ABSTRACT
Four nitrogen mustards have been used in this study to examine protein-DNA interactions in intact human cells, specifically at the locus control region hypersensitive site-2 (LCR HS-2) of the human [beta]-globin locus. Three of these nitrogen mustards are DNA-targeted by attachment of an acridine or amsacrine intercalating chromophore, while the fourth (chlorambucil) is a non-targeted mustard. The ligation-mediated PCR technique was used to determine the sites of damage at base pair resolution on DNA sequencing gels. A densitometric comparison was made between DNA damaged in intact erythroid K562 cells and in purified DNA. The intensity of DNA damage sites in the LCR HS-2 were found to differ significantly between intact K562 cells and purified DNA. At the NF-E2/AP-1 motif, pronounced damage protection was observed in DNA derived from drug treated cells. The nuclear factor- erythroid 2 (NF-E2) protein factor is thought to bind at this NF-E2/AP-1 motif in K562 cells. Other sites of protection and enhancement that corresponded to known transcription factor binding sites were also detected. These nitrogen mustards are therefore very effective compounds for detection of transcription factor binding to DNA in intact cells and are superior to other commonly used agents. The sequence selectivity of the compounds was determined using plasmid DNA and compared to that found in intact cells. The acridine-based nitrogen mustard had a preference for forming adducts at guanine bases, while the two amsacrine-based nitrogen mustards and chlorambucil formed adducts at both guanine and adenine bases.
It is now generally accepted that chromatin structure plays a role in controlling the expression of genes. Nucleosomes, the fundamental unit of chromatin structure, have been thought to act negatively on gene expression by preventing the access of transcription factors to the regulatory regions of genes (1 -4 ). The removal of nucleosomes from chromatin leads to the generation of a DNase I hypersensitive site (DHS) characteristic of actively transcribed genes. The [beta]-globin locus has a number of DHSs in the Locus Control Region (LCR) and in the promoter regions of transcriptionally active [beta]-like genes (5 ). The LCR confers tissue specific expression to the [gamma]- and [epsilon]-globin genes, gives rise to an `open' chromatin structure and acts like a traditional enhancer. The LCR HS-2 is 10 kb upstream from the [epsilon]-globin gene and 55 kb from the [beta]-globin gene. A number of transcription factors have been shown to bind in the LCR HS-2 including nuclear factor-erythroid 2 (NF-E2) (5 ,6 ). More recently the concept of nucleosomes acting negatively on gene expression has been modified by the discovery of transcription factors that interact with DNA even in the presence of nucleosomes (7 ). In addition the Swi/Snf proteins have been proposed to alter the interaction of nucleosomes with DNA (utilising ATP) to allow the interaction of transcription factors with DNA (8 ,9 ). Histone acetylation is also thought to play a role in this process (10 ).
The aim of the work reported here was to use the nitrogen mustards as DNA damaging agents to probe protein-DNA interactions in the chromatin of intact human cells. The presence of a protein bound to DNA can hinder the interaction of a damaging agent with DNA. Hence these protected regions can be used to build up a picture of protein-DNA interactions in an intact cell. The use of the ligation-mediated Polymerase Chain Reaction (LMPCR) technique (11 ,12 ) enables the sites of DNA damage to be examined at base pair resolution. In K562 cells, expression of the [beta]-globin genes can be manipulated by the addition of an inducing agent to elevate the levels of [epsilon]- and [gamma]-globin gene expression. (The [beta]-globin gene is not expressed in K562 cells.) Thus protein-DNA interactions can be determined for several different states of gene expression. In this paper the LCR hypersensitive site-2 (HS-2) was used as the target DNA sequence to examine the interaction of four nitrogen mustards with DNA in intact cells. This region of DNA in the LCR HS-2 is expected to be free of nucleosomes.
The LMPCR technique was first developed in 1989 (11 ,12 ) and allows the sequence specificity of a DNA damaging agent to be determined at base pair resolution in a mammalian single copy gene. The technique involves the extension of a gene specific oligonucleotide up to the site of damage, ligation of a double-stranded linker oligonucleotide, PCR using a 32P-labelled gene specific oligonucleotide and an oligonucleotide complementary to the linker oligonucleotide, gel electrophoresis and autoradiography. The size of the PCR product reflects the position of the site of damage that can be precisely determined using Maxam and Gilbert (13 ) sequencing reactions.
Nitrogen mustards have been used as anti-tumour drugs for a number of years and chlorambucil is an example of a widely used agent (14 ,15 ). The DNA sequence specificity of chlorambucil damage has been determined in plasmid and it was found to prefer runs of consecutive guanines using a cleavage assay (16 ,17 ) or at GA and AG using an inhibition of transcription assay (18 ).
Since DNA is thought to be the biological target for the anti-tumour activity of nitrogen mustards, it has been postulated that attachment of a DNA binding group to the molecule could improve the efficiency of these compounds (19 ). The concept is that the DNA-affinic group will rapidly place the compound in contact with DNA so that the nitrogen mustard can react more quickly and the lesions inside a cell will be directed towards DNA rather than proteins or other cell constituents. Several series of such `DNA-directed' nitrogen mustard analogues have been developed by attaching a nitrogen mustard to an amsacrine or acridine intercalating group and shown to have altered patterns of DNA alkylation and enhanced anti-tumour activity (20 -23 ).
Figure 1 shows the structure of the nitrogen mustards used in this study. In C3-AA the mustard is attached to 9-aminoacridine by a C3 chain, while C2O-AMSA and C5O-AMSA have the mustard attached to the amsacrine moiety by chains of two different lengths. Chlorambucil was used as a control that lacked an intercalating chromophore.
The synthesis of nitrogen mustards was carried out as previously described for C3-AA (22 ), C2O-AMSA and C5O-AMSA (23 ). Chlorambucil was obtained from Sigma.
K562 cells were grown in RPMI 1640 medium with the addition of 10% foetal calf serum and 0.01% penicillin, streptomycin and fungizome using conventional techniques. Hemin induction was carried out as described (24 ) by the addition of 50 [mu]M hemin (final concentration) 4 days before harvest of the cells.
Harvested K562 cells were washed with phosphate buffered saline and then incubated for 3 h at 37oC in the dark with the appropriate concentration of nitrogen mustards. The cells were washed twice with phosphate buffered saline and resuspended in 50 mM Tris-HCl, pH 7.5, 20 mM EDTA. DNA purification was carried out as described in Murray and Martin (25 ).
Purified human DNA and plasmid pUC19 were treated with the appropriate concentration of nitrogen mustard in 2 mM HEPES, pH 7.5, 10 mM NaCl, 10 [mu]M EDTA for 2 or 3 h at 37oC in subdued light. The DNA was ethanol precipitated and washed twice in ethanol and dissolved in 10 mM Tris-HCl, pH 8.8, 0.1 mM EDTA. In the plasmid experiments C3-AA was present at 30 [mu]M; C5O-AMSA, 45 [mu]M; C2O-AMSA, 30 [mu]M and chlorambucil, 200 [mu]M.
For the plasmid pUC19, the linear amplification technique and densitometry were performed as described (26 ) using the REV primer 5'-aacagctatgaccatg-3'.
LMPCR and densitometry on the LCR HS-2 were carried out as described in Cairns and Murray (27 ). Briefly in this procedure the first strand synthesis oligonucleotide (coding strand 5'-ATATGTCACATTCTGTCTCAGGCATCC-3'; non-coding strand 5'-TTCCTGTTACATTTCTGTGTGTCTC-3') is extended up to the damage site; this DNA is ligated to the double-stranded linker oligonucleotide (5'-GAAGAGAAGGT-3' and 5'-CCAAACGCCATTTCCACCTTCTCTTC-3'); PCR is carried out using a 32P-labelled PCR oligonucleotide (coding strand 5'-GTCTCAGGCATCCATTTTCTTTATG-3'; non-coding strand 5'-CATTAGTGACCTCCCATAGTCCAAGCA-3') and the linker oligo- nucleotide 5'-CCAAACGCCATTTCCACCTTCTCTTC-3'. These products are then analysed on a DNA sequencing gel. Maxam and Gilbert (13 ) sequencing reactions are subjected to the same procedure to pinpoint the precise site of damage (28 ).
We initially determined the DNA sequence specificity of the nitrogen mustards using the plasmid pUC19 as the target sequence. The data is summarised in Table 1 where the 12 most intense damage sites are shown. There were significant differences in the sequence selectivity of the nitrogen mustards. Considering the 12 most intense damage sites in Table 1 , C3-AA damaged exclusively at guanines (12/12) with runs of consecutive guanines being especially favoured. C5O-AMSA, however, had a preference for adenines (8As/12), while C2O-AMSA had a slight preference for guanines (7Gs/12), as did chlorambucil (7Gs/12). Table 1 also shows the sequences on either side of the site of damage. No obvious consensus sequence emerged from examining these sequences except that damage to the four compounds occurred mainly in runs of consecutive purines especially guanines.
Table 1 Sequence specificity of nitrogen mustard damage in double stranded plasmid pUC19 DNAa
The DNA sequence selectivity was also determined for the nitrogen mustards in the single copy human [beta]-globin cluster at the LCR HS-2 using the LMPCR technique. This data is shown in Table 2 for each compound, on both strands and the 10 most intense sites of damage are included for each strand. With respect to these 20 most intense damage sites for purified human DNA, C3-AA and C2O-AMSA predominantly damaged at guanines (18Gs/20) and (17Gs/20), respectively; while C5O-AMSA and chlorambucil damaged mainly at guanines (13Gs/20) and (14Gs/20), respectively. Table 2 shows the neighbouring sequences at the damage site. No obvious consensus sequence was found except that damage sites were particularly common at runs of consecutive purines especially guanines. In induced and uninduced K562 cells, the presence of transcription factor binding motifs modulated the degree of damage as shown in Table 2 .
Table 2 DNA sequence specificity of nitrogen mustard damage in the LCR HS-2a
Both strands of the [beta]-globin LCR HS-2 were examined from bp 8560 to 8800 using the LMPCR technique as shown in Figures 2 and 3 gel images. Control lanes with undamaged DNA were included to gauge the background level in these experiments. Damage in the nitrogen mustard treated lanes was significantly greater than in the control lanes.
Three DNA-targeted nitrogen mustard analogues have been used in this study, two with an attached amsacrine chromophore (C2O-AMSA and C5O-AMSA) and one with an attached acridine moiety (C3-AA). These three compounds were compared to chlorambucil, that is similar in structure but lacks an attached intercalating chromophore. The three DNA-targeted nitrogen mustard analogues were able to produce similar levels of damage at much lower concentrations, ~10-fold for plasmid and cells.
The DNA sequence specificity of the four nitrogen mustards was examined in plasmid DNA as well as in intact cells. The data with plasmid DNA revealed that the specificity of the acridine-based C3-AA was at guanine bases. Prakash et al. (20 ) also examined the sequence specificity of C3-AA in plasmid DNA and found specificity for guanine bases. Chlorambucil and the amsacrine-containing C2O-AMSA had a preference for guanine over adenine bases; while the C5O-AMSA had a preference for adenine over guanine. C5O-AMSA has a longer alkyl linker than C2O-AMSA and the length of the alkyl linker chain is correlated with preference for adenine over guanine base damage (20 ,26 ). In intact cells using the LMPCR technique, the sequence specificity was similar except that damage at guanine was more prevalent. With respect to the neighbouring sequences at the damage sites, no consensus sequence emerged from the analysis except that the sites of most intense damage were at runs of consecutive purines (especially guanines).
The mutagenesis spectrum for chlorambucil in an SV40-based shuttle plasmid has been determined in human cells (29 ). Interestingly a hotspot for mutagenesis (5'-GAAGGTT-3') is also the same sequence as the two most intense sites of damage for chlorambucil on the non-coding strand (Table 2 ). Amsacrine is thought to act as an anti-tumour agent by interacting with topoisomerase II. This enzyme preferentially interacts with the sequence 5'-CNGY/NGKT-3', where N is any base, Y is a pyrimidine, K is G or T and / is the site of cleavage (30 ). However, these sequences were not found in the list of most damaged sequences for the amsacrine analogues used here (Table 2 ).
In this paper the use of four nitrogen mustards as probes of protein-DNA interactions in intact mammalian cells has been examined. The human single-copy [beta]-globin gene cluster was used as the target DNA sequence, in particular the LCR HS-2 was investigated in detail.
With intact cells there were areas of DNA sequence where the intensity of damage was very different between intact cells and purified DNA. The main such site occurred at bp 8658-8678 where the transcription factor NF-E2 is expected to bind in erythroid cells. Using the DNA damaging agents, DMS (31 -34 ), bleomycin (27 ) and hedamycin (Cairns, M.J. and Murray, V. - manuscript in preparation), a `footprint' was also seen in this area. DMS gave the poorest `footprint', followed by (in increasing degrees of effectiveness), chlorambucil, hedamycin, C2O-AMSA, C5O- AMSA, bleomycin and C3-AA. The acridine-containing C3-AA gave the most `even' damage ratio (Fig. 4 ) and a clear `footprint' at the NF-E2 site; this is possibly the best `footprinting' agent to use. The two amsacrine-based compounds were extremely sensitive to local variations in protein-DNA structure as large changes are seen between damage in purified DNA and intact cells. These compounds are best used when high sensitivity is needed. The nitrogen mustards are superior to bleomycin since they damage at more sites and give an relatively even level of damage. Bleomycin, however, gave a much greater footprint at the NF-E2 site and a factor of 8-fold was seen (27 ) compared to 6-fold for C2O-AMSA.
Other sites of protection and enhancement that corresponded to known transcription factor binding sites were also detected and these included the GATA-1 and both CACC elements. At these potential transcription factor binding sites there were no clear `footprints', in fact enhanced regions were mainly seen. An alternative explanation concerning these enhanced regions is that they are part of the NF-E2 `footprint'. Using bleomycin, weak protection was seen at the GATA-1 and the distal CACC elements and enhancement at the proximal CACC site (27 ). Using C5O-AMSA a novel `footprint' was seen at bp 8622-8635 that has the sequence 5'-TTCTGTGTAACCT-3', no transcription factor is known to bind at this position.
At the edges of these protein `footprints', sites of enhanced damage occurred. At the border between a protein bound to DNA and naked DNA, large distortions in the structure of DNA can occur, including bending. For the DNA-targeted nitrogen mustards, the chromophore will intercalate between the DNA base pairs before damage is caused by alkylation. At a grossly distorted DNA region, both the intercalation and damage could be greatly enhanced. In addition the presence of amino acid side chains could assist in nitrogen mustard damage to DNA in these regions.
A three-dimensional X-ray structure of a leucine zipper comprising minimal elements from c-fos/c-jun has been elucidated (35 ). The protein binds to DNA in the major groove, which could interfere with a nitrogen mustard reacting with the N-7 of guanine. Since NF-E2 is a member of the bZIP family and is structurally related to the c-fos/c-jun AP-1 leucine zipper, a similar mode of binding to DNA might be expected.
A comparison was made between damage in induced and uninduced K562 cells for all four compounds. Only very small differences were found near the GATA-1 site. Hence our data implies that the protein-DNA interactions are not significantly altered by hemin induction. With bleomycin as the damaging agent, no significant differences were seen between induced and uninduced cells but with DMS significant changes were seen in the LCR HS-2 (27 ,32 ). Unlike DMS, the DNA-targeted nitrogen mustards and bleomycin are expected to intercalate into DNA before damage occurs. However, although chlorambucil does not intercalate into DNA and is expected to act similarly to DMS, it does not show changes between induced and uninduced cells.
There is the possibility that some of the effects are due indirectly to the cytotoxicity of the nitrogen mustards. However, these effects are expected to be minimal because in purified DNA, where cytotoxicty cannot be operating, damage was similar (in areas unaffected by protein-DNA interactions) to that in cells.
There are probably several reasons why the DNA-targeted nitrogen mustards are good `footprinting' agents in intact cells. They are able to enter cells easily, are large molecules and will intercalate into DNA before damaging the DNA. The size of the molecule is important as smaller molecules, such as DMS and chlorambucil, are not as restricted by steric hindrance in accessing DNA bound to a protein. In addition the process of intercalation is likely to be perturbed by the presence of a protein bound to DNA. The use of DNase 1 as a `footprinting' agent requires the preparation of nuclei and crucial factors could be lost during this preparation process. Bulky DNA damaging agents that can enter intact cells, such as those reported here, do not suffer from this drawback and have potential uses as probes of protein-DNA interactions.
This work was supported by the Australian Research Council, the Auckland Division of the Cancer Society of New Zealand and by an Australian Postgraduate Award to MJC.
*To whom correspondence should be addressed. Tel: + 61 2 9385 2028; Fax: + 61 2 9385 1483; Email: v.murray@unsw.edu.au
REFERENCES