Mass Spectrometry Studies of the Binding of the Minor Groove

146

Chem. Res. Toxicol. 2009, 22, 146–157

Mass Spectrometry Studies of the Binding of the Minor Groove-Directed Alkylating Agent Alkamin to AT-Tract Oligonucleotides Amin M. S. Abdul Majid,*,†,‡ George Smythe,§ William A. Denny,| and Laurence P. G. Wakelin† School of Molecular Sciences, Department of Pharmacology, and Bioanalytical Mass Spectrometry Facility, UniVersity of New South Wales, NSW 2052, Australia, School of Pharmacy, UniVersiti Sains Malaysia, Penang, Malaysia, 11800, and Auckland Cancer Society Research Centre, Faculty of Medicine and Health Science, UniVersity of Auckland, New Zealand, 1142 ReceiVed July 30, 2008

Minor groove binding alkylating agents, which have potential as cancer drugs, generate cytotoxic DNA adducts that are relatively resistant to repair as a consequence of locating covalent attachment at purine N3 nitrogen atoms. Recently, we used electrospray and matrix-assisted laser desorption ionization mass spectrometry to study the binding of the minor groove-directed polybenzamide bis-half-mustard alkamin, and its monofunctional analogue alkamini, to the oligonucleotide d(CGCGAATTCGCG)2, identifying a number of inter- and intrastrand alkamin cross-links involving the GAATTC sequence [Abdul Majid, A. M. S., Smythe, G., Denny, W. A., and Wakelin, L. P. G. (2007) Mol. Pharmacol. 71, 1165-1178]. Here, we extend these studies to d(CGCAAATTTGCG)2, A3T3, and d(CGCAAAAAAGCG)·d(CGCTTTTTTGCG), A6/T6, in which the opportunity for both inter- and intrastrand cross-linking is enhanced. We find that both ligands alkylate all adenines in the longer AT-tracts, as well as the abutting guanines, whether they are in the same strand as the adenines or not, in a manner consistent with covalent attack on purine N3 atoms from the minor groove. Alkamin forms intrastrand cross-links involving A4 and A6 and A6 and G10 in A3T3 and all of the purines in the A6/T6 purine tract, including G10. In addition, it forms interstrand cross-links between A4, A5, A6 and A4′, A5′, A6′, between G10 and the latter adenines in A3T3, and between G22 and adenines A5 and A6 in A6/T6. The reactivity of the abutting guanines provides unexpected opportunities for both inter- and intrastrand cross-linking by alkamin, such as the interstrand cross-link in the CAAAAAAG sequence. We conclude that positioning monofunctional mustard groups on either end of a minor groove-directed polybenzamide has the capacity to enhance interstrand cross-links at all manner of AT-tracts, including most in which the adenines are all in one strand. Introduction Nitrogen mustard alkylating agents make an important contribution to cancer chemotherapy. However, in addition to forming cytotoxic DNA interstrand cross-links, they are susceptible to hydrolysis, form complexes with proteins, can be deactivated by cellular thiols, and yield mutagenic and carcinogenic monofunctional adducts (1-3). One approach to overcoming these deficiencies has been the development of DNA-targeted mustards, achieved by attaching alkylating groups to minor groove binding (MGB)1 ligands and to intercalating agents (4, 5). The major objective has been to increase dose potency and to enhance interstrand cross-link formation at the * To whom correspondence should be addressed. Tel: +6046532232. Fax: +6046570017. E-mail: [email protected]. † Department of Pharmacology, University of New South Wales. ‡ University of Science Malaysia. § Bioanalytical Mass Spectrometry Facility, University of New South Wales. | University of Auckland. 1 Abbreviations: MGB, minor groove binding; ESI, electrospray ionization; MALDI, matrix-assisted laser desorption ionization; TOF, time-offlight; MS/MS, tandem mass spectrometry; A2T2, d(CGCGAATTCGCG)2; A3T3, d(CGCAAATTTGCG)2; A6/T6, d(CGCAAAAAAGCG)·d(GCGTTTTTTCGC); for abbreviations describing molecular ion assignments, see the Supporting Information, Figures S1-S3 and Table S1.

expense of monofunctional reaction (4, 6). Several classes of MGB-directed aniline mustards have been described incorporating oligopyrrole peptide, bisbenzimidazole, bis-quaternary ammonium heterocycle, and polybenzamide carriers (see Figure 1 and ref 4), and some, such as tallimustine (7, 8), have been evaluated in the clinic. Attaching mustards to MGBs in this manner redirects alkylation from the N7 of guanine in the major groove to the N3 of adenine in the minor groove (3, 4, 9). In seeking to maximize cross-link frequencies at palindromic AT-tracts, Denny and colleagues have synthesized ligands that incorporate single-armed mustard groups arranged at either end of a terephthalic acid framework (Figure 1 and refs 10 and 11), the intent of which was to alkylate adenines in complementary strands. DNA sequencing experiments show that alkamin, the parent bifunctional polybenzamide mustard (Figure 1), binds to AT-tracts in a manner consistent with this design principle (10, 12) and that its monofunctional analogue, alkamini, shares this binding mode and sequence selectivity (12). Alkamin causes interstrand cross-links in vitro, is potently cytotoxic against P388 mouse leukemia cells, and has in vivo activity against this tumor model (10, 11). In contrast, alkamini is 35-fold less cytotoxic and lacks in vivo activity, pointing to the importance of alkamin cross-links as a determinant of its biological activity (10, 11). Recently, we have used matrix-assisted laser desorption ioniza-

10.1021/tx800276h CCC: $40.75  2009 American Chemical Society Published on Web 12/29/2008

Alkamin Binding to A3T3 and A6/T6 DNAs

Chem. Res. Toxicol., Vol. 22, No. 1, 2009 147

Figure 1. Structures of minor groove-targeted alkylating agents.

tion (MALDI) and electrospray ionization (ESI) mass spectrometry to characterize the binding of these agents to the oligonucleotide d(CGCGAATTCGCG)2, A2T2, and have found that alkamin forms a variety of monofunctional adducts and interstrand and intrastrand cross-links (13). Here, we extend this analysis to two related dodecamers in which the AT-tract is modified. In the first, d(CGCAAATTTGCG)2, A3T3, the longer self-complementary AT-run provides an increased potential for the formation of adenine-adenine interstrand cross-links, and in the second, d(CGCAAAAAAGCG)· d(CGCTTTTTTGCG), A6/T6, the extended A-tract provides the opportunity for enhanced intrastrand cross-links. We find that both ligands alkylate all adenines in the extended AT-tracts, as well as the abutting guanines, whether they are in the same strand as the adenines or not, in a manner consistent with covalent attack on purine N3 atoms from the minor groove. Alkamin forms intrastrand cross-links involving A4 and A6 and A6 and G10 in A3T3 and all of the purines in the A6/T6 purine tract, including G10. In addition, it forms interstrand cross-links between A4, A5, A6 and A4′, A5′, A6′, between G10 and the latter adenines in A3T3, and between G22 and adenines A5 and A6 in A6/T6. The reactivity of the abutting guanines provides unexpected opportunities for both inter- and intrastrand cross-linking by alkamin, such as the interstrand cross-link in the CAAAAAAG sequence. We conclude that positioning monofunctional mustard groups on either end of a minor groovedirected polybenzamide is an excellent strategy for enhancing the formation of interstrand cross-links at all manner of ATtracts, including most in which the adenines are all in one strand. Lastly, we note that our findings further illustrate the power of mass spectrometry to provide a detailed description of the multifarious bound forms of MGB alkylating agents complexed with AT-tract DNAs. The method can provide direct information about the points of attachment in inter- and intrastrand crosslinked adducts and can distinguish between them at base resolution, issues that are difficult to resolve by other means. In the present work, we also extend the type of DNA-ligand complex that can be identified, since the spectra contain evidence that alkamini, but not alkamin, is also able to alkylate phosphates. Specifically, alkamini forms a phosphotriester with the cytosine phosphate in the sequence 5′-TTTGC in both oligonucleotides studied.

Materials and Methods Materials. Alkamin, MW 714.6, and alkamini, MW 609.0, were synthesized as the dihydrochloride salts as previously described (10, 14), were dissolved to a concentration of 2.0 mM in a mixture

of isopropanol and acetonitrile to prevent hydrolysis, and were stored at -20 °C. The dodecanucleotides d(CGCAAATTTGCG), d(CGCAAAAAAGCG), and d(CGCTTTTTTGCG), designated A3T3, A6, and T6, respectively, were purchased from Oswel DNA Services (United Kingdom). All were synthesized as the ammonium salt and purified by HPLC; their molecular masses were 3656.4, 3683.3, and 3629.3 Da, respectively. A3T3 was dissolved to a concentration of 4 mM in duplex in deionized water, diluted to 2 mM in 100 mM ammonium acetate buffer, pH 7.0, and annealed by heating to 90 °C. After it was slowly cooled to room temperature, it was further diluted to 100 µM in 20 mM ammonium acetate. The A6T6 duplex, being nonself complementary, was made by annealing the two complementary dodecamers at 90 °C in a 1:1 molar ratio to give 2 mM duplex in 100 mM acetate buffer. The mixture was cooled to room temperature and diluted to 100 µM in 20 mM ammonium acetate buffer to give a duplex of sequence A6/T6. All DNA samples were stored at -20 °C. For the selfcomplementary sequences, bases have been numbered from the 5′end on the Watson strand, for example, C1G2C3A4A5 etc., and have been denoted as their prime on the Crick strand, for example, C1′G2′C3′A4′A5′ etc., whereas the A6/T6 duplex has been numbered 1-12 on the Watson strand and 13-24 on the Crick strand. Preparation of DNA-Ligand Complexes. DNA-ligand complexes were prepared in 20 mM ammonium acetate buffer, pH 7.0, at a stoichiometry of one drug molecule per duplex by mixing 2 µL of 2 mM ligand in isopropanol with 38 µL of 100 µM DNA in 20 mM NH4OAc solution. Thus, the final conditions were 100 µM ligand, 95 µM DNA, and 2.5% isopropanol in 20 mM NH4OAc, pH 7.0. Complexes were prepared in 1.5 mL Eppendorf tubes and were incubated in a heating block at 37 °C for 24 h. Control DNA samples, in which the DNA was incubated without drug, were subjected to the same procedures. Freshly prepared complexes were analyzed immediately after diluting the DNA concentration to 20 µM with matrix or organic solvent. Mass Spectrometry. Electrospray spectra were collected using an Applied Biosystems MDS Sciex Q-Star Pulsar i mass spectrometer fitted with a NanoSpray ion source, and MALDI spectra were collected using a Voyager Elite XL Biospectrometry Workstation equipped with a pulsed nitrogen laser (337 nm). Mass spectrometry measurements were performed as previously described in ref 12, which gives a comprehensive account of sample preparation and instrument settings. Briefly, samples for MALDI measurements were prepared by mixing 2 µL DNA-ligand samples with 8 µL of a freshly prepared 3-hydroxypicolinic acid matrix (50 mg of 3-hydroxypicolinic acid dissolved in 1 mL of a 50% aqueous solution of acetonitrile, mixed with 1 mL of a 50 mg/mL aqueous solution of ammonium citrate). A total of 4 µL of this mixture was spotted onto a stainless steel sample plate, and positive ion spectra were recorded with the spectrometer in reflector mode with delayed extraction. All spectra were calibrated manually with reference to the mass of the individual unreacted single-stranded

148

Chem. Res. Toxicol., Vol. 22, No. 1, 2009

DNA peaks. Electrospray analyses were carried out using freshly prepared solutions containing 8 µL of a 50% water/acetonitrile mixture and 2 µL of the aqueous DNA-ligand sample. Two microliter samples were injected over a period of 5 min with a flow rate of 20 nL per min. The Q-Star instrument was calibrated against a mixture of ALILTLVS peptide and CsI, and spectra were recorded in both negative and positive ion reflector modes from 100 to 2000 m/z. Data were processed using Analyst QS; the primary mass-to-charge ratio spectra containing singly and multiply charged species were reconstituted into a simple mass spectrum using the Bayesian Peptide Reconstruction (BPR) function. All peaks in the reconstituted spectra were checked for authenticity by visually inspecting their multiply charged components in the primary spectra. Where possible, negative ion tandem mass spectrometry (MS/MS) measurements were made on DNA-ligand complexes using the MS/MS mode of the Q-Star spectrometer. A time-of-flight (TOF)-MS scan was acquired (m/z 400-2500, 1 s) and accumulated for ∼2 min. Precursor masses determined from the TOF-MS scan were selected by Q1 for MS-MS analysis. Nitrogen was used as the collision gas, and an optimum collision energy was chosen for each precursor ion.

Results General Considerations. To present our findings as succinctly as possible, yet to make available the full range of experimental data, we have placed the complete MALDI and ESI spectra for each drug-DNA complex, together with their detailed species analysis, and tandem MS/MS spectra used to substantiate important assignments in the Supporting Information. In the text proper, we provide illustrative spectra and tables of the essential species that identify the critical DNA-ligand molecular interactions. The Supporting Information also contains a description of the types of chemical species that accompany alkylation, the nomenclature for describing the observed species, and a description of the tandem MS/MS gas-phase degradation pathways. In positive ion MALDI spectra, DNA and its fragment ions generally appear as the singly charged free acid, and in our systems, ions of m/z 7000-8000 represent duplex forms and related species, such as unreacted DNA, intact duplex-drug adducts, and depurinated duplex ions. Those between m/z 3600 and 4500 include unreacted single-stranded DNA and singlestranded DNA-drug adducts. The peaks in the range m/z 3000-3600 represent deadenylated and deguanylated singlestranded DNAs resulting from depurination of the quaternized bases, and between m/z 1000 and 3000, we see the fragmented DNA products derived from phosphate hydrolysis at the alkylated and depurinated sites. Finally, in the low molecular mass region m/z 500-1000 are found peaks representing ligand-base adducts resulting from depurination, hydrolyzed unreacted ligand, and an assortment of matrix-derived materials. The transformed negative ion ESI spectra generally follow the same form as found in the MALDI measurements, save for the observation of fewer duplex species. Structure of the Alkamini-A3T3 Complex. The species assignments directly relevant to defining the nature of the complex are displayed in Table 1, and a representative MALDI spectrum is shown in Figure 2a; the species abbreviations are defined in Table S1 of the Supporting Information. The data imply that alkamini forms a variety of 1:1 and 2:1 complexes with A3T3 DNA, despite the initial stoichiometry being 1 drug molecule added per duplex. For the 1:1 complexes, there is abundant evidence that alkamini alkylates adenines A4, A5, and A6 and guanine G10. Support for this includes observation of the single-stranded DNA-ligand adduct, SS-L; deadenylated and deguanylated single-stranded DNAs, SS deA and SS deG; RHFs for A4, A5, and A6; LHFs for the three adenines and

Abdul Majid et al. Table 1. Species Assignments That Define the Binding Modes for the A3T3 Complexes of Alkamini and Alkamina alkamini species assignments DS-L deA DS DNA DS deA DS 2deA SS-2 L SS-L-G SS-L-A SS-L-OH SS)L SS-L SS-L deA SS-L deG SS DNA SS deA SS deG SS deA deA SS deA deG CGCAAATTTG LHF G10 RHF A4 RHF A5 RHF A5 deG RHF A6 LHF A6* LHF A5* L-pCG LHF A4* A-L-G A-L-A G-L-OH A-L-OH L-G L-A L-OH

MALDI (+ve ion)

ESI (-ve ion)

alkamin MALDI (+ve ion)

ESI (-ve ion)

X X X NA NA NA NA X X X X X X X X X X X X X X NA NA NA NA X X X

X NA NA NA NA X X X X X X X

X X X NA X X X X NA NA NA X X X X X

NA X X X NA NA NA X X X

X X X X X

X X

X

X X X

X X X

X NA NA NA NA X (+ve) X (+ve) X (+ve)

X X X X NA NA NA

X (+ve) X (+ve) X (+ve) X (+ve) NA NA NA

a X indicates the species observed, NA means not applicable, a blank means species not observed, and * signifies the presence of sugar ring variants but not the parent ion. The full data, including observed and calculated masses, are recorded in the Supporting Information, Tables S2 and S3.

guanine G10; and the base-ligand ions for reaction with adenine, L-A, and guanine, L-G (Tables 1 and S2 of the Supporting Information and Figures 2a and S4 and S5 of the Supporting Information). Figure 3b shows a schematic illustration of the four 1:1 binding modes for base alkylation in the alkamini-A3T3 complex. With the mustard group attached at, or near, the 5′-end of the AT-tract, the benzamide backbone can sit in the AT-segment covering 3-4 base pairs toward the 3′-end, whereas when covalently attaching to guanine G10, the ligand binds in the reverse orientation. The data also strongly indicate the presence of 2:1 complexes involving two ligand molecules bound simultaneously to one DNA strand, the ligands attaching covalently at guanine G10 and at adenine A5. The evidence for these complexes includes observation of the intact single-stranded DNA-ligand adduct bearing two ligand molecules, SS-2 L; a ligand-bound deadenylated duplex ion, DS-L deA; ligand-bound deadenylated and deguanylated single-stranded DNAs, SS-L deA and SS-L deG; doubly deadenylated and deguanylated single-stranded DNA, SS deA deG; and an RHF A5, which is also deguanylated, RHF A5 deG (Tables 1 and S2 of the Supporting Information and Figures 2a, 3b, and S4 and S5 of the Supporting Information). The data are essentially silent about the potential presence of 2:1 complexes in which a ligand is attached to an adenine in each strand. Such complexes may exist but are difficult to identify because their single-stranded DNA species are indis-



Figure 2. Positive ion MALDI-TOF spectra for ligand complexes with A3T3 DNA. (a) Alkamini and (b) alkamin. Full spectral data for both MALDI and ESI measurements are shown in Figures S4-S7 of the Supporting Information, and species assignments are recorded in Tables S2 and S3 of the Supporting Information.

tinguishable from those of genuine 1:1 complexes. Indeed, there are two “smoking guns” in the data, which suggest such a possibility, namely, the presence of doubly deadenylated duplex ions, DS 2deA, and the exceptionally low levels of intact singlestranded DNA in both MALDI (4%) and ESI (14%) spectra (Figures S4 and S5 and Table S2 of the Supporting Information). In the absence of evidence for intrastrand 2:1 complexes involving two adenines (Table 1 and Table S2 and Figures S4 and S5 of the Supporting Information), the DS 2deA ion can only originate from an interstrand 2:1 complex. Similarly, it is difficult to account for loss of substantially more than 50% of unreacted single-stranded DNA unless there has been a simultaneous covalent reaction on each strand. To help substantiate the above assignments, we also made tandem MS/MS ESI measurements selecting SS-L and SS-2 L (Figure S6a,b of the Supporting Information) and SS-L deG and SS-L deA (Figure S6c of the Supporting Information) as the parent ions. Figure S6a of the Supporting Information reveals unequivocal evidence for reaction in the 1:1 complex with A4,

A5, A6, and G10 and shows base-ligand adducts for reaction with adenine and guanine and a large peak for deadenylated DNA. Taken at face value, the areas for the RHFs suggest that in the 1:1 complexes, alkamini binds preferentially in the order A5 > A4 > A6 > G10, the preference for A5 over A4 being 3-4-fold (Figure S6a of the Supporting Information). With respect to the 2:1 complexes, the SS-2 L MS/MS spectrum, Figure S6b of the Supporting Information, contains fragmentation products that derive from both the 2:1 complexes themselves and the variety of 1:1 complexes that result from collision-induced disruption of a single ligand-purine N3 bond. Thus, the spectrum is complex and difficult to interpret, but it is clear that all four purines, A4, A5, A6, and G10, are involved in the 2:1 complexes as revealed by the standard RHF and LHF fragments. The most telling peaks are associated with fragments that define the predominant positions of simultaneous attachment of two alkamini molecules, and these are the RHF A5(a)-L (m/z 2690.8) and RHF A5(a) deG (m/z 2039.8), which, taken with the other available evidence, strongly suggests that A5 and G10

150


Figure 3. Schematic diagram of modes of alkamini binding to A2T2, A3T3, and A6/T6 DNA. (a) A2T2: In the 1:1 complex, alkamini binds either to G4, A5, or A6, and several 2:1 combinations are also found; see ref 13. (b) A3T3: In the 1:1 complex, alkamini binds to A4, A5, A6, and G10 and also forms a triester with the phosphate of C11; 2:1 complexes involve A5 and G10. (c) A6/T6: In the 1:1 complex, alkamini binds to A4 through G10 and G22 and may form a triester with the phosphate of C23. A variety of 2:1 complexes exist, as decribed in the text.

are the principal sites, as deduced from the original spectra. To help confirm this assertion and to minimize the complicating consequences of DNA-ligand adduct dissociation in 2:1 complexes, we also scrutinized MS/MS spectra for SS-L deA and SS-L deG (Figure S6c of the Supporting Information), which indeed show peaks for RHF A5(a)-L and RHF A5(a) deG. Finally, for the alkamini-A3T3 complex, we note the presence of a new type of reaction where the ligand appears to alkylate a phosphate group to form a phosphotriester. The MALDI spectrum has a peak at m/z 3028.2, which we assign as the intact DNA fragment CGCAAATTTG with 5′- and 3′hydroxyls, and a peak at m/z 1136.1, which we identify as the species L-pCG, in which the ligand is attached to the 5′phosphate of the dinucleotide phosphate pCG (Table 1 and Table S2 and Figure S5 of the Supporting Information). These fragments derive from hydrolysis of the triester formed at the cytosine C11 phosphate and represent the products from cleavage of the G10 C3′-O3′-P-O5′-C11 moiety at the O3′-P bond where the O3′ oxygen cleaves with the sugar (see the Discussion and Figure 6). There is also evidence for the hydrolyzed ligand, L-OH, which is a potential product of hydrolysis of the ligand-phosphotriester bond. However, in the MALDI spectrum of the complete ligand-DNA reaction mixture, this species could also derive from ligand that simply failed to alkylate the DNA. Unfortunately, the phosphotriester fragments do not appear in the ESI spectrum, preventing direct confirmation of the assignments using MS/MS methods. Accordingly, we attempted to find supporting evidence in the MS/ MS spectra described above (Figure S6a,b of the Supporting Information), anticipating that the intact triester would be represented in SS-L or SS-2 L, but have had only minimal success. In no case do we observe the expected DNA-containing

Abdul Majid et al.

fragments, but we do find low levels of the hydrolyzed mustard, as [HL-O]- at m/z 517.5, which can only derive in a gas-phase MS/MS experiment involving the intact DNA-ligand complex as the parental ion, from fragmentation of a ligand-oxygen adduct, such as would be found in a triester. This binding mode is depicted in Figure 3b with the polybenzamide backbone of alkamini lying in the AT-tract and the mustard group reacting with the phosphate 3′ to the adjacent guanine G10. The formation and hydrolysis of such phosphotriesters are addressed in the Discussion. Structure of the Alkamin-A3T3 Complex. Species assignments for the alkamin-A3T3 complex are displayed in Table 1, and a representative MALDI spectrum is shown in Figure 2b. The data reveal that alkamin forms a variety of 1:1 complexes with A3T3 DNA involving monofunctional adducts and inter- and intrastrand cross-links, but there is no evidence for 2:1 complexes. The proportion of unreacted single-stranded DNA in both MALDI and ESI spectra, 25-33% (Table S3 of the Supporting Information), suggests that the interstrand crosslink frequency is about 1 in 2 (i.e., 50% of the original duplex DNA is involved in interstrand cross-links). As found for alkamini, there is much evidence that alkamin monofunctionally alkylates adenines A4, A5, and A6 and guanine G10, but there is no evidence for reaction with a phosphate. The main indicators of monofunctional reaction include intact single-stranded DNA bearing a hydrolyzed ligand, SS-L-OH; intact single-stranded DNA, SS DNA; deadenylated and deguanylated single-stranded DNAs, SS deA and SS deG; a deadenylated duplex DNA, DS deA; RHFs for reaction with A4, A5, and A6; LHFs for adenines A5 and A6; and base-ligand adducts for A-L-OH and G-LOH (Tables 1 and S3 of the Supporting Information and Figures 2b and S7 and S8 of the Supporting Information). These assignments are substantiated by tandem MS/MS ESI measurements on the parent ion SS-L-OH (m/z 845.2, z ) -5), which show daughter ions of the LHF(a)s and RHF(a)s for A4, A5, and A6 and the RHF(a) for G10, together with A-L-OH and G-L-OH (Figure S9a of the Supporting Information). Figure 4b shows the schematic diagram for these binding modes in which alkamin lies within the AT-tract in the same manner as alkamini (Figure 3b), with the hydrolyzed mustard lying to the 3′-end of the duplex. As before, binding to guanine G10 is with the reversed orientation (Figures 3b and 4b). The areas of the RHFs suggest that, like alkamini in its 1:1 A3T3 complexes, alkamin binds preferentially to adenine A5; however, its degree of preference for A5 over the other purines is reduced to about 2-fold (Figure S9a of the Supporting Information). The principal evidence for intrastrand cross-linked complexes includes the intact single-stranded DNA bearing an alkamin molecule that lacks a hydrolyzed mustard, SS)L; doubly deadenylated single-stranded DNA, SS deA deA; a mixed deadenylated and deguanylated DNA, SS deA deG; and the base-ligand fragments A-L-G and A-L-A (Tables 1 and S3 of the Supporting Information and Figures 2b and S7 and S8 of the Supporting Information). Unfortunately, SS)L is not observed in the ESI spectrum, which precludes the use of tandem MS/MS to identify unequivocally which specific purines are involved in the cross-link. However, structural considerations suggest that the A-L-G cross-link can be assigned to adenine A6 and guanine G10, and molecular models indicate that this 1f5 cross-link requires a fully extended alkamin conformation. Similar structural arguments suggest that the adenine-adenine cross-link must be a 1f3 linkage between A4 and A6, where the mustard groups are oriented in their most proximal position. Figure 4b shows a schematic description of these binding modes.



Figure 4. Schematic diagram of modes of alkamin binding to A2T2, A3T3, and A6/T6 DNA. (a) A2T2: Alkamin forms monofunctional adducts with G4, A5, and A6, an intrastrand cross-link between G4 and A6, and interstrand cross-links between G4 and A5′ or A6′ and between A5/A6 and A5′/A6′; see ref 13. (b) A3T3: Alkamin forms monofunctional adducts with A4, A5, A6, and G10, intrastrand cross-links between A4 and A6 and between A6 and G10, and interstrand cross-links between G10 and A5′ or A6′ and between A5 or A6 and A5′ or A6′. (c) A6/T6: Alkamin forms monofunctional adducts with A4 through G10 and with G22; 2:1 monofunctional complexes involving two adenines on A6; two classes of intrastrand cross-links involving all possible adenine-adenine linkages from A4 to A9 and a guanine to adenine cross-link involving G10 and A6, A7, or A8; and interstrand cross-links between G22 and A5 or A6.

Interstrand cross-links are revealed by the finding of intact single-stranded DNAs bearing an alkamin molecule to which is attached a guanine or adenine base (SS-L-A, SS-L-G), the purine-ligand adducts A-L-G and A-L-A, and all of the usual DNA-associated derivatives resulting from fragmentation of SSL-A and SS-L-G (Tables 1 and S3 of the Supporting Information

and Figures 2b and S7 and S8 of the Supporting Information). The latter patterns are indistinguishable from fragmentation of the monofunctional adduct, indicating that the interstrand crosslinks necessarily involve A4, A5, A6, and G10 and their complementary bases A4′, A5′, A6′, and G10′. To help resolve the exact points of attachment, tandem MS/MS measurements

152


Abdul Majid et al.

Table 2. Species Assignments That Define the Binding Modes for the A6/T6 Complexes of Alkamini and Alkamina alkamini species assignments DS A6/T6-2(L-OH) DS A6/T6)L DS A6/T6-2 L SS A6-2(L-OH)* SS T6-L-A SS A6-2 L SS A6-L-OH SS A6)L SS T6-L-OH SS A6 deA-L-OH SS A6 DNA SS T6 DNA SS A6-L SS T6-L SS A6-L deA SS A6 deA SS T6 deG SS A6 deA deA CGCTTTTTTG RHF A4 RHF A5 RHF A5 deA RHF A6 RHF A6 deA RHF A7 LHF A6* LHF A5* RHF A8 L-pCG RHF A9 A-L-G A-L-A G-L-OH A-L-OH L-G L-A RHF G10/G22 L-OH HO-L-OH

MALDI (+ve ion) NA NA NA NA X NA NA NA NA X X X X X X X X X X X X X X X X X X NA NA NA NA X X X NA

ESI (-ve ion) NA NA X NA NA X NA NA NA NA X X X X X X

alkamin MALDI (+ve ion)

NA X NA X X X X NA NA NA X X

X

ESI (-ve ion) X X NA X NA X X X X X NA NA NA X X X

X X X

X

X X X NA X

NA NA NA NA X (+ve) X (+ve) X X (+ve) NA

X X NA NA NA

X X NA X (+ve) X (+ve) X (+ve) X (+ve) NA NA X NA X (+ve)

a X indicates the species observed, NA means not applicable, a blank means species not observed, and * signifies the presence of sugar ring variants, or an ion pair, but not the parent ion. The full data, including observed and calculated masses, are recorded in the Supporting Information, Tables S4 and S5.

with SS-L-A as the parent ion (m/z 870.6, z ) -5) revealed reaction with all four purines, both LHF(a)s and RHF(a)s being located for each, and the base-ligand adducts A-L-A and A-L-G (Figure S9b of the Supporting Information). This result indicates that adenine-adenine interstrand cross-links involve all three adenines in each strand, the area of the RHF A5(a) peak, which is by far the largest, suggesting that the main cross-link is a 1f4 linkage between A5 and A5′ (Figure S9b of the Supporting Information). The tandem data for the adenine-guanine crosslink, which is confirmed by this measurement to involve G10, cannot define which complementary adenine is involved. Attempts to resolve this issue by tandem measurements using SS-L-G as the parent ion have been thwarted because of its low intensity. Figure 4b displays a schematic of the interstrand cross-links, showing a variety of 1f3, 1f4, and 1f5 adenine-adenine linkages and 1f3 and 1f4 adenine-guanine adducts. Structure of the Alkamini-A6/T6 Complex. Species assignments for the alkamini-A6/T6 complex are reported in Table 2, and a representative MALDI spectrum is shown in Figure 5a. As found for A3T3, the data indicate that alkamini forms a variety of 1:1 and 2:1 complexes with A6/T6 DNA,

notwithstanding the input stoichiometry being 1:1. In accord with this stoichiometry and given that all of the target adenines are on one strand, no trace of intact A6 DNA remains in either the MALDI or the ESI spectrum (Figures 5a, and S10 and S11 of the Supporting Information). For this complex, there is a practical difference between the two spectra, in so much as the DNA fragments resulting from cleavage of the depurinated DNAs are largely missing from the ESI spectrum (Figures S10 and S11 of the Supporting Information and Tables 2 and S4 of the Supporting Information). However, taking both spectra together, the evidence shows that alkamini forms 1:1 complexes with alkylation of every adenine on the A6 DNA strand and with guanine G10 (Figure 3c). This assertion is supported by observation of a peak for the intact A6 DNA bearing a ligand, SS A6-L, for the deadenylated A6 strand, for the RHFs for A4 through to A9, and for the ligand-base adducts L-A and L-G (Table 2 and Figure 5a). Interestingly, the LHFs for reaction at the purines are missing from both spectra, and in the MALDI, the normal RHFs are all accompanied by companion fragments in which the phosphate has cleaved to the left-hand fragment, so yielding RHFs terminating in a 5′-hydroxyl (Figures S10 and S11 and Table S4 of the Supporting Information). Although this behavior has been seen before (13), it has never been so consistent, implying that phosphate cleavage of depurinated DNA in the long oligopurine tract is more finely balanced between hydrolysis of the P-O3′ and P-O5′ bonds (see Figure S1 of the Supporting Information) at the apurinic sugar. With respect to reaction with G10, the ESI spectrum gives a peak that is ambiguous in assignment since it could derive from reaction with G10 or G22. However, a tandem MS/MS ESI measurement with SS A6-L as the parent ion (Figure S12a of the Supporting Information) unequivocally shows reaction with G10, and the full spectrum supports the above assignments. Figure S12a of the Supporting Information also reveals that fragmentation in the tandem experiment is more resistant to collision-induced decay for this DNA complex than found previously for A2T2 (13) and A3T3 adducts, so much so that the dominant ions in the spectrum are the intact complex and the deadenylated DNA. The resistance of the latter to fragmentation results in only a limited number of peaks that usefully define the binding sites. Raising the collision energy did not resolve the issue, since the daughter fragments have been found to decompose further to granddaughter ions that obscure meaningful interpretation. Evidence that alkamini also reacts with the T6 DNA strand is provided by observation of a peak for the intact T6 DNA bearing a ligand, SS T6-L, for the deguanylated T6 strand and for a RHF derivative of G10/G22 (Table 2 and Figure 5a). Tandem MS/MS ESI measurements with SS T6-L and SS T6 deG (Figure S12b,c of the Supporting Information) as the parent ions confirm reaction with G22 and provide direct evidence for the formation of a G-L adduct from this same guanine. Thus, as shown in the schematic Figure 3c, we can identify eight 1:1 alkamini A6/T6 complexes, seven on the A6 strand and one on the T6 strand. Molecular models suggest that the complexes involving A4, A5, and A6 have the ligand bound with the polybenzamide backbone lying toward the 3′-end of the A-tract and that the orientation of those bound to A8, A9, and G10 is reversed. The ligand attached to A7 could lie in either direction. On the T6 strand, alkamini is attached to G22 and surely lies with its backbone positioned toward the 3′-end of the duplex, so providing a mirror image of the ligand bound at G10. Finally, with regard to the 1:1 complexes, we note that the MALDI spectrum contains a peak at m/z 1136.4 that we assign to the



Figure 5. Positive ion MALDI-TOF spectra for ligand complexes with A6/T6 DNA. (a) Alkamini and (b) alkamin. Full spectral data for both MALDI and ESI measurements are shown in Figures S10, S11, S13, and S14 of the Supporting Information, and species assignments are recorded in Tables S4 and S5 of the Supporting Information.

species L-pCG (Tables 2 and S4 of the Supporting Information and Figures 5a and S10 and S11 of the Supporting Information), which is one of the hydrolytic fragments to be expected from alkylation of the C11/C23 phosphate, that is, the phosphate immediately 3′ to guanine G10/G22. We also observe hydrolyzed alkamini, which is another potential product of hydrolysis of the triester, but we have failed to find the third component, CGCAAAAAAG/CGCTTTTTTG, that would confirm phosphate alkylation. However, careful scrutiny of the ESI spectrum does reveal a peak for CGCTTTTTTG (Tables 2 and S4 of the Supporting Information) with an area of just under 1% of the largest peak but none of the other potential ions. Accordingly, taking these findings together, we tentatively propose that alkamini also alkylates the C23 phosphate in the T6 strand, in a similar manner to how it alkylates the C11 phosphate in A3T3 (see above), as shown in Figure 3c. Unfortunately, the tandem MS/MS experiment on the SS T6-ligand ion (Figure S12b of

the Supporting Information) has failed to provide evidence for L-pCG, CGCTTTTTTG, or L-OH. The 2:1 alkamini-A6/T6 complexes are revealed by the finding of peaks representing duplex and A6 single-stranded DNAs bearing two ligand molecules (DS A6/T6-2 L, SS A6-2 L), deadenylated A6 DNA bearing a ligand (SS A6-L deA), doubly deadenylated A6 DNA, a deadenylated RHF of A5, and a deadenylated RHF of A6 (Tables 2 and S4 of the Supporting Information and Figures 5a and S10 and S11 of the Supporting Information). Unfortunately, it has proven impossible to isolate the parent DS A6/T6-2 L and SS A6-2 L ions for tandem MS/ MS analysis, which would have helped to define the nature of the 2:1 complexes. However, with the data available, we can confidently assert that the A6 DNA strand bears two ligands, each attached to an adenine, and that one of these complexes has a ligand bound at A5 with another at a 3′-adenine, and that a second complex has a ligand fixed at A6 and another also

154


Abdul Majid et al.

Figure 6. Hydrolysis schemes for phosphotriester fragmentation. Hydroxide ion attack on the phosphorus-oxygen-ligand linkage at the P-O bond (1) results in the release of hydrolyzed ligand, L-OH, and regeneration of intact DNA. Attack on the C3′-O3′-P moiety at the O3′-P bond, 2, causes cleavage of the sugar-phosphate backbone, yielding an intact 5′-DNA fragment terminating in a 3′-hydroxyl, 2(a), and a 3′-DNA fragment bearing a 5′-phosphate group to which the ligand remains attached, 2(b). Attack on the third phosphotriester bond at P-O5′-C5′, 3, yields the complementary cleavage pattern to give a 5′-DNA fragment with a 3′-phosphate group bearing the ligand, (3a), and a 3′-DNA fragment with a 5′-hydroxyl group, (3b). Thus, two of the hydrolytic pathways result in DNA strand breakage, whereas the third leads to reversible loss of the ligand.

bound at a 3′-adenine. Lastly, we note that it is conceivable that some of the 2:1 complexes could also involve G22 since 2:1 complexes with a guanine adduct on each strand would fail to leave a discernible trace at the single-stranded DNA level, and in this regard, we note that the DS A6/T6-2 L ion is compatible with this notion. Structure of the Alkamin-A6/T6 Complex. Species assignments for the alkamin-A6/T6 complex are recorded in Table 2 with a representative MALDI spectrum shown in Figure 5b. The data indicate the formation of a variety of 1:1 and 2:1 monofunctional adducts, as well as both inter- and intrastrand cross-links. With respect to the monofunctional adducts, the spectrum of reaction for alkamin is the same as found for alkamini; thus, alkamin alkylates adenines A4-A9 and guanine G10 on the A6 DNA strand and also guanine G22 on the T6 strand, as shown in Figure 4c. The evidence for these complexes is again found as usual. Thus, the 1:1 complexes are distin-

guished by peaks for SS A6-L-OH, SS T6-L-OH, SS A6 deA, SS T6 deG, the RHFs of adenines A4-A9, the RHF of guanine G10/G22, and the base-ligand adducts A-L-OH and G-L-OH (Tables 2 and S5 of the Supporting Information and Figures 5b and S13 and S14 of the Supporting Information). Observed LHFs are confined to products of reaction with A5 and A6 (Tables 2 and S5 of the Supporting Information), the main species being the doubly dehydrated furan (Figures S13 and S14 and Table S5 of the Supporting Information). These assignments are supported by ESI tandem MS/MS measurements on SS A6-L-OH (m/z 850.7, z ) -5) and SS T6 deG (m/z 696.0, z ) -5) as the parent ions, where the former reveals LHFs for A4, A5, and A6 and RHFs for A7, A8, A9, and G10, as well as a peak for A-L-OH (Figure S15a,b of the Supporting Information), and the latter confirms reaction at G22 (Figure S15b of the Supporting Information). The 2:1 mono adducts are revealed by peaks for both single-stranded A6 DNA and


the full duplex bearing two ligand-OH moieties, SS A6-2(LOH) and DS A6/T6-2(L-OH), and by doubly deadenylated A6 DNA and its precursor SS A6 deA-L-OH (Tables 2 and S5 of the Supporting Information and Figures 5b and S13 and S14 of the Supporting Information). Unfortunately, tandem MS/MS measurements on the SS A6-2(L-OH) ion failed due to sensitivity issues, so the specific identities of the 2:1 complexes remain unknown but clearly involve two adenines on the A6 DNA strand. Once again, we cannot rule out the possibility of the involvement of the T6 strand in these bis mono adducts. With respect to the cross-linked adducts, the interstrand linkage is unequivocally revealed by a peak for the full length T6 DNA bearing a ligand attached to an adenine, SS T6-L-A, which can only have come from depurination of an adenine adduct on the A6 strand (Tables 2 and S5 of the Supporting Information and Figures 5b and S13 and S14 of the Supporting Information). Such a cross-link is supported by the finding of peaks for the intact duplex bearing a bisalkylating alkamin moiety, DS A6/T6)L, and for a guanine-alkamin-adenine adduct, G-L-A, although these latter species can also be derived from intrastrand cross-links (see below). That the cross-link on the T6 strand involves guanine G22 is made clear by a tandem MS/MS measurement on the parental ion SS T6-L-A, which yields daughter ions for the RHF(a) and LHF(a) of G22 and for a deguanylated full length T6 strand (Figure S15c of the Supporting Information). Figure 4c indicates two possible binding modes for the interstrand cross-link, where one arm of alkamin is attached to G22 on the T6 strand and the other is attached to adenines A5 or A6, to give a 1f3 or 1f4 crosslink with the ligand spanning the sequence CAAA. It is conceivable that the cross-link may also extend to the next adenine in the sequence, A7, to give a 1f5 linkage, in which case the binding site would be CAAAA. The hallmark evidence for intrastrand adenine-adenine cross-links, in this setting, includes the peak for single-stranded A6 DNA bearing the bifunctional alkamin moiety, SS A6)L, and the base-ligand adduct A-L-A; supporting evidence includes the duplex species DS A6/T6)L, mentioned above; doubly deadenylated A6 DNA, SS A6 deA deA; and the plethora of RHFs for adenines A4-A9 (Tables 2 and S5 of the Supporting Information and Figures 5b and S13 and S14 of the Supporting Information). Tandem MS/ MS measurements on the parent ion of the intrastrand crosslinked adduct, SS A6)L (m/z 846.9, z ) -5), Figure S15d of the Supporting Information, confirm the involvement of all adenines in the cross-link, there being clear evidence for the LHF(a)s of A4, A5, and A6 and for RHFs of A4, A5, A7, A8, and A9, the RHFs being notable for the variety in their positions of phosphate cleavage. In addition, we see an abundance of the A-L-A base ligand adduct and the appearance of the doubly deadenylated full length A6 DNA. Molecular models indicate that alkamin can form 1f3, 1f4, and 1f5 intrastrand crosslinks but that its mustard groups cannot come close enough to form 1f2 linkages of this type (cf. discussion of interstrand cross-links in ref 13). This suggests that the experimental evidence is consistent with many potential cross-linked species in which the ligand binds in one of three complex classes spanning A4-A8, A5-A9, and A6-A9, as shown in Figure 4c. Lastly, the tandem MS/MS experiment indicates the involvement of guanine G10 in a guanine-ligand-adenine intrastrand cross-link, as confirmed by peaks for the RHF(a) of G10 and for the mixed purine base-ligand adduct A-L-G (Figure S15d of the Supporting Information). Structural considerations imply that this cross-link is of the 1f3, 1f4, or


1f5 type, involving G10 and adenines A6, A7, and A8, as shown in Figure 4c.

Discussion Base Alkylation. The data presented here and in Abdul Majid et al. (13) enable a direct comparison of the modes of binding of alkamin and alkamini to the range of AT-tracts represented by A2T2, A3T3, and A6/T6 DNAs, to which end Figures 3 and 4 also include the binding data for A2T2. In A2T2, alkamini alkylates G4, A5, and A6; in A3T3, it binds to A4, A5, A6, and G10; and in A6/T6, it reacts with A4-A9 and with G10 and G22 (Figure 3). Thus, it is apparent that, across the series, alkamini reacts with all adenines in the AT-tracts and with the guanines of abutting GC base pairs. Reaction with the latter is not restricted to guanines that continue the purine run in one strand, and cross-strand reactions are allowed for alkamin binding, for example. However, there is no evidence among the three DNAs that alkamin or alkamini reacts with any other guanines, indicating the effectiveness of the polybenzamide moiety in directing alkylation specifically to the minor groove, as originally intended (10). Unfortunately, our inability to quantify the relative extent of reaction at each site with confidence (see the discussion in ref 13) precludes a detailed assessment of the effects of the fine details of DNA structure on reactivity within, and adjacent to, the various AT-tracts. Interestingly, it is a universal finding that alkamini also forms 2:1 complexes with these DNAs, notwithstanding the input stoichiometry being one ligand molecule per duplex. Thus, alkylation by alkamini is evidently a cooperative process, so that adduct formation by one ligand enhances reaction of a second. The origins of the cooperativity could lie in the initial reversible binding process, or covalent attachment of the first ligand could cause conformational and/or electrostatic changes in the DNA that enhance the nucleophilicity of the second target purine. Alkamin forms a variety of monofunctional and bifunctional adducts with A2T2, A3T3, and A6/T6, the monofunctional complexes being of the same kind as those of alkamini (Figure 4). Thus, it seems that the addition of the second mustard arm has no effect on the reactivity, or specificity, of reaction of the first. The only material difference between the monofunctional adducts of the two drugs is that alkamin fails to yield complexes of 2:1 stoichiometry with A2T2 and A3T3. Presumably, this is a result of steric effects, so that it is not possible to pack two bulkier alkamin ligands into the 4- or 6-nucleotide AT-tracts. With A6/T6, however, alkamin is able to form 2:1 complexes on the A6 strand, and here, the run of seven purines appears to be sufficiently long to accommodate two ligands. Alkamin forms a single intrastrand cross-link with A2T2 (G4fA6, Figure 4), two such cross-links with A3T3 (A4fA6 and A6fG10, Figure 4), and a multitude of intrastrand linkages with A6/T6 involving all six adenines, as well as a cross-link involving guanine G10 and adenines A6, A7, or A8 (Figure 4). Thus, in the promiscuous manner of alkamini, we find that alkamin takes the opportunity to react bifunctionally with all purines in the same strand of the AT-tract, including the abutting guanines that present themselves in the appropriate geometrical arrangement. In A2T2, this produces a 1f3 linkage; in A3T3, this produces a 1f3 and a 1f5 connectivity; and in A6/T6, the data are consistent with all combinations between 1f3, 1f4, and 1f5 (Figure 4). Interestingly, we find the same general behavior with respect to interstrand cross-links, where, once again, alkamin reacts bifunctionally with all of the purines, abutting guanines included, appropriately dispersed between the two DNA strands

156


within the AT-tract. Thus, in A2T2, we find cross-links between adenines A5, A6, A5′, and A6′ within the 1f2 (equivalent to positions 2f3 at the binding site), 1f3, and 1f4 family of linkages and a 1f4 and 1f5 linkage between guanine G4 and adenines A6′ and A5′ (Figure 4). In A3T3, there are cross-links between adenines A4, A5, A6, A4′, A5′, and A6′ encompassing the full linkage family, plus a guanine G10 to adenine A4′, A5′, or A6′ cross-link; and in A6/T6, there is a single interstrand cross-link between guanine G22 and adenine A4, A5, A6, or A7. Taking the area of the RHF peaks in the tandem spectra as a guide to reactivity, it seems that the 1f4 interstrand crosslinks in A2T2 and A3T3, that is, those between the A5 and the A5′ adenines, are by far the most abundant. While this degree of complexity of reaction with the adenines in AT-tracts may have been anticipated for alkamin, the general finding of extensive involvement of abutting guanines is a surprise. Given the difficulties of quantifying reaction at each site within the AT-tract mentioned above, the most reliable measure of overall reactivity of the site is given by the amount of intact single-stranded DNA remaining at the completion of reaction. By this criterion, for both alkamini and alkamin, the general order of reactivity of the three AT-tracts is A6/T6 > A3T3 > A2T2. This ranking presumably reflects statistical issues associated with the longer AT-tract in the first two examples (thus promoting 2:1 complexes) and may include the effects of an enhanced nucleophilicity of the adenine N3 atoms as a consequence of electrostatic and adenine-adenine stacking interactions. The molecular electrostatic potential (MEP) in the DNA minor groove is known to be greatest in AT-tracts, because of the narrow width of the groove in these sequences and the lack of the positive potential contributed by the 2-amino group of guanine (15). The magnitude of this effect will be AT-tract length dependent because of summation effects (see ref 16). The enhanced negative MEP can be expected to raise the energy of the highest occupied molecular orbital (HOMO) of adenine, which is associated with the N3 lone pair, thus making it more nucleophilic. It is also possible that the HOMO may be elevated directly as a result of adenine-adenine stacking interactions, in an analogous manner to the consequences of orbital mixing observed with guanine in G-tracts (15, 17). The enhanced negative MEP in the longer AT-tracts is also likely to impact the affinity of the reversible binding of the polybenzamide mustards, and this may be reflected in greater covalent reaction as a consequence of a “tighter” precovalent adduct complex. Thus, for this variety of reasons, we can rationalize the observed overall reactivity of the three dodecamers, although it is not possible to indicate which, if any, is the most dominant effect. Phosphate Alkylation. Alkamini, but not alkamin, shows evidence for alkylation of a phosphate group in A3T3 and A6/ T6 DNAs with formation of a phosphotriester. Phosphotriester formation by nitrogen mustards is usually a feature of more reactive alkylating agents such as ethyl and chloroethyl nitrosoureas (18), ethyl and diethyl sulfate (19), 3-(2-haloethyl)aryltriazenes (20), nitroimidazole aziridines such as RSU-1069 (21), and the aziridinylpyrrolo-benzimidazoles (22). Phosphotriesters are stable in acid solution but are vulnerable to hydrolysis in alkaline conditions (23). Figure 6 shows the expected hydrolytic products of a DNA fragment bearing an alkylated phosphate. Hydroxide ion attack on the phosphorus-oxygen-ligand linkage at the P-O bond, (1), results in the release of hydrolyzed ligand, L-OH, and regeneration of the intact DNA. Attack on the C3′-O3′-P moiety at the O3′-P bond, 2, causes cleavage of the sugar-phosphate backbone, yielding an intact 5′-DNA fragment terminating in a 3′-hydroxyl,

Abdul Majid et al.

2(a), and a 3′-DNA fragment bearing a 5′-phosphate group to which the ligand remains attached, 2(b). Attack on the third phosphotriester bond at P-O5′-C5′, 3, yields the complementary cleavage pattern to give a 5′-DNA fragment with a 3′-phosphate group bearing the ligand, (3a), and a 3′-DNA fragment with a 5′-hydroxyl group, (3b). Thus, two of the hydrolytic pathways result in DNA strand breakage, whereas the third leads to reversible loss of the ligand. In principle, all three pathways are available to a given phosphotriester, but in practice, the fragmentation pattern is expected to be dominated by the weakest P-O bond. The latter is likely to depend on the nature of the ligand and possibly the DNA sequence. In the case of alkamini and A3T3 and A6/T6 DNAs, the evidence suggests that the phosphotriesters hydrolyze principally via attack on the C3′-O3′-P moiety since we see the equivalent of fragments (2a) and (2b) in the MALDI spectra [CGCAAATTTG3′-OH and L-pCG (Tables 1 and S2 of the Supporting Information and Figures 2a and S4 and S5 of the Supporting Information) and CGCTTTTTTG3′-OH and L-pCG (Tables 2 and S4 of the Supporting Information and Figures 2a and S10 and S11 of the Supporting Information)]. By contrast, there is no evidence for reaction at P-O5′-C5′, but the presence of L-OH suggests phosphotriester reversal, although the latter may also have its origins in simple hydrolysis of unreacted ligand. A characteristic feature shared by the two phosphotriester complexes of alkamini is that the ligand alkylates both the terminal guanine and its 3′ phosphate, which provides a 3′-end to an extended AT-tract in the sequence 5′-TTTGC (see Figure 2b,c). In this sequence, the polybenzamide backbone will lie in the narrow minor groove extending in the 5′-direction over the runs of thymines, and it is notable that within this particular DNA strand, the mustard group has no access to adenines. The crystal structures of A3T3 and A6/T6 DNAs (24, 25) show that an oxygen atom of the 3′-phosphate of the guanine is in close juxtaposition, about 4 Å, to the guanine N3 nitrogen, thus indicating that an alkamini molecule, positioned as described, is ideally situated to alkylate both base and phosphate by rotating the mustard from the bottom of the groove toward the sugar-phosphate backbone. This maneuver could be achieved by rotating around the mustard nitrogen-benzene ring bond, or about the benzamide phenyl-amide NH bond (see Figure 1), and would be assisted by the fact that in the GC segment the minor groove width has expanded from a value of about 10 Å in the preceding AT-tract to about 12.5 Å, thereby providing space for the postulated flexions. Thus, it would seem that, within the alkamini class of targeted mustard, sequences such as (T)nGC, where n g 3, are potential sites for phosphate alkylation. We see no obvious reasons why alkamin should not also form such complexes, perhaps appearing as interstrand cross-links involving an adenine, but we find no evidence for their formation, even at higher input stoichiometries (data not shown). Comparison with Other Work and Biological Implications. Our findings for the reactivity of alkamin and alkamini with A3T3 and A6/T6 DNAs are consistent with the sequencing studies of Turner et al. (12) who identified AAAAA and its derivative AmTn-tracts, many having the capacity for adenine-adenine interstrand cross-linking, as the most reactive sites. Such sequencing studies are not able to provide direct evidence for cross-linking, whether interor intrastrand; nevertheless, Turner et al. (12) concluded, by comparison of reactivity in complementary strands, that interstrand cross-link frequencies at such sites are very low


for alkamin. In contrast, the mass spectrometry method provides unequivocal evidence for both forms of cross-link at base pair resolution and shows that alkamin forms a complex variety of cross-links at AT-tracts in high yield. Perhaps the most surprising finding is that guanines abutting the AT-tracts are also reactive, and this results in unexpected intrastrand and interstrand cross-links at these sequences. Thus, for example, CAmG-tracts (cf. A6/T6) would appear to offer novel sites for interstrand cross-linking as well as the expected sites for intrastrand cross-links, and in fact, the only sequences that would not permit interstrand cross-linking would be GAmG. Taken together, this work and that of Abdul Majid et al. (13) illustrate that positioning monofunctional mustard groups on either end of a polybenzamide MGB ligand is an excellent strategy for enhancing the formation of interstrand cross-links at all manner of AT-tracts, including most in which the adenines are all in one strand. The importance of cross-links to biological activity is made clear in the potent cytotoxic and experimental antitumor activity of alkamin, properties that the monofunctional alkamini lacks (10, 11). Acknowledgment. We acknowledge financial support from University Science Malaysia (A.M.S.A.M.), the Australian National Health and Medical Research Council (L.P.G.W.), and the Auckland Division of the Cancer Society of New Zealand (W.A.D.). Mass spectrometry was carried out at the Bioanalytical Mass Spectrometry Facility, UNSW, and was supported in part by grants from the Australian Government Systemic Infrastructure Initiative and Major National Research Facilities Program (UNSW node of the Australian Proteome Analysis Facility). We thank Dr. Mark Raftery for technical discussions concerning the mass spectrometers used. Supporting Information Available: Main chemical processes leading to ligand-induced fragmentation of DNA (Figures S1 and S2) and the abbreviations used to describe the observed species (Table S1). Tandem MS/MS gas-phase degradation pathways used to identify key molecular entities (Figure S3). Full MALDI and ESI spectra for the alkamini-A3T3 complex (Figures S4-S6) and the corresponding spectra for the alkamin-A3T3 complex (Figures S7-S9). MALDI and ESI spectra for the alkamini-A6/T6 complex (Figures S10-S12) and the correspondening data for the alkamin-A6/T6 complex (Figures S13-S15). Species assignments for the alkamini- and alkamin-A3T3 complexes (Tables S2 and S3) and the assignments for the A6/T6 complexes (Tables S4 and S5). This material is available free of charge via the Internet at http:// pubs.acs.org.


(6)

(7)

(8)

(9) (10)

(11)

(12) (13)

(14)

(15)

(16) (17) (18) (19) (20) (21)

References (1) Gamcsik, M. P., Dolan, M. E., Andersson, B. S., and Murray, D. (1999) Mechanisms of resistance to the toxicity of cyclophosphamide. Curr. Pharm. Des. 5, 587–605. (2) Panasci, L., Paiement, J. P., Christodoulopoulos, G., Belenkov, A., Malapetsa, A., and Aloyz, R. (2001) Chlorambucil drug resistance in chronic lymphocytic leukemia: The emerging role of DNA repair. Clin. Cancer Res. 7, 454–461. (3) Drablos, F., Feyzi, E., Aas, P. A., Vaagbo, C. B., Kavli, B., Bratlie, M. S., Pena-Diaz, J., Otterlei, M., Slupphaug, G., and Krokan, H. E. (2004) Alkylation damage in DNA and RNAsRepair mechanisms and medical significance. DNA Repair 3, 1389–1407. (4) Denny, W. A. (2001) DNA minor groove alkylating agents. Curr. Med. Chem. 8, 533–544. (5) Gourdie, T. A., Valu, K. K., Gravatt, G. L., Boritzki, T. J., Baguley, B. C., Wakelin, L. P., Wilson, W. R., Woodgate, P. D., and Denny, W. A. (1990) DNA-directed alkylating agents. 1. Structure-activity

(22) (23) (24)

(25)

relationships for acridine-linked aniline mustards: Consequences of varying the reactivity of the mustard. J. Med. Chem. 33, 1177– 1186. Gravatt, G. L., Baguley, B. C., Wilson, W. R., and Denny, W. A. (1991) DNA-directed alkylating agents. 4. 4-Anilinoquinoline-based minor groove directed aniline mustards. J. Med. Chem. 34, 1552– 1560. Weiss, G. R., Poggesi, I., Rocchetti, M., DeMaria, D., Mooneyham, T., Reilly, D., Vitek, L. V., Whaley, F., Patricia, E., Von Hoff, D. D., and O’Dwyer, P. (1998) A phase I and pharmacokinetic study of tallimustine [PNU 152241 (FCE 24517)] in patients with advanced cancer. Clin. Cancer Res. 4, 53–59. Arcamone, F. M., Animati, F., Barbieri, B., Configliacchi, E., D‘Alessio, R., Geroni, C., Giuliani, F. C., Lazzari, E., Menozzi, M., and Mongelli, N. (1989) Synthesis, DNA-binding properties, and antitumor activity of novel distamycin derivatives. J. Med. Chem. 32, 774–778. Wemmer, D. E., and Dervan, P. B. (1997) Targeting the minor groove of DNA. Curr. Opin. Struct. Biol. 7, 355–361. Prakash, A. S., Valu, K. K., Wakelin, L. P., Woodgate, P. D., and Denny, W. A. (1991) Synthesis and anti-tumour activity of the spatially-separated mustard bis-N,N′-[3-(N-(2-chloroethyl)-N-ethyl)amino-5-[N,N-dimethylamino)methyl)-aminophenyl]-1,4-benzenedicarboxamide, which alkylates DNA exclusively at adenines in the minor groove. Anti-Cancer Drug Des. 6, 195–206. Atwell, G. J., Yaghi, B. M., Turner, P. R., Boyd, M., O’Connor, C. J., Ferguson, L. R., Baguley, B. C., and Denny, W. A. (1995) Synthesis, DNA interactions and biological activity of DNA minor groove targeted polybenzamide-linked nitrogen mustards. Bioorg. Med. Chem. 3, 679–691. Turner, P. R., Ferguson, L. R., and Denny, W. A. (1999) Polybenzamide mustards: Structure-activity relationships for DNA sequencespecific alkylation. Anti-Cancer Drug Des. 14, 61–70. Abdul Majid, A. M., Smythe, G. A., Denny, W. A., and Wakelin, L. P. (2007) Structure of the d(CGCGAATTCGCG)2 complex of the minor grove binding alkylating agent alkamin studied by mass spectrometry. Mol. Pharmacol. 71, 1165–1178. Gourdie, T. A., Valu, K. K., Gravatt, G. L., Boritzki, T. J., Baguley, B. C., Wakelin, L. P., Wilson, W. R., Woodgate, P. D., and Denny, W. A. (1990) DNA-directed alkylating agents. 1. Structure-activity relationships for acridine-linked aniline mustards: consequences of varying the reactivity of the mustard. J. Med. Chem. 33, 1177– 1186. Pullman, B., Perahia, D., and Cauchy, D. (1979) The molecular electrostatic potential of the B-DNA helix. VI. The regions of the base pairs in poly (dG.dC) and poly (dA.dT). Nucleic Acids Res. 6, 3821– 3829. Dean, P. M., and Wakelin, L. P. (1980) Electrostatic components of drug-receptor recognition. II. The DNA-binding antibiotic actinomycin. Proc. R. Soc. London, Ser. B 209, 473–487. Hartley, J. A., Gibson, N. W., Kohn, K. W., and Mattes, W. B. (1986) DNA sequence selectivity of guanine-N7 alkylation by three antitumor chloroethylating agents. Cancer Res 46, 1943–1947. Carter, C. A., Kirk, M. C., and Ludlum, D. B. (1988) Phosphotriester formation by the haloethylnitrosoureas and repair of these lesions by E. coli BS21 extracts. Nucleic Acids Res. 16, 5661–5672. Sun, L., and Singer, B. (1975) The specificity of different classes of ethylating agents toward various sites of HeLa cell DNA in vitro and in vivo. Biochemistry 14, 1795–1802. Lown, J. W., and Singh, R. (1982) Mechanism of action of antitumor 3-(2-haloethyl) aryltriazenes on deoxyribonucleic acid. Biochem. Pharmacol. 31, 1257–1266. Silver, A. R., O’Neill, P., and Jenkins, T. C. (1985) Induction of DNA strand breaks by RSU-1069, a nitroimidazole-aziridine radiosensitizer. Role of binding of both unreduced and radiationreduced forms to DNA, in vitro. Biochem. Pharmacol. 34, 3537– 3542. Skibo, E. B., and Schulz, W. G. (1993) Pyrrolo[1,2-a]benzimidazolebased aziridinyl quinones. A new class of DNA cleaving agent exhibiting G and A base specificity. J. Med. Chem. 36, 3050–3055. Shooter, K. V. (1976) The kinetics of the alkaline hydrolysis of phosphotriesters in DNA. Chem.-Biol. Interact. 13, 151–163. Woods, K. K., Maehigashi, T., Howerton, S. B., Sines, C. C., Tannenbaum, S., and Williams, L. D. (2004) High-resolution structure of an extended A-tract: [d(CGCAAATTTGCG)]2. J. Am. Chem. Soc. 126, 15330–15331. Nelson, H. C., Finch, J. T., Luisi, B. F., and Klug, A. (1987) The structure of an oligo(dA).oligo(dT) tract and its biological implications. Nature 330, 221–226.

TX800276H

Mass Spectrometry Studies of the Binding of the Minor Groove

Recommend Documents