ, Mtb H 37 Ra, MAC NJ211 and/or Mtb H 37 Rv, tubulin polymerization, and in a preliminary cell cytotoxicity assay against BJ cells, an immortalized normal human foreskin fibroblast cell line. In addition to the presented structure-activity development of the Sulindac scaffold, we also followed up a potent and previously reported screening hit, Zantrin Z2, which showed potent activity in our preliminary screens (see S1 Appendix in Supporting Information for results).
abides by the USDA Animal Welfare Act and the Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Animal care and euthanasia
The CSU animal assurance welfare number is A3572-01 under file with NIH. All animals are cared for by the Colorado State Lab Animal Resources, headed by two experienced veterinarians and a large number of support staff. The mice are observed twice daily by our Research Associates and Lab Animal Resources (LAR) personnel. A log is kept recording any untoward behavior. Sick animals are reported to the Staff Veterinarian on a morbidity form. Investigators receive a copy of the morbidity form with initial physical exam findings, diagnostics, and possible diagnoses. The LAR at Colorado State University has procedures in place to control animal pain. Specific health and comfort parameters (activity and temperament, feeding behavior, appearance) are used to monitor pain and severity of disease of M. tuberculosis infected mice, and these conditions are monitored daily to determine the condition of the mice. When issues are evident (distress, severe physiologic changes, weight loss), mice are required to be euthanized within the day. The method used for euthanasia of mice is carbon dioxide inhalation. This method, approved by the CSU Regulatory Compliance Office uses either tank CO2 or house CO2 This method is approved by the AVMA Panel of Euthanasia, and was chosen for its reliability, ease of use, and minimization of distress to animals.
Preparation of α-methyl Sulindac sulfide amide began with esterification of commercially available Sulindac sulfide in the presence of MeOH/thionyl chloride to provide Sulindac sulfide methyl ester in quantitative yield. The introduction of a methyl group at the α-position of the ester group was carried out using LDA and CH 3 I to afford α-methyl Sulindac sulfide ester in 93% yield as a mixture of enantiomers. [34] α-Methyl Sulindac sulfide ester was hydrolyzed to give the corresponding acid in excellent yield. Finally, the coupling of α-methyl Sulindac sulfide with various amines in the presence of HATU [33] as the coupling agent afforded compounds 44-55 in good yields (Fig 2). Typically, when the racemic mixture of the acid was treated with a chiral amine, an inseparable mixture of diastereomeric amides was formed.
N,N'-Dimethylaminoethyl amides of Sulindac sulfide and α-methyl Sulindac sulfide (23,48) were treated with Boc-L-Valine under the same coupling reaction conditions described in Fig 1 and Fig 2 followed by treatment with acid for the removal of the Boc protecting group to produce 56-57 in good yields (Fig 3).
The synthetic strategy used to prepare the target compounds 58-60 is outlined in Fig 4. Synthesis began with the esterification of Sulindac sulfide to form its methyl ester. The ester was then converted to its corresponding aldehyde by treating with DIBALH. Finally, reductive amination of the aldehyde with various substituted amines afforded compounds 58-60 in moderate yields. [35] Compounds 61-63 were prepared by methods given above (Fig 1). [31,32] Further synthetic experimental details are provided in S2 Appendix under Supporting information.
Compounds. All compounds were purchased commercially (Aldrich, ChemBridge, or CalBiochem) and used as provided excepting the Sulindac analogs that were prepared as

Biological studies
In vitro screen versus Mtb H 37 Ra and MAC NJ211. The initial screening set and selected other compounds were tested for their inhibitory activity against Mtb H 37 Ra (ATCC 25177) and MAC NJ211 strains. Screening was performed at 1.28 and 12.8 μg/mL in Middlebrook 7H9 broth supplemented with 0.2% glycerol and ADC enrichment using a colorimetric (Alamar blue) microdilution broth assay by reported methods. [15] Active compounds ( 12.8 μg/ mL) were re-tested using twofold dilutions to obtain the actual MIC 99 . The MIC 99 is defined as the lowest drug concentration that completely inhibited growth.
In vitro qHTS assay against Mtb H 37 Rv and cytotoxicity evaluations. The Sulindac set was screened against Mtb H 37 Rv to determine the IC 90 (the concentration that inhibits bacterial growth by 90%) by reported methods. [36] These compounds were also counter screened at St. Jude Children's Research Hospital for a comparative toxicity assessment against an established mammalian cell line relative to the in vitro Mtb H 37 Rv results. BJ, a normal human foreskin fibroblast cell line was purchased from the American Type Culture Collection, ATCC, Manassas, VA, USA and was cultured according to recommendations. [37] Cell culture medium was also purchased from ATCC. Cells were routinely tested for mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza). Approximately 1,000 BJ exponentially growing cells were plated per well (30 μL) in white polystyrene flat bottom sterile 384-well tissue culture treated plates (Corning, Tewksbury, MA, USA), and incubated overnight at 37°C in a humidified 5% CO 2 incubator. Compound solutions (DMSO) were pintransferred (V&P Scientific, San Diego, CA, USA) the following day. Cytotoxicity was determined following a 72-hr incubation of BJ cells with each compound at a concentration of 10 μM using Promega Cell Titer Glo Reagent according to the manufacturer's recommendations. Luminescence was measured on an Envision plate reader (Perkin-Elmer).
Inhibition of Mtb FtsZ polymerization in vitro. Selected available compounds that were active in vitro against Mtb H 37 Rv were examined for their ability to inhibit the hypothetical target, Mtb FtsZ, and its mammalian homolog tubulin using methods described previously. [15] Compounds were initially screened at 100 μM in duplicate. Those that showed <20% activity at this concentration were deemed inactive. All other samples (! 20% inhibition) were screened in a dose response format to obtain IC 50 values. Any compounds that were insoluble in the assay medium at ! 25 μM did not advance. Compounds considered active in the Mtb FtsZ assay were further screened in a tubulin polymerization assay as reported. [15] In vitro Z-ring inhibition Studies: Bacterial strains and growth conditions. Construction of M. tuberculosis 41 (Mtb 41) FtsZ reporter strain was previously described. [38] It is a Mtb H 37 Ra derivative strain where the chromosomal ftsZ is deleted by homologous recombination and contains a copy of Pami::ftsZ-gfp. The ftsZ-gfp in this strain is expressed from the amidase promoter (Pami) and the reporter strain requires 0.2% acetamide for growth. [38] Mtb 41 was grown in Middlebrook 7H9 broth supplemented with oleic acid, albumin, dextrose, sodium chloride and 0.2% acetamide. In some experiments, actively growing cultures of Mtb 41 were exposed to the small-molecule inhibitors at 75 μM (ca. the H 37 Ra MIC 99 ) and growth was followed for several days after exposure by monitoring the absorbance at 600 nm.
Microscopy. M. tuberculosis 41 strain was grown as described in the absence or presence of the Sulindac analogs for 72 hours, at indicated time points samples were harvested by centrifugation, washed in phosphate-buffered saline, fixed in 1% paraformaldehyde, and stored at 4°C until microscopy. Bacteria were examined by bright-field and fluorescence microscopy with a Nikon E600 microscope equipped with a 100X Nikon Plan Fluor oil immersion objective with a numerical aperture of 1.4 and a standard FITC filter set from Chroma. Images were acquired with a Photometrics Coolsnap ES camera and MetaMorph 6.2 imaging software (Universal Imaging Corporation). Images were optimized with Adobe Photoshop CS5.

Computational methods
The Mtb FtsZ crystal structure (PDB code 1RLU) determined at 2.08 Å resolution was refined applying Prime preparation and refinement tools of the Protein preparation wizard as implemented in Schrödinger software. After the addition of hydrogens and detection of disulfide bonds the structure was optimized applying default parameters of the Impref utility using the OPLS2001 force field. The maximum allowed root-mean-square deviation between the refined structure and the input crystal structure was 0.3. Ligand structures were prepared using the LigPrep utility at pH 7.4. Ligands were docked into the interdomain cleft site using Induced Fit docking protocol in XP precision mode. Default parameters were applied except for setting the site enclosing box size to 33 Å. The Induced Fit docking method combines Glide docking with Prime structural refinement tools in the Schrödinger software to account for the flexibility of protein side chains within 5 Å of the ligand during docking.

Results and Discussion
A structurally diverse compound set selected for screening A diverse set of compounds and natural products was selected for screening against Mtb FtsZ and its mammalian homolog tubulin. In order to assess whole bacterial activity, MICs against the non-pathogenic mycobacterial strains Mtb H 37 Ra and MAC NJ211 were determined and inhibitory activities evaluated against a potential target, FtsZ, that is highly homologous between Mtb strains and essential in mycobacteria. Selected results and structures are given in Table 1 and Fig 5, respectively.
Colchicine was included as a known, albeit modest, inhibitor of Mtb FtsZ. [15] Rhein, Quercetin, Myrcetin, Tetracycline, and Minocycline (Aldrich Chemical Company) contain planar aromatic systems and bear resemblance to Viriditoxin, [39] a reported natural product that is active against E. coli FtsZ. Rhein and the isoflavones have also been reported to have antibacterial/antitubercular activity. [40] The Zantrins (Z1-5) were first reported by Margalit as inhibitors of bacterial FtsZ GTPase activity. [11] Z1-Z4 purchased from ChemBridge were included as a test to determine how well the reported GTPase activity correlates with inhibition of FtsZ polymerization as there is an apparent kinetic disconnect between GTPase effects and inhibition of polymerization. [41] In some respects, Zantrin Z2 is similar to the aromatic cationic compounds Sanguinarine and Pseudojatorrizine. Sanguinarine has been reported to show antibacterial activity and to inhibit E. coli FtsZ assembly. [42] Tamoxifen is a known antiestrogen drug that has effects on human tubulin and is reported to show antitubercular activity in a screen of the Prestwick collection of natural products and FDA approved drugs (see http:// pubchem.ncbi.nlm.nih.gov). [43] Compound 12, a 4-aminofurazan analog available from ChemBridge, is a reported inhibitor of E. coli FtsZ GTPase activity as well as protein polymerization in a standard light scattering assay. [44] Compound 12 is also reported to alter Z-ring assembly in whole bacterial cells and to cause bacterial filamentation in E. coli cells. [44] Totarol is a reported natural product that inhibits proliferation of a variety of bacteria and is reported to have activity against Mtb FtsZ in vitro as well as causing bacterial filamentation although recently reported as a promiscuous aggregator. [9,45] The taxanes are known drugs that alter tubulin polymerization dynamics, and analogs have been reported to show inhibition of Mtb growth in vitro and are reported to show specificity for Mtb FtsZ. [8] Compound AG-825 is a known inhibitor of HER-2 available from Calbiochem, and the compound bears striking resemblance to a reported class of bacterial FtsZ inhibitors out of the Stokes laboratory. [46]  New analog series based on the Sulindac scaffold We found that SSA (18), an early sulindac analog shows modest inhibition of Mtb FtsZ polymerization, inhibits bacterial growth but has no effect on human tubulin at 100 μM (Table 1). Further we explored Mtb FtsZ activities of several series of compounds designed based on the Sulindac scaffold (as described under Methods). The scope of these studies was to probe the effect of targeted modifications of this scaffold, the development of structure-activity relationships, and the identification of analogs with the most potent inhibition of FtsZ polymerization without affecting tubulin polymerization, coupled with potent whole cell antitubercular activity. Results of these studies are presented herein. Whole cell antitubercular and cytotoxicity results. Fig 6 presents the Mtb H 37 Rv and BJ cell line cytotoxicity data of Sulindac sulfide amide derivatives 18-37. Compounds 18-23 with basic acyclic acetamide linker at the C-3 position showed significant activity against the virulent strain H 37 Rv. Among these six compounds, compound 22 with two n-butyl groups at the terminal nitrogen was found to be more active than corresponding methyl and ethyl substituted compounds. It is notable from the screening data of compounds 20 and 21 that the carbon chain length between the acetamide and the terminal nitrogen has marginal influence on the activity. Either an additional substituent at the acetamide nitrogen or removal of a substituent from the terminal nitrogen did not improve the activity (19 and 23) dramatically. We also  We then expanded our study to include a benzothiazol-2-ylmethylene group at the C-1 position (38-43, Fig 7) while varying the amide. These analogs demonstrated significant activity against Mtb H 37 Rv that was similar to their 4-methylthiobenzylidene analogs. Fig 8 lists the Mtb H 37 Rv and BJ cell data of α-methyl Sulindac sulfide amide derivatives 44-55. Incorporation of a methyl group at the α-position of the amide group did not cause significant alterations in activity. The N,N-dimethylaminoethyl analog 44 was 2-fold less active than its desmethyl analog 18. The N,N'-dimethylaminoethyl acetamide derivative 48, in contrast, had 4-fold more activity than 23. In the case of the cyclic basic amide series, the α-methyl analogs displayed similar or slightly better activity than their corresponding desmethyl analogs.
We also prepared a small set of extended Sulindac sulfide amide derivatives 56-57 from 23 and 48 by coupling with L-Valine. The α-methyl analog 57 displayed an IC 90 of 0.74 μM against H 37 Rv, 6-fold more active than its desmethyl analog 56 (Fig 9).
Analogs exploring the amine linkage at the C-3 position are presented in Fig 10. Compounds 58-59 with a benzyl or furan-2-ylmethyl linkage at the aminoethyl group at the C-3 position displayed excellent activity as compared to their inactive amide counterparts (34 and 36). The piperidinylethyl analog 60 also showed 1.6-fold more activity than its corresponding amide analog 24. Finally, we prepared a small set of related E-conformation compounds 61-63[31, 32] to further test activity against whole bacteria and Mtb FtsZ. These compounds and their activity are reported in Fig 11. Comparison of Mtb H 37 Rv activity between compounds 18 and 61 suggests no significant difference between the E and Z forms of this compound.
Activity of Sulindac Analogs against Mtb FtsZ and Tubulin. Selected compounds that were available in sufficient quantities were examined for their ability to inhibit both Mtb FtsZ and its mammalian homolog tubulin. The MIC for Mtb H 37 Ra was also determined for this set. Ten compounds demonstrated significant activity against Mtb FtsZ while not affecting tubulin polymerization and these are reported in Table 2. It is notable that the H 37 Ra and H 37 Rv (MIC 99 versus IC 90 ) data are relatively well correlated for these compounds.
To examine if the Sulindac analogs would also affect FtsZ polymerization in vivo, we utilized a Mtb FtsZ reporter strain, Mtb 41, wherein the FtsZ tagged with GFP serves as the sole source of FtsZ. In this strain the chromosomal ftsZ gene is deleted and a copy of ftsZ-gfp expressed from the inducible amidase promoter is integrated at the attB site. [38] Exponential cultures of Mtb 41 were exposed to 62 and 63 (75 μM) for 48 hours and the bacteria examined by brightfield and fluorescence microscopy (Fig 12 -data are given for 63). Exposure to drug led to a clear increase in bacterial cell length (Fig 12, compare panel a to c). In addition, distinct midcell FtsZ-GFP rings seen in control cultures were absent in treated cultures (Fig 12, compare panel  b with d). Instead, the compound treated bacteria showed weak FtsZ-GFP fluorescent spots or aggregates (Fig 12, panel d). Exposure to 62 and 63 also led to a similar growth inhibition of M. tuberculosis 41 (Fig 12B). Since FtsZ is essential for growth and viability of Mtb, [38] 62 and 63 inhibition of FtsZ ring assembly likely led to growth inhibition of Mtb 41.
Modeling Mtb FtsZ inhibitor binding. Sulindac sulfide is a known human COX-1/COX-2 inhibitor. Replacement of its carboxylate group with positively charged moieties such as N, N-dimethylamino ethyl amide in compound 18 results in analogs that have greatly reduced activity at COX-1/COX-2. [24] Tolerance and pharmacokinetic studies have shown that compound 18 shows a greatly improved toxicity profile compared to Sulindac sulfide. [24] While Sulindac sulfide is tolerated in mice up to 50 mg/kg over 20 days, compound 18 shows tolerance up to 300 mg/kg over 17 days. Also, compound 18 maintains~4 μM plasma levels for 8 hours after gastric gavage at a dosage of 200 mg/kg. [24] Compounds presented in Table 2 are analogs of compound 18 containing the core substructure of Sulindac and positively charged amine moieties and therefore may be expected to have favorable pharmacokinetic and toxicity profiles as well. While Mtb FtsZ inhibitory activities of the presented Sulindac analogs are moderate, compounds from these series may serve as starting leads for the design of further chemical modifications to achieve better FtsZ inhibitory potencies. To provide insight into the possible binding site of the presented compounds we modeled representative analogs and known inhibitors from Table 1 into the crystal structure of Mtb FtsZ. Small molecule inhibitors may bind FtsZ in the GTP binding site or the interdomain cleft region (Fig 13), which is equivalent to the Taxol binding region in tubulin and also contains the co-crystallized PC190723 in Staph. aureus FtsZ (PDB code 4DXD). These two sites have been targeted in ligand binding studies of FtsZ inhibitor libraries. [47] Another pocket near loop T7 has also been investigated recently in case of a B. subtilis FtsZ inhibitor, BT-benzo-29.
[48] Inhibitory effects of BT-benzo-29 on the assembly, GTPase activity of single mutant FtsZ constructs compared to wild type provided evidence that residues L272, V275 play an important role in the inhibition of FtsZ by this ligand. The corresponding residues in the Mtb FtsZ crystal structure (L269, I272, respectively) along with loop T7 define a new ligand binding site that is distinct from the interdomain cleft site. Further, docking studies suggested that the FtsZ inhibitors Scopoletin and CCR-11 may bind the same pocket. [49,50] In Mtb FtsZ the corresponding binding site is smaller compared to B. subtilis and we could not successfully dock    Table 2 containing a variety of non-polar groups are accommodated in poses similar to that of 24. The varying non-polar group in this series forms analogous, non-polar interactions with Ile295 while their positively charged amine is salt bridging with Asp297. The similarity of the binding poses and analogous non-polar interactions with Ile295 are consistent with our finding that these substitutions resulted in only a modest variation of Mtb FtsZ polymerization inhibition, as IC 50 s range between 22.9 and 44.6 μM within this series ( Table 2).
Colchicine is a low potency inhibitor of Mtb FtsZ assembly at 104 μM and inhibits tubulin assembly at 6.5 μM. [15] While the Colchicine binding pocket of tubulin has no structurally analogous site in Mtb FtsZ, our docking results predict that Colchicine can also be accommodated in the interdomain cleft site of FtsZ. Similarities between the docked pose of compound 24 and Colchicine are illustrated in Fig 14B and 14C. Colchicine forms several favorable hydrogen bonding interactions with backbone atoms of Asp297, Ser298, Gly31, Lys33, and the sidechain of Lys33. Binding poses of Colchicine and compound 24 displayed simultaneously in Fig  14C suggest that the indene of compound 24 and the Colchicine core ring structure occupy overlapping space while the carbonyl and O-Me groups of the cyclohexyl moiety of Colchicine are in proximity to the amide carbonyl of analog 24, both ligands participating in similar hydrogen bonding interactions with the D297 backbone.
Evidence supporting the interaction between inhibitors and residues in the interdomain cleft binding site region has been presented in the case of several inhibitors. Table 3 lists examples, showing implicated amino acids and the corresponding residues as mapped onto the Mtb FtsZ crystal structure (PDB code 1RLU). The E. coli FtsZ residues R33, D187 and E305 (corresponding Mtb G31, E185, E302, respectively) are involved in the binding of doxorubicin (Table 3). While G31 is in the interdomain cleft site, the other two Mtb FtsZ residues are located near the boundary of the interdomain cleft site region on the side opposite to loop T7. Ligands bound in the interdomain cleft side may possibly extend into this region near the Mtb residues E185, E302. Interestingly, a number of residues within the interdomain cleft binding site have been implicated in functional motions that affect FtsZ polymerization and activity. [51] Out of the known FtsZ inhibitors from Table 1 Table, respectively. The complete coordinates of the model are supplied in S4B Appendix.

Conclusions
The series of compounds initially acquired and screened were selected based on established antibacterial activity and/or reported bacterial FtsZ inhibition, as well as similarity to other reported FtsZ inhibitors. All acquired samples were first screened in house against Mtb FtsZ and Mtb H 37 Ra and MAC NJ211, BSL-2 pathogens. Active samples were then tested against tubulin polymerization for selectivity determination. Table 1 presents results obtained for a number of compounds showing modest antitubercular activity versus Mtb H 37 Ra, coupled with activity against FtsZ in an IC 50 range of about 20-50 μM. Results suggest further optimization strategies for some of these lead scaffolds that might improve FtsZ activity and potentially concurrent antibacterial activity against the bacillus. The Tetracycline (Tetracycline-5 and Minocycline-6) class appears to show interesting activity against FtsZ, but the known primary target of this drug class is protein synthesis inhibition by binding the 30S ribosomal subunit, while the FtsZ activity is likely only an interesting sideline. Among the remaining actives, three in particular stand out for different reasons. First, Zantrin Z2 (compound 8) is approximately ten fold more potent against FtsZ than the range of activity of most other FtsZ actives in Table 1 (20-50 μM). Furthermore, it has the best antibacterial activity in the set with an MIC 99 of 1 μM against Mtb H 37 Ra while showing no activity against tubulin up to 100 μM. Hence, we reasoned that initial results for this compound warranted further evaluation, and a sample was screened in a mammalian cell line toxicity assay (Vero cells) and a rapid PK model available through the NIH TAACF screening program. Unfortunately (see S1 Appendix and Table A in S1 Appendix, in Supporting Information for experimental details and results), Zantrin Z2 was equally toxic against Vero cells as it was active against Mtb H 37 Ra. Furthermore, at the dose of 300 mg/kg in a rapid PK mouse model, the sample showed no ill effects to the mice at a single dose but also showed no bioavailability by oral gavage, suggesting that it is likely a poor development candidate.
Two other compounds (11 and 18) stood out from the remainder of the active samples as purely synthetic molecules showing reasonable FtsZ activity and significant activity against Mtb H 37 Ra. Both compounds have advantages over the relatively more complex and less synthetically tractable natural product scaffolds that showed activity in Table 1. Additionally, Tamoxifen (11) is a clinically used agent suggesting reasonable starting bioavailability for a lead pharmacophore, and compound 18 is a simple Sulindac analog that has been reported as an active compound in vivo in a murine xenograft assay adapted for chemoprevention studies. It is interesting to note that several 1,1,2-triarylethenes akin to Tamoxifen show significant (nM) COX-2 activity, [55] demonstrating a plausible structural link of the NSAID Sulindac to Tamoxifen and a basis for their similar FtsZ inhibition as well as suggesting that certain of the related NSAIDs that effectively bind cyclooxygenases may have off-target FtsZ activity and possibly, depending on substitution, tubulin. In many respects the two scaffolds are similar in that they contain relatively planar aromatic ring systems with an attached basic amino function that could be protonated at physiological pH, suggesting that a positively charged group may be important for FtsZ activity. In fact, both of these features (aromaticity and cationic character) are common features within the screening actives, and are important parts of the phamacophore model developed in this program and discussed in this paper. While both samples appear to be reasonable starting points for further SAR, we chose to focus this work on the Sulindac scaffold due to our considerable experience with that template. In short, we prepared a small series of Sulindac derivatives designed to test several substitution points on the scaffold. New synthetic analogs were screened against Mtb H 37 Rv, and BJ cells (an immortalized human foreskin fibroblast cell line) for a preliminary assessment of overt cytotoxicity. Selected compounds were also tested against Mtb FtsZ, its mammalian homolog tubulin and Mtb H 37 Ra. Several observations can be made based on the modest SAR work. First, as a class, these compounds show good, lead activity against Mtb H 37 Rv in vitro with certain analogs showing IC 90 values below 1.0 μM. Although these compounds showed no activity in the BJ cell toxicity assay at the modest screening cutoff, the Sulindac scaffold and analogs have been reported to have mammalian cell cytotoxicity albeit giving modest selectivity for cancer cell lines over normal cells in vitro. [24,25] Finally, of the Sulindac analog set for which FtsZ activity was measured (Table 2), the H 37 Rv whole cell potency was consistently better than Mtb FtsZ inhibition activity in this series, which might be due to cell wall penetration/transport issues as well as possible activity at other Mtb targets. The difference between whole cell and FtsZ activities may be specific to the core scaffold shared within this series, since for example another scaffold, Zantrin Z2 shows about 10-fold more potent activity in the FtsZ assay than the average range of these analogs. Furthermore, the inhibition in this Sulindac series is in a modest IC 50 range of approximately 23-45 μM in spite of the variety of substitutions, suggesting that we were not varying the structure enough to probe the true extent of the binding site pocket as demonstrated by retrospective docking results of representative analogs of this series.
In order to test that these samples actually had a true effect on FtsZ polymerization under normal bacterial growth conditions, we turned to the GFP-labeled Z-ring assay in Dr. Malini Rajagopalan's laboratory. In a GFP labeled FtsZ reporter strain Mtb 41, [38] Sulindac analogs 62 and 63 showed significant growth inhibition at the MIC 99 determined for those compounds in Mtb H 37 Ra. Furthermore, these growth effects translated into a clear visual inhibition of Zring formation on analysis of the fluorescently labeled protein aggregation and ring formation with time of exposure to both compounds. These results indicate that the Sulindac analogs 62 and 63 can have direct inhibitory effects on Z ring assembly, cell division and growth around the determined MIC 99 in a related strain Mtb H 37 Ra.
Our docking results suggest that Sulindac analogs target an interdomain cleft site which is distinct from the GTP binding site in Mtb FtsZ. The presented model offers rational approaches to modify the scaffold to improve FtsZ binding through exploring this site more extensively while potentially improving Mtb versus host cell selectivity (e.g. against tubulin and off-target enzymes). Based on binding models of a structurally diverse set of known FtsZ inhibitors and their overlapping pharmacophoric features, we developed a composite pharmacophore model that describes shared interactions present in the same active site region. Another potential application is screening known Mtb phenotypic actives that are available on Pub-Chem (http://www.ncbi.nlm.nih.gov/pcsubstance) [17,36] against the pharmacophore model in order to identify other potential FtsZ inhibitor scaffolds with known selective activity versus Mtb H 37 Rv. Coordinates of this pharmacophore model are available (S4 Appendix, Supporting Information) for future applications for the design of chemical modifications of inhibitor scaffolds for the improvement of their binding affinities. Overall, our observations indicate that Sulindac derivatives remain a promising development scaffold for the design of new FtsZ inhibitors against M. tuberculosis. Promising analogs in the various series presented are candidates for future medicinal chemistry optimization efforts and for further validation in followup assays.