A defined chromatin construct was employed to investigate the effect of nucleosome cores on the DNA damage caused by nitrogen mustard and cisplatin analogues. DNA damage in reconstituted and non-reconstituted (free) DNA was detected using the linear amplification procedure and analysed on polyacrylamide DNA sequencing gels using densitometry. These methods allowed the precise sites of DNA damage to be determined and indicated the relative intensity of damage at each site. For each compound, comparisons between damage levels in the free and reconstituted DNA (L/C ratio) permitted the influence of chromatin structure on drug–DNA interactions to be assessed. Other workers have found a number of compounds that preferentially damage the linker DNA in reconstituted nucleosomes, and these include aflatoxin B1 (32), benzo[a]pyrene diol epoxide (33), bleomycin, neocarzinostatin and melphalan (34).
Effect of molecular weight on L/C ratios
Most of the 14 DNA-damaging agents examined within this project displayed a capacity for footprinting nucleosomes (Table 1). However, the relative degree of damage inhibition conferred by nucleosome cores (L/C ratios) varied between compounds. Student’s t-tests provided a statistical method for assessing the relative significance of each footprint, and generally reflected the trends already established by the L/C ratios (Table 1). Table 1 revealed several interesting relationships between the physical and biological properties of the agents and their ability to damage nucleosomal DNA (discussed in the following text). The implications of such findings may eventually influence the future design and development of improved chemotherapeutics.
One of the most prominent associations observed was the relationship between each compound’s molecular weight and nucleosome footprinting capacity (L/C ratio magnitude). Table 1 lists all 14 DNA-damaging agents in order of descending average L/C ratio, and includes information regarding the basic physical properties of each agent. In general, larger compounds were associated with higher L/C ratios, while smaller compounds exhibited lower values. Furthermore, the compounds could be roughly divided into two broad groups: those with lower L/C ratios (ranging from 0.9 to 1.4) and those with higher L/C ratios (ranging from 1.5 to 3.5). The average molecular weight of the six compounds in the lower L/C ratio category was approximately 354. In contrast, compounds with higher L/C ratios (excluding bleomycin and DNase I) had an average molecular weight of approximately 631 – that is, almost twice the average size of the lower L/C ratio compounds. This implies that histone steric hindrance is a major feature in determining whether a compound can access DNA in a nucleosome.
Of the 14 DNA-damaging agents, chlorambucil and dimethylsulphate were the only compounds that did not produce detectable nucleosome footprints. Consequently, the corresponding L/C ratios were very small (approximately 1.0). These findings thus implied that neither chlorambucil nor DMS preferentially target the linker or core regions of reconstituted chromatin. For both compounds, this effect has also been described in previous investigations (32,35). Furthermore, these two agents also had the lowest molecular weight (MW) values of all compounds examined. Although chlorambucil and cisplatin are similar in size, differences between their DNA-damaging mechanisms may account for the different nucleosome footprinting capacities observed here. DMS, on the other hand, has been successfully employed in footprinting analyses to study a range of protein–DNA interactions (30,36–38). However, several transcription factor binding studies have also reported DMS to be a less-effective footprinting agent than larger compounds, such as bleomycin and various nitrogen mustard analogues (24,25). Together, these observations suggest that DMS may not be an effective probe for examining all protein–DNA interactions.
Other drugs classified within the lower L/C ratio category included cisplatin, carboplatin, cis-[PtCl2(C6H11NH2)2] and 2AcC3PtenCl2. Unlike chlorambucil and DMS, however, these compounds produced distinct nucleosome footprints which signified their preference for binding to the linker region of the nucleosome. Interestingly, the clinically successful anti-tumour drugs, cisplatin and carboplatin, both had L/C ratios of 1.3. Hence, it may be advantageous for new anti-tumour drugs based on cisplatin to have a similar L/C ratio. On the other hand, the bis-cyclohexylamine platinum analogue (cis-[PtCl2(C6H11NH2)2]) produced a similar L/C ratio of 1.2 and yet has not been shown to exhibit clinically effective anti-tumour properties (39).
The acridine-tethered analogue, 2AcC3PtenCl2 (L/C ratio 1.3), was the only DNA-targeted compound (12) classified into the ‘low’ L/C ratio category. This compound is significantly larger than cis-[PtCl2(C6H11NH2)2] and at least twice the molecular weight of cisplatin. While 2AcC3PtenCl2 induces substantial damage levels in purified DNA, it has not exhibited significant anti-tumour activity in animal models (11,28,40). In comparison, other DNA-targeted analogues in this study produced higher L/C ratios (>1.5) and have previously demonstrated significant cytotoxic activity. These include 4AcC3PtenCl2, 9-aminoAcC3PtenCl2 (40), trans-diamminePtCl phenazine-1-carboxamide (41) and the nitrogen mustard analogues C3-AA (14,16), C20-AMSA and C50-AMSA (13). This observation suggests that the chemotherapeutic potential of compounds with DNA-targeting mechanisms may be enhanced when their L/C ratios are higher. Interestingly, while the L/C ratio for 4AcC3PtenCl2 was over 1.5 times the magnitude of the L/C ratio for 2AcC3PtenCl2, the only point of difference between these analogues is the position of the tethered acridine group. This infers that the site of attachment of an intercalating moiety may directly affect the L/C ratio and biological activity of a DNA-targeted drug.
Larger compounds that were classified into the higher L/C ratio category included bleomycin, DNase I, DNA-targeted cisplatin analogues (excluding 2AcC3PtenCl2) and all DNA-targeted nitrogen mustard analogues. Table 1 shows that the DNA-targeted cisplatin complexes (trans-diammine PtCl phenazine-1-carboxamide, 9-aminoAcC3PtenCl2 and 4AcC3PtenCl2) collectively produced the highest L/C ratios. These were closely followed in rank by bleomycin and DNase I (the established nucleosome footprinting agents), and then the DNA-targeted nitrogen mustard analogues (C50-AMSA, C3-AA and C20-AMSA). Within this group of eight compounds, a diverse range of molecular weights, chemical structures and mechanisms of action were represented. A common feature, however, was their strong preference for binding to inter-nucleosomal DNA. In the case of bleomycin and DNase I, their propensity for targeting linker DNA has already been documented comprehensively (5,34,42–44). In the context of the current study, the magnitude of their size warranted a pronounced distinction between their binding preferences and those of the remaining compounds, which were significantly smaller. However, the L/C ratios of bleomycin and DNase I were not the maximum values established in this study and instead were ranked directly between the ratios of DNA-targeted cisplatin and nitrogen mustard analogues (see Table 1). Thus, the extent to which the mode of action of these agents dictated their capacity for damaging nucleosomal core DNA was considered in further detail.
Effect of an intercalator moiety on L/C ratios
Although a significant relationship between compound size and nucleosome footprinting capacity was observed in this study, L/C ratios did not increase proportionally with increasing molecular weights (see Table 1). However, a broader correlation between the compound classes, their DNA-damaging mechanisms and L/C ratios was distinct. For example, DNA-targeted compounds collectively gave rise to the highest L/C ratios and most prominent nucleosome footprints. For several of these agents, previous investigations have already demonstrated their ability to act as excellent probes of protein–DNA interactions in intact human cells (24,25,45–47). As these compounds intercalate with DNA via the tethered DNA-binding group, a large disruption of protein–DNA interactions would be required for binding to occur in the nucleosome core (48–50). Thus, DNA damage is expected to be significantly inhibited at the sites of positioned nucleosome cores compared to the inter-nucleosomal linker DNA (51,52).
Nitrogen mustard analogues were investigated in this to generate information about DNA-damaging agents with a different mechanism of action to that of the platinum-based compounds. Despite their structural differences, the three DNA-targeted nitrogen mustard analogues displayed very similar damage trends in the presence of positioned nucleosome cores. Table 1 shows the way in which the L/C ratios of these complexes are clustered together with respect to the ratios of other groups of compounds. This suggests that the DNA-damaging mechanism of action of these DNA-targeted nitrogen mustard analogues has a common effect on their interaction with reconstituted nucleosomal DNA, regardless of the type of DNA-intercalating moiety attached. In general, the damage data obtained for these agents also correlates well with observations made in previous investigations (24,25). Furthermore, in human cells, Temple and colleagues (25) found that C20-AMSA, C50-AMSA and C3-AA all produced more distinct protein footprints than either DMS or chlorambucil. This finding is consistent with qualitative and quantitative observations made in the current investigation (Table 1).
A prominent feature of the acridine and amsacrine-tethered nitrogen mustard analogues was that they gave rise to altered sequence specificities compared to the cisplatin analogues. Other DNA-damaging characteristics unique to each of the nitrogen mustard analogues were also observed. These differences were most readily apparent in the corresponding densitometry plots for each complex. In general, C3-AA produced ‘deeper’ and more even footprints than the amsacrine analogues, which gave rise to footprints containing greater fluctuations in the damage intensity ratios. Interestingly, this phenomenon was also described implicitly by Temple and colleagues (1997) who observed that C3-AA gave more even damage ratios and clearer footprints than C50-AMSA and other agents. In this study, however, C3-AA treatments also incurred greater ‘end effects’ in the resulting damage patterns, than either C20-AMSA or C50-AMSA. This was evidenced by distinct damage enhancement peaks within nucleosomal regions towards the ends of the construct, particularly in the SEQ strand (Figure 3). At these locations, the template DNA may not be as closely associated with nucleosomal proteins because of some degree of ‘unwinding’. The degree of constraint imposed on DNA at the periphery of the nucleosome structure is sometimes found to be sequence specific (34), and in this case, may facilitate the intercalation of acridine-rather than amsacrine-tethered complexes. Alternatively, binding of the acridine moiety may incur a higher level of nucleosome core dissociation and thus damage at these sites compared to amsacrine interactions. As L/C ratios alone give no indication of these damage trends, such effects should be considered when evaluating an agent’s capacity for damaging nucleosome core DNA.
In summary, these DNA modification studies have helped to construct a more detailed model for the mechanism of action of various DNA-damaging drugs in the presence of nucleosome structures. The findings reported here ultimately show that drugs with different chemical and chemotherapeutic properties can also display similar binding preferences in the presence of specific nucleosome cores. This information should be taken into account when designing novel anti-tumour drugs based on cisplatin or other DNA-damaging agents. Cisplatin (L/C ratio – 1.32), carboplatin (L/C ratio – 1.34) and bleomycin (L/C ratio – 1.76) were the only compounds examined in this project that are used clinically, but have very different L/C ratios. Hence, it could be argued that cisplatin analogues should have a low L/C ratio of about 1.3, while bleomycin analogues should have a higher L/C ratio of about 1.8. The development of chemotherapeutic agents with improved efficacy and reduced toxic side effects could benefit from such a rational approach.