Synthesis, antitumor evaluation and computational study of thiazolidinone derivatives of dehydroabietic acid-based B ring-fused-thiazole

In an attempt to search for new natural product-based antitumor agents, a series of novel thiazolidinone derivatives of dehydroabietic acid-based B ring-fused-thiazole were designed and synthesized. The primary antitumor tests showed that compounds 5m exhibited almost the best inhibitory activity against the tested cancer cells. The computational study suggested NOTCH1, IGF1R, TLR4, and KDR were the core targets of the title compounds, and the IC 50 of SCC9 and Cal27 is strong correlation with the binding ability of TLR4 and compounds.


Introduction
Chemotherapy is still a main method for cancer treatment [1] .However, The tumor cells could resistant to some chemotherapeutic drugs intrinsically, and show acquired drug resistance after chemotherapy, which makes the tumor cells become insensitive to similar drugs later [2,3] .For this reason, the development of the drug with novel structure and higher therapeutic e ciency is still an important mission for medicinal chemists.
Natural products play an important role in drug development due to their chemical structure diversity, better biocompatibility and drug-like properties [4] .Dehydroabietic acid (DHA) is a natural tricyclic diterpenoid, isolated from pine resins.DHA and its derivatives were reported possessing a broad spectrum of biological activities [5] .Recent reports indicated that DHA and its derivatives exhibited antitumour activity with a variety of mechanisms, such as survivin inhibition [6] , DNA binding [7] , EGFR or MEK1 kinase binding [8,9] , PI3K/AKT/mTOR pathway signaling inhibition [10] .Therefore, using DHA as a promising starting material to synthesize leading compounds for the development of antitumor agents has received growing interest from medicinal chemists.
In addition, thiazolidinone scaffolds are privileged structures in drug design, some of its derivatives made signi cant contribution in clinical treatment [11] .In the development of antitumor agents, thiazolidinone scaffolds are also widely used as pharmacophore to form new candidate compounds, such as some synthetic indole-thiazolidinones could trigger DNA damage and induce apoptosis of human tumor cells [12] , some 2,3-disubstituted 1,3-thiazolidin-4-ones could prevent mitotic divisions of the human renal tumor cell though G1 cell cycle arrest [13] , some hybrid molecules integrated Ciminalum and thiazolidinone moieties show signi cant anticancer activity against different tumor cells [14] .
For the reasons given above, an effective strategy, molecular-hybridization approaches [15] , was applied in this work for the development of novel DHA antitumor derivatives.Thus, a series of title compounds were synthesized through integrating thiazolidinone moieties with our previous synthesized DHA based B ringfused-thiazole amine [10] .The in vitro antitumor activities of the compounds against the oral cavity cancer cells lines (SCC9, Cal27 and SAS) were evaluated.The action mechanisms of the title compounds against tumor cell line were studied using a representative compound by the prediction of therapeutic targets, protein-protein interaction network, and molecular docking.

Chemistry
The synthetic approaches adopted to afford the title compounds are outlined in Scheme 1. Compounds 1 were prepared according to the literatures [16,17] .Compounds 2 were prepared according to our previous work [10] , and further acylated with chloroacetyl chloride at room temperature to formed compound 3 in 80-90 % yield.The rst thiazolidinone 4 was prepared by the reaction of compound 3 with potassium thiocyanate in ethanol using excess triethylamine as an acid-binding agent.The title compounds 5a -5n were synthesized in good yields by the condensation reaction of the thiazolidinone 4 with different aryl aldehydes in a mixture of triethylamine and ethanol.
The identities of the title compounds were con rmed using physicochemical analytical methods. 1 H NMR spectra revealed that the protons of NH which belong to the imino-amino tautomers exhibited 2 single signals in the 12.32 -12.18 and 10.40 -9.24 ppm range.Due to this tautomer, some H signals were showed cleavages with varying degrees especially the H on the conjugate plane.The singlets at 8.00-7.67ppm were assigned to the protons of the ole nic bond.The aromatic protons of the dehydroabietic acid scaffold showed as signals at 7.81 -7.63 and 7.24 -7.14.The methyl ester group protons showed a signal in the 3.70 -3.64 ppm range.The 13 C NMR spectra signals of all the target compounds were match with the number of the carbon type.The high-resolution mass spectra con rmed the expected molecular weights of the nal products.

Antitumor activity
The synthesized title compounds 5a -5n were evaluated for their in vitro antitumor activities by CCK-8 assay against the 3 oral cavity cancer cells lines (SCC9, Cal27 and SAS), which were belong to tongue squamous cell carcinoma.The results are summarized in Table 1.
It was found that the synthesized compounds showed effective antitumor activities against SCC9 and Cal27, most of them exhibited better inhibitory ability than thiazolidinone 4, which almost showed as the best compound against SAS cells line.These results indicated that integrated arylidene groups could enhance the antitumor activities against SCC9 and Cal27, but showed opposite effect against SAS cell line.The compounds with nitrogenous heterocyclic arylidene groups, 5m, 5k, and 5l showed the lowest half-inhibitory concentration (IC 50 ) in SCC9 and Cal27 assays.It suggested that the lone electron pair on the aryl groups might act as important role to interact with treatment targets.Compound 5m was chosen as representative for the further computational action mechanism study due to its excellent IC 50 of 0.2 ± 0.1, 0.2±0.1 and 7.3 ± 0.4, against SCC9, Cal27, and SAS, respectively.Analysis of the core targets and network pharmacology of 5m against the cancer cells tested A total of 16 potential therapeutic targets of the representative compound 5m was imported into the STRING database to obtain the data of PPI, summarized in Supplementary Table S4, and the visualized PPI network, shown in Figure 1, the nodes in the network represent the potential targets, and the edges represent the interaction between the targets.The network was comprised of 16 nodes and 16 edges, and the average node degree was 2.0.The targets with the degree greater than 1.9 times median and could found in the 145 common targets of compounds 5a -5n, NOTCH1, IGF1R, TLR4, and KDR, were selected as the core targets.
The gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enrichment analysis were performed using the 16 potential targets of compound 5m.87 GO-terms of biological process (BP), 1 GO-term of molecular function (MF) and 5 pathways of KEGG pathways were enriched signi cantly, summarized in Supplementary Table S5 -S7.Besides the top 5 GO-terms of BP with the lowest false, the terms of MF and KEGG were shown in Table 2.These results showed that compound 5m inhibited the cancer cells tested might mainly though participating in the cancer related pathways and regulating the biological process of cells death, apoptosis, proliferation, and so on.In an attempt to study the possible interaction mode between synthesized compounds 5a -5n and the core targets predicted, NOTCH1, IGF1R, TLR4, and KDR (PBD ID: 2f8x, 1igr, 3fxi, and 1y6a, respectively), molecular docking was conducted using AutoDock Tools.Each docking result contained 50 conformations predicted, and the conformation with the lowest binding energy was chosen as representative to analysis.The correlation between the antitumor activity of compounds and the binding ability to the targets were investigated by Spearman's rank correlation coe cient analysis through compare the IC 50 with the binding energies.The results were shown in Table 3.
Table 3 The binding energies and their correlation coe cient between the IC 50 of compound 5m The binding mode prediction between compound 5m and NOTCH1 were performed at the binding pocket of MAML, the co-activator protein of NOTCH1 [18] , and shown in Fig. 2. Compound 5m possibly binds to the pocket through 10 non-bonding interactions with residues of Glu2072, Tyr381, Pro366, Val2039, Val2042, Asp2043, Asn2041, Ser2074, Glu2076, Gly2073 and 2 hydrogen bonds interacted residues of Arg382 and Asn2040.(Fig. 2a, b).
The correlation study result showed that the IC 50 against SCC9 and Cal27 exhibited moderate positive correlation with the lowest binding energies of IGF1R at signi cant level of 0.05, and exhibited positive strong correlation with the lowest binding energies of TLR4 at signi cant level of 0.01.However, in SAS assay, the IC 50 showed moderate negative correlation with the lowest binding energies to IGF1R and KDR at signi cant level of 0.05.These results indicated that the binding ability of compounds to TLR4 might be an important factor effected the antitumor activity against SCC9 and Cal27.

Conclusions
In summary, 15 new thiazolidinone derivatives of dehydroabietic acid-based B ring-fused-thiazole were synthesized and their in vitro cytotoxic activities were investigated by CCK-8 assay against SCC9, Cal27 and SAS.Most of the synthesized compounds exhibited better antitumor activity than the intermediate thiazolidinone 4 against SCC9 and Cal27, while compound 5m (R = 1H-pyrrol-2-) showed excellent IC 50 of 0.2 ± 0.1, 0.2 ± 0.1 and 7.3 ± 0.4, against SCC9, Cal27, and SAS, respectively.NOTCH1, IGF1R, TLR4, and KDR were extracted as core targets from the predicted potential therapeutic targets of these new thiazolidinones againts the cells tested, besides, TLR4 was suggested as an important target of compounds against SCC9 and Cal27 due to the strong correlation between the IC 50 and the lowest binding energies.These results indicated that the synthetic strategy in this study is feasible.The compound 5m might serve as a leading compound for potent antitumor agents and the TLR4 inhibitor.
Cells were harvested during logarithmic growth phase and seeded in 96-well plates at a density of 5×10 3 cells/well, and cultured at 37 • C in a humidi ed incubator (5% CO 2 ) for 24 h, followed by exposure to various concentrations of compounds for 24 h.Subsequently 10 µL of CCK-8 was added to each well, the cells were then incubated for an additional 1 h at 37 • C. Cell growth inhibition was determined by measuring the optical density value (OD) at λ = 450 nm using a microplate reader.Three independent experiments were performed.Cell growth inhibition was calculated according to the following equation: The IC 50 were obtained from liner regression analysis of the concentration -response curves plotted for each tested compound.

Computational Study
Prediction 5a -5n therapeutic targets in the 3 cancer cell lines The structure of 5a -5n was imported into the database of PharmMapper Server (http://www.lilabecust.cn/pharmmapper/)for prediction [22] .The results, whose scores of Norm Fit were greater than zero, were chosen as the predicted targets.Their information, including the gene names and gene ID, were further extracted using UniProtKB (http://www.uniprot.org) [23]."Tongue squamous cell carcinoma" were used as keywords to retrieve the 3 cancer cell lines associated targets in Online Mendelian Inheritance in Man database (OMIM, https://omim.org)and Genecards database (https://www.genecards.org) [24].
The intersection of 5a -5n respective targets and the 3 cancer cell lines associated targets were chosen as the treatment targets of 5a -5n against tumor call tested.
Analysis of the core targets and network pharmacology of 5m against the cancer cells tested The potential targets gene list was imported into the STRING database (https://string-db.org),with the species as "Homo sapiens" and a con dence score greater than 0.4, to obtain the data of PPI and the visualized PPI network [25] .The targets whose degrees were greater than 1.9 times median, and can nd in all the treatment targets were selected as the core targets .
The biological functions of the target genes were assessed through GO and KEGG pathways enrichment analysis.The predicted results were also obtained for the STRING database.
Molecular docking study of the core targets All docking procedures were performed using AutoDock 4.2.6 software (ADT) according to the reported paper [26] .All the proteins in the PDB les were cleaned by removing small molecules, ions, and original ligand in the crystal of the proteins.Then, polar hydrogen atoms were added, Gasteiger charges were computed, and the atoms were set as "Assign AD4 type" in ADT.The compounds for docking program were drawn using ChemDraw 8.0 and automatically set the torsional bonds by the AUTOTORS module in ADT.The docked sites of the protein were set in the place of the original ligand or according the literature [18] .The binding energy between the docked compound and the protein was calculated using the AutoGrid program with a grid spacing of 0.375 Å by the Lamarckian genetic algorithm as a searching method.The interaction between compounds and the amino acid residues of the proteins was found by LigPlot + 2.1 and visualized by PyMOL 2.4.0a0software.
The correlation between the antitumor activity and the binding energies were analysed using spearman rank correlation coe cient.The strength of the correlation were guided as very weak, weak, moderate, strong and very strong by the correlation coe cient in absolute value in the intervals of [0, 0.2), [0.2, 0.4), [0.4,0.6), [0.6, 0.8), and [0.8, 1.0], respectively [27] .

Figure 1 Network 2 a
Figure 1

Table 2
The top 5 GO-terms of BP, GO-terms of MF and KEGG pathways prediction of compound 5m The docking results showed the lowest binding energy of compounds to NOTCH1, IGF1R, TLR4 and KDR were -6.67 --8.15, -5.94 --7.04, -7.04 --8.24, and -7.71 --8.19 kcal/mol, respectively.These result indicated that the compounds might possess certain binding ability to the 4 targets, which were found overexpression in caner cells.