Potent antitumor activity of novel taxoids in anaplastic thyroid cancer

Anaplastic thyroid cancer (ATC) is the most aggressive form of thyroid cancers and it is rapidly fatal without any effective therapeutic regimens. There are some clinical trials showing that paclitaxel-based chemotherapy for ATC can achieve a relatively high response rate and low incidence of adverse reaction. The aim of this study was to evaluate potential therapeutic activity of novel taxoids in ATC cells. We evaluated antitumor activity of five novel 3′-difluorovinyltaxoids (DFV-taxoids) in anaplastic thyroid cancer cells by a series of in vitro and in vivo experiments. Besides, we also explored the potential mechanism underlying the difference among the taxoids and paclitaxel by molecular docking and tubulin polymerization assays. Our data showed that these novel DFV-taxoids were more effective than paclitaxel in ATC cell lines and xenografts, as reflected by the inhibition of cell proliferation, colony formation and tumorigenic potential in nude mice, and the induction of G2/M phase arrest and cell apoptosis. Using tubulin polymerization assays and molecular docking analysis, we found that these DFV-taxoids promoted more rapid polymerization of β-tubulin than paclitaxel. Our data demonstrate that these novel taxoids exhibit stronger antitumor activity in ATC cells than paclitaxel, thereby providing a promising therapeutic strategy for the patients with ATC.


Introduction
From 2000 to 2011, the incidence rate of thyroid cancers has risen the fastest, and it is the fourth most common cancer in China based on the latest statistics [1]. Anaplastic thyroid cancer (ATC) accounts for 1-2% of all thyroid cancers. Although rare, it is a major cause of death in thyroid cancer patients due to its aggressive behavior and resistance to treatment [2]. Thus, there is an urgent need to develop effective therapeutic strategies against ATC.
Paclitaxel is one of the most active single chemotherapeutic agents used clinically for the treatment of head and neck cancers, including thyroid cancer [3]. Paclitaxel promotes tubulin polymerization, causing cell cycle arrest and apoptotic death by disrupting normal microtubule dynamics required for cell division and vital interphase processes [4]. Despite their potent antitumor activity, paclitaxel can cause undesirable side effects and drug resistance [5]. Thus, it was apparent in the early 1990s that it would be essential to develop new taxoids with fewer side effects, enhanced activity, improved solubility, and superior pharmacological properties [6].
Due to the limited availability and side effects of paclitaxel, numerous taxoids were designed and synthesized based on the structure of paclitaxel in the past decades [6,7]. Ojima group developed a series of novel generation taxoids with systematic modifications at the C2, C10, C3′, and C3′ N positions of the tetracyclic diterpene skeleton, 10-deacetylbaccatin (DAB) [7,8]. There was also a study to develop taxoids based on 14 β-hydroxy-10-deacetylbaccatin (14-OH-DAB) [9]. A key feature of the second-generation taxoids is their possession of an isobutenyl or isobutyl group at the C3′ position in place of a phenyl group, which has been demonstrated to enhance potency, especially the enhancement of tubulin polymerization [7,9]. In recent years, novel taxanes have been developed, presenting a new generation of taxoids [10]. Some of them have been found to be more effective than paclitaxel in the treatment of different types of cancer [11].
In this study, we systematically evaluated therapeutic potential of five novel 3′-difluorovinyltaxoids (DFV-taxoids) in ATC cells by a series of in vitro and in vivo experiments.

ATC cell lines
Human thyroid cancer cell lines 8505C and 8305C were kindly provided from Dr. Haixia Guan (Guangdong Provincial People's Hospital, Guangzhou, China) and routinely cultured at 37°C in RPMI-1640 medium with 10% fetal bovine serum (FBS). Paclitaxel and novel taxoids were dissolved in dimethyl sulfoxide (DMSO), aliquoted and stored at −80°C until use. The concentration of DMSO was 0.1% as vehicle control.

Cell proliferation assay
Cells were treated with paclitaxel or taxoids at ten concentrations (0 nM, 0.01 nM, 0.05 nM, 0.1 nM, 0.5 nM, 1 nM, 2 nM, 5 nM, 10 nM and 100 nM) and vehicle control (0.1% DMSO) for 48 h, and the MTT assay was then performed to assess their effect on cell proliferation in a dosedependent manner. Their IC 50 values were then calculated as described previously [12]. Next, we treated 8505C or 8305C cells with 5 nM taxoids for 0 h, 6 h, 12 h, 24 h, 48 h and 72 h to determine their effect on cell proliferation in a time-dependent manner.

The LDH release assay
Cells were first treated with paclitaxel or taxoids at different concentrations (0.5 nM, 1 nM, 2 nM, 5 nM and 10 nM) and time points (0 h, 6 h, 12 h, 24 h, 48 h and 72 h), and the LDH levels were measured to evaluate their in vitro cytotoxicity according to the manufacturer's protocol.

Colony formation assay
Cells were treated with various concentrations of paclitaxel or taxoids for 48 h, followed by culturing in RPMI-1640 medium with 10% FBS for 5-6 days. Colonies were then fixed with methanol and stained with crystal violet. Each experiment was performed in triplicate.

Cell cycle assay
Cells were serum starved for 12 h, and then treated with 50 nM paclitaxel or taxoids and vehicle control for the indicated time. Next, cells were harvested and fixed in icecold 70% ethanol for at least 4 h. After washing twice with PBS, propidium iodide (PI) solution (50 μg/mL PI, 50 μg/ mL RNase A, 0.1% Triton-X, 0.1 mM EDTA) was added to stain the cells. Cells were then subjected to flow cytometer analysis to determine cell cycle distributions.

Cell apoptosis assay
Cells were treated with 50 nM paclitaxel or taxoids and vehicle control for 24 h and then stained with Annexin V-FITC/PI Apoptosis Detection Kit according to the manufacturer's protocol. Apoptotic cells were measured by flow cytometry. Each experiment was run in triplicate.

Western blot analysis
Cells were treated with 50 nM paclitaxel or taxoids and vehicle control for the indicated times. Next, cells were lysed, and the lysates were subjected to western blot analysis. Antibody information was shown in Table S1.

Protein expression and purification
Human TUBB2B gene was cloned to pET-28a plasmid and transformed into E. coli BL21 strain. After induced by 0.2 mM IPTG for 6 h, bacterial cells were harvested and lysed by ultrasonication. The fusion protein TUBB2B was expressed in inclusion bodies. After complete denaturation by 8 M urea, the protein was rapidly diluted by seven-fold Tris-buffer followed by dialysis with the same buffer for four times. After folding, the protein was purified by Histag with a Ni-NTA column, desalting and concentrated by a Millipore ultrafiltration centrifuge tube. The expressed and purified proteins were analyzed by SDS-PAGE and RP-HPLC. RP-HPLC was performed at 40°C on a Waters XBridge C18 column (4.6 × 150 mm, 3.5 μm) running a 45 min, 5-95% linear gradient of acetonitrile in water containing 0.1% TFA at a flow rate of 1 mL/min.

Tubulin polymerization assay
Tubulin polymerization assay was carried out with 10 µM paclitaxel, taxoids and vehicle control in MES buffer containing 1 mM EGTA, 0.5 mM MgCL 2 , 1 mM GTP, and 10% glycerol. The reaction system was kept at room temperature before addition of tubulin and shifting to 37°C to react [13]. Assembly of TUBB2B proteins was monitored spectrophotometrically by recording changes in turbidity at 350 nm at 37°C [14,15].

Molecular docking
Molecular docking of the DFV 3rd-generation taxoids was carried out using AutoDock Vina [16] and the available cocrystal structure of paclitaxel in β-tubulin (PDB: 1JFF). Structures of the new taxoids were adapted from the coordinates of paclitaxel in 1JFF, which conserves the baccatin core and relevant ring conformations. Modified groups were minimized using the Avogadro molecule builder [17] and relaxed under the MMFF94 forcefield [18][19][20][21]. Partial charges of each atoms were calculated using the Gastieger method [22,23] and non-polar hydrogen atoms were merged to their respective heavy atoms. Calculations were done using a grid box size of 40 A 3 and the default scoring function for AutoDock Vina. Visualization of the docked poses were done using the UCSF Chimera [24], where the T-taxol-type [25] structure was identified from a total of 20 docking solutions for each molecule.

Animal studies
Five-to six-week-old female athymic nude mice were purchased from SLAC Laboratory Animal Co. Ltd. (Shanghai, China), and housed in a specific pathogen-free environment. Tumor xenografts were established by subcutaneously injecting 8505C (5 × 10 6 ) cells into right iliac fossa region of nude mice. When the tumors grew to~5 mm in diameter, the mice were randomly divided into three groups (five mice per group), and were then administered with vehicle control (<1%) and 5 mg/kg paclitaxel or taxoid SB-T-1285406 by intraperitoneal injection three times a week. Tumor volumes were measured by a Vernier Caliper every other day and calculated by the following formula: Tumor volume (mm 3 ) = length × width 2 × 0.5. All mice were sacrificed 4 h after the last treatment, and tumors were harvested and weighted.
Xenograft tumors were fixed in 10% formalin for 48 h, embedded in paraffin, sectioned at 5 μm, deparaffinized and rehydrated. Immunohistochemistry (IHC) assay was performed to evaluate Ki-67 levels in these tumors. Hematoxylin-Eosin (H&E) staining was used to observe histologic structure and cell morphology of hepatic and renal tissues. The serum levels of alkaline phosphatase (AKP), aspartate transaminase (AST), alanine transaminase (ALT), blood urea nitrogen (BUN), creatinine (CRE) were measured by spectrophotometry according to the manufacturer's protocol. The TUNEL assay was used to determine cell apoptosis in tumor tissues by using the TUNEL Andy Fluor ™ 488 Apoptosis Detection Kit according to the manufacturer's protocol. All animal experiments were conducted and approved by the Laboratory Animal Center of Xi'an Jiaotong University.

Statistical analysis
Student's t test, two-way ANOVA with Bonferroni post-test and one-way ANOVA with Dunnett's post-test were used to compare the data between two groups. The data were expressed as mean ± standard deviation (SD). P < 0.05 was considered statistically significant. All statistical analyses were conducted using the SPSS statistical package (16.0, SPSS Inc. Chicago, IL).

Taxoids inhibit the proliferation and colony formation of ATC cells
To determine in vitro growth response of ATC cells to these novel taxoids (Fig. 1a) and paclitaxel, we first treated 8505C and 8305C cells with increasing concentrations of these five taxoids and paclitaxel from 0.01 nM to 100 nM for 48 h to calculate IC 50 values. The results showed that these taxoids and paclitaxel significantly inhibited the proliferation of ATC cells in a dose-dependent manner, and the taxoids were more effective than paclitaxel with a 2-9-fold increase in sensitivity (Fig. 1b). Next, we treated 8505C and 8305C cells with 5 nM of taxoids and paclitaxel for 0 h, 6 h, 12 h, 24 h, 48 h and 72 h to determine their effect on cell proliferation in a time-dependent manner. The results further supported that these taxoids more significantly inhibited cell proliferation than paclitaxel (Fig. 2). Besides, we treated these cells with different concentrations of taxoids and paclitaxel, and performed colony formation assay. The results showed that these taxoids exhibited more potent inhibitory effect on colony formation of ATC cells than paclitaxel (Fig. 3). Altogether, our data indicate that these novel taxoids are more effective cytotoxic antitumor agents than paclitaxel against ATC cells.
Next, we treated 8505C and 8305C cells with paclitaxel or taxoids at different concentrations and time points, and the LDH levels were then measured to evaluate their in vitro cytotoxicity. The result demonstrated that the LDH release of each taxoid was similar to paclitaxel (Fig. S1), indicating that these taxoids are safe enough.

Taxoids induce G2/M phase arrest in ATC cells
Paclitaxel, one of the broadest-spectrum antitumor agents, performs its anti-mitotic and anti-cancer activity by targeting microtubule [26]. The consequent arrest of cell cycle at mitotic phase has been considered the cause of paclitaxelinduced cytotoxicity [27]. In this study, we treated 8505C and 8305C cells with 50 nM paclitaxel or five taxoids and vehicle control for 48 h, and analyzed their cell cycle distributions. As shown in Fig. 4a and Fig. S2, both paclitaxel and the taxoids caused a significant G 2 /M phase arrest compared to the control. The percentage of G 2 /M phase cells increased from 23.66 ± 2.01% to 38.99 ± 1.10% in paclitaxel-treated 8505C cells, while the percentage of G 2 / M phase cells increased from 23.66 ± 2.01% to 65.02 ± 1.62%, 66.02 ± 3.07%, 59.77 ± 1.28%, 62.59 ± 2.29% and 62.14 ± 1.13% in taxiods-treated 8505C cells relative to DMSO-treated 8505C cells. Similarly, paclitaxel treatment increased the percentage of G 2 /M phase cells from 21.58 ± 1.59% to 26.27 ± 1.19% in 8305C cells, while taxiods treatment increased the percentage of G 2 /M phase cells from 21.58 ± 1.59% to 46.29 ± 0.80%, 46.23 ± 4.53%, 41.51 ± 1.31%, 46.68 ± 0.20% and 45.59 ± 0.94% compared to the control. These results indicate that these novel taxoids have a more prominent advantage in inducing cell cycle arrest than paclitaxel, further supporting the above conclusion.
Considering that paclitaxel and taxoids can induce G 2 /M phase arrest of ATC cells, thus we determined their effect on the expression of cell cycle regulatory components at the G 2 /M boundary, such as p-Rb, Weel, Cyclin B1, Cyclin E, CDK1 and p21. First, we treated 8505C and 8305C cells with 50 nM paclitaxel or SB-T-1285406, one representative of these five taxoids, for 24 h and 48 h, respectively. Cells were lysed and subjected to western blot analysis. The results showed that paclitaxel or SB-T-1285406 treatment resulted in a marked decrease in the expression of p-Rb, Weel, Cyclin B1, Cyclin E, CDK1 and p21 both in 8505C and 8305C compared to the control, particularly after 48 h treatment (Fig. 4b). These results indicate the G 2 /M arrest induced by paclitaxel or SB-T-1285406 is closely associated with a marked alteration in the expression of G 2 /M cell cycle regulatory proteins.

Taxoid SB-T-1285406 inhibits tumor growth in nude mice
To determine in vivo antitumor activity of SB-T-1285406, we established 8505C cell-derived tumor xenografts in nude mice and treated these mice with vehicle control, 5 mg/kg paclitaxel or SB-T-1285406 for every other day for Fig. 2 Inhibitory effect of taxoids and paclitaxel on the proliferation of ATC cells. ATC cell lines 8505C (a) and 8305C (b) were treated with 5 nM of paclitaxel or taxoids and vehicle control for the indicated times, and the MTT assay was performed to evaluate cell proliferation. ***,P < 0.001 2 weeks. As shown in Fig. 6a, both SB-T-1285406 and paclitaxel significantly inhibited the growth of xenograft tumors compared to the control, and the former showed better effect than the latter. At the end of the experiments, the tumors were isolated and weighted. The results showed that mean tumor weight was significantly lower in SB-T-1285406-and paclitaxel-treated groups compared to control group, and SB-T-1285406 treatment exhibited better inhibitory effect than paclitaxel treatment (Fig. 6b).
Ki-67 is present in all proliferating cells, but absent in resting cells, thus it serves as a proliferation marker [28]. In this study we evaluated Ki-67 levels in xenograft tumors by IHC assay. As shown in Fig. 6c, the percentage of Ki-67 positive cells was significantly decreased in SB-T-1285406and paclitaxel-treated xenograft tumors compared to control tumors. Similarly, SB-T-1285406 showed more potent inhibitory effect than paclitaxel. Next, we also evaluated cell apoptosis in xenograft tumors by TUNEL assay. As shown in Fig. 6d, the number of apoptotic cells was significantly increased in SB-T-1285406-and paclitaxeltreated tumors compared to control tumors. Similar to in vitro results, we did not find significant difference between SB-T-1285406-and paclitaxel-treated tumors. Importantly, there was no significant difference with respect to serum levels of AKP, ALT, AST, BUN, CRE and H&E staining of hepatic and renal tissues among three treatment groups ( Fig. 6e and Fig. S3). Besides, SB-T-1285406 and paclitaxel treatment almost did not affect body weight of mice compared to the control (Fig. S4). Collectively, our data indicate that novel taxoids such as SB-T-1285406 have more potent antitumor activity than paclitaxel in ATC cells, and exhibit sufficient safety like paclitaxel.
Taxoids promote more rapid polymerization of tubulin than paclitaxel We next attempted to reveal potential mechanism by which novel taxoids are superior to paclitaxel. It is well-known  [29]. Thus, we first cloned human TUBB2B gene to pET-28a plasmid and transformed it into E. coli BL21 strains to express TUBB2B proteins (Fig. S5). After purifying these proteins, we evaluated the activities of these five taxoids and paclitaxel by in vitro tubulin polymerization assay. This assay will provide a direct measure of turbidity, indicating the speed and extent of tubulin polymerization. As shown in Fig. 7a, these taxoids universally promoted rapid polymerization of β-tubulin at a faster rate than paclitaxel. The turbidity of tubulin solution treated by these taxoids quickly reached a plateau relative to paclitaxel. Briefly, it took about 10 mins for paclitaxel to reach the plateau, while for taxoids, it took only 3-5 mins to reach the peak. These observations imply that there is a difference in structure between microtubules formed with these taxoids and those with paclitaxel.
Paclitaxel exerts potent antitumor activity by promoting tubulin polymerization and inhibiting depolymerization of formed microtubules via its binding to β-tubulin components of microtubules [30,31]. To explain why the ability of these taxoids binding to microtubules are better than paclitaxel, we performed molecular docking analysis using paclitaxel-bound α, β-tubulin crystal structure 1JFF from protein database (PDB) (http://www.rcsb.org/) [30]. In all cases, the "T-taxol"-type binding mode was found as a reasonable solution. For example, an excellent overlay of paclitaxel and SB-T-1285406 shown in Fig. 7b (1) exhibited a hydrogen bond between the NH of Gly362 and the C2'-OH group of paclitaxel and the DFV-taxoid with distances of 2.69 Å and 2.39 Å, respectively. A previous study has indicated the expected binding orientation of the C2 benzoate substituents [32]. The unique property of the difluoromethoxy group results in the orientation that is placed in a hydrophobic binding site, which reduces solvent exposure and bolsters the protein-drug contact interface. Moreover, Fig. 7b (2) showed the interactions of the difluoromethoxy group of SB-T-1285406 with His229 and Phe272. The difluoromethoxy group is oriented to make van der Waals contacts with the proximal surface of the protein consisting of Phe272 while simultaneously forming a hydrogen bond (OCF 2 H---N) with the delta nitrogen of His229.
The introduction of the DFV group did not change the overall "T-taxol"-type conformation. However, there were DFV-specific interactions of the two fluorine atoms with surrounding amino acid residues, Ser236, Ala233 and Phe272, which paclitaxel did not have (Fig. 7b (3)). The introduced fluorine atoms exhibited a broader interaction profile than the phenyl group presents in paclitaxel, allowing it to act as a hydrogen bond acceptor with the hydroxyl group of Ser236 and also interacting with nearby hydrophobic amino acid residues such as Ala233 and Phe272.  Since the two fluorine atoms of the 3′-DFV group was located substantially closer to Phe272, as compared to the 3′-phenyl group of paclitaxel, this van der Waals interaction would make the binding of the DFV-taxoid tighter than paclitaxel. Heavy atom distances (i.e., C---F) of the DFV group of SB-T-1285406 with Ser236, Ala233, and Phe272 were 2.87 Å, 3.36 Å, and 3.51/3.88, Å, respectively (Fig.  7b (3)).
In general, both organofluorine groups are observed to interact with the deep binding pocket behind the protruding residue His229, to which all of DFV-taxoids wrap around. The electrostatic surface representation (Fig. 7b (4)) showed the hydrophobicity of this pocket composed of Phe272 and Ala236 to which these organofluorine moieties can more tightly and favorably bind, as compared to paclitaxel. The above findings suggest that these unique interactions may . Data are presented as means ± SD. Scale bars, 200 μm. d TUNEL assay was performed to evaluate cell apoptosis in xenograft tumors from DMSO-, SB-T-1285406-and paclitaxel-treated mice. Green color represents target TUNEL staining, and blue color represents Hoechst33342 staining for nuclei. Scale bars, 100 μm. e Serum activity of alkaline phosphatase (AKP), aspartate transaminase (AST), alanine transaminase (ALT), blood urea nitrogen (BUN) and creatinine (CRE) were measured by spectrophotometry in the indicated tumors. *,P < 0.05; **,P < 0.01;***,P < 0.001 be essential for enhancing the ability of these novel taxoids binding to β-tubulin, thereby contributing to their better anti-cancer effect than paclitaxel.

Discussion
Although rare, ATC is the most lethal histotype of thyroid cancer [33]. ATC patients generally have a palpable, size variable and irregular rigid mass in the thyroid with or without hoarseness, dysphagia and dispones. Clinical pathology of ATC is characterized by rapid cell growth and the ability to metastasize to other organs of the body, especially lungs, bones and brain. These features contribute to its high mortality. At present, the principle of ATC treatment is total thyroidectomy combined with postoperatively systemic chemotherapy, radiotherapy and other therapeutic strategies. Although multimodality strategies have been demonstrated to be better than monotherapy, the vast majority of ATC patients still have a very poor survival because they are unresponsive to conventional antitumor agents. Thus, it is urgent to develop more effective therapeutic drugs for ATC treatment.
For the last decades, paclitaxel emerged as one of the most important compounds broadly used in the therapy of breast, ovarian, lung, head, and neck cancers [4]. A phase 2 clinical trial first proposed the efficacy of continuous infusion paclitaxel every 3 weeks for 1-6 cycles in 20 patients with ATC [34]. Besides, another study analyzed the clinical data of 13 patients undergoing paclitaxel treatment every week, and finally observed 1 patient as complete response (CR) [35]. These observations indicate that paclitaxel treatment may be beneficial to prolong the survival of ATC patients. Meanwhile, there are studies demonstrating that antitumor effect of paclitaxel in ATC can be further enhanced when combined with lenvatinib, sorafenib or Fig. 7 The effect of DFV-taxoids and paclitaxel on tubulin polymerization in vitro. a Tubulin polymerization assay was carried out in a system containing 1 mg/mL β-tubulin, 1 mM GTP, 10 µM taxoids or paclitaxel, the assembly of human microtubule protein TUBB2B was monitored spectrophotometrically by recording changes in turbidity at 350 nm every min. b Molecular docking of paclitaxel and DFVtaxoids with β-tubulin. (1) Overlay of SB-T-1285406 (cyan) and paclitaxel (light gray) in β-tubulin, showing a conserved T-taxol binding mode. (2) The van der Waals contacts of the CHF 2 O group in the C2 benzoate with nearby hydrophobic amino acid residues, Ala233 and Phe272. (3) The van der Waals contacts of the C3′-DFV group with nearby amino acid residues, Ala233, Ser236, and Phe272. (4) Electrostatic surface model, illustrating the hydrophobic interface of β-tubulin where the DFV and tri/difluoromethoxy groups are located radiotherapy [36,37]. However, traditional paclitaxel often causes serious side-effects besides drug resistance. Thus, it is essential to develop new taxoids with better pharmacological properties and improved activity in order to overcome the problems with traditional paclitaxel in cancer therapy. Evidently, new taxanes can effectively control advanced cancers including ATC [35][36][37][38][39], indicating that it may be a promising therapeutic strategy to prolong the survival of ATC patients without serious toxic side-effects.
In this study, we evaluated five novel DFV-taxoids (SB-T-1285106, SB-T-1285205, SB-T-1285406, SB-T-1285506, and SB-T-1285605), which were modified at the C2, C10 positions based on chemo-structure of paclitaxel. We next compared their antitumor activity with paclitaxel by a series of in vitro and in vivo experiments. The results showed that all of these taxoids exhibited exceptionally high potency than paclitaxel against ATC cells, as reflected by their lower IC 50 values and more potent inhibition of cell proliferation, colony formation and tumorigenic potential in nude mice. Besides, these DFV-taxoids also exhibited more apparently G 2 /M phase arrest in ATC cells compared to paclitaxel. More importantly, we did not observe serious toxic side-effects, indicating that these DFV-taxoids such as SB-T-1285406 is a safe and effective agent for ATC therapy.
To clarify the mechanism of these taxoids showing better antitumor effect than paclitaxel, we performed in vitro tubulin polymerization assay, and found that these taxoids induced GTP-independent tubulin polymerization faster than paclitaxel, implying that there exists difference in structure between microtubules formed with these taxoids and paclitaxel. These results were in consistent with our previous study [8], besides the difference in binding rates, our previous study also mentioned rapid tubulin polymerization of new generation taxoids producing numerous short microtubules compared with paclitaxel, which may also be another difference between the DFV-taxoids and paclitaxel. This may be the main reason why these taxoids exhibit potent antitumor effect than paclitaxel.
To further look into the mechanism behind this difference, we performed molecular docking to analyze their interaction with β-tubulin. In fact, we have found that the DFV-taxoids possess unique and specific interactions with β-tubulin through organofluorine groups in the molecule, which do not exist in paclitaxel. The difluoromethoxy group of SB-T-1285406 interacts with His229 and Phe272 (Fig.  7b (2)). The difluoromethoxy group makes van der Waals interactions with the proximal surface of the protein consisting of Phe272, while simultaneously forming a hydrogen bond (OCF 2 H---N) with the delta nitrogen of His229. In the case of trifluoromethoxy-taxoids, one of the fluorine atoms can form a hydrogen bond with the amine hydrogen of His229 (NH---F-CF 2 CO) (Fig. S6). The introduction of the DFV group did not change the overall "T-taxol"-type conformation, while there are DFV-specific interactions of the two fluorine atoms with surrounding residues Ser236, Ala233 and Phe272, which paclitaxel does not have (Fig. 7b  (3)). Since the 3′-DFV group is located substantially closer to Phe272, as compared to the 3′-phenyl group of paclitaxel, this van der Waals contact would make the binding of the DFV-taxoid tighter than paclitaxel. As the electrostatic surface representation in Fig. 7b (4) indicates, the hydrophobicity of this pocket allows the tighter and more favorable binding of these organofluorine moieties, as compared to paclitaxel.
In summary, by a series of extensive in vitro and in vivo experiments, we demonstrate that these five novel DFVtaxoids exhibit promising antitumor effect in ATC cells. They exhibit better antitumor activity than paclitaxel, thereby causing shorten therapeutic periods and less sideeffects during several cycles of chemotherapy. Thus, our data indicate that these taxoids as monotherapy or in combination with other antitumor agents will provide a relatively safe and effective strategy for the treatment of ATC patients.

Data availability
The data included in this study are available on request from the corresponding authors.