Synthesis and Reactivity of Aryl Nitrogen Mustard

Dec 15, 1997 - Mustard-Oligodeoxyribonucleotide Conjugates. Michael W. Reed,* Eugeny A. Lukhtanov, Vladimir Gorn, Igor Kutyavin, Alexander Gall,...
0 downloads 0 Views 292KB Size
Download PDF
64

Bioconjugate Chem. 1998, 9, 64−71

Synthesis and Reactivity of Aryl Nitrogen Mustard-Oligodeoxyribonucleotide Conjugates Michael W. Reed,* Eugeny A. Lukhtanov, Vladimir Gorn, Igor Kutyavin, Alexander Gall, Ansel Wald, and Rich B. Meyer Epoch Pharmaceuticals, Inc., 1725 220th Street, S.E., #104, Bothell, Washington 98021. Received July 23, 1997; Revised Manuscript Received October 6, 1997X

A versatile method is described for preparing aryl nitrogen mustard-oligodeoxyribonucleotide (mustard-ODN) conjugates under anhydrous conditions. The chemistry uses DMSO soluble triethylammonium or tributylammonium salts of the ODNs. A G/A motif triplex forming ODN was chosen for study since it had been shown earlier to bind with high affinity and specificity to a duplex DNA target. A 5′-hexylamine derivative of this ODN was reacted with three different 2,3,5,6-tetrafluorophenyl ester derivatives of aryl nitrogen mustards which were designed to have different alkylation rates. An HPLC assay was used to determine reaction rates of these mustard-ODNs under various conditions. The reactivity of the mustard groups depended on chloride concentration and the presence of nucleophiles. Conjugation of mustards to G/A-containing ODNs decreased their aqueous stability. Hydrolysis and alkylation rates of these agents were consistent with reaction via an aziridinium intermediate. Rates of sequence specific alkylation within a triplex were determined by denaturing gel electrophoresis and shown to depend on inherent reactivity of the mustard group. The improved synthesis and chemical characterization of mustard-ODNs should facilitate their use as sequence specific alkylating agents and as probes for nucleic acid structure.

INTRODUCTION

Aryl nitrogen mustards are one of the oldest and most studied classes of cancer chemotherapy agents (1). Their cytotoxic effects are attributed to random alkylation of vital macromolecular targets, especially the nucleophilic N7 position of the purine bases in genomic DNA (ref 2 and references therein). Formation of interstrand crosslinks in DNA by these bifunctional alkylating agents is believed to be especially lethal, but monoalkylation, depurination, and strand scission may also contribute to cytotoxic and mutagenic effects (3). Our goal is to deliver these (and other) alkylating agents to specific genetic targets as triplex-forming oligodeoxynucleotide (ODN) conjugates in order to induce sequence specific mutations or gene knockouts. We have recently shown (4) sequence specific targeting and covalent modification of human genomic DNA by aryl nitrogen mustard-ODN conjugates (mustard-ODNs). Mustard-ODNs are also valuable probes for nucleic acid structure. For example, the mechanisms of RecA-mediated synaptic joint formation (5) and RNA polymerase-catalyzed transcription (6) have recently been explored using mustard-ODNs. As shown in Figure 1, the alkylation mechanism of the aryl nitrogen mustards is believed to occur mainly via spontaneous formation of an electrophilic aziridinium cation intermediate (7). Although aromatic aziridinium cations have not been directly detected, the first-order reactions observed for this class of mustards are in accord with the proposed mechanism. Mustards have been attached to ODNs with appropriate length linkers that position the short-lived aziridinium cation in the proximity of a * Corresponding author. Telephone: (425) 485-8566. Fax: (425) 486-8336. E-mail: [email protected]. X Abstract published in Advance ACS Abstracts, December 15, 1997.

Figure 1. Mechanism for reaction of aryl nitrogen mustardODN conjugates. The major pathway is via the aziridinium cation intermediate. Alkylation of N7 in dG is shown, but other nucleophilic centers in DNA can also be alkylated. The second (2-chloroethyl)amine group reacts in a similar manner.

dG residue after hybridization to a single-stranded (ss) or double-stranded (ds) DNA target (ref 8 and references therein). Targeted alkylation of a single DNA strand in dsDNA by a triplex-forming mustard-ODN can occur in high yield (4), and efficient interstrand cross-linking has been accomplished by conjugating a mustard group to each end of the ODN (9). Nonetheless, there are competing side reactions such as hydrolysis and nonsequence specific alkylation that can decrease alkylation efficiency in some triplex systems. To understand effects of sequence-directed DNA alkylation in more complex biological systems, we have explored the aqueous stability and reactivity of mustard-ODNs. Mustard-ODNs have been used for many years, but

S1043-1802(97)00134-1 CCC: $15.00 © 1998 American Chemical Society Published on Web 01/12/1998

Nitrogen Mustard−Oligonucleotide Conjugates

synthesis and characterization of these reactive ODNs remains a challenging problem (10). ODNs have poor solubility in organic solvents, and the (2-chloroethyl)amino functionality of biologically relevant nitrogen mustards is rapidly hydrolyzed. In addition, self-reaction of mustard groups with the ODN carrier is possible. In the course of our studies, we have developed a general method for preparation of reactive ODN conjugates in anhydrous organic solvents. In this paper, we describe synthesis of three aryl nitrogen mustard-activated esters and conjugation to a G/A motif triplex-forming ODN. These mustard groups were designed to have different inherent reaction rates. The ability to vary the reactivity of the mustard-ODNs is especially important for biological applications where uptake and degradation rates of the ODN carrier are a concern. Rates of reaction of mustard-ODNs under various conditions were analyzed by reverse phase HPLC, and alkylation of duplex DNA targets was determined by gel electrophoresis. EXPERIMENTAL PROCEDURES

Nitrogen mustards are potent alkylating agents and were handled with great care. DMSO solutions are especially hazardous, and disposable nitrile gloves were used for these operations. Excess reagents were decomposed in water. 1H and 13C NMR spectra were recorded on a Varian Gemini 300 MHz spectrometer. Infrared spectra were obtained with a Perkin-Elmer 783 spectrophotometer. Elemental analyses were performed by Quantitative Techonologies Inc. (Boundbrook, NJ). Melting points were determined on a Mel-Temp melting point apparatus in open capillary tubes and are uncorrected. All air and water sensitive reactions were carried out under a slight positive pressure of argon. Flash chromatography was performed on 230-400 mesh silica gel. Analytical thinlayer chromatography was carried out on EM Science F254 aluminum-backed, fluorescent indicator plates. 4-[Bis(2-chloroethyl)amino]benzenebutanoic acid (chlorambucil) was purchased from Fluka Chemical Co. 4-[Bis(2chloroethyl)amino]phenylacetic acid and 3-[bis(2-chloroethyl)amino]benzoic acid were made by variations of the literature method (11). 2,3,5,6-Tetrafluorophenyl trifluoroacetate (TFP-TFA) was prepared as described earlier (12). Triethylammonium bicarbonate (TEAB, 0.1 M) and tributylammonium bicarbonate (TBAB, 0.1 M) were prepared by sparging a heterogeneous mixture of the appropriate amine with CO2 until the organic layer disappeared. 2,3,5,6-Tetrafluorophenyl 4′-[Bis(2-chloroethyl)amino]benzenebutanoate (1). To a solution of 0.25 g (0.82 mmol) of chlorambucil and 0.3 g (1.1 mmol) of TFPTFA in 5 mL of dry dichloromethane was added 0.2 mL of dry triethylamine. The mixture was stirred under argon at room temperature for 0.5 h and evaporated. The residual oil was purified by column chromatography on silica gel with hexane/chloroform (2:1) as the eluting solvent to give 0.28 g (75% yield) of 1 as a colorless oil: TLC (CHCl3) Rf ) 0.6; IR (CHCl3) 3010, 1780, 1613, 1521, 1485 cm-1; 1H NMR (CDCl3) δ 7.11 (d, J ) 8.8 Hz, 2H), 7.06-6.94 (m, 1H), 6.65 (d, J ) 8.8 Hz, 2H), 3.8-3.6 (m, 8H), 2.7-2.6 (m, 4H), 2.1-2.0 (m, 2H); 13C NMR (CDCl3) δ 147.69, 144.54 (dt), 144.30 (dt), 142.30 (dd), 138.5 (dd), 129.88, 112.22, 103.17 (t), 53.61, 40.52, 33.63, 32.64, 30.99, 26.64. Anal. (C20H19Cl2F4NO2) C, H, N, F. 2,3,5,6-Tetrafluorophenyl 4′-[Bis(2-chloroethyl)amino]phenylacetate (2). 4-[Bis(2-chloroethyl)amino]phenylacetic acid was converted to TFP ester 2 using a

Bioconjugate Chem., Vol. 9, No. 1, 1998 65

procedure similar to that described for synthesis of 1 to give 0.45 g (74% yield) of 2 as a white crystalline solid: mp 99-100 °C (from hexanes/ethyl acetate); TLC (2:1 hexanes/ethyl acetate) Rf ) 0.68; 1H NMR (CDCl3) δ 7.26 (d, J ) 8.8 Hz, 2H), 6.99 (m, 1H, TFP), 6.71 (d, J ) 8.8 Hz, 2H), 3.89 (s, 2H), 3.75 (m, 4H), 3.65 (m, 4H); 13C NMR (CDCl3) δ 167.97, 147.69 (m), 144.40 (m), 142.35 (m), 139.11 (m), 145.61, 130.64, 121.00, 112.17, 103.24 (t), 53.53, 52.06, 40.45, 40.07. Anal. (C18H15NO2Cl2F4) C, H, N. 2,3,5,6-Tetrafluorophenyl 3′-[Bis(2-chloroethyl)amino]benzoate (3). To a suspension of 2.62 g (10 mmol) of 3-[bis(2-chloroethyl)amino]benzoic acid in 20 mL of dry dichloromethane was added 3.37 g (13 mmol) of TFP-TFA followed by 2 mL (14.4 mmol) of dry triethylamine. A mild exothermic reaction was observed, and the mixture formed a clear solution within 5 min. The mixture was stirred under argon for an additional 15 min and evaporated. The residual oil was purified by column chromatography on silica gel using dichloromethane as the eluting solvent. The solid was recrystallized from 15 mL of hexanes to give 3.04 g (74% yield) of 3 as a white solid: mp 56-59 °C; TLC (CHCl3) Rf ) 0.7; 1H NMR (CDCl3) δ 7.65-7.6 (m, 1H), 7.5-7.4 (m, 2H), 7.1-7.0 (m, 2H), 3.9-3.96 (m, 4H); 13C NMR (CDCl3) δ 147.75 (dt), 146.57, 144.47 (dt), 142.50 (dd), 139.23 (dd), 130.26, 128.50, 119.73, 117.91, 113.41, 103.32 (t), 53.41, 40.30. Anal. (C17H13Cl2F4NO2) C, H, N, F. (R/S)-3-[4′-[[Bis(2-chloroethyl)amino]phenyl]acetamido]-1,2-propanediol (4). To a solution of 241 mg (2.65 mmol) of 3-amino-1,2-propanediol in 5 mL of dry DMF was added 226 mg (0.53 mmol) of TFP ester 2. After 1 h, the reaction mixture was evaporated and applied onto a silica gel column (2 × 28 cm). Elution with methylene chloride/methanol (95:5) followed by concentration of the proper fractions gave 174 mg (94% yield) of 4 as a white solid: mp 76-78 °C; TLC (9:1 dichloromethane/methanol) Rf ) 0.42; 1H NMR (CDCl3) δ 7.12 (d, J ) 6.6 Hz, 2H), 6.66 (d, J ) 6.6 Hz, 2H), 5.88 (m, 1H), 3.8-3.3 (m, 9H), 1.7 (br s, 2H). Anal. (C15H22N2O3Cl2) C, H, N. Synthesis and Purification of Oligodeoxynucleotides. ODNs were prepared on an Applied Biosystems model 394 synthesizer using the 1 µmol protocols supplied by the manufacturer. Protected β-cyanoethyl phosphoramidites, CPG supports, deblocking solutions, cap reagents, oxidizing solutions, and tetrazole solutions were purchased from Glen Research (Sterling, VA). The sequences of the 65-mer dsDNA target and the triplexforming ODN (ODN-NH2) are shown in Figure 2. A 3′hexanol modification was introduced into ODN-NH2 using 2 µmol of hexanol-modified CPG support (12), and a 5′-aminohexyl linker was introduced using an N-MMThexanolamine phosphoramidite (Glen Research). Preparative HPLC purification, detritylation, and butanol precipitation of the synthetic ODNs were carried out as previously described (13). Aliquots (0.2 mg) of the 65mer ODNs were further purified by preparative gel electrophoresis for triplex cross-linking experiments. Characterization of ODNs. The concentrations of all ODNs were determined from the UV absorbance at 260 nm in phosphate-buffered saline (pH 7.2). An extinction coefficient for each ODN was estimated using a nearest neighbor model (14). For the triplex-forming ODN used in these studies, a value of 29.1 µg/mL per A260 unit was used. All modified ODNs were analyzed by reverse phase HPLC (C18 HPLC) using a 250 × 4.6 mm C18 column equipped with a guard column (Rainin Dynamax, 10 µm particle size, 300 Å pore size). A

66 Bioconjugate Chem., Vol. 9, No. 1, 1998

gradient of 5 to 65% solvent B over 30 min was used (flow rate ) 1 mL/min), where solvent A was 0.1 M triethylammonium acetate (TEAA, pH 7.5) and solvent B was acetonitrile; detection was by UV absorbance at 260 nm. Unless otherwise noted, all modified ODNs were greater than 95% pure by C18 HPLC. Typical Synthesis of Mustard-ODN Conjugates (ODN I). Dried, detritylated ODN-NH2 from a 2 µmol scale synthesis was dissolved in 0.5 mL of water and injected onto a Hamilton PRP-1 column (Reno, NV) that was equilibrated with 0.1 M TEAB (pH 7.2). The TEA salt of the ODN was eluted from the column using a gradient of 0 to 60% acetonitrile over 30 min, with a flow rate of 2 mL/min. The desired peak (∼15 min) was collected and dried in vacuo on a centrifugal evaporator. The residue was dissolved in 0.5 mL of water, and the concentration was determined to be 2.36 mg/mL. C18 HPLC analysis showed one major peak (90% purity). A 1 mg aliquot (424 µL, 71 nmol) was redried in a 1.7 mL Eppendorf tube, and the residue was dissolved in 0.2 mL of DMSO with 5 µL of ethyldiisopropylamine. A 20 mg/ mL solution of TFP ester 1 in DMSO was prepared, and 31.5 µL (0.63 mg, 1.4 µmol) was added to the ODN. The mixture was shaken for 3 h at room temperature, and then the ODN was precipitated by adding 10 mL of 2% NaClO4/acetone in a 14 mL polypropylene tube. The mixture was centrifuged at 3000 rpm for 5 min, and the pellet was sonicated with 2 mL of acetone and recentrifuged. The pellet was dried in vacuo for 15 min, and the crude product was stored at -20 °C. The crude reaction mixture was analyzed by C18 HPLC. Purification by C18 HPLC used the same gradient and column specified above. The peak eluting at 19 min was collected in ∼1 mL of TEAA/acetonitrile and immediately precipitated by adding 100 µL of 3 M sodium acetate and 4 mL of ice cold absolute ethanol. The mixture was centrifuged at 3000 rpm for 5 min, and the pellet was sonicated with 2 mL of ethanol and recentrifuged. The pellet was dried in vacuo for 15 min, and the purified product was dissolved in 0.25 mL of water. A 5 µL aliquot was removed for C18 HPLC analysis, and another 5 µL aliquot was removed for concentration determination. The bulk solution of ODN I was immediately frozen and stored at -20 °C for future use. HPLC analysis showed 98% purity. The concentration was 1.76 mg/mL (0.44 mg, 44% yield). ODN I was prepared in similar yield and purity using TBAB for the initial salt exchange. The more lipophilic TBA salt was eluted from the PRP-1 column using a gradient of 0 to 100% acetonitrile over 30 min, with a flow rate of 2 mL/min. The desired peak (∼23 min) was collected and dried in vacuo on a centrifugal evaporator as usual. ODN II and ODN III were prepared in similar yields and purities from the TBA salts. HPLC Assay for Kinetics Studies. Reverse phase HPLC analysis of mustard-ODNs was carried out using a Rainin Gradient HPLC system (Emeryville, CA) equipped with 10 mL pump heads and a Rainin Dynamax PDA-1 photodiode array detector. For kinetics studies, 10 µL of each sample was injected onto a 4.6 × 150 mm Rainin Microsorb C18 column and eluted using a gradient of 5 to 65% acetonitrile in 0.1 M triethylammonium acetate (pH 7.5) over 20 min (flow rate of 1 mL/min). ODN products were detected by UV absorbance at 260 nm, and data were integrated and analyzed using Rainin Dynamax software. Aqueous Stability of Mustard Conjugates. One hundred microliters of a 0.1 mM solution of the conjugate of interest was prepared in 20 mM HEPES buffer (pH

Reed et al.

7.2) with 140 mM KCl and 10 mM MgCl2. For the chloride-free experiments, 20 mM HEPES buffer (pH 7.2) was used with no additional salts. The 0.1 mM stock solution was immediately aliquoted to six Eppendorf tubes that were submerged in a 37 °C bath. For ODN I, ODN II, and the aminopropanediol conjugate 4, aliquots were removed at 30, 45, 60, 90, 120, and 240 min. For ODN III, aliquots were removed at 2, 4, 6, 8, and 12 h. The aliquots were immediately frozen (-20 °C) and thawed just prior to HPLC analysis. Intact mustardODNs eluted at ∼10 min, and a mixture of degradation products eluted at 5-8 min. Intact compound 4 eluted at ∼14 min. After integration, the percent of intact conjugate was plotted vs time and the half-life for disappearance of starting material was determined from the line of best fit using an equation for exponential decay. Reaction of Mustard Conjugates with a Model Nucleophile. One hundred microliters of a 0.1 mM solution of ODN I, II, or III was prepared in 20 mM HEPES buffer (pH 7.2) with 140 mM KCl and 10 mM MgCl2. Sodium thiosulfate (10 mM, 100 equiv) was added to the mixtures. The half-life for disappearance of the starting conjugate was determined as described for the stability studies. Determination of DNA Alkylation Rates for Triplex-Forming Mustard-ODNs. The duplex used in these studies was 65 bp long and contained a 33 bp homopurine-homopyrimidine run. The purine-rich strand of the duplex was 5′-end labeled by treatment with T4 polynucleotide kinase and [γ-32P]ATP under standard conditions (15). Labeled ODN was purified using a Nensorb column (NEN Research Products) and had a specific activity of ∼6000 cpm/fmol. Duplexes were formed by annealing 20 nM purine-rich strand with 40 nM complementary pyrimidine-rich strand in 20 mM HEPES (pH 7.2) with 140 mM KCl, 10 mM MgCl2, and 1 mM spermine using an incubation profile of 1 min at 95 °C and 30 min at 37 °C. After annealing, coralyne chloride (Sigma) was added to give a final concentration of 10 µM. The concentration of the labeled duplex was kept constant (20 nM), while the concentration of TFO was varied (0.2-2 µM). Triplexes were formed in capped and siliconized polypropylene microcentrifuge tubes (0.65 mL) at 37 °C in a final volume of 25 µL. The labeled duplex (22.5 µL) was combined with 2.5 µL of ODN I or ODN II (2 or 20 µM), and the solutions were incubated at 37 °C. Aliquots (2.5 µL) were removed at 0, 30, 60, 120, 240, 360, and 1320 min and stored frozen in 4 µL of loading buffer (80% formamide, 0.01% xylene cyanol, and bromphenol blue). The aliquots were thawed, and alkylated products were electrophoretically resolved in a denaturing 8% polyacrylamide gel. The labeled bands were visualized by autoradiography and quantified using a BioRad GS-250 Phosphorimager. The percent of monoalkylated labeled strand was plotted vs time, and the half-life for alkylation was determined from the line of best fit using an equation for exponential growth. RESULTS AND DISCUSSION

Triplex-Forming Mustard-ODN Model System. The specific ODN chosen for this study (ODN-NH2) is a 21-mer G/A motif triplex-forming ODN, designed to hybridize to a homopurine run in the human DQβ1*0302 allele (16). Inheritance of this allele predisposes individuals to insulin-dependent diabetes mellitus. This endogenous genetic target provides an ideal system for comparison of different triplex-forming motifs. In a

Nitrogen Mustard−Oligonucleotide Conjugates

Bioconjugate Chem., Vol. 9, No. 1, 1998 67

Figure 2. Design of triplex-forming ODN (ODN-NH2) and the dsDNA target. ODN-NH2 was reacted with the aryl nitrogen mustard TFP esters 1-3 to give ODN conjugates I-III as shown in Scheme 2. The position of the alkylated dG residue in the purine-rich strand is indicated by an arrow.

separate study (17), triplex formation in this model system was examined using a gel retardation assay. In a physiologically relevant buffer (pH 7.2 HEPES, 140 mM KCl, 10 mM MgCl2, and 1 mM spermine), we found that the G/A motif triplex-forming ODN binds with the highest association constant. This buffer was designed to model salt concentrations expected to be found in the nucleus of cells (18). Although some self-association was observed with this G-rich ODN, it did not appear to affect triplex formation, especially in the presence of the triplex specific intercalating agent coralyne (19). ODN-NH2 was prepared with a 5′-hexylamine linker in order to provide a site for attachment of the mustards. The linker length was not optimized for this target system but provides a long enough tether to allow each of the mustards studied herein to find a suitable conformation for alkylation of the targeted guanine in the homopurine-containing strand. The sequence of ODN-NH2, the site of triplex formation, and the position of the targeted dG in the purine-rich strand are shown in Figure 2. Since cell culture experiments were planned, a 3′-hexanol modification was incorporated into the ODN in order to block the 3′-end of the ODN from nuclease degradation (12). Conjugation Chemistry. Many simple conjugate groups such as amine-containing linkers can be conveniently added to the growing ODN chain during automated synthesis (20). However, the harsh deprotection conditions require postsynthetic addition of more sensitive conjugate groups. This generally involves treatment of an aqueous solution of an ODN containing a nucleophilic linker group with an electrophilic form of the conjugate group. For example, amine-modified ODNs react with “activated esters” to give amide bonds. We have found that 2,3,5,6-tetrafluorophenyl (TFP) esters are particularly convenient to make and use for ODN conjugation reactions. These compounds can be purified by flash chromatography with little decomposition, and they hydrolyze slowly in relation to some other activated esters. In addition, the chemical shift of the H4 proton (δ ) 7.0 ppm) can be monitored by 1H NMR to give an indication of the purity of these compounds.Three different aryl nitrogen mustard TFP esters were prepared as shown in Scheme 1. In our earlier work (9), we used TFP ester 1 prepared from the clinically used anticancer compound chlorambucil. The reactivity of the aniline mustards can be “tuned” by judicious selection of the substituents on the aromatic ring. Increasing the electron density of the aryl group increases the nucleophilicity of the aniline nitrogen, thus increasing the rate of cyclization to the aziridinium cation. Phenyl acetate mustard (2) and benzoate mustard (3) were prepared as described for 1 and investigated as slower-reacting analogues of 1. Early work had shown the relative rates of hydrolysis of the carboxylic acid precursors to 1, 2, and 3 to be 27, 17, and 4 when measured in 50% acetone (11).Synthesis of mustard-ODNs is problematic due to the rapid hydrolysis of the (chloroethyl)amino functional groups in neutral or basic aqueous conditions. Thus, anhydrous conditions were developed for synthesis of

Scheme 1. Synthesis of Aryl Nitrogen Mustard TFP Esters 1-3a

a

TFP is 2,3,5,6-tetrafluorophenyl.

Scheme 2. Synthesis of Mustard-ODN Conjugatesa

a The structure of ODN-NH is shown in Figure 2. Three 2 different mustard-ODNs (ODNs I-III) were prepared using TFP esters 1-3, respectively. Structures of the different aryl nitrogen mustard groups (R) are shown in Scheme 1.

mustard-ODNs as illustrated in Scheme 2. Others have used organic soluble cetyltrimethylammonium salts of the ODNs (21), but this method can be troublesome since the nonvolatile cetyltrimethylammonium chloride is not easily removed from the reaction mixtures. We recently found that triethylammonium (TEA) or tributylammonium (TBA) salts of ODNs are soluble in polar aprotic solvents such as DMSO. As the first step in mustardODN synthesis (Scheme 2, step 1), alkylammonium salts were prepared by loading the sodium form of the ODN onto a reverse phase HPLC column (equilibrated with TEA or TBA bicarbonate) and eluting with a gradient of acetonitrile. The excess volatile TEAB or TBAB is removed in vacuo to give the organic salt form of the

68 Bioconjugate Chem., Vol. 9, No. 1, 1998

Reed et al.

Figure 3. Reaction of ODN-NH2 with TFP ester 1. The progress of the reaction was followed by C18 HPLC as described in Experimental Procedures. Unreacted starting ODN was removed by HPLC to give purified ODN I.

ODN. Most ODNs are soluble in DMSO as the TEA salts (∼5 mg of ODN per milliliter of DMSO), but some modified ODNs require the extra lipophilicity provided by the TBA salts. We used TEA and TBA salts interchangeably for preparation of ODNs I-III. Preparation of the mustard-ODNs (Scheme 1, step 2) involved treatment of the DMSO solution of ODN-NH2 (TEA salt) with 20 equiv of activated ester (1, 2, or 3) and excess ethyldiisopropylamine. Progress of the reaction from ODN-NH2 to the more lipophilic mustard-ODN conjugate (I, II, or III) was easily monitored by reverse phase (C18) HPLC as shown in Figure 3. Isolation and purification of the mustard-ODN (Scheme 2, step 3) was accomplished by first precipitating the crude product from DMSO solution by adding sodium perchlorate in acetone. The resulting pellet was dissolved in TEAA buffer and immediately purified by reverse phase HPLC and precipitation from the desired fraction. Although some yield was lost during the final HPLC and precipitation steps, ODNs I-III were routinely prepared with >95% purity. Aqueous Stability of Mustard-ODNs. Reverse phase HPLC was used to study the aqueous reactivity of ODNs I-III. Mustard-ODNs were prepared at 100 µM in an aqueous buffer (pH 7.2) that contained a physiological concentration of chloride ion. Typical chromatograms that illustrate the kinetics for degradation of ODN II are shown in Figure 4A. A complex mixture of less lipophilic ODN degradation products was observed. Hydrolyses of ODN I and ODN III showed similar mixtures of degradation products. The large separation in retention times between degradation products and starting mustard-ODN allowed disappearance of ODNs I-III to be determined as shown in Figure 4B. As expected, the concentration of mustardODNs decayed exponentially, consistent with first-order reaction via a reactive aziridinium cation intermediate. Apparent residual ODN I at 240 min is due to baseline noise in the HPLC assay. Half-lives for hydrolysis of ODNs I-III were approximately 27, 57, and 216 min, respectively. Aqueous decomposition of mustard-ODNs indicated other degradation products in addition to simple hydrolysis products (Figure 1, Nu- ) OH-). Observed products for these G-rich ODNs most likely result from intramolecular or intermolecular self-alkylation. G/A-containing ODNs have been shown to self-associate in solution, presumably as parallel strands held together through G-G and A-A hydrogen bonds (22). Analysis of degra-

Figure 4. Aqueous stability of ODN II. (A) Decomposition of II in chloride-containing buffer (pH 7.2) at 37 °C was followed by C18 HPLC at 0, 30, 60, 120, and 240 min. A mixture of ODN degradation products resulted. (B) Plot of the percentage of unreacted ODN vs time for each of the mustard-ODNs. The percentage of intact ODN was calculated from integration of chromatograms.

dation products in a related mustard-ODN by gel electrophoresis provided further evidence of self-alkylation (23). Reactivity of Mustard-ODNs with a Model Nucleophile. To further characterize the reactivity of mustard ODNs, we used the same HPLC assay and followed reaction of ODNs I-III with 100 equiv (10 mM) of sodium thiosulfate. As shown in Figure 5A, only two major reaction products were observed for reaction of ODN II with this strong nucleophile. Stepwise formation and disappearance of the 7.8 min peak is indicative of a mono-thiosulfate adduct, and steady appearance of the final reaction product at 6.6 min is indicative of a bisthiosulfate adduct. The same product profile was observed for ODN I and III. As shown in Figure 5B, relative rates for reaction of ODNs I-III with excess thiosulfate depend on the inherent reactivity of the nitrogen mustard conjugate group. Half-lives for reaction of ODNs I-III with thiosulfate were approximately 18, 52, and 192 min, respectively. Effects of Buffer Composition on Reaction Rate. It appeared from our preliminary studies that adding thiosulfate to the buffer mixture shortened half-lives of mustard-ODNs. To determine the significance of this effect (and accuracy of our HPLC assay), we performed more careful kinetic measurements on ODN II. We also examined the reactivity of ODN II in a chloride-free buffer. Log[ODN II] vs time was plotted, and linear regression analysis gave first-order rate constants as shown in Figure 6. Adding a strong nucleophile (thiosulfate) slightly increased the reaction rate, presumably by trapping the aziridinium intermediate. An increase

Nitrogen Mustard−Oligonucleotide Conjugates

Figure 5. Reaction of ODN II with a model nucleophile. (A) Reaction of II with 100 equiv of sodium thiosulfate at 37 °C was followed by C18 HPLC at 0, 30, 60, 120, and 240 min. Stepwise conversion to mono- and bis-thiosulfate adducts resulted. (B) Plot of the percentage of unreacted ODN vs time for each of the mustard-ODNs. The percentage of intact ODN was calculated from integration of chromatograms.

in the mustard-ODN half-life in the presence of chloride ion was observed, consistent with earlier studies on the reactivity of chlorambucil (24). Chloride ion can react with the aziridinium ion intermediate to regenerate the starting (2-chloroethyl)amine functional group. It should be noted that ionic strength was not held constant in our studies, and Debye-Huckel or specific salt effects may also be affecting reaction rate. Effects of ODN on Reaction Rate. As described above, mixtures of self-alkylation products result from decomposition of G-rich mustard-ODNs in the absence of a competing nucleophile. The reactive aziridinium cation can be trapped by appropriately positioned nucleophilic DNA residues in a complex. To further explore the effects of the ODN carrier on reactivity of the phenyl acetate mustard, a small molecule analogue (4) was prepared by reacting TFP ester 2 with 1-amino-2,3propanediol. The aqueous stability of 4 was measured using the HPLC assay described above. As expected, two major decomposition products were observed for 4, consistent with simple hydrolysis of the (chloroethyl)amine groups. The t1/2 for hydrolysis of this derivative was 92 min as compared to 57 min for ODN II in the same buffer. Aminodiol analogue 4 reacted with thiosulfate in the same stepwise fashion with a t1/2 of 48 min as compared to 52 min for ODN II in the same buffer. These results show that the ODN carrier can significantly affect the aqueous stability of the mustard conjugate group. The effects may be sequence-dependent, thus making the stability of mustard-ODNs difficult to predict accurately. Relative half-lives of ODNs I-III under various conditions are summarized in Table 1.

Bioconjugate Chem., Vol. 9, No. 1, 1998 69

Figure 6. Kinetics plots for reaction of ODN II at 37 °C in different buffers. Buffers are described in Table 1. The HPLC assay described in Figures 5 and 6 was repeated in triplicate for each buffer. Linear regression analysis of log(percentage of intact ODN) vs time gave the rate constants (k) as indicated from the slope. Table 1. Half-Life (Minutes) for Reaction of Nitrogen Mustard Conjugatesa bufferc

ODN I

ODN II

ODN III

4b

KCl thiosulfate no chloride

27 18 -

57 52 44

216 192 162

92 48 -

a Structures of nitrogen mustards are shown in Scheme 1. t 1/2 for the disappearance of the starting compound was determined b by HPLC assay. Compound 4 is an aminopropanediol derivative of TFP ester 2. In the reactions, 0.1 mM solutions of the indicated compound were used at 37 °C. c Reactions of nitrogen mustards were in pH 7.2 HEPES buffer (20 mM). KCl contains 140 mM KCl and 10 mM MgCl2. Thiosulfate indicates addition of 10 mM Na2S2O3 to the KCl buffer. No chloride used only HEPES buffer. Except for ODN II, results are from a single representative experiment.

DNA Alkylation Rates for Triplex-Forming Mustard-ODNs. Rates of DNA alkylation by ODNs I and II within a triplex were measured by denaturing gel electrophoresis using a synthetic 65-mer dsDNA target as shown in Figure 7. Alkylation of the labeled purinerich strand in the duplex by ODN I or II gave a species with two linked strands that was observed as a slowerrunning band in the gel. This slow-running band is a result of monoalkylation of the dsDNA target. In separate sequencing experiments, the alkylated base was determined to be the dG directly 3′ of the triplex region as shown in Figure 2. Trace amounts (