Adriamycin-induced DNA adducts inhibit the DNA interactions of transcription factors and RNA polymerase.

Adriamycin is known to specifically induce DNA interstrand cross-links at 5′-GC sequences. Because 5′-GC sequences are a predominant feature of 5′-untranslated regions (transcription factor-binding sites, promoter, and enhancer regions), it is likely that adriamycin adducts at GC sites would affect the binding of DNA-interacting proteins. Two model systems were chosen for the analysis: the octamer-binding proteins Oct-1, N-Oct-3 and N-Oct-5, which bind to ATGCAAAT and TAATGARAT recognition sites, and Escherichia coli RNA polymerase binding to the lac UV5 promoter. Electrophoretic mobility shift studies showed that adriamycin adducts at GC sites inhibited the binding of octamer proteins to their consensus motifs at drug levels as low as 1 μM, but no effect was observed with a control sequence lacking a GC site. Adriamycin adducts at GC sites also inhibited the binding of RNA polymerase to the lac UV5 promoter. Adriamycin may therefore function by down-regulating the expression of specific genes by means of inactivation of short but critical motifs containing one or more GC sites.

The mode of action of the anticancer drug adriamycin has been examined extensively over the past 20 years. Many different studies have cited the interaction of the inherently reactive drug with cell membranes, DNA, proteins, metal ions, and molecular oxygen, leading to an apparently complex interplay of the mechanism of antitumor action, the major determinants of which may differ according to the properties of target cancer cells (Myers et al., 1988). Although the reactivities of adriamycin with a range of cellular constituents are well known, the specific cellular mechanisms involved and the ultimate cause of tumor cell death are still to be elucidated. One possible physiological action in tumor cells involves the altered regulations of DNA-binding proteins in actively transcribed DNA (i.e. the open DNA regions in nuclear matrix attachment sites (Ciejek et al., 1983)). Covalent attachment of the drug chromophore to specific DNA consensus sequences required for recognition by DNA-binding proteins may lead to altered levels and modes of binding by these proteins.
The majority of adriamycin administered to sensitive tumor cells is known to rapidly localize in the nucleus (Gigli et al., 1988). The drug has a high affinity for DNA, thus providing the driving force for further nuclear uptake. It is well known that intercalation is the immediate form of interaction, and there is an extensive body of evidence to show that one of the first cellular responses is the impairment of topoisomerase II activity (see for example reviews by Liu (l989); Holm et al. (l99l); Capranico and Zunino (l992); Sinha (l995)). However, upon reductive activation the drug can also bind covalently to DNA (Moore, 1977;Sinha, 1980). The sequence specificity of this interaction in vitro is highly GC selective (Cullinane and Phillips, 1990), and it appears that an interstrand cross-link occurs at this site (Cullinane et al., 1994b;Cutts and Phillips, 1995). Adriamycin-induced cross-links have also been reported in HeLa S 3 cells (Skladanowski and Konopa, 1994a), but it remains to be seen whether the same in vitro sequence selectivity applies, although studies with other sequence-specific damaging agents suggest that this is likely (Murray and Martin, 1985;Hartley et al., 1992;Cullinane and Phillips, 1994) .
Once a DNA-drug adduct is formed, it is widely accepted that the nature of the interaction impedes cellular functions that involve DNA (i.e. replication and transcription), particularly in the event of damage to both strands by the formation of an interstrand cross-link (Hopkins et al., 1991). It follows that DNA-binding proteins would be affected to varying degrees by their modified substrate. Although many intercalating drugs have been repeatedly shown to disrupt the actions of topoisomerase II, recently some covalently binding agents have also been shown to disrupt the binding of transcription factors to their specific consensus sequences (Broggini and D'Incalci, 1994;Welch et al., 1994;Sun and Hurley, 1994). If these adducts prevent binding of transcription factors to DNA in tumor cells, then the sequence selectivity of the particular drug will determine which transcription factors are affected, and hence which genes are inhibited. The net effect of this process is that gene-specific inhibition may occur, depending on the sequence specificity of the particular drug adducts.
The availability of in vitro electrophoretic mobility shift assay (EMSA) 1 systems for the octamer family of transcription factors provides a model system for the study of adriamycininduced inhibition of protein-DNA binding and avoids the denaturation, heating, or alkaline treatments required in other types of DNA damage detection, all of which contribute to instability of adriamycin adducts (van Rosmalen et al., 1995). The consensus recognition sequence for these proteins is an 8-bp ATGCAAAT element and involves a single GC dinucleotide within the octamer element that participates directly in binding of the Oct proteins through the POU-specific region of the POU DNA-binding domain (Klemm et al. (1994); reviewed in Herr and Cleary (1995)). An alternative recognition sequence for some octamer-binding proteins is the TAATGARAT motif present in herpes simplex virus immediate early gene promoters (O'Hare and Goding, 1988;Preston et al., 1988;Baumruker et al., 1988), but because this site does not include a GC dinucleotide in the binding site, it serves as an excellent control to assess the significance of adriamycin-induced interstrand cross-links at GC sites.
DNA-dependent polymerases are also potential targets for inhibition when site-specific DNA adducts are formed in promoter regions. Simple gel retardation assays for the detection of binding of eukaryote RNA polymerases are not available. However, the interaction between Escherichia coli RNA polymerase and the lac UV5 promoter has been extensively studied, and a simple assay for this interaction is well documented (Straney and Crothers, 1985;Gray and Phillips, 1993). Although there is a high frequency of GC sites in this region, they do not appear to play a role in direct contact with the RNA polymerase (von Hippel et al., 1984;Brodolin et al., 1993). This system therefore provides an opportunity to study the indirect effect of adducts on protein binding within the promoter region.
In this study we show that the covalent binding of low concentrations of adriamycin to GC sequences interferes with the ability of octamer proteins to bind at these sites. The binding of prokaryote RNA polymerase was also inhibited in its binding to the lac UV5 promoter.

EXPERIMENTAL PROCEDURES
Materials-Adriamycin hydrochloride was a gift from Farmitalia Carlo Erba, Milan, Italy. The Klenow fragment of DNA polymerase I and the radiochemicals [␣-32 P]dATP and dGTP were purchased from Amersham Corp. E. coli RNA polymerase and DNase free bovine serum albumin were obtained from Pharmacia Biotech Inc., and the restriction enzymes EcoRI, HindIII, and PvuII were from New England Biolabs. Phenol was obtained as a solid from International Biotechnologies Incorporated, lambda exonuclease was from Life Technologies, Inc., and calf thymus DNA was from Worthington Biochemical Corporation.
Plasmids and Nuclear Extracts for EMSA-The plasmids pUC119 H2B-boxϩ and pBSOA25 have been described previously (Baumruker et al., 1988;Bendall et al., 1993). These plasmids were digested with EcoRI and HindIII to release restriction fragments containing the following motifs: ATGCAAAT, the wild-type H2B gene octamer-binding sequence, and TAATGAATT, the OA25 high affinity binding site for the N-Oct-3-and N-Oct-5-binding proteins. The fragments were purified by electrophoresis through 2% agarose and then electroeluted in a biotrap apparatus (Schleicher & Schuell). The fragments were labeled with [␣-32 P]dATP using the Klenow fragment of DNA polymerase I and separated from unincorporated nucleotides and proteins using Nensorb 20 cartridges (DuPont NEN). Nuclear extracts from the secondary malignant melanoma cell line A2058 were prepared as described previously (Sturm et al., 1991), and the protein content was determined by the Bradford assay (Bio-Rad) as containing 2-4 g of protein/l.
Adriamycin Reactions-Drug-DNA reactions were performed in a volume of 5 l. These reactions typically contained 7 mM dithiothrietol, approximately 500 cpm of labeled DNA fragment, 25 M bp linearized pSP64 plasmid DNA, 20 M FeCl 3 , and 10 M adriamycin in a transcription buffer consisting of 40 mM Tris, 3 mM MgCl 2 , 100 mM KCl, 0.1 mM EDTA, pH 8.0. The reactions were supplemented with pSP64 at the optimal concentration (25 M bp) for the formation of adriamycin adducts with DNA (Cullinane et al., 1994a), and the amount of labeled DNA fragment added was assumed to be negligible. Reactions were allowed to proceed for up to 96 h. Samples were either used immediately in the EMSA assay or stored at Ϫ20°C for up to 1 week. The EMSA assay was performed following the conditions outlined elsewhere (Sturm et al., 1991), and the samples were characterized by nondenaturing polyacrylamide gel electrophoresis.
Exonuclease Assay-The OA25 and H2B fragments were 3Ј endlabeled at the HindIII site by using [␣-32 P]dGTP following standard procedures. Drug reactions were set up in 10-l volumes and included the components as described previously. The reaction was allowed to proceed for 40 h, after which samples were made up to 30 l with transcription buffer. The DNA was then precipitated with ethanol and resuspended in 5 l of a solution containing 0.36 units/l lambda exonuclease, 50 g/l bovine serum albumin, 67 mM glycine-KOH, 2.5 mM MgCl 2 , pH 9.4. Exonuclease digestion was allowed to proceed for 1 h at 37°C, and then an equal volume of 90% formamide loading buffer was added and samples were heat denatured at 90°C for 5 min before being subjected to electrophoresis using a 12% denaturing polyacrylamide sequencing gel. Maxam-Gilbert G sequencing lanes were performed as described elsewhere (Maxam and Gilbert, 1977).
RNA Polymerase Assay-A 188-bp fragment containing the lac UV5 promoter was isolated from the plasmid pCC1  using the restriction enzymes EcoRI and PvuII. The DNA was purified and 3Ј end labeled using [␣-32 P]dATP. After lyophilization, the fragment was resuspended in Tris-EDTA buffer, and calf thymus DNA was added to generate a 175 M bp stock of DNA. Drug reactions were performed as described for the octamer assay. After the reaction, 40 l of transcription buffer was added and the samples were extracted twice with an equal volume of Tris-saturated phenol and once with chloroform. Samples were precipitated with ethanol using glycogen as an inert carrier of the DNA, and samples were resuspended in 5 l of transcription buffer. The samples were then exposed to RNA polymerase as described previously (Gray and Phillips, 1993) using a final concentration of 615 nM RNA polymerase, 125 g/l bovine serum albumin, 10 mM dithiothrietol. This solution was incubated for 15 min at 37°C, and heparin was subsequently added for a further 5 min to quench nonspecific binding of the RNA polymerase. An equal volume of loading buffer (50% glycerol, 20 mM dithiothrietol) was added to samples prior to characterization by nondenaturing electrophoresis through a 5% Tris-glycine gel.

Adriamycin Adducts with the ATGCAAAT Octamer Consensus
Motif-Initial studies were performed to establish at what levels adriamycin-induced adducts would prevent octamer proteins binding to their consensus motif. The crystal structure of Oct-1 bound to ATGCAAAT has revealed major contacts of the POU specific domain of Oct-1 directly to the GC within this sequence (Klemm et al., 1994). It was therefore anticipated that the tetracyclic structure of adriamycin covalently bound at a GC site would pose a direct blockage to this interaction. A2058 human melanoma cells were chosen as the source of nuclear extract because they contain three different octamer proteins that produce markedly different mobilities in an EMSA assay. The octamer motif is important for promoter activation mediated through the generally expressed Oct-1 protein, such as the histone H2B and snRNA genes (reviewed by Herr (1992)). The octamer element is also a target-binding site for the N-Oct-3 and N-Oct-5 octamer factors (Schreiber et al., 1990;Sturm et al., 1991) encoded by the Brn-2 POU domain gene (Schreiber et al., 1993;Thomson et al., 1995). N-Oct-3 and N-Oct-5 are commonly present in neuroectodermal derived tissues (Schreiber et al., 1992;Thomson et al., 1994) including all human melanoma cell lines tested (Thomson et al., 1993). Fig. 1 shows that there is an adriamycin concentration-dependent inhibition of octamer binding with the N-Oct-3, N-Oct-5, and Oct-1 proteins, with all following similar inhibition patterns. At 10 M adriamycin, the binding of Oct-1, N-Oct-3, and N-Oct-5 is inhibited by approximately 25, 40, and 30% respectively (Fig. 1B). To confirm that this effect was due to covalent binding of the drug chromophore to the consensus sequence (or whether it was merely an effect of drug intercalation), 10 M of adriamycin was reacted with the DNA probe for increasing times prior to exposure to the nuclear extract (Fig. 2, A and B). The yield of adriamycin-DNA adducts is known to be optimal after a reaction time of approximately 40 h with a half-life of formation of adducts being about 15 h (Cullinane et al., 1994b), and this effect is consistent with that shown in Fig. 2B where inhibition of protein binding is maximal after 60 -80 h with a half-life of inhibition of all three octamer proteins being less than 20 h. Because the percentage inhibition of protein binding increased with drug-DNA reaction time, this confirms that the inhibition is due to specific adriamycin-induced adducts rather than other nonspecific effects such as intercalation. All three octamer-binding proteins were also inhibited (but to a lesser extent) by exposure of the DNA to 1 M adriamycin for 40 h (Fig. 2, C and D.).
The extent of formation of adriamycin-induced adducts with DNA is known to be catalyzed by the presence of FeCl 3 with approximately 5-fold more adducts being formed when DNA is reacted with adriamycin for 50 h in the presence of 10 M FeCl 3 compared with the absence of added FeCl 3 (Cullinane et al., 1994b). This catalytic role is shown in Fig. 2E where Oct-1, N-Oct-3, and N-Oct-5 proteins were scarcely affected when FeCl 3 was absent but resulted in extensive inhibition of binding capacity of all three proteins when FeCl 3 was present during the reaction of drug with DNA (Fig. 2, A and B). Overall, the inhibition of binding of octamer proteins correlates well with the known characteristics of formation of adriamycininduced adducts. Although these trends are quite clear and highly reproducible, there is some variation in the absolute level of formation of octamer protein-DNA complexes (5-17% in Fig. 2, B, D, and F), and this is due dominantly to the varying concentration of transcription factors present in individual nuclear extracts (Thomson et al., l993).
Adriamycin Treatment of a TAATGARAT-like Octamer-binding Site-Adriamycin cross-links (which comprise a large proportion of total adducts) occur almost exclusively at 5Ј-GC dinucleotide sites (Cullinane and Phillips, 1990). To confirm the effect of adducts at the GC position of the wild-type octamer sequence, an alternative binding motif was utilized that lacked the central GC site. The OA25 site containing the TAATGA-RAT-like motif TAATGAATT was isolated in a binding site selection protocol using antibodies to the Oct-1 protein (Bendall et al., 1993) and shows higher affinity for the Brn-2-derived proteins N-Oct-3 and N-Oct-5 than Oct-1. Moreover, these proteins show greater affinity for this site than the wild-type octamer sequence (Thomson et al., 1994). Fig. 3 shows the results of a concentration-dependent reaction of adriamycin with the recognition fragment before exposure to nuclear extract. There was no discernable effect of increasing drug concentration up to 50 M (the apparent decrease of binding in the 100 M adriamycin lane is due to extensive aggregation of the DNA, which is then largely trapped in the loading well).
DNA Sequence Specificity of Adducts in Octamer-binding Sequences-To ensure that adducts were located specifically within the octamer recognition site and not in an alternative motif, the location of adducts in the H2B and OA25 fragments were probed by digestion with lambda exonuclease. Fig. 4 shows the blockages to exonuclease digestion of the drugtreated DNA fragments (three drug levels, lanes 6 -8) for both the GC containing octamer sequence (H2B) and the control octamer sequence lacking the GC site (OA25). The location of these blockages (with respect to GC sites and the recognition sequences) is shown in Fig. 5 for the DNA reacted with 1 M adriamycin. Digestion of DNA by lambda exonuclease is typically stalled in a staggered manner 1-3 nucleotides prior to adduct sites. A major adduct site is clearly identified in Fig. 5 directly within the octamer-binding site of the wild-type motif and is located at the GC position. However, within the TAAT-GARAT-like motif (lacking a GC site), no adducts are evident, and this result is consistent with the documented GC specificity of adriamycin-induced adducts. With increasing adriamycin concentration and also in the presence of 20 M FeCl 3 (Fig. 4, lane 8) exonuclease digestion was limited due to the presence of a greater amount of adducts on both DNA fragments, and this results in all downstream adduct sites being greatly underestimated. However, in lane 8 of Fig. 4, the amount of radiolabel associated with the first adduct site is a good indication of the extent of adducts present at all other similar (GC) sites.
Effect of Adriamycin Adducts in the lac UV5 Promoter-When E. coli RNA polymerase binds to its recognition sequence, it has been shown by DNase I footprinting studies that it covers the Ϫ50 to ϩ20 nucleotide region when in the open promoter complex form (Carpousis and Gralla, 1985). Adriamycin 5Ј-GC adduct sites are well represented in this sequence but not at known direct contact sites. It was therefore of interest to establish if adducts at these sites would affect the interaction between RNA polymerase and the promoter region. Initial studies were conducted by exposing DNA to adriamycin for various times followed by the addition of the RNA polymerase. This resulted in complete inhibition of the ability of RNA polymerase to bind, probably due to unwinding of the promoter region by the presence of intercalated drug. Therefore, to probe the more specific effect of adriamycin adducts on formation of the transcription process, a rigorous phenol (2ϫ) followed by a chloroform cleanup of the DNA adducts was employed. The extent of complex formed between E. coli RNA polymerase and end-labeled DNA fragment (containing the lac UV5 promoter) was probed by a gel retardation assay.
The time dependence of formation of adriamycin adducts FIG. 1. Gel retardation assay of inhibition of binding by octamer proteins with increasing adriamycin concentration. A, the H2B probe was exposed to 0.5-100 M adriamycin in the presence of 20 M FeCl 3 for an incubation period of 24 h. Control samples were incubated in the presence of 20 M FeCl 3 (denoted Fe) or in the absence of both FeCl 3 and adriamycin (denoted C). The reacted probe was exposed to nuclear extract from A2058 cells. Electrophoretically retarded bands denote protein-DNA complexes due to Oct-1, N-Oct-3, and N-Oct-5 proteins present in the nuclear extract. B, PhosphorImager quantitation of the Oct-1 (q), N-Oct-3 (f), and N-Oct-5 (å) complexes show the trends of protein binding with increasing adriamycin concentration. Band intensities were calculated relative to the Oct-1 band in the 0.5 M adriamycin lane.
(prior to exposure to polymerase) under these conditions is shown in Fig. 6A. There is a decreasing amount of the DNA-RNA polymerase complex with increasing time of reaction of adriamycin with the DNA fragment (Fig. 6A) or with increasing adriamycin concentration (Fig. 7A), and the relative amount of DNA existing in the complexed form is shown in Figs. 6B and 7B, respectively. The extent of binding is reduced 5-fold with saturation at 10 M drug concentration. To confirm that the promoter region was being alkylated, the DNA was probed by digestion with lambda exonuclease for DNA reacted for increasing times with 5 M adriamycin. Sequence-specific blockages are evident in the image shown in Fig. 8, and the location of the blockages are shown in Fig. 9. Every potential adriamycin GC cross-linking site is alkylated, as indicated by inhibition of digestion several nucleotides 5Ј to each GC site. More importantly, adducts are evident at every GC site in the RNA polymerase-binding site.

Adriamycin Facilitated Disruption of Octamer Protein-DNA
Interactions-From the gel shift assays shown in Figs. 1-3, it is clear that micromolar levels of adriamycin form a sufficient number of adducts with the wild-type octamer sequence to inhibit the binding of N-Oct-3, N-Oct-5 and Oct-1 proteins.
Although there is no structural information available on the interaction of the N-Oct-3 and N-Oct-5 proteins with the octamer site, an identical methylation interference pattern is seen on the wild-type octamer sequence with all three of these proteins (Schreiber et al., 1990). The interaction of the Oct-1 POU domain with the octamer element is now well defined, with the crystal structure of the complex being resolved at 3.0 Å resolution (Klemm et al., 1994). From this structure it is evident that the Oct-1 POU specific domain contacts the 5Ј half of the site (ATGC), whereas the POU-homeo domain contacts the 3Ј half, with the domains binding on opposite sides of the DNA duplex. The adriamycin GC cross-link site in the motif plays a crucial role in the binding mode of the POU-specific domain as Arg-49 of the POU-specific domain forms a hydrogen bond with the O-6 of guanine and also makes contact with the O-6 and N-7 positions of the guanine on the opposite strand. Because these guanine bases are integral components of the cross-link as well as for Oct-1 protein binding, it is highly likely that there will be steric constraints for protein binding when an adriamycin cross-link (or adduct) is present. These contact sites have been directly demonstrated with the use of a variety of other FIG. 2. Inhibition of octamer protein binding as a function of drug reaction time. The H2B probe was reacted for 0 -96 h with 10 M adriamycin and 20 M FeCl 3 (A), 1 M adriamycin and 20 M FeCl 3 (C), or 10 M adriamycin in the absence of FeCl 3 before exposure to the A2058 nuclear extract (E). Band intensities were quantitated for A, C, and E as a percentage of the total band intensity (i.e. total end-labeled probe) in each lane, and the results are shown in B, D, and F, respectively. The time-dependent inhibition was fitted to a single exponential decay. chemical protection and interference assays, with dimethyl sulfate methylation and diethylpyrocarbonate carbethoxylation modification (Sturm et al., 1987;Baumruker et al., 1988;Pruijn et al., 1988;Verrijzer et al., 1990) of the guanine nucle-otides or ethylation of the phosphate backbone (Pruijn et al., 1988) of either strand each preventing Oct-1 DNA-protein interactions.
Importance of the GC Specificity of Adriamycin Adducts-Interestingly, the GC site of an octamer motif does not provide critical contacts in the recognition of DNA-binding sites by octamer-binding proteins. As has been discussed previously by Baumruker et al. (1988), flanking sequences may be crucial in binding site recognition with possibly few obligatory contact nucleotides within a defined site. Mutation analyses have shown that the GC site can be altered to TT and still retain good binding capacity. However, the Arg-49 that makes con-  Fig. 4 (10 M adriamycin). Only bands representing greater than 2% of the mole fraction of total radioactivity in the lane are shown. The boxed sequences represent the complementary strand to the consensus sequence of the alternative motif (A) and the sequence of the wild-type consensus motif (B). The numbering scheme employed is arbitrary. tacts with the site seems to be important in making flexible contacts with the DNA region , and this site therefore has been established as integral to the binding process. The bulky adduct caused by an agent such as adriamycin is unlikely to be able to be accommodated in such an interaction.
In the TAATGARAT motif, chemical modification and mutation analyses suggest that the Oct-1 POU-specific domain contacts the 3Ј-GARAT sequence (R indicates purine), whereas the 5Ј-TAAT sequence is contacted by the Oct-1 POU homeodomain. As the exonuclease digestion pattern shows (Fig. 5A), there are no adducts in this region to cause direct disruption of binding. However, DEPC alkylation (which forms adducts with adenine and guanine bases) does result in disruption of the interaction between the Oct-1 protein and the TAATGARAT binding sequence . These results collectively show that DEPC alkylation of the alternative motif probably interferes with octamer binding in a similar manner to adriamycin inhibition of octamer binding to ATGCAAAT. However, because the exonuclease digestion indicates that adriamycin has no capacity to form adducts at the TAATGAATT sequence, the adriamycin-induced inhibition must be directly due to adducts at the 5Ј-GC sequence. Previous studies of the characteristics of adriamycin-induced DNA adducts show that the nature of this sequence specificity to be due to an interstrand cross-link .
Overall, the results from this study indicate that the covalent adducts induced by adriamycin in DNA do have a potential inhibitory role in vivo with respect to protein binding to sequences containing 5Ј-GC sites. Because the octamer-binding proteins are able to recognize a variety of motifs with varying affinities of binding, it is likely that a selective drug-induced inhibition of some octamer proteins would also arise in vivo. It is questionable whether the extent of inhibition of octamer protein binding facilitated by adriamycin would be significant enough to lead to detectably altered regulation patterns of specific cellular proteins, but it is possible that the effects of adducts could be further magnified when the complex interactions of proteins required in transcription initiation are considered.
Inhibition of RNA Polymerase-von Hippel (1984) has proposed that RNA polymerase can be seen as a complementary surface to that exposed by the promoter region of DNA and that recognition is determined predominantly by hydrogen bonding, whereas the stability may be determined by the electrostatic interaction between basic residues of the protein and the DNA FIG. 6. Gel retardation assay of the inhibition of formation of the RNA polymerase-promoter complex by adriamycin. A, endlabeled 188-bp fragment containing the lac UV5 promoter was incubated with 5 M adriamycin and 40 M FeCl 3 for 0 -48 h (denoted by ϩAdr) before a phenol cleanup and exposure to RNA polymerase. Samples reacted in the absence of adriamycin (C) for 0 and 48 h are also shown. DNA probe not exposed to RNA polymerase is indicated by P. B, the percentage of labeled probe associated with the RNA polymerasepromoter complex with increasing time of reaction with adriamycin is shown. phosphate backbone, as well as interaction between hydrophobic groups of protein and methyl groups of thymine. Although there are five potential adriamycin adduct sites within the DNA sequence recognized by RNA polymerase, they do not interfere with the major contacts in the Ϫ10 and Ϫ35 regions. Formaldehyde cross-linking studies of the RNA polymerase bound to the lac UV5 promoter have detected DNA-protein interactions at the Ϫ50 to Ϫ49, Ϫ5 to Ϫ10, ϩ5 to ϩ8, and ϩ18 to ϩ21 regions of the DNA sequence (Brodolin et al., 1993). Adriamycin cross-links or monoadducts are not present at any of these positions but are within 2 nucleotides of the ϩ5 to ϩ8 interaction, and this may affect the extent of contact at this site. The adducts are also likely to influence nonspecific interactions of the polymerase with its recognition sequence and to inhibit transition to the open complex. Because significant bending of the DNA seems to be required for polymerasepromoter interaction (Cartenberg and Crothers, 1991), it is also possible that adriamycin adducts disturb the DNA bending potential in the promoter region.
Significance of Adducts-In order for transcription to proceed, transcription factors and RNA polymerase must be able to recognize and bind to their target sequences, and additional processes that involve large regions of DNA being bent bring distant components of the transcription machinery together (van der Vlet and Verrijzer, 1993). Therefore it is conceivable that adducts could inhibit gene transcription on three different levels: 1) by inhibiting DNA-binding proteins through steric constraints, rendering DNA unrecognizable due to adduct induced bending, or through subsequent mutation of the DNA bases required in the consensus sequence (Broggini and D'Incalci, 1994); 2) by inhibiting the potential of the DNA to bend into the required conformation to bring distant regions together and initiate transcription (van der Vlet and Verrijzer, 1993); and 3) by posing a direct blockage to the path of RNA polymerase (Cullinane and Phillips, 1990). Overall, the extent of inhibition of transcription facilitated by DNA adducts is potentially much greater than that predicted by the simple in vitro assays outlined in this study.
It is significant that many genes in the mammalian genome (and especially those genes such as protooncogenes associated with proliferation) have a high occurrence of GC runs in their 5Ј regions . For example the c-Ha-ras oncogene contains seven GC runs in the 5Ј-flanking region, FIG. 8. Lambda exonuclease digestion of adriamycin reacted lac UV5 promoter region. The 188-bp fragment was reacted with 5 M adriamycin and 40 M FeCl 3 for 0 -48 h, and the DNA was extracted twice with phenol and once with chloroform prior to ethanol precipitation. Samples were then digested with lambda exonuclease and characterized by electrophoresis through an 8% denaturing polyacrylamide gel. Maxam-Gilbert sequencing lanes are denoted as G.
FIG. 9. Location of adducts in the lac UV5 promoter region. Only bands comprising greater than 1% of the mole fraction of lane 4 of Fig.  8 are shown. The boxed sequence section represents the region covered by RNA polymerase when bound to the promoter in the open complex (Carpousis and Gralla, 1985). some of these runs being contained in Sp1 binding enhancer regions . GC modifying drugs such as adriamycin are likely to favor these regions due to the sequence content and accessibility of these regions (the tight packaging of DNA in nucleosome structures appears to become relaxed and accessible when assembled on nuclear matix attachment regions for transcriptional processes to occur) (Workman and Buchman, 1993;Krajewska, 1992;Wolffe, 1992;Ciejek et al., 1983). There are a number of transcription factors that require a GC in their consensus DNA sequences, including CTF/NF1, AP2, SP1, and Oct-1 and Oct-2 proteins (Wingender, 1990;Williams and Tijian, 1991), and all of these sequences are therefore potential targets for adriamycin alkylation.
It is now important to establish if there is a correlation between adriamycin treatment and selective targeting of genes that are high in GC content. Although adriamycin has been a major inclusion in chemotherapy regimes for 20 years and appears to act as a topoisomerase II inhibitor, a major research effort is still required directed toward elucidating its basic modes of action in tumor tissue. It has recently been shown that the ability of a range of adriamycin derivatives to induce interstrand cross-links correlated well with their cytotoxicity (Skladanowski and Konopa, 1994b), and it has generally been assumed that this effect is related essentially to the impairment of the processive movement of DNA polymerase and RNA polymerase. The present results suggest that interstrand crosslinks may well exert a separate earlier effect in inhibiting the binding of proteins such as transcription factors and RNA polymerase to the modified DNA and because of the short sequences (e.g. octamer) involved in promoter activity may confer gene specificity to adriamycin action.