Redox Pathway Leading to the Alkylation of DNA ... - ACS Publications

14 downloads 0 Views 535KB Size Report
dithiothreitol, the oxidizing agent hydrogen peroxide, or the alkylating agent ... hydrogen peroxide oxidation of Tris [tris(hydroxymethyl)aminomethane] present in ...
1276

J. Med. Chem. 1997, 40, 1276-1286

Redox Pathway Leading to the Alkylation of DNA by the Anthracycline, Antitumor Drugs Adriamycin and Daunomycin Dylan J. Taatjes,† Giorgio Gaudiano,†,‡ Katheryn Resing,† and Tad H. Koch*,† Department of Chemistry and Biochemistry, University of Colorado, Boulder, Colorado 80309-0215, and Istituto di Medicina Sperimentale, CNR, Viale Marx, 15-43, 00137 Rome, Italy Received December 9, 1996X

Reaction of the anthracycline, antitumor drugs adriamycin and daunomycin with the selfcomplementary DNA oligonucleotide GCGCGCGC, (GC)4, in the presence of the reducing agent dithiothreitol, the oxidizing agent hydrogen peroxide, or the alkylating agent formaldehyde gives a similar mixture of DNA-drug adducts. Negative ion electrospray mass spectra indicate that adduct formation involves coupling of the DNA to the anthracycline via a methylene group and that the major adduct is duplex DNA containing two molecules of anthracycline, each bound to a separate strand of the DNA via a methylene group. The source of the methylene group is formaldehyde. A molecular structure with each anthracycline intercalated at a 5′CpG-3′ site and covalently bound from its 3′-amino group to a 2-amino group of a 2′deoxyguanosine nucleotide is proposed based upon spectral data and a relevant crystal structure. The reaction of (GC)4 with the anthracyclines and formaldehyde forms an equilibrium mixture with DNA-drug adducts which is shifted toward free DNA by dilution. The results suggest a pathway to the inhibition of transcription by reductively activated adriamycin and daunomycin. Reductive activation in the presence of oxygen yields hydrogen peroxide; hydrogen peroxide oxidizes constituents in the reaction mixture to formaldehyde; and formaldehyde couples the drug to DNA. In this regard, hydrogen peroxide reacts with adriamycin via Baeyer-Villiger reactions at the 13-position to yield 2, 3, and formaldehyde. Formaldehyde also results from hydrogen peroxide oxidation of Tris [tris(hydroxymethyl)aminomethane] present in transcription buffer and spermine, a polyamine commonly associated with DNA in vivo, presumably via the Fenton reaction. Introduction Adriamycin (doxorubicin) continues to be one of the most important antitumor drugs in the clinic. It is a broad spectrum drug particularly useful in the treatment of Hodgkin’s disease, non-Hodgkin lymphomas, acute leukemias, sarcomas, and solid tumors of the breast, lung, and ovary.1 Daunomycin (daunorubicin), another anthracycline antitumor drug which differs from adriamycin only by the absence of a hydroxyl group at the 14-position, has been used primarily for the treatment of leukemia; however, a new liposomal formulation, DaunoXome, may extend its application.2 Both drugs have been investigated extensively with respect to mechanism of action. The drugs are strong intercalators in DNA with the A-ring and amino sugar moieties having important interactions with the minor groove. This activity results in topoisomerase-induced DNA strand breaks.3,4 The quinone functionality is redox active and catalyzes the production of reactive oxygen species in the presence of a reducing agent and molecular oxygen.5,6 Production of reactive oxygen species has been linked to both tumor cell cytotoxicity and cardiotoxicity. In the absence of molecular oxygen, reduction leads to glycosidic cleavage with the production of a quinone methide transient (Scheme 1) which shows mild electrophilic and nucleophilic reactivity.7 Alkylation of DNA has been proposed as a possible cytotoxic event; however, no structural evidence for its occurrence has been reported. The quinone methide † ‡ X

University of Colorado. Istituto di Medicina Sperimentale. Abstract published in Advance ACS Abstracts, March 15, 1997.

S0022-2623(96)00835-7 CCC: $14.00

from reductive activation of the related anthracycline, menogaril, has been shown to react slowly with 2′deoxyguanosine at its 2-amino substituent in a strictly anaerobic medium.8 The predominant reaction of the quinone methide from reduction of adriamycin and daunomycin is with the proton, leading to the respective, inactive 7-deoxyaglycons. Recent experiments by Phillips and co-workers have refocused attention on the possibility of reductive activation leading to the alkylation of DNA as an important cytotoxic event. They noticed that adriamycin in the presence of ferric ion and dithiothreitol caused transcriptional blockages in DNA.9,10 Adriamycin is a strong chelator of ferric ion, and complexation was proposed as necessary to achieve reduction with DTT. Transcriptional blockages were specific for 5′-GpC-3′ sites in the DNA and appeared to be the result of covalent bond formation at the 2-amino group of the deoxyguanosines.9,11 The covalent bonds were labile upon isolation of the DNA followed by redissolution in buffer.12 They proposed the quinone methide as a possible reactive intermediate based upon the precedent of reaction of reductively activated menogaril with 2′deoxyguanosine. An inconsistency with the quinone methide as a reactive intermediate was an insensitivity to the presence of molecular oxygen.9 In a communication,13 we reported that the covalent bonds between adriamycin or daunomycin and DNA at 5′-GpC-3′ sites were the result of formaldehyde Schiff base chemistry linking the 3′-amino group of the anthracyclines to the 2-amino substituent of deoxyguanosines. Further, the formaldehyde could result from anthracycline redox chemistry. We now report more © 1997 American Chemical Society

Redox Pathway Leading to Alkylation of DNA

Journal of Medicinal Chemistry, 1997, Vol. 40, No. 8 1277

Scheme 1

Table 1. Formation of DNA-Drug Adducts at Ambient Temperature as a Function of Reaction Conditions entry

drug

1

adriamycin

2

adriamycin

3

adriamycin

4

adriamycin

5

adriamycin

6

daunomycin

7

daunomycin

buffer reagent (equiv) Tris (pH 8.0)/KCl/EDTA/MgCl2 FeCl3/DTT (∼400 equiv) Tris (pH 7.4)/KCl/EDTA/MgCl2 H2O2 (∼30 equiv) phosphate (pH 8.0)/MgCl2 H2O2 (∼20 + 30 equiv)b phosphate (pH 7.0)/MgCl2 H2O2 (∼50 equiv) Tris (pH 7.4)/KCl/EDTA/MgCl2 formaldehyde (10 equiv) Tris (pH 8.0)/KCl/EDTA/MgCl2 H2O2 (∼12 equiv) Tris (pH 7.4)/KCl/EDTA/MgCl2 formaldehyde (5 equiv)

reaction time

DNA ratios (GC)4:1:(2 + 3):4:(5)a

4 days

0:19:35:45

4 days

40:5:20:35

14 days

83:8:6:1:(2)

12 days

60:11:16:5:(8)

1 day

0:8:30:62

9 days

trace:12:36:48

1 hc

40:9:34:17

a DNA ratios were determined from integrals of chromatograms for detection at 260 nm. They are uncorrected for differences in response. Hydrogen peroxide was added in two portions with the second portion added after 6 days. c Complete disappearance of the DNA was observed after 5 h with adducts 3 and 4 as the major products.

b

data and the full experimental details for the earlier experiments and the results of some additional relevant experiments, and we discuss the results more extensively with respect to other observations in the literature. Results Formation of DNA-Adriamycin and DNADaunomycin Adducts. The DNA selected for these studies was (GC)4 because it is self-complementary, bears multiple 5′-GpC-3′ sites, and is small enough for mass spectral characterization of labile drug-DNA adducts. Further, (GC)4 has a melting temperature sufficiently high to maintain a predominantly doublestranded structure in both the reaction medium and the eluent used for HPLC analyses and isolations. The predominant reaction medium was the transcription buffer employed by Phillips and co-workers, pH 8 Tris/ KCl/EDTA/MgCl2, and the HPLC eluent was 90% 20 mM triethylammonium acetate/10% acetonitrile. The respective (GC)4 melting temperatures were 61 and 45 °C. Reaction of (GC)4 with adriamycin in pH 7.4 or 8.0 Tris buffer containing ferric chloride and dithiothreitol at 25 °C was complete in 4 days and gave a product

mixture which showed four HPLC peaks designated 1-4, corresponding to DNA-drug adducts 1-4 (see Figure 4 in Supporting Information for an example of a chromatogram). Peaks for adducts 2 and 3 were usually poorly resolved and, consequently, integrated together. The ratio of DNA to adducts is reported for pH 8.0 in Table 1 (entry 1). Neither the extent of reaction as a function of time nor the adduct ratios showed significant dependence on the pH in the region 7-8. UV-vis spectra of the adducts showed a λmax for the drug at 510 nm consistent with the chromophore intercalated in double-stranded DNA (dsDNA). The relative intensities of the absorptions at 260 and 480 nm indicated that adducts 1 and 2 contained one drug chromophore per dsDNA and adducts 3 and 4, two drug chromophores per dsDNA. Since adriamycin redox chemistry is very sensitive to the presence of molecular oxygen, we surmised that the earlier conclusion of Phillips and coworkers that molecular oxygen was not a factor in the reaction was incorrect.9 A reaction similar to entry 1 of Table 1 but degassed under high vacuum, using freeze-thaw techniques, showed no reaction of the DNA after 4 days. With exposure to air, the DNA-drug adduct mixture appeared after a subsequent reaction period. The requirement of molecular oxygen indicated

1278

Journal of Medicinal Chemistry, 1997, Vol. 40, No. 8

Taatjes et al.

Scheme 2

Figure 1. Negative ion electrospray mass spectrum (ESMS) of DNA-adriamycin adduct 4 from reaction of (GC)4 with adriamycin and hydrogen peroxide in Tris buffer. Peaks representing assigned ions bearing one or more sodium or potassium ions are not labeled to simplify the figure. Calculated mass to charge ratios appear in Table 4 in Supporting Information.

that the catalytic production of reactive oxygen species by adriamycin was important. When the reaction was performed with hydrogen peroxide in place of FeCl3/ DTT, the same four HPLC peaks for DNA-drug adducts were observed, although the yield was lower after a comparable time period (see Table 1, entry 2). This was an early clue that the role of iron went beyond raising the reduction potential of adriamycin as originally proposed.9 Replacement of Tris buffer with phosphate buffer made the reaction even slower and produced a fifth DNA-drug adduct, designated 5, as shown in Table 1, entries 3 and 4. Hence, Tris appeared to play a role beyond that of a buffer. At this point mass spectral analysis of the predominant DNA-drug adduct (4) from reaction in Tris buffer with hydrogen peroxide was employed to characterize the molecular nature of the covalent bond. A major problem was the reported and observed instability of the DNA-drug adducts with respect to isolation. HPLC separation followed by centrifugal evaporation of the eluent, acetonitrile/pH 6 triethylammonium acetate, yielded only recovered (GC)4 and adriamycin. Addition of pH 7 phosphate buffer prior to centrifugal evaporation stabilized the DNA-drug adducts. The negative ion electrospray mass spectrum (ESMS) is displayed in Figure 1. It shows multiple ions corresponding to double-stranded DNA bound to two molecules of adriamycin each by a methylene unit, single-stranded DNA bound to one molecule of adriamycin by a methylene, and DNA. The calculated masses corresponding to each

peak are provided in Table 4 which appears in Supporting Information. We interpret the spectrum in terms of adduct 4 being double-stranded DNA covalently bound to two molecules of adriamycin, each by a methylene group. The other peaks result from decomposition in the inlet system of the mass spectrometer during ionization. The interpretation was further supported by MS/MS fragmentation experiments on m/z 740.6 which yielded ions for DNA, DNA bound to a single carbon, and adriamycin bound to a single carbon. Subsequently, mass spectra of DNA-drug adducts 1, 2 + 3, and 4 from reaction in Tris buffer with FeCl3/DTT were obtained; they are shown in Figure 5 of Supporting Information. They all show peaks for DNA bound to adriamycin by a methylene group as well as a peak for adriamycin bound to a single carbon. An important clue to the nature of the covalent bond linking adriamycin to DNA came from parallel studies of the reaction of adriamycin with DTT and with hydrogen peroxide. Both reactions gave two products, isolated and identified as the carboxylic acid 2 and the fully aromatized aglycon 3 (Scheme 2). The structure for the carboxylic acid was established from spectral comparison with a sample prepared by periodate oxidation14 and for the aglycon 3, from spectral comparison with literature data.15 The reactions were characteristic of the Baeyer-Villiger oxidation at the 13-position via transients 4 and 5 as proposed in Scheme 2. The Baeyer-Villiger mechanism predicted that the byproduct of formation of 2 was formaldehyde. In fact, formaldehyde was detected in the reaction mixture using the Hantzsch reagent.16 The corresponding Baeyer-Villiger oxidation of daunomycin with hydrogen peroxide yielded the aglycon 3 but not the carboxylic acid 2. Quite by accident, Wang and co-workers discovered that formaldehyde couples daunomycin to DNA via Schiff base chemistry involving the 3′-amino group of the drug and the 2-amino group of 2′-deoxyguanosines.17 In fact, they obtained single crystals of a DNA-drug adduct using (CG)3 for the DNA and performed an X-ray structural characterization. The adduct has two molecules of drug bound to double-stranded DNA, one to each strand via a one-carbon linkage (Scheme 3). This structure suggests a parallel structure for one of the DNA-drug adducts, probably adduct 4, from the reaction of (GC)4 with adriamycin and hydrogen peroxide

Redox Pathway Leading to Alkylation of DNA

Journal of Medicinal Chemistry, 1997, Vol. 40, No. 8 1279

Scheme 3

Table 2. Formation of Formaldehyde from Reaction of Tris Buffer (40 mM Tris/100 mM KCl/3 mM MgCl2) with 20 mM Hydrogen Peroxide as a Function of Conditions and Time [formaldehyde] (µM) at time buffer

2h

10 h

Tris (pH 7.4) Tris (pH 8.0)/EDTA (0.10 mM) Tris (pH 7.4)/Fe2+ (50 µM) Tris (pH 8.0)/EDTA (0.10 mM)/ Fe2+ (50 µM) Tris (pH 7.4)/Fe3+ (50 µM) Tris (pH 8.0)/EDTA (0.10 mM)/ Fe3+ (50 µM)

6.0

8.4

80 1.5 9.1 3.0

110 38 6.0 18

69 h 26 1.0 180 1600 84 1500

as shown in Scheme 3. Other possible adduct structures bearing one and two adriamycins based upon the Wang model are proposed in Figure 9 of Supporting Information. Wang and co-workers subsequently solved the crystal structures of several additional DNA-anthracycline adducts formed with formaldehyde.18,19 Reaction of daunomycin with (GC)4 and hydrogen peroxide in Tris buffer also yielded four DNA-drug adducts as shown in Table 1 (entry 6). A mass spectrum of adduct 4 appears in Figure 1 of the earlier communication13 and shows peaks for double-stranded DNA bound to two molecules of daunomycin, each by a methylene unit. However, daunomycin does not yield formaldehyde upon Baeyer-Villiger oxidation. What is the source of the formaldehyde? The answer lies in the enhanced DNA-drug adduct formation in Tris buffer, especially with iron present (Table 1, entries 1-4). In fact, reaction of Tris with hydrogen peroxide in the presence of Fe(II) or Fe(III) produced a significant amount of formaldehyde and even a larger amount when EDTA was present, as shown in Table 2. This appears to be a Fenton reaction and is precedented by the Cu(II)/EDTA-catalyzed oxidation of Tris to formaldehyde.20 Reaction of (GC)4 with adriamycin or daunomycin with excess formaldehyde rapidly gave the same four DNA-drug adducts as indicated by HPLC (Table 1, entries 5 and 7). Both reactions were complete in less than 1 day using 5-10 equiv of formaldehyde. Isolation and mass spectral analysis indicate similar structures as shown for daunomycin in Figure 6 of Supporting Information. One exception is the appearance of peaks corresponding to single-stranded DNA bound to two molecules of daunomycin, each by a methylene, in the spectra of adducts 3 and 4. Possibly, more than four adducts are present in these reactions because of the

Figure 2. Concentration of DNA and DNA-daunomycin adducts 1-4 as a function of time for the reaction of (GC)4 with daunomycin and formaldehyde in pH 7.0 phosphate buffer, starting with 168 µM (GC)4, 337 µM daunomycin, and 337 µM formaldehyde.

abundance of formaldehyde. In fact, a shoulder was observed on the HPLC peak representing adduct 4. The reactions of (GC)4 with adriamycin and hydrogen peroxide in phosphate buffer (Table 1, entries 3 and 4) showed an additional DNA-drug adduct, designated 5. The mass spectrum of adduct 5 showed peaks for singlestranded DNA bound to one molecule of carboxylic acid 2 via a methylene group as shown in Figure 7 of Supporting Information. Both products of hydrogen peroxide oxidation of adriamycin, 2 and 3, were also observed in the reaction mixture. Recall that formaldehyde is released upon formation of carboxylic acid 2. In Tris buffer bearing iron ion, neither 2, 3, nor DNAdrug adduct 5 were observed. This result is interpreted in terms of Tris/iron, present in high concentration, diverting all of the hydrogen peroxide from the BaeyerVilliger reaction to the Fenton oxidation of Tris. The corresponding reaction of (GC)4 with daunomycin and hydrogen peroxide in phosphate buffer yielded almost no DNA-drug adducts. This is consistent with the lack of formaldehyde formation from Baeyer-Villiger oxidation of daunomycin. Reversibility of Adduct Formation. As indicated above, attempts to isolate DNA-drug adducts can lead to decomposition back to starting DNA and drug. Similar decomposition of DNA-drug adducts was reported by Phillips and co-workers.9,12 This observation suggested that the DNA-drug adducts might actually exist in equilibrium with each other and with free DNA and drug. Equilibration was established by performing the same reaction at two different states of dilution. One reaction mixture contained (GC)4, daunomycin, and formaldehyde at 10 times higher concentrations than the other. Both reactions were maintained at 25 °C and monitored for DNA-drug adduct formation as a function of time. The results are shown in Figures 2 and 3. Both reactions appeared to approach equilibrium after about 150 h of reaction time. At equilibrium, the more concentrated reaction solution showed almost complete

1280

Journal of Medicinal Chemistry, 1997, Vol. 40, No. 8

Taatjes et al.

Chart 1

mine was approximately 100-fold lower than the concentration of Tris in buffer. Discussion

Figure 3. Concentration of DNA and DNA-daunomycin adducts 1-4 as a function of time for the reaction described in the legend to Figure 2 but with reagents at a 10-fold dilution. Table 3. Formation of Formaldehyde from Reaction of 5 × 10-4 M Spermine with 2.5 × 10-2 M Hydrogen Peroxide in pH 7.4 0.10 M Phosphate Buffer as a Function of Time [formaldehyde] (µM) at time buffer

17 h

40 h

4 days

PO4 PO43-/EDTA (0.15 mM) PO43-/Fe2+ (50 µM) PO43-/EDTA (0.15 mM)/ Fe2+ (50 µM)