Structural Refinement of 2,4-Thiazolidinedione Derivatives as New Anticancer Agents Able to Modulate the BAG3 Protein

The multidomain BAG3 protein is a member of the BAG (Bcl-2-associated athanogene) family of co-chaperones, involved in a wide range of protein–protein interactions crucial for many key cellular pathways, including autophagy, cytoskeletal dynamics, and apoptosis. Basal expression of BAG3 is elevated in several tumor cell lines, where it promotes cell survival signaling and apoptosis resistance through the interaction with many protein partners. In addition, its role as a key player of several hallmarks of cancer, such as metastasis, angiogenesis, autophagy activation, and apoptosis inhibition, has been established. Due to its involvement in malignant transformation, BAG3 has emerged as a potential and effective biological target to control multiple cancer-related signaling pathways. Recently, by using a multidisciplinary approach we reported the first synthetic BAG3 modulator interfering with its BAG domain (BD), based on a 2,4-thiazolidinedione scaffold and endowed with significant anti-proliferative activity. Here, a further in silico-driven selection of a 2,4-thiazolidinedione-based compound was performed. Thanks to a straightforward synthesis, relevant binding affinity for the BAG3BD domain, and attractive biological activities, this novel generation of compounds is of great interest for the development of further BAG3 binders, as well as for the elucidation of the biological roles of this protein in tumors. Specifically, we found compound 6 as a new BAG3 modulator with a relevant antiproliferative effect on two different cancer cell lines (IC50: A375 = 19.36 μM; HeLa = 18.67 μM).


Introduction
The multimodular BAG3 protein (Bcl-2-associated athanogene 3) is a member of the human BAG family of co-chaperones that interacts, through its conserved BAG domain (BD) located at the C-terminus of the protein, with the ATPase (Adenosine Triphosphatase) domain of heat shock protein 70 (Hsp70), modulating a variety of physiological and pathological functions [1][2][3][4]. In addition to BD, BAG3 also includes proline rich (PxxP) and WW modules responsible for its interaction with several signaling factors highly involved in cancer development. Through the so-called M-domain, located between PxxP and WW motifs, BAG3 can also mediate the recruitment of many protein kinases implicated in various signal-transduction pathways. Moreover, the BAG3 role as a key player in cellular macroautophagy, mediated by the two conserved IPV (Ile-Pro-Val) motifs, has been reported [5]. Under physiological conditions, normal cells express only low basal levels of BAG3, while its expression is highly induced under several stressful stimuli, where it is essential for maintaining cellular proteostasis by regulating the two key cellular protein degradation pathways such as proteasomal delivering and autophagy [6]. In addition,

In-Silico Screening
Starting from the encouraging results obtained in our previous work [18], and with the aim of identifying novel 2,4-thiazolidinedione-based BAG3BD modulators, a fast virtual screening campaign was performed. Specifically, we re-evaluated the virtual library of 2,4-thiazolidinedione-based compounds, built following the related chemical route and accounting the commercial availability of the starting reagents (for further details, please see the section Materials and Methods and [18]). This starting set of compounds was filtered selecting those featured esters and amide chemical functions at N-3 that were considered essential structural motifs for BAG3BD interaction, as emerged in our previously identified modulators [18]. Afterward, this subset of molecules was further refined selecting those featured promising pharmacokinetic properties (see Materials and Methods). Then, we focused on a restricted collection of 2,4-thiazolidinedione derivatives exploiting the convenient synthetic route, and the commercial availability of cheap synthons.
Eventually, with the aim of selecting further promising BAG3BD modulators, we here performed an in-silico screening following a fast ligand-based approach. Specifically, the focused subset of compounds was screened against the lead compound 1 computing the related shape similarity values (Phase software) (Schrodinger, LLC, New York, NY, USA) [19]. In this way, a restricted set of compounds was selected from the original library with a shape comparable to that of the lead compound 1, following the principle that similar molecules could likely show similar binding modes on the protein counterpart, as demonstrated in several virtual screening studies (Figure 1) [20,21]. Specifically, this protocol envisaged a first conformational search round on each item of the accounted library of 2,4-thiazolidinedione-based compounds and, afterwards, the obtained conformers were aligned and compared to the reference compound 1. In detail, the "shape similarity" can range from zero (no one atom matching between screened and reference compounds) to one (all atoms matching), and the screened subset of compounds were here further selected only saving those features a shape similarity higher than 0.500. Using this further filter, the starting library of~2.3 × 10 4 2,4-thiazolidinedione-based compounds was reduced to a small set of eight promising molecules (2)(3)(4)(5)(6)(7)(8)(9), also chosen in order to define some structural information coupled with the outcomes of subsequent biological investigations ( Figure 2).

Synthesis of 2,4-Thiazolidinedione Derivatives 2-9
To obtain the newly selected derivatives (2-9) of our previously described lead compound 1, we took advantage of the synthetic strategy already optimized by us, shown in Scheme 1. Thus, a Knoevenagel condensation has been carried out, in basic conditions for piperidine, between the commercially available 2,4-thiazolidindione 10 and the selected aromatic aldehydes (a-d), affording the (Z)-5-arylidene-2,4 thiazolidinediones 10a-10d in high yields (Scheme 1) [22]. Treatment of 10a-10d with halides (e-g), in presence of NaH and DMF (Dimethylformamide) dry, provided the desired compounds 2-4 and 6-9 along with the intermediate 11 [23]. Finally, for the synthesis of compound 5, a next step in which 11 was coupled with morpholine, in the presence of HOBt and N,N-diisopropylcarbodiimide (DIC), as coupling agent, was necessary (Scheme 1) [24,25].

Biophysical Assays
The synthesized molecules 2-9 have been subjected to biophysical screening to evaluate their binding affinities to BAG3. A surface plasmon resonance (SPR) assay [18], using a recombinant BAG3 protein, was performed and, to explore the potential selectivity of our potential binders, the affinity for BAG4-the most closely BAG3 related protein-was also evaluated. Moreover, the binding affinity towards the BAG3BD domain was also assessed. The lead compound 1 was used in the same experimental condition in SPR assay, as reported in Table 1. Table 1. SPR assays of compounds 1-9 on rBAG3 protein full length, BAG3BD, and BAG4.

Biological Assays
Encouraged by the binding affinities profile obtained for this second collection of 2,4-thiazolidinedione derivatives as BAG3 modulators, we decided to further explore their potential anticancer activity.
Firstly, the antiproliferative activity of lead compound 1, 2 and 5-9 against selected human cell lines was assessed. No effect on human PHA (Phytohaemagglutinin)-stimulated proliferating non-tumor human cell line (PMBC) has been detected for compound 1 tested at different concentrations (10-50 µM) after 72 h of exposure [18]. In more detail, all the synthesized compounds have been tested on human melanoma cancer A375 and HeLa cancer cell lines, which are known to overexpress BAG3 protein, by an MTT cell viability assay, using different concentrations (5-50 µM). The IC 50 values have been calculated after 48 h of treatment with test compounds (Table 2) [26]. Table 2. Antiproliferative activity of 1,2 and 5-9: IC 50 (µM) ± SD, 48 h; human melanoma cell line (A375) and human cervical adenocarcinoma cell line (HeLa).

Compound A375 Cell Line IC 50 ± SD (µM)
HeLa Cell Line IC 50 ± SD (µM) 1 (lead) 15 28.40 ± 0.9 31.28 ± 1.0 IC 50 = half maximal inhibitory concentration; SD = standard deviation. The different colours have been used in order to highlight the different profile of the lead compound and the most promising molecule of the new collection.
Since BAG3 possesses an antiapoptotic activity, the effects of 6 on A375 and HeLa cell cycle distribution, by flow cytometry analysis, were analyzed. The cells were incubated for 48 h with 1 and 6 at concentrations 5-10-25µM. Upon treatment with 6, a dose-dependent accumulation in the G 2 phase was observed in both HeLa and A375 cells (Figures 3a and 4a). Concerning 1, an accumulation in the G 2 phase was observed in A375 cells, while no evident effect was found in HeLa cells (Figure 3a). In agreement with these results, in both cell lines compound 6 induced a significant increase of the apoptotic response in a dose-dependent manner, as depicted in Figures 3b and 4b displaying the hypodiploid nuclei. A dose-dependent increase in hypodiploid nuclei was also observed for 1 in A375 cells (Figure 3b).  Furthermore, the expression of caspases, the key mediators of programmed cells death, was examined in HeLa cells after the treatment with compound 6. The flow cytometry analysis showed a significant and dose-dependent activation of both caspase 3 and caspase 9 levels in cells following the incubation with compound 6 (Figure 5a,b) [27,28]. Finally, an expression decrease of BAG3 was observed upon treatment with compound 6 of HeLa cells. This result confirms the direct interference of 6 with BAG3 which is known to self-regulate its levels through a positive feedback mechanism ( Figure 6) [29]. Taking together, all these data point out the important role of 6 in BAG3 modulation, emerging, thus, as an attractive candidate for drug development and a useful tool for further biological investigations of this multifaceted co-chaperone. Hence, by using an integrated approach exploiting computer-aided procedures, biophysical techniques, and biological evaluation, the 2,4-thiazolidindione scaffold was successfully confirmed as a promising chemical platform able to selectively modulate BAG3 activity.

In-Silico Screening
The starting library of 2,4-thiazolidinedione-based compounds (~2.3 × 10 4 items) was initially subjected to a filtering step in order to select those featured esters and amide functions at the N-3 position, using the Ligand Filtering utilities implemented in Schrödinger Suite (4242 output compounds). Afterwards, the pharmacokinetic properties of these compounds were computed using QikProp software (Schrodinger, LLC, New York, NY, USA), related to: (a) absorption, distribution, metabolism, and excretion (ADME); (b) reactive functional groups types, excluding those that may cause false positives in high-throughput screening (HTS) assays, and/or decomposition.
Specifically, the following parameters were accounted: (a) molecular weight < 400 g/mol; (b) #stars: number of property or descriptor values that fall outside the 95% range of similar values for known drugs (range or recommended values: 0-5); (c) rtvFG: number of reactive functional groups, which can lead to false positive in high-throughput screening (HTS) assays and to decomposition, reactivity, or toxicity problems in vivo (range or recommended values: 0-2); (d) #donorHB: estimated number of hydrogen bonds that would be donated by the solute to water molecules in an aqueous solution. Values are averages taken over a number of configurations, so they can be non-integer (range or recommended values: 0.0-6.0); (e) #accptHB: estimated number of hydrogen bonds that would be accepted by the solute from water molecules in an aqueous solution. Values are averages taken over a number of configurations, so they can be non-integer (range or recommended values 2.0-20.0); (f) #QPlogPo/w: predicted octanol/water partition coefficient (range or recommended values −2.0-6.5).
After further "qualitative" filtering steps (i.e., compatibility with synthetic procedure, commercial availability of the synthons, etc., see Results and Discussion), a fast "shape screening" alignment was then performed, using the 3D structure of the reference BAG3 lead inhibitor 1 as shape query ligand. Phase software (Schrodinger, LLC, New York, NY, USA) was employed [19]; in particular, for the screened compounds, the sampling was performed allowing the conformers around the amide bond to vary freely, and finally, 1000 maximum number of conformers were considered for the shape computation. Once associated a shape similarity value for each compound computed against 1, a ranking from the best to worst values was obtained, saving those featured having a shape similarity value higher than 0.500.
Further filters, specifically aimed to select a focused set of compounds useful for both assessing the predicted binding with the protein counterpart and to facilitate the definition of a minimal structure-activity relationship, were applied, leading to eight items (compounds 2-9) as promising 2,4-thiazolidinedione-based compounds for the subsequent chemical synthesis and biological evaluation steps.

Biophysical Assay
The SPR analysis were performed on Biacore 3000 TM equipped with a research-grade CM5 sensor chip (GE Healthcare TM , Milan, Italy). Human recombinant BAG3 (rBAG3) protein was purchased from MyBioSource© (San Diego, CA, USA), BAG3 domain (rBAG3BD) was purchased from ARETA International S.r.l. (Gerenzano, VA, Italy), and human recombinant BAG4 protein was purchased from Abnova, Co© (Taipei, Taiwan) Our previously synthesized lead compound 1 was used as a positive control [17]. The proteins were coupled to the surface of a CM5 sensor chip using standard amine-coupling protocols, according to the manufacturer's instructions. Three separates recombinant rBAG proteins surfaces and one unmodified reference surface were prepared for simultaneous analyses. rBAG proteins (BAG3 and BAG4) and rBAG3BD (100 µg mL −1 in 10 mM CH 3 COONa, pH 4.5) were immobilized on the sensor chip surface at a flow rate of 5 µL min −1 to obtain densities of 5-9 kRU. Compounds 1-9 were dissolved in 100% DMSO to obtain 50 mM solutions and diluted 1:100 (v/v) in PBS (phosphate-buffered saline: 10 mM NaH 2 PO 4 , 150 mM NaCl, pH 7.4) to a final DMSO (dimethyl sulfoxide) concentration of 1.0%. For each molecule, a six-point concentration series was conducted in triplicate (0-0.25-1-10-20-100 µM). SPR (Surface Plasmon Resonance) experiments were performed at 25 • C, using a flow rate of 10 µL min −1 , with 60-s monitoring of association and 300-s monitoring of dissociation. Sensorgram elaborations were performed using the BIAevaluation software (GE Healthcare TM , Milan, Italy). provided by GE Healthcare TM . Simple interactions were suitably fitted to a single-site bimolecular interaction model (A + B = AB), yielding a single K D .

Cell Culture
Cell lines of human cervical carcinoma (HeLa) and human malignant melanoma (A375) were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, supplemented with 100 U/mL each of penicillin and streptomycin, and 2 mM/L glutamine and were grown at 37 • C under 5% CO 2 air humidified atmosphere.

Cell Viability Assay
Cell viability was evaluated using a colorimetric assay based on MTT ((3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide)) assay. Cells were grown in 96well tissue culture plates (3,5 × 10 3 cells/well) and after 24 h, were treated with different concentrations of 1, 2 and 5-9 (50, 25, 10, and 5 µM) and incubation was performed for 48 h at 37 • C in an atmosphere containing 5% CO 2 . At the end of treatment, the plates were centrifuged at 1200 rpm for 5 min, the medium was removed and 100 µL of MTT (1 mg/mL) were added to each well, and cells were incubated at 37 • C for the time necessary to allow the formation of purple formazan precipitate. The solution was then removed from each well, and the formazan crystals within the cells were dissolved with 100 µL of DMSO [31]. The optical density (OD) of each well was measured with a microplate spectrophotometer (Multiskan Spectrum Thermo Electron Corporation reader) (Thermo Fisher Scientific, Rome, Italy) equipped with a 620-nm filter. Cell mortality was calculated as: % mortality = 100 − [100 × (OD treated/OD control)] [32].

Apoptosis and Cell Cycle Analysis
The effect of compound 1 and 6 on cell death was analyzed by propidium iodide (PI) (Merck, Darmstadt, Germany) incorporation in permeabilized cells and flow cytometry. Briefly, cells were plated at a density of 2 × 10 4 cells/well in a 24-well plate. After 24 h, 1 and 6 (25-10-5 µM) were added for 48 h. Cells were washed twice with PBS and incubated in 500µL of a solution containing 0.1% Triton X-100, 0.1% sodium citrate, and 50 mg/mL propidium iodide (PI), at 4 • C for 30 min in the dark. The PI-stained cells were subsequently analyzed by flow cytometry by FACSCalibur flow cytometer (Becton Dickinson, Milan, Italy) using CellQuest software (4.0, Becton Dickinson, North Ryde, NSW, Australia). Results are expressed as the percentage of cells in the hypodiploid region. Cellular debris were excluded from the analysis by raising the forward scatter threshold and the DNA content of the nuclei was registered on a logarithmic scale. Cell cycle profiles were evaluated by DNA staining with a PI solution using a flow cytometer [33]. Results from 5000 events per sample were collected, and the relative percentage of the cells in G 0 /G 1 , S, G 2 /M, and sub-G 0 /G 1 phases of the cell cycle was determined using the ModFit LT version 3.3 analysis software (BD Biosciences, North Ryde, NSW, Australia).

Measurement of Caspase 3 and Caspase 9 Levels
Cells were plated into 12-well plated (2 × 10 4 cells/well) and after 24 h were treated with 6 (50-25-10-5 µM) for 48 h to assess caspase 3 and caspase 9 levels. Cells were collected, washed with PBS, then incubated in fixing buffer (containing 4% formaldehyde, 0.1% NaN 3 , and 2% FBS in PBS) for 20 min and then permeabilized with fix perm solution (fixing buffer containing 0.1% Triton X) for 30 min. Afterward, anti-caspase 3 or anti-caspase 9 antibodies were added for a further 30 min. Cells were then incubated with anti-rabbit Texas Red antibody as a secondary antibody for 1 h at 4 • C. Cell fluorescence was evaluated using a fluorescence-activated cell sorter (FACSscan; Becton Dickinson, Milan, Italy) and analyzed with the Cell Quest software (4.0, Becton Dickinson, North Ryde, NSW, Australia). Data were shown as a percentage of positive cells.

Data Analysis
Data are reported as mean ± S.E.M. values of independent experiments, performed at least three times, with three or more independent observations. Statistical analysis was performed by Student's t-test. Differences with p < 0.05 were considered statistically significant.

Conclusions
Starting from the first disclosed BAG3 modulator 1, previously discovered by us, a new collection of differently decorated compounds featuring a 2,4-thiazolidinedione core was synthesized. The affinity towards the target of interest was evaluated by performing SPR assays using both the full-length BAG3 protein and the isolated BAG domain. By a deeply biological evaluation of the selected molecules, compound 6 was disclosed as a new interesting antiproliferative agent able to interfere with BAG3 functions. These promising results may stimulate further research toward the identification of more potent and selective BAG3 binders as attractive candidates for the development of new anticancer drugs.