Correlation of doxorubicin footprints with deletion endpoints in lacO of E. coli

Fundamentaland Molecular Mechanismsof Mutagenesis

ELSEVIER

Mutation Research 326 (1995) 17-27

Correlation

of doxorubicin footprints with deletion endpoints in lac0 of E. coli

W.D. Sedwick *, R.D. Anderson, J. Baxter, S. Donover, S. Schneiter, M.L. Veigl Department of Medicine, Case Western Reserve University, The Case Western Reserve Veterans Hospital and The Ireland Cancer Center of University Hospitals, UCRC Bldg. 2, Suite 200, II001 Cedar Rd., Cleveland, OH 44106, USA

Received 15 March 1994; revision received 28 July 1994; accepted 3 August 1994

Abstract This study explored the possibility that the sequence location of doxorubicin-induced deletion endpoints might relate to DNA structural alterations caused by doxorubicin binding to DNA. The 3’-OH endpoints of doxorubicin-induced deletions terminating in the 35-bp region of lac0 appear to distribute differently from spontaneous deletion endpoints. Doxorubicin-induced deletions focus in the 26-bp palindrome which is separated by a 9-bp region with no reverse complementary, whereas spontaneous deletion 3’-OH endpoints are found distributed throughout the operator region. In order to explore the mechanism of deletion induction by doxorubicin, drug footprinting studies were carried out with DNA labeled at the 5’ end of each of the complementary DNA strands encompassed by lac0. Doxorubicin protected the 9-bp region between the palindromic sequences from DNase I cutting and caused enhanced DNase I cleavage at symmetrical sites in the palindrome, which were inherently resistant to the nuclease in the absence of the drug. These symmetrical sites also define regions in which the occurrence of deletion endpoints is enhanced 6-fold in the presence of doxorubicin. This enhanced cutting and mutation occur in regions of the palindrome that are flanked by expected doxorubicin binding sites, but are not themselves binding sites of the drug. Similarly, other sites where the frequency of deletion endpoints increased in response to doxorubicin occurred directly adjacent to regions where doxorubicin appeared to inhibit cutting by DNase I. These results suggest that the

binding of doxorubicin in the palindrome directs both the frequency and the specificity of deletion formation in this gene region. Keywords:

Doxorubicin;

Deletion endpoint;

E. coli lac0

1. Introduction

Directing damage to specific DNA sequences provides a potential strategy for selectively killing tumor cells and may be an important mechanism

* Corresponding author.

underlying the activity of chemotherapeutic drugs which are currently in use. The important chemotherapeutic agent, doxorubicin, provides a relevant model compound which can potentially cause DNA sequence specific damage through a number of mechanisms. These may include: (1) enhanced DNA damage at drug specified topoisomerase II cleavable complex sites (Tewey et

0027-5107/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDZ 0027-5107(94)00155-3

18

W.D. Sedwick et al. /Mutation Research 326 (1995) 17-27

al., 19841, (2) specified location of DNA repair activity (Sancar and Rupp, 1983; Anderson et al., 1993b) or (3) targeted DNA damage at specific doxorubicin binding sequences (Chaires et al., 1990; Bachur et al., 1979; Cebula, 1986; Imlay and Linn, 1987). To determine whether preferential drug binding to DNA may underlie doxorubicin-induced DNA damage manifested as mutations, drug footprinting studies with doxorubicin, DNase I and the luc operator DNA sequence were undertaken. Footprints constitute sites arising from inhibited or enhanced DNA cutting by DNase I at DNA structural distortions caused by doxorubicin binding. In the present study, drug footprints are compared with deletion endpoints induced by doxorubicin, which were found in previous investigations (Anderson et al., 1991, 1993b) to focus in the fuc0 palindrome. The current study demonstrates that doxorubicin enhances DNase I cutting in DNA sequences immediately adjacent to or at the sites of doxorubicin-induced deletion endpoints.

2. Methods Genetic analysis of deletion endpoints

The bacterial strains and genetic analyses used for identification of doxorubicin-induced deletion endpoints compared in this study are fully described in previous publications (Anderson et al., 1991, 1993b; Halliday and Glickman, 1991; Halliday, 1992). Doxorubicin footprinting analysis

The two double-stranded DNA segments containing the lac operator region used in these studies were prepared by the polymerase chain reaction (PCR). For each analysis, one of the primers bracketing the DNA sequence region to be amplified by the PCR was pre-labeled with [y- 32P]ATP using polynucleotide kinase. The first segment was 398 bp long and was labeled on the 5’-end of a 21-bp primer that initiated synthesis from position 103 in 1acZ (Kunkel and AIexander, 1986), resulting in location of the labeled end of the 5’ terminus at a position that was 76 bp

from the lac0 palindrome. Similarly, the second segment was 262 bp long and was labeled on the 5’ end of a 20-bp primer that initiated synthesis from position 1098 in lacl (Farabaugh et al., 19781, resulting in a labeled endpoint that was 92 bp from the lac0 palindrome. Before use as a substrate in doxorubicin footprinting experiments, end-labeled operator-containing segments were treated with E. coli DNA polymerase I and cold dNTP (Sambrook et al., 1989) to eliminate any nicks or incompletely extended primers in the PCR product. After heat inactivation at 70°C for 10 min, end-labeled operator-containing segments were purified by ultrafiltration in Centricon-30 micro-concentrator tubes. The end-labeled DNA fragment (0.1 pg) was then incubated with different concentrations of doxorubicin and subsequently digested with 0.2 pg/ml of DNase I (United States Biochemicals, Cleveland, OH) for 30 s. In addition to doxorubicin, the 10 ~1 reaction mixes contained 10 mM Tris, pH 7.9, 10 mM KCl, 2 mM Na,CH,COOH, 10 mM MgCl,, 5 mM CaCl,, and 0.1 mM dithiothreitol (Chaires et al., 1990). Reactions were stopped by addition of 2.5 ~1 of a solution containing 0.25 M EDTA and 3 M ammonium acetate followed by a 15 min incubation at 70°C and precipitation with ethanol. The DNA fragments were then separated by electrophoresis on a polyacrylamide gel and visualized by autoradiography. A Zeineh laser densitometer was employed to quantitate the various end-labeled DNA fragments. For reduction of digital data collected from a Zeineh laser densitometer, a computer program was written in Basic and used to align peaks in gel scans relative to the DNA sequence and to equalize the distance between each base pair. Profiles obtained by analyzing the doxorubicin footprints on each complementary DNA strand could then be directly analyzed and compared. In experiments contrasting the efficiency of DNase I scission in the presence and absence of doxorubicin, all scans were also normalized for loading by reference to cleavage sites which were unaffected by doxorubicin binding (see Fig. 2). Similar results were obtained when several different bands were used independently for normalization. Therefore, the choice of band used for nor-

W.D. Sedwick et al. /Mutation Research 326 (I 99s) 17-27

malization did not bias the data obtained the analyses.

from

DNA sequencing

DNA sequencing was carried out according to the dideoxy chain termination method for PCR products described by Anderson et al. (1993a) with the same DNA template and primers used to make the operator-containing segments for footprinting analysis.

3. Results Fig. 1 illustrates the distribution and the frequency of 3’-OH deletion endpoints in the 35bp region of the Zuc operator from four sets of spontaneously occurring or doxorubicin-induced mutations in E. cd. The lac operator encompasses the binding site for the LAC repressor protein and contains palindromic sequences which could form a potential stem-loop structure. Fig. 1 illustrates the point made in previous communications (Anderson et al., 1991, 1993b) that doxorubicin appears to cause deletion endpoints to

-t ,CeG

4.G

Imp

6.6

+.A 6.6 ;.A

d.C

-I wild type

I

Ratio

C-7 * +6) (+16 + +28)

i q.***.

(xl09

P”’

(+7 + +15)

‘A

C;.t $.b, +.A A.f-(+20) (.1)-L+ dt

(-IO)

localize in the palindromic sequences of lac0, while the endpoints of spontaneous deletions are distributed evenly throughout this region. The autoradiograph from a typical DNase I footprinting experiment is shown in Fig. 2. In the region of interest within the luc operator there is good resolution of individual bands which differ in length by one nucleotide. The intensity of any individual band reflects the DNA cleavage frequency of DNase I at a specific site in the luc0 sequence. There are very few background bands appearing in the undigested control, confirming that the PCR method used to prepare template for these experiments (see Methods) produced little aberrant template. In the DNase I digested control lanes from samples containing no drug, it is clear that cutting is not homogeneous throughout the sequence. For example, the adeninethymidine rich motif (between - 1 and + 5; between + 17 and + 23) is especially resistant to cleavage in the absence of doxorubicin. This result is consistent with the known behavior of DNase I (Drew and Travers, 1984). Despite uneven cutting, the DNase I cleavage patterns clearly vary as a function of treatment with dox-

MutationEkqwney

'G

G

19

I

I

I 3

I 4

I

1 I

fl

f.;\

Fig. 1. 3’ Deletion endpoints in the palindrome region of the Zac operator. The operator palindrome is illustrated in a putative stem-loop configuration. The frequency (rounded to the nearest whole number) of 3’ deletion endpoints falling within the loop versus the palindrome sequences comprising the stem is tabulated from four sets of mutations which were analyzed previously at the DNA sequence level. The four sets include spontaneously occurring deletions in wild type (Halliday and Glickman, 1991) and uurB- E. coli (Halliday, 1992) and doxorubicin-induced deletions in uurB_ (Anderson et al., 1991) and wild type organisms (Anderson et al., 1993b). The ratio of the frequency of deletion endpoints in the palindrome versus the loop in the foldback structure is also shown.

W.D. Sedwick et al. /Mutation Research 326 (1995) 17-27 TAC(:

cont.

0.1

DNaseIDigestion

30

Dox

a05

30

0.02

0.01

0.005

0

0

0

(mg/ml)

33

30

30

5

30

0

kec)

Fig. 2. Autoradiogram showing the dose response for DNase I cleavage of la&-containing DNA in the presence of doxorubicin. Labeled DNA encompassing the lac0 sequence was treated with the indicated concentrations of doxorubicin and digested with DNase I, as described in Methods. Samples were then loaded on a standard DNA sequencing polyacrylamide gel, along with the dideoxy sequencing reaction products generated from the lac0 template (TACG). After electrophoresis the gel was dried and exposed to X-ray film. A portion of the lac0 sequence is depicted to the left of the autoradiogram extending from the top of the figure to the bottom in the 5’ --f 3’ orientation. This represents the sense strand relative to the lacl and 1acZ genes flanking lac0. The boxed sequence containing letters in outline type represents self-complementary sequence within the lac0 palindrome distinguished from sequence which is not self-complementary and falls within the loop of the potential stem-loop structure designated in normal type face.

The apparent inconsistency in dose response between the lanes treated with 0.02 and 0.01 mg/ml doxorubicin is due to a variation in sample loading in these gel lanes. Fig. 3A illustrates the doxorubicin concentration-dependent effects on DNA cleavage by DNase I in 1ucU. The densitometer analysis of the autoradiogram from Fig. 2, without correction orubicin.

for gel loading as described in Methods is depicted. Individual peaks in the densitometer tracings represent individual bands from the autoradiogram. Areas under the peaks correlate with band intensities. A general reduction of peak intensities occurs at the highest concentration of doxorubicin (0.1 mg/ml) suggesting generalized inhibition of cleavage by DNase I. The intensity

W D. Sedwick et al. /Mutation

of individual bands, which are representative of the degree of DNase I cutting between specific adjacent nucleotides, illustrates that DNase I cutting can be enhanced or inhibited by doxorubicin. The concentration dependence of doxorubicin enhancing sensitivity to DNase I cutting at a specific site (E) is contrasted with another site exhibiting doxorubicin-induced inhibition of DNase I scission (I) in Fig. 3B. For this type of analysis these values were determined after correction of the scans for total loaded sample. A

21

Research 326 (1995) 17-27

relatively steep response for both inhibition and enhancement is observed up to a doxorubicin concentration of 20 pg/ml. The enhanced cutting effect in band E is attenuated by higher concentrations of doxorubicin (SO pg/ml), presumably reflecting the generalized inhibitory effects of high concentrations of doxorubicin on DNase I cutting. The correlation between occurrence of doxorubicin-induced deletion endpoints and doxorubicin-modulated sensitivity to DNase I cutting in

01 0

20

40

Doxorubicin

60

60

100

[rJg/ml]

Fig. 3. Doxorubicin-induced modulation of DNase I digestion in LacO. Individual lanes from the autoradiogram depicted in Fig. 2 were scanned by laser transmission densitometry (panel A). Uncorrected densitometer tracings are displayed from left to right in the 5’ to 3’ direction relative to the sense strand of the lacl and IacZ genes flanking lac0. The actual sequence for this region of DNA is depicted beneath the densitometric tracings. Peaks represent DNase I cut sites in the lac0 DNA. The actual sites of cleavage can be determined by correlation with the sequence of the DNA generated from the sequence ladder in the autoradiogram. The arrows E -+ and I + illustrate doxorubicin-induced enhancement and inhibition of DNase I cleavage, respectively, relative to DNA digested for an identical time period (30 s) in the absence of doxorubicin. DNAs digested for 5 s (lane 1) and 30 s (lane 2) in the absence of doxorubicin demonstrate that the pattern of DNase I cleavage is only quantitatively affected by increasing DNase I digestion time within the time period chosen for this experiment. Lanes 3-7 depict the pattern of DNase I cleavage of ZucO in the presence of increasing amounts of doxorubicin, 5 pg/ml, 10 pg/ml, 20 pg/ml, 50 pg/ml and 100 pg/ml, respectively. The dose-response relationships for the doxorubicin-enhanced cleavage at site E (solid circles) and doxorubicin-inhibited cutting at site I (open triangles) by DNase I are compared in pane1 B. Before quantitatively evaluating enhanced and inhibited cutting at specific sites, each trace is corrected for sample load as described in the Methods section. The degree of cutting was determined by integrating the area under each peak. The y axis depicts these integrated areas as numbers of OD units, which reflect relative degree of cutting at a given site by DNase I. This relative cleavage is then plotted as a function of doxorubicin concentration.

WD. Sedwick et al. /Mutation Research 326 (1995) 17-27

22

luc0 DNA is illustrated in Fig. 4. This figure shows net doxorubicin-induced enhancement and inhibition of DNase I strand cutting after subtraction of results obtained from the control strands digested with DNase I in the absence of doxorubicin (see Methods). The alternating

purine-pyrimidine regions encompassing positions - 7 to - 2 and positions + 24 to + 28 were generally protected from DNase I cutting in the presence of doxorubicin. Doxorubicin also induced two regions where enhanced and inhibited cutting by DNase I are juxtaposed on opposite

r--------7T-LzJ 20 m/ml Doxorubicin Footprint

INHIBITED \I

C-5) 5’3’-

c-t’

(+5) (+? GDGCOCRTR I- !==#I I I CTC(ICCT*T

(+I I

I

I (-)

Fig. 4. Correlation of the doxorubicin-modulated DNase 1 footprint with doxorubicin-induced deletion endpoints in lac0. Corrected, complementary DNase I footprinting profiles generated with 0.02 mg/ml of doxorubicin are depicted in register with histograms representing the DNA sequence specific frequency of 3’ deletion endpoints recovered in lac0 from doxorubicin-treated and untreated cultures of E. coli. Doxorubicin-modulated DNase I footprinting profiles were corrected by subtracting the corresponding profiles obtained in the absence of doxorubicin. In order to compare DNase I cutting in either of the complementary DNA strands encompassing lac0, separate footprint experiments were performed in which one or the other strand was end-labeled as described in Methods. The bold face line represents results obtained from the DNA template 5’ labeled in lacZ (- strand, antisense orientation with respect to the lacl and 1acZ genes). The unenhanced line represents results obtained from the DNA template 5’ labeled in lacl (+ strand, sense orientation with respect to the lucl and lacZ genes). Matching peak positions in the complementary strands are directly compared after computer aided alignment and depicted 5’ to 3’ with respect to the sense strand. The frequencies of 3’ deletion endpoints at given sites in the lac0 sequence are depicted as histograms in register with the footprinting profiles of both the sense and antisense strands in the lac0 DNA sequence region. Spontaneous 3’ deletion endpoints are from Halliday and Glickman (1991) and Halliday (1992); doxorubicin-induced 3’ deletion endpoints are from Anderson et al. (1991, 1993b) for uurB_ and uurb+ wild type organisms, respectively. The luc0 sequence is depicted at the bottom of the figure in register with both the footprinting profiles and mutation frequency histograms. Expected doxorubicin binding sites, based on previous crystallographic and footprinting studies (see Discussion), are indicated in this sequence by the symbols which vertically span the complementary DNA strands. The vertical bars indicate the likely steps in the DNA sequence where the doxorubicin chromophore might intercalate. The boxes linked to the vertical bars by a line indicate the likely orientation of the sugar side chain of doxorubicin which lies in the minor groove. Based on studies with daunorubicin (Chaires et al., 1990), the largest symbol represents the site of highest expected doxorubicin binding frequency; the intermediate sized symbols indicate sequence triplets where binding may possibly be less frequent, while the smallest symbols show the additional possible binding sites which were indicated in a study of doxorubicin-induced transcriptional stop sites (Trist and Phillips, 1989).

FKD. Sedwick et al. /Mutation Research 326 (1995) 17-27

DNA strands. These regions are symmetrically located on the antiparallel strands of the DNA palindrome between positions - 2 and + 3 and between positions + 18 and + 24. The majority of doxorubicin-induced 3’-OH deletion endpoints occurred within or next to these two symmetric regions where one DNA strand is enhanced for cutting and the other strand is inhibited for cutting by DNase I. There are two additional sites where cutting by DNase I is reduced by doxorubicin in the central portion of the antisense strand (between positions + 2 and + 9 and between positions + 9 and + 16). Doxorubicin-induced deletion endpoints also occur at the bases immediately adjacent to these regions (3’ to the protected region, between positions + 8 and + 9, and 5’ to the protected region, between positions + 15 and + 16).

4. Discussion Studies with a number of anthracycline congeners have linked toxicity of these agents with their ability to bind DNA (Arcamone, 1981; Gelvan and Samuni, 1986). As a consequence of binding by intercalation, anthracyclines cause DNA unwinding, DNA stiffening and elongation (Reinert, 1983), induce DNA strand scission (Goldenberg et al., 1986; Ross et al., 1979) and interfere with topoisomerase II activity (Nelson et al., 1984; Tewey et al., 1984). Daunorubicin and doxorubicin are distinguished from many other DNA intercalators by the kinetics of their binding to DNA (Chaires et al., 1982; Chaires, 1985a,b; Britt et al., 1986). Anthracycline binding to DNA appears to take place in three distinct steps: (1) rapid binding to the outside of the DNA helix, (2) intercalation, and (3) conformational alteration of both the bound anthracycline and the DNA (Chaires, 1985b). Stop flow kinetic analysis of the strong association of anthracyclines with DNA under physiological conditions has shown that these drugs are much slower to dissociate from DNA and show complex kinetics when compared with other toxic simple DNA intercalators such as phenanthridine and acridine (Krishramoorthy et al., 1986). Daunorubicin also exhibits a striking

23

preference for B- versus Z-DNA and an ability to shift the equilibrium of left-handed Z-DNA to right-handed B-DNA (Chaires, 1985b). The effects of doxorubicin binding to the luc operator region were successfully analyzed using a DNase I drug footprinting technique. The results of this analysis are consistent with the previously delineated DNA binding characteristics of doxorubicin. In addition, footprinting revealed symmetrical regions of DNA structural changes which appear to be associated with doxorubicin binding to the la0 sequence. The correlation between these unique regions and doxorubicininduced 3’-OH deletion endpoints in fuc0 suggests that doxorubicin-induced structural alterations in DNA direct mechanisms leading to deletions. In these studies, footprinting of the 1~0 region was facilitated by use of the PCR for generation of labeled DNA template. To generate the two selectively strand-labeled 1~0 DNA segments for footprinting analysis, individual primers were 5’ end-labeled prior to production of each segment by the PCR. Internal nicks in these end-labeled DNA segments were then removed by nick translation with E. coli DNA polymerase I. Published techniques require restriction enzyme sites which flank the sequence of interest to allow selective labeling of sense (+ > or antisense (-) strands (Bailly et al., 1993). However, since PCR primers can be synthesized to optimally bracket any sequence of interest, there is no need to subclone sequences into plasmid vectors to generate restriction sites. Optimal primer sites can be chosen to create an end-labeled DNA fragment of an appropriate size which will allow single nucleotide step resolution on DNA sequencing gels of the sequence region to be analyzed. This rapid method yielded results which are consistent with previous footprinting analyses. The known sequence specific characteristics of anthracycline binding to DNA facilitate interpretation of DNase I footprinting results in lac0. Fluorescence quenching studies suggest that doxorubicin has an affinity for alternating purinepyrimidine sequences Wedaldi et al., 1982) while spectroscopic methods show the drug intercalates at GpC sequences of native DNA (Manfait et al.,

24

W.D. Sedwick et al. /Mutation

1982). Crystallographic analysis of daunorubicin binding to the hexamers S-d[CGTACG] (Wang et al., 1987) and 5’-d[CGATCG] (Moore et al., 19891 demonstrates that the chromophore intercalates between the GpC bases with the daunosamine side chain lying in the minor groove adjacent to the A. T base pair. Doxorubicin is unusual among intercalators in that its tight binding characteristics and the long half-life of its binding complex have allowed the drug to be footprinted even though its preferred binding complexes involve only three bases (Chaires et al., 1990). High resolution footprinting studies suggest the triplet sequences 5’-(A/T)GC and S-(A/T)CG (where A/T means either A or T) represent the highest frequency sites of daunorubicin binding. In these studies, S-(A/T)CA appeared to be a secondary binding site, consistent with the loss of a potential hydrogen bond between the guanine base and daunorubicin upon substituting an A for the G (Chaires et al., 1990). Earlier studies showed that footprinting patterns were similar for doxorubicin and daunorubicin (Chaires et al., 1987); however, additional RNA polymerase inhibition studies suggest 5’-TCA may also be a preferred site for doxorubicin binding (Trist and Philips, 1989). In the present study, DNA sequences in fac0 protected from cutting by DNase I are consistent with the expected doxorubicin binding sites. Inhibition of DNase I activity at or adjacent to sites of drug binding may arise from a variety of mechanisms. DNase I is not a sequence specific enzyme but its ability to cleave the phosphodiester bonds of DNA can be markedly affected by alterations in DNA helical structure (Neidle et al., 1987; Drew and Travers, 1984). Doxorubicin leads to stiffening of DNA (Reinert, 1983) and flexibility variation in DNA has been directly correlated with DNA sequence specific cutting of DNase I (Hogan et al., 1989). Other changes influencing activity include indirect alteration of helical configuration in the sequence region caused by a combination of drug intercalation and minor groove binding. Finally, DNase I inhibition may reflect a direct blockade of the enzyme resulting from association of the side chain sugar of doxorubicin with the minor groove. Putative doxorubicin binding sites in lac0 are indicated symboli-

Research 326 (19951 17-27

tally in Fig. 4. A single representative, 5’-AGC, of the preferred triplet binding class (5’-(T/A) GC or CG) appears in the central portion of the sequence. There are also six 5’-ACA triplets which are apparent secondary binding sites (5’-(A/ TIC(A/T)l for the drug (Chaires et al., 1990). Four of these S-ACA triplets in lac0 symmetrically appear in the palindrome of the operator region. Moreover, two additional putative doxorubicin binding sites (5’-TCA) are also separately highlighted in Fig. 4. These 5’-TCA sites have been described as the major triplet class of binding sites for doxorubicin by assay of doxorubicin’s ability to block transcription (Trist and Philips, 19891-a result which stands in contrast to studies monitoring interference with DNase I binding, which highlight the importance of the 5’-A/T(GC or CG) triplets (Chaires et al., 1990). In the present analysis, except for a possible drug binding site at position 17/18, all putative sites of drug binding correlate with inhibition of DNase I cutting on one or both DNA strands. The pattern of enhanced cutting by DNase I emphasizes the symmetrical DNA structural alterations induced by doxorubicin in the LAC operator sequence. As in previous studies (Chaires et al., 1987, 1990; Drew and Travers, 1984; Fox and Waring, 19841, enhanced cutting occurs adjacent to drug binding sites which are otherwise refractory to cutting in the absence of the drugs. In particular two potential doxorubicin binding sites at each of the alternating 5’(ACACA) segments and a single 5’-ACA binding site flank the 5’-AAIT(T) segments in the lac0 palindrome. Doxorubicin binding at these sites may alter the major and minor groove dimensions from a DNase I refractory narrow and deep minor groove and flat major groove configuration to create a more optimal groove configuration just adjacent to the drug binding sites which can be easily cleaved by DNase I. Similar phenomena have been attributed to propagation of DNA unwinding adjacent to sites of DNA intercalation (Drew and Travis, 1984; Suck and Oefner, 1986). The occurrence of enhanced cutting opposite inhibited cutting on complementary DNA strands as observed between positions + 17 and + 24 is not common, but has been documented in other

W.D. Sedwick et al. /Mutation Research 326 (1995) 17-27

DNA segments for both daunorubicin and ellipticine (Bailly et al., 1993). The direct correspondence of enhanced cutting on one DNA strand and inhibited cutting on its complementary strand may reflect influence from an additional 5’-TCA doxorubicin binding site which exists because of a single mismatch on one side of the lac0 palindrome. This binding site would allow bound doxorubicin to flank the TTTAA sequence on one side of the palindrome more closely than the TTAA site on the other side of the palindrome and may more sharply focus the pattern of DNase I cutting in this region. In both cases, however, symmetrical doxorubicin binding sites lie directly adjacent to both sides of the region of enhanced cutting. Further, the orientation of the sugar side chain of doxorubicin which lies in the minor grove, as well as the intrusion of the aglycon portion of the doxorubicin chromophore into the major groove of the DNA helix, may orient DNase I association and lead to the observed enhanced cutting of the opposite DNA strands of the palindrome at the two symmetrically enhanced cutting sites (Fig. 4). The accumulation of 3’-OH deletion endpoints adjacent to sites of drug enhanced and inhibited cutting by DNase I suggests that doxorubicin-induced helical alterations may play an important role in the mechanism of DNA deletion induction. Observed symmetric foci of deletion endpoints (between - 1 and + 5 and between + 18 and + 22) are unlikely to have resulted from a simple drug-induced redistribution of the endpoints of spontaneous mutants to sites which are unprotected by doxorubicin since these events occur at much higher frequency in the presence of doxorubicin. The proteins which mediate deletions in response to these helical perturbations, however, have yet to be identified. The overlap of the pattern of 3’-OH lacI/lacO deletion endpoints and the DNase I footprints suggests that mechanisms leading to deletion are most likely to be potentiated by the same DNA structural distortions. The overlap between the distribution of mutations and DNase I cutting cannot be attributed to formation of unusual DNA structures such as cruciforms or Z-DNA which may potentially form in lac0 (Wells, 1988), since none of

25

these structures would have been favored for formation in the linear DNA segments used in the footprinting reactions. Finally, proteins which normally bind lac0 in vivo, such as the LAC repressor, are unlikely to have determined the DNA sequence specificity of doxorubicin-induced deletion endpoints. Although these proteins are present in E. coli, doxorubicin-induced deletion endpoints correlated with footprints generated in the absence of proteins which may associate with the operator. In addition, preliminary analysis of a doxorubicin-induced spectrum under conditions where the LAC repressor binding was inhibited by the LAC inducer (IPTG) indicates that repressor binding does not prevent doxorubicin-induced focus of deletion endpoints in lac0. However, LAC repressor binding to lac0 does appear to have other generalized permissive and inhibitory effects on mutation induction in lac0 (unpublished observation). One specific aspect of the data presentation for the mutant distributions must also be mentioned. Deletion endpoints depicted in Fig. 4 include a ‘hot spot’ for deletion mutations which has been found in most spontaneous and induced mutation sets encompassing the lac0 region. This deletion, which extends between position 1108 in lacl and position + 4 of lac0 (Farabaugh et al., 1978), is bracketed by a 7-bp direct repeat. Therefore, its endpoints may have occurred anywhere in this 7-bp region. This mutation accounts for two of six deletions in doxorubicin-treated wild type cells and four of six doxorubicin-induced deletions in uvrB- E. coli, with endpoints designated in Fig. 4 at the + 4 site. The placement of the endpoints of these deletions at the + 4 position of lac0 and the 1108 position of lacl was based on the existence of an adjacent optimal binding site for doxorubicin at the 5’ end of these deletions, which correlated with a probable consensus sequence for many of the 5’ endpoints of the doxorubicin deletions in our earlier studies (Anderson et al., 1991, 1993b). On the other hand, based on the probable doxorubicin binding sites, the drug binding site (5’-AGC) within the repeat sequence region for this deletion (i.e. at positions + 4- + 11 of lac0) could also focus

26

W.D. Sedwick et al. /Mutation Research 326 (1995) 17-27

doxorubicin deletion endpoints between positions 7 and 8 or 8 and 9 in the repeat region flanking this deletion. However, including the endpoints of this deletion in either position maintains the general distribution of deletion endpoints depicted in Fig. 4. Distinction between these or other possibilities cannot be discerned in deletions with flanking direct repeats. These studies provide evidence that biological effects can be directed by the DNA sequence specificity of drug binding and the associated alterations in local DNA structure. Specifically, comparison of doxorubicin-induced deletion endpoints with doxorubicin-induced modulation of DNase I cleavage sites suggests that doxorubicin binding specificities may underlie DNA sequence specific mutation events in E. culi. These results provide an impetus for continued work on the design of drugs directed to biologically important DNA targets. In addition, strategies for minimizing the toxic effects of drug-induced mutation may be tested. For example, combining low concentrations of topoisomerase II interactive agents with different DNA sequence specificities may help to minimize focused mutation in genes responsible for secondary leukemia induction (Negrini et al., 1993). Future studies will elucidate the mechanisms mediating the sequence specificity of deletion induction by DNA intercalating drugs.

Acknowledgements

This work was supported by grant ROlCA52683 from the NIH (W.D.S. and M.L.V.), NC1 Clinical Investigator Award K08 CA01616 (R.D.A.) and in its initial stages by a grant from Glaxo, Inc. Additional support came from an NC1 Cancer Center Core Grant to the Case Western Reserve Cancer Center (No. P3OCA 4370301).

References Anderson, R.D., M.L. Veigl, J. Baxter and W.D. Sedwick (1991) DNA sequence specificity of doxorubicin-induced

mutational damage in uurB_ Escherichia coli, Cancer Res. 51,3930-3937. Anderson, R.D., C.-Y. Bao, D.T. Minnick, J. Baxter, M.L. Veigl and W.D. Sedwick (1993a) Sequencing of doublestranded DNA polymerase chain reaction products for mutation analysis, Mutation Res., 288, 181-185. Anderson, R.D., M.L. Veigl, J. Baxter and W.D. Sedwick (1993b) Excision repair reduces doxorubicin-induced genotoxicity, Mutation Res., 294, 215-222. Arcamone. F. (1981) Doxorubicin Anticancer Antibiotics, Academic Press, New York. Bachur, N.R., S.L. Gordon, M.U. Gee and H. Koa (1979) NADPH cytochrome P-450 reductase activation of quinone anticancer agents to free radicals, Proc. Nat]. Acad. Sci. USA, 76, 954-957. Bailly, C., A.W. Cuthbert, D. Gentle, M.R. Knowles and M.J. Waring (1993) Sequence-selective binding of amiloride to DNA, Biochemistry, 32, 2514-2524. Britt, M., F. Zunino and J.B. Chaires (1986) The interaction of the beta-anomer of doxorubicin with B and Z DNA, Mol. Pharmacol., 29, 74-80. Cebula, T.A. (1986) Genetic and physiological modulation of anthracycline-induced mutagenesis in Salmonella typhimurium, Environ. Mutagen., 8, 672-692. Chaires, J.B. (1985a) Thermodynamics of the daunomycinDNA interaction: Ionic strength dependence of the enthalpy and entropy, Biopolymers, 24, 403-491. Chaires, J.B. (1985bJ Long-range allosteric effects on the B to Z equilibrium by daunomycin. Biochemistry, 24, 74797486. Chaires, J.B., N. Dattagupta and D.M. Crothers (1982) Studies on the interaction of anthracycline antibiotics and deoxyribonucleic acid: Equilibrium binding studies on interaction of daunomycin with deoxyribonucleic acid, Biochemistry, 21, 3933-3940. Chaires, J.B., K.R. Fox, J.E. Herrera, M. Britt and M.J. Waring (1987) Site and sequence specificity of the daunomycin-DNA interaction, Biochemistry, 26, 8227-8236. Chaires, J.B., J.E. Herrera and M.J. Waring (1990) Preferential binding of daunomycin to 5’-ATCG and 5’-ATGC sequences revealed by footprinting titration experiments, Biochemistry, 29, 6145-6153. Drew, H.R. and A.A. Travers (1984) DNA structural variations in the E. coli tyrT promoter, Cell, 37, 491-502. Farabaugh, P.J., U. Schmeissner, M. Hofer and J.H. Miller (1978) Genetic studies of the lac repressor. VII. On the molecular nature of spontaneous hotspots in the lacl gene of Escherichia coli, J. Mol. Biol., 126, 847-863. Fox, K.R. and M.J. Waring (1984) DNA structural variations produced by actinomycin and distamycin as revealed by DNase I footprinting, Nucleic Acids Res., 15, 491-507. Gelvan, D. and A. Samuni (1986) Cellular targets of adriamycin-induced damage in Escherichiu co&, B&hem. Pharmacol., 35, 3267-3275. Goldenberg, G.J., H. Wang and G.W. Blair (1986) Resistance to adriamycin: Relationship of cytotoxicity to drug uptake and DNA single- and double-strand breakage in cloned

W.D. Sedwick et al. /Mutation Research 326 (1995) 17-27

cell lines of adriamycin-sensitive and -resistant P388 leukemia, Cancer Res., 46, 2978-2983. Halliday, J.A. (19921 Mechanisms of Spontaneous Mutations in Escherichia coli, Doctoral Thesis, York University, Toronto. Halliday, J.A. and B.W. Glickman (19911 Mechanisms of spontaneous mutation in DNA repair-proficient Escherichia coli, Mutation Res., 250, 28-44. Hogan, M.E., M.W. Roberson and R.H. Austin (19891 DNA flexibility variation may dominate DNase I cleavage, Proc. Natl. Acad. Sci. USA, 86, 9273-9277. Imlay, J.A. and S. Linn (1987) DNA damage and oxygen radical toxicity, Science, 240, 1302-1309. Krishramoorthy, C.R., S.F. Yen, J.C. Smith, J.W. Lown and W.D. Wilson (19861 Stopped-flow kinetic analysis of the interaction of anthraquinone anti-cancer drugs with calf thymus DNA, poly[d(G-C)].poly[d(G-Q] and polyId(A-Tl] +polyfd(A-Tl], Biochemistry, 255933-5940. Kunkel, T.A. and P.S. Alexander (19861 The base substitution fidelity of eucaryotic DNA polymerases. Mispairing frequencies, site preferences, insertion preferences and base substitution by dislocation, J. Biol. Chem., 261, 160-166. Manfait, M., A.J. Alix, P. Jeanneson, J.C. Jardillier and T. Theophanides (1982) Interaction of adriamycin with DNA as studied by resonance Raman spectroscopy, Nucleic Acids Res., 10, 3803-3816. Moore, M.H., W.N. Hunter, B. Langlois dEstaintot and 0. Kennard (19891 DNA-drug interactions. The crystal structure of d(CGATCG) complexed with daunomycin, J. Mol. Biol., 206, 693-705. Negrini, M., CA. Felix, C. Martin, B.J. Lange, T. Nakamura, E. Canaani and C.M. Croce (1993) Potential topoisomerase II DNA-binding sites at the breakpoints of a t(9;ll) chromosome translocation in acute myeloid leukemia, Cancer Res., 53,4489-4492. Neidle, S., L.H. Pearl and J.F. Skelly (1987) DNA structure and perturbation by drug binding, Biochem. J., 243, 1-13. Nelson, E.M., K.M. Tewey and L.F. Liu (1984) Mechanism of

27

antitumor drug action: Poisoning of mammalian DNA topoisomerase II on DNA by 4’-(9-acridinylaminolmethanesulfon-m-anisidide, Proc. Nab. Acad. Sci. USA, 81, 1361-1365. Reinert, ICE. (19831 Anthracycline-binding induced DNA stiffening, bending and elongation; stereochemical implications from viscometric investigations, Nucleic Acids Res., 11,3411-3430. Ross, W.E., D. Glaubiger and K.W. Kohn (1979) Qualitative and quantitative aspects of intercalator-induced DNA strand breaks, Biochim. Biophys. Acta, 562, 41-50. Sambrook, J., E.F. Fritsch and T. Maniatis (19891 Molecular Cloning: A Laboratory Manual, Cold Spring Laboratory Press, Cold Spring Harbor, NY. Sancar A. and W.D. Rupp (19831 A novel repair enzyme: UvrABC excision nuclease of Escherichia cofi cuts a DNA strand on both sides of the damaged region, Cell, 33, 249-260. Suck, D. and C. Oefner (1986) Structure of DNase I at 2.0 A resolution suggests a mechanism for binding to and cutting DNA, Nature, 321,620-625. Tewey, K.M., T.C. Rowe, L. Yang, B.D. Halligan and L.F. Liu (1984) Adriamycin-induced DNA damage mediated by mammalian DNA topoisomerase II, Science, 226,466-468. Trist, H. and D.R. Phillips (1989) In vitro transcription analysis of the role of flanking sequence on the DNA sequence specificity of adriamycin, Nucleic Acids Res., 17, 36733688. Vedaldi, D., F. Dall’Acqua, S. Caffieri and G. Rodighiero (1982) Sequence specificity in DNA for the interaction with adriamycin or daunomycin, Farmaco, 37, 571-581. Wang, A.H.-J., G. Ughetto, G.J. Quigley and A. Rich (1987) Interactions between an anthracycline antibiotic and DNA: Molecular structure of daunomycin complexed to d(CpGpTpApCpG) at 1.2-A resolution, Biochemistry, 26, 1152-1163. Wells, R.D. (1988) Unusual DNA structures, J. Biol. Chem., 263, 1095-1098.