New Synthetic Analogs of Nitrogen Mustard DNA Interstrand Crosslinks and Their Use to Study Lesion Bypass by DNA Polymerases

Abstract

Nitrogen mustards are a widely used class of antitumor agents that exert their cytotoxic effects through the formation of DNA interstrand crosslinks (ICLs). Despite being among the first antitumor agents used, the biological responses to NM ICLs remain incompletely understood. We have previously reported the generation of NM ICL mimics by incorporation of ICL precursors into DNA using solid-phase synthesis at defined positions, followed by a double reductive amination reaction, but these mimics deviated from the native NM ICLs. Using further development of our approach, here we report a new class of NM ICL mimics that only differ from their native counterpart by substitution of dG with 7-deaza-dG at the ICL. Importantly, this approach allows for the synthesis of diverse NM ICLs, illustrated here with a mimic of the adduct formed by chlorambucil. We used the newly generated ICLs in reactions with replicative and translesion synthesis DNA polymerase to demonstrate their stability and utility for functional studies. These new NM ICLs will allow for the further characterization of the biological responses to this important class of antitumor agents.

Graphical Abstract

INTRODUCTION

DNA interstrand crosslinks (ICLs) are formed by a covalent linkage between the two strands of a DNA duplex and inhibit DNA strand separation, a prerequisite for replication and transcription.^{1, 2} Therefore, ICL-inducing agents, in particular cisplatin and nitrogen mustards, are widely used as chemotherapeutic agents to target rapidly dividing tumor cells. The blockage of replication also triggers repair pathways that remove ICLs, constituting an important resistance pathway to therapies.^{3, 4} Therefore, ICL-inducing agents are particularly effective against cells with defects in DNA repair pathways, such as BRCA1/2 deficient tumors.^{5, 6} A number of mechanisms exist in human cells to address ICLs and they involve a combination of factors from the Fanconi anemia, translesion synthesis, homologous recombination, and base and nucleotide excision repair pathways. The exact pathway deployed depends on the cell cycle, the context of replication, and the structure of the ICL.

To facilitate studies of ICL repair, defined and site-specific ICLs are needed. The direct treatment of DNA duplexes with ICL-inducing bis-reactive electrophiles results in mixtures of ICLs and side products such as intrastrand crosslinks and mono-adducts, with ICLs typically making up a very small percentage of the product.⁷ To overcome the poor specificity of ICL formation, we and others have developed ways to generate site-specific ICLs. Such approaches are broadly based on the incorporation of an ICL precursor at defined positions in two complementary strands of a DNA duplex, followed by the formation of an ICL using a using post-synthetic coupling reaction.^8-17

Our studies have been primarily focused on nitrogen mustards (NMs), the first group of FDA-approved chemotherapy agents.¹⁸ NMs form ICLs through the bis-(2-chloroethyl)-amine moiety, which primarily reacts with N7 of 2’-deoxyguanosine (dG) through two sequential aziridine intermediates to form ICL 1 (Scheme 1A).^{19, 20} The “R” group is responsible for modulating the electrophilicity of the aziridinium ion (and hence the reactivity of the drug) and for improving its pharmacological profile.²¹ NMs preferentially form ICLs in 5’-GNC-3’ sequences, and given that the length of the crosslink (7.5 Å) is shorter than the natural distance between two dGs in 5'-GNC-3' sequence (8.9 Å), they induce a bend in the duplex.^22-24 In addition to forming ICLs, NMs also form a mixture of intrastrand crosslinks and mono-adducts as well as DNA-protein crosslinks.^{25, 26} NM ICLs can furthermore undergo spontaneous depurination or ring-opening to the Fapy-dG adduct due to the presence of a positive charge in the purine ring stemming from the alkylation of the N7 position.^{27, 28}

Scheme 1. Native and synthetic nitrogen mustard (NM) DNA interstrand crosslinks (ICLs).

A. NM ICL formed by reacting a nitrogen mustard with DNA bears a positive charge in the purine ring of the two crosslinked dG residues. B. NM ICL mimic formed through a double reductive amination reaction of two alkyl-aldehyde-deazaguanine residues incorporated by solid-phase synthesis yields an uncharged and stable ICL.

We reported the generation of NM-like ICLs via a reductive amination reaction between two aldehyde-modified dG residues (2), in which a substitution of dG with 7deaza-dG provided greater chemical stability of the crosslink and a versatile platform for functionalization (Scheme 1B).⁸ The ICLs were generated via a double reductive amination reaction between two aldehyde groups that were introduced into complementary strands of a duplex masked as protected aldehydes. This strategy allowed us to generate stable NM ICL analogs with various lengths of the crosslink (Fig. S1).^{29, 30} However, our first attempt to generate ICL 3 through a coupling of two 2-carbon precursors with ammonia or methylamine was unsuccessful (Fig. S1A). Since we were able to form ICLs with N,N’-dimethylethylenediamine and hydrazine (4 and 6 in Fig S1B), which have a longer linkage between the two crosslinked bases, we attributed the failure of the formation of 3 to the inability of the imine formation and reductive amination to induce a bend in the DNA (Fig. S1B).⁸ Our next attempt allowed for the generation of the 5-atom NM ICL 6 through a coupling of a 1-carbon aldehyde precursor (5) and a 2-carbon aldehyde precursor (2) with hydrazine. We attributed this success to the higher activity of hydrazine versus a simple amine (Fig. S1C).²⁹ Although our molecular modeling studies suggested that 6 was a structurally suitable NM ICL mimic, we found it to be chemically unstable over prolonged periods of time, limiting its usefulness for biochemical and cell applications.

Inspired by our recent synthesis of a “single-nucleotide NM ICL” 8 employing a reductive amination reaction between an aldehyde precursor on DNA and a dG nucleoside derivative with a primary amine 7 (Fig. S1D),³¹ we revisited our earlier efforts for the generation of isosteric NM ICL mimics. We report here the generation of a new class of ICL mimics 3 that differs from the native ICL only by the substitution of dG with 7deaza-dG (Scheme 1). Our approach for the synthesis of additional NM ICL derivatives mimicking those formed by clinically important ICLs is illustrated here using chlorambucil as an example. We demonstrate the suitability of these ICLs for biological studies in reactions with replicative and TLS polymerases.

MATERIALS AND METHODS

Reagents and Materials:

1-Chloro-2-deoxy-3,5-di-O-toluoyl-α-_D-ribofuranose was purchased from Carbosynth (Newbury, Berkshire, UK). 2-Cyanoethyl N,N,N′,N′-tetraisopropylphosphorodiamidite and 2’-deoxyguanosine monohydrate were purchased from ChemGenes (Wilmington, MA, USA). dU-CE phosphoramidite and controlled pore glass (CPG) supports were purchased from Glen Research (Sterling, VA, USA). Solid-phase extraction cartridges (Agilent TOP-DNA) were purchased from Agilent Technologies (Santa Clara, CA, USA). Centrifugal filters (Amicon^® Ultra 0.5 mL, MWCO 3,000) and HPLC solvents were purchased from Merck Millipore (Burlington, MA, USA). Ultrapure water, ethylenediaminetetraacetic acid (EDTA) solution, Trizma^® hydrochloride (Tris-HCl) buffer, tris-borate-EDTA (TBE) buffer, and 1,4-dithiothreitol (DTT) were purchased from Biosesang (Seongnam, South Korea). Klenow fragment (3’-->5’ exo⁻), NEBuffer 2, dNTP mix, and bovine serum albumin (BSA) were purchased from New England Biolabs (Ipswich, MA, USA). Polyacrylamide gel electrophoresis (PAGE) supplies and electroelution supplies were purchased from Bio-Rad Laboratories (Hercules, CA, USA). DNA sequencing gel apparatus and its supplies were purchased from Analytik Jena AG (Jena, Germany). SYBR^™ Gold nucleic acid gel stain was purchased from Invitrogen (Carlsbad, CA, USA). F254-coated silica gel TLC plates were purchased from Merck KGaA (Darmstadt, Germany). All other chemical reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA). The following HPLC-purified oligonucleotide was purchased from Integrated DNA Technologies (Coralville, IA, USA): Primer: 5’[Cy5]-CACTGACTCTATGATG-3’

Oligonucleotide Synthesis:

7-Deaza-7-(2,3-diacetoxy-propyl)-5’-O-(4,4’-dimethoxytrityl-N(2)-isobutyryl)-2’-deoxyguanosine-3’-[(2-cyanoethyl)-N,N-diisopropylphosphoramidite] was synthesized as previously reported with slight modifications; 2,4-Diamino-6-hydroxypyrimidine was used as a starting material to make 7-deazaguanine intermediate, and 2-cyanoethyl N,N,N′,N′-tetraisopropylphosphorodiamidite and 4,5-dicyanoimidazole were used to make the phosphoramidite.^{8, 32} The phosphoramidite was dissolved in 3:1 DCM/acetonitrile and introduced into the oligonucleotide sequences shown in Table 1 by Expedite^™ 8909 Nucleic Acid Synthesis System (PerSeptive Biosystems, Framingham, MA, USA). Standard DNA synthesis conditions and 1 μmol 1000Å CPG supports were used during the synthesis, except that coupling time was extended to 15 minutes for the modified phosphoramidite.

Table 1. Oligonucleotide sequences.

X denotes position of ICL precursor (9) or ICL (3)

The oligonucleotides were deprotected in 30% ammonium hydroxide at 55°C for 12 hours, dried by vacuum centrifugation at room temperature, and then purified by HPLC (Agilent Technologies 1260 Infinity II) equipped with a C18 reverse-phase LC column (Phenomenex Clarity^® 5 μm Oligo-RP^™, 150 x 4.6 mm) using a 100 mM triethylammonium acetate (TEAA) to acetonitrile gradient. Before sample injection, the column was equilibrated with 100 mM TEAA with a flow rate of 1 mL/min. After sample injection, the acetonitrile gradient was initially maintained at 5% for 5 minutes. Oligonucleotide samples were eluted by linearly increasing the acetonitrile gradient from 5% to 40% in 20 minutes. Under these conditions, DMT-on oligonucleotides eluted between 30-40%. The gradient was then increased from 40%-95% acetonitrile in 1 minute, kept at 95% for 2 minutes before the column was re-equilibrated with 100 mM TEAA/5% acetonitrile. The DMT groups were removed by incubating the dried oligonucleotide in 1 mL of 80% acetic acid at 25 °C for 1 hour. Alternatively, both the oligo purification and the detritylation could be done using a solid-phase extraction cartridge (Agilent Technologies TOP-DNA) following the manufacturer's instructions. The purified and freeze-dried oligonucleotides were resuspended in ultrapure water and the concentrations were measured by microvolume UV spectrophotometer (Thermo Scientific^™ NanoDrop). The yields of purified oligonucleotides were approximately 30 to 50%, depending on the sequence and the batch.

Synthesis of 20mer and 6mer ICLs:

The 20-mer diols (9-1 and 9-2, 30 nmol) were converted to aldehydes (2-1 and 2-2) at 4°C for 2 hours (dark) in the presence of 50 mM sodium periodate (NaIO₄) and 100 mM sodium phosphate (pH 5.4, final volume 200 μL). The reaction product was subject to centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters) to remove excess NaIO₄. Amine solutions were prepared by making 500 mM working stocks of methylamine (MA) in water and (3-(4-Aminophenyl) propionic acid (APA) in DMSO. An aliquot of 7 (7.5 nmol) was converted to one of the amines (10A and 10B) at 25°C overnight (dark) in the presence of 20 mM amine, 20 mM sodium cyanoborohydride (NaBH₃CN), and 100 mM sodium phosphate (pH 5.4, final volume 45 μL). The reaction product was subjected to centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters) to remove excess reagents. LC-MS analyses of the crude mixtures indicated that both the diol cleavage and the reductive amination reactions went to completion. The 20-mer amine (10A, 10B, 7.5 nmol) and an equal amount of 21-mer aldehyde (2-2, converted from 9-2) were annealed at 25°C overnight (dark) in the presence of 100 mM sodium chloride (NaCl) (final volume 100 μL). The annealed oligonucleotides were concentrated by centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters) at 4°C and crosslinked at 25°C overnight (dark) in the presence of 20 mM NaBH₃CN and 100 mM sodium phosphate (pH 5.4, final volume 70 μL). After centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters), the concentrated reaction mixture was mixed with an equal volume of formamide loading buffer (95% formamide, 20 mM EDTA, 0.1% bromophenol blue), heated at 92°C for 2 min, and then ice-chilled immediately. The solution was loaded on a 15% polyacrylamide gel (17 cm x 15 cm x 1.5 mm) containing 7 M urea and 1X TBE and resolved by gel electrophoresis (300V). The gel was visualized by UV shadowing on an F254-coated silica gel TLC plate, and the bands corresponding to ICLs were excised. The excised gel slabs were finely chopped and the DNA isolated by electroelution using a Bio-Rad model 422 device, equipped with MWCO 3,500 membranes. The eluted ICLs were desalted by centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters), lyophilized, and resuspended in ultrapure water. The purities of PAGE-purified ICLs (3A and 3B) were analyzed by LC-MS, PAGE, and HPLC. The 20-mer and 39-mer sequences shown in Table 1 were used to afford 39/20-mer with MA 13A and APA 13B for the DNA polymerase assays. The isolated yields of the purified ICLs, measured by microvolume UV spectrophotometer (Thermo Scientific^™ NanoDrop), varied depending on crosslinkers and scales. The isolated yield of the MA ICL was 4%, while the isolated yield of the APA ICL was 9%. The reactions were tried in different scales, ranging from 7.5 nmol to 50 nmol. Higher yields up to 30% were obtained on larger reaction scales. The DMEDA ICLs for the DNA polymerase assays were generated as previously reported.⁸

To generate the resected 6 bp ICLs, the purified 20 bp ICLs were digested with the USER enzyme mix (uracil DNA glycosylase and endonuclease VIII, NEB M5505), which cleaved the phosphodiester backbone at the position of the uracil residues. The 6 bp ICLs were purified by 12–15% denaturing PAGE, followed by electroelution as described above.

HPLC Analysis:

Oligonucleotide samples were desalted by centrifugal filtration (Merck Amicon^® Ultra 3kDa MWCO filters), dried by lyophilization, and resuspended in ultrapure water. An Agilent Technologies 1260 Infinity II system composed of a quaternary pump, multisampler, column oven, and a diode array UV detector (DAD) was used for the analysis. C18 reverse-phase LC column (Phenomenex Clarity^® 5 μm Oligo-RP^™, 150 x 4.6 mm) was used for the stationary phase. Before sample injection, the column was equilibrated with 100 mM TEAA at 50°C with a flow rate of 1 mL/min. After sample injection, the acetonitrile gradient was initially maintained at 0% for 5 minutes. Oligonucleotide samples were eluted by linearly increasing the acetonitrile gradient from 0% to 40% in 30 minutes. Under these conditions, 20/21-mer oligonucleotides eluted between 17– 18 minutes, and ICLs were eluted between 18 – 19 minutes. After 30 minutes, 100% acetonitrile gradient was applied for 2 minutes before the column was re-equilibrated with 100 mM TEAA.

UPLC-ESI⁻-MS Analysis:

UPLC-ESI⁻-MS analyses were conducted with a Thermo Scientific^™ Q Exactive^™ Focus hybrid quadrupole-Orbitrap mass spectrometer interfaced with a Thermo Scientific^™ UltiMate 3000 UHPLC system. A heated electrospray ionization (HESI-II) probe was operated in negative ionization mode. The analyzer was operated in full-MS mode with a scan range of m/z 200 – 2000. A Thermo Scientific^™ Hypersil GOLD^™ (100 x 2.1 mm, 1.9 μm) column was conditioned with 15 mM ammonium acetate and acetonitrile at 40°C at a flow rate of 50 μL/min. DNA oligonucleotides (50 pmol) were eluted with a gradient that was initially kept at 2% acetonitrile for 2 minutes, then linearly increased to 80% acetonitrile over 16 minutes, kept at 80% acetonitrile for 2 minutes, decreased to 2% acetonitrile over 2 minutes, and finally re-equilibrated at 2% acetonitrile for 13 minutes. Under these conditions, DNA oligonucleotides eluted between 14–16 minutes. For MS analysis, electrospray ionization was achieved at a spray voltage of 3000 V and a capillary temperature of 320 °C. Sheath gas was set at 35, auxiliary gas was set at 10, and sweep gas was set at 1. Spectral data was processed and deconvoluted using Thermo Scientific^™ BioPharma Finder 3.1 software.

Enzymes:

Klenow (exo−) enzyme (5 U/μl equivalent to 3.6 μM) was purchased from NEB (M0212). The protein was diluted in 25 mM Tris-Cl (pH 7.4), 1 mM DTT, 0.1 mM ethylene- diamine-tetraacetic acid (EDTA), and 50% glycerol to the indicated concentrations for use in polymerase assays. Human pol η (with a C terminal His tag) was prepared as described previously, yielding a preparation with a concentration of 0.3 mg/ml (~4 μM).³³ The protein was diluted in 40 mM Tris-HCl (pH 8.0), 10 mM dithiothreitol (DTT), 0.1 mg/ml bovine serum albumin (BSA), and 30% glycerol to the indicated concentrations for use in polymerase assays.

Primer extension with Klenow exo−:

Fluorescent-labeled primer (100 nM), DNA templates (200nM), sodium acetate (NaOAc, 100mM final concentration), and ultrapure water were mixed together, and the primer and template were annealed by programmed incubation in a thermal cycler (65°C 5 min, 37°C 30 min, 25°C 20 min, then 4°C 20 min). A master mix (108 μL) was made for each ICL substrate by mixing annealed primer:template solution (12 μL), 10X NEBuffer 2 (12 μL, 500 mM NaCl, 100 mM Tris-HCl, pH 7.9, 100 mM magnesium chloride, 10 mM DTT), 1 mM dNTP mix (12 μL), and ultrapure water (72 μL). The Klenow fragment (3’->5’ exo⁻) stock solution was diluted to 5 nM using polymerase dilution buffer (50 mM Tris-HCl, pH 7.8, 0.5 mg/mL BSA, 1 mM DTT, 5% glycerol). To each reaction tube containing 18 μL of the master mix on ice, polymerase was added (2 μL), and then the reaction tube was incubated at 37°C. The reactions were stopped at different time points by adding 20 μL of stop solution (95% formamide, 20mM EDTA, 0.1% orange G), heated at 92°C for 3 min, followed by ice-chilling. Aliquots (20 μL) were loaded to 12% polyacrylamide sequencing gel containing 7 M urea and 1X TBE and electrophoresed for 2 hours at 60W. The labeled primers, indicating the polymerase extension products, were visualized on an Amersham^™ Typhoon^™ biomolecular imager (GE Healthcare Life Sciences).

Primer extension with DNA Polymerase η:

A 10X reaction buffer (400 mM Tris-HCl, pH 8.0, 100 mM DTT, 1 mg/mL BSA, and 25% glycerol) was freshly prepared on ice and used to prepare mixture A (10 nM annealed substrate, 50 nM of polymerase, and 1X reaction buffer) and mixture B (20 mM MgCl₂, 200 μM dNTP mix, and 1X reaction buffer). The reaction was initiated by combining 5 μL mixture A and 5 μL mixture B, followed by incubation at 37°C for either 1 min, 3 min, or 20 min. The final reaction mixture contained 5 nM substrate, 25 nM polymerase, 10 mM magnesium chloride (MgCl₂), 100 μM dNTP mix, 25 mM NaOAc, 40 mM Tris-HCl, pH 8.0, 10 mM DTT, 100 μg/mL BSA, and 2.5% glycerol. The reactions were stopped by adding 10 μL of stop solution (95% formamide, 20 mM EDTA, 0.1% orange G), analyzed by gel electrophoresis, and products visualized and quantified as described for the Klenow reactions.

Results

Generation of mechlorethamine and chlorambucil ICL mimics.

Although our previous attempts to generate a stable 5-atom NM mimic by direct reductive amination were not successful (Fig S1), we sought to revisit and optimize this strategy. We were encouraged by our observation that an aldehyde-containing oligonucleotide can react with a 7-(2-aminoethyl)-deazaguanosine 7 to generate a single nucleotide ICL 8 (Fig S1D).³¹ Therefore, we examined the possibility that a crosslink may be formed by sequential reductive aminations of two aldehyde precursor oligonucleotides by generating an amine species in one of the two strands prior to annealing. We first generated a pair of complementary 20/21-mer oligonucleotides containing diol ICL precursors 9 (Fig.1, Table 1). Both of them were oxidized to the aldehyde 2 with periodate and the excess oxidizing agent was removed. One of the oligonucleotides was then treated with methylamine (MA) in the presence of cyanoborohydride to yield the secondary amine 10A, which was annealed to the aldehyde-containing complementary strand and the ICL 3A was formed by reductive amination (Figure 1). Analysis of ICL formation by denaturing PAGE gel revealed the formation of a new band with slower mobility, indicative of ICL formation (Fig 1B, lane 2). Although formed in moderate yield, the product was readily purified by PAGE and isolated by electroelution (Fig 1B, lane 4). We also monitored the progression of the reaction by UPLC-ESI⁻-MS and confirmed all intermediates (aldehyde 2, methylamine 10A, crude ICL reaction mixture, Fig. 2A - D) and the final purified product ICL (Fig. 2E) based on exact mass measurements, unambiguously establishing the identity of the purified product 3A. The reaction progress could also be monitored by HPLC, where each of the oligonucleotides in the reaction sequence eluted with a distinct retention time (Fig. S2).

Figure 1. Synthesis of mechlorethamine and chlorambucil NM ICL mimics.

A. Scheme for ICL synthesis. Diol precursors 9-1 and 9-2 in 20/21-mer oligonucleotides were oxidized to the aldehyde 2-1/2-2 and reacted with methylamine or 3(4-aminophenyl) propionic acid and NaBCH₃CN to form amines 10A and 10B. Annealing of 10A and 10B with 2-2 followed by reduction yielded ICLs 11A and 11B. B. Analysis and pufication of the ICL products by 20% denaturing PAGE analysis. lane 1: 20-mer ssDNA marker; lanes 2 and 3: crude reaction mixures of 11A and 11B; lanes 4 and 5: analysis of products 11A and 11B following purification of the product from the gel and electroelution.

Figure 2. UPLC-ESI⁻-MS analysis of MA reaction intermediates and ICL product.

A. – C. ssDNA intermediates, D. crude ICL reaction mixture, E. purified ICL product. Masses are indicated as m/z.

The ability to generate an ICL with a secondary amine provided an opportunity to generate ICLs with substitutions at the nitrogen in the crosslink. Using the same protocol and 3(4-aminophenyl) propionic acid (APA), we generated ICL 3B, a close mimic of the adduct formed by chlorambucil missing only one methylene group between the phenyl ring and the carboxylic acid. The APA ICL was formed with a slightly higher yield than that of the MA ICL, as indicated by the higher intensity of the slower migrating band (Fig. 1B, lane 3). As for the MA ICL, it could readily be purified by PAGE (Fig 1B, lane 5) and reaction progress could also be monitored by HPLC (Fig. S3A-C). The intermediate amine monoadduct (10B), ICL reaction mixture, and the purified product could be unambiguously identified by UPLC-ESI⁻-MS (Fig. S3D-F). Although the ICL formation proceeded with only moderate yields, our approach allowed for the generation of nanomole quantities of ICL-containing oligonucleotides, a large supply for biochemical and potential structural studies. Importantly, these new NM ICLs only differ from their native counterparts by the substitution of 7-deaza-dG for dG and represent the most structurally similar NM ICL that is stable over extended periods of time.

DNA polymerase Klenow does not bypass NM ICLs 20mer duplex.

We then tested the stability and suitability of the ICLs for biochemical studies. We first assessed the stability of the MA and APA ICLs with exonuclease-deficient Klenow Fragment (KF^exo−), a bacterial replicative polymerase.³⁴ For this purpose, the ICLs were incorporated in templates for primer extension assays that we have previously employed.^{31, 35} For the first set of assays, we synthesized MA (3A) and APA (3B) ICLs in a duplex with a 39-mer template strand and an ICL in a 20-mer duplex (Table 1). For comparison, we used a template with a NM mimic that contains an 8-atom ICL (DMEDA-ICL, 4) that we had used in previous studies with DNA polymerases. We annealed a fluorescently labeled primer to the ICL templates, as well as a single-stranded control, and subjected them to a time-course extension reaction with KF^exo−. The ssDNA substrate was extended to the full-length product in a time-dependent manner (Fig. 3, lanes 1–6). For the MA and APA ICL substrates, KF^exo− briefly paused at the beginning of the duplex, but was able to open the duplex and advance up to the base before the ICL (Fig.3, lanes 7–19, see Fig. S3 for quantification). Prolonged reaction time made the polymerase eventually insert a nucleotide opposite to the ICL (Figure 3, lanes 12,18, and 24), but it was not able to bypass it. These results were qualitatively similar to those obtained with the 8-atom DMEDA ICL (4) (Figure 3, Lanes 19-24).³¹ The lack of bypass shows that the MA and APA ICLs are stable and suitable for biochemical studies.

Figure 3. Reaction of KF^exo− with NM ICLs in 20-mer and 6-mer duplexes.

KF^exo− (0. 5nM) was incubated with the annealed primer-template (10 nM) in the presence of Mg²⁺ (10 mM) and dNTP (100 μM) for the indicated time. Following addition of stop solution, samples were analyzed on a 10% denaturing gel. A. lanes 1-6: unmodified ssDNA; lanes 7-12: 20nt MA ICL; lanes 13-18: 20nt APA ICL; lanes 19-24: 20nt DMEDA ICL. B. lanes 1-6: unmodified ssDNA; lanes 7-12: 6nt MA ICL; lanes 13-18: 6nt APA ICL; lanes 19-24: 6nt DMEDA ICL. Position of bands on the right of the gel, sequences are in Table 1. See Fig S4 for quantification.

DNA polymerase Klenow can bypass NM ICLs in a short duplex.

We and others have previously observed that bypass of an ICL by TLS polymerases occurs more readily if the ICL is embedded in a shorter duplex.^{31, 35, 36} Since it is currently unknown where incisions around the ICL take place during ICL repair,³⁷ we have employed ICLs in a 6-mer duplex as model templates for “unhooked” ICLs for translesion synthesis. To obtain the 6-mer ICLs, we synthesized ICLs in 20-mer duplexes containing two uracil residues (see Table 1), which were digested with UDG/EndoVIII to generate substrates with the ICL embedded in a 6-mer duplex.³⁵ These substrates were incubated with KF^exo− in a time-course reaction. Although the polymerase stalled at the ICLs initially, both the MA and APA ICLs were bypassed at later time points (Fig. 3C, lanes 7-18, quantification in Fig. S3), suggesting that they did not provide an absolute block to this replicative polymerase. The larger aromatic group on the APA ICL delayed the bypass and extension compared to the methyl group in the MA ICL. By comparison, the stalling was more pronounced with a DMEDA ICL with an 8-atom ICL used in earlier studies,³¹ although it was bypassed at later time points as well (Fig. 3B, lanes 19-24, Fig. S3). These studies show that MA and APA NM ICLs are suitable for biological studies and raise the possibility that NM ICLs may be bypassed by replicative polymerases in the context of a short duplex.

Bypass of NM ICL by human DNA polymerase η.

It is thought that TLS polymerases are essential for the bypass of ICLs during replication-coupled repair of most ICLs.^37-39 Here we used human DNA polymerase η, a key TLS polymerase, to test its activity on MA and APA NM ICLs. We have previously shown that Pol η can bypass DMEDA NM ICL mimics in a 6-mer duplex.^{31, 35} We first tested Pol η in a time-course experiment with the MA (3A) and APA (3B) ICLs (39-mer template with a 20-mer duplex). While the ssDNA was extended to the full-length product in a time-dependent manner (Fig. 4, lanes 1-6, quantification in Fig. S4), the polymerase reaction stalled at the MA and APA ICLs; initially at the −1 position (lane 7), gradually extending the primer to the 0, +1 and + 2 positions. Only a small amount of the primer (>5%) was extended to the full-length product. As for Klenow, the insertion and extension were slightly slower for the APA ICL compared to the MA ICL. By contrast, the 8-atom DMEDA ICL primer stalled at the −3 position, eventually inserting a nucleotide at the −2 and −1 position, with no insertion opposite the ICL or extension to the full-length product even at later time points.

Figure 4. Bypass of NM ICLs by DNA polymerase η.

Pol η (25 nM) was incubated with the annealed primer-template (5 nM) in the presence of Mg²⁺ (10 mM) and dNTP (100μM) for the indicated time. Following addition of stop solution, samples were analyzed on a 10% denaturing gel. A. lanes 1-6: unmodified ssDNA; lanes 7-12: 20nt MA ICL; lanes 13-18: 20nt APA ICL; lanes 19-24: 20nt DMEDA ICL. B. lanes 1-6: unmodified ssDNA; lanes 7-12: 6nt MA ICL; lanes 13-18: 6nt APA ICL; lanes 19-24: 6nt DMEDA ICL. Position of bands on the right of the gel, sequences are in Table 1. See Fig S4 for quantification.

We next tested the activity of Pol η on the substrates with ICLs in 6-mer duplexes as models for unhooked substrates in ICL repair. We have previously shown that Pol η can bypass NM-like ICLs.³¹ In a time-dependent reaction, Pol η efficiently bypassed the 6-mer MA and APA ICLs with only a minor stalling point being observed at the insertion steps (Fig 4B, lanes 7-18, quantification in Fig S4). We similarly found that the DMEDA ICL was efficiently bypassed by Pol η. Taken together, these studies show that Pol η can efficiently bypass NM ICLs closely mimicking the adducts formed by mechlorethamine and chlorambucil.

DISCUSSION AND CONCLUSIONS

We have previously reported the synthesis of NM-like ICL mimics, based on the strategy to incorporate ICL precursors in the form of alkyl-aldehyde substituted 7-deazaguanine residues into DNA using solid-phase DNA synthesis, followed by a double reductive amination reaction to furnish the ICLs (Fig. 1 and S1).^{8, 27, 29, 30} This approach eliminated two main challenges for the generation of site-specific ICLs suitable for biological studies: The poor specificity of ICL formation when treating oligonucleotides directly with nitrogen mustards and the limited stability of the positively charged NM ICLs toward depurination and ring-opening. Up to the work presented here, we were unable to form ICLs most closely mimicking the NM ICL (Fig 1B) with a 5-atom bridge as in the native NM ICLs using this approach. We have attributed this to the need to induce a bend in the DNA to form the ICL, and hypothesized that the initially reversible imine formation would not provide enough energy to form this ICL.⁸ By carefully reexamining the reaction conditions, here we were finally able to generate this NM ICL. We believe that the key to this achievement was the sequential reductive amination of the two aldehydes, generating an alkylamino group on one of the two strands prior to annealing, and the careful purification of this intermediate to remove any oxidizing agent used in the generation of the aldehyde from the diol precursor. One possible alternative explanation, that aldehydes could be reduced before imine formation could be ruled out, as the aldehydes in DNA were not reduced upon incubation with NaBH₃CN (Fig. 2D and data not shown). Although ICL formation was only moderately efficient, it allowed for the generation of sufficient amounts of ICLs for biochemical and (if desired) structural studies. Importantly, the use of a simple amine (rather than hydrazine or dimethylethylenediamine) allows for the generation of clinically relevant NM ICLs, illustrated here with the example of chlorambucil. The ICLs generated here differ from their native counterparts only by the substitution of dG with 7-deaza-dG that brings with it a loss of positive charge in the purine ring, which has the potential to alter the structure of duplex. However, our molecular modeling²⁹ and limited biochemical studies with NM ICLs with native purine bases, in which the positive charge was stabilized by a fluoride substitution in the deoxyribose,²⁷ suggest that the effect of the loss of charge is minor.

We carried out some preliminary studies of the new ICLs with two DNA polymerases, the bacterial replicative Klenow polymerase, and the human TLS polymerase η. We gained two main pieces of insight from these studies. First, we observed that our new NM ICL mimics were stable under the reaction conditions, as evidenced by a lack of bypass of the ICLs in a 20-mer duplex by Klenow polymerase (Fig 3A). This finding is important as we had previously observed that a 5-atom NM ICL mimic using a hydrazine linkage (Fig. S1C) was unstable under prolonged incubation conditions, in particular during the generation of longer templates for DNA polymerase reactions that enable the study of eukaryotic DNA polymerases in the presence of PCNA and other accessory factors (YKC and ODS, unpublished).⁴⁰

Second, as previously reported in our study with different NM ICL mimics, including DMEDA NM ICL that we included here for comparison,³¹ we were intrigued to see that both TLS polymerases and Klenow were able to bypass NM ICLs in a 6-mer duplex upon prolonged incubation time. Although surprising at first glance, it is worth noting that 7-deazaG nucleotides with fluorescent groups attached through linkers at the 7-position are readily used by DNA polymerase in automated DNA sequencing reactions.⁴¹ This is consistent with the notion that replicative polymerases check for base-pairing through specific contacts in the minor groove, and there is some flexibility for substitution in the major groove as long as they do not disrupt base pairing or distort duplex structure.⁴²

It has already been reported that multiple repair pathways exist for ICLs in mammalian cells and that ICL structure is one of the key determinants of pathway choice .^{39, 43, 44, 45} With our improved approach allowing for the synthesis of structurally diverse and clinically relevant ICLs, the stage is now set to investigate how ICL structure affects biological responses at a detailed level.

ABBREVIATIONS:

7deazaG

2’-deoxy-7deazaguanosine

APA

(3-(4-Aminophenyl) propionic acid

BSA

bovine serum albumin

dG

2’-deoxyguanosine

DMEDA

N’N’-Dimethylethylenediamine

DMT

4,4’-dimethoxytrityl

EDTA

ethylenediaminetetraacetic acid

Fapy

formamidopyrimidine

MgCl₂

Magnesium chloride

HPLC

high performance liquid chromatography

ICL

interstrand crosslink

MA

methylamine

MWCO

molecular weight cutoff

NM

nitrogen mustard

Pol η

DNA polymerase eta

TEAA

triethylammonium acetate

TLS

translesion synthesis