From recent research on the human genome, many diseases, including cancers and malignant lymphomas, are better understood at a genomic level. Thus, many areas of cancer science, such as diagnosis, treatment and prevention, are changing dramatically.(1) For instance, new anticancer agents that target mutated gene products are showing great promise, such as Glivec, which targets the Abelson leukemia viral oncogene kinase in patients with chronic myelogenous leukemia.(2) DNA-targeting agents, such as cisplatin, bleomycin and mitomycin C, are used routinely in cancer treatment. However, these drugs are extremely toxic, attacking non-cancerous cells and inducing severe side effects. One important question to consider is whether the introduction of sequence selectivity to DNA-targeting agents can improve their efficacy as anticancer agents. To address this question, we have designed and synthesized a series of sequence-specific alkylating agents. N-methyl Py–Im hairpin polyamides bind in the minor groove of DNA, with antiparallel paired Im/Py uniquely recognizing G-C base pairs and Py/Py pairs recognizing either A-T or T-A base pairs.(3,4) Duocarmycin A, a minor groove-binding antitumor antibiotic produced by Streptomyces spp., alkylates adenine N3 at the 3′ end of sequences of three or more consecutive A-T base pairs in DNA.(5–7) CBI is a synthetic analog of the alkylating moiety of duocarmycin A that has increased stability in aqueous solution.(8) Seco-CBI is a precursor of CBI, and in aqueous solution, seco-CBI derivatives change spontaneously to CBI derivatives. The mechanism of the formation of CBI from seco-CBI and the subsequent alkylation of DNA by CBI derivatives is shown in Fig. 1. We have demonstrated that hybrids between DNA-alkylating agents and Py–Im hairpin polyamides selectively alkylate at the matched sequences according to the recognition rules of the Py–Im polyamides.(9–11) In the present study, the growth inhibition of the alkylating moiety (1) itself with three-A-T base pair recognition, and conjugated compounds with various hairpin Py–Im polyamides (2–6) with six-base pair recognition were compared using 10 human and mouse cell lines.
DNA-targeting agents, including cisplatin, bleomycin and mitomycin C, are used routinely in cancer treatments. However, these drugs are extremely toxic, attacking normal cells and causing severe side effects. One important question to consider in designing anticancer agents is whether the introduction of sequence selectivity to DNA-targeting agents can improve their efficacy as anticancer agents. In the present study, the growth inhibition activities of an indole-seco 1,2,9,9a-tetrahydrocyclopropa[1,2-c]benz[1,2-e]indol-4-one (CBI) (1) and five conjugates with hairpin pyrrole-imidazole polyamides (2–6), which have different sequence specificities for DNA alkylation, were compared using 10 different cell lines. The average values of – log GI50 (50% growth inhibition concentration) for compounds 1–6 against the 10 cell lines were 8.33, 8.56, 8.29, 8.04, 8.23 and 8.83, showing that all of these compounds strongly inhibit cell growth. Interestingly, each alkylating agent caused significantly different growth inhibition patterns with each cell line. In particular, the correlation coefficients between the – log GI50 of compound 1 and its conjugates 2–6 showed extremely low values (R < 0). These results suggest that differences in the sequence specificity of DNA alkylation lead to marked differences in biological activity. Comparison of the correlation coefficients between compounds 6 and 7, with the same sequence specificity as 6, and MS-247, with sequence specificity different from 6, when used against a panel of 37 human cancer cell lines further confirmed the above hypothesis. (Cancer Sci 2006; 97: 219–225)
green fluorescent protein
50% growth inhibition concentration
polymerase chain reaction
root mean square.
Materials and Methods
HLC2, HeLa, HEK293, WI38, NIH3T3 and M5S cell lines, which are human lung carcinoma, human cervical epithelial carcinoma, human kidney epithelial, human embryonic lung fibroblast, mouse fibroblast and mouse near-diploid fibroblast cell lines, respectively, were cultured in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, 100 IU/mL penicillin and 100 µg/mL streptomycin at 37°C in a humidified atmosphere of 5% CO2 in 95% air. HL60, Jurkat, Raji and HCT116 cell lines, which are human acute myeloid leukemia, human T-cell leukemia, human Burkitt lymphoma and human colon carcinoma cell lines, respectively, were cultured in RPMI-1640 medium supplemented with 10% heat-inactivated fetal bovine serum, 100 IU/mL penicillin and 100 µg/mL streptomycin. Analysis of growth inhibition of 37 human cancer cell lines was carried out at the Cancer Chemotherapy Center, Japanese Foundation for Cancer Research (Japanese Foundation for Cancer Research, Tokyo, Japan). The 37 human cancer cells were plated in RPMI-1640 with 5% fetal bovine serum at appropriate densities in 96-well plates and allowed to attach overnight. The cells were exposed to compound 6 for 48 h.
Evaluation of growth inhibition and analysis
The compounds indole-secoCBI (1), PyPyPyPy-γ-ImPy-indoleCBI (2), PyImPyPy-γ-ImPy-indoleCBI (3), ImPyPyPy-γ-ImPy-indoleCBI (4), ImImPyPy-γ-PyPy-indoleCBI (5) and ImImPyPy-γ-ImPy-indoleCBI (6) were synthesized by procedures already reported.(11) Detailed synthetic procedures will be published elsewhere. Colorimetric assays using WST-8 (Dojindo, Kumamoto, Japan) were carried out in 96-well plates. The cells were plated in each well at approximately 15% confluence in 50 µL of culture medium. One day later, when cells were in the logarithmic growth phase, the medium was changed to 100 µL of fresh medium containing various concentrations of the compounds and 0.1% DMF. After treatment with the compounds for 48 h, 10 µL of WST-8 reagent was added into each well and incubated for 2 h at 37°C. Absorbance was then measured at 450 and 600 nm using an MPR-A4I microplate reader (Tosoh, Tokyo, Japan). The absorbance of the control well (C), the treated wells (T) and the treated wells at time 0 (T0) were measured. The GI50 was calculated as 100 × [(T − T0)/(C − T0)] = 50.
All primers for DNA amplification were purchased from Proligo (Boulder, CO, USA). To obtain the DNA fragment encoding the GFP gene, PCR amplification was carried out in a final volume of 50 µL containing 1 ng of pAce-Green N1 vector (Novagen, Darmstadt, Germany), 200 nM of the primer set GFP forward (5′-AAGCTTCCACCATGAGCAAGG-3′) and GFP reverse (3′-AAGCTTTCAGCTCATCTTGTA-5′), 200 µM deoxynucleotide triphosphates (Sigma Aldrich, St Louis, MO, USA), 2 IU of Taq DNA polymerase and 1 × ThermoPol Reaction Buffer (New England BioLabs, Beverly, MA, USA). Amplification cycles were carried out using an iCycler (Bio-Rad, Hercules, CA, USA). The reaction mix was incubated at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 60°C for 30 s and 72°C for 60 s, with a final extension step of 72°C for 7 min. Products were identified following separation in 1% Tris-borate-ethylenediaminetetracetic acid agarose gels with 0.5 µg/mL ethidium bromide using a 100-bp ladder marker (New England BioLabs) and visualization under UV illumination.
Cloning of amplified DNA fragments
The PCR products were ligated into the pGEM-T Easy vectors (Promega, Madison, WI, USA). Escherichia coli DH5α competent cells (Toyobo, Osaka, Japan) were then transformed and cultured on LB plates with 100 µg/mL ampicillin and 32 µg X-gal/400 µg isopropyl-d-thiogalactoside overnight at 37°C. White colonies were identified by colony-direct PCR using the primer set of GFP reverse and a T7 promoter primer (5′-TAATACGACTCACTATAGGG-3′) in the reaction mixture described above. The reaction mixtures were incubated at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 60 s, with a final extension step of 72°C for 7 min. Appropriate colonies were selected for transfer to 5 mL of LB medium with 100 µg/mL ampicillin and cultured overnight at 37°C. The plasmids with inserts were extracted using the GenEluteTM Plasmid Miniprep Kit (Sigma Aldrich) and identified by PCR (program and reaction mixtures as above).
Preparation of 5′-Texas Red-modified DNA fragments and high-resolution gel electrophoresis
5′-Texas Red-modified DNA fragments of GFP were prepared by PCR using the primer set of GFP reverse and a 5′-Texas Red labeled T7 promoter primer, 1 ng of pGEM-T Easy vector with inserted GFP, and the program and reagents described above. The fragments were purified with a GenEluteTM PCR Clean-up Kit (Sigma Aldrich) and their concentration determined by UV absorption. The 5′-Texas Red-labeled DNA fragments (10 nM) were alkylated by various concentrations of compound 1 in 10 µL of 5 mM sodium phosphate buffer (pH 7.0) containing 10% DMF at room temperature for 10 h. The reaction was quenched by the addition of 1 µg calf thymus DNA and heating for 5 min at 90°C. The DNA was recovered by vacuum centrifugation. The pellet was dissolved in 3.5 µL of loading dye (formamide with fuchsin red), heated at 95°C for 20 min and then placed immediately on ice. A 2-µL aliquot was subjected to electrophoresis on a 6% denaturing polyacrylamide gel using a Hitachi SQ5500-E DNA Sequencer (Hitachi, Tokyo, Japan).
Molecular modeling studies
Minimizations were carried out with the Discover program (MSI, San Diego, CA, USA) using consistent force field force-field parameters. The starting structure was built based on the nuclear magnetic resonance structure of the ImPyPy-γ-PyPyPy-d(GCTGTCCAGC)/d(GCTGGACAGC) complex and the duocarmycin A–distamycin A–d(CAGGTGGT)/d(ACCACCTG) complex.(12) The connecting parts between them were built using standard bond lengths and angles. The CBI unit of the assembled initial structure was energy minimized using a distance-dependent dielectric constant of ɛ = 4r (r stands for the distance between atoms i and j) and with convergence criteria having a RMS gradient of less than 0.001 kcal/mol Å. Eighteen Na+ cations were placed at the bifurcating position of the O–P–O angle at a distance of 2.51 Å from the phosphorus atom. The resulting complex was soaked in a 10-Å layer of water. The whole system was minimized with no constraints, to the stage where the RMS was less than 0.001 kcal/mol Å.
Indole-CBI and its hairpin Py–Im polyamide conjugates
With the goal of developing sequence-specific antitumor agents, we recently developed new types of Py–Im CBI conjugates (2–6) with indole linkers (Fig. 2a). Because indole-CBI (1) has better chemical stability under basic and acidic conditions, the conjugates were readily synthesized by coupling CBI-indoles and Py–Im polyamides prepared by automated solid-phase synthesis.
We first examined the sequence specificity of compound 1 using denaturing polyacrylamide gel electrophoresis. Sequences of the alkylated regions were determined by thermal cleavage of the DNA strand at the alkylated sites. All N3-alkyl adducts are cleaved quantitatively to produce cleavage bands under the heating conditions used. The sequences of the DNA fragments alkylated by 1 are shown in Fig. 2b. Alkylation by compound 1 occurred specifically at the 3′ end of the A in A-T-rich sequences, which is the same as the known specificity of duocarmycin A and DU-86.(13) It was demonstrated that alkylation by conjugates 2–6 occurred specifically at the A in the matched Py–Im polyamide sequences according to the recognition rule summarized in Fig. 3a.(12) DNA alkylation activities of compounds 1–6 clearly indicate that they bind to the minor groove of DNA. Energy-minimized binding models of the compound 1–d(GCGTATACGC)/d(GCGTATACGC) and 6–d(GCTGTCCAGC)/d(GCTGGACAGC) complexes based on the alkylation experiments are shown in Fig. 4.
Effects of compounds 1–6 on the growth of 10 mammalian cell lines
To evaluate the cytotoxic potency of indole-CBI (1) and the five alkylating Py–Im polyamides (2–6), which recognize specific 6-base pair DNA sequences, we examined the growth inhibition effects of compounds 1–6 on 10 human and mouse cell lines. The cells were treated with 10−10 to 10−6 M of the agent for 48 h, and then living cells were counted. Graphs of the mean GI50 values are shown in Fig. 3b. The average of –log GI50 of compound 1 was estimated as 8.33. We found that HLC2 was the most sensitive cell line (–log GI50= 8.72) and M5 (7.59) the most resistant to compound 1. The average –log GI50 values for treatment with compounds 2, 3, 4, 5 and 6 were determined to be 8.56, 8.29, 8.04, 8.23 and 8.83, respectively. It was found that Jurkat (9.46), HL60 (9.28), HL60 (9.30), HL60 (8.76) and Jurkat (9.47) were the cell lines most sensitive to compounds 2, 3, 4, 5 and 6, respectively. In contrast, M5S (7.59), HEK293 (7.66), HeLa (7.28), NIH3T3 (7.25), HeLa (7.39) and WI38 (8.05) cells were most resistant to compounds 2, 3, 4, 5 and 6, respectively. The alkylating Py–Im polyamides appeared to be more effective against HL60 and Jurkat cells, although these cells were slightly resistant to compound 1. The Raji and HLC2 cells showed a tendency for resistance against compounds 6 and 3, respectively. HCT116 cells displayed resistance against compounds 4 and 6. HeLa cells showed resistance against compounds 2, 3, 4 and 5, whereas this cell line was sensitive to compound 1. HEK293 cells were sensitive to compound 6, whereas this cell line showed resistance to the other alkylating polyamides (compounds 2–5). WI38 cells displayed resistance against all of the alkylating Py–Im polyamides, while the –log GI50 value of compound 1 was close to the average. NIH3T3 cells were slightly sensitive to compound 5 and resistant to compound 4. M5S cells were resistant to compounds 1 and 2. Thus, we observed that all of these compounds showed differential patterns of growth inhibition in the 10 cell lines tested. The correlation coefficients between compounds 1 and 6 against 10 mammalian cell lines are shown in Fig. 3c.
Growth inhibitory effects of compound 6 in 37 human cancer cell lines
Compound 6 was the most effective growth inhibitor among the compounds 1–6, as shown in Fig. 3. Therefore, we investigated the GI50 profiles of compound 6 against 37 human cancer cell lines.(14,15) The growth inhibition pattern of compound 6 is shown in Fig. 5, and is compared with the reported growth inhibition pattern of compound 7 (ImImPyPy-γ-ImPyPy-vinylCBI)(11) and MS-247.(14) It was found that the growth inhibition patterns of compounds 6 and 7, which have exactly the same sequence specificity, were consistent with a high correlation coefficient (R) of 0.85. However, the growth inhibition of MS-247, which has A-T specificity, showed a different pattern compared with those of compounds 6 and 7, with a correlation coefficient of 0.61.
The results in the present study revealed different growth inhibition patterns induced by the simple DNA-alkylating agent indole-CBI (1) and its Py–Im polyamides conjugates (2–6) against 10 mammalian cell lines. Although compounds 1–6 showed strong growth inhibitory effects (average GI50 <1.0 × 10−8 M), the results suggest that conjugation of Py–Im polyamides to compound 1 does not impart outstanding growth inhibition potency. However, analysis of the graphs in Fig. 3c revealed that compound 1, with a three-A-T base pair recognition site, did not correlate with the conjugates with six-base pair recognition sites (R < 0). It is generally accepted that higher correlation coefficients (R > 0.75) are observed for anticancer agents possessing the same reaction mechanism.(15) For example, the correlation coefficients between triethylenethiophosphoramide and many DNA alkylators, such as triethylenemelamine, uracil nitrogen mustard, yoshi 864, piperazine alkylator, piperazinedione, pipobroman and hepsulfam, are well matched (R > 0.90), although the chemical structures of these compounds are quite different.(16) We cannot compare the results directly due to the small number of cell lines examined; however, the negative correlation coefficients observed between compound 1 and compounds 2–6 suggest that growth inhibition results from different mechanisms. One attractive possibility is that the difference is due to different sequence specificities: compound 1 alkylates A-T-rich sequences, whereas compounds 2–6 have six-base pair recognition sites containing G-C sequences. The lower correlation coefficients between compounds 2–6 also support this argument. In addition, we examined the cytotoxicity pattern of compound 6 against a panel of 37 cancer cell lines. Comparison of the growth inhibition pattern of compound 6 with that of compound 7, which has exactly the same sequence specificity as compound 6, showed a high correlation coefficient (R = 0.85). In sharp contrast, the R-value for comparison between compound 6 and MS-247,(14,15) which has A-T specificity, is relatively low (R = 0.61). These results further support the premise that the mechanism of growth inhibition due to DNA alkylation may depend on sequence specificity.
We demonstrated recently that alkylation of the template strand in a coding region results in the production of truncated mRNA, effectively inhibiting transcription in vitro.(17) We also demonstrated that the selective inhibition of renilla and firefly luciferase expression in living mammalian cells is induced by sequence-specific alkylation of the template strand in the coding region.(18) These results suggest that the alkylation site of each compound may cause different silencing of genes in cells, resulting in differences in growth inhibition. Although a direct relationship between alkylation sites and growth inhibition effects is not known, the cells tested may have different crucial sequences for alkylation.
In the present study, we compared the cytotoxicities of DNA alkylating agents that recognize different sequences. The results clearly demonstrate that the growth inhibitory effects of indole-CBI (1) are altered dramatically by conjugation with Py–Im polyamides, resulting in sequence-specific alkylation. Recently, we found that alkylating Py–Im polyamides, which differ only in that the C–H bond is substituted by an N atom in the second ring, showed significantly different cytotoxicities in human cancer cell lines.(19) These results suggest that DNA sequence specificity may contribute to the cytotoxic potency of alkylating agents. The results also suggest the intriguing possibility that DNA-alkylating agents recognizing longer base pair sequences may provide a promising approach for developing new types of tailor-made antitumor drugs. We are currently in the process of elongating the sequence specificity of alkylating polyamides to more than 10 base pairs.
We thank the Screening Committee of New Anticancer Agents supported by a Grant-in-Aid for Scientific Research on Priority Area ‘Cancer’ from The Ministry of Education, Culture, Sports, Science and Technology, Japan.