Directing Quinone Methide-Dependent Alkylation and Cross-Linking of Nucleic Acids With Quaternary Amines.

Abstract

Polyamine and polyammonium ion conjugates are often used to direct reagents to nucleic acids based on their strong electrostatic attraction to the phosphoribose backbone. Such non-specific interactions do not typically alter the specificity of the attached reagent, but polyammonium ions dramatically redirected the specificity of a series of quinone methide precursors. Replacement of a relatively non-specific intercalator based on acridine with a series of polyammonium ions resulted in a surprising change of DNA products. Piperidine stable adducts were generated in duplex DNA that lacked the ability to support a dynamic cross-linking observed previously with acridine conjugates. Minor reaction at guanine N7, the site of reversible reaction, was retained by a monofunctional quinone methide-polyammonium ion conjugate but a bisfunctional analogue designed for tandem quinone methide formation modified guanine N7 in only single-stranded DNA. The resulting intrastrand cross-links were sufficiently dynamic to rearrange to interstrand cross-links. However, no further transfer of adducts was observed in duplex DNA. An alternative design that spatially and temporally decoupled the two quinone methide equivalents neither restored the dynamic reaction nor cross-linked DNA efficiently. While di- and triammonium ion conjugates successfully enhanced the yields of cross-linking by a bisquinone methide relative to a monoammonium equivalent, alternative ligands will be necessary to facilitate the migration of cross-linking and its potential application to disrupt DNA repair.

Graphical Abstract

INTRODUCTION

Many cancer treatments depend on low molecular weight compounds that alkylate and cross-link DNA with little selectivity. Despite the advent of drugs that can target specific enzymes and receptors, these nonspecific treatments remain widely prescribed. To enhance the efficacy of nucleobase modification, reactive components such as nitrogen mustards have been conjugated to intercalators or other ligands with general affinity to duplex DNA.^1,2 Additional specificity can be achieved by conjugating reactive species to ligands that recognize specific nucleotide sequences through association with the major or minor grooves of helical DNA.^3–6 Typically, the appendages conferring covalent chemistry are selected on the basis of their mild reactivity to ensure sufficient time for target binding prior to reaction.⁷ To avoid this constraint, reagents may instead be chosen for their inducible reactivity as illustrated by the photochemical activation of psoralen.⁸ Perhaps most valuable are alternative reagents that are activated by the very environment to which they are targeted. Only limited examples of these are known and include an aziridine-containing analogue of cytosine,⁹ cyclopropapyrroloindole and related species,¹⁰ and derivatives of 2-amino-6-vinylpurine.^11,12 A complementary approach for selective alkylation and cross-linking derives from the reversible covalent chemistry of certain quinone methide intermediates (Scheme 1A).¹³ The dynamics of these systems support intra- to interstrand transfer of self-adducts without need of an external method of activation (Scheme 1B).^14–17

Scheme 1.

(A) Reversible quinone methide reaction supports (B) sequence specific alkylation and (C) cross-link migration within duplex DNA.

Quinone methides (QMs) and related azaquinone methides are highly electrophilic and often quite transient and yet still very useful in the synthesis of natural products and new materials,^18–20 modification of proteins,^21–23 and design of releasable linkers.^24–26 Many new opportunities based on the versatile properties of QMs have also been projected for the future.²⁷ Compounds that generate QM under metabolic conditions are additionally under investigation for their chemoprotective and anticancer properties.^28–30 Quinone methide precursors (QMPs) can be designed for induction under a wide range of conditions including photochemistry,^22,31–33 oxidation,^29,34 reduction,^26,35 and hydrolysis,^24,28,30 Our laboratory has conjugated QMPs to ligands directed to single-stranded DNA,^{14–16,36–38} as well as the major^17,39,40 and minor⁴¹ grooves of duplex DNA and noncanonical structures.⁴² Others have extended the use of QMs to target RNA.⁴³ Additionally, the reversible chemistry of quinone methides allows for oligonucleotide and protein nucleic acid self-adducts to transfer their QM spontaneously to a bound target without need of an activation step.^14–17,44 In the absence of a target, QM conjugates rapidly reform their self-adducts and avoid trapping by competing nucleophiles.

The dynamic nature of QM reaction with nucleic acids also supports the unusual ability of cross-links to regenerate during strand exchange and to migrate along duplex DNA (Scheme 1C) as illustrated previously with a bisquinone methide-generating acridine conjugate (bisQMP-Acr, Scheme 2).^45,46 These features have the potential to extend the lifetime of their cross-links and confound DNA repair processes more effectively than lesions generated by irreversible reaction. However, the kinetics of cross-link migration by this conjugate are currently too slow for cellular application. A random walk of a cross-link by Brownian motion may only traverse 9 base pairs over 7 days despite the relatively short half-life of the guanine N7-QM bond (< 2.5 h) that predominates in reaction between duplex DNA and bisQMP-Acr.^39,45,47,48 For this example, mobility may be limited by the attached acridine that likely requires substantial conformational reorganization prior to QM migration. However, acridine or another DNA ligand is not easily avoided since the bisQMP derivatives lacking DNA affinity react with very low efficiency.^39,49

Scheme 2.

Substitution of acridine with alkyl ammonium ions in a bisquinone methide conjugate (bisQMP) to switch DNA binding modes.

As reported here, the acridine component of bisQMP-Acr has now been replaced with polyammonium ions to maintain an affinity for DNA but minimize the conformational barrier for QM migration. The high affinity of polyamines for DNA is primarily based on strong electrostatic interactions that are non-specific and do not impede the mobility of the amines on the surface of duplex DNA.^50,51 Prior success with polyamines and ammonium derivatives was also encouraging. Polyamine conjugates of a nitrogen mustard enhanced alkylation efficiency by 3- to 4-orders of magnitude and did not affect target specificity.⁵² Similarly, addition of an alkylammonium group to CC-1065 CBI analogues increased alkylation yields by 100-fold without necessarily changing specificity.⁵³ As described below, polyammonium ions were capable of achieving similar cross-linking efficiency as that generated by bisQMP-Acr but, in contrast to the precedence above, the sites of reaction were altered dramatically.

RESULTS AND DISCUSSION

The nascent bisfunctional QMP (bisQMP) was retained for the polyammonium ion conjugates since the transient exo-methylene groups at the 2- and 6-position generated high yields of DNA cross-linking in the analogous acridine conjugate (Scheme 2).³⁹ To identify the optimum number of ammonium groups for our applications, the mono-, di- and triammonium species were generated. An obvious starting material was tyramine but this and its Boc-protected derivative were not compatible with conditions for hydroxymethylation of the phenol. Hydroxymethylation of hydroxyphenethyl bromide was also unsuccessful. Instead, 3-(4-hydroxyphenyl)propanoic acid (1, Scheme 3) was selected as the common starting material for the target bisQMPs.

Scheme 3.

General strategy for preparing a bisquinone methide precursor conjugated to an alkylammonium group: (a) NaOH, formaldehyde, (b) tert-butyldimethylsilyl chloride (TBDMS-Cl), (c) BH₃/THF, (d) methanesulfonyl chloride (MsCl), (e) (CH₃)₂NH, Et₃N, (f) para-toluenesulfonic acid (TsOH), (g) AcCl, Et₃N and (h) MeI.

Preparation of conjugates.

Hydroxymethylation and silyl protection to form the tri-tert-butyldimethylsilyl (TBDMS) derivative (2) followed the same procedure developed for bisQMP-Acr.³⁹ Borane reduction of 2 provided a primary alcohol that was subsequently activated for substitution by addition of a mesyl group (3). Treatment of 3 with dimethylamine generated the desired intermediate but standard substitution of the benzylic silyl ethers with acetate using FeCl₃^39,54 was complicated by the presence of the amine. This issue was further exacerbated for the di- and triammonium products. In a complementary approach, mild treatment with para-toluenesulfonic acid (TsOH) selectively hydrolyzed the benzylic TBDMS groups to yield 4.⁵⁵ Final acetylation and methylation was accomplished with acetyl chloride/triethylamine and methyl iodide respectively to form bisQMPN₁. Synthesis of bisQMPN₂ followed the same strategy except N,N,N’-trimethylethyldiamine replaced dimethylamine (Scheme S2). A similar procedure was not possible for bisQMPN₃ since the analogous triamine could react at both the terminal and central nitrogens. As an alternative, 3 was treated with 3-methylamino-1-propanol and the product’s primary alcohol was then converted to its mesylate ester before substitution with N,N,N’-trimethylethyldiamine (Scheme S3). The remaining procedures followed those used for generating bisQMPN₁ and bisQMPN₂. Experimental details and physical characterization are included in the supporting information.

Efficiency of DNA cross-linking with bisQMPN_x (x = 1, 2, 3).

Previous reports on nitrogen mustard conjugates containing polyamines did not suggest that intramolecular reaction diminished DNA modification.^2,52 Accordingly, the amino derivatives of bisQMPN_x (x=1,2) were evaluated for their cross-linking ability prior to the final methylation step. Interference by the amines was not anticipated since these should be fully protonated by protonation under conditions used for cross-linking (pH 7).⁵¹ Surprisingly, only low yields of DNA cross-linking were observed after treatment with as much as 250 μM of the monoamine derivative (Figure S1) and a general loss of sample was observed when the monoamine concentration was further increased. Performance of the diamine was even worse. No cross-linking was apparent after treatment with 50 μM of the diamine and higher concentrations (≥150 μM) only generated aggregates with little gel mobility (Figure S1). Thus, the quarternary alkylammonium derivatives were prepared for subsequent study to block reaction of the polyamine.

The relative efficiency of each conjugate was compared using a range of concentrations (0 – 500 μM) to cross-link a model DNA duplex (Figure 1). QM generation was initiated by TBDMS deprotection using fluoride and the yield of cross-linking was determined by the ratio of slow migrating DNA relative to the total signal in each lane of a denaturing polyacrylamide gel. Studies with bisQMP-acr had previously confirmed a strict requirement for such deprotection since addition of chloride in place of fluoride generated no DNA products.³⁹ The monocationic derivative, bisQMPN₁, yielded very low amounts of cross-linking. A maximum of 16% was formed after an extended incubation (8 h) with a high concentration of bisQMPN₁ (500 μM). The activity of this derivative was not significantly better than that of bisQMPs lacking affinity for DNA.^39,49 In contrast, the model duplex was fully consumed in 2 h after incubation with 100 μM of the alkyl diammonium derivative bisQMPN₂ (Figure 1). Aggregation of the DNA was also evident after treatment with higher concentrations of bisQMPN₂. The subtle and continuous loss of gel mobility relative to the parent strand is typical of monoalkylation and distinct from the discontinuity in migration caused by cross-linking as described previously with bisQMP-Acr.³⁹

Figure 1. DNA cross-linking by ammonium-bisquinone methide conjugates.

(A) Nucleotide sequences of OD1 and OD2 and scheme to assess cross-linking efficiency (symbolized by the black lines). (B) Reaction was initiated by addition of the indicated QMP in CH₃CN (final concentration of 20%) to 5’-[³²P]-OD1 (2.8 μM) and OD2 (3.0 μM) in 10 mM MES pH 7 and 10 mM NaF. Samples were incubated under ambient conditions for either 8 h (bisQMPN₁) or 2 h (bisQMPN₂ and bisQMPN₃). Products were separated by denaturing polyacrylamide gel electrophoresis (PAGE, 20%) and detected by phosphoimagery. Yields of cross-linking are reported (%) relative to total [³²P] per lane.

The alkyl triammonium derivative was most active and fully consumed the DNA duplex after 2 h in the presence of 50 μM bisQMPN₃. Nearly full cross-linking was achieved in this same period with 100 μM bisQMPN₃ (Figure 1). At concentrations greater than 100 μM, the DNA aggregated and had little mobility in the polyacrylamide gel. Time-dependence of reaction was also monitored for the di- and triammonium conjugates (100 μM) and revealed that bisQMPN₂ and bisQMPN₃ approached their maximum cross-linking of ~ 75% and 100% respectively after 4 h (Figure S2). The yield after 2 h in this case showed only 75% cross-linking for bisQMPN₃ rather than the 98% evident in Figure 1 and illustrates the potential variability within each examination. Qualitatively, bisQMP-Acr and bisQMPN₂ exhibit similar efficiencies of cross-linking and only bisQMPN₃ achieves full cross-linking.³⁹ In this example, the alkyl di- and triammonium conjugates exhibited an enhanced activity expected from the precedence of comparable nitrogen mustard conjugates.⁵² Unlike the mustard conjugates, however, the amines of bisQMPN_x required full protection as their quaternary alkyl ammonium ions to prevent aggregation. No further studies were pursued with the inefficient monoammonium conjugate bisQMPN₁ but characterization of the other two conjugates was continued to determine the dynamics of their DNA cross-links. Reaction overall seemed to benefit from the expected increase in general affinity between the cationic conjugates and anionic DNA. However, binding constants were not determined since they would only reflect the composite interactions and not necessarily reflect the sub-population that is positioned in a productive orientation for alkylation of the DNA.

Intra- to interstrand migration of cross-links generated by bisQMPN_x (x = 2, 3).

In general, QMs react efficiently with DNA, but the most notable feature is the reversibility of their covalent attachment to certain sites within the nucleobases. This reversibility allows the QM adducts to redistribute in response to changes in reaction conditions and components.^44–46 The dynamics of bisQMPN₂ and bisQMPN₃ were first examined by their proficiency to transfer cross-links from intra- to interstrand configurations (Figure 2A). Similar to studies with bisQMP-Acr, the polyammonium ion conjugates were initially incubated with single-stranded DNA (OD3) for 24 h.^44–46 This period is sufficient to consume all bisQMP and generate a mixture of DNA alkylation, cross-linking and hydrolyzed derivatives. Only the fraction forming dynamic cross-links is then capable of the subsequent QM transfer to form interstrand cross-links. This latter process was monitored over time by addition of a labeled acceptor strand ([³²P]-OD4) that complements the donor strand. For bisQMPN₂, a slow transfer from intra- to interstrand cross-linking was observed over 72 h by the generation of a species with low mobility during PAGE (Figure 2B). In this period, only 10% of the target strand was cross-linked. Cross-link transfer by an equivalent migration of bisQMPN₃ was more efficient and produced a 34% yield after 72 h (Figure 2B). These experiments relied on a labeled acceptor strand to minimize handling of radioactivity and comparable results can be expected if the donor strand had been radiolabeled based on prior investigations with bisQMP-acr.⁴⁶

Figure 2. Trapping a reversible quinone methide by intrastrand cross-linking and its subsequent transfer to form interstrand cross-linking.

(A) Nucleotide sequences of OD3 and OD4 and scheme to assess efficiency of intra- to interstrand transfer of cross-linking (symbolized by the black lines) through reversible regeneration of a quinone methide intermediate. (B) The indicated bisQMP (100 μM) in CH₃CN (final concentration of 20%) was added to OD3 (3 μM) in 10 mM MES pH 7 with 10 mM NaF and maintained under ambient conditions for 24 h. 5’-[³²P]-OD4 was then added and incubation was extended for the specified time before analysis by denaturing PAGE (20%) and detection by phosphoimagery. Yields of cross-linking are reported (%) relative to total [³²P] per lane.

These polyammonium conjugates did not perform as well as the original bisQMP-Acr that supported complete intra- to interstrand migration over a shorter period of 24 h.⁴⁵ More importantly, neither bisQMPN₂ nor bisQMPN₃ generated cross-links that remained dynamic in duplex DNA. Previously, cross-links formed by an acridine conjugate had demonstrated their reversibility in two assays, one involved the equivalent of exchanging OD3 with a full complement of OD4. The second involved adding a third oligonucleotide to base pair with the remaining single-stranded region of OD4 that extends the duplex structure to accept the mobile cross-link. No migration of the original interstrand cross-link formed alternatively by bisQMPN₂ and bisQMPN₃ was observed with these two strategies (Figures S3 and S4). Thus, the cross-link formed on single-stranded DNA persisted in a dynamic state for transfer to an interstrand complex but the subsequent interstrand cross-link was static and non-transferable. Such limitations of QM migration had previously been overcome by enhancing the rate of QM regeneration with electron rich substituents.^17,45,56 Accordingly, this same approach was tested with the polyammonium ion conjugates.

Enhancing the dynamics of the bisQMPN_x conjugates by substituent effects.

QMs are intrinsically electron deficient intermediates and are highly sensitive to substituent effects.^48,57,58 Electron withdrawing groups attached to the QM suppress the rate of its formation and activate the rate of its consumption. Conversely, electron donating groups have the opposite effect and may extend QM lifetime. Replacement of a methylene linker to a QM with a more electron donating ether had previously been necessary before an oligonucleotide-QM self-adduct successfully alkylated a target duplex via triplex formation and before a bisQM-based cross-link migrated through duplex DNA.^17,45 The original bisQMPN₂ and bisQMPN₃ were therefore redesigned with an ether linker to the QM for stimulating their dynamics of cross-linking (ebisQMPN₂ and ebisQMPN₃, Scheme 4A).

Scheme 4.

(A) Electron-rich quinone methide precursors. (B) Preparation of a common synthetic intermediate used to generate the electron rich quinone methides: (a) NaOH, formaldehyde, (b) TBDMS-Cl, (c) BH₃/THF, (d) MsCl.

The design and synthesis of these electron rich species followed an earlier strategy used to convert bisQMP-Acr into its ether-linked derivative.⁴⁵ Both began with hydroxymethylation followed by silyl protection of 2-(4-hydroxyphenoxy)acetic acid (5) to generate 6 (Scheme 4B). The carboxylic acid was reduced by borane and the resulting alcohol was esterified with mesyl chloride to provide the common intermediate 7 for both ebisQMPN₂ and ebisQMPN₃. For the alkyl diammonium derivative ebisQMPN₂, the mesyl group of 7 was displaced with N,N,N’-trimethylethyldiamine and then elaborated under conditions analogous to those used to transform 3 to bisQMPN₂ (Scheme S5). Similarly, 7 was converted to ebisQMPN₃ following the protocol for 3 to bisQMPN₃ (Scheme S6). Experimental details are again provided in the supporting information.

Trends in the efficiency of cross-linking duplex DNA were similar for both pairs of bisQMPs. For each example, the triammonium conjugates produced greater yields of cross-linking at lower concentrations relative to the diammonium conjugates. More specifically, a cross-linking yield of 90% was achieved with 100 μM of ebisQMPN₃ but similar results required 250 μM of ebisQMPN₂ (Figure 3). No aggregation was detected after treatment with 250 μM of these conjugates in contrast to the original designs of bisQMPN₂ and bisQMPN₃ (Figure 1). High concentration of the electron rich conjugates likely generated multiple modifications per strand as evident from the progressively slower migration of products during PAGE. Results for the electron rich species illustrated in Figure 3 represent final yields of cross-linking based on a 24 h incubation in contrast to Figure 1 that represents the initial formation of cross-links after 2 h with the original bisQMPN₂ and bisQMPN₃. Unexpectedly, the electron rich conjugates ebisQMPN₂ and ebisQMPN₃ generated cross-links at approximately 50% of the rate of the original conjugates (Figures S2 versus S5). In contrast, the electron rich derivative of bisQMP-Acr cross-linked duplex DNA approximately 4-fold more rapidly than its parent bisQMP-Acr as predicted.⁴⁵ These unanticipated observations may in part be due to differences in the target sequences and the nucleophiles participating in reaction. The strongest nucleophiles of DNA may effectively compete with surrounding water for coupling to the acridine conjugates but the weaker nucleophiles targeted by the polyammonium ion conjugates (see below) may compete less efficiently. The resulting quenching may be further exacerbated by the enhanced rate of QM formation from the electron rich derivatives or even possibly by the enhanced rate of their regeneration from adducts that could be too transient to detect by PAGE.

Figure 3. DNA cross-linking by electron-rich alkyl ammonium-bisquinone methide conjugates.

(A) Nucleotide sequence of duplex OD4/OD5. (B) Reaction was initiated by addition of the indicated QMP in CH₃CN (final concentration of 20%) to 5’-[³²P]-OD4 (2.8 μM) and OD5 (3.0 μM) in 10 mM MES pH 7 and 10 mM NaF and maintained under ambient conditions for 24 h. Products were separated by denaturing polyacrylamide gel electrophoresis (PAGE, 20%) and detected by phosphoimagery. Yields of cross-linking are reported (%) relative to total [³²P] per lane.

Once the full quenching of ebisQMPN₂ and ebisQMPN₃ was confirmed to require less than 8 h in the absence of DNA (Figure S6), migration of cross-links from intra- to interstrand DNA could be monitored in analogy to that illustrated in Figure 2 for bisQMPN₂ and bisQMPN₃. No intra- to interstrand cross-link transfer was evident after addition of OD4 to a pre-incubated mixture of OD5 and ebisQMPN₂. Only a slight yield of QM transfer (9%) was evident after 24 from an equivalent pre-incubation of OD5 and ebisQMPN₃ (Figure S7). Despite the paucity of interstrand cross-linking, the labeled acceptor sequence (OD4) did progressively lose mobility in a manner that is consistent with one or more alkylations.³⁹ These results could again reflect an enhanced competition by water to quench the transient QM and possibly enhance isomerization of reversible to non-reversible adducts. Additionally, the polyammonium ion conjugates may not secure the QM within the duplex to the extent supported by acridine. Consequently, the QM intermediates may be more readily accessible to quenching by water. Such a result was recently observed when a diffusible QM conjugate was activated by an electron-donating substituent.⁵⁹

Nucleobase targets of ebisQMPN₃.

The parent ortho QM reacts at many of the nitrogens within the nucleobases of DNA and generates an initial profile of kinetic and reversible products including adducts of guanine N7 and adenine N1. These species then dissipate over time to reveal a set of stable and irreversible products formed in low yields by reaction with the exocyclic purine amines.^47,60 Still, these profiles are highly susceptible to manipulation by conjugation with DNA-binding ligands. For example, adducts formed at guanine N7 dominate and persist when certain QMs are held by acridine and triplex-forming oligonucleotides in the major groove of duplex DNA.^39,40 Reaction at guanine N7 and adenine N1 is also crucial to maintain the reversibility of alkylation for cross-link migration since these sites represent strong nucleophiles for QM capture and stable nucleofugal groups for QM regeneration.⁴⁷ The use of polyammonium ions reported here was chosen in part to replace acridine with the expectation that sites of reaction would remain unperturbed based on previous studies with nitrogen mustards.⁵² The validity of this extrapolation to QMs was examined by characterizing the nature of DNA products formed by ebisQMPN₃. More specifically, piperidine fragmentation of cross-linked DNA was used to assess the extent of reaction at guanine N7.

The parent duplex OD4:OD5 was fully converted to its cross-link product after a 24 h incubation with ebisQMPN₃ (50 μM) and this conversion was confirmed by its diminished mobility during PAGE (Figure 4). An equivalent sample was then treated with piperidine and no strand scission generating fragments of the parent strand was detectable above background. Thus, no guanine N7 adducts were likely formed. Similar analysis for samples incubated for short periods also revealed no evidence for guanine N7 reaction and indicated that products of guanine N7 reaction did not even accumulate transiently during the formation of cross-links. The lack of such adducts thus explains the lack of dynamic exchange and migration of the polyammonium ion conjugates. Piperidine treatment did, however, transform the heterogeneous cross-linked products to distinct species migrating slightly slower than the parent strand. These products are consistent with the formation of irreversible adducts containing progressively more polyammonium groups that retard gel mobility by partially neutralizing of the polyanionic DNA. Thus, the polyammonium ions direct ebisQMPN₃ to sites in DNA that lack the reversibility to support QM migration (such as the exo amino groups of adenine and guanine)⁴⁷ and differ from those targeted by the acridine conjugates.

Figure 4. Piperidine lability of DNA cross-links formed by ebisQMPN₃.

Reaction was initiated by addition of ebisQMPN₃ (50 μM) in CH₃CN (final concentration of 20%) to 5’-[³²P]-OD5 (2.8 μM) and OD4 (3.0 μM) in 10 mM MES pH 7 and 10 mM NaF and maintained under ambient conditions for the indicated time before reaction was quenched by flash freezing in liquid N₂. Samples were thawed and one aliquot of each was combined with loading dye for direct analysis by denaturing PAGE (20%). A second aliquot was treated with piperidine and heat before comparable analysis. All materials were detected by phosphoimagery. The cross-link standard (xl) was generated with [³²P]-OD5/OD4 and ebisQMPN₃ (100 μM, 24 h).

Decoupling QM generation for cross-linking.

Many of the intrinsic constraints of the bisQMP system above were subsequently removed in an attempt to regain the desired dynamics of DNA alkylation. To date, cross-linking of DNA in our laboratory has derived from tandem formation of QM intermediates in which the electrophilic centers are ortho to the parent phenol. This design precludes simultaneous formation of two QMs and allows only one electrophilic site to be generated at a time. Additionally, a strict geometry and distance of reactive sites is maintained and these could only be altered by constructing different bisQMPs as illustrated by others.^32,61 As an alternative, cross-linking may rely on two individual QMPs tethered through a flexible linker to sample a wide range of distances and geometries.^23,62 This approach also releases the constraint of sequential QM generation and each QMP converts to its QM independently. This complementary approach has now been extended to related polyammonium ion conjugates for comparison to the bisQMPs above (Scheme 5). An analogous strategy had previously been applied to metalation of DNA by polynuclear platinum reagents.⁶³ These species were capable of long range cross-linking with many fewer constraints than those formed by the common mononuclear cis platinum drugs and they also supported migration of the platinum between guanine residues.⁶⁴

Scheme 5.

Relaxing the constraints of timing and geometry for quinone methide formation and subsequent nucleophilic addition (For R, see Scheme 2)

Both the alkyl diammonium (diQMPN₂) and triammonium (diQMPN₃) conjugates were designed for comparison to the bisQMP configurations above that were restricted by tandem formation and regeneration of QM intermediates. These additional conjugates were synthesized from a common intermediate 8 (Scheme 6) that in turn was produced by a single hydroxymethylation of 4-(2-bromoethyl)phenol as controlled by complexation with phenylboric acid (Scheme S7).⁶⁵ The two monofunctional precursors were then coupled by alternative reaction with 0.5 equivalents of N,N’-dimethyl-1,3-propane diamine and N,N-bis(3-methylamino)propyl)methylamine (Schemes S8 and S9). Benzylic deprotection, acetylation and final methylation above to form diQMPN₂ and diQMPN₃ followed the same strategy described for the bisQMPN_x derivatives above (Scheme 6 and Supporting Information).

Scheme 6.

Generation of di- and trialkylammonium conjugates from a common precursor.

Treatment of a model duplex DNA over 24 h with diQMPN₂ and diQMPN₃ demonstrated that the diammonium species was significantly less efficient as a cross-linking reagent than the triammonium species and also much less efficient than the original bisQMPN₂ with a fixed geometry (Figures 1 and S8). In contrast, the diQMPN₃ and bisQMPN₃ expressed relatively similar yields of cross-linking and both promoted DNA aggregation and precipitation when added in excess of 100 μM. The heterogenous electrophoretic migration of the labeled oligonucleotide after treatment with diQMPN₂ in particular suggests progressive accumulation of monoadducts rather than cross-links. Cross-linking remained minor even after the concentration of diQMPN₂ was raised to 500 μM. Time-dependence of reaction between duplex DNA and both conjugates confirmed the trends above and the modest potential for cross-linking with 500 mM diQMPN₂ and slightly higher efficiency with 50 μM diQMPN₃ (Figure 5). The lower concentration of diQMPN₃ was selected to minimize DNA aggregation but even these conditions still resulted in a general loss of the target DNA by 24 h. Cross-linking was evident in a yield of no more than ~ 50%. The remaining products migrated as expected for a series of alkylated derivatives³⁹ resulting from one of the QM equivalents of the diQM precursor reacting with DNA and the other reacting with water. The polyammonium ion once again appears to direct QM reaction away from guanine N7 and other reversible nucleophiles to alternative sites that formed irreversible adducts.

Figure 5. Rate of DNA reaction with diquinone methide conjugates.

Reaction was initiated by alternative addition of diQMPN₂ (500 μM) and diQMPN₃ (50 μM) in CH₃CN (final concentration of 20%) to 5’-[³²P]-OD4 (2.8 μM) and OD5 (3.0 μM) in 10 mM MES pH 7 and 10 mM NaF. After incubation under ambient conditions for the indicated times, products were separated by denaturing PAGE (20%) and detected by phosphoimagery. Control samples ss and xl were generated by 24 h incubations in the respective absence and presence of bisQMPN₃ (100 uM).

Reaction of both duplex and single-stranded DNA with diQMPN₃ did not yield piperidine induced fragmentation of the labeled strand as expected from involvement of guanine N7 (Figure S9). Instead, a series of products migrating slightly slower than starting material was generated after piperidine treatment in analogy to that observed for ebisQMPN₃ (Figure 4). The similarity of species formed by single- and double-stranded DNA confirms the lack of interstrand cross-linking. The general retardation of their gel mobility was again likely caused by a progressive accumulation of individual polyammonium adducts on the DNA that diminish its net charge (Figure S9). Thus, the diQMP configuration joined by a central polyammonium ion was unable to restore selectivity for the reversible nucleophiles of DNA that are necessary to support cross-link migration.

DNA alkylation by a monofunctional quinone methide precursor conjugated to an alkyl diammoniun appendage.

A monofunctional analogue of bisQMPN₂ and diQMPN₂ was prepared for comparison to a corresponding monofunctional analogue of bisQMP-Acr to determine the basis of their different specificity for DNA products (Figure 6A). The monofunctional acridine conjugate (QMP-Acr) had been prepared previously in an attempt to overcome the slow migration of DNA cross-links. In this case, the covalent anchor that may have limited the distance of each migration step was removed by eliminating one of the two QM equivalents of bisQMP-Acr.⁵⁹ This design did indeed facilitate QM transfer but also allowed its dissociation from the parent duplex. Release of the QM intermediate from the confines of a duplex enhanced its exposure and susceptibility to quenching by solvent nucleophiles that are not typically competitive when QM intermediates remain associated to DNA.^14,44 In contrast to the polyammonium ion conjugates above, QMP-Acr still shared the ability of bisQMP-Acr to alkylate guanine N7.⁵⁹ This was again illustrated by the piperidine-dependent fragmentation of DNA induced after reaction between duplex DNA and QMP-Acr (Figure 6).

Figure 6. Rate of DNA alkylation by the monofunctional quinone methide conjugates QMP-Acr and QMPN₂.

(A) Monofunctional QMPs and a model DNA duplex. (B) Alkaline-labile sites of alkylation were detected after treatment with piperidine (oval marks indicate alkylation). (C) 5’-[³²P]-OD7/OD8 (3.0 mM) was alternatively incubated with QMP-Acr (120 μM) and QMPN₂ (240 μM) in the presence of 10 mM MES pH 7.0 and 10 mM NaF for the indicated times before quenching, treating with 10 % aq. piperidine at 90 °C for 30 min, analyzing by denaturing PAGE (20%) and detecting with phosphoimagery.

The monofunctional QMP with an alkyl diammonium appendage (QMPN₂) was generated by a strategy similar to that used for its bisfunctional analogue (Scheme S10 and Supporting Information). Reaction of the acridine and polyammonium ion conjugates with duplex DNA was monitored over 72 h and maximum yields were obtained by 12 – 24 h in the presence of 120 μM QMP-Acr and 240 μM QMPN₂ (Figure 6). Under these conditions, the two conjugates consumed approximately equal quantities of the parent duplex and equivalent results were also obtained when examining the concentration dependence of reaction (Figure S10). QMPN₂ generated the same piperidine-stable adducts as those of the related ebisQMPN₂ that have a retarded migration based on the presence of stable polyamommium adducts. However, in contrast to the action of ebisQMPN₂ (Figure 4), piperidine-labile adducts of guanine N7 were also evident after treatment with QMPN₂. Thus, the unanticipated change in target specificity from guanine N7 to sites stable to piperidine for the various cross-linking conjugates containing polyammonium ions is a result of both these ions and the bifunctional nature of the QMP. The ability of a monofunctional QM such QMPN₂ to react at guanine N7 is even masked when two of these conjugates are covalently tethered to generate diQMPN₃ (Figure S9).

Minor reaction is also evident at adenine residues to create DNA fragments after piperidine treatment but these diminish over time for QMP-Acr as expected from their transient existence.^42,59 Surprisingly, fragmentation at adenine A7 induced by QMPN₂ persisted throughout the 72 h incubation and suggests a different site of alkylation since the nature of the electrophilic center of both conjugates remained constant (Figure 6A). Alternatively, the polyammonium ion may stabilize its adenine adduct through electrostatic interactions that are not available to the acridine conjugate.

The overall yield of alkylation at guanine N7 generated by QMPN₂ decreased significantly for single-stranded versus duplex DNA and a similar decrease was also evident for QMP-Acr (Figure S10). In contrast, formation of the piperidine stable polyammonium adducts formed by QMPN₂ increased for single-stranded versus duplex DNA as evident from the series of products migrating more slowly than the parent strand. Their differences in migration again likely reflect the extent of charge neutralization associated with their varying numbers of adducts containing polyammonium ions. These products are readily formed by access to a multiplicity of sites for reaction with single-stranded DNA. In contrast, piperidine labile adducts of guanine N7 require the formation duplex DNA and its suppression of reaction at other sites. These observations may reflect the preference for the acridine of QMP-Acr to intercalate into helical DNA rather than associate with single-stranded DNA in a random coil. In contrast, the polyammonium ion may easily condense equally well around the phosphoribose backbone of structured and unstructured DNA.

The transfer of adducts formed by both QMPN₂ and QMP-Acr from one DNA strand to its complement was equivalent and quite modest. First, their ability to alkylate DNA was lost after a 24 h preincubation in the absence of an acceptor DNA (Figures S11). This was expected from the ability of solvent to react with the transient QM to form irreversible adducts. Second, equivalent preincubations in the presence of an acceptor strand of DNA traps a fraction of the QM intermediates as reversible adducts including that formed by alkylation of guanine N7.^45,46 These adducts remain available for transfer and can be observed after their migration to a complementary DNA added subsequently. Accumulation of guanine N7 adducts in this complementary strand increased slowly over time as measured by the increase of piperidine induced fragmentation at guanine (Figure S12). The rate and yield of QM transfer (~ 10–12% alkylation of the complementary strand over 50 h) was equivalent for adducts formed by QMPN₂ and QMP-Acr and consistent with the similarly low yields of guanine N7 alkylation formed directly by the original acceptor as a single strand (Figure S10). Piperidine stable adducts would not be expected to participate in transfer since these remain inert under even the harsh conditions used to fragment guanine N7 adducts. The similarity in transfer of a monofunctional QM when alternatively conjugated to a polyammonium ion and acridine demonstrates that the dynamics is a function of the reversible adducts and independent of the attached ligand. However, these ligands do still control the type of adducts generated. The polyammonium ion conjugates described here direct alkylation primarily to nucleophiles in duplex DNA that act irreversibly in contrast to the reversible adducts formed by related acridine conjugates. An equivalent redirection of target sites was not observed previously for nitrogen-mustard conjugates^2,52 and the basis of these differences is currently under investigation.

CONCLUSION

Electrophiles are typically directed with precision to specific sites in DNA when conjugated to sequence selective ligands but even ligands without such selectivity may dramatically affect the profile of adducts. For example, a simple ortho QM reacts predominantly with cytosine N3 and adenine N1 when incubated with either nucleosides or DNA as a single strand or duplex.^47,66 However, an acridine conjugate of this same QM overwhelmingly diverts reaction to guanine N7.⁵⁹ Polyamines primarily interact with duplex DNA by non-specific electrostatic interactions and have been described as very mobile on the surface of DNA.^50,51 By localizing near the backbone of DNA, polyamine conjugates may be expected to position an attached reactant for access to both grooves. Thus, the profile of resulting adducts would presumably remain a function of the reactant’s innate chemical specificity. Such a scenario was supported by nitrogen mustards in which the parent compound and its polyamine conjugates all targeted guanine N7. An equivalent result was not observed for the QM conjugates described here. The general efficiency of DNA alkylation was increased when QMs were conjugated to alkyl di- and triammonium derivatives but their apparent specificity was diverted from guanine N7 to other sites on DNA that are not sensitive to piperidine treatment. This shift did not prevent DNA cross-linking by bisfunctional QMs that generated electrophilic sites with a fixed geometry and distance (bisQMPN₂ and bisQMPN₃, Figure 1; ebisQMPN₂ and ebisQMPN₃, Figure 3). However, dynamic cross-linking analogous to that produced by bisQMP-Acr was not maintained since the polyammonium ion conjugates alkylated nucleophiles acting irreversibly. This result may originate from direct delivery of the electrophiles to such nucleophiles or through an indirect accumulation after rapid migration of QMs from nucleophiles reacting reversibly.

Migration of intrastrand cross-links from a single-stranded donor to interstrand cross-links with a complementary acceptor strand was very limited with the bisQMPN_x conjugates (Figure 2) and no further migration was evident after subsequent addition of secondary acceptor strands (Figures S3 and S4). Reversible adduct formation and efficient QM transfer could not be recovered by use of activated QM derivatives containing electron donating groups. Similarly, the migratory ability was not recovered with diQMPN_2,3, derivatives in which the two QM equivalents were decoupled spatially and temporally. Instead, the increased degrees of freedom available to diQMPN_2,3 diminished their capacity to cross-link duplex DNA (Figure 5). Consequently, the bisQMP design remains ideal for dynamic cross-linking with an ultimate goal of disrupting its repair. Selection of the appropriate DNA ligand is also critical for maintaining these dynamics. The next generation of conjugates will revisit the use of intercalators to enhance the kinetics of QM transfer but candidates will be chosen with a moderate affinity for DNA to minimize their interference with cross-link migration.

EXPERIMENTAL PROCUDURES

Detailed procedures are described in Supporting Information.

General Materials.

Starting materials, reagents, and solvents were obtained commercially and used without further purification. Water was purified to a resistivity of 18.2 MΩ-cm using a Barnsted GenPure system. Oligonucleotides were purchased from IDT with standard desalting and labeled on the 5’ terminus with γ-[³²P]-ATP (Perkin Elmer) and T4 polynucleotide kinase (New England Biolabs) using manufacturer protocols. DNA concentrations were calculated by their A₂₆₀ (Agilent 8453 UV-Vis spectrometer) and A₂₆₀ values provided by IDT.

General Methods.

Standard reactions of DNA were initiated by addition of a CH₃CN (final concentration 20%) solution of the indicated QMP to oligonucleotide samples in 10 mM 2-(N-morpholino)ethanesulfonate (MES) pH 7 and 10 mM NaF. Studies with duplex DNA contained one equivalent of the radiolabeled strand (2.8 – 3.0 μM) and 1.1 equivalents of the complementary strand to ensure analysis derived from the hybridized species. Reactions were incubated under ambient conditions and quenched at the desired time by flash freezing in liquid N₂. Samples were then thawed in the presence of an equal volume of loading dye (0.05% bromophenol blue and 0.05% xylene cyanol FF in formamide) and separated by denaturing polyacrylamide (20%) gel electrophoressis (PAGE). Products were detected by phosphoimagery and yields were calculated relative to the total signal for each lane. To detect adducts at guanine N7, reaction samples were dried under vacuum rather than frozen. Their residues were then dissolved in aq. piperidine (10%, 30 μL), heated at 90 °C for 30 min and then dried at high vacuum. Samples were dissolved in water (30 μL) and dried in three successive repetitions to remove all traces of piperidine.