Design, Synthesis, and In Vitro and In Vivo Biological Studies of a 3′-Deoxythymidine Conjugate that Potentially Kills Cancer Cells Selectively

Thymidine kinases (TKs) have been considered one of the potential targets for anticancer therapeutic because of their elevated expressions in cancer cells. However, nucleobase analogs targeting TKs have shown poor selective cytotoxicity in cancer cells despite effective antiviral activity. 3′-Deoxythymidine phenylquinoxaline conjugate (dT-QX) was designed as a novel nucleobase analog to target TKs in cancer cells and block cell replication via conjugated DNA intercalating quinoxaline moiety. In vitro cell screening showed that dT-QX selectively kills a variety of cancer cells including liver carcinoma, breast adenocarcinoma and brain glioma cells; whereas it had a low cytotoxicity in normal cells such as normal human liver cells. The anticancer activity of dT-QX was attributed to its selective inhibition of DNA synthesis resulting in extensive mitochondrial superoxide stress in cancer cells. We demonstrate that covalent linkage with 3′-deoxythymidine uniquely directed cytotoxic phenylquinoxaline moiety more toward cancer cells than normal cells. Preliminary mouse study with subcutaneous liver tumor model showed that dT-QX effectively inhibited the growth of tumors. dT-QX is the first molecule of its kind with highly amendable constituents that exhibits this selective cytotoxicity in cancer cells.


Introduction
With cancers being the leading cause of death world-wide, developing safe and effective anticancer agents remains in urgent need. Molecularly targeted therapy has been the recent focus for anticancer drug development, as seen in the example of Sorafenib [1,2]. Sorafenib is a multi-tyrosine kinase inhibitor that can potentially minimize adverse effects such as hepatotoxicity caused by other anticancer drugs including 5-fluorouracil and doxorubicin [3,4]. Similar to tyrosine kinases, thymidine kinases (TKs) have been considered another potential anticancer target [5][6][7][8]. TKs are the first phosphorylating enzymes in the thymidine salvage pathway converting thymidine to its 59-phosphate form for DNA synthesis. In normal mammalian cells, cytosolic TKs are only present at a low level in the S phase of cells; whereas elevated level of TKs have been observed in virus infected cells and rapidly proliferating cancer cells [5][6][7][8], e.g., lung tumors and breast cancer tissues [9,10].
Nucleobase analogs targeting TKs such as AZT and acyclovir have been shown effective antiviral activity [11,12]. However, poor cancer-selectivity and strong side effects have been associated with nucleobase analogs including neutron capture agent 5thymidine boron conjugate and radioisotopic iodinated deoxyuridine [13,14]. Recently, a combination therapy using predelivered thymidine kinase followed by nucleobase analogs has been investigated in liver cancer patients with limited effects achieved [15]. This is likely due to that TKs oriented nucleobase analogs such as AZT are quickly removed by nucleobase repair processes after they are incorporated in the DNA synthesis of cancer cells via the thymidine salvage pathway [16][17][18]. 5-Fluorouracil, another thymine analog, is a dihyrofolate reductase inhibitor and directly causes cytotoxicity in normal hepatocytes [3,19]. Therefore, alternative design of more effective nucleobase analogs targeting TKs is needed.
We report here a novel 39-deoxythymidine phenylquinoxaline conjugate (dT-QX, Figure 1) that selectively kills a variety of cancer cells, but not normal hepatocytes. Structurally, dT-QX is a 39-triazole-39-deoxythymidine linked to a DNA intercalating quinoxaline moiety. dT-QX was designed to target the thymidine salvage pathway based on the fact that 39-triazole thymidine derivatives are recognized by human thymine kinases [20,21]. The purpose of adding the quinoxaline moiety into dT-QX structure was to block cellular removal of thymidine analogs [16][17][18] via DNA intercalation in the DNA synthesis of cancer cells, because quinoxaline moiety is a known DNA intercalator [22]. In addition, the quinoxaline structure can be conveniently synthesized by one step condensation of a diketone compound 1 [23] and an ortho-phenylenediamine ( Figure 1). More importantly, the quinoxaline structure is highly amendable for chemical modifications with a variety of substituents for advanced structure activity study to optimize the potential biological activity. The anticancer activity of dT-QX was then evaluated using a panel of cancer cells. The effects of dT-QX on inhibiting DNA synthesis and inducing cellular oxidative stress were also investigated. Finally, the inhibition of tumor growth by dT-QX was assessed in mice bearing subcutaneous liver tumors.

Biological Studies
The animal protocol was approved by the Reviewing Committee of College of Life Science and Technology at Huazhong University of Science and Technology. Cancer cell lines HepG2, Hep3B, MCF-7 and C6 were obtained from ATCC (VA, USA). Human liver cells HL-7702 and murine liver cancer cells H22 were from Shanghai Institute of Life Science Cell Culture Center (Shanghai, China). Cells were maintained in high glucose DMEM or RPMI-1640 medium (Invitrogen, CA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 25 mM HEPES, 2 mM L-glutamine, 0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, 50 U/mL penicillin, and 50 mg/mL streptomycin at 37uC and 5% CO 2 . Stock solutions of dT-QX, Pip-QX or AZT were prepared in DMSO, and doxorubicin hydrochloride salt was directly dissolved in water. All biological chemicals were obtained from Sigma Aldrich (WI, USA) unless specified otherwise. All of the experiments were independently repeated at least three times.
Cell viability MTT assay. Cells (5,000 per well) were seeded on 96-well plates in growth media overnight before each treatment in triplicates. dT-QX, Pip-QX or AZT was at a final concentra-tion of 0, 0.1, 1, 10, 20, 50, 100, or 200 mM with total DMSO less than 0.2%. Doxorubicin was used as a comparison at a final concentration of 0.05, 0.1, 0.2, 0.5, 1.0 or 2.0 mM. After 72 h, MTT assay was carried out as reported [23]. The cell viability of each treatment was plotted with GraphPad Prism program, and IC 50 values were obtained using sigmoidal dose-response analysis provided by the software (version 4.00, GraphPad Software, CA, USA).
Anti-BrdU fluorescence assay. Hep3B, HepG2 or HL-7702 cells (50,000 per well) were seeded in the growth media overnight on 48-well plates and treated with 0, 10 or 50 mM dT-QX for 5 h. Anti-BrdU assay was carried out according to the manufacturer's recommendation (BD Biosciences, NJ, USA). Briefly, BrdU solution (0.1 mg/mL) was added to each well, and cells were incubated at 37uC for 3 h. Cell medium was removed and cells were fixed with 3.7% formaldehyde in PBS. Cells were permeablized with 0.1% triton-100 in PBS and blocked with 3% FBS in PBS solution. Cellular DNA was denatured by DNaseI (0.3 mg/mL) in PBS solution. The incorporated BrdU was stained with Alexa FluorH488 anti-BrdU monoclonal antibody (BD Biosciences, NJ, USA). The nuclei were counter-stained with Hoechst 33342 solution. Cell images of each treatment were    Subcutaneous liver tumor study in mice. Murine liver cancer cells H22 were initially maintained in the RPMI-1640 growth medium and then grow in the BALB/c mice (Hubei Provincial Laboratory Animal Center, China) intraperitoneally. Mice were euthanized after 4 days and H22 cells were harvested with PBS solution. H22 cells were washed once with sterile PBS and were injected subcutaneously (3x10 6 cells per mouse) at the lower back of naive BALB/c mice. Once tumors reached an average size of (8x8 mm, 340 mm 3 ), nine mice were randomly divided into three groups (3 per group) and injected with 100 mL of saline (with 0.1% DMSO), AZT (50 mM in sterile PBS with 0.1% DMSO) or dT-QX (50 mM in sterile PBS with 0.1% DMSO) on the tumor site on day 1 and day 7. The tumor growth and body weight were monitored daily. On day 12 followed the first injection, all mice were euthanized and images of tumors were recorded. Statistic analysis (one way ANOVA with Tukey's multiple comparison test) of the treatments was performed using GraphPad Prism software.

Results and Discussion
Synthesis of dT-QX was achieved by coupling AZT with phenylquinoxaline 4 via copper(I)-catalyzed click reaction ( Figure 2). Intermediate 3 was obtained from nucleophilic substitution of bromoalkyne 2 with oxidized terpenone 1. Phenylquinoxaline 4 was obtained by condensation with odiaminobenzene and silica gel under reflux in toluene, which was much better than acetonitrile, DMF or ethanol. The conjugation of AZT with phenylquinoxaline 4 to dT-QX was accomplished in 75% yield using the in situ reduction of copper(II) by ascorbate [20,21]. Phenylquinoxaline 4, a reference molecule, was found to be poorly soluble in 1% DMSO aqueous solution and hence was modified to a N-methylpiperazine derivative Pip-QX as shown in Figure 2. Pip-QX was synthesized in a manner similar to dT-QX via conversion of 1 to the piperazine derivative 5, followed by condensation with o-diaminobenzene in an overall yield of 61%. Pip-QX served as one of reference compounds in following biological studies. Multiple batches of dT-QX and control compounds were synthesized, and all the batches performed consistently not only in chemistry but also in biological studies.
The biological activity of dT-QX was first assessed using cell viability assay on a panel of cancer cell lines including human liver carcinoma HepG2 and Hep3B, mouse liver carcinoma H22, breast adenocarcinoma MCF-7, and rat brain glioma C6 cells. Human liver HL-7702 cell line was used as a representative of normal cells in the study. HL-7702 cells are transformed human normal hepatocytes with low expression of cancer markers [24]. Treatment of HL-7702 with dT-QX did not resulted in any significant cytotoxicity at concentrations as high as 200 mM (Figure 3a). The EC 50 of dT-QX on all cancer cells was found to range from 6.6 to 42.1 mM. The most pronounced cytotoxicity was observed on human hepatocellular carcinoma Hep3B cells with more than 80% cell death at 20 mM, followed by breast adenoma MCF-7 and brain glioma C6 cells. In stark contrast, Pip-QX non-selectively killed all cell lines including HL-7702 at 50 mM (Figure 3b). The second reference compound, AZT as a 39-deoxythymidine analog, was found to exhibit low cytotoxicity against these cell lines (Figure 3c), while co-treatment of AZT plus Pip-QX also non-selectively killed all cell lines (Figure 3d), similar to that of Pip-QX treatment alone. These results suggested that the selective killing of cancer cells over normal liver cells by dT-QX was due to the unique covalent conjugation of cytotoxic phenylquinoxaline moiety with 39-deoxythymidine. In comparison, the anticancer drug doxorubicin was found highly toxic to all these cells. In fact, doxorubicin was even more toxic toward normal liver 7702 cells than liver cancer Hep3B or H22 cells at concentrations above 0.5 mM (Figure 3e).
To further understand the selective cytotoxicity of dT-QX towards cancer cells, we first investigated cell proliferation using anti-BrdU fluorescence assay. BrdU (5-bromo-39-deoxyuridine) is a thymidine analog that is incorporated into cell genome as it replicates [25]. Thus, the level of BrdU in cell nucleus reflects the level of cell division and proliferation. Considering the above cell viability results, human HL-7702, HepG2 and Hep3B cells were investigated as representatives of normal and cancer cells. The anti-BrdU assay was carried out at 5 h after dT-QX treatment to allow enough accumulation of compound in cells and to determine its impact on cellular DNA synthesis. The level of BrdU in cell nuclei was measured using a fluorescent anti-BrdU conjugate [25] (green color) with cell nuclei counter-stained by Hoechst 33342 (blue color) as shown in Figure 4. In normal human HL-7702 cells, treatments of DMSO and dT-QX resulted in similar levels of BrdU incorporation in cell nuclei, which indicated the presence of dT-QX did not interfere with cell division and proliferation. However, significant loss of green fluorescence was observed in cancerous HepG2 and Hep3B cells with dT-QX treatment as compared to that of DMSO control (Figure 4a). These results indicated that dT-QX selectively inhibited cellular DNA synthesis of HepG2 and Hep3B cells at the early stage of treatment resulting in selective killing both of cancer cells. The inhibition of BrdU incorporation was much more pronounced on Hep3B cells with the complete loss of green fluorescence in cell nuclei (Figure 4a). This was consistent with cell viability results that dT-QX was more effective at killing Hep3B than HepG2 cells. The more pronounced cytotoxicity on Hep3B cell than HepG2 may be due to the fact that Hep3B is originated from hepatitis B infection [26].
To elucidate which chemophore of dT-QX was responsible for the inhibition of DNA synthesis, the anti-BrdU fluorescence assay was carried out with AZT or Pip-QX on Hep3B cells (Figure 4b). Treatment of AZT resulted in no significant inhibition of DNA synthesis, while Pip-QX significantly inhibited the DNA synthesis in cells, similar to that of dT-QX, indicating phenylquaxinoline chemophore was responsible for the inhibition of DNA synthesis observed. In combination with cell viability results, it suggested that conjugation with 39-deoxythymidine uniquely modified the cytotoxic phenylquinoxaline chemophore to make it more selective toward cancer cells than normal hepatocytes. These results also confirmed that conjugating the quinoxaline moiety with 39-triazole-39-deoxythymine led to selectively targeting cancer cells by blocking their DNA synthesis.
The selective inhibition of nuclear DNA synthesis by dT-QX in cancer cells led to the hypothesis that dT-QX might selectively inhibit the mitochondrial DNA synthesis and cause mitochondrial dysfunction as well. Unfortunately, the anti-BrdU fluorescent assay was not sensitive enough to observe such inhibition in cells. Alternatively, mitochondrial dysfunction due to DNA depletion by nucleoside analogs has been reported to occur concurrently with elevated superoxide level in mitochondria after treatment for 4 days [27]. Therefore, the impact of dT-QX on the mitochondrial function was assessed using a fluorescence based mitochondrial superoxide imaging assay. Hep3B cells were investigated as the representative of cancer cell model because of the pronounced inhibition of DNA synthesis at early stage by dT-QX observed in anti-BrdU assay (Figure 4a).
Significant red fluorescence indicative of superoxide production was detected in Hep3B cells after treatment of 50 mM dT-QX for 8 h as compared to that DMSO ( Figure 5). The cytosolic presence of red staining of superoxide was confirmed by counter-staining cell nuclei with a Hoechst dye. It was also found that treatment of Hep3B cells with dT-QX for a shorter time such as 3 or 5 h did not induce significant red fluorescence, suggesting the production of mitochondrial superoxide was accumulative and was likely due to the inhibition of cellular DNA synthesis. In addition, a positive control with a Mn(III)-chelator acting as a NADPH/GSH:O 2 2 oxidoreductase in cells [28] resulted in a similar level of red fluorescence in Hep3B cells to that treated with dT-QX (see Figure S8). In contrast, low background red fluorescence was observed in human normal HL-7702 cells treated with dT-QX, similar to that treated with DMSO ( Figure 5). Thus, these results suggested that dT-QX could cause mitochondrial dysfunction by inducing mitochondrial superoxide stress in cancer cells following the inhibition of DNA synthesis.
A preliminary in vivo antitumor study with dT-QX was carried out in a mouse model. Subcutaneous tumors were established with murine H22 liver cancer cells [29]. Tumors were allowed to reach an average size of 8.868.8 mm. The relative low solubility of dT-QX in PBS solution limited the route of administration to be intratumor injections. Thus, AZT or dT-QX (100 mL650 mM each) was injected directly into tumors on day 1 and day 7. Regardless of treatments, neither mice died nor was there a significant body weight loss observed during the 12-day study. The tumor growth profile over 12 days after the first injection is shown in Figure 6a. Tumors in mice treated with AZT grew rapidly, similar to those of negative control (saline); whereas the growth of tumors was significantly inhibited in dT-QX treated mice (Figure 6a). On day 12 after the first injection, the sizes of tumors in mice injected with dT-QX were clearly smaller than those with AZT or saline (Figure 6b). Statistic analysis also confirmed that the treatment of dT-QX was significantly different from the other two (P,0.01). These results suggested that intra-tumor injection of dT-QX at a low dosage (equivalent to 0.13 mg/kg per mouse) was quite effective at inhibiting the growth of subcutaneous tumors in a mouse model.
In conclusion, dT-QX as a novel thymidine analog exhibited an excellent selective cytotoxicity toward a variety of cancer cells but not on normal human liver HL-7702 cells. The selectivity was achieved through selective inhibition of cellular DNA synthesis by dT-QX, resulting in mitochondrial superoxide stress which may lead to mitochondrial dysfunction. The covalent linkage with 39-deoxythymidine uniquely directed the cytotoxicity of phenylquinoxaline moiety more toward cancer cells than normal liver cells. Preliminary in vivo study also demonstrates that dT-QX could be a potential drug candidate for anti-cancer agents. dT-QX is the first molecule of its kind that exhibits this high selective cytotoxicity and excellent in vivo potency. Considering the highly amendable structural constituents for better anti-cancer activity and low hepatotoxicity, the potential of dT-QX as a lead compound for anti-cancer drug development is therefore very promising.