Synthesis, structure and in vitro antiproliferative effects of alkyne-linked 1,2,4-thiadiazole hybrids including erlotinib- and ferrocene-containing derivatives

Chemotherapy is an indispensable tool to treat cancer, therefore, the development of new drugs that can treat cancer with minimal side effects and lead to more favorable prognoses is of crucial importance. A series of eleven novel 1,2,4-thiadiazoles bearing erlotinib (a known anticancer agent), phenylethynyl, ferrocenyl, and/or ferrocenethynyl moieties were synthesized in this work and characterized by NMR, IR and mass spectroscopies. The solid-phase structures were determined by single-crystal X-ray diffraction. Partial isomerisation of bis(erlotinib)-1,2,4-thiadiazole into its 1,3,4-thiadiazole isomer, leading to the isolation of a 3 : 2 isomer mixture, was observed and a plausible mechanism for isomerisation is suggested. The in vitro cytostatic effect and the long-term cytotoxicity of these thiadiazole-hybrids, as well as that of erlotinib, 3,5-dichloro-1,2,4-thiadiazole and 3,5-diiodo-1,2,4-thiadiazole were investigated against A2058 human melanoma, HepG2 human hepatocellular carcinoma, U87 human glioma, A431 human epidermoid carcinoma, and PC-3 human prostatic adenocarcinoma cell lines. Interestingly, erlotinib did not exhibit a significant cytostatic effect against these cancer cell lines. 1,2,4-Thiadiazole hybrids bearing one erlotinib moiety or both an iodine and a ferrocenethynyl group, as well as 3,5-diiodo-1,2,4-thiadiazole demonstrated good to moderate cytostatic effects. Among the synthesized 1,2,4-thiadiazole hybrids, the isomer mixture of bis-erlotinib substituted 1,2,4- and 1,3,4-thiadiazoles showed the most potent activity. This isomer mixture was proven to be the most effective in long-term cytotoxicity, too. 3,5-Diiodo-1,2,4-thiadiazole and its hybrid with one erlotinib fragment were also highly active against A431 and PC-3 proliferation. These novel compounds may serve as new leads for further study of their antiproliferative properties.


Introduction
The various types of cancer are highly devastating diseases worldwide, in most cases with poor prognosis and low survival rates. 1,2 It is no doubt that chemotherapy is one of the most indispensable tools for the treatment of malignancies. However, the clinical efficacy of most anticancer chemotherapies is substantially decreased by a variety of factors including multidrug resistance (MDR) 3,4 and severe adverse effects that strongly make it necessary to develop more potent novel drugs with enhanced activity and selectivity towards tumours. One of the most promising new strategies in chemotherapy is based on the design and synthesis of hybrid compounds, by coupling different pharmacophore fragments. [5][6][7] Such hybrid drugs having more than one molecular target at the cellular level can be considered to be anticancer agents of enhanced efficiency triggering cell death along multiplied pathways.
Hybrid anticancer agents possess real potential to overcome certain disadvantages of single cancer drugs, including MDR and adverse effects; therefore, based on this promising strategy, the aim of the present work was to develop novel anticancer agents for the potential treatment of human glioblastoma, 8 melanoma, 9 non-melanoma skin carcinoma, 10,11 hepatocarcinoma 12 and prostatic adenocarcinoma 13 by coupling the were simplied to phenylethynyl groups, which were then further modied by replacing phenyl substituent(s) with threedimensional and redox-active ferrocenyl group(s) of enhanced lipophilicity. Accordingly, in this paper, we report the synthesis and structural characterisation of a selection of novel thiadiazole-hybrids (4-15: Schemes 1 and 2) along with their evaluation in antiproliferative studies carried out in vitro on A2058, A431, U87, HepG2 and PC-3 cells using 1, 2 and 3, as controls.

Synthesis and identication of compounds
Considering the ethynyl functional group of 1, ethynylferrocene and phenylacetylene, it seemed to be straightforward to couple these compounds with 3,5-dihalo-1,2,4-thiadiazoles, 2 and 3, using Sonogashira-type cross-coupling reactions (Scheme 1; note that the bromo-derivative of these thiadiazoles is not known yet). Sonogashira-coupling was indeed a good choice, however, the product yield was very sensitive to reaction conditions, namely to solvent, base, catalyst, and temperature (see Tables S1 and S2, ESI †). The best results for coupling ethynylferrocene and phenylacetylene to thiadiazole were obtained by using toluene as the solvent, diisopropylamine (DIPA) as the base, and PdCl 2 (PPh 3 ) 2 / CuI as the catalyst at a slightly elevated temperature of 50 C. However, the combination of DMF, K 3 PO 4 , and Pd[P(t-Bu) 3 ] 2 /CuI at 80 C provided the most efficient conditions for the coupling with erlotinib. Using 1.1 or 3 equivalents of the alkyne component (1, ethynylferrocene or phenylacetylene) under optimised conditions, the cross-coupling reactions led smoothly and selectively to thiadiazole products monosubstituted at position ve (C5) or to the disubstituted products, respectively, in isolated yields varied between 66 and 87%. In the subsequent step the monosubstituted 1,2,4-thiadiazoles were then converted into the target disubstituted products (Scheme 1). The selectivity of the coupling reaction in the rst step allowed the synthesis of thiadiazole-hybrids with two different substituents by selecting the sequence of the introduction of the selected alkyne component. The regioselectivity of the rst reaction step can be explained by the lower electron density of the thiadiazole ring at C5 and thus higher reactivity with nucleophiles, compared to position C3. 14 This regioselectivity was unambiguously conrmed by the MO analysis of 2 and 3 disclosing LUMO concentrated at C5-N4 region and LUMO+1 delocalised over C3-N2 region in both models (Fig. 1). Moreover, the energetic data calculated for these orbitals are also in good agreement with the experimentally observed relative reactivity of 3 and 2, indicating that the diiodo derivative is more reactive than its dichloro-substituted counterpart. To investigate the effect of the ethynyl spacer between the thiadiazole moiety and ferrocenyl group, this latter was also directly linked to 2 applying Suzuki-type coupling. Reaction conditions were studied and good yield was obtained using dioxane, K 2 CO 3 , and Pd(OAc) 2 /PPh 3 at reux conditions (see Table S3, ESI †). Since the removal of the ethynyl spacer has not resulted in improved antiproliferative properties, this research direction has not been further investigated. All synthesized compounds have been unambiguously identied and characterized by mass ( Fig. S1-S9, ESI †), IR (Fig. S10 †), and NMR spectroscopies. Molecular masses were determined by ESI-MS, and IR spectroscopy provided a characteristic ngerprint for the ethynyl group(s) in the 2200-2227 cm À1 spectral region. The constitution of all newly synthesized thiadiazole-hybrids was conrmed by 1 H and 13 C NMR data ( Fig. S17-S30, ESI †) assigned on the basis of connectivities disclosed by 2D-HSQC-and HMBC measurements ( Fig. S31-S37, ESI †). In addition, the solid phase structures of hybrids 4, 5, 6, 8, 10 and 11 have also been determined by singlecrystal X-ray diffraction as discussed later in details.
It is of interest that Sonogashira coupling reactions aiming at the synthesis of bis-erlotinib derivative 14 led to the isolation of an approximately 3 : 2 mixture of the targeted product and its isomer 15 featuring a [1,3,4]thiadiazole-centered symmetrical constitution (Scheme 2), as detected and identied by extended NMR studies including 2D-HSQC and 2D-HMBC measurements.
The formation of 15 can be interpreted in terms of the partial isomerisation of 14, the primary product of the Sonogashira reaction when it was conducted at 95 C for prolonged reaction times (for 15 h by Procedure A and for 20 h by Procedure B) representing harsher conditions than those employed for the other conversions investigated in our research that were conducted at 50-80 C for 6-12 h (see: Experimental). The mechanism proposed by us for transformation 14 / 15 involves the initial p-complexation of a Pd(0) species with the alkyne residue at C5-position and a concomitant coordination of a Cu(I) species to the N2 ring atom to construct bimetallic cationic complex 16a (Scheme 2a). By means of the coordination of the Cu(I) centre to the proximal S1 atom and the development of a bonding interaction between alkyne-coordinated Pd(0) and the skeletal N4 atom in 16a, the two metal centres promote the ssion of the S1-N2 and C3-N4 bonds proceeding via transition state TS (16)(17) to generate a ve-membered palladacycle intermediate with pending tricoordinated copper(I)-centre (17). In the subsequent steps involving separated intermediates 18 and 19, complex 17 is supposed to isomerize to complex 20 featuring a N-Cu-N coordination mode. To terminate the rearrangement process, 20 undergoes electrocyclisation through a transition state TS(20-21b), with a quasi-six membered ring, nally constructing 21b stabilized by a Cu(I)/Pd(0) interaction as evidenced by DFT calculations performed on simplied models (Fig. 2). Regarding the initial elementary step it must be noted here thataccording to the aforementioned theoretical studiesthe cleavage of S1-N2 and C3-N4 bonds in regioisomer complex 16b, stabilized by Cu(I)/Pd(0) interaction, is not feasible to advance the process TS(20*-21b*)] comprising single ethynyl group and appropriately coordinated neutral and cationic metal-containing simplied fragments PdPH 3 and CuPH 3 , respectively (Scheme 2b). The intermediates and transition states were identied as local minima and rst order saddle points, respectively, on the potential energy surface (PES). Transition states were localized by QST2 method. 24 The energetic prole of the overall transformation was characterized by the changes in Gibbs free energy (G) accompanying the assumed elementary steps and the activation barriers of the ring opening and ring closing processes. The free energy values of optimised structures were obtained by correcting the computed total energy with zero-point vibrational energy (ZPE) and thermal corrections calculated at the same level. In the initial stage of the conversion, the ssion of the 1,2,4thiadiazole ring was disclosed as the rate-limiting endothermic elementary step (DG ¼ +39.1 kcal mol À1 ) proceeding via a high barrier (DG ‡ ¼ 60.0 kcal mol À1 ) followed by an endothermic isomerization of the palladacycle intermediate (17* / 20*) taking place by the copper-centred decoordination-coordination sequence via separated Cu(I)-and Pd(0) fragments 18* and 19*. Finally, 20* was identied as the intermediate which can be connected by transition state TS(20*-21b*) exclusively with 21b* in accord with qualitative structural considerations, in the 1,3,4thiadiazole-forming cyclization. The multistep isomerization of bimetal complex 16a* into 21b* is an endothermic process as indicated by the change in the free energy calculated for the overall conversion (DG ¼ +6.2 kcal mol À1 ). Moreover, the change in the Gibbs free energy [DG(15*-14*) ¼ +14.5 kcal mol À1 ] calculated for vacuum by modelling simplied metal-free isomer pair 2-ethynyl-1,3,4-thiadiazole(15*)/5-ethynyl-1,2,4-thiadiazole (14*) would suggest thatin generalthe transformation of 5alkynyl-substituted 1,2,4-thiadiazoles into the 1,3,4-thiadiazole counterpart is not a feasible process however, the relative energetics of structures 14 and 15 and their appropriate metal complexes under real experimental conditions with bulky phosphine ligand [P(t-Bu) 3 ] of outstanding donor strength and dimethylformamide as solvent of signicant coordination-and solvation ability, might signicantly differ from the calculated values allowing the development of an equilibrium system containing 15 as the minor component. In accord with this view, upon further prolongation of the reaction time (24 h) practically no change in the isomer ratio was discernible in the isolated mixture of products 14 and 15.
Finally, the abovementioned Cu(I)/Pd(0) contact was disclosed by MO analysis of the simplied models 16b* and 21b*. The enhanced stability of these complexes relative to their coordination isomers 16a* [DG(16a*-16b*) ¼ +5.0 kcal mol À1 ] and 21a* [DG(21a*-21b*) ¼ +4.2 kcal mol À1 ], respectively, can be attributed to this type of interaction which is demonstrated by the delocalisation of two-two bonding orbitals in between the metal centres ( Fig. 2). Our attempts to separate 14 from 15 was not successful, therefore the mixture 14/15 was used in biological studies.

Crystal structure of ferrocenethynyl-and phenylethynylhybrids
The structures of compounds 4, 5, 6, 8, 10, and 11 in the solid phase were determined by solving X-ray diffraction data collected from single crystals. We note that our attempts to grow  X-ray quality single crystals from erlotinib derivatives were to no avail. The structure of 7 was published in our previous paper. 25 The crystal structures of 4, 5, 6, 8, 10, and 11 are shown in Fig. 3 and 4 with geometric parameters listed in Tables S4-S9, ESI. † Compound 5 crystallizes in the triclinic space group P 1, all others in the monoclinic space group P2 1 /c. All investigated hybrids exhibit a quasi-planar arrangement of the thiadiazole and connected phenyl or cyclopentadienyl rings; small deviations can be explained by solid-phase effects. A characteristic motif of the crystal packing is the formation of molecular dimers, by weak p/p interactions between aromatic rings of oppositely oriented molecules (see Fig. 4), and their arrangement into columns in the crystal ( Fig. S11-S16, ESI †). Intermolecular interactions in crystals, in general, are weak as indicated by the relatively large distance between ring centroids and interatomic distances (Tables S10-S12, ESI †).
Antiproliferative activity of compounds 1-13 and 3/2 mixture of 14/15 The effects of 1-13 and 14/15 on the growth of ve human cancer cell lines were investigated using three different protocols. Protocol 1 has involved the treatment of cells with 1-14/15 for 20 h, consecutive washing, and incubating cells for an additional 72 h. Half maximal inhibitory concentration (IC 50 ) values are presented in Tables 1 and 2. Interestingly, 1 has not shown a signicant cytostatic effect on U87, A2058, A431, HepG2, and PC-3 cells (IC 50 > 50 mM). Linking 1 to the 1,2,4thiadiazole ring, however, resulted in enhanced cytostatic effect; compound mixture 14/15 comprising the 1,2,4-and 1,3,4thiadiazole isomers with two erlotinib moieties exhibited a signicant anti-tumour effect on all ve cell lines (IC 50 < 2 mM in case of U87, A2058, HepG2, and PC-3; IC 50 < 6 mM in case of A431). PC-3 cells were the most sensitive to 14/15 (IC 50 ¼ 0.4 mM). Iodine substituent on the 1,2,4-thiadiazole ring was found to be effective in increasing cytostatic effect; the diiodo-1,2,4-thiadiazole (3) was observed to be equally effective against A431 cells as its mono and di-substituted erlotinib derivatives (13 and 14/15). 3 and all erlotinib-hybrids were effective against PC-3. The linking of ferrocene, ethynylferrocene or ethynylbenzene to the thiadiazole frame has not resulted in a marked anti-tumour effect, namely with an IC 50 value lower than 10 mM (Table 1), however, several compounds exhibited notable cytostatic effect with IC 50 values between 10 and 30 mM (see Table 1).
The long term cytotoxicity (cytotoxic activity; the direct killing of cancer cells) of 1 and three thiadiazole hybrids (3, 13, and 14/15), that were effective in cytostasis experiments, was also studied on previously used four cell lines. Two setups of experiments were conducted: cells were treated with thiadiazole derivatives for 72 h (protocol 2) or treated 3-times with compounds, without interstitial washing, for 24 h (protocol 3). 1 produced an effect on A431 cells aer 72 hours, and on A431 and U87 cells following the 3 Â 24 h treatment. 14/15 was the most effective in these experiments too.
Currently, the combination of dabrafenib/trametinib, vemurafenib/cobimetinib, and encorafenib/binimetinib are approved for treating melanoma. 9 Although there is a rapid early response and high response rate to these combined agents, the progression of disease occurs at a median of eleven months, due to drug resistance; therefore, novel drugs and drug combinations are needed. 26 Thiadiazole-hybrids might be potential candidates. A2058 cells are found to be more sensitive to 14/15 than to vemurafenib (IC 50 ¼ 5.93 mM). 27 If chemotherapy is applied for the treatment of non-melanoma skin cancers, 5-uorouracil (e.g. in the form of its oral prodrug capecitabine or in creams) may be used. 10 )). Docetaxel is the mainstay of chemotherapy for prostate cancer with cabazitaxel as second-line drug. 13 1 has also been investigated in clinical trials as a potential chemotherapeutic agent for prostate cancer treatment. 31 Nine of the synthesized thiadiazole-hybrids proved to be more effective on PC-3 cells than 1 ( Table 2). Thiadiazole derivatives studied in this work, especially compounds 14/15, 13, and 3 with the lowest IC 50 values, may serve as new leads for further study of their antiproliferative properties. We note that the cytostatic effect of these compounds, especially that of 14/15, is outstanding compared to the reference anticancer drug 1.
Finally, it is of crucial importance to provide evidence for that the intact hybrid molecules rather than any of their decomposition-derived fragments are the species which induce evolution of the antiproliferative effect in the course of the biological assays employing long-term treatment of the cells. According to our rst observations the synthesized compounds in solid state were stable at ambient conditions; decomposition or colour change were not observed over a couple of months. Moreover, these compounds were not sensitive to air, moisture or light when handled at ambient conditions, and can generally be stored in closed vial at room temperature in the dark without decomposition as proved by IR-and MS measurements. The long-term stability of the compounds in solution was also checked by registering their 1 H-NMR spectra in DMSO-d 6 aer 72 h following the preparation of the liquid samples stored under air at room temperature. Supporting our abovementioned observations regarding the stability of the novel hybrids, their spectra did not show any detectable change in their structures, although the solvent was contaminated with a substantial amount of HDO.  3 ] 2 ], and triphenylphosphine, PPh 3 ; purity $ 98%), bases (diisopropylamine, triethylamine, potassium carbonate, K 2 CO 3 , and tripotassium phosphate, K 3 PO 4 ; purity $ 98%), and anhydrous solvents (1,4-dioxane, toluene, N,N-dimethylformamide; purity $ 99.8%) were purchased from commercial sources (Sigma-Aldrich, Fluorochem, VWR) and used, except solvents, without further purication. Solvents were dried according to published methods 32 and distilled before use. Compounds 3, 7 and 9 were synthesized as described recently. 25 Synthetic reactions were monitored by thinlayer chromatography (TLC) using Merck Silica gel 60 F254 TLC plates, and plates were visualised under a Camag model UV lamp 4 dual wavelength 254/366 nm. Column chromatography was performed using Merck silica gel 60 (0.063-0.200 mm) using a column (diameter 2.5 cm) with sintered glass disc.

Experimental
The 1 H-and 13 C-NMR spectra of the synthesized compounds were recorded on a Bruker DRX-500 MHz spectrometer at 500 MHz and 125 MHz, respectively, at room temperature using the deuterium signal of the solvent as the lock and tetramethylsilane (TMS) as the internal standard. The assignment of all 1 H-and 13 C-NMR data necessary for exact structural elucidation of the compounds was based on the cross-peak correlations discernible in 2D-HSQC and HMBC spectra obtained by standard Bruker pulse programs. Mass spectroscopic measurements were done using an Esquire 3000 + (Bruker) ion trap mass spectrometer and electrospray ionization (ESI). Exact mass measurements for samples 12, 13, and 14/15 were taken on a high-resolution Waters Q-Tof Premier mass spectrometer equipped with an ESI ion source (3000 V capillary voltage, 350 C desolvation temperature, 650 L h À1 nitrogen as desolvation gas). The samples were dissolved in methanol (10 mg mL À1 ) and 5 mL were injected in a continuous ow of methanol (400 mL min À1 , contain 0.1% formic acid). Each compound was analysed twice, the rst spectrum was recorded without consideration of temperature variations. The second measurement was processed aer the correction of temperature variations with a reference compound.
Protocol 1: the cells were grown to conuency and were distributed into 96-well tissue culture plates with an initial cell number of 5.0 Â 10 3 per well. Aer 24 h of incubation at 37 C, the cells were treated with the compounds in 200 mL nal volume containing 1.0 v/v% DMSO. The cells were incubated with the compounds at 0.4-50 mM concentration range for 20 h, whereas control cells were treated with serum-free medium (RPMI-1640) only or with DMSO (c ¼ 1.0 v/v%) at 37 C for 20 h. Aer incubation, the cells were washed twice with serum-free RPMI-1640 medium. To determine the in vitro cytostatic effect, the cells were further cultured for 72 hours in 10% serum-containing medium. 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide solution, MTT-solution, (45 mL, 2 mg mL À1 , nal concentration: 0.37 mg mL À1 ) was added to each well. The respiratory chain 60,61 and other electron transport systems 62 reduce MTT and thereby form non-water-soluble violet formazan crystals within the cell. 63 The amount of these crystals can be determined spectrophotometrically and serves as an estimate for the number of mitochondria and hence the number of living cells in the well. 64 Aer 3 hours of incubation the cells were centrifuged for 5 minutes at 900 g and the supernatant was removed. The obtained formazan crystals were dissolved in DMSO (100 mL) and optical density (OD) of the samples was measured at l ¼ 540 and 620 nm, respectively, using ELISA Reader (iEMS Reader, Labsystems, Finland). OD 620 values were subtracted from OD 540 values. The percent of cytostasis was calculated by using the following equation: Cytostatic effect (%) ¼ [1 À (OD treated /OD control )] Â 100 Values OD treated and OD control correspond to the optical densities of the treated and the control cells, respectively. In each case, two independent experiments were carried out with 4 parallel measurements. The 50% inhibitory concentration (IC 50 ) values were determined from the dose-response curves. The curves were dened using Microcal™ Origin2018 soware: cytostasis was plotted as a function of concentration, tted to a sigmoidal curve, and based on this curve, the half-maximal inhibitory concentration (IC 50 ) value was determined. IC 50 represents the concentration of a compound that is required for 50% inhibition in vitro and expressed in micromolar units. Protocol 2 and 3: cells were divided into 96 well tissueculture plates in 200 mL culture medium with the initial cell number of 5000 cells per well. The compounds were dissolved in DMSO and then diluted with fresh culture medium (nal DMSO concentration was 1% in each well) and they were added to the cells at 0.016, 0.08, 0.4, 2.0 and 10 mM nal concentration. Cells were incubated with the compounds at 37 C for 72 hours (protocol 2). The same layout of plates was parallelly treatedculture medium was removed, then compounds dissolved in medium containing 2.5% FBS were added to the wells, without washing -3 times, in every 24 hours (protocol 3). Aer that, cell viability was determined by MTT-assay using 0.37 mg mL À1 nal concentration of MTT, in each well. Aer 3 hours of incubation with MTT the absorbance was measured with ELISA-reader (Labsystems MS Reader) at 540 nm and 620 nm as reference wavelengths. IC 50 values were determined from the dose-response curves using the same method as described in Protocol 1.

Conclusions
In summary, we have demonstrated a exible route for the synthesis of 1,2,4-thiadiazole hybrids by cross-coupling halogen derivatives of 1,2,4-thiadiazoles with erlotinib, ethynylferrocene, phenylacetylene, or ferroceneboronic acid, as well as investigated the in vitro antiproliferative and cytotoxic activity of these compounds on ve tumorous cell lines (U87, A2058, A431, HepG2 and PC-3). The structures of all investigated compounds were conrmed by NMR, IR and mass spectroscopy, as well as single crystal X-ray diffraction. Ten compounds of the investigated fourteen exhibited cytostatic effect against at least one of the investigated cell lines with IC 50 value below 50 mM.

Conflicts of interest
There are no conicts to declare.