Novel Homo-Bivalent and Polyvalent Compounds Based on Ligustrazine and Heterocyclic Ring as Anticancer Agents

Bivalent and polyvalent inhibitors can be used as antitumor agents. In this experiment, eight ligustrazine dimers and seven ligustrazine tetramers linked by alkane diamine with different lengths of carbon chain lengths were synthesized. After screening their antiproliferation activities against five cancer cell lines, most ligustrazine derivatives showed better cytotoxicity than the ligustrazine monomer. In particular, ligustrazine dimer 8e linked with decane-1,10-diamine exhibited the highest cytotoxicity in FaDu cells with an IC50 (50% inhibiting concentration) value of 1.36 nM. Further mechanism studies suggested that 8e could induce apoptosis of FaDu cells through the depolarization of mitochondrial membrane potential and S-phase cell cycle arrest. Inspired by these results, twenty-seven additional small molecule heterocyclic dimers linked with decane-1,10-diamine and nine cinnamic acid dimers bearing ether chain were synthesized and screened. Most monocyclic and bicyclic aromatic systems showed highly selective anti-proliferation activity to FaDu cells and low toxicity to normal MCF 10A cells. The structure-activity relationship revealed that the two terminal amide bonds and the alkyl linker with a chain length of 8–12 carbon were two important factors to maintain its antitumor activity. In addition, the ADMET calculation predicted that most of the potent compounds had good oral bioavailability.


Introduction
Natural products play a vital role in the development of the drug, especially anticancer drugs [1]. Ligustrazine (2,3,5,6-tetramethylpyrazine, TMP), an important component of the Chinese traditional medicinal herb Chuanxiong (Ligusticum chuanxiong Hort), has been of wide clinical use for cardiovascular and cerebrovascular diseases [2]. Recently, TMP has been reported to possess anticancer activity in vitro and in vivo, inducing cancer cell apoptosis and cell cycle modulation [3,4]. More importantly, the introduction of different substituted groups into ligustrazine has helped to improve the antitumor activity. Many ligustrazine derivatives with potent antitumor activity have been developed in the past few years, like monocarbonyl ligustrazine-curcumin hybrids [5], ligustrazine-betulinic acid hybrids [6], and ligustrazine-rhein derivatives [7]. This advancement stimulated our interest in using TMP as the scaffold to synthesize new antitumor agents.
In the interaction between host and guest molecules, both bivalency and polyvalency could enhance the affinity and activity of monovalent ligand [8,9]. This phenomenon has attracted attention in drug design. Many effective bivalent anti-cancer drugs have been found, such as artemisinin-derivative dimers [10], jesterone dimer [11], indole-3-carbinol dimer [12], bis-daunorubicin [13]. Previous work in this field has designed and synthesized several ligustrazine dimers linked by cyclohexanone and oxime [14], curcumin [15], or triterpenes [16], which had a good cytotoxic effect on human cancer cells. The result showed that the dimerization of ligustrazine could obviously improve the antitumor potency of its monomer.
With an effort to developing new antitumor agents, we designed and synthesized a series of novel TMP dimers and tetramers linked with different alkyl diamines and screened their anti-proliferation potential on a panel of human cancer cell lines, including HeLa, Hep G2, , FaDu, and A549. Based on these experimental results, according to the importance of small molecular heterocyclic rings in drug discovery [17], additional dimers series of aromatic rings instead of ligustrazine, a total of 51 new compounds were synthesized, screened, and their structure-activity relationship was briefly discussed. Meanwhile, through morphological observation and flow cytometric analysis, the antitumor mechanisms of the most potent one were preliminarily discussed in this study. ADMET properties of all compounds were also evaluated to explore the drug-likeness.

Chemistry
In this study, 51 novel designed compounds were synthesized. Series 6a-g and 8a-f were obtained according to the synthetic method described in Schemes 1 and 2. The important intermediate 2-chloromethyl-3,5,6-trimethyl-pyrazine 5 was prepared from ligustrazine by the tandem reaction: Boekelheide reaction [18] and deprotection and the chlorination reaction [19]. The key intermediate 2-carboxylic acid-3,5,6-trimethyl-pyrazine 7 was synthesized by the one-pot reaction, as previously reported [20]. To explore the length effect, the TMP moieties, with 2-12 methylene alkyl chain as a linker, were synthesized.
HeLa, , FaDu, A549, and MCF 10A, respectively. In light of these results, we replaced TMP with aromatic heterocycles similar to ligustrazine in electronic space. The novel small aromatic molecule and heterocyclic dimers 9e-35e were synthesized, and their anti-proliferative ability was evaluated. Compounds 9e, 10e, 11e, 17e, 18e, 21e, 22e, 23e, and 25e, aromatic acid dimers (cinnamoyl, nicotinoyl, benzoyl, quinoline-3-carboxyl, quinoline-6-carboxyl, benzofuran-2-carboxyl, benzothiophene-2-carboxyl, indol- 3-carboxyl, chlorinated cinnamoyl), were linked by decane-1, 10-diamine, which had potent anti-proliferative ability in FaDu or A549 cell lines. Their IC50 in FaDu cell lines were less than 1 µM, in the range of 20 nM (benzothiophene- 2-carboxyl) and 697 nM (benzofuran- 2-carboxyl). Also, they had nearly no toxicity in normal cells like MCF 10A. In series 9dn, compounds 9d-f (IC50 < 1 µM), two cinnamic acids linked by C8, C10, C12 chain with two terminal amide bonds, were more effective than compounds 9g-i (IC50 > 20 µM), two cinnamic acids linked by the same alkyl chain with two terminal ester bonds instead of amide bonds. This demonstrated the vital role of two-terminal amide bonds in antitumor activity. Compared with compounds 9d-f, the inhibition of compounds 9j-m, two cinnamic acids linked by the similar length ether chain instead of alky chain, was weaker in tumor cells (IC50 > 20 µM), which confirmed that the alkyl chain played an important role in the structure. Then, we evaluated the position effect of the substitution at the ortho, meta, or para positions of the same aromatic ring. Compounds 25-35e and the unsubstituted compounds 9e and 11e were all derivatives of cinnamic acid or benzoic acid with different substituents and electronic effects. The results of the compounds 26-27e or 28-29e confirmed that the substituents had a general effect on the activity, -OH, -OCH3, -NH2, and other electron donor groups could significantly reduce the activity, with the influence of -OH greater than that of -OCH3. However, the electron attraction group like -Cl (25e) had no obvious effect. The compounds with para-substituted analogs (31e, para position, IC50 = 5.594 µM) exhibited better cell growth inhibitory activities in FaDu cells than ortho-and meta-substituted analogs (30e, 32e, ortho and meta position, IC50 > 20 µM). It is worth noting that para -OCH3 substitution could enhance anticancer activity. The IC50 value of compound 34e was 5.853 ± 0.408 µM, the IC50 value of para-OCH3 was half that of the parent compound 33e, and the IC50 value of compound 33e was 10.393 ± 0.949 µM. Generally, there were three crucial elements in the structure-activity relationship of these systems: alkyl chain linker, two-terminal amide bonds, and heteroaromatic substituents. This SAR (structure-activity relationship) study would help to discover more effective compounds in the future. HeLa, MCF-7, FaDu, A549, and MCF 10A, respectively. In light of these results, we replaced TMP with aromatic heterocycles similar to ligustrazine in electronic space . The novel small aromatic  molecule and heterocyclic dimers 9e-35e were synthesized, and their anti-proliferative ability was  evaluated. Compounds 9e, 10e, 11e, 17e, 18e, 21e, 22e, 23e, and 25e, aromatic acid dimers (cinnamoyl,  nicotinoyl,  benzoyl,  quinoline-3-carboxyl,  quinoline-6-carboxyl,  benzofuran-2-carboxyl,  benzothiophene-2-carboxyl, indol-3-carboxyl, chlorinated cinnamoyl), were linked by decane-1, 10-diamine, which had potent anti-proliferative ability in FaDu or A549 cell lines. Their IC50 in FaDu cell lines were less than 1 µM, in the range of 20 nM (benzothiophene- 2-carboxyl) and 697 nM (benzofuran- 2-carboxyl). Also, they had nearly no toxicity in normal cells like MCF 10A. In series 9dn, compounds 9d-f (IC50 < 1 µM), two cinnamic acids linked by C8, C10, C12 chain with two terminal amide bonds, were more effective than compounds 9g-i (IC50 > 20 µM), two cinnamic acids linked by the same alkyl chain with two terminal ester bonds instead of amide bonds. This demonstrated the vital role of two-terminal amide bonds in antitumor activity. Compared with compounds 9d-f, the inhibition of compounds 9j-m, two cinnamic acids linked by the similar length ether chain instead of alky chain, was weaker in tumor cells (IC50 > 20 µM), which confirmed that the alkyl chain played an important role in the structure. Then, we evaluated the position effect of the substitution at the ortho, meta, or para positions of the same aromatic ring. Compounds 25-35e and the unsubstituted compounds 9e and 11e were all derivatives of cinnamic acid or benzoic acid with different substituents and electronic effects. The results of the compounds 26-27e or 28-29e confirmed that the substituents had a general effect on the activity, -OH, -OCH3, -NH2, and other electron donor groups could significantly reduce the activity, with the influence of -OH greater than that of -OCH3. However, the electron attraction group like -Cl (25e) had no obvious effect. The compounds with para-substituted analogs (31e, para position, IC50 = 5.594 µM) exhibited better cell growth inhibitory activities in FaDu cells than ortho-and meta-substituted analogs (30e, 32e, ortho and meta position, IC50 > 20 µM). It is worth noting that para -OCH3 substitution could enhance anticancer activity. The IC50 value of compound 34e was 5.853 ± 0.408 µM, the IC50 value of para-OCH3 was half that of the parent compound 33e, and the IC50 value of compound 33e was 10.393 ± 0.949 µM. Generally, there were three crucial elements in the structure-activity relationship of these systems: alkyl chain linker, two-terminal amide bonds, and heteroaromatic substituents. This SAR (structure-activity relationship) study would help to discover more effective compounds in the future. To study the selective antiproliferative activities of bivalent and polyvalent inhibitors in normal cells and cancer cell lines, the cytotoxicity of compounds in FaDu cells and normal mammary epithelial MCF 10A cells was measured. The results listed in Table 1 showed that although the ligustrazine dimers 8d and 8e had high cytotoxicity in MCF 10A, their IC50 values were 2.69 ± 0.46 and 0.047 ± 0.008 µM, while the IC50 values in FaDu cells were 110 ± 30 and 1.36 ± 0.035 nM, respectively, which were far lower than that in MCF 10A. The selective index (SI) between MCF10A and Fadu (IC50 MCF 10A /IC50 FaDu ) was 24.45 and 34.56, respectively. In addition, the heteroaromatic acid dimers 9e-12e, 21e-23e, and 25e had the similar activity, which was stronger than doxorubicin in FaDu cells, and had high selectivity in FaDu cells and MCF 10A cells with SI (IC50 MCF 10A /IC50 FaDu ) values over 28.69. In contrast, the SI (IC50 MCF 10A /IC50 FaDu ) value of doxorubicin was 0.43. These results encouraged us to further investigate the possible cellular mechanisms.    To study the selective antiproliferative activities of bivalent and polyvalent inhibitors in normal cells and cancer cell lines, the cytotoxicity of compounds in FaDu cells and normal mammary epithelial MCF 10A cells was measured. The results listed in Table 1 showed that although the ligustrazine dimers 8d and 8e had high cytotoxicity in MCF 10A, their IC50 values were 2.69 ± 0.46 and 0.047 ± 0.008 µM, while the IC50 values in FaDu cells were 110 ± 30 and 1.36 ± 0.035 nM, respectively, which were far lower than that in MCF 10A. The selective index (SI) between MCF10A and Fadu (IC50 MCF 10A /IC50 FaDu ) was 24.45 and 34.56, respectively. In addition, the heteroaromatic acid dimers 9e-12e, 21e-23e, and 25e had the similar activity, which was stronger than doxorubicin in FaDu cells, and had high selectivity in FaDu cells and MCF 10A cells with SI (IC50 MCF 10A /IC50 FaDu ) values over 28.69. In contrast, the SI (IC50 MCF 10A /IC50 FaDu ) value of doxorubicin was 0.43. These results encouraged us to further investigate the possible cellular mechanisms. To study the selective antiproliferative activities of bivalent and polyvalent inhibitors in normal cells and cancer cell lines, the cytotoxicity of compounds in FaDu cells and normal mammary epithelial MCF 10A cells was measured. The results listed in Table 1 showed that although the ligustrazine dimers 8d and 8e had high cytotoxicity in MCF 10A, their IC50 values were 2.69 ± 0.46 and 0.047 ± 0.008 µM, while the IC50 values in FaDu cells were 110 ± 30 and 1.36 ± 0.035 nM, respectively, which were far lower than that in MCF 10A. The selective index (SI) between MCF10A and Fadu (IC50 MCF 10A /IC50 FaDu ) was 24.45 and 34.56, respectively. In addition, the heteroaromatic acid dimers 9e-12e, 21e-23e, and 25e had the similar activity, which was stronger than doxorubicin in FaDu cells, and had high selectivity in FaDu cells and MCF 10A cells with SI (IC50 MCF 10A /IC50 FaDu ) values over 28.69. In contrast, the SI (IC50 MCF 10A /IC50 FaDu ) value of doxorubicin was 0.43. These results encouraged us to further investigate the possible cellular mechanisms.

Anti-Proliferative Activity In Vitro
For the newly synthesized compounds, their anti-proliferative activity on human cancer cell lines, HeLa (cervical carcinoma), Hep G2 (hepatoma carcinoma), MCF-7 (breast carcinoma), FaDu (head and neck carcinoma), A549 (lung carcinoma), and normal mammary epithelial cell line MCF 10A, was screened by MTT assay [21] with doxorubicin (DOX) as the positive control. The IC 50 values of these compounds are summarized in Table 1. Among the ligustrazine derivatives, 6a-g and 8a-f, the tetramers 6f and 6g had broad-spectrum cytotoxic activities in all cell lines (IC 50  6.57-20.83 µM), and the cytotoxicity activities increased with the increase of carbon chain length. The dimeric ligustrazine 8d, 8e, 8f showed the most promising anticancer activity in most tested tumor cell lines with IC 50 values between 0.00136 and 6.35 µM, a much better result than DOX and TMP. In particular, compound 8e, two ligustrazine rings linked with decane-1,10-diamine, exhibited the best anti-proliferative ability in cells (except Hep G2), and compared with other linkers, its IC 50 values were 1.42 ± 0.71 µM, 0.037 ± 0.001 µM, 0.00136 ± 0.00035 µM, 1.05 ± 0.05 µM, and 0.047 ± 0.008 µM against HeLa, MCF-7, FaDu, A549, and MCF 10A, respectively. In light of these results, we replaced TMP with aromatic heterocycles similar to ligustrazine in electronic space. The novel small aromatic molecule and heterocyclic dimers 9e-35e were synthesized, and their anti-proliferative ability was evaluated. Compounds 9e, 10e, 11e, 17e, 18e, 21e, 22e, 23e, and 25e, aromatic acid dimers (cinnamoyl, nicotinoyl, benzoyl, quinoline-3-carboxyl, quinoline-6-carboxyl, benzofuran-2-carboxyl, benzothiophene-2-carboxyl, indol-3-carboxyl, chlorinated cinnamoyl), were linked by decane-1,10-diamine, which had potent anti-proliferative ability in FaDu or A549 cell lines. Their IC 50 in FaDu cell lines were less than 1 µM, in the range of 20 nM (benzothiophene- 2-carboxyl) and 697 nM (benzofuran- 2-carboxyl). Also, they had nearly no toxicity in normal cells like MCF 10A. In series 9d-n, compounds 9d-f (IC 50 < 1 µM), two cinnamic acids linked by C8, C10, C12 chain with two terminal amide bonds, were more effective than compounds 9g-i (IC 50 > 20 µM), two cinnamic acids linked by the same alkyl chain with two terminal ester bonds instead of amide bonds. This demonstrated the vital role of two-terminal amide bonds in antitumor activity. Compared with compounds 9d-f, the inhibition of compounds 9j-m, two cinnamic acids linked by the similar length ether chain instead of alky chain, was weaker in tumor cells (IC 50 > 20 µM), which confirmed that the alkyl chain played an important role in the structure. Then, we evaluated the position effect of the substitution at the ortho, meta, or para positions of the same aromatic ring. Compounds 25-35e and the unsubstituted compounds 9e and 11e were all derivatives of cinnamic acid or benzoic acid with different substituents and electronic effects. The results of the compounds 26-27e or 28-29e confirmed that the substituents had a general effect on the activity, -OH, -OCH 3 , -NH 2 , and other electron donor groups could significantly reduce the activity, with the influence of -OH greater than that of -OCH 3 . However, the electron attraction group like -Cl (25e) had no obvious effect. The compounds with para-substituted analogs (31e, para position, IC 50 = 5.594 µM) exhibited better cell growth inhibitory activities in FaDu cells than ortho-and meta-substituted analogs (30e, 32e, ortho and meta position, IC 50 > 20 µM). It is worth noting that para -OCH 3 substitution could enhance anticancer activity. The IC 50 value of compound 34e was 5.853 ± 0.408 µM, the IC 50 value of para-OCH 3 was half that of the parent compound 33e, and the IC 50 value of compound 33e was 10.393 ± 0.949 µM. Generally, there were three crucial elements in the structure-activity relationship of these systems: alkyl chain linker, two-terminal amide bonds, and heteroaromatic substituents. This SAR (structure-activity relationship) study would help to discover more effective compounds in the future.
To study the selective antiproliferative activities of bivalent and polyvalent inhibitors in normal cells and cancer cell lines, the cytotoxicity of compounds in FaDu cells and normal mammary epithelial MCF 10A cells was measured. The results listed in Table 1 showed that although the ligustrazine dimers 8d and 8e had high cytotoxicity in MCF 10A, their IC 50 values were 2.69 ± 0.46 and 0.047 ± 0.008 µM, while the IC 50 values in FaDu cells were 110 ± 30 and 1.36 ± 0.035 nM, respectively, which were far lower than that in MCF 10A.

Colony Formation Assay
As ligustrazine dimer 8e had the most potent antiproliferative activity in all synthesized compounds, the colony formation assay in FaDu cells was conducted. As shown in Figure 2, FaDu cells treated with compound 8e at concentrations 0.75 nM, 1.50 nM, 3 nM, 6 nM for 10 days formed smaller and fewer colonies compared to untreated control group, indicating that compound 8e had a dose-dependent inhibition on the proliferation of FaDu cells, and the colony inhibition rate reached around 85.4% at 6 nM.
As ligustrazine dimer 8e had the most potent antiproliferative activity in all synthesized compounds, the colony formation assay in FaDu cells was conducted. As shown in Figure 2, FaDu cells treated with compound 8e at concentrations 0.75 nM, 1.50 nM, 3 nM, 6 nM for 10 days formed smaller and fewer colonies compared to untreated control group, indicating that compound 8e had a dose-dependent inhibition on the proliferation of FaDu cells, and the colony inhibition rate reached around 85.4% at 6 nM.

Live/Dead Staining
Live-dead double staining experiment was conducted to visually evaluate the cell toxicity of compound 8e, the most potent one obtained by MTT assay. Propidium iodide (PI)-stained dead cells

Morphological Observation by Hoechst 33,342 Staining
To investigate whether the anti-proliferation behavior of 8e was related to the apoptosis pathway, the interaction between 8e-and FaDu cells was analyzed through Hoechst 33,342 staining with a laser scanning confocal microscope. As shown in Figure 4, cells without treatment as the

Morphological Observation by Hoechst 33,342 Staining
To investigate whether the anti-proliferation behavior of 8e was related to the apoptosis pathway, the interaction between 8e-and FaDu cells was analyzed through Hoechst 33,342 staining with a laser scanning confocal microscope. As shown in Figure 4, cells without treatment as the control group exhibited almost negligible apoptotic characteristics with homogenous and round blue-stained nuclei. On the contrary, the typical apoptotic nuclei with irregular nuclear morphology and condensation of chromatin (much brighter stained cells) were observed, and most FaDu cells were shrunken and rounded up from the culture dish after exposure to 8e. The result demonstrated that inducing apoptosis in FaDu cells might be one cause of cell growth inhibition by compound 8e.

Apoptosis Analysis by Annexin V-FITC/PI Staining
To substantiate whether the cell death induced by the ligustrazine dimer 8e was associated with apoptosis, the interaction of FaDu cells with 8e was further analyzed by an annexin V-FITC/PI staining, and the apoptosis ratios were quantitated by flow cytometry. The cells were divided into four quadrants: Q1 represented necrotic cells (annexin V − /PI + ), Q2 represented late apoptotic cells (annexin V + /PI + ), Q3 represented early apoptotic cells (annexin V + /PI − ), Q4 represented living cells (annexin V − /PI − ). As shown in Figure 5, after being exposed to different concentrations of 8e (2, 5, 10 nM) for 24 h in FaDu cells, the apoptotic cells (including the early and late apoptosis ratios) increased gradually from 8.23% of the control to 34.2%, 57.7%, 79.5%, respectively. The results revealed that 8e could induce FaDu cells' apoptosis in a dose-dependent manner. This was consistent with other reports showing TMP and its derivatives could induce apoptosis of cancer cells [22][23][24].

Apoptosis Analysis by Annexin V-FITC/PI Staining
To substantiate whether the cell death induced by the ligustrazine dimer 8e was associated with apoptosis, the interaction of FaDu cells with 8e was further analyzed by an annexin V-FITC/PI staining, and the apoptosis ratios were quantitated by flow cytometry. The cells were divided into four quadrants: Q1 represented necrotic cells (annexin V − /PI + ), Q2 represented late apoptotic cells (annexin V + /PI + ), Q3 represented early apoptotic cells (annexin V + /PI − ), Q4 represented living cells (annexin V − /PI − ). As shown in Figure 5, after being exposed to different concentrations of 8e (2, 5, 10 nM) for 24 h in FaDu cells, the apoptotic cells (including the early and late apoptosis ratios) increased gradually from 8.23% of the control to 34.2%, 57.7%, 79.5%, respectively. The results revealed that 8e could induce FaDu cells' apoptosis in a dose-dependent manner. This was consistent with other reports showing TMP and its derivatives could induce apoptosis of cancer cells [22][23][24].
(annexin V + /PI + ), Q3 represented early apoptotic cells (annexin V + /PI − ), Q4 represented living cells (annexin V − /PI − ). As shown in Figure 5, after being exposed to different concentrations of 8e (2, 5, 10 nM) for 24 h in FaDu cells, the apoptotic cells (including the early and late apoptosis ratios) increased gradually from 8.23% of the control to 34.2%, 57.7%, 79.5%, respectively. The results revealed that 8e could induce FaDu cells' apoptosis in a dose-dependent manner. This was consistent with other reports showing TMP and its derivatives could induce apoptosis of cancer cells [22][23][24].

Mitochondrial Membrane Potential (∆Ψm) Analysis
Mitochondria play an essential role in cell apoptotic progression by regulating mitochondrial membrane potential (∆Ψm). Depolarization of ∆Ψm is considered an indicator of cell apoptosis [25]. To better understand the mechanism of 8e-induced FaDu cells apoptosis, the depolarization of ∆Ψm was quantitated by JC-1 (tetraethylbenzimidazolylcarbocyanine iodide) dye. As a mitochondrial-specific dual-fluorescence probe, JC-1 selectively entered the mitochondria. In healthy cells with high ∆Ψm, it accumulated in the mitochondrial matrix as J-aggregates with red fluorescence, while in the reduced ∆Ψm cells, it remained in the cytoplasm as a monomer with green fluorescence. The change of ∆Ψm could be easily detected by the change of fluorescence color. Flow cytometric analysis of FaDu cells treated with different concentrations of 8e (2, 5, 10 nM) for 24 h revealed that the ratio of green to red fluorescence was significantly higher than that of the control group in a concentration-dependent manner, indicating that the number of lower ∆Ψm cells increased ( Figure 6). These results suggested that 8e could induce apoptosis of FaDu cells by decreasing ∆Ψm.
reduced ΔΨm cells, it remained in the cytoplasm as a monomer with green fluorescence. The change of ΔΨm could be easily detected by the change of fluorescence color. Flow cytometric analysis of FaDu cells treated with different concentrations of 8e (2, 5, 10 nM) for 24 h revealed that the ratio of green to red fluorescence was significantly higher than that of the control group in a concentrationdependent manner, indicating that the number of lower ΔΨm cells increased ( Figure 6). These results suggested that 8e could induce apoptosis of FaDu cells by decreasing ΔΨm.

Cell Cycle Analysis
In order to investigate whether 8e induced cell cycle disturbances in FaDu cells, the flow cytometric analysis was performed with PI staining. The fluorescence intensity was proportional to the content of double-stranded DNA. The separation of cells in G0/G1, S, and G2/M was based upon the distribution of DNA content. Cytometric profiles of the PI-stained DNA showed that after being treated with 8e at different concentrations (2, 5, 10 nM) for 24 h, the cell cycle of FaDu cells was arrested. As shown in Figure 7, the percentage of cells in the S phase (33.7%, 44.3%, and 54.7%) was markedly improved compared with the control group (28.1%), and this effect was concentrationdependent in FaDu cells. The population of the G0/G1 phase decreased from 51.9% to 41.8%, 44.2%, and 32.4%, respectively. These data indicated that 8e significantly arrested cell cycle at S phase. This was different from the previously reported cell cycle arrest in the G0/G1 phase induced by TMP [4], which revealed that the dimerization of TMP could change the modulation of the cell cycle of its monomer.

Cell Cycle Analysis
In order to investigate whether 8e induced cell cycle disturbances in FaDu cells, the flow cytometric analysis was performed with PI staining. The fluorescence intensity was proportional to the content of double-stranded DNA. The separation of cells in G0/G1, S, and G2/M was based upon the distribution of DNA content. Cytometric profiles of the PI-stained DNA showed that after being treated with 8e at different concentrations (2, 5, 10 nM) for 24 h, the cell cycle of FaDu cells was arrested. As shown in Figure 7, the percentage of cells in the S phase (33.7%, 44.3%, and 54.7%) was markedly improved compared with the control group (28.1%), and this effect was concentration-dependent in FaDu cells. The population of the G0/G1 phase decreased from 51.9% to 41.8%, 44.2%, and 32.4%, respectively. These data indicated that 8e significantly arrested cell cycle at S phase. This was different from the previously reported cell cycle arrest in the G0/G1 phase induced by TMP [4], which revealed that the dimerization of TMP could change the modulation of the cell cycle of its monomer.

In Silico ADMET Prediction
ADMET prediction is becoming increasingly crucial as most drugs fail in clinical trials owing to poor pharmacokinetic parameters [26,27]. Table 2 shows the ADMET calculated descriptors for our synthesized compounds and two biplots shown in Figure 8. All compounds had 0-4 HBD (hydrogen bond donor) and 4-10 HBA (hydrogen bond acceptor), which were under Lipinski's rule of five (HBD < 5, HBA < 10). PPB (plasma protein binding) was one of the key properties related to drug efficacy. Except for compounds 10-15e, 17-24e, and 9d, most of the compounds had weak (<90%) binding with plasma proteins in the blood, which were predicted to be non-inhibitors of cytochrome P450 2D6, and participated in the good-metabolism of phase I metabolism. T1/2 of the target molecules was in the range of 1. 63-2.29 h, showing the characteristic of rapid metabolism. The HT (hepatotoxicity) values estimated the hepatoxicity of the chemical compounds. HT values for series 6a-i, 13-14e, 20e, 29e, 31-35e, 9f, 9i, 9k, and 9m were zero, and there was no hepatoxicity, while other compounds might be hepatotoxic. Moreover, the LD50 (median lethal dose) predicted that series 8a-f, 9-35e, and 9d-m were over 500 mg/kg, implying they were less toxic. In addition, an ADMET model formed with descriptors 2D polar surface area (PSA) and AlogP98 could predict human intestinal absorption and the blood-brain barrier penetration (BBB) at 95% and 99% confidence level. In the absorption plot, compounds 6a-f, 6h-i, 8a-f, 9-11e, 13-19e, 23e, 30-33e, 9d, 9g, and 9j-m were predicted to be easily absorbed by the intestine at 99% confidence level. While in the BBB plot, the most potent compounds 8e and 8f were fallen out of the 99% ellipse (undefined), suggesting its poor blood-brain barrier penetration ability. Therefore, these molecules might possess low or no side effects of the central nervous system (CNS). In conclusion, ADMET prediction of the compounds might be helpful for further designing new drugs with favorable oral bioavailability.

In Silico ADMET Prediction
ADMET prediction is becoming increasingly crucial as most drugs fail in clinical trials owing to poor pharmacokinetic parameters [26,27]. Table 2 shows the ADMET calculated descriptors for our synthesized compounds and two biplots shown in Figure 8. All compounds had 0-4 HBD (hydrogen bond donor) and 4-10 HBA (hydrogen bond acceptor), which were under Lipinski's rule of five (HBD < 5, HBA < 10). PPB (plasma protein binding) was one of the key properties related to drug efficacy. Except for compounds 10-15e, 17-24e, and 9d, most of the compounds had weak (<90%) binding with plasma proteins in the blood, which were predicted to be non-inhibitors of cytochrome P450 2D6, and participated in the good-metabolism of phase I metabolism. T 1/2 of the target molecules was in the range of 1. 63-2.29 h, showing the characteristic of rapid metabolism. The HT (hepatotoxicity) values estimated the hepatoxicity of the chemical compounds. HT values for series 6a-i, 13- 14e, 20e,  29e, 31-35e, 9f, 9i, 9k, and 9m were zero, and there was no hepatoxicity, while other compounds might be hepatotoxic. Moreover, the LD 50 (median lethal dose) predicted that series 8a-f, 9-35e, and 9d-m were over 500 mg/kg, implying they were less toxic. In addition, an ADMET model formed with descriptors 2D polar surface area (PSA) and AlogP98 could predict human intestinal absorption and the blood-brain barrier penetration (BBB) at 95% and 99% confidence level. In the absorption plot, compounds 6a-f, 6h-i, 8a-f, 9-11e, 13-19e, 23e, 30-33e, 9d, 9g, and 9j-m were predicted to be easily absorbed by the intestine at 99% confidence level. While in the BBB plot, the most potent compounds 8e and 8f were fallen out of the 99% ellipse (undefined), suggesting its poor blood-brain barrier penetration ability. Therefore, these molecules might possess low or no side effects of the central nervous system (CNS). In conclusion, ADMET prediction of the compounds might be helpful for further designing new drugs with favorable oral bioavailability.  showing the 95% and 99% confidence limit ellipses corresponding to the intestinal absorption and blood-brain barrier models.  showing the 95% and 99% confidence limit ellipses corresponding to the intestinal absorption and blood-brain barrier models.

Conclusions
Using ligustrazine, aromatic, or heteroaromatic acids as the main raw materials and linked by alkyl or ether chain, bivalent antitumor agents were synthesized and screened from five different human cancer cell lines and one normal mammary epithelial cell line. SAR revealed that the two terminal amide bonds and the alky chain were the dominant factors with 10 C atom chain showing the most potent antitumor effect. Among different aromatic moieties, the ligustrazine dimer 8e was the best candidate; its IC 50 was 1.36 nM in FaDu cell, it showed a broad-spectrum antitumor activity with IC 50 in the range of 0.00136-1.42 µM, and the selective ratio MCF 10A/FaDu was 34.56. The subsequent fluorescence staining and flow cytometry analysis indicated that 8e could induce apoptosis through depolarization of the mitochondrial membrane potential in FaDu cells. Further mechanism investigation showed that 8e could arrest cell cycle at the S phase in FaDu cells. Furthermore, ADMET predicted that most of the potent compounds followed Lipinski's law and became 'drug-like' molecules, further confirming that the dimerization of ligustrazine could improve the antitumor activity of its monomer. The highly selective inhibition of these dimers on the proliferation of FaDu cells might be related to the effect on a specific target in FaDu cells. Based on the characteristics of FaDu cells, further targeted screening would give insight into the selectivity of such potent compounds.

Chemistry
1 H-NMR and 13 C-NMR spectra were recorded on a Mercury Vx-300 (300 MHz) or AVANCE III (400 MHz). Chemical shifts were reported in ppm (δ) using the residue solvent line as the internal standard (for TMS: 0 ppm, 1 H and 13 C; for CHCl 3 -d: 7.26 ppm, 1 H and 77.16 ppm, 13 C). Coupling constants were given in Hertz (J). High-resolution MS (HRMS) were recorded on an Agilent 6520 Q-TOF LC/MS. Melting points were uncorrected and were determined on a digital melting point apparatus (Shenguang WRS-1B, shanghai, China). Flash column chromatography was carried out by using silica gel (200-300 mesh). Reactions were monitored by thin-layer chromatography (TLC) on silica gel GF254 plates (Qingdao Haiyang Chemical Plant, Qingdao, China). Reagents and solvents were used as purchased from commercial sources without further purification. The purity of the synthesized compounds was over 95% analyzed by HPLC (Waters 2695 Alliance system, Waters Corp., Milford, MA, USA), with the Kromasil C18 column eluted by methanol/water (80/20) containing 0.1% trifluoroacetic acid at a flow rate of 1 mL/min. Compounds 4 and 5 were prepared, as reported previously [19]. 1 H-NMR, 13 C-NMR and HR-MS spectra of compounds 6a-9m can be seen in supplementary materials.

General Synthetic Procedure for 6a-g
A mixture of alkane-diamine (10 mmol), TMP-Cl (50 mmol), KOH (60 mmol), and dimethylformamide (35 mL) was heated to reflux. The reaction progress was monitored via TLC. After 2 h, the reaction mixture was cooled to room temperature. Subsequently, the solid was filtered off, and the filtrate was added with H 2 O (100 mL) and extracted with ethyl acetic acid three times. The organic layers were combined and washed with H 2 O and brine, and then separated and dried over anhydrous Na 2 SO 4 for 8 h. The solvent was removed under reduced pressure; the resulting residue was separated on flash column chromatography eluted by a mixture of petroleum ether/acetone in a volume ratio of 7:1-3:1, and further recrystallized from acetone to give pale white solid.  [C 44 H 69 N 10 ] + 737.5707.

1,4-Bis((3,5,6-Trimethylpyrazin-2-yl)Methyl)Piperazine (6h)
A mixture of piperazine (0.800 g, 9.29 mmol), TMP-Cl (3.80 g, 22.29 mmol), KOH (1.56 g, 27.66 mmol), and dimethylformamide (20 mL) was heated under reflux for 2 h. The reaction was monitored by TLC. The reaction was quenched, filtered, and the filtrate was added with 100 mL H 2 O and extracted with ethyl acetate three times. The organic layers were combined, washed with H 2 O and brine, and then dried over anhydrous Na 2 SO 4 for 8 h. The solvent was evaporated in vacuo. The resulting residue was separated on flash column chromatography with a mixture of petroleum ether/acetone 7:1-3:1(volume ratio) as eluent, and then further recrystallized from acetone to afford compound 6h.

General Synthetic Procedure for 8a-f
Compound 7 was gained according to the method described by Wu [20]. A mixture of tetramethylpyrazine acid (4.40 mmol), EDCI (4.72 mmol), triethylamine (12 mmol), DMAP (2.40 mmol), and anhydrous dichloromethane (10 mL) was stirred to dissolve, then diamino-alkane (2 mmol) was added and stirred at room temperature until the reaction was finished (monitored by TLC). The reaction mixture was washed with brine and water, extracted with dichloromethane (20 mL), dried over anhydrous sodium sulfate for 8 h, filtered, and the solvent was evaporated. The resulting residue was purified by flash column chromatography with an eluent of petroleum ether/acetone = 15:1-3:1(volume ratio) to give pale white solid.  13 C-