Sequence preferences of DNA interstrand cross-linking agents: dG-to-dG cross-linking at 5'-CG by structurally simplified analogues of mitomycin C

The nucleotide sequence preferences of the DNA interstrand cross-linking agents dehydroretronecine diacetate (DHRA), 2,3-bis(acetoxymethyl)-1-methylpyrrole (BAMP), dehydromonocrotaline, and dehydroretrorsine were studied by using synthetic DNA duplex fragments and polyacrylamide gel electrophoresis (PAGE). These agents have structural features in common with the reductively activated aziridinomitosene of mitomycin C (MC). Like MC, they preferentially cross-linked DNA duplexes containing the duplex sequence 5'-CG. For DHRA and BAMP interstrand cross-linked DNA duplexes, PAGE analysis of iron(II)-EDTA fragmentation reactions revealed the interstrand cross-links to be deoxyguanosine to deoxyguanosine (dG-to-dG), again analogous to DNA cross-links caused by MC. Unlike MC, DHRA could be shown to dG-to-dG cross-link a 5'-GC sequence. Furthermore, the impact of flanking sequence on the efficiency of interstrand cross-linking at 5'-CG was reduced for BAMP, with 5'-TCGA and 5'-GCGC being equally efficiently cross-linked. Possible origins of the 5'-CG sequence recognition common to all of the agents are discussed. A model is presented in which the transition state for the conversion of monoadducts to cross-links more closely resembles ground-state DNA at 5'-CG sequences.     

Sequence-specific DNA interstrand cross-linking by an aziridinomitosene in the absence of exogenous reductant

The aziridinomitosene derivative (1S,2S)-6-desmethyl(methylaziridino)mitosene (4) was shown to alkylate plasmid DNA at pH 7.4 in the absence of a reducing agent [Vedejs, E., Naidu, B. N., Klapars, A., Warner, D. L., Li, V. -s., Na, Y., and Kohn, H. (2003) J. Am. Chem. Soc. 125, 15796-15806], an activity not found in the parent mitomycins. We sought to evaluate aziridinomitosene 4 for the presence of DNA interstrand cross-linking activity using nonreductive reaction conditions. Radiolabeled DNA treated with 4 was analyzed by denaturing polyacrylamide gel electrophoresis (DPAGE), a technique that readily separates the less mobile cross-linked ds DNA from the more mobile ss DNA products. Nonreduced 4 produced an interstrand cross-link (ICL) in duplex DNA containing 5'-d(CG) sites, and the yield of ICL was comparable to that obtained from reduced MC under similar conditions. A ds DNA having the central tetranucleotide 5'-d(ACGT) provided the greatest ICL yield from both nonreduced 4 and reduced MC. Substitution of 5'-d(CG) with the inverted sequence 5'-d(GC) completely abolished interstrand cross-linking by 4, revealing 5'-d(CG) as its specific site of ICL formation. Replacement of dG at 5'-d(CG) with 2'-deoxyinosine (dI), which lacks the exocyclic C2 amino group present in dG, also prevented DNA ICL formation by 4, revealing an essential role for the dG C2 amino group in the interstrand cross-linking reaction between 4 and duplex DNA. This report directly demonstrates the presence of bifunctional alkylating activity in a nonreduced aziridinomitosene and clearly shows that unreduced 4 alkylates residues in the minor groove of ds DNA, cross-linking with the same 5'-d(CG) sequence specificity displayed by reduced MC.     

DNA cross-linking by intermediates in the mitomycin activation cascade

We have assayed the cross-linking of oligonucleotides containing repeated mitomycin-reactive CpG sites in order to assess the factors that enhance activation of the carbamoyl function at C10, yielding efficient mitomycin cross-linking. Drugs studied include mitomycin C (MC), N-methylmitomycin A (NMA), and the aziridinomitosene of NMA (MS). Drugs were reduced both by catalytic hydrogenation and by diothionite. We find that cross-linking by fully reduced NMA can be increased severalfold by addition of either excess dithionite reductant or the oxidant FeCl3. Enhancement by FeCl3 is not seen with MC or MS, but excess dithionite increases cross-linking by all three compounds. We explain the action of Fe3+ by postulating production of the semiquinone of the monoadduct of mitomycin reacted at the C1-position; according to this mechanism, departure of the carbamate from C10 is more efficient for the semiquinone than for the hydroquinone. However, our results imply that the hydroquinone can also function as a cross-linking agent. Excess dithionite, beyond that required for stoichiometric reduction, activates the carbamate 2-3-fold for cross-linking. We find that the fully reduced leucoaziridinomitosene is highly unstable in solution, yet it produces efficient cross-liking. Hence, this compound is highly reactive in DNA alkylation and a good candidate for the role of primary alkylating agent.     

C5 cytosine methylation at CpG sites enhances sequence selectivity of mitomycin C-DNA bonding

We have established that UvrABC nuclease is equally efficient in cutting mitomycin C (MC)-DNA monoadducts formed at different sequences and that the degree of UvrABC cutting represents the extent of drug-DNA bonding. Using this method we determined the effect of C5 cytosine methylation on the DNA monoalkylation by MC and the related analogues N-methyl-7-methoxyaziridinomitosene (MS-NMA) and 10-decarbamoylmitomycin C (DC-MC). We have found that C5 cytosine methylation at CpG sites greatly enhances MC and MS-NMA DNA adduct formation at those sites while reducing adduct formation at non-CpG sequences. In contrast, although DC-MC DNA bonding at CpG sites is greatly enhanced by CpG methylation, its bonding at non-CpG sequences is not appreciably affected. These cumulative results suggest that C5 cytosine methylation at CpG sites enhances sequence selectivity of drug-DNA bonding. We propose that the methylation pattern and status (hypo- or hypermethylation) of genomic DNA may determine the cells' susceptibility to MC and its analogues, and these effects may, in turn, play a crucial role in the antitumor activities of the drugs.     

Recognition between mitomycin C and specific DNA sequences for cross-link formation

An extensive series of oligodeoxyribonucleotides was reacted with reductively activated mitomycin C (MC), and the resulting cross-linked drug-oligonucleotide complexes were isolated by reverse-phase HPLC and characterized by nucleoside and MC-nucleoside adduct analysis. HPLC also served for assay of the yield of cross-linked oligonucleotides. AT-rich duplex oligonucleotides, containing a single central CG.CG, gave high yields of cross-links between the two guanines while those having GC.GC, instead, gave none. In another series, the central sequences CGC.GCG and CGC.ICG both yielded 50% cross-link while CGC.GCI was completely resistant. Cross-linking was conducted also in two steps: Oligonucleotides substituted monofunctionally by MC at guanine at either a CG or GC sequence were annealed with their complementary strands followed by reductive reactivation of the bound MC to form a cross-link. The CG oligomers were cross-linked quantitatively while the GC ones were again resistant. These results show unambiguously that the MC cross-link is absolutely specific to the CG.CG duplex sequence, confirming our previous finding [Chawla, A.K., Lipman, R., & Tomasz, M. (1987) in Structure and Expression, Volume 2: DNA and Its Drug Complexes (Sarma, R.H., & Sarma, M.H., Eds.) Adenine Press, Guilderland, NY]. Evidence is presented that this specificity is due to the specific orientation of the monofunctionally attached MC in the minor groove. Superimposed on the CG.CG requirement, a four-base-pair sequence preference was observed at PuCGPyr.PuCGPyr sequences. This suggests that the guanine N2 atom of GpPyr is more reactive toward the drug than that of GpPu, due to the favorable effect of the negative dipole of the O2 of the Pyr on the reaction; in accordance, GpT was more reactive than GpC.(ABSTRACT TRUNCATED AT 250 WORDS)     

The effect of C(5) cytosine methylation at CpG sequences on mitomycin-DNA bonding profiles

Recent studies have documented that cytosine C(5) methylation of CpG sequences enhances mitomycin C (1) adduction. The reports differ on the extent and uniformity of 1 modification at the nucleotide level. We have determined the bonding profiles for mitomycin monoalkylation in two DNA restriction fragments where the CpG sequences were methylated. Three mitomycin substrates were used and two different enzymatic assays employed to monitor the extent of drug modification at the individual base sites. Drug DNA modification was accomplished with I and 10-decarbamoylmitomycin C (2) under reductive (Na2S2O4) condilions and with N-methyl-7-methoxyaziridinomitosene (3) under nonreductive conditions. The UvrABC incision assay permitted us to quantitate the sites of drug adduction, and the lambda-exonuclease stop assay provided a qualitative estimation of drug-DNA modification consistent with the UvrABC data. We learned that C(5) cytosine methylation (m5C) enhanced the extent of overall DNA modification. Using the UvrABC endonuclease assay, we found that modification by 1 increased 2.0 and 7.4 times for the two DNA restriction fragments. Analysis of the modification sites at the nucleotide sequence level revealed that guanine (G) was the only base modified and that the overall increased level of DNA adduction was due to enhanced modification of select m5CpG* (G* = mitomycin (mitosene) adduction sites) loci compared with CpG* sites: the largest differences reached two orders of magnitude. Significantly, not all CpG* sites underwent increased drug adduction upon C(5) cytosine methylation. The effect of C(5) cytosine methylation on the drug adduction profiles was less pronounced for G* sites located within dinucleotide sequences other than CpG*. We observed that DNA methylation often led to slightly diminished adduction levels at these sites. The different m5CpG* adduction patterns provided distinctive sequence-selective bonding profiles for 1-3. We have attributed the large differences in guanine reactivity to DNA structural factors created, in part, by C(5) cytosine methylation. The significance of these findings in cancer chemotherapy is briefly discussed.     

Duplex oligodeoxyribonucleotides cross-linked by mitomycin C at a single site: synthesis, properties, and cross-link reversibility

Oligodeoxyribonucleotides cross-linked by reductively activated mitomycin C (MC) were prepared and purified for the first time. The cross-linked products were structurally characterized by nucleoside and MC-nucleoside adduct analysis. Optimal conditions were established for the cross-linking reaction, resulting in high yields, typically in the 20-50% range. Nuclease digests of the cross-linked oligonucleotides yielded the same bifunctional MC-deoxyguanosine adduct as that previously isolated from DNA exposed to MC in vitro and in vivo [Tomasz et al. (1987) Science 235, 1204]. The cross-linked oligonucleotides displayed broad thermal melting profiles, greatly increased Tm, and complex circular dichroism spectra. Phosphodiester linkages at the cross-link were resistant to spleen exonuclease, nuclease P1, and TaqI and ClaI restriction endonucleases; snake venom diesterase action was uninhibited. The cross-links are stable to heat at neutral pH but are removed by treatment in hot piperidine or by the reducing agents Na2S2O4 and dithiothreitol. Mechanisms are proposed for these reactions. These studies define optimal methods for introducing mitomycin cross-links into DNA fragments at a specific site, providing a versatile tool to study the effects of the MC cross-links on DNA structure and function.     

Mutations of different molecular origins exhibit contrasting patterns of regional substitution rate variation

Transitions at CpG dinucleotides, referred to as “CpG substitutions”, are a major mutational input into vertebrate genomes and a leading cause of human genetic disease. The prevalence of CpG substitutions is due to their mutational origin, which is dependent on DNA methylation. In comparison, other single nucleotide substitutions (for example those occurring at GpC dinucleotides) mainly arise from errors during DNA replication. Here we analyzed high quality BAC-based data from human, chimpanzee, and baboon to investigate regional variation of CpG substitution rates.

We show that CpG substitutions occur approximately 15 times more frequently than other single nucleotide substitutions in primate genomes, and that they exhibit substantial regional variation. Patterns of CpG rate variation are consistent with differences in methylation level and susceptibility to subsequent deamination. In particular, we propose a “distance-decaying” hypothesis, positing that due to the molecular mechanism of a CpG substitution, rates are correlated with the stability of double-stranded DNA surrounding each CpG dinucleotide, and the effect of local DNA stability may decrease with distance from the CpG dinucleotide.

Consistent with our “distance-decaying” hypothesis, rates of CpG substitution are strongly (negatively) correlated with regional G+C content. The influence of G+C content decays as the distance from the target CpG site increases. We estimate that the influence of local G+C content extends up to 1,500~2,000 bps centered on each CpG site. We also show that the distance-decaying relationship persisted when we controlled for the effect of long-range homogeneity of nucleotide composition. GpC sites, in contrast, do not exhibit such “distance-decaying” relationship. Our results highlight an example of the distinctive properties of methylation-dependent substitutions versus substitutions mostly arising from errors during DNA replication. Furthermore, the negative relationship between G+C content and CpG rates may provide an explanation for the observation that GC-rich SINEs show lower CpG rates than other repetitive elements.

     

Use of oligonucleotides to define the site of interstrand cross-links induced by Adriamycin.

It has been known for several years that Adriamycin forms adducts and interstrand cross-links when reacted for long periods of time with bacterial and mammalian DNA in vitro, with the cross-link being restricted to 2 bp elements containing GpC sequences. The self-complementary 20mer deoxyoligonucleotide TA4T4GCA4T4A has been used in this study as a model of the apparent G-G cross-linking site at GpC sequences. The rate of formation of cross-links, as well as the dependence on both Adriamycin and Fe(III) concentration, were similar with this oligonucleotide as compared with calf thymus DNA. The cross-linking was demonstrated on both denaturing and non-denaturing sequencing gels. The half-life of the G-G cross-link was 40 h, consistent with that implied with high molecular weight, heterogeneous sequence DNA. Exonuclease III digests of adducts formed with 20mer deoxyoligonucleotides containing single, central G-G, G-I and I-I potential cross-links revealed that a guanine residue is required at both ends of the cross-link. No cross-linking was observed with a similar oligonucleotide containing only a single central (G.C) bp.     

Collagen type IX: Evidence for covalent linkages to type II collagen in cartilage

A major site of pyridinoline cross-linking in bovine type IX collagen was traced to a tryptic peptide derived from one of the molecule's HMW chains. This peptide gave two amino acid sequences (in 2/1 ratio) consistent with it being a three-chained structure. The major sequence matched exactly that of the C-telopeptide of type II collagen from the same tissue. A second HMW chain that contained pyridinoline cross-links also gave two amino-terminal sequences, one from its own amino terminus, the other matching exactly the N-telopeptide cross-linking sequence of type II collagen. We conclude that type IX collagen molecules are covalently cross-linked in cartilage to molecules of type II collagen, probably at fibril surfaces.     

CpG islands in genes showing tissue-specific expression

Patterns of DNA methylation at CpG dinucleotides and their relations with gene expression are complex. Methylation-free CpG clusters, so-called HTF islands, are most often associated with the promoter regions of housekeeping genes, whereas genes expressed in a single-cell type are usually deficient in these sequences. However, in the human carbonic anhydrase (CA) gene family, both the ubiquitously expressed CAII and the muscle specific CAIII appear to have such CpG islands although erythrocyte-specific CAI does not. The CAII island is quantitatively more CpG rich than that of CAIII, with a CpG:GpC ratio of 0.94 compared with 0.82 for CAIII. Estimation of CpG:GpC ratios in the proximal-promoter regions of 44 vertebrate genes suggest that 40% of genes with tissue-specific or limited tissue distribution may show methylation-free CpG clusters in their promoter regions. In many cases the CpG:GpC ratio is less than that found in housekeeping genes and this may reflect variation in the interaction of CpG clusters with regulatory factors that define different patterns of tissue expression.     

Isolation and structure of an intrastrand cross-link adduct of mitomycin C and DNA.

A new covalent mitomycin C-DNA adduct (4) was isolated from DNA exposed to reductively activated mitomycin C (MC) in vitro. The MC-treated DNA was hydrolyzed enzymatically under certain conditions, and the new adduct was isolated from the hydrolysate by HPLC. Its structure was determined by ultraviolet and circular dichroism spectroscopy and chemical and enzymatic transformations conducted on microscale. In the structure, a single 2" beta, 7"-diaminomitosene residue is linked bifunctionally to two guanines in the dinucleoside phosphate d(GpG). The guanines are linked at their N2 atoms to the C1" and C10" positions of the mitosene, respectively. A key to the structure was a finding that removal of the mitosene from the adduct by hot piperidine yielded d(GpG); another was that the adduct was slowly converted to the known interstrand cross-link adduct 3 by snake venom diesterase and alkaline phosphatase. Adduct 4 represents an intrastrand cross-link in DNA formed by MC. Of the two possible strand-polarity isomers of 4, 4a in which the mitosene 1"-position is linked to the 3'-guanine of d(GpG) is designated as the proper structure, on the basis of the mechanism of the cross-linking reaction. The same adduct 4 was isolated from poly(dG).poly(dC), synthetic oligonucleotides containing the GpG sequence, and Micrococcus luteus and calf thymus DNAs. The relative yields of interstrand and intrastrand cross-links (3 and 4) were determined under first-order kinetic conditions; an average 3.6-fold preference for the formation of 3 over that of 4 was observed. An explanation for this preference is proposed.(ABSTRACT TRUNCATED AT 250 WORDS)     

Identification of tyrosine 204 as the photo-cross-linking site in the DNA-EcoRI DNA methyltransferase complex by electrospray ionization mass spectrometry

We describe a highly sensitive strategy combining laser-induced photo-cross-linking and HPLC-based electrospray ionization mass spectrometry to identify amino acid residues involved in protein-DNA recognition. The photoactivatible cross-linking thymine isostere, 5-iodoracil, was incorporated at a single site within the sequence recognized by EcoRI DNA methyltransferase (GAATTC). UV irradiation of the DNA-protein complex at 313 nm results in a >60% cross-linking yield. SDS-polyacrylamide gel electrophoresis and mass spectrometry were used to analyze the covalent cross-linked complex. The total mass is consistent with covalent bond formation between one strand of DNA and the protein with 1:1 stoichiometry. Protease digestion of the cross-linked complex yields several peptide-DNA adducts that were purified by anion-exchange column chromatography. A combination of mass spectrometric analysis and amino acid sequencing revealed that tyrosine 204 was cross-linked to the DNA. Electrospray mass spectrometric analysis of the peptide-nucleoside adduct confirmed this assignment. Tyrosine 204 resides in a peptide motif previously thought to be involved in AdoMet binding and methyl transfer. Thus, amino acids within loop segments but outside of "DNA binding" motifs can be critical to DNA recognition. Our method provides an accurate characterization of picomole quantities of DNA-protein complexes.     

Phenylalanines that are conserved among several RNA-binding proteins form part of a nucleic acid-binding pocket in the A1 heterogeneous nuclear ribonucleoprotein

We have studied the domain structure of the A1 heterogeneous nuclear ribonucleoprotein using both partial proteolysis and photochemical cross-linking to oligodeoxynucleotides. Both the intact A1 protein and its proteolytic fragment, the UP1 protein, can be cleaved by Staphylococcus aureus V-8 protease to produce two polypeptides of 92 amino acids. These two polypeptides correspond to the internal repeat sequence previously noted by us to occur in UP1. The two polypeptides can be purified via single-stranded DNA cellulose chromatography and independently cross-linked to [32P]p(dT)8, indicating that each domain can bind to single-stranded nucleic acids. Purification and sequencing of A1 tryptic peptides that had been cross-linked to oligothymidylic acid revealed that 4 phenylalanine residues, phenylalanines 16, 58, 107, and 149 are the sites of covalent adduct formation, with phenylalanine 16 being the major site of cross-linking. These phenylalanine residues are internally homologous when the repeat sequences in A1 are aligned, that is, phenylalanines 16 and 107 occupy analogous positions in the 91-residue repeat, as do phenylalanines 58 and 149. An examination of the primary structures of a variety of eucaryotic RNA-binding proteins with sequence homology to A1 reveals that the cross-linked phenylalanines in A1 are highly conserved among all of these proteins. Our results provide the first experimental evidence that conserved residues in the 90-amino acid repeating domains shared by A1 and other single-stranded nucleic acid binding-proteins form part of an RNA-binding pocket.     

Bioreductive activation of mitomycin C by DT-diaphorase.

The role of DT-diaphorase (DTD, EC 1.6.99.2) in the bioreductive activation of mitomycin C was examined using purified rat hepatic DTD. The formation of adducts with reduced glutathione (GSH), binding of [3H]mitomycin C to DNA, and mitomycin C-induced DNA interstrand cross-linking were used as indicators of bioactivation. Mitomycin C was metabolized by DTD in a pH-dependent manner with increasing amounts of metabolism observed as the pH was decreased from 7.8 to 5.8. The major metabolite observed during DTD-mediated reduction of mitomycin C was 2,7-diaminomitosene. GSH adduct formation, binding of [3H]mitomycin C and mitomycin C-induced DNA interstrand cross-linking were observed during DTD-mediated metabolism. In agreement with the pH dependence of metabolism, increased bioactivation was observed at lower pH values. Temporal studies and experiments using authentic material showed that 2,7-diaminomitosene could be further metabolized by DTD resulting in the formation of mitosene adducts with GSH. DNA cross-linking during either chemical (sodium borohydride) or enzymatic (DTD) mediated reduction of mitomycin C could be observed at pH 7.4, but it increased as the pH was decreased to 5.8, showing the critical role of pH in the cross-linking process. These data provide unequivocal evidence that the obligate two-electron reductase DTD can bioactivate mitomycin C to reactive species which can form adducts with GSH and DNA and induce DNA cross-linking. The use of mitomycin C may be a viable approach to the therapy of tumors high in DTD activity, particularly when combined with strategies to lower tumor pH.     

Modulation of mitomycin cross-linking by DNA bending in the Escherichia coli CAP protein-DNA complex

We have examined the comparative reactivity of mitomycin cross-linking sites in DNA molecules either free in solution or complexed with Escherichia coli CAP protein. Sites in the region to which the protein is bound show strongly variable cross-linking by the drug. The reactivity of a CpG site located where the minor groove is narrowed by bending toward the protein was decreased by about 4-fold, compared to free DNA. The reactivity of a site placed so that the minor groove is widened by the bend was reduced by about 25%, and the reactivity of a (CpG)3 sequence facing primarily away from the protein was reduced 25-fold by CAP binding. These results support the view that local DNA structure plays a critical role in determining the efficiency of cross-linking.     

Interstrand cross-links of cisplatin induce striking distortions in DNA

In the reaction between cellular DNA and cisplatin, different bifunctional adducts are formed including intrastrand and interstrand cross-links. The respective role of these lesions in the cytotoxicity of the drug is not yet elucidated. This paper deals with the current knowledge on cisplatin interstrand cross-links and presents results on the formation, stability and structure of these adducts. A key step in the studies of these lesions is the recent determination of solution and crystallographic structures of double-stranded oligonucleotides containing a unique interstrand cross-link. The DNA distortions induced by this adduct exhibit unprecedented features such as the location of the platinum residue in the minor groove, the extrusion of the cytosines of the cross-linked d(GpC).d(GpC) site, the bending of the helix axis towards the minor groove and a large DNA unwinding. In addition to a detailed determination of the distortions, the high resolution of the crystal structure allowed us to locate the water molecules surrounding the adduct. The possible implications of this structure for the chemical properties and the cellular processing of cisplatin interstrand cross-links are discussed.