Region-specific DNA damage by AT-specific DNA-reactive drugs is predicted by drug binding specificity

Bizelesin and adozelesin are DNA-reactive antitumor drugs that alkylate adenines at the 3' ends of their preferred binding sites [5'T(A/T)(4)A3'and 5'(A/T)(3)(-4)A3', respectively]. We used these drugs to examine the determinants for region-specific damage of human genomic DNA. The distribution of bizelesin binding motifs in several regions analyzed "in silico" correlated well with the experimentally determined lesions in these regions assessed by quantitative polymerase chain reaction (QPCR) stop assay. In contrast to the typically low motif density, clusters of potential bizelesin binding sites were found in the matrix-associated regions (MAR domains) of the c-myc and apolipoprotein B (apoB) genes. Accordingly, lesions induced by bizelesin in these domains (2.13 and 7.06 lesions kbp(-1) microM(-1), respectively) markedly exceeded lesions in bulk DNA (0.87 lesions kbp(-1) microM(-1)) or in regions with typically low motif density (e.g., 0.75 and 0.87 lesions kbp(-1) microM(-1) in a beta-globin gene and c-myc origin of replication regions, respectively). Consistent with the more frequent, less localized adozelesin motif, actual lesions induced by adozelesin exceeded by severalfold lesions by bizelesin in four selected regions (within the c-myc and HPRT loci). Whereas adozelesin is likely to affect similar regions as bizelesin, adozelesin's more promiscuous binding probably compromises its relative specificity for such targets. In contrast, findings for bizelesin provide for the first time a proof of principle that a small molecular weight drug can preferentially damage specific regions in cellular DNA. Targeting of critical repetitive sequences, such as AT-rich MAR domains, which allow for clustering of drug binding motif, can be the paradigm for region specificity of small molecular weight agents.     

Induction of AT-specific DNA-interstrand crosslinks by bizelesin in genomic and simian virus 40 DNA

Bizelesin is a bifunctional AT-specific DNA alkylating drug. Our study characterized the ability of bizelesin to induce interstrand crosslinks, a potential lethal lesion. In genomic DNA of BSC-1 cells, bizelesin formed from approx. 0.3 to 6.03+/-0.85 interstrand crosslinks per 106 base pairs, at 5-100 nM drug concentration, respectively, comparable to the number of total adducts previously determined in the same system (J.M. Woynarowski, M.M. McHugh, L.S. Gawron, T.A. Beerman, Biochemistry 34 (1995) 13042-13050). Bizelesin did not induce DNA-protein crosslinks or strand breaks. A model defined target, intracellular simian virus 40 (SV40) DNA, was employed to map at the nucleotide level sites of bizelesin adducts, including potential interstrand crosslinks. Preferential adduct formation was observed at AT tracts which are abundant in the SV40 matrix associated region and the origin of replication. Many sites, including each occurrence of 5'-T(A/T)4A-3', co-mapped on both DNA strands suggesting interstrand crosslinks, although monoadducts were also formed. Bizelesin adducts in naked SV40 DNA were found at similar sites. The localization of bizelesin-induced crosslinks in AT-rich tracts of replication-related regions is consistent with the potent anti-replicative properties of bizelesin. Given the apparent lack of other types of lesions in genomic DNA, interstrand crosslinks localized in AT-rich tracts, and to some extent perhaps also monoadducts, are likely to be lethal effects of bizelesin.     

AT-rich islands in genomic DNA as a novel target for AT-specific DNA-reactive antitumor drugs

Interstrand cross-links at T(A/T)4A sites in cellular DNA are associated with hypercytotoxicity of an anticancer drug, bizelesin. Here we evaluated whether these lethal effects reflect targeting critical genomic regions. An in silico analysis of human sequences showed that T(A/T)4A motifs are on average scarce and scattered. However, significantly higher local motif densities were identified in distinct minisatellite regions (200-1000 base pairs of approximately 85-100% AT), herein referred to as "AT islands." Experimentally detected bizelesin lesions agree with these in silico predictions. Actual bizelesin adducts clustered within the model AT island naked DNA, whereas motif-poor sequences were only sparsely adducted. In cancer cells, bizelesin produced high levels of lesions (approximately 4.7-7.1 lesions/kilobase pair/microM drug) in several prominent AT islands, compared with markedly lower lesion levels in several motif-poor loci and in bulk cellular DNA (approximately 0.8-1.3 and approximately 0.9 lesions/kilobase pair/microM drug, respectively). The identified AT islands exhibit sequence attributes of matrix attachment regions (MARs), domains that organize DNA loops on the nuclear matrix. The computed "MAR potential" and propensity for supercoiling-induced duplex destabilization (both predictive of strong MARs) correlate with the total number of bizelesin binding sites. Hence, MAR-like AT-rich non-coding domains can be regarded as a novel class of critical targets for anticancer drugs.     

The genome factor in region-specific DNA damage: the DNA-reactive drug U-78779 prefers mixed A/T-G/C sequences at the nucleotide level but is region-specific for long pure AT islands at the genomic level

Bizelesin is the first anticancer drug capable of damaging specific regions of the genome with clusters of its binding sites T(A/T)(4)A. This study characterized the sequence- and region-specificity of a bizelesin analogue, U-78779, designed to interact with mixed A/T-G/C motifs. At the nucleotide level, U-78779 was found to prefer runs of A/Ts interspersed with 1 or 2 G/C pairs, although 25% of the identified sites corresponded to pure AT motifs similar to bizelesin sites. The in silico computational analysis showed that the preferred mixed A/T-G/C motifs distribute uniformly at the genomic level. In contrast, the secondary, pure AT motifs (A/T)(6)A were found densely clustered in the same long islands of AT-rich DNA that bizelesin targets. Mapping the sites and quantitating the frequencies of U-78779 adducts in model AT island and non-AT island naked DNAs demonstrated that clusters of pure AT motifs outcompete isolated mixed A/T-G/C sites in attracting drug binding. Regional preference of U-78779 for AT island domains was verified also in DNA from drug-treated cells. Thus, while the primary sequence preference gives rise to non-region-specific scattered lesions, the clustering of the minor pure AT binding motifs seems to determine region-specificity of U-78779 in the human genome. The closely correlated cytotoxic activities of U-78779 and bizelesin in several cell lines further imply that both drugs may share common cellular targets. This study underscores the significance of the genome factor in a drug's potential for region-specific DNA damage, by showing that it can take precedence over drug binding preferences at the nucleotide level.     

Targeting critical regions in genomic DNA with AT-specific anticancer drugs

Cellular DNA is not a uniform target for DNA-reactive drugs. At the nucleotide level, drugs recognize and bind short motifs of a few base pairs. The location of drug adducts at the genomic level depends on how these short motifs are distributed in larger domains. This aspect, referred to as region specificity, may be critical for the biological outcome of drug action. Recent studies demonstrated that certain minor groove binding (MGB) drugs, such as bizelesin, produce region-specific lesions in cellular DNA. Bizelesin binds mainly T(A/T)(4)A sites, which are on average scarce, but occasionally cluster in distinct minisatellite regions (200-1000 bp of approximately 85-100% AT), herein referred to as AT islands. Bizelesin-targeted AT islands are likely to function as strong matrix attachment regions (MARs), domains that organize DNA loops on the nuclear matrix. Distortion of MAR-like AT islands may be a basis for the observed inhibition of new replicon initiation and the extreme lethality of bizelesin adducts (<10 adducts/cell for cell growth inhibition). Hence, long AT-islands represent a novel class of critical targets for anticancer drugs. The AT island paradigm illustrates the potential of the concept of regional targeting as an essential component of the rational design of new sequence-specific DNA-reactive drugs.     

Combinatorial identification of a novel consensus sequence for the covalent DNA-binding polyamide tallimustine

Many agents successfully used in cancer chemotherapy either directly or indirectly covalently modify DNA. Examples include cisplatin, which forms a covalent adduct with guanines, and doxorubicin, which traps a cleavage intermediate between topoisomerase II and torsionally strained DNA. In most cases, the efficacy of these drugs depends on the efficiency and specificity of their DNA binding, as well as the discrimination between normal and neoplastic cells in their handling of the drug-DNA adducts. While much is known about the chemistry of drug-DNA adducts, little is known regarding the overall specificity of their formation, especially in the context of a whole human genome, where potentially billions of binding sites are possible. We used the combinatorial selection method restriction endonuclease protection, selection, and amplification (REPSA) to determine the DNA-binding specificity of the semisynthetic covalent DNA-binding polyamide tallimustine, which contains a benzoic acid nitrogen mustard appended to the minor groove DNA-binding natural product distamycin A. After investigating over 134 million possible sequences, we found that the highest affinity tallimustine binding sites contained one of two consensus sequences, either the expected distamycin hexamer binding sites followed by a CG base pair (e.g., 5'-TTTTTTC-3' and 5'-AAATTTC-3') or the unexpected sequence 5'-TAGAAC-3'. Curiously, we found that tallimustine preferentially alkylated the N7 position of guanines located on the periphery of these consensus sequences. These findings suggested a cooperative binding model for tallimustine in which one molecule noncovalently resides in the DNA minor groove and locally perturbs the DNA structure, thereby facilitating alkylation by a second tallimustine of an exposed guanine on another side of the DNA.     

DNA interstrand cross-links of the novel antitumor trinuclear platinum complex BBR3464. Conformation,recognition by high mobility group domain proteins,and nucleotide excision repair

The novel phase II antitumor polynuclear platinum drug BBR3464 ([(trans-PtCl(NH(3))(2))(2)(mu-trans-Pt(NH(3))(2)(NH(2)(CH(2))(6)NH(2))(2))](NO(3))(4)) forms intra- and interstrand cross-links (CLs) on DNA (which is the pharmacological target of platinum drugs). We examined first in our recent work how various intrastrand CLs of BBR3464 affect the conformation of DNA and its recognition by cellular components (Zehnulova, J., Kasparkova, J., Farrell, N., and Brabec, V. (2001) J. Biol. Chem. 276, 22191-22199). In the present work, we have extended the studies on the DNA interstrand CLs of this drug. The results have revealed that the interstrand CLs are preferentially formed between guanine residues separated by 2 base pairs in both the 3' --> 3' and 5' --> 5' directions. The major 1,4-interstrand CLs distort DNA, inducing a directional bending of the helix axis and local unwinding of the duplex. Although such distortions represent a potential structural motif for recognition by high mobility group proteins, these proteins do not recognize 1,4-interstrand CLs of BBR3464. On the other hand, in contrast to intrastrand adducts of BBR3464, 1,4-interstrand CLs are not removed from DNA by nucleotide excision repair. It has been suggested that interstrand CLs of BBR3464 could persist considerably longer in cells compared with intrastrand adducts, which would potentiate the toxicity of the interstrand lesions to tumors sensitive to this polynuclear drug.     

Interstrand cross-links and sequence specificity in the reaction of cis-dichloro(ethylenediamine)platinum(II) with DNA

Characterization of the adducts produced in DNA by the cancer chemotherapeutic drug cis-diamminedichloroplatinum(II) and a radiolabeled analogue, [3H]-cis-dichloro(ethylenediamine)platinum(II) ([3H]-cis-DEP) was recently reported [Eastman, A. (1983) Biochemistry 22, 3927]. Both drugs reacted at identical sites in DNA, most of which produced intrastrand cross-links. DNA-interstrand cross-links, which represent less than 1% of total platination, have now been characterized. DNA containing interstrand cross-links was enriched for on the basis of its renaturability after boiling. This DNA was digested to deoxyribonucleosides, and the adducts were separated by high-pressure liquid chromatography. A cross-link between two deoxyguanosines was observed to be the most prominent adduct. It is proposed that the major sequence in which this cross-link occurs is 5'-CG-3'. DNA that was incubated with [3H]-cis-DEP for 1 h showed low levels of interstrand cross-links. After removal of unreacted drug, their frequency increased significantly over 6 h with a maximum occurring at about 12 h. A similar phenomenon was seen in the case of intrastrand cross-links that contained adenine, in particular when the cross-link was between the terminal bases in an ANG trinucleotide sequence (N is any nucleotide). The primary site of reaction is at guanine, with a slow subsequent cross-link to the adenine. A model is presented that is consistent with the observation that adenine is always at the 5' terminus of these adducts. The proportion of adducts at ANG sequences also increased at elevated temperatures. This is discussed with regard to potential significance during hyperthermia treatment of patients.(ABSTRACT TRUNCATED AT 250 WORDS)     

Repair of DNA-protein cross-links in mammalian cells

DNA-protein cross-links are generated by both endogenous and exogenous DNA damaging agents, as intermediates during normal DNA metabolism, and during abortive base excision repair. Cross-links are relatively common lesions that are lethal when they block progression of DNA polymerases. DNA-protein cross-links may be broadly categorized into four groups by the DNA and protein chemistries near the cross-link and by the source of the cross-link: DNA-protein cross-links may be found (1) in nicked DNA at the 3' end of one strand (topo I), (2) in nicked DNA at the 5' end of one strand (pol beta), (3) at the 5' ends of both strands adjacent to nicks in close proximity (topo II; Spo 11), and (4) in one strand of duplex DNA (UV irradiation; bifunctional carcinogens and chemotherapeutic agents). Repair mechanisms are reasonably well-defined for groups 1 and 3, and suggested for groups 2 and 4. Our work is focused on the recognition and removal of DNA-protein cross-links in duplex DNA (group 4).     

Repair synthesis by human cell extracts in cisplatin-damaged DNA is preferentially determined by minor adducts.

During reaction of cis-diamminedichloroplatinum(II) (cis-DDP) with DNA, a number of adducts are formed which may be discriminated by the excision-repair system. An in vitro excision-repair assay with human cell-free extracts has been used to assess the relative repair extent of monofunctional adducts, intrastrand and interstrand cross-links of cis-DDP on plasmid DNA. Preferential removal of cis-DDP 1,2-intrastrand diadducts occurred in the presence of cyanide ions. In conditions where cyanide treatment removed 85% of total platinum adducts while approximately 70% of interstrand cross-links remained in plasmid DNA, no significant variation in repair synthesis by human cell extracts was observed. Then, we constructed three types of plasmid DNA substrates containing mainly either monoadducts, 1,2-intrastrand cross-links or interstrand cross-links lesions. The three plasmid species were modified in order to obtain the same extent of total platinum DNA adducts per plasmid. No DNA repair synthesis was detected with monofunctional adducts during incubation with human whole cell extracts. However, a two-fold increase in repair synthesis was found when the proportion of interstrand cross-links in plasmid DNA was increased by 2-3 fold. These findings suggest that (i) cis-DDP 1,2-intrastrand diadducts are poorly repaired by human cell extracts in vitro, (ii) among other minor lesions potentially cyanide-resistant, cis-DDP interstrand cross-links represent a major lesion contributing to the repair synthesis signal in the in vitro assay. These results could account for the drug efficiency in vivo.     

UvrABC incision of N-methylmitomycin A-DNA monoadducts and cross-links

The Escherichia coli UvrABC endonuclease is a multisubunit enzyme that initiates the repair of a wide variety of DNA lesions in vivo by making dual incisions on a damaged strand at the eighth or ninth phosphodiester bond 5' and the fourth or fifth phosphodiester bond 3' to the modified base. It has been hypothesized that UvrABC is able to recognize a broad spectrum of lesions because it does not recognize the lesion per se but rather gross helical distortions that the lesion induces in the DNA. Several lesions have recently been studied which are thermal stabilizing and are not believed to distort the DNA grossly, including the CC-1065-N-3-adenine and anthramycin-N-2-guanine adducts. We have studied the activity of UvrABC in vitro on another thermal stabilizing and nondistortive adduct, N-methylmitomycin A (NMA), a bifunctional DNA-alkylating agent that reacts with guanine on the side facing the minor groove, yielding either monoadducts or interstrand cross-links. NMA adducts increase the thermal stability of DNA, and theoretical calculations indicate that NMA adducts do not grossly distort the DNA helix. Our results show that UvrABC makes incisions at the eighth phosphodiester bond 5' and the fifth phosphodiester bond 3' to an NMA monoadduct, consistent with the incision pattern observed for the majority of other lesions that are also recognized by UvrABC. DNA containing a site-specific NMA cross-link was also recognized and incised by UvrABC. The rate of incision of NMA cross-linked DNA was about 200-fold higher in supercoiled molecules than in relaxed molecules, whereas the rate of incision of DNA containing NMA monoadducts was stimulated approximately 2-fold by supercoiling. The signal for UvrABC recognition and incision of damaged DNA is discussed in relation to the ability of UvrABC to incise NMA adducts as well as other nondistortive lesions.     

Matrix attachment region (MAR) properties and abnormal expansion of AT island minisatellites in FRA16B fragile sites in leukemic CEM cells

AT-rich minisatellites (AT islands) are sites of genomic instability in cancer cells and targets for extremely lethal AT-specific drugs, such as bizelesin. Here we investigated the AT islands in the FRA16B fragile site region for their possible roles in the organization of DNA on the nuclear matrix. The FRA16B AT island nominally spans ~3 kb of mostly >90% A/T DNA. In silico analysis indicates that this domain exhibits characteristics of nuclear matrix attachment regions (MARs): an exceptionally intense computed ‘MAR potential’ and profound duplex destabilization and flexibility. FRA16B repeats specifically bind to isolated nuclear matrices, which indicates their in vitro MAR function. This binding is several-fold greater than that of a known MAR in the c-myc gene. AT islands in fragile sites FRA16B and FRA16D are significantly more abundant in CEM cells that are hypersensitive to bizelesin compared to normal WI-38 cells. FRA16B overabundance in CEM is due to an ~10-fold expansion of FRA16B repeats. The expanded FRA16B minisatellites in CEM cells preferentially localize to the nuclear matrix-associated DNA indicating their in vivo MAR function. The unexpanded repeats in WI-38 cells localize to the loop DNA. The c-myc MAR is also matrix-associated in CEM cells while localizing to loop DNA in WI-38 cells. These results are the first to demonstrate that AT islands in fragile sites can function as MARs both in vitro and in vivo. The ability of FRA16B-mediated MAR sites to rearrange depending on the repeat expansion status could be relevant to both genomic instability of cancer cells and their sensitivity to AT-island targeting drugs.     

Mechanism of the formation of DNA-protein cross-links by antitumor cisplatin

DNA–protein cross-links are formed by various DNA-damaging agents including antitumor platinum drugs. The natures of these ternary DNA–Pt–protein complexes (DPCLs) can be inferred, yet much remains to be learned about their structures and mechanisms of formation. We investigated the origin of these DPCLs and their cellular processing on molecular level using gel electrophoresis shift assay. We show that in cell-free media cisplatin [cis-diamminedichloridoplatinum(II)] forms DPCLs more effectively than ineffective transplatin [trans-diamminedichloridoplatinum(II)]. Mechanisms of transformation of individual types of plain DNA adducts of the platinum complexes into the DPCLs in the presence of several DNA-binding proteins have been also investigated. The DPCLs are formed by the transformation of DNA monofunctional and intrastrand cross-links of cisplatin. In contrast, interstrand cross-links of cisplatin and monofunctional adducts of transplatin are stable in presence of the proteins. The DPCLs formed by cisplatin inhibit DNA polymerization or removal of these ternary lesions from DNA by nucleotide excision repair system more effectively than plain DNA intrastrand or monofunctional adducts. Thus, the bulky DNA–protein cross-links formed by cisplatin represent a more distinct and persisting structural motif recognized by the components of downstream cellular systems processing DNA damage considerably differently than the plain DNA adducts of this metallodrug.     

Mechanism of Formation of Novel Covalent Drug·DNA Interstrand Cross-Links and Monoadducts by Enediyne Antitumor Antibiotics

The potent enediyne antitumor antibiotic C1027 has been previously reported to induce novel DNA interstrand cross-links and drug monoadducts under anaerobic conditions [Xu et al. (1997) J. Am. Chem. Soc. 119, 1133-1134]. In the present study, we explored the mechanism of formation of these anaerobic DNA lesions. We found that, similar to the aerobic reaction, the diradical species of the activated drug initiates anaerobic DNA damage by abstracting hydrogen atoms from the C4', C1', and C5' positions of the A1, A2, and A3 nucleotides, respectively, in the most preferred 5'GTTA1T/5'ATA2A3C binding sequence. It is proposed that the newly generated deoxyribosyl radicals, which cannot undergo oxidation, likely add back onto the nearby unsaturated ring system of the postactivated enediyne core, inducing the formation of interstrand cross-links, connecting either A1 to A2 or A1 to A3, or drug monoadducts mainly on A2 or A3. Comparative studies with other enediynes, such as neocarzinostatin and calicheamicin gamma1I under similar reaction conditions indicate that the anaerobic reaction process is a kinetically competitive one, depending on the proximity of the drug unsaturated ring system or dioxygen to the sugar radicals and their quenching by other hydrogen sources such as solvent or thiols. It was found that C1027 mainly generates interstrand cross-links, whereas most of the anaerobic lesions produced by neocarzinostatin are drug monoadducts. Calicheamicin gamma1I was found to be less efficient in producing both lesions. The anaerobic DNA lesions induced by enediyne antitumor antibiotics may have important implications for their potent cytotoxicity in the central regions of large tumors, where relative anaerobic conditions prevail.     

Interstrand cross-linking by adriamycin in nuclear and mitochondrial DNA of MCF-7 cells

Activation of Adriamycin by formaldehyde leads to the formation of drug–DNA adducts in vitro and these adducts stabilise the DNA to such a degree that they function as virtual interstrand cross-links. The formation of these virtual interstrand cross-links by Adriamycin was investigated in MCF-7 cells using a gene-specific interstrand cross-linking assay. Cross-linking was measured in both the nuclear-encoded DHFR gene and in mitochondrial DNA (mtDNA). Cross-link formation increased linearly with Adriamycin concentration following a 4 h exposure to the drug. The rate of formation of Adriamycin cross-links in each of the genomes was similar, reaching maximal levels of 0.55 and 0.4 cross-links/10 kb in the DHFR gene and mtDNA respectively, following exposure to 20 µM Adriamycin for 8 h. The interstrand cross-link was short lived in both DNA compartments, with a half-life of 4.5 and 3.3 h in the DHFR gene and mtDNA respectively. The kinetics of total Adriamycin adduct formation, detected using [¹⁴C]Adriamycin, was similar to that of cross-link formation. Maximal adduct levels (30 lesions/10 kb) were observed following incubation at 20 µM drug for 8 h. The formation of such high levels of adducts and cross-links could therefore be expected to contribute to the mechanism of action of Adriamycin.     

Unique cross-link and monoadduct repair characteristics of a xeroderma pigmentosum revertant cell line

Monoadducts and cross-links formed in DNA of human cells by a psoralen derivative, 4'-hydroxy-methyl-4,5',8-trimethylpsoralen (HMT), have been measured by a new, simple method, based on S1 nuclease digestion of 3H-labeled adducts in DNA, that provides rapid information on the repair of both classes of lesions. Normal human fibroblasts and cells from patients with dyskeratosis congenita and xeroderma pigmentosum (XP) group C were capable of removing both monoadducts and cross-links, whereas XP groups A and D failed to remove either. An XP revertant, isolated from a group A cell line on the basis of an acquired mutagen-induced resistance to ultraviolet light, has the unique property of being capable of removing cross-links but not monoadducts. Consistent with this property, the XP revertant was found to be resistant to cell killing by the cross-linking psoralen derivative, HMT, but as sensitive as its parental cell line to a monofunctional psoralen derivative, 5-methylisopsoralen.