Synergistic binding of actinomycin D and echinomycin to DNA mismatch sites and their combined anti-tumour effects

Abstract

Combination cancer chemotherapy is one of the most useful treatment methods to achieve a synergistic effect and reduce the toxicity of dosing with a single drug. Here, we use a combination of two well-established anticancer DNA intercalators, actinomycin D (ActD) and echinomycin (Echi), to screen their binding capabilities with DNA duplexes containing different mismatches embedded within Watson-Crick base-pairs. We have found that combining ActD and Echi preferentially stabilised thymine-related T:T mismatches. The enhanced stability of the DNA duplex–drug complexes is mainly due to the cooperative binding of the two drugs to the mismatch duplex, with many stacking interactions between the two different drug molecules. Since the repair of thymine-related mismatches is less efficient in mismatch repair (MMR)-deficient cancer cells, we have also demonstrated that the combination of ActD and Echi exhibits enhanced synergistic effects against MMR-deficient HCT116 cells and synergy is maintained in a MMR-related MLH1 gene knockdown in SW620 cells. We further accessed the clinical potential of the two-drug combination approach with a xenograft mouse model of a colorectal MMR-deficient cancer, which has resulted in a significant synergistic anti-tumour effect. The current study provides a novel approach for the development of combination chemotherapy for the treatment of cancers related to DNA-mismatches.

INTRODUCTION

The mismatch repair (MMR) mechanism can efficiently identify and correct DNA mismatches in order to maintain the fidelity of genetic processes (1,2). However, many cancer cells have increased mutation rates due to MMR defects, leading to an increase in the number of base pair mismatches in their genomes (3,4). The mismatched sites are thermodynamically and structurally distinct from canonical Watson–Crick base pairs, which makes them attractive targets for anti-cancer small-molecule ligands (5–7). Several families of small-molecule compounds have been used for targeted therapeutic applications or for detection of DNA damage through targeting mismatches. For example, bis-naphthyridine based analogues have been shown to preferentially bind mismatches in DNA associated with neurological diseases (8). Metal complexes with DNA mismatches selectively show enhanced toxicity in MMR-deficient cancer cells (9). We have recently shown the ability of a quinoxaline antibiotic drug, echinomycin (Echi) to recognise homo-thymine mismatches for treating defective mismatch-repair system related cancers (10). However, due to the complexity of human cancers, the limited DNA targeting ability of a single drug, and the development of drug resistance, it is essential to improve the therapeutic efficacy for existing DNA-binding drugs. Combinations of small molecule drugs can be very useful for enhancing specificity and therapeutic ability against several cancers (11–13). The combined use of two or more DNA-targeting drugs has several advantages: it can exert a synergistic effect, thereby improving the efficacy, as well as reducing the side effects and toxicity of a single drug since individual drug concentrations can be reduced. For example, it was found that the anticancer drug actinomycin D (ActD), when used in combination with other DNA binding agents, becomes more effective in the treatment of non-metastatic Ewing's sarcoma (14). Although many studies have been published on combined anti-cancer drug therapy, the molecular mechanisms underlying these effects often remain unclear.

ActD and Echi are natural antibiotics produced by Streptomyces species that show anticancer effects due ultimately to their DNA-binding abilities (Figure 1). The phenoxazone ring of ActD can be intercalated into DNA at 5'-GpC sites, while its cyclic peptide rings are bound in the minor groove (15,16). The strong preference for the GpC sequence is due to hydrogen bonds between the N2/N3 atoms of the guanine base and the C=O/NH atoms in the threonine residue of ActD. In contrast to the mono-intercalation shown by ActD, Echi is a bis-intercalator with two quinoxaline rings that intercalate into 5'-CpG site in DNA, while its two depsipeptide rings also bind in the DNA minor groove (17,18). The CpG preference of Echi is due to the hydrogen bonds between the N2/N3 atoms of the guanine base and the alanine amide and carbonyl groups of Echi. The biological activities of these antibiotics are mainly due to their ability to bind DNA and thereby they disrupt cellular processes such as replication and transcription (19,20). Echi has also been shown to inhibit the DNA-binding activity of hypoxia-inducible factor-1α (HIF-1α) in vitro, thus acting as a potential HIF-1α inhibitor (21). The binding affinities of ActD and Echi at GpC and CpG sites are influenced by their flanking sequences. Previous studies have shown that ActD and Echi bind preferentially to the GpC and CpG sequences, respectively, that flank a single T:T mismatch, compared to other canonical base pairs at the same position (10,22,23). Thus, both ActD and Echi have the potential to serve as mismatch-binding agents for the treatment of diseases related to DNA mismatches, including cancers.

Figure 1.

Chemical structures of (A) actinomycin D (ActD) and (B) echinomycin (Echi). The abbreviations shown in the figure represent phenoxazone (PXZ), l-threonine (THR), d-valine (DVA), l-proline (PRO), sarcosine (SAR), N-methyl-l-valine (MVA), quinoxaline (QUI), d-serine (DSN), l-alanine (ALA), N-dimethyl-l-cysteine (N2C) and N-methyl-l-cysteine (NCY). The numbers indicate the order of the cyclic peptides in ActD and Echi structures.

We have examined in this study a combination of ActD and Echi, to screen DNA sequences with different mismatches within canonical Watson–Crick base pairs. Our initial melting temperature results show that these two drugs can simultaneously recognise a thymine-related T:T mismatch flanked by GpC and CpG steps, with greater selectivity than other mismatches or canonical base pair DNA sequences. We have solved the crystal structures of the DNA sequences d(AGCTCGT)/d(ACGTGCT) and d(AGCTCGT)/d(ACGCGCT) containing a central T:T and T:C mismatch complexed with ActD and Echi. Detailed structural analyses have explained how two distinct intercalators can cooperatively recognise mismatched DNA by inducing specific conformational changes and forming an extensive interaction network. Cell-based assays using MMR-deficient and proficient cells have shown that the two-drug combination can result in synergistic effects against MMR-deficient cancer cells. In vitro knockdown of the MMR-related gene MLH1 in SW620 MMR-proficient cells using short hairpin RNAs further confirms that the synergistic effects of two drug combination is maintained in knockdown cells. Finally, we evaluated the in vivo anti-cancer activity of the two-drug treatment approach in a HCT116 xenograft mice model, which is consistent with a relationship between the mismatched DNA duplex–drug interactions and the possible biological consequences. Our study thus aims to provide a molecular mechanism for a new combination chemotherapy treatment strategy for mismatched DNA-related cancers.

MATERIALS AND METHODS

Chemicals and oligonucleotides

DNA oligonucleotides were synthesized by MD Bio (Taipei, Taiwan). All oligonucleotides were purified by polyacrylamide gel electrophoresis. DNA oligonucleotide concentrations were determined according to Beer's law: A = ϵbc (where A is the absorbance at 260 nm, ϵ the molar extinction coefficient, b the cell path length and c the molar concentration), using a JASCO-v630 UV–visible spectrophotometer (JASCO International, Tokyo, Japan). The DNA Calculator software of Vladimír Čermák from https://molbiotools.com/dnacalculator.html was used to obtain an approximate value of ϵ_260nm for each oligonucleotide. ActD and Echi were purchased from Sigma-Aldrich (St. Louis, Missouri, USA) and stocks were prepared in dimethyl sulfoxide. The optical density at 440 nm (ϵ_{440 nm} = 24500 M⁻¹ cm⁻¹) and 325 nm (ϵ_{325 nm} = 11500 M⁻¹ cm⁻¹) corresponding to the maximum absorbance of ActD and Echi, respectively was used to determine the drug concentrations (24).

Melting temperature measurements

As described in previous studies, the melting temperature (T_m) was determined by measuring the absorbance of different DNA sequences at 260 nm in the presence or absence of drugs (25). DNA oligonucleotides (3 μM) were prepared in a buffer containing 100 mM sodium cacodylate (pH 7.2) and 100 mM NaCl. The oligonucleotides were denatured at 95°C for 5 min and annealed on ice for 30 min. Then, the drugs were added to the oligonucleotide solution at various DNA:drug molar ratios. The samples were incubated at 4°C for 24 h to allow for complex formation. Afterwards, 650 μl of the sample was added to the quartz cuvette and covered with a layer of silicone oil. The sample was then equilibrated at 4°C for 10 min using a spectrophotometer. UV absorbance versus temperature profiles were recorded by changing the temperature from 4°C to 95°C at a rate of 1°C min⁻¹ and the optical density at 260 nm was recorded every 0.2 min for three experiments. The T_m values were determined by polynomial fitting to the observed curves for DNA oligonucleotides using Varian\Cary WinUV Thermal v3.00 (Agilent, Santa Clara, CA, USA).

CD spectroscopy

The T:T mismatch-containing DNA duplex oligonucleotides (10 μM) were prepared in a buffer containing 100 mM sodium cacodylate (pH 7.2) and 100 mM NaCl with different ratios of ActD, Echi and a combination containing equal concentrations of ActD and Echi. The sample preparation protocol was similar to that described for the melting temperature experiments. CD spectra were collected between 340 and 220 nm at 1 nm intervals using a Chirascan™ circular dichroism spectrometer with the Pro-Data software suite (version 4.8.3.0). All the spectra were recorded as the average of three runs at room temperature. The method used for CD spectra analyses have been described previously (26,27).

Crystallization of T:T and T:C mismatch complexes with ActD and echi

To grow T:T mismatched DNA and drug complex crystals, d(ACGTGCT/AGCTCGT) oligomers were co-crystallized with ActD and Echi in a molar ratio of 1:1:2 using the vapor diffusion sitting-drop method. First, 0.125 mM single-strand oligonucleotides were heated to 95°C for 5 min, then annealed on ice for 30 min to enable duplex formation to take place, and incubated with 0.250 mM Echi at 4°C for 48 h. Then, 0.125 mM ActD was added and the sample was incubated for another 24 h. Within a week, small needle-shaped, yellow-colored crystals were obtained in a 5 μl drop containing 2.5 mM sodium cacodylate (pH 7.0), 1 mM MgCl₂, 1 mM spermine tetrahydrochloride, 1.5 mM ZnCl₂, and 1% PEG200 equilibrated with 500 μl 30% polyethylene glycol (PEG)-200 at 20°C. To obtain T:C mismatched DNA and drug complex structures, the d(ACGCGCT/AGCTCGT) DNA oligomer was co-crystallized with a DNA:ActD:Echi ratio of 1:1:2 using a method similar to that described above. Long, yellow-colored crystals grew after 1–2 weeks in a 5 μl drop containing 2.5 mM sodium cacodylate (pH 7.0), 1 mM MgCl₂, 2 mM spermine tetrahydrochloride, 2 mM ZnCl₂, and 1% PEG200 equilibrated against 500 μl 30% PEG200 at 20°C.

X-ray data collection, phasing and structure refinement

The X-ray diffraction data from a single crystal were collected in a synchrotron radiation facility at the National Synchrotron Radiation Research Center (Taiwan) using the NSRRC BL15A1 and NSRRC TPS 05A beamlines. The HKL-2000 program package was used to integrate and reduce diffraction data (28). The phase of the T:T mismatched ActD-Echi complex was determined by molecular replacement (MR) of phaser-MR in a Python-based hierarchical environment for integrated xtallography (PHENIX, v1.10.1) using the partial structures of Echi-d(ACGTCGT)₂ (PDB ID: 5YTZ) and ActD-d(ATGCTGCAT)₂ (PDB ID: 1MNV) complexes as templates (29). The initial model building was performed using the crystallographic object-oriented toolkit (COOT) v0.8.9.2 based on the electron density map (Supplementary Figure S1) (30). Structural refinements were performed using phenix.refine in PHENIX. The final refined T:T mismatched complex was used to determine the phases of the T:C mismatched ActD-Echi complex. The final crystallographic and refinement statistics of the complexes are listed in Supplementary Table S1. The 2F_o – F_c electron density maps were generated using the fast Fourier transform in CCP4i, and PyMOL v2.2.3 was used to draw graphical representations of the refined structures. LigPlot⁺ was used to analyze the van der Waals interactions between DNA and drug molecules (31). DNA base pair and base pair step parameters were analyzed using the online servers Web 3DNA and CURVES+ (32,33). The values of DNA torsion angles, local base pair and base pair step parameters are given in Supplementary Tables S2–S4.

Fluorescence polarisation (FP) assay

FAM-labeled DNA oligonucleotides (100 nM) were heated at 95°C for 5 min and allowed to cool on ice for 30 min. The samples were mixed with a different concentration of one drug, while the concentration of the other drug was kept constant in a buffer containing 100 mM sodium cacodylate and 100 mM NaCl at pH 7.2, and incubated for 24 h at 4°C. The fluorescence polarisation signal was measured using a SpectraMax Paradigm Multi-Mode Microplate Reader (Molecular Devices, San Jose, CA, USA), with excitation and emission wavelengths set at 485 and 535 nm, respectively. The FP is defined as:

where is the intensity of the emitted light parallel to the plane of excitation light, and is the intensity of the emitted light perpendicular to the plane of excitation (34). The data were fitted to calculate the binding affinity in terms of the dissociation constant (K_d) using a one-site total binding equation in GraphPad Prism v5.01 (San Diego, CA, USA).

Cell lines, culture conditions, and cell viability

HCT116 and HCT116 + ch3 cells were cultured in Dulbecco's modified Eagle's medium F-12 (DMEM/F-12) containing 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic solution. SW620 cells were maintained in DMEM with 10% FBS and 1% antibiotic-antimycotic. The cell lines were kindly provided by Prof. Thomas Kunkel (NIEHS/NIH) and were maintained in a 5% CO₂ growth medium at 37°C. A colorimetric assay based on the dye 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was used to measure cell viability (35). The cells were first washed in PBS, trypsinized, and then seeded in a 96-well cell culture plate at 5 × 10³ cells per well and incubated overnight. ActD and Echi were prepared in fresh medium and added to each well at a varying concentration range for single drug treatment or a combination of two drugs each. The cells were further incubated for 48 h and 100 μl MTT (0.5 mg/ml) solution was added to each well. After 3 h of incubation, the MTT-containing medium was removed and 100 μl dimethyl sulfoxide solution was added to each well. A colorimetric measurement was performed at 570 nm to determine cell viability, relative to untreated control cells. Each experiment was repeated at least three times independently for statistical significance. The combination index (CI) has been widely used to determine the degree of drug interactions (36,37). The synergistic effect of the combination therapy of ActD and Echi was determined by calculating the CI, as previously described by Chou (38). The CI is defined as:

where, a is a small dose of drug 1 to be combined with a small dose of drug 2 (b) to get a desired cytotoxic effect and A and B are large doses of first and second drugs respectively, that causes the same effect. A CI value <1 indicates synergy, whereas a value greater than 1 indicates antagonism of the combination. A CI value of 1 indicates that the two drugs have an additive effect on the cells. To calculate CI, we used CompuSyn v1.0, and the cell viability values of ActD, Echi, and a combination of both at varying concentrations (39).

Lentivirus infection

The lentiviral particles sh-scramble (clone ID- ASN0000000003) and sh-MLH1 (clone ID-TRCN0000288642) were purchased from the National RNAi Core Facility at Academia Sinica, Taiwan. Both cloning vectors were pLKO_TRC005. The target sequence of sh-MLH1 was d(CCAAGTGAAGAATATGGGAAA) (NM_000249). For lentivirus infection, 2 × 10⁴ SW620 cells were seeded in 24-well plates and incubated overnight. The cells were then treated with medium containing lentivirus (M.O.I. = 10) and 10 μg/ml protamine sulfate (P4020, Sigma-Aldrich) and centrifuged at 2000 rpm for 10 min. After 24 h of infection, the medium was replaced with 2 μg/ml puromycin (P8833, Sigma-Aldrich) containing medium and the cells were further incubated for 48 h. A fresh medium was substituted after selection. Experiments were performed after at least two cell passages.

RT-qPCR analysis

Total RNA was isolated using Trizol™ reagent (Invitrogen, 15596026). The cDNA synthesis was performed using the PrimeScript™ RT Reagent Kit (TaKaRa, RR037A) according to the manufacturer's protocol. KAPA SYBR FAST qPCR Master Mix (2×) Universal (KK4600) was used for the Real-Time System (Applied Biosystems StepOne™). The qPCR protocol was performed as follows: 95°C for 10 min and 95°C for 10 s, 60°C for 30 s, for 40 cycles. Gene expressions were normalized to human GAPDH expression. The sequences of the primers used in this experiment were:

HIF-α: forward 5'-GAAAGCGCAAGTCCTCAAAG-3' and reverse 5'-TGGGTAGGAGATGGAGATGC-3'; MLH1: forward 5'-GCACCGGGATCAGGAAAGAA-3' and reverse 5'-GCACCGGGATCAGGAAAGAA-3'; GAPDH: forward 5'-GGCATCCTGGGCTACACTGA-3' and reverse 5'-GGAGTGGGTGTCGCTGTTG-3'.

In vivo tumour growth inhibition in HCT116 xenograft mice

The animal experiments were conducted in accordance with the guidelines and protocols (IACUC No. 109-024^R) approved by the Institutional Animal Care and Use Committee (IACUC) at the National Chung Hsing University. The mice were kept in individually ventilated cages and provided with standard diet and water. Male 7- to 8-week-old BALB/_C/Athymic NCr-nu/nu mice (National Laboratory Animal Center, Taiwan) were used for the experiment. 5 × 10⁶ HCT116 cells were suspended in PBS and injected subcutaneously into the right flank of the mice. After 12 days of inoculation, the tumour volume reached an average of > 50 mm³. The mice were then randomly divided into four treatment groups with the following number of mice per dosing group: Control (n = 5), ActD (n = 6), Echi (n = 6) and a two-drug combination group (n = 7). All drugs were prepared in 5 mg/ml BSA and administered intraperitoneally, so that the control group received only vehicle (5 mg/ml BSA), ActD (60μg/kg), Echi (10μg/kg) and the combination treatment (ActD (60 μg/kg) + Echi (10 μg/kg)) four times at one-day intervals. The body weights of the mice and the tumour volume were measured twice a week. Tumour volume (TV) was determined by measuring the width (W) and length (L) of the tumour with a digital calliper and calculated using the following formula TV = (L × W²) × 0.5. After a treatment for 21 days, the mice were sacrificed using carbon dioxide. The tumours and the livers of the mice were then removed to determine their weights.

Hematoxylin and eosin (H&E) stain and tunel assay

Excised tumour samples were fixed in 4% paraformaldehyde solution and embedded in paraffin. The dewaxed paraffin sections were stained with H&E for histological analysis. Apoptosis cells were detected using the Tunel assay according to the manufacturer's protocol (One-step TUNEL In Situ Apoptosis Kit, catalogue number E-CK-A320D, AMSBio Inc). The cell nuclei of the tumour sections were stained by mounting medium with DAPI (Abcam, ab104139) and sealed with a coverslip. Fluorescence images were taken with an Olympus IX73 inverted microscope. Apoptosis cells were quantified manually using ImageJ software.

Statistical analysis

Statistical analyses between groups were evaluated using a two-tailed t-test from GraphPad Prism v5.01 (San Diego, CA, USA). One-way ANOVA was used for the qPCR experiments. P-values smaller than 0.05 were considered statistically significant.

RESULTS

The combination of ActD and echi can preferentially stabilise thymine-related DNA duplexes containing T:T mismatched base pair

To understand the stabilising effect of the combination of ActD and Echi, we designed several DNA duplexes containing 5'-GpC and 5'-CpG steps separated by a mismatch or canonical Watson-Crick base pair (Figure 2A). The thermal denaturation profiles of DNA duplexes in different stoichiometry ratios with ActD, Echi and a combination of ActD and Echi showed a well-defined sigmoidal change with a steep slope implying the formation of drug-DNA complexes (Supplementary Figure S2). The values of melting temperatures (T_m in °C) for each DNA duplex in the presence and absence of drugs are shown in Table 1. We use the fixed DNA:ActD:Echi stoichiometry ratio of 1:1:1 and 1:2:2 to determine the differences in melting temperature (ΔT_m) of each duplex as ΔT_m1 and ΔT_m2, respectively for the two drug combination treatment. We then compared the stability of each DNA duplex in the presence of two antibiotics for the 8 mismatches and 2 Watson-Crick base-pairs by measuring their ΔT_m values. Under identical conditions, the addition of equimolar concentrations of two drugs in the ratio of 1:1:1 DNA:ActD:Echi can significantly stabilise the T:T mismatched DNA duplex (ΔT_m1 value of 35.2 ± 0.2°C), followed by the T:C mismatched DNA duplex (ΔT_m1 value of 23.7 ± 0.6°C) compared to other mismatches or Watson–Crick DNA duplexes (Figure 2B). A similar trend is observed even at higher DNA:ActD:Echi ratios of 1:2:2. On the other hand, the ΔT_m values for completely canonical A:T and G:C duplexes are relatively low, ranging from 12.1 ± 0.3 to 22.2 ± 0.1°C for the combination treatment. Notably, the combination has greater stabilising effects on the A:T duplex over the G:C. Previous study has shown that, under physiological conditions, the A:T base pair is easy to adopt Hoogsteen conformation which reduces the interstrand C1'–C1' distance than that of the G:C base pair (40). The reduction in the C1'–C1' distance usually enhances the binding affinity of the ligand (41,42), explaining the greater stability of A:T duplexes upon ActD and Echi binding. From the melting temperature analysis, it is also evident that the individual drug treatment of either ActD or Echi at 1:1 or 2:1 drug:DNA ratios, the stability of DNA duplexes is relatively low compared to the combination drug treatment. These results suggest that DNA duplexes with alternating 5'-GpC and 5'-CpG sites separated by a single T:T or T:C mismatch appear to provide a suitable environment for the simultaneous intercalation of ActD and Echi drugs compared to other base pairing at same position.

Figure 2.

(A) Schematic diagram showing 11-mer DNA duplex used in melting temperature analysis. The central X-X represents any of the four bases A, T, G or C used to generate various mismatches and Watson-Crick base pair containing duplexes. ActD and Echi binding sites are represented in yellow and green colours, respectively. (B) The stabilising effects of ActD and Echi on various DNA duplexes (3 μM) is shown as melting temperature changes (ΔT_m) for different mismatches and Watson-Crick DNA duplexes in the presence of ActD (3 μM) and Echi (3 μM) in a buffer containing 100 mM sodium cacodylate (pH 7.2) and 100 mM NaCl. Error bars represent the standard deviations of three independent experiments. (C) CD spectra of the T:T mismatch containing 11-mer DNA duplex (10 μM) in presence of different ratios of ActD and Echi combination obtained by subtracting out the spectra of drugs in 100 mM sodium cacodylate (pH 7.2) and 100 mM NaCl. The change in CD intensities at 295 nm as a function of the molar equivalents of drug-DNA at 25°C is shown on the right. The two solid lines represent the initial binding curve of ActD and Echi to duplex DNA and the curve that reaches the plateau. The vertical dashed line indicates the plateau reached when two equivalents of ActD and Echi combination are added to duplex DNA.

Table 1.

The effects of ActD, Echi and a combination of ActD and Echi at various concentrations on melting temperature (T_m) of different mismatches and Watson-Crick base pair containing DNA duplexes at 3 μM

	T m (°C)
DNA	DNA alone	DNA: Echi [1:1]	DNA: Echi [1:2]	DNA: ActD [1:1]	DNA: ActD [1:2]	DNA: ActD:Echi [1:1:1]	DNA: ActD:Echi [1:2:2]	ΔTm1 (°C)a	ΔTm2 (°C)b
A:T	47.5 (0.5)	53.9 (1.1)	56.5 (0.5)	52.9 (0.1)	55.2 (1.3)	68.3 (0.4)	69.7 (0.6)	20.8 (0.5)	22.2 (0.1)
G:C	54.0 (0.4)	60.6 (0.3)	61.6 (0.7)	61.5 (1.1)	63.9 (0.1)	66.1 (0.2)	69.6 (0.8)	12.1 (0.3)	15.6 (0.4)
T:T	33.1 (0.2)	47.3 (0.3)	50.2 (1.0)	44.6 (0.6)	48.6 (0.4)	68.3 (0.4)	68.8 (0.7)	35.2 (0.2)	35.7 (0.5)
T:C	30.2 (0.3)	38.1 (1.5)	40.5 (0.8)	40.3 (0.8)	44.2 (0.4)	53.9 (0.9)	57.2 (0.9)	23.7 (0.6)	27 (0.6)
C:C	30.7 (0.7)	40.6 (0.6)	42.6 (1.2)	45.8 (0.1)	48.0 (0.5)	51.7 (0.5)	55.3 (0.2)	21.0 (0.5)	24.6 (0.6)
A:G	40.9 (0.6)	54.2 (1.0)	58.4 (0.6)	51.7 (0.7)	56.8 (0.6)	57.7 (0.5)	61.6 (0.8)	16.8 (0.2)	20.7 (0.5)
G:T	43.8 (0.4)	53.8 (0.7)	53.5 (0.6)	51.8 (0.8)	54.9 (0.6)	57.4 (0.7)	60.1 (0.9)	13.6 (0.4)	16.3 (0.5)
G:G	43.1 (0.5)	49.0 (0.7)	53.8 (0.7)	50.1 (0.5)	53.1 (0.6)	56.6 (0.9)	60.3 (1.0)	13.5 (0.4)	17.2 (0.5)
A:A	34.2 (0.4)	46.1 (1.3)	49.1 (1.2)	40.9 (0.8)	43.9 0.4)	55.1 (0.4)	57.3 (1.0)	20.9 (0.2)	23.1 (0.7)
A:C	38.0 (0.8)	43.4 (0.9)	46.1 (0.9)	45.4 (0.6)	48.8 (1.0)	57.3 (0.7)	61.2 (0.6)	19.3 (0.3)	23.2 (0.2)

^aΔT_m1 was calculated as the difference between the T_m of DNA alone and T_m of DNA:ActD:Echi (1:1:1). ^bΔT_m2 was calculated as the difference between the T_m of DNA alone and T_m of DNA:ActD:Echi (1:2:2). The T_m values are averages from three experiments. Values in parentheses indicate the standard deviations from three independent experiments.

We determined the stoichiometry of the binding of ActD and Echi to the DNA duplex containing T:T mismatch using CD spectroscopy (Figure 2C and Supplementary Figure S3). The CD spectra of the DNA duplex (indicated by a blue line in the respective figure) showed a negative and a positive peak at 250 and 275 nm, which is characteristic of B-DNA conformation. After adding different concentrations of ActD and Echi, the spectra show a decrease in the intensity of the negative peak at about 250 nm and a red shift from ∼275 to ∼292 nm, indicating a conformational change of the DNA after drug binding. As described previously, we plotted the change in intensity of CD at 295 nm as a function of the concentration of the added drugs (43). The plot shows that a plateau is reached when two equivalents of ActD and Echi were added in combination. The intersections of the two extrapolated lines representing the saturation values indicate a 2:1 stoichiometry between drugs and DNA for the combination treatment (Figure 2C). On the other hand, single drug treatment reaches a plateau when about 1.5 equivalents of ActD or Echi are added to the DNA, indicating that single drug treatment is unable to fully and specifically bind to the T:T mismatch containing duplex (Supplementary Figure S3).

The binding of ActD and echi generates an intrinsic preference for a T:T mismatch over a T:C mismatch and A:T and G:C Watson–Crick DNA duplexes

In addition to the T:T mismatch, a thymine-related T:C mismatch duplex is also stabilised by the binding of ActD and Echi in combination. We determined the binding affinity in terms of the dissociation constant (K_d) for T:T and T:C mismatch and A:T and G:C Watson–Crick base pair containing duplexes in presence of ActD, Echi and a combination of both drugs by fluorescence polarisation (FP) binding analysis. Supplementary Figure S4 shows a sequence design and FP response curves for different DNA duplexes in the presence of ActD, Echi and a combination of two drugs. For all DNA duplexes, the K_d values for ActD or Echi when titrated separately are ranged between 2.24 ± 0.31 μM to 3.69 ± 0.32 μM (Table 2). When titrated in combination, the dissociation constant for each drug in these DNA duplexes decreases significantly. For a T:T mismatch, the K_d values for ActD and Echi in combination are 0.18 ± 0.02 and 0.13 ± 0.00 μM, respectively, which is about 14- to 19-fold lower than the K_d values of individual ActD or Echi and about 7- and 5-fold lower than the K_d values of ActD and Echi in combination for T:C mismatched duplexes (1.22 ± 0.06 and 0.69 ± 0.13 μM, respectively). Under similar conditions, the perfect Watson–Crick A:T and G:C DNA duplexes showed about 6 to 10 times higher K_d values for ActD and Echi in combination, implying a significantly lower binding affinity compared to T:T mismatch. These results suggests that the combination of ActD and Echi is more likely to bind thymine-related mismatches in a DNA-drug complex with highest binding affinity of two drug combination for T:T mismatch flanked between GpC and CpG sites, respectively.

Table 2.

Binding affinity of ActD, Echi and a combination of ActD and Echi to T:T and T:C mismatch DNA duplexes and A:T and G:C Watson–Crick DNA duplexes measured by fluorescence polarisation (FP) binding assay

DNA duplex	Ligand	K d (μM)#	R 2
T:T mismatch	ActD alone	2.52 ± 0.22	0.96
	Echi alone	2.57 ± 0.21	0.96
	Combination (ActD)	0.13 ± 0.00	0.98
	Combination (Echi)	0.18 ± 0.02	0.98
T:C mismatch	ActD alone	2.96 ± 0.54	0.97
	Echi alone	2.50 ± 0.28	0.97
	Combination (ActD)	1.22 ± 0.06	0.98
	Combination (Echi)	0.69 ± 0.13	0.97
A:T Watson-Crick	ActD alone	2.24 ± 0.31	0.97
	Echi alone	2.49 ± 0.33	0.97
	Combination (ActD)	1.09 ± 0.23	0.96
	Combination (Echi)	1.38 ± 0.08	0.97
G:C Watson-Crick	ActD alone	2.82 ± 0.32	0.98
	Echi alone	3.69 ± 0.32	0.98
	Combination (ActD)	1.32 ± 0.16	0.96
	Combination (Echi)	1.00 ± 0.12	0.97

^#The data were fitted to a one-site total binding curve equation in GraphPad Prism (version 5.01). Best-fit values with standard errors for triplicates are shown.

Intercalation of ActD and echi into thymine-mismatched DNA duplexes induces substantial single-strand distortion in the T:T mismatched region

To understand the basis for the preference of ActD and Echi combination for binding thymine-related T:T mismatch, we have determined the crystal structure of a T:T mismatch-containing DNA duplex complexed with ActD and Echi in a tetragonal P4₁2₁2 space group at 2 Å resolution. Each asymmetric unit contains a DNA duplex with ActD and Echi molecules intercalated at the 5'-GpC and 5'-CpG sites, respectively, and separated by a single wobble-type T:T mismatch (Figure 3A). The entire structure is stabilised by end-to-end crystal packing interactions mediated by π–π stacking between the terminal residues of each duplex and side-to-side interactions of Zn²⁺ metal ions and water molecules in the major groove (Supplementary Figure S5A). In the side-to-side interactions, two Zn²⁺ ions are found; one Zn²⁺ metal ions coordinating the intermolecular interactions between the two guanine N7 atoms of G10 and G10* (* indicates a symmetry-related position) through a common octahedral coordination geometry (44,45). Yet another Zn²⁺ display an unusual square-pyramidal geometry through the N7 atom of guanine G6 and the OP1 of T4* as well as two water molecules and one chloride ion (Supplementary Figure S5B).

Figure 3.

The overall structure of the T:T mismatch duplex in complex with ActD and Echi. (A) Assembly of a crystal structure containing a central T:T mismatch in a DNA duplex complexed with ActD and Echi shown in front view (left) and side view (right). The bases in the complementary strand are numbered from A1 to T7 and A8 to T14. The T:T mismatch bases are shown enlarged on the right-hand side. The skeletal models show the intercalation sites of (B) ActD (left) and (C) Echi (right). The phenoxazone (PXZ) ring of ActD is parallel to the horizontal axis in DNA while the quinoxaline (QUI) of Echi tilts about 16°. (D) Various non-bonding interactions between the drugs and DNA stabilize the central T4:T11 mismatch. The π-π stacking interaction between QUI0 and T4 is indicated by a thick red dashed line. The van der Waals interactions of D-serine (DSN1) and N-methylvaline (MVA8) with T4 and T11 are indicated by a black dashed line. The C–H•••π contact between proline (PRO9) of ActD and T4 oxygen is indicated by a red dotted line. (E) The close intermolecular contacts between ActD and Echi cyclic peptides stabilize the complex structure. The C-H•••O interactions between the carbonyl oxygen of Echi-DSN1 with the MVA11 side chain and the N-methyl of sarcosine (SAR10) of ActD are shown by a red dotted line. The black dashed line represents the van der Waals contacts between MVA8 and DSN1 of Echi with the SAR10 and MVA11 residues of ActD.

In the complex structure, the phenoxazone ring (PXZ) of ActD is intercalated between the G2:C13/C3:G12 base pair step, whereas the two bulky cyclic peptide rings (α and β) enlarge the width of the minor groove and partially unfold the duplex (Figure 3B). The two quinoxaline rings (QUIs) of Echi are bis-intercalated into the T4:T11/C5:G10 and G6:C9/T7:A8 base pair steps, while the adjacent QUI of Echi and the cyclic peptide ring of ActD on both sides of the mismatches are interleaved to avoid steric hindrance. The methyl groups of MeVal (MVA) from the α- and β-rings of ActD are wedged between the A1pG2 and T11pG12 steps, respectively, resulting in asymmetric twists in the respective backbones of these two steps. The QUI0 of Echi is inserted into the T4pC5 step and tilted about 16° from the horizontal axis in the duplex to chain A, causing this step to be overwound with a high twist angle. These large changes in the twist and tilt angles at the central G10pT11pG12 residues in chain B result in substantial single-stranded backbone distortion to create a ‘chair-like shape’ in this DNA strand, while keeping the other chain A aligned with the helical axis (Supplementary Figure S6A). The PXZ ring of ActD is parallel to Watson-Crick hydrogen bonds of adjacent base pairs and formed extensive stacking interactions with adjacent G2 (G12) bases on either side of the ring. Short intermolecular hydrogen bonds are formed between the 2-amino groups of PXZ and the O4' atoms of the C13 nucleotide. The hydroxyl groups of PXZ are involved in bifurcated hydrogen bonding with the N2 and O2 atoms of G2 (G12) and C13 (C3) nucleotides, respectively. In addition, the CO/NH backbone atoms of the Thr1 and Thr7 residues of ActD and the N2/N3 atoms of G2 and G12 nucleotides exhibit good hydrogen-bonding interactions in both complex structures (Supplementary Figure S6B). The cyclic peptide ring of ActD is found to be far from the DNA base pairs on the minor groove side, resulting in a lack of direct contacts between the peptides and DNA bases. However, the proline (PRO) side of the peptide ring appears to push the cytosine bases (C3 and C13) toward the major groove, reducing their stacking interactions.

Unlike the planar intercalation of ActD, the QUI ring in the Echi complex, on the other hand, tilts towards the ‘A chain’, forming two consecutive triple stacking interactions between segments T4-QUI0-G10-C9 and A8-QUI9-G6-C5, providing further stability to the complex structures (Figure 3C). The CG base pairs within the quinoxaline intercalation sites are considerably buckled due to the interactions between the βC atoms of Echi alanine (ALA) with the nucleotide sugars, which favors intermolecular hydrogen bonding between the CO/NH of the two ALA residues and the N2 and N3 atoms of the G6 and G10 bases (Supplementary Figure S6C). The central T:T mismatch is stabilised by weak interactions between QUI0, D-serine (DSN1), and MVA8 residues of Echi and PRO9 of ActD. QUI0 forms a π-π stacking interaction with the thymine T4, while the carbonyl groups of DSN1 and MVA8 interact with T4 and T11 bases through van der Waals interactions. The β/γ carbon atoms of PRO9 form unique C–H···π contacts with T4 to further stabilise the mismatched pair (Figure 3D). In addition to these drug-DNA interactions, the proximity of ActD and Echi in the minor groove also results in close contact between the two cyclic peptide rings. Further, the carbonyl oxygen atom of Echi-DSN1 forms a weak C–H···O interaction with the MVA11 side chain atoms and the N-methyl group of sarcosine (SAR10) of ActD. The bulky methyl groups of Echi-MVA8 and the carbonyl atom of DSN1 residues thus participate in extensive van der Waals interactions with the SAR10 and MVA11 residues of ActD (Figure 3E). Moreover, an opening angle of ∼10°, together with a negative base pair stretch (–1.8 Å), causes the central T4:T11 mismatches to move apart (Supplementary Figure S7A). In particular, T11 is pushed towards the major groove side, which leads to the formation of a cavity in the minor groove near O2 of T11 to avoid crowding of the N-methyl group of MVA11 of ActD on the G12pX11 step for the T:T mismatch. The T4 sugar pucker has an O1'-endo conformation and results in an intrastrand phosphate-phosphate distance of 6 Å. T11 has a C2'-endo conformation with an intrastrand phosphate group distance of 6.5 Å (Supplementary Figure S7B). This arrangement of the sugar pentose atoms at T4 and T11 further contributes to the significant distortion twisting of individual DNA strands. Asymmetric distortions in the DNA backbone are revealed by the torsion and base-pair helical parameters, discussed in detail in the Supplementary data (Supplementary Figures S8).

Comparison of DNA–drug complexes containing a central T:T mismatch and a central T:C mismatch

To understand why ActD and Echi prefer to bind with T:T mismatched DNA duplexes over a T:C mismatch, we solved the high resolution crystal structure of a ternary complex containing d(AGCTCGT/ACGCGCT) intercalated with an ActD and Echi in a similar space group (P4₁2₁2) to that of the T:T complex (Supplementary Figure S9A). In this complex, the central T:C mismatch forms a wobble-type pairing via two hydrogen bonds. Octahedral coordination of two Mg²⁺ ions to symmetry-related T4* and G10* residues, including water molecules via the N7 atoms of G6 and G10 residues, stabilises the whole complex through side-by-side interactions. In addition, end-to-end π–π stacking interactions between the terminal bases of the two symmetry-related complexes further stabilises the structure, in a way that resembles the T:T complex (Supplementary Figure S9B). Overall, the T:T and T:C complexes display significant similarities in DNA conformations upon intercalation of the two drugs. However, when the two complexes are superimposed, the root-mean-square (r.m.s.) deviation is 0.4 Å (409 DNA atoms), suggesting that although these complexes are generally very similar, they still exhibit some local differences (Figure 4A). The overall structural comparison are discussed in the Supplementary data (Supplementary Figure S10).

Figure 4.

Comparison between T:T and T:C mismatch complex structures. (A) Superimposition of T:T (blue) and T:C (red) complexes with a root-mean-square deviation of 0.4 Å for all DNA atoms, indicating that the two complexes are globally similar but locally slightly different. Superimposition of T:T and T:C mismatch (right) shows local differences in bases T11 and C11 in the two complexes. (B) Comparison between the geometries of central T4:T11 and T4:C11 mismatches in the two complex structures. Both the T:T and T:C mismatches form wobble pairs, where T:T has four H-bonded interactions mediated by a water molecule and T:C has two hydrogen bonds. A reduction of the C1'-C1' distance in both mismatched complexes is also shown. (C) Residues of ActD [A] and Echi [E] form van der Waals interactions between T4:T11 in the T:T complex and T4:C11 mismatch in the T:C complex, analysed with LIGPLOT + .

The main difference between these structures lies in the environment of the central mismatched T4:X11 pair (X represents T or C) and the interactions between the drug and the DNA bases. In the T:T complex, the central T4:T11 pair is formed by two hydrogen bonds between the thymine bases and by two extra hydrogen bonds formed by the coordination of a water molecule with the two carbonyl oxygen atoms on carbon four of thymine (left of Figure 4B). In the T:C complex, however, the presence of an amino group (rather than a carbonyl group) on carbon four of base C11 appears to cause a larger base pair opening and the ability to bind a water molecule is lost, resulting in only two hydrogen bonding interactions at the mismatch site (right of Figure 4B). The presence of wobble-type T:T and T:C mismatches also leads to a reduction in DNA diameter, as shown by a reduction in C1'-C1' distances and a negative stretch value at the central mismatch sites. However, LigPlot+ analysis shows that compared to the T:C complex, the central T11 in the T:T complex appears to contribute a greater number of van der Waals interactions, which may be due to the presence of an additional methyl group at carbon (C5) of the thymine (Figure 4C). T4 in both complexes adopts the O1’-endo sugar pucker conformation. T11 in the T:T complex adopts a C2’-endo sugar pucker while C11 in the T:C complex adopts a C2’-exo sugar conformation. These local differences suggest that the X11 base in the T:T and T:C mismatch complexes plays a key role in causing the differential stability of pyrimidine mismatches following ActD and Echi intercalation. On the other hand, the T4 base is important in acting as a support and providing a consecutive parallel stacking interaction with Echi-QUI0 to stabilise the overall mismatched structure.

Combination of ActD with echi shows synergistic effects against MMR-deficient HCT116 cancer cells

Our structural and biophysical experimental results show that the combination of ActD and Echi can preferentially recognise thymine-related T:T mismatched DNA duplexes in a 1:1 molar ratio. Therefore, we hypothesized that this combination would increase selectivity and cytotoxicity to those cancer cells characterized by aberrant DNA mismatch repair. To determine the therapeutic potential of the combined two drugs approach in vitro, we used the human colon cancer cell line HCT116, which lacks normal MMR and transcription-coupled excision repair activity, and therefore retains a high proportion of mismatches in its genome (46). Also, HCT116 cells are widely used to study the anti-proliferative effects of various DNA intercalators (47). Other human colon cancer cell lines HCT116 + ch3 and SW620, which retain normal mismatch repair were used for comparison.

Both the MMR-deficient HCT116 and proficient HCT116 + ch3 cells show similar effects when treated with ActD or Echi individually, in a dose-dependent manner. However, the combined treatment of two drugs (1:1 molar ratio) significantly inhibits cell growth in HCT116 cells. Equimolar concentrations of ActD and Echi result in a combination index (CI) value of ∼0.5, indicating synergistic effects of the combination treatment on HCT116 MMR-deficient cells. On the other hand, for HCT116 + ch3 proficient cells, the CI value ranges from 0.9 to 1.1, indicating an additive effect of the drug combination (Figure 5A and Supplementary Figure S11). To further confirm the synergistic effect of ActD and Echi on MMR deficient cells, we knocked down the MMR-related gene MLH1 in SW620 MMR-proficient cells using short hairpin RNAs (shRNAs) (Supplementary Figure S12). The results show that sh-MLH1 SW620 cells become more sensitive to the combination treatment compared to sh-scrambled SW620 cells. However, individual drug treatments show no obvious differences. The CI values for sh-MLH1 SW620 also show strong synergy (CI range, 0.52–0.58), while the control group shows moderate synergy (CI range between 0.70 and 0.85) (Figure 5B and Supplementary Figure S13). Overall, these results indicate that MMR-deficient cancer cells are more sensitive to treatment with the two drugs, which we suggest may due to their strong stabilisation of DNA mismatches.

Figure 5.

The combination of ActD and Echi exhibits a synergistic effect on inhibiting the growth of MMR deficient cells. (A) HCT116 + ch3 (MMR proficient) and HCT116 (MMR deficient) cells; (B) sh-scramble (MMR proficient) and sh-MLH1 (MMR-deficient) cells were treated with ActD, Echi and a combination of ActD and Echi in a dose dependent manner for 48 h. Cell viability was evaluated by the MTT assay. IC₅₀ values of ActD and Echi treatment were defined by nonlinear curve-fitting analysis using Graphpad Prism v5.01. Data are represented as mean ± SD. Statistical analysis performed by two-tailed test. *P < 0.05, **P ≤ 0.025, ***P ≤ 0.001. CI, combination index; at IC₅₀, half-maximal inhibitory concentration; IC₇₅, 75% of maximal inhibitory concentration; IC₉₀, 90% of maximal inhibitory concentration were calculated by CompuSyn v1.0.

Combination of ActD with echi reduces in vivo tumour volumes in HCT116 xenograft mice model

To investigate the combination treatment in vivo, nude mice bearing subcutaneously inoculated HCT116 cells were treated with ActD (60 μg/kg) (48–51), Echi (10 μg/kg) (52–54) and in combination (ActD (60 μg/kg)+Echi (10 μg/kg)) by i.p injection. The xenograft mice protocol is shown in Figure 6A. The individual drug treatment group shows a reduction in tumour volume growth curve in only a few of the mice cohort while tumour weights had no significant differences compared to the control group. However, the combination group significantly decreases tumour volume and tumour weight for all treated mice (Figure 6B, C and Supplementary Figure S14). H&E staining revealed increased cellular necrosis on tumor tissue from the combination group. To detect apoptosis cells in the tumours, we examined TUNEL-positive cells from tumour sections. The combination increased apoptosis cell numbers in tumours to a greater extent than the individual drug treatment (Figure 6D). Adverse toxicological effects of ActD (55–57) and Echi (58–60) have been previously extensively reported. In our current model, we observed that ActD causes loss of appetite and weight loss on mice during the treatment period but did not reduce mouse activity. However, mouse weights recovered in the latter part of the experiment (Figure 6E). No mice died as a result of the treatment. After mouse sacrifice, we found that liver weight has no significant change (Figure 6F). These results show that the combination treatment suppresses tumour growth to a significantly greater extent than single drug treatment.

Figure 6.

Combination treatment inhibits tumour growth in HCT116 mice xenografts. (A) Schematic representation of the xenograft mice model experiment. (B) Averaged tumour growth curves and image of the tumor tissues of mice in control (n= 5), ActD (n= 6), Echi (n= 6) and combination (n= 7) group treatment. All data are represented as mean ± SD. Scale bars, 10 mm. (C) Statistics of tumour weights obtained at the end of experiment. (D) Representative H&E-stained images of mice tumour post-treatment at day 21 after the treatment. Scale bars, 50 μm. TUNEL (green) and DAPI (blue) staining of tumour sections are shown. Scale bars, 50 μm. Quantification of the number of TUNEL-positive cells by calculating four high power fields from tumour area (n= 5 per group) is shown on the right. (E) The body weight change plot for mice from treatment start time of four treatment groups (red arrow indicate execution of the treatment). (F) Liver/body weight ratio of post-treatment for day 21. All data are represented as mean ± SD. Statistical analysis performed by two-tailed test. *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001.

DISCUSSION

Combination chemotherapy has become an important approach for improving the otherwise poor survival rates of many cancer patients (61). Combination chemotherapy is based on the use of two or more drugs that can enhance lethal effects on cancer cells while minimising toxicity to normal cells. In addition, compared with the treatment using only a single drug, combination chemotherapy can provide an enhanced range of drug activity against cancer cells, as well as minimising drug resistance, providing synergistic therapeutic effects, and limiting the toxicity of a single drug (62–64). Many studies have described drug combinations resulting in minimal side effects and higher efficacy (65–67). The chemotherapeutic agents used in combination can act on different targets, such as proteins and nucleic acids. For example, the combination of two different enzyme inhibitors, ixabepilone and capecitabine, has shown enhanced efficacy in treating breast cancer through a synergistic effect (68). Xia et al. demonstrated the potential of combining lipophilic bisphosphonates with rapamycin as a novel strategy to inhibit farnesyl and geranylgeranyl diphosphate synthases and mTOR pathway proteins respectively, in the treatment of lung adenocarcinoma (12). The simultaneous binding of two different drugs (ethidium and proflavine) to a multi-drug binding protein demonstrates the synergistic binding mechanism of a target protein through structural rearrangements (69). Another DNA-targeting anticancer agent, cisplatin, has been widely used in combination with a variety of DNA-binding compounds and protein inhibitors, to show enhanced anti-cancer activity and the overcoming of platinum resistance (70). The combination of two putative DNA-binding agents clofazimine and diminazene aceturate has potential applications in the treatment of piroplasmosis (71).

Despite extensive studies on drug combinations, the molecular basis of their effects frequently remains poorly understood, particularly from a structural point of view. Detailed knowledge about the intermolecular interactions between small molecules and their targets could open new opportunities for drug discovery (72–74). Therefore, to elucidate the mechanism of combining DNA-targeting drugs, we have determined the ternary structures of mismatched DNA sequences with two well-known anticancer drugs ActD and Echi. Recently, we have shown that the intercalation of two different drugs in different base-pair context resulted in polymorphic DNA structures (75). Consistent with these observations, intercalation of ActD and Echi into duplex DNA also causes only one DNA strand to be significantly distorted while the other strand remains parallel to the DNA helical axis, at least in the current structures. Detailed structural analysis shows that distortions of the DNA backbone are distinct from those produced by either drug alone. Many studies have reported that the binding of a first small molecule to its target occurs via a paired cooperative binding model, which facilitates the binding of a second molecule to recognise the target sequence with higher affinity (10,76,77). The ternary complex presented here reveals evidence of the cooperative effects of the ActD and Echi combination in recognising T:T mismatched DNA. We compared the current structure with the reported structures of two ActD molecules complexed with a T:T mismatch flanked by two consecutive GpC steps (PDB:1MNV) or with two Echi molecules complexed with a T:T mismatch flanked by two consecutive CpG steps (PDB: 5YTZ) (10,22). Superimposing these three structures highlights the conserved binding mode of ActD and Echi with the major differences in DNA backbone structure (Figure 7A). The alternate twist angles at the CpT/TpG (∼12°) and TpC/GpT (∼42°) steps in the current T:T complex structure cause the backbone to bend. This will generate lower values of roll (∼8° and –10°, respectively) in the base pairs along its long axis and compresses the major groove, while widening the minor groove of the double helix, which would facilitate the insertion of a second drug moiety (Figure 7B). Furthermore, in the current T:T mismatch structure, the 16° tilt angle between the planes of C5 (C9) and G6 (G10) residues is significantly higher than the 12° angle in the earlier Echi complex (PDB: 5YTZ). The C3 base is tilted by 3.5° toward the central T4, which is significantly different compared to the base pair tilt of –9° (between G3 and T4) observed in the previous Echi complex. The differences between these tilts increase stacking interactions in the current complex structure compared to the previous T:T mismatch complex. In addition, the central T4:T11 mispair opens at an angle of 10°, which is notably smaller than the 15° and 25° openings in the previous Echi and ActD complexes, respectively (Figure 7C). This lower opening reduces the interstrand C1'–C1' distances which aligns the central mismatch with other polynucleotides and drug residues in the same plane to maximize stacking interactions and provide further stability to the complex. Other major differences within these three T:T mismatch structures are found in the degree of staggering of drug molecules. In the current T:T mismatched structure, the ActD and Echi molecules are staggered at an angle of about 55°, whereas in the structures with two separate ActD or Echi, these orientation angles are about ∼40° and ∼45°, respectively (Supplementary Figure S15A). This staggering causes the cyclic peptides of ActD to come in closer contact with Echi, thereby further stabilizing the complex with minimal steric clashes. For the two-Echi molecule complex, the DSN and MVA residues of the depsipeptide form two van der Waals contacts and an oxygen-mediated head-to-head carbonyl group interaction between the MVA residues, with an average distance of about 4.4 Å (Supplementary Figure S15B, left). In the two-ActD molecule complex, the MVA and SAR side chains formed two van der Waals interactions within a distance of about 4.1 Å (Supplementary Figure S15B, center). In contrast to these structures, the SAR10 and MVA11 of ActD and DSN1 and MVA8 side chains of Echi are in proximity to form two C•••C van der Waals contacts and two C–H•••O interactions, with a separation of 3.9 Å (Supplementary Figure S15B, right). Compared to two separate ActD or Echi drugs bound to a mismatched DNA, this shorter interaction network provides greater stability for the simultaneous binding of ActD and Echi and reflects superior tightness. These observations are also supported by our biophysical experiments. For example, the equimolar combination of ActD and Echi showed a significantly higher stabilising effect compared to ActD or Echi alone with the T:T mismatch DNA. The melting temperature curve profiles show a steeper slope in the linearly increasing region when 1 equivalent of Echi is used to stabilise the T:T mismatch duplex, indicating a 1:1 stoichiometry for the DNA:Echi complex. However, in the case of ActD, the slope became steeper as the equivalent of ActD increased from 1 to 2, indicating a 1:2 stoichiometry for the DNA:ActD complex. Such non-specific binding of ActD to sites other than the GpC sites has also been observed previously (78,79). In the combination treatment, however, the steepness of the slope does not show significant differences in the 1:1:1 or 1:2:2 ratios of DNA:ActD:Echi treatments, suggesting that 1 equivalent of each of ActD and Echi can sufficiently stabilise the T:T mismatch duplex. In the CD spectra, the isobestic point at 295 nm is observed when one equivalent each of ActD and Echi is titrated in combination against the T:T mismatch duplex, supporting the 1:1:1 stoichiometry of DNA:ActD:Echi in solution. Thus, these results suggest that together the two distinct drugs have a more profound impact on T:T mismatch DNA structures flanked between GpC and CpG sites than a single drug alone.

Figure 7.

Structural comparison of different T:T mismatched complexes in the presence of ActD and Echi. (A) The superimposition of the current T:T mismatched complex containing ActD and Echi each (PDB ID: 7DQ0) with the previous ActD-T:T complex containing two ActD molecules per duplex (PDB ID: 1MNV) and Echi-T:T complex containing two Echi molecules per duplex (PDB ID: 5YTZ) (front and side views). Comparison between the DNA helical parameters in the current T:T mismatched complex (PDB ID: 7DQ0) and the previous T:T mismatch in the Echi-T:T complex (PDB ID: 5YTZ) and ActD-T:T complex (PDB ID: 1MNV). (B) Large structural variations were observed in the twist (in °) and roll (in °) DNA base pair step parameters, resulting in a profound impact on drug binding. (C) Variation in the DNA tilt (in °) and base pair openings (in °) in the three complex structures are significantly different.

At present, combination chemotherapy is one of the more effective treatment methods for various cancers. Combining several therapeutic agents has shown promising results in alleviating tumour diseases (80,81). We have identified the basis for applying two different intercalators that can synergistically recognise mismatched DNA by inducing specific changes in the DNA backbone. Based on detailed structural analysis, we propose three general rules to support the synergistic effect for combination tumour chemotherapy.

1. The two drugs when bound must avoid steric hindrance; to prevent steric clashes between the drug residues, it is necessary to intercalate and stagger two different drugs.

2. Binding synergy is usually mediated by ligand-ligand interactions, as observed in protein–DNA and drug–DNA complexes. Here, we find that the synergistic binding effect of two different intercalators to DNA is more effective than the single-drug bound complexes mediated by drug-to-drug interaction network. The degree of staggering keeps the two drugs within an optimum distance to exhibit strong interactions between the drugs.

3. The two-drug approach maximizes the binding force of the two drugs to the DNA duplex by forming three consecutive triple stacking interactions. This is contrary to the single-drug intercalation mode, such as the two Echi or two ActD molecular approaches, where only two consecutive stacks interact to stabilise the DNA duplex. Such increased continuous stacking interactions resulted from drug combinations can firmly stabilise long stretches of DNA.

Meeting these three general rules will lead to additional effects on DNA that cannot otherwise be achieved with only a single drug.

Our results show that the two-drug combination is more effective than either drug alone in inhibiting the growth of HCT116 MMR-deficient cancer cells. It has been reported that in MMR-deficient and HeLa cells, the repair efficiency of thymine-related T:T and G:T mismatches is low, suggesting that these mismatches may occur more frequently in cancer cell genomes (82,83). In the current study, we have determined the structural basis for combining these two different intercalators to simultaneously identifying the thymine-related T:T mismatch, which has a higher priority than other mismatches and Watson–Crick base-paired DNA. Consistent with these observations, the in vitro cell toxicity results also show that MMR-deficient cells are more sensitive to treatment with the combination drug. The possible mechanism of cell death is likely to be a direct consequence of the two drugs binding strongly to DNA in MMR-deficient cancer cells. Echi has been shown to inhibit HIF-1α expression by binding to the hypoxia response elements sequence of HIF-1α in the DNA promoter region (21). To rule out the possibility that the synergism between the combination of Echi and ActD is due to the effect of Echi on HIF-1α, we treated MMR-deficient cancer cells with Echi at different concentrations. Under normal oxygen conditions, the relative expression levels of HIF-1α showed no effect, indicating that the inhibition of cell growth was not influenced by HIF-1α expression (Supplementary Figure S16). Furthermore, the xenograft model reveals that combination treatment significantly reduces tumour growth to a greater extent than treatment with Act or Echi alone. In clinical trials, treatment with ActD (84–86) or Echi (87–89) alone had a low anti-cancer effect in various cancers. As a result, different combination therapies have been developed, for example, Act or Echi plus methyl-CCNU (90), dacarbazine (DTIC), vincristine (91), trimetrexate (92) to treat advanced cancers. However, these combination therapies have not explored possible synergies. Our study takes advantage of molecular structure insights into drug development. In vitro and in vivo data uncover that combination treatment is a promising approach for the treatment of cancers such as MMR-deficient cancer. We anticipate that synergistic therapy can minimise off-target and generalized cytotoxicity and maximise anti-cancer efficacy. It is notable that the doses of ActD and Echi used here in the in vivo therapeutic experiments are both 10-fold lower than those previously reported when used as single agents (93,94). These reports also found that significant toxicity accompanied anti-tumour activity, with low numbers of survivors after treatment, in marked contrast to the present findings for the synergistic combination. Therefore, the ActD + Echi combination, with its 10-fold lower dose, may have clinical potential as an effective and reduced toxicity chemotherapeutic regimen.

In summary, the current study has elucidated the structural basis for the simultaneous recognition of T:T mismatched DNA by two different DNA-binding intercalative anticancer agents. We have also shown that this combination acts synergistically on a xenograft model for colorectal cancer and has potential application in the treatment of other MMR-deficient cancers. It is also hoped that the structural insights from this study will help guide the development of future generations of combined DNA-targeted chemotherapy.

DATA AVAILABILITY

The atomic coordinates and structure factors for the reported crystal structures have been deposited in the Protein Data Bank under the accession numbers 7DQ0 for the T:T mismatched complex and 7DQ8 for T:C mismatched complex structures.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Ministry of Science and Technology, Taiwan, R.O.C. [109–2628-M-005–001-MY4, 109–2311-B-005–007-MY3 to M.-H.H.]; National Chung Hsing University and Chung Shan Medical University [NCHU-CSMU 10904, NCHU-CSMU 10803 to M.-H.H.]. The open access publication charge for this paper has been waived by Oxford University Press – NAR Editorial Board members are entitled to one free paper per year in recognition of their work on behalf of the journal.

Conflict of interest statement. None declared.