Chem. Res. Toxicol. 1994, 7, 666-672
666
DNA-DNA Interstrand Cross-Linkingby 2,5-Bis(l-aziridinyl)-3,6-bis( carbethoxyamino)1,Cbenzoquinone: Covalent Structure of the dG-to-dG Cross-Links in Calf Thymus DNA and a Synthetic DNA Duplex Stephen C. Alley and Paul B. Hopkins* Department of Chemistry, University of Washington, Seattle, Washington 98195 Received May 10, 1994@ The products of the alkylation reaction of reduced 2,5-bis(l-aziridinyl)-3,6-bis(carbethoxyamino)-1,4-benzoquinone(AZQ, la)with duplex DNA were studied using calf thymus DNA and a synthetic oligodeoxynucleotide. Reaction of calf thymus DNA with a mixture of AZQ and ascorbic acid followed by enzymatic digestion of the sugar phosphate backbone afforded numerous AZQ-derived products including substances identified as monoadducts of AZQ with both dG and dA (with the former in greater abundance) and two diadducts, as would be expected for intrastrand or interstrand cross-links, with one containing two dG residues per AZQ and the other one each of dG and dA (with the former adduct in greater abundance). The nucleotide connectivity and covalent structure of the dG-to-dG interstrand cross-links were studied in greater detail using a synthetic DNA duplex containing the nucleotide sequence B’-d(GGG€CC), where it appeared that the predominant interstrand cross-links bridged dG residues on opposite strands and were separated by two intervening base pairs [5‘-d(GNNC)l. The covalent structure of this lesion was tentatively identified as 2b,in which the N7 atoms of two dG residues have been alkylated by the aziridine functions of AZQ, based upon the results of piperidine fragmentation and characterization of the enzymatic and acidic hydrolysates of the cross-linked DNA.
Introduction
bis(ethy1amino) alcohol resulting from the hydrolysis of AZQ was found to be inactive against L1210 murine The bifunctional alkylating agents constitute an imleukemia cells (12). In vitro studies of DNA-DNA crossportant subset of the clinically useful antitumor sublinking by AZQ have demonstrated that AZQ must be in stances, including such members as cisplatin, mitomycin its reduced form in order to form DNA interstrand crossC, the nitrosoureas, and the nitrogen mustards (1). Some links efficiently (13-15). fifty years have now elapsed since the discovery of the The covalent structure of the AZQ-derived DNA-DNA clinical efficacy of the first member of this class, the interstrand cross-link is unknown. Hydrolysis of AZQ nitrogen mustard mechlorethamine (2). During this in acidic and neutral aqueous solution results in scission time, many structural variants of mechlorethamine have of both aziridine rings, affording the corresponding bisbeen synthesized and tested as potential antitumor [(2-hydroxyethyl)aminol-substitutedquinone (16). Preagents (3-5). Among these are the diaziridinyl quinones liminary evidence favoring analogous chemistry of the (6, 7). The most promising member of this family is 2,5reduced form of AZQ has been reported (17). Postlabeling bis(l-aziridinyl)-3,6-bis(carbethoxyamino~-l,4-benzoqui- experiments on deoxynucleotides revealed that AZQ can none (AZQ,’ la). A recent phase I11 clinical trial has form abundant adducts with 3’-dAMP, 3’-dCMP, and 3’established AZQ as equal to BCNU in efficacy against dGMP. With calf thymus DNA, the majority of the AZQ human brain tumors (8). adducts were attributed to reaction with deoxyguanosine As with many other bifunctional alkylating agents residues (18). The covalent structures of these adducts which possess antitumor activity, DNA is believed to be were not established. the biologically important target of AZQ. AZQ has been We report herein on the covalent structure of dG-toshown to cause DNA strand scission and to form DNAdG, DNA-DNA interstrand cross-links formed in duplex DNA interstrand cross-links ( 9 , lO),as well as to produce DNA by the reduced form of AZQ. This study follows on free radicals (11)in vivo. DNA strand scission and free recent studies involving the simplest of the diaziridinyl radical formation did not correlate with cell death, but quinones (DZQ, lb). This substance, whose high toxicity DNA-DNA cross-linking did, implicating cross-linking has precluded its use as an antitumor agent (191, was as the toxic event. The aziridine moieties of AZQ have shown t o produce in DNA dG-to-dG, interstrand crossbeen shown t o be critical for its biological activity, as the links possessing the substructure 2a (20). The nucleotide sequence preferences of these interstrand cross-links for DZQ in the oxidized form were shown to be 5’-d(GN,C), @Abstractpublished in Advance ACS Abstracts, August 15, 1994. n = 0, 1, and 2 in short GC-rich (20) and AT-rich (21) Abbreviations: AZQ,2,5-bis(l-azridinyl)-3,6-bis(carbetho~-o)1,4-benzoquinone;BCNU, N,iV’-bis(Z-chloroethyl)-N-nitrosourea;DZQ, oligodeoxynucleotideduplexes. DZQ in the reduced form 2,5-bis(l-aziridinyl)-l,4-benzoquinone; ESIMS, electrospray ionization has been shown to cross-link DNA at the sequences 5’mass spectrometry; DPAGE, denaturing polyacrylamide gel electrophoresis. d(GN,C), n = 0 and 2 (22). The carbethoxyamino groups ~~~~~~
Q893-228x/94I27Q7-Q666$04.5QlQ 0 1994 American Chemical Society
DNA-DNA Interstrand Cross-Linking by AZQ of AZQ modulate its reactivity, greatly increasing its halflife in aqueous solution relative to DZQ [ca. 250 min for AZQ (26) and ca. 4 min for DZQ (23) at pH 41. These groups also increase the steric bulk around the aziridine ring, which may influence the nucleotide sequence preference of its DNA interstrand cross-linking reaction. Herein, we present evidence that, in calf thymus DNA, reduced AZQ cross-links deoxyguanosine residues. In short oligodeoxynucleotide duplexes, we show that this lesion can be formed as an interstrand cross-link at the sequence 5'-d(GGGCCC), with the cross-links likely bridging deoxyguanosine residues on opposite stands and separated by two intervening base pairs [B'-d(GNNC)I. Strong evidence indicates that the covalent structure of the AZQ-derived DNA interstrand cross-link is 2b,the analog of 2a formed in DNA by DZQ.
N7
LN*RR
0
l a AZQ l b DZQ
RENHCOOCH~CH) R=H
NHZ
W Z
20 DZQ R=H Pb AZQ
RrNHCOOCH2CH3
Experimental Section Materials and Methods. Materials and methods were as described previously (20) except for the following: AZQ was obtained from the Drug Synthesis Branch of the National Cancer Institute (Bethesda, MD); bacterial alkaline phosphatase (Escherichia coli) and calf thymus DNA were obtained from Sigma (St. Louis, MO). Loading buffer was 90% aqueous deionized formamide containing 10 mM Tris/Tris.HCl (pH 7.5), 0.1%xylene cyanol, and 0.1 mM NazEDTA. Hydrolysis buffer was 50 mM Tridhis-HCl (pH 8.5) and 10 mM MgC12. HPLC was performed on an Alltech, 5-pm, C18, 250 mm x 4.6 mm Econosphere column using an SSI 200Bl220B dual-pump system with an SSI controller and sequential SSI 500 UVlvis (output to a Linear 255/M recorder and an HP 3390A electronic integrator) and Rainen UV-D I1 (output to a Linear 156 recorder and an HP 3390A electronic integrator) detectors. Solvent gradients were run a t 1 mumin. Gradient A solvent A, 10 mM ammonium acetate; solvent B, 100%acetonitrile; isocratic 92%A for 7 min, a 13-min linear gradient to 70%A, a 10-min linear gradient to 60%A, and a 10-minlinear gradient to initial conditions. Gradient B: solvent A, 100 mM ammonium acetate; solvent B, 50%acetonitrile; isocratic 97%A for 12 min, an 8-min linear gradient to 70%A, and a 10-min linear gradient to initial conditions. Isolation and Identification of Lesions from AZQTreated Calf Thymus DNA. Calf thymus DNA (3.4 OD; 0.32 pmol of deoxynucleotides;final concentration 4 mM deoxynucleotides) was dissolved in 28 pL of HzO and treated sequentially with 8 pL of 1 M sodium acetate buffer (pH 4.0),4pL of 50 mM AZQ (stock solution in MezSO; 0.20 pmol; final concentration 2.5 mM), and 40 pL of 20 mM ascorbic acid (0.80 pmol; final concentration 10 mM). After 3 h a t 25 "C, the cross-linking mixture was ethanol precipitated, the supernatant removed, and the pellet dried in vacuo. To the pellet were added 42 pL of water and 10 p L of l o x hydrolysis buffer. This mixture was sonicated for 15 min at 25 "C to redissolve the pellet. Once
Chem. Res. Toxicol., Vol. 7,No.5, 1994 667 dissolved, 100 units of DNase I, 50 units of DNase 11, 10 units of snake venom phosphodiesterase I, 10 units of calf intestinal alkaline phosphatase, and 10 units of bacterial alkaline phosphatase were added, bringing the total volume to 100 pL. After 18 h a t 25 "C, the hydrolysate was separated by HPLC using gradient A. Peaks a t 20.0, 20.9, 21.4, 22.3, 24.1, and 24.9 min were collected and directly analyzed by ESIMS. The following ions were observed: 20.0 min: mlz 534 (guanine monoadduct), 667 (guanine-guanine cross-link); 20.9 min: mlz 534, 651 (guanine-adenine cross-link), 667; 21.4 min: mlz 518 (adenine monoadduct), 534, 667; 22.3 min: mlz 518, 534, 651, 667; 24.1 min: mlz 518, 534; 24.9 min: d z 518, 552/554 (guanine monoadduct opened by chloride). Alternatively, the eluent between 16 and 26 min was collected and directly analyzed by ESIMS. No additional ions were observed beyond those described above which corresponded to the full mass of AZQ and one or two nucleic acid bases. General Conditions for Preparation, Isolation, and Characterization of Interstrand Cross-LinkedDNA Duplexes. DNA was synthesized, purified, and radiolabeled as previously described (20). For analytical-scale reactions, radiolabeled duplex DNA (0.1 OD; 0.25 nmol; 10 pM final concentration) was dissolved in 19.5 pL of HzO and treated sequentially with 2.5 p L of 1 M sodium acetate buffer (pH 4.0), 0.5 pL of 50 mM AZQ (stock solution in MezSO; 25 nmol; 1 mM final concentration), and 2.5 pL of 20 mM ascorbic acid (50 m o l ; 2 mM final concentration). After 3 h at 25 "C, the cross-linking mixture was ethanol precipitated, the supernatant removed, and the pellet dried in vacuo. The pellet was dissolved in 10 p L of loading buffer and analyzed by denaturing polyacrylamide gel electrophoresis (DPAGE) as described previously (20).Autoradiography was used to visualize the reaction products, and phosphorimagery was used to determine the yield of cross-linked products. For preparative-scale reactions, unradiolabeled DNA (55 OD; 0.14 pmol; 0.1 mM final concentration) was dissolved in 600 p L of water and sequentially treated with 100 pL of 1 M sodium acetate buffer (pH 4.01, 50 pL of 50 mM AZQ (stock solution in MeZSO; 2.5 pmol; 2.5 mM final concentration), and 250 p L of 20 mM ascorbic acid (5.0 pmol; 5.0 mM final concentration). After 3 h at 25 "C, the cross-linking mixture was split into 4 samples of 250 pL each and ethanol precipitated, the supernatant removed, and the pellet dried in vacuo. Pelleted DNA was redissolved in 30 pL of 98% aqueous deionized formamide containing 10 mM NazEDTA and subjected to DPAGE. Isolated cross-linked DNA possessed a UV spectrum with maximum absorbances a t 260 and 348 nm. Piperidine Cleavage of Radiolabeled, Cross-Linked DNA. Previously isolated, radiolabeled, cross-linked DNA was dissolved in 100 pL of 1 M aqueous piperidine and heated to 90 "C in a capped microfuge tube for 45 min. After concentration to dryness, the mixture was sequentially concentrated to dryness twice from 20 pL of water, dissolved in 7 pL of loading buffer, and analyzed by DPAGE as previously described (20). The gel was dried onto filter paper and quantified by phosphorimagery. Enzymatic Hydrolysis of Cross-LinkedDNA. Isolated cross-linked DNA GGGCCC (0.5 OD) was hydrolyzed using 100 pL of hydrolysis buffer, 100 units of DNase I, 50 units of DNase 11, 10 units of snake venom phosphodiesterase I, and 10 units of calf intestinal alkaline phosphatase. For the cross-link of higher electrophoretic mobility, 10 units of bacterial alkaline phosphatase was added. After 18 h a t 25 "C,the hydrolysate was separated by HPLC using gradient A. The peak a t 22 min was collected and concentrated to dryness, the residue redissolved in water, and the UV spectrum measured. On a subsequent run, the peak was collected and directly analyzed by ESIMS. Alternatively, the eluent between 19 and 25 min was collected and directly analyzed by ESIMS. These measurements were all consistent with its assignment as structure 3. W 258, 348 nm; ESIMS (120-V inlet voltage, 4400-V needle voltage, 3 pL/min, sum of 50 scans): mlz 667 (M2++ H+ - 2 deoxyribosyl+;37%of base peak), 516 (M2+- deoxyribosyl+ -
Alley and Hopkins
668 Chem. Res. Toxicol., Vol. 7, No. 5, 1994
Table 1. Presumed Identity of Selected Ions in ESIMS Analysis of AZQ-Calf Thymus DNA Adducts ,
CHSCHZOCONHV
N
0
ion (mlz)
Figure 1. HPLC profile of the enzymatic hydrolysate of AZQtreated calf thymus DNA. The six peaks which were collected for ESIMS analysis are indicated. Retention time increases to the right.
deoxyguanosine; 5%), 365 (M2++ H+ - 2 deoxyguanosine; 28%), 152 (Gua + H+; 100%). Isolation of Formamide-Derivatized 7-(2-Aminoethyl)guanine from Cross-Linked DNA. Cross-linked DNA GGGCCC was hydrolyzed and derivatized as previously described (20). The formylated hydrolysate was separated by HPLC using gradient B. Previously prepared authentic 5b (20)was shown to coelute with 5b from hydrolyzed, formylated, cross-linked DNA GGGCCC in gradient B.
Results Alkylation of Calf Thymus DNA. No lesion in which two components of DNA are attached to a single AZQ residue has been previously reported. Postlabeling has shown that deoxyguanosine is the likely target of AZQ (18). However, these lesions were not structurally characterized. To investigate these lesions further, calf thymus DNA (4 mM deoxynucleotides) was exposed to 2.5 mM AZQ and 10 mM ascorbic acid in 100 mM sodium acetate buffer at pH 4.0 for 3 h at 25 “C. Hydrolysis of the sugar-phosphate backbone by a mixture of DNase I, DNase 11, phosphodiesterase I, calf intestinal alkaline phosphatase, and bacterial alkaline phosphatase and analysis by reverse-phase HPLC (Figure 1)revealed the four common deoxynucleosides as well as at least 10 more strongly retained, presumably more hydrophobic substances. These latter substances all possessed appreciable absorbance at 348 nm, indicating that they likely contain the quinone function of AZQ. Six of these peaks, which in total constituted 78% of the total stronglyretained 348-nm-absorbing material, were individually collected and analyzed by electrospray ionization mass spectrometry (ESIMS). These samples afforded ions which corresponded to the full mass of AZQ as well as one or two DNA bases. The glycosyl linkages of these lesions appeared to have solvolyzed during the ionization process. All but one of the ions appeared in a t least two HPLC peaks, suggesting that the sugar-phosphate backbone was not completely hydrolyzed, increasing the apparent complexity of the mixture due to the presence of lesions in more than one state of oligomerization. Guanine and adenine monoadducts as well as guanineguanine and guanine-adenine cross-links were observed (see Tables 1and 2). Further inspection of these spectra gave no evidence of abundant ions of masses predicted for corresponding monoadducts of the other two nucleic acid bases nor for any of the other eight possible nucleic acid diadducts. Because this complex mixture was not completely separated, and authentic standards are not presently available, it is not possible based on these data to
W
Y
H
Y
X
adeninyl
518 534 5521554 651 667
Time
NHCOOCHZCH,
guaninyl
OH OH
guaninyl
c1
guaninyl
adeninyl
guaninyl
guaninyl
Table 2. Relative Intensity (%) of Selected Ions in ESIMS of AZQ-Calf Thymus DNA Adductsu ion (mlz) HPLCpeak 518 534 5521554 651 667 1 70 30 2 93 2 5 3 18 59 23 4 24 35 14 27 5
6 total adductsb
90
10
9 39
91 28
16 a Total for each row normalized to 100%. Combined late11
6
eluting, 348-nm-absorbing material from HPLC. See text. quantitate the mixture unequivocally. However, based upon the reasonable assumption that these adducts all possess comparable extinction coefficients at 348 nm (i.e., that total HPLC peak intensity is proportional to the mole fraction of the 348-nm-absorbing material in each peak) and the more speculative assumption that the ion current in the ESIMS for a given adduct is proportional to the mole fraction of each adduct in that sample, tentative conclusions can be drawn. This model suggests that guanine-only derived monoadducts and cross-links are formed in preference to adenine-containing monoadducts and cross-links. Inspection of the data in Table 2 reveals that only one of the six HPLC peaks contained predominantly an adenine-derived lesion, the monoadduct of adenine (peak 5). Two other peaks contained nonnegligible amounts of adenine-derived materials, but guanine-only adducts were in the majority. The remaining three peaks contained almost exclusively guanineonly adducts. The same conclusion was drawn following an experiment in which the total strongly-retained, 348nm-absorbing material from the HPLC was collected as a single fraction and analyzed by ESIMS. Here the guanine-only ions outnumbered the adenine-containing ions by 4.51. Consistent with previous postlabeling studies (18), we find that, in the alkylation of calf thymus DNA by AZQ, deoxyguanosine residues are the major sites of alkylation in both monoadducts and cross-links. Nucleotide Connectivity of Cross-Links. A short synthetic DNA duplex was employed to investigate the nucleotide connectivity of the AZQ-derived DNA interstrand cross-links. Because the calf thymus DNA studies showed deoxyguanosine to be the predominant site of alkylation, the selected duplex was deoxyguanosine-rich and contained a nucleotide sequence which had previously been shown to be cross-linked by DZQ. This selfcomplementary 12-mer, 5’-d(TATGGGCCCATA), is referred to hereafter as “GGGCCC”. This DNA (10 pM duplex), in either 5’- or 3’-32P-radiolabeled form, was
Chem. Res. Toxicol., Vol. 7, No. 5, 1994 669
DNA-DNA Interstrand Cross-Linking by AZQ AzQ Ascorbicacid
+
+ -
+ +
123
(A) ’isTATGGGCCCATA
Piperidine
1
32PTATGP
ATACCCGGGTATs
Interstrand Cross-links
H20* A
321 123
(e)
532pTATGGGCCCATA
Piperidine
123
5 32pTATp
H20.A
12
32PTATGGP ATACCCGGGTATs 321
Single Strands
Figure 2. DPAGE analysis of DNA interstrand cross-linking reactions using AZQ and ascorbic acid at pH 4.0.
Figure 3. Nucleotide connectivity of interstrand cross-links as defined by cleavage fragment lengths following treatment with hot aqueous piperidine. The numbering scheme for the deoxyguanosine residues used in the text is indicated. (A) More electrophoreticallymobile cross-link;(B) less electrophoretically mobile cross-link.
I, exposed to 1 mM AZQ and 2 mM ascorbic acid in 100 mM sodium acetate buffer a t pH 4.0 for 3 h a t 25 “C. Following ethanol precipitation, the products derived from DNA-DNA interstrand cross-linking were separated from non-cross-linked products by denaturing polyacrylamide gel electrophoresis (DPAGE), visualized by autoradiography, and quantitated by phosphorimagery (Figure 2). DPAGE analysis revealed predominantly material with electrophoretic mobility comparableto unalkylated single strands. This material, which likely contains some monoadducts, was not further studied. In addition to these substances with high electrophoretic mobility, there were two bands with mobility about half that of unalkylated single strands present in a 1:l ratio and ca. 1% yield each. We assign these products, based on both their electrophoretic mobility and their lability toward hot aqueous piperidine, as DNA-DNA interstrand crosslinks. Piperidine treatment of these interstrand crosslinked products resulted in nearly quantitative strand cleavage, releasing radiolabeled products whose electrophoretic mobilities corresponded to the products of the Maxam-Gilbert G reaction (24). This is consistent with the alkylation of N7 atoms of deoxyguanosine residues, further evidence of which is presented below. The identities of the fragments released following piperidine treatment provide information about the sequence specificity of the cross-linking reaction. The more electrophoretically mobile of the two interstrand cross-linked products from DNA GGGCCC (in both 5’and 3’-labeled forms) yielded cleavage in the majority a t G2 (85%of the total cleavage; Figure 3), with the balance being minor cleavage a t G1 and G3 in a 1:l ratio. This result is exclusively consistent with a predominant G2to-G2, 5’-d(GNNC) cross-link. The less electrophoretically mobile interstrand cross-linked product from DNA GGGCCC (in both 5’- and 3’-labeled forms) afforded cleavage in the majority a t G1 and G3 in a 1:lratio (90% of the total cleavage) with the remainder at G2. The simplest interpretation of these data is that, like the more electrophoretically mobile cross-link, this interstrand cross-link is formed at the sequence 5’-d(GNNC), likely connecting G1 to G3. However, the more complex possibility of the presence, in some amount, of a 1:lmixture of G3-to-G3, 5’-d(GC) and G1-to-G1, 5’-d(GNNNNC)
200
250
300
350
400
450
Wavelength (nm)
Figure 4. Ultraviolet spectrum of AZQ interstrandcross-linked DNA GGGCCC isolated by DPAGE.
cross-links cannot be excluded. The minor cleavage products from each cross-linked DNA corresponded to the major cleavage products from the other, suggesting that the minor cleavage products derive from incomplete resolution of the cross-linked DNAs by DPAGE. Covalent Structure of the Cross-Link. The covalent structure of the AZQ-derived DNA interstrand crosslink was studied by techniques previously developed in the study of the DZQ-derived DNA interstrand cross-link. The U V spectrum of isolated interstrand cross-linked DNA GGGCCC possessed an absorbance maximum a t 260 nm due to the DNA as well as a maximum at 348 nm, presumably due to the quinone chromophore of AZQ (Figure 4). This absorbance is not coincident with the maximal absorbance of AZQ (344 nm), but has been redshifted. A red shift was also observed when DZQ was hydrolyzed to its corresponding bis(ethy1amino) alcohol (23)or exposed to DNA under acidic or neutral conditions (25,26). The less electrophoretically mobile interstrand crosslinked product of DNA GGGCCC, previously isolated by DPAGE, was treated with a mixture of DNase I, DNase 11, phosphodiesterase I, and calf intestinal alkaline phosphatase (the cross-link with higher electrophoretic mobility required bacterial alkaline phosphatase in addition to the other enzymes), and the hydrolysate was analyzed by reverse-phase HPLC (Figure 5). In addition to the four common deoxynucleosides, several more hydrophobic substances were observed. The most abundant of these substances was collected and analyzed by W spectroscopy and electrospray ionization mass spectrometry. The W spectrum of this substance possessed absorbance maxima a t 258 and 348 nm, consistent with N7 alkylated deoxyguanosine and 2,5-bis(ethylamino)-
670 Chem. Res. Toxicol., Vol. 7, No. 5, 1994
Alley and Hopkins
dG
dCdT dA
I
I
Time
Figure 5. HPLC profile (detection at 260 nm, upper, and 348 nm, lower) of the enzymatic hydrolysate of AZQ interstrand cross-linked DNA GGGCCC. The peak containing substance 3 is indicated. Retention time increases to the right.
220
270
320
370
420
470
Wavelength (nm)
4
R=NHCOOCH,CH:,
was treated with acetic-formic anhydride to form the formamide derivative 5b. Analysis by HPLC revealed the four common DNA bases as well as an extra peak which coeluted with a previously prepared authentic sample of Sb (20).Both of the DNA interstrand crosslink bands of DNA GGGCCC gave similar results. These data together strongly support the conclusion that the dG-to-dG interstrand cross-link formed by reduced AZQ in this synthetic oligodeoxynucleotide duplex possesses the substructure 2b.
Figure 6. Ultraviolet spectrum of substance 3 isolated by HPLC from an enzymatic hydrolysate of AZQ interstrand crosslinked DNA GGGCCC.
3,6-bis(carbethoxyamino)-l,Cbenzoquinone moieties, respectively, as shown in 3 (Figure 6). The electrospray ionization mass spectrum of this substance possessed a prominant ion at mlz 667, consistent with solvolysis of the glycosyl linkages in 3 to form protonated 4. The sodium (mlz 689) and potassium (mlz 705) ion adducts of 4, ions corresponding to loss of one (mlz 516) and two (mlz 365) guanines from protonated 4, and protonated guanine (mlz 152) were also present. These fragments were analogous to those found for the DZQ-derived DNA interstrand cross-link (20)and demonstrated that the full masses of AZQ and two guanines are contained in the DNA interstrand cross-link, as shown in 4. Similar results were obtained for the interstrand cross-linked material possessing higher electrophoretic mobility. During later experiments, all of the strongly retained peaks were collected in one fraction and analyzed by ESIMS. The spectrum contained the same ions as were observed above for the single peak, but these ions were reduced in intensity due to dilution. We speculate that the other peaks observed in the HPLC chromatogram contain substances less completely enzymatically digested than 3, but that all of these substances contain 2b as the nucleus of the DNA interstrand cross-link. In order to demonstrate conclusively that the DNA interstrand cross-link was formed at the N7 position of deoxyguanosine residues, the vinylogous amide functions of 2b were hydrolyzed in acid to release 742-aminoethy1)guanine (5a). Previously isolated, cross-linked DNA GGGCCC was treated with 88% formic acid for 2 h at 140 "C. To avoid coelution of 5a with thymine in the subsequent HPLC analysis, the formic acid hydrolysate
2b
R=NHCOOCH#~J
I ) H3O' 2) (CWO(COCHd, Wr c 3) NH3
H
N
HSI A
N
5
iHR
N
'N'
mR=H 5b
LCHO
Discussion We provide herein further structural information on DNA adducts of AZQ, the most clinically useful member of the diaziridinyl quinone family of antitumor substances, with particular attention to the lesion responsible for the DNA-DNA interstrand cross-link. HPLC analysis of enzymatic hydrolysates of AZQ-treated calf thymus DNA afforded several substances whose absorbance spectra were consistent with incorporation of the quinone function of AZQ. Electrospray ionization mass spectra revealed that this hydrolysate contained two monoadducts, with adenine or guanine and AZQ in a 1:l molar ratio, and two diadducts, one with guanine and AZQ in a 2:l molar ratio and one with adenine, guanine, and AZQ in a 1:l:l molar ratio. The combined HPLC data and the electrospray ionization mass spectra suggested that some 80% of these lesions contained exclusively guanine as the nucleic acid component of the lesion. A synthetic oligodeoxynucleotide duplex containing the deoxynucleotide sequence B'-d(GGGCCC) afforded inter-
DNA-DNA Interstrand Cross-Linking by AZQ
Figure 7. Molecular model of B-form DNA 5’-d(TATGGGCCCA), viewed from the major groove, generated by the Biopolymer module of Insight I1 (v. 2.3.0) on a Personal IRIS 4D-25 workstation. The N7 atoms of deoxyguanosine residues are labeled as in Figure 3.
strand cross-linked products upon treatment with reduced AZQ. The simplest interpretation of strand scission assays on these interstrand cross-linked substances was that the linkages bridged two deoxyguanosine residues predominantly a t the sequence 5’-d(GNNC). Enzymatic hydrolysis of the isolated, interstrand crosslinked oligodeoxynucleotide duplex afforded a substance with W and mass spectra consistent with structure 3. Formic acid hydrolysis followed by treatment with aceticformic anhydride afforded formamide derivative 5b, strongly suggesting the alkylation sites on guanine (N7) and AZQ (aziridinyl carbon). These data establish structural analogy in the chemistries of AZQ and DZQ in their interstrand cross-linking reactions with duplex DNA. With DZQ, it has been demonstrated that these linkages preferentially bridge deoxyguanosineson opposite strands which possess a 5’offset relative to one another. In the present case, we show only that the same linkage pattern can be formed by AZQ. Inspection of three-dimensional models of righthanded, duplex DNA reveals that cross-linking of sites such as the N7 atoms of closely spaced deoxyguanosine residues with a 5’-offset causes the bridging tether to trace the path of a left-handed helix, the shortest path between the two sites. The dimensions of the bridging tether derived from AZQ appear capable of bridging the observed sites on B-DNA without any mandatory structural distortions, a situation which stands in contrast to that observed with nitrogen mustard (27) and cisplatin (28, 291, agents which also bridge the N7 atoms of deoxyguanosine residues on opposite strands. We comment in closing on an interesting feature of the nucleotide sequence specificity of the interstrand crosslinking reaction reported herein. Figure 7 illustrates the spatial proximity of the nucleophilic N7 atoms of deoxyguanosine residues a t the sequence 5’-d(GGGCCC). The predominant spacing of cross-links observed herein was 5’-d(GNNC), corresponding to linkage of atoms on opposite strands labeled 1and 3, for example. Structural models of AZQ suggest that the bridging atoms can accommodate this spacing without the necessity for substantial distortion of B-DNA. This spacing appears,
Chem. Res. Toxicol., Vol. 7, No. 5, 1994 671 however, not to be universally the preference of AZQ. When a synthetic DNA duplex containing the sequence 5’-d(GGGGC) surrounded by dA and dT residues was treated with AZQ and ascorbic acid, interstrand crosslinks were formed. However, in contrast to the outcome with the 5’-d(GGGCCC)-containing DNA [where crosslinks a t 5’-d(GNNC) apparently predominated], analysis of the piperidine fragmentation products suggested that all possible pairs of deoxyguanosine residues had been interstrand cross-linked [5’-d(GN,C), n = 0-31. One of the products ( n = 0) comprised 5% of the cross-link mixture, with the balance being equally distributed among the other three products ( n = 1-3). Unfortunately, it was not possible to separate these structurally isomeric interstrand cross-linked DNAs, precluding a more thorough analysis of their nucleotide connectivity and covalent structure. How is it possible that two duplex DNAs afford such seemingly disparate connectivity patterns with reduced AZQ? We speculate that monoadducts of reduced AZQ a t deoxyguanosine may preferentially close to form crosslinks a t the sequence 5’-d(GNNC), but that if this sequence is not available, transition states leading to other connectivities are energetically accessible. This explains why all monoadducts a t the sequence 5’d(GGGCCC),where every deoxyguanosine residue is part of a 5’-d(GNNC) sequence, could yield essentially exclusively 5’-d(GNNC) cross-links, while in the DNA containing the subsequence 5’-d(GGGGC), in which only one 5’d(GNNC) is present, other connectivities are found. In other words, non-sequence-selective monoalkylation in the 5’-d(GGGGC)-containing DNA commits AZQ to interstrand cross-links a t sites other than its “preferred” 5’-d(GNNC) sequence. The validity of this speculation remains to be experimentally tested.
Acknowledgment. We thank the NIH for generous financial support (GM 32681). S.C.A. was an NIH predoctoral fellow (GM 08437). P.B.H. was a Cope Scholar. Phosphorimagery was performed by the Phosphorimagery Facility, Markey Molecular Medicine Center, The University of Washington. We thank Mr. H. Huang, Ms. S. M. Rink, and Mr. G. M. Lee for technical advice and Prof. N. H. Andersen for use of the IRIS workstation.
References Pratt, W. B., and Rudden, R. W. (1979) The Anticancer Drugs, Oxford University Press, New York. Gilman, A., and Phillips, F. S. (1946) The biological actions and therapeutic applications of /3-chlorethylamines and sulphides. Science 103,409-415. Moore, M. J. (1991) Clinical pharmacokinetics of cyclophosphamide. Clin. Pharmacokinet. 20, 194-208. Furner, R.L.,and Brown, R. K.(1980) L-Phenylalaninemustard (L-PAM): the first 25 years. Can. Treat. Rep. 64, 559-574. Genka, S., Deutsch, J., Shetty, U. H., Stahle, P. L., John, V., Lieberburg, I. M., Ali-Osman, F., Rapoport, S. I., and Grieg, N. H. (1993) Development of lipophilic anticancer agents for the treatment of brain tumors by the esterification of water-soluble chlorambucil. Clin. Exp. Metastasis 11, 131-140. Domagk, G., and Gauss, W. (1954) A contributionto experimental chemotherapy of tumors. 2. Krebsforsch. 59, 617-622. Khan, A. H.,and Driscoll, J. S. (1976) Potential central nervous system antitumor agents. Aziridinylbenzoquinones. 1. J. Med. Chem. 19,313-317. Schold, S . C.,Jr., Herndon, J. E., Burger, P. C., Halperin, E. C., Vick, N. A., Cairncross, J. G., Macdonald, D. R., Dropcho, E. J., Morawetz, R., Bigmer, D.D., and Mahaley, M. S., Jr. (1993)
672 Chem. Res. Toxicol., Vol. 7, No. 5, 1994 Randomized comparison of diaziquone and carmustine in the treatment of adults with anaplastic glioma. J . Clin. Oncol. 11, 77-83. Szmigiero, L., Erickson, L. C., Ewig, R. A., and Kohn, K. W. (1984) DNA strand scission and cross-linking by diaziridinylbenzoquinone (diaziquone) in human cells and relation to cell killing. Cancer Res. 44,4447-4452. King, C. L., Hittelman, W. N., and Loo, T. L. (1984) Induction of DNA strand breaks and cross-links by 2,5-diaziridinyl-3,6-bis(carboethoxyamino)-1,4-benzoquinone in Chinese hamster ovary cells. Cancer Res. 44,5634-5637. Lea, J. S., Garner, H. J., Butler, J., Hoey, B. M., and Ward, T. H. (1988) The lack of correlation between toxicity and free radical formation of two diaziridinyl benzoquinones. Biochem. Pharmucol. 37, 2023-2025. Egorin, M. J., Fox, B. M., Spiegel, J. F., Gutierrez, P. L., Friedman, R. D., and Bachur, N. R. (1985) Cellular pharmacology in murine and human leukemic cell lines of diaziauone (NSC 182986). Cancer Res. 45, 992-999. (13) King. C. L.. Wone. S.-K.. and Loo. T. L. (1984) Alkylation of DNA by ;he new agticancer agent 3,6-diaziridinyi-2,5-bis(carboethoxyamino)-l,4-benzoquinone(AZQ). Eur. J . Cancer Clin. Oncol. 20, 261-264. (14) Lee, C.-S., Hartley, J . A., Berardini, M. D., Butler, J., Siegel, D., Ross, D., and Gibson, N. W. (1992) Alteration in DNA crosslinking and sequence selectivity of a series of aziridinylbenzoauinones after enzvmatic reduction bv DT-diaDhorase. Biochem&try 31, 3019-3Ok (15) Hartlev. J. A.. Berardini. M., Ponti. M., Gibson, N. W., Thompson, A. S., ?hurston, D. E., Hoey, B. M., and Butler, J. (1991) DNA cross-linking and sequence selectivity of aziridinylbenzoquinones: a unique reaction a t 5’-GC-3’ sequences with 2,5-diaziridinyl-1,4-benzoquinone upon reduction. Biochemistry 30,1171911724. (16) Poochikian, G. K., and Cradock, J. C. (1981) 2,5-Diaziridinyl-3,6bis(carboethoxyamino)-1,4-benzoquinoneI: kinetics in aqueous solution by high-performance liquid chromatography. J.Pharm. Sca. 70, 159-167. (17) Fisher, G. R., Donis, J., and Gutierrez, P. L. (1992) Reductive metabolism of diaziquone (AZQ) in the S9 fraction of MCF-7 cells. Bwchem. Pharmacol. 44,1625-1635. (18) Gupta, R. C., Garg, A., Earley, K., Aganval, S. C., Lambert, G. R., and Neslow, S. (1991) DNA adducts of the antitumor agent diaziquone. Cancer Res. 51, 5198-5204. (19) Domagk, G. (1956) Histological modification of experimental and human tumors after dosage of a cytostatic agent. Deut. Med. Wochenschr. 21, 801-806.
Alley and Hopkins (20) Alley, S. C., Brameld, K B., and Hopkins, P. B. (1994) DNA interstrand cross-linking by 2,5-bis(l-aziridinyl)-1,4-benzoquinone: nucleotide sequence preferences and covalent structure of the dG-to-dG cross-links at 5’- d(GN,C) in synthetic oligonucleotide duplexes. J. Am. Chem. Soc. 116, 2734-2741. (21) Berardini, M. D., Souhami, R. L., Lee, C.-S., Gibson, N. W., Butler, J., and Hartley, J. A. (1993) Two structurally related diaziridinylbenzoquinones preferentially cross-link DNA a t different sites upon reduction with DT-diaphorase. Biochemistry 32,3306-3312. (22) Haworth, I. S., Lee, C.-S., Yuki, M., and Gibson, N. W. (1993) Molecular dynamics simulations provide a structural basis for the experimentally obsellred nucleotide preferences for DNA interstrand cross-links induced by aziridinylbenzoquinones.Biochemistry 32, 12857-12863. (23) Kusai, A., Tanaka, S., and Udea, S. (1981) The stability of carboquone in aqueous solution I. Kinetics and mechanisms of degradation of 2,5- diethylenimino-l,4-benzoquinone in aqueous solution. Chem. Pharm. Bull. 29,3671-3679. (24) Maxam, A. M., and Gilbert, W. (1980) Sequencing end-labeled DNA with base-specific chemical cleavages. Methods Enzymol. 65, 499-560. (25) Lusthof, K. J., De Mol, N. J., Janssen, L. H. M., Verboom, W., and Reinhoudt, D. N. (1989) DNA alkylation and formation of DNA interstrand cross-links by potential antitumour 2,5-bis(laziridinyl)-1,4-benzoquinones. Chem.-Biol. Interact. 70,249-262. (26) Lusthof, K. J., De Mol, N. J., Janssen, L. H. M., Egberink, R. J. M., Verboom, W., and Reinhoudt, D. N. (1990) Covalent binding studies on the “C-labeled antitumour compound 2,5-bis(l-aziridinyl)-1,4-benzoquinone. Involvement of semiquinone radical in binding to DNA, and binding to proteins and bacterial macromolecules in situ. Chem.-Bid. Interact. 76, 193-209. (27) Rink, S. M., Solomon, M. S., Taylor, M. J., Rajur,S. B., McLaughlin, L. W., and Hopkins, P. B. (1993) Covalent structure of a nitrogen mustard-induced DNA interstrand cross-link: an N7to-”I linkage of deoxyguanosine residues at the duplex sequence 5’-d(GNC).J . Am. Chem. SOC.115, 2551-2557. (28) Lemaire, M. A., Schwartz, A,, Rahmouni, A. R., and Leng, M. (1991) Interstrand cross-links are preferentially formed a t the d(GC) sites in the reaction between cis-diamminedichloroplatinum (11) and DNA. Proc. Natl. Acad. Sci. U S A . 88, 1982-1985. (29) Hopkins, P. B., Millard, J . T., Woo, J., Weidner, M. F., Kirchner, J . J., Sigurdsson, S. Th., and Raucher, S. (1991) Sequence preferences of DNA interstrand cross-linking agents: importance of minimal DNA structural reorganization in the cross-linking reactions of mechlorethamine, cisplatin, and mitomycin C. Tetrahedron 47, 2475-2489.
