Interaction of Chemically Modified Antisense Oligonucleotides with

Interaction of Chemically Modified Antisense Oligonucleotides with Sense DNA: A Label-Free Interaction Study with Reflectometric Interference Spectros...
0 downloads 0 Views 120KB Size
Anal. Chem. 1999, 71, 2850-2857

Interaction of Chemically Modified Antisense Oligonucleotides with Sense DNA: A Label-Free Interaction Study with Reflectometric Interference Spectroscopy Matthias Sauer,† Andreas Brecht,§ Klaus Charisse´,‡ Martin Maier,‡ Michael Gerster,‡ Ivo Stemmler,† Gu 1 nter Gauglitz,*,† and Ernst Bayer‡

Institute for Physical and Theoretical Chemistry, Auf der Morgenstelle 8, and Institute for Organic Chemistry, Auf der Morgenstelle 18, University of Tu¨bingen, D-72076 Tu¨bingen, Germany

Antisense oligonucleotides (ON) are regarded as potential therapeutic agents for controlling gene expression at the mRNA level. The strength of the interaction with the target sequence is one critical factor for the therapeutic efficiency of an ON. Herein, the results of studies on antisense 15mer and 20mer ONs against mdr1b-mRNA are described. The mdr1b is a member of the group that encodes the P-glycoprotein (Pgp), responsible for the phenomenon of multidrug resistance. The effects of backbone modification (DNA, phosphorothioate (PTO)), terminal modifications (hexadecyl, cholesteryl, tocopherol, polyethylenglycol, 2′-O-methyl-modified RNA) and base sequence misalignments (1 to 3 bases) on interaction kinetics and binding strength were investigated. The interaction of an immobilized sense strand with the dissolved antisense ON was monitored with a label-free optical transducer based on thin film interference (RIfS). Association kinetics were detected at a low density of immobilized ON. Thermodynamics were investigated by homogeneous phase titration of sense and antisense ON and subsequent quantification of equilibrium concentrations of unbound ON at a transducer highly loaded with sense ON. Association rate constants varied from 3.1 ((0.2) × 104 M-1 s-1 (poly(ethylene glycol)-modified DNA strand) to 4.3 ((0.1) × 104 M-1 s-1 (hexadecyl-modified strand). Binding constants varied from 1.9 (( 0.1) × 108 M-1 (cholesteryl modification) to 5 ((0.4) × 107 M-1 (tocopherol modification). Phosphorothioate ON showed a reduction in binding strength of more than 1 order of magnitude. The data presented give valuable information for the efficiency of modified antisense oligonucleotides. Over the past several years, substantial interest in the use of oligonucleotide (ON) analogues as drugs has developed.1 To * Corresponding author. † Institute for Physical and Theoretical Chemistry. ‡ Institute for Organic Chemistry. § Present address: Swiss Federal Institute of Technology, Physical Chemistry, 1015 Lausanne, CH. (1) Vlassos, V. V.; Vlassova I. E.; Pautova L. V. Prog. Nucleic Acid Res. Mol. Biol. 1997, 57, 95-143.

2850 Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

exploit the considerable therapeutic potential in the antisense2 and antigene3 approach, a wide variety of structural oligonucleotide analogues with modifications of the backbone or the terminal regions have been developed.4,5 The introduction of structural modifications in antisense oligonucleotides changes the kinetic and thermodynamic properties of the molecules in a manner which is not readily predictable on the basis of theoretical considerations. To optimize the design of modified ONs for maximal antisense activity, a reliable and effective method for the characterization of the influence of structural modifications on the antisense-sense interaction is vital. Common methods for the detection of DNA-DNA interaction are based on labeling of one, or even both, DNA strands with different labels, especially fluorescent dyes.6 Undoubtedly, these methods can be quite elegant ways for detecting DNA-DNA interaction but have the disadvantage that the strength of the duplex is affected by the presence of the label. Label-free detection of biomolecular interaction can be based on quartz-crystal balances or optical methods such as surface plasmon resonance. Another successful approach, reflectometric interference spectroscopy (RIfS), was introduced as a highly sensitive and robust technique for direct label-free monitoring.7,8 Binding detection with RIfS shows small temperature dependence, allows avoidance of gold layers, and gives perspectives for highthroughput screening.9 The technique is based on the spectral distribution of white-light reflectance from transparent thin layers. A distinct reflectance pattern with alternating maxima and minima results from interference of beams partially reflected from each interface of the interference layer (Figure 1). (2) Crooke, S. T.; Bennett C. F. Annu Rev. Pharmacol. Toxicol. 1996, 36, 107129. (3) Helene, C.; Toulme, J. Biochim. Biophys. Acta 1990, 1049, 99-125. (4) Miller, P. S. Prog. Nucleic Acid Res. Mol. Biol. 1996, 52, 261-291. (5) Milligan, J. F.; Matteucci, M. D.; Martin, J. C. J. Med. Chem. 1993, 36, 1923-1937. (6) Yu, H.; Chao, J.; Patek, D., Mujumdar, R.; Waggoner, A. S. Nucleic Acids Res. 1994, 15, 3226-3232. (7) Gauglitz, G.; Krause-Bonte, J.; Schlemmer, H.; Matthes, A. Anal. Chem. 1988, 60, 2609-2612. (8) Brecht, A.; Gauglitz, G.; Polster, J. Biosens. Bioelectron. 1993, 8, 387-392. (9) Rothmund, M.; Schu ¨ tz, A.; Brecht, A.; Gauglitz, G.; Berthel, G.; Gra¨fe, D.; Fresenius J. Anal. Chem. 1997, 359, 115-122. 10.1021/ac981057v CCC: $18.00

© 1999 American Chemical Society Published on Web 06/11/1999

Figure 1. (A) Principle of the detection of affinity interactions by RIfS (n and d are the refractive index and the physical thickness of the layer, I1 and I2 are the intensities of the light beams, reflected at the interfaces of the layer. ∆φ is the difference in the optical path length of the two beams, and φ is the phase shift upon reflection). (B) Spectral reflectance pattern due to constructive and destructive interference of the reflected radiation.

Biological material deposited at the surface during a binding event increases the optical thickness of the interference layer, leading to a shift in the interference pattern. This approach allows on-line monitoring of binding reactions with high resolution.10 The capability of this transducer to detect low-molecular-weight compounds bound to an immobilized receptor has been demonstrated recently.11 The simple test format and high sensitivity makes RIfS attractive for detecting and characterizing the interaction of antisense molecules with the target DNA. In this paper, we report on the application of RIfS to study hybridization of antisense ON to its complementary target sequences (in the following referred to as sense ON) and especially to investigate the interaction of terminal- and backbonemodified antisense ONs targeted to multidrug resistance (mdr) 1b-mRNA. Multidrug resistance is a phenomenon which was originally noticed in cultured tumor cells. Following the development of resistance to a single applied anticancer agent, these cells can rapidly become resistant to a broad spectrum of chemically diverse anticancer agents.12 A 15mer ON sequence of the mdr1bmRNA has been the target for antisense ONs, to inhibit the expression of P-glycoprotein, a transporter protein which mediates the active efflux of anticancer drugs in the resistant cells.13 3′-Modifications of ON with tocopherol, hexadecyl, poly(ethylene glycol), cholesteryl, and ON with both mixed RNA and DNA bases were investigated. The RNA 2′O-methyl modification was designed with the goal of increasing the affinity constant. The other modifications at the 3′ end are aimed at increasing the uptake of the antisense ON in the cell. The antisense molecules studied are based on a 15mer phosphorothioate backbone. Phosphodiester and phosphorothioate ON with different sequence misalignments were investigated. The effect of the various modifications of antisense molecules on both binding kinetics and thermodynamics was studied systematically in two sets of experiments. In the first set of experiments, the kinetic constants for the interaction of the sense ON with the antisense ON were detected (10) Gauglitz, G.; Brecht, A.; Kraus, G.; Nahm, W. Sens. Actuators, B 1993, 11, 21-27. (11) Piehler, J.; Brecht, A.; Gauglitz, G. Anal. Chem. 1996, 68, 139-143. (12) Doyle, L. A.; Gao, Y.; Yang, W.; Ross, D. D. Int. J. Cancer 1995, 62, 593598. (13) Alahari, S. K.; Dean, N. M.; Fisher, M. H.; Delong, R.; Manoharan M.; Tivel, K. L.; Juliano, R. L. Mol. Pharmacol. 1996, 50, 808-819.

by directly monitoring the binding of sense ON to antisense ON, immobilized on the transducer surface at low density. These binding events take place at the liquid-solid interface and are therefore influenced by surface effects such as rebinding. The second set of experiments was aimed at determining affinity constants for the interaction of the different modified antisense molecules with the 15mer target sequence of the mdr1bmRNA. The affinity constants were measured by a homogeneous titration assay in order to avoid any influence arising from the transducer surface. To this end, the sense ONs were preincubated with different concentrations of antisense ON, after which the equilibrium concentration of free sense ON was quantified by direct optical detection of diffusion-controlled specific binding to the immobilized antisense ON in a flow system. Diffusion control was obtained by using a transducer highly loaded with sense ON. Under these conditions the slope of the binding curve is directly proportional to the concentration of free antisense ON without influence of the affinity of the sense ON for the immobilized antisense ON, the surface concentration of the immobilized ON, and surface effects such as rebinding. Thermodynamic and kinetic data for a system with clear therapeutic relevance is presented. MATERIALS AND METHODS Chemicals and Biochemicals. Phosphate-buffered saline solution (150 mM NaCl; 50 mM KH2PO4) was prepared in Millipore water and titrated to pH 7.4 with 2 M NaOH. General chemicals were purchased from Fluka (Neu-Ulm, Germany). Antisense Molecules Investigated Are Listed in Table 1. (Mismatches are underlined and bold, subscript O is used for phosphodiester and subscript S for phosphorothioester.) DNA Synthesis and Purification. Oligodeoxyribonucleotides and 2′-O-methyl oligoribonucleotides were assembled on an Applied Biosystems 394 DNA/RNA Synthesizer (Applied Biosystems, Perkin-Elmer Corp., Foster City, CA) using standard phosphoramidite chemistry and TentaGel (Rapp Polymere, Tu¨bingen, Germany) or controlled pore glass (CPG) (Perseptive Biosystems, Wiesbaden, Germany) as solid support.14,15 Modified protocols adapted for the synthesis on polystyrol-polyethylenglycol (PS-PEG) tentacle polymers were used.16,17 3′-Amino-modified ONs was synthesized using 3′-C7-aminoderivatized CPG purchased from Eurogentec (Darmstadt, Germany). 3′-Conjugated with poly(ethylene glycol) ONs was prepared as recently reported.18 Cholesterol, tocopherol, and hexadecyl were introduced at the 3′-end of the ON using PS-PEG supports functionalized with 2-(4,4′-dimethoxytrityloxymethyl)-6-cholesterylmethyloxycarbonylaminohexane-1-succinate,19 1-(4,4′-dimethoxytrityloxymethyl)-2-succinyl-3-D,L-R-tocopheryl-glycerol,20 and 1-(4,4′dimethoxytrityloxymethyl)-2-succinyl-3-hexadecyl-glycerol19 for solidphase synthesis, respectively. The derivatized resins were prepared according to a procedure described elsewhere. (14) Ko ¨ster, H. Tetrahedron Lett. 1972, 1527-1530. (15) Hjerte´n, S. J. Chromatogr. 1985, 347, 191-198. (16) Bayer, E.; Bleicher, K.; Maier M. Z. Naturforsch 1995, 50b, 1096-1100. (17) Gerster M.; Maier M.; Clausen N.; Schewitz J.; Bayer E. Z. Naturforsch, 1997, 52b, 110-116. (18) Bayer, E.; Maier, M.; Bleicher, K.; Gaus, H-J. Z. Naturforsch, B: Chem. Sci. 1995, 50B, 671-676. (19) McKellar, C.; Graham, D.; Will, D. W.; Burgess, S.; Brown T. Nucleic Acids Res. 1992, 20, 3411-3417. (20) Will, D. W.; Brown, T. Tetrahedron Lett. 1992, 33, 2729-2731.

Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

2851

Table 1. Antisense Molecules Investigated phosphodiester (O): sense (s) ON smdr 20mer O20 smdr 20mer O20 2 mismatches NH2-smdr 20mer O20 for immobilization phosphodiester (O): antisense (a) ON amdr 15mer O20 amdr 20mer O20 NH2-amdr 20mer O20 for immobilization PTO (S): antisense (a) ON amdr 20mer S20 amdr 20mer S20 1 mismatch amdr 20mer S20 2 mismatches amdr 20mer S20 3 mismatches PTO (S): antisense (a) ON with terminal modification amdr 15mer S15 3′-polyethylenglycol amdr 15mer S15 3′-tocopherol amdr 15mer S15 3′-hexadecyl amdr 15mer S15 3′-cholesteryl RNA modified: antisense (a) ON amdr 15mer-RNA-DNA

Chemicals for solid-phase synthesis were purchased from Perseptive Biosystems (Wiesbaden, Germany), Applied Biosystems (Weiterstadt, Germany), and Roth (Karlsruhe, Germany). Tetraethylthiuramdisulfide (TETD) purchased from Applied Biosystems (Weiterstadt, Germany) was used as the sulfurization reagent for PTOs. Purification of the ON was performed by ion-pair-reversedphase HPLC on a Nucleosil C18, 5-µm (Grom, Herrenberg, Germany) column (25 × 0.46 cm) using a linear gradient from 20 to 50% CH3CN in 0.1 M triethylammoniumacetate (TEAA), pH 7.0, within 20 min. The characterization21,22 and the purity of the ON were assessed by electrospray mass spectrometry with an API III TAGA 6000 E mass spectrometer (Sciex, Toronto, Canada). RIfS Technique and Setup. Binding events were monitored by the use of RIfS. The experimental setup for detection of affinity interactions is shown in Figure 2. White light from a tungsten light source (20 W) was guided to the transducer using bifurcated fiber optics (PMMA; 1-mm diameter from Microparts, Dortmund, Germany). The reflected light was collected by the same fiber and spectrally detected by a diode array spectrometer (MCS 410, 512 diodes, 350-780 nm, and 16-bit nominal resolution; Carl Zeiss, Jena, Germany). The transducer is mounted in a flow cell of approximately 200-nL volume, 50-µm depth, and 2-mm width. The fiber was matched to the rear of the transducer with the help of glycerol. Sample handling was carried out by flow injection analysis (sample loop of 500 µL) using an ASIA system from Ismatec (Zu¨rich, Switzerland), including an autosampler module. Data acquisition and evaluation, as well as sample handling control, were carried out with a PC. Binding curves were recorded as apparent optical thickness nd of the interference layer vs time. The optical thickness was determined from the interference spectrum by polynomial regression of a part of the curve as described previously.10 The performance of the detection system was estimated from the noise of the baseline. A typical baseline interval (200 s) was fitted by (21) Bayer, E.; Bauer, T.; Schmeer, K.; Bleicher, K.; Maier, M.; Gaus, H-J. Anal. Chem. 1994, 66, 3858-3863. (22) Bleicher, K.; Bayer, E. Biol. Mass Spectrom. 1994, 23, 320-322.

2852 Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

5′-AGA GAA AGA GCA GAC AAG AA-3′ 5′-AGA TAA AGA GCA TAC AAG AA-3′ 5′-AGA GAA AGA GCA GAC AAG AA-3′-NH2 5′-TTC TTG TCT GCT CTT-3′ 5′-TTC TTG TCT GCT CTT TCT CT-3′ 5′-TTC TTG TCT GCT CTT TCT CT-3′-NH2 5′-TTC TTG TCT GCT CTT TCT CT-3′ 5′-TTC TTG TCT TCT CTT TCT CT-3′ 5′-TTC TTG TAT GCT CTT TAT CT-3′ 5′-TCC TTG TAT GCT CTT TAT CT-3′ 5′-TTC TTG TCT GCT CTT-3′-PEG 5′-TTC TTG TCT GCT CTT-3′-tocopherol 5′-TTC TTG TCT GCT CTT-3′-hexadecyl 5′-TTC TTG TCT GCT CTT-3′-cholesteryl 5′-UUOCO USTOGS USCOTS GsCSUS COUSUO-3′

Figure 2. Schematic setup for monitoring of affinity interactions by RIfS.

linear regression (data not shown). The standard deviation of the residuals from the fit was taken as an estimation for the rms noise of the thickness determination. It was found to be 0.8 pm. As a signal of 1 nm corresponds to a surface loading of about 1 ng/ mm2 DNA, changes in surface loading of less than 3 pg/mm2 can be resolved by RIfS. The low drift of the signal (less than 0.005 pm/s) demonstrates the high stability of the setup. Surface Modification. Interference layers (10 nm of Ta2O5 and 330 nm of SiO2 on float glass) were manufactured by Schott (Mainz, Germany) using a plasma-impulse chemical vapor deposition CVD process. Prior to use, the layers were cleaned in a freshly prepared mixture of 60% H2SO4 and 40% H2O2. This procedure results in a highly hydrophilic and acidic surface. For the covalent

immobilization of NH2 terminated ON, the following procedure was used: In a first step, the surface was silanized by a 1-h reaction of (3-glycidoxypropyl)trimethoxysilane (GOPTS). After rinsing with acetone the resulting surface was modified with 6-aminohexanoic acid with a solution containing 1 mg of 6-aminohexanoic acid per 1.25 µL Millipore water. The surface was activated using equal volumes of a 1 M solution of N-hydroxysuccinimide in dimethylformamide and a 1.5 M solution of N,N′-diisopropylcarbodiimide in dimethylformamide mixed directly on the surface and incubated for 4 h. After rinsing with acetone, the covalent immobilization of the ON was carried out with an ink-jet system (Microdrop, Norderstedt, Germany). The system consists of a glass capillary mounted concentrically in a piezoelectric actuator. By applying high voltage pulses to the piezo actuator, the NH2 terminated ON (2 mg/mL in ultrapure water) was delivered to an area of 2 mm2 of the activated transducer in a series of droplets. A droplet frequency of 0.5 kHz was used. The experimental details for this procedure will be published elsewhere.23 For measurements of the kinetics and thermodynamics of its interaction with the mdr antisense sequence, the sense NH2-smdr 20mer was immobilized. Conversely, the antisense NH2-amdr 20mer was immobilized in binding experiments in which ON with the mdr sense sequence was investigated. Measurement Procedure and Data Analysis. Thermodynamics. The affinity constants for the interaction of antisensesense oligonucleotides were measured in the liquid phase by a homogeneous phase titration scheme.24 A constant concentration of sense ON of typically 16 nM was used throughout a titration experiment. The concentration of antisense ON in the mixture was systematically varied to match the affinity of each sample (typically 0-400 nM). The investigated antisense and sense ON were preincubated in phosphate-buffered saline solution for 10 min to obtain hybridization equilibrium. The equilibrium concentration of free unhybridized sense ON was probed kinetically by direct optical detection of specific binding to the immobilized antisense strand with RIfS. From a set of measurements, the equilibrium concentration of free sense ON was obtained as a function of the concentration of antisense ON. The affinity constant K of the sense-antisense ON pair was derived by fitting a Marquart-Levenberg nonlinear least-squares fit with the affinity constant as a free fit parameter of the model function (1)24 to these data. The capabilities of this model function and fitting procedure were investigated elsewhere.24

cR,bind ) c0,R -

[

c0,R + c0,L +

1 K

2

x

(c

0,R

+

+ c0,L + 4

1 K

2

) -c

]

0,Rc0,L

(1 - Cd) (1)

The following notation is used: cR,bind is the concentration of free, not hybridized receptor (sense oligonucleotide), c0,L is the initial concentration of ligand (antisense oligonucleotide), c0,R is (23) Stemmler, I.; Sauer, M.; Gauglitz, G.; Brecht, A. Nucleic Acids Res., in press. (24) Piehler, J.; Brecht, A.; Giersch, T.; Hock, B.; Gauglitz, G. J. Immunol. Methods 1997, 201, 189-206.

the initial concentration of receptor (sense oligonucleotide), K is the affinity constant, and Cd is an additional parameter which considers disturbance of equilibrium. For the homogeneous titration measurements, the 3′ aminolabeled sense mdr oligonucleotide (3′-NH2-smdr) was immobilized as described above. Transducers with a maximum binding capacity of at least 0.8 ng/mm2 of sense strand were used to ensure mass transport limited binding of the free antisense oligonucleotide. Under these conditions, maximum binding rates independent of the affinity of the target oligonucleotide for the immobilized complementary strand were guaranteed. The preincubated samples were injected for 20 s at a flow rate of 160 µL/min. The binding of the free target oligonucleotide to the immobilized 3′-NH2-smdr was recorded at a flow rate of 30 µL/min for 200 s. Linear binding curves, proving constant binding rates, were observed. For regeneration of the surface between each cycle a pulse of 100 s of 10 mM HCl at a flow rate of 160 µL/min was used. Kinetics. The association rate constant ka was derived by fitting the binding curves to a model, assuming pseudo-first-order association kinetics.25 The surface coverage Γ as a function of time is given by

Γ ) Γ0(1 - e-(ks)t)

(2)

Here, ks is the exponential time constant and ks ) kac + kd, where kd is the dissociation rate constant of the binding event. Equation 2 holds only under conditions where diffusion plays no role in the binding process. Transducers with low surface capacity (maximum surface coverage, Γmax, was 250 pg/mm2) were used to ensure that the determined binding follows the abovementioned type of reaction kinetics. Under these conditions, the binding to the immobilized ON is slower than the diffusion process. Therefore, the concentration of antisense ON in close vicinity to the surface is constant and identical to the bulk concentration. The model function, 2, was fitted using a nonlinear least squares algorithm (software, ORIGIN from MicroCal, Northampton, MA). The exponential time constants, ks, were plotted against the antisense-oligonucleotide concentrations. The association rate constant was determined from the slope of this curve, and the dissociation rate constants were obtained by fitting an exponential decay to the dissociation part of the binding curve. RESULTS AND DISCUSSION Oligonucleotide Immobilization. The immobilization of the 20mer sense mdr phosphodiester oligonucleotide with no mismatches (smdr-O20-0MM) on a surface of 6-aminohexanoic acid with the microdrop system resulted in a maximum binding capacity of 1 ng/mm2. The maximum binding capacity was derived from the maximum signal of a binding curve obtained by injecting the binding partner (antisense ON) under saturating conditions. This maximum binding did not increase with further increase in the concentration of the ON complementary to the immobilized strand. A typical binding curve of the complementary 20mer antisense mdr PTO oligonucleotide with no mismatches (amdrS20-0MM) to the immobilized smdr-O20-0MM is shown in Figure (25) Eddowes, M. Biosensors 1987, 3, 1-15.

Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

2853

Figure 4. Binding of amdr-S15-RNA to smdr-O20-0MM.

Figure 3. Binding curve amdr-S20-0MM to immobilized smdr-P200MM (250 nM).

3. The hybridization resulted in a fast increase of optical thickness by about 450 pm during the association phase. After the incubation period the rinsing of the flow cell lead to a decrease of optical thickness according to the dissociation of the DNA double strand. Complete regeneration of the surface was achieved with 20 mM HCl. At least 30 regeneration cycles were possible without significant loss of surface-binding capacity. To exclude interference of nonspecific binding with the detected hybridization, all antisense molecules studied were tested for nonspecific binding before kinetic and thermodynamic experiments were carried out. To this end, the antisense ON was injected in a flow cell with immobilized ON of the same sequence so that specific hybridization was not possible. During the incubation period no changes of optical thickness were detected. Interaction Measurements for Kinetics. Monitoring the binding events at different concentrations of free antisense oligonucleotide on transducers with low binding capacity allowed kinetic characterization of the interaction process. Kinetic studies were carried out with 13 antisense oligonucleotides designed with different modifications and sequence misalignments by recording binding curves at concentrations between 5 and 600 nM. Typical binding curves of an RNA-modified 15mer PTO (amdr-S15-RNA) for the concentrations 21, 53, 106, 212, 318, and 530 nM are shown in Figure 4. The increase in optical thickness up to 480 s corresponds to the association phase of the binding event. The following decrease in optical thickness up to 600 s corresponds with the dissociation phase during injection of buffer. Subsequent regeneration with HCl leads to an abrupt drop of the binding curve around 610 s. A second regeneration pulse corresponds to the drop of the curve near 740 s. After further rinsing with buffer the signal returns to the baseline. For determination of the association rate constant, the binding curves were evaluated by fitting eq 2, assuming pseudo-first-order kinetics for the hybridization between antisense and sense strands. The exponential time constants, ks, of the association phase of the binding events, determined by the fitting procedure, were plotted versus the antisense concentration, as shown in Figure 5. The association rate constants, ka, were obtained from the slopes of the secondary plot curves. 2854 Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

Figure 5. Determination of the association rate from the exponential time constants ks; the constant ka of 3.2 ( 0.3 × 104 M-1 s-1 was the result for amdr-S15-RNA.

The resulting association constants for the different antisense molecules are summarized in Table 2. The determination of dissociation rate constants was influenced by the effect of immediate rebinding of the dissociated compound to the immobilized complementary ON during the dissociation phase. This rebinding allows only the estimation of a lower limit for the dissociation rate constant, kd, as summarized in Table 2. The decrease of the association rate constants in the series from amdr-S20-0MM to amdr-S20-3MM-Var2 is significantly lower than the decrease of the affinity constants. Obviously the decrease of the affinity constants originates from a corresponding increase of the dissociation rate constants. Interaction Measurements for Thermodynamics. The approach to assessing the affinity constants by a homogeneous titration assay, followed in this work, can be viewed as a binding inhibition assay, where the inhibitor (antisense ON) is the same molecule as that immobilized. The homogeneous titration assay is based on the determination of the concentration of free sense molecule in equilibrium with the antisense strand by detecting sense ON binding to an immobilized complementary antisense ON. As described in the section with the subheading of Thermodynamics, the free sense molecule was titrated by adding increasing antisense-strand concentrations in successive samples. This strategy gives access to affinity constants without interference from surface effects such as rebinding and without influences

Table 2. Association Rate Constants ka, Dissociation Rate Constants kd, and Affinity Constants K compound

ka (M-1 s-1) (determined at the liquid-solid interface)

kd (s-1) (determined at the liquid-solid interface)

K (M-1) (determined in the homogeneous phase)

amdr-S15-RNA amdr-S15-3′-toc amdr-S15-3′-hexadecyl amdr-S15-3′-chol amdr-S15-3′-PEG amdr-S20-0MM amdr-S20-1MM amdr-S20-2MM amdr-S20-3MM-Var1 amdr-S20-3MM-Var2 amdr-S15-0MM amdr-O20-0MM smdr-O20-2MM

(3.2 ( 0.3) × 104 (3.7 ( 0.1) × 104 (4.3 ( 0.1) × 104 (4.1 ( 0.2) × 104 (3.1 ( 0.2) × 104 (3.1 ( 0.2) × 104 (2.4 ( 0.18) × 104 (1.95 ( 0.08) × 104 (1.07 ( 0.07) × 104 (1.11 ( 0.05) × 104 (3.1 ( 0.15) × 104 (3 ( 0.1) × 104 (1.9 ( 0.2) × 104

>0.001 >0.003 >0.002 >0.001 >0.0024 >0.005 >0.013 >0.012 >0.007 >0.014 >0.009 >0.001 >0.009

(1.5 ( 0.1) × 108 (5 ( 0.4) × 107 (1.3 ( 0.1) × 108 (7.5 ( 0.3) × 107 (1.9 ( 0.1) × 108 (3.2 ( 0.7) × 108 (7 ( 0.2) × 107 (3.3 ( 0.15) × 107 (2.5 ( 0.12) × 107 (7.4 ( 0.32) × 105 (1.5 ( 0.1) × 108 (2.0 ( 0.12) × 109 (7.0 ( 0.2) × 107

Figure 6. Inhibition of binding of smdr-O20-0MM to immobilized amdr-O20-0MM through titration with amdr-S20-0MM in various concentrations.

caused by the affinity and the surface concentration of the immobilized antisense ON. The change of sensor response due to titration of the free binding sense molecule with several concentrations of target antisense strand is shown in Figure 6 for smdr-O20-0MM and amdrS20-0MM. The slope of the binding curves decreases with increasing concentration of the target strand. The drop of the curve to below the baseline corresponds to the refractive-index effect resulting from addition of HCl. The binding curves return to the original baseline level after injecting buffer. The titration curve obtained from the slope of these binding curves, normalized to the blank value, and the fit of the model function, eq 1, are shown in Figure 7. The affinity constants obtained in the other titration experiments are summarized in Table 2. Backbone Effects. Comparison of the affinity constants of fully complementary 20mer phosphodiester (amdr-O20-0MM) and of a PTO of the same sequence (amdr-S20-0MM) shows an affinity reduction of as much as 1 order of magnitude for the PTO (from 2 × 109 M-1 to 3.2 × 108 M-1). This decrease in affinity seems to be the price for a better metabolic stability of antisense molecules based on the phosphorothioate backbone modification. Association rate constants were found to be nearly identical for thioates and diesters. Terminal Modification Effects. The affinity constant of the unmodified 15mer amdr-S15-0MM was found to be 1.5 × 108 M-1.

Figure 7. Titration curve of amdr-S20-1MM with smdr-O20-0MM (16 nM); an affinity constant of 7 ( 0.2 × 107 M-1 was determined by the fit.

Figure 8. Influence of structural modification on affinity constant and association rate constant.

The results for the other modified antisense molecules are compared to this affinity constant in Figure 8. All affinity constants were obtained by the homogeneous titration assay as described in the section Measurement Procedures and Data Analysis. The association rate constants of the modified ON relative to those of the unmodified ON are also shown in Figure 8. They were measured independently from the affinity constants by evaluation of the association part of the binding curves at the heterogeneous phase as described in the section Measurement Procedures and Data Analysis. According to Figure 8 structural modifications have a pronounced impact on both affinity and the kinetic constant of the interaction of the antisense molecule with the target sequence. Because a high affinity constant is a basic requirement for any biological antisense effect, these data allow preselection of promising multidrug resistance antisense candidates prior to Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

2855

Figure 9. Reduction of affinity constant as seen in dependence on increasing number of sequence mismatches in amdr-S20.

biological antisense testing. As can be seen in Figure 8, the RNAmodified ON shows the same affinity constant and association rate constant as the unmodified ON. The RNA modification is therefore not suitable to increase the affinity to the target sequence as intended during the design of the molecule. A profound reduction (66%) in the affinity constant was observed in the 3′tocopherol-modified phosphorothioester, whereas the association rate constant shows only a 19% increase compared with that of the unmodified molecule. The poly(ethylene glycol) modification of the 3′ terminus resulted in a 50% reduction of affinity compared with that of the unmodified amdr-S20. The association rate constant was not influenced by this modification. The association rate constant of hexadecyl-modified amdr-S20 increased by 38% despite a reduction of the affinity constant to 86% of that of the unmodified molecule. Because of the theoretical relation, K ) ka/kd, between the affinity constant, K, and the association and dissociation rate constants, ka and kd, the reduction of the affinity constant must result from a corresponding increase of the dissociation rate constant. The relation K ) ka/kd, cannot be used to obtain valid data for the affinity constant, K, from the experimental values for ka and kd (Table 2) because the kinetic constants were measured at the liquid-solid interface and are therefore influenced by surface effects. The difference between the affinity constants, K, obtained in the homogeneous phase (Table 2), and the upper limit of K, calculated according to the relation K ) ka/kd, with the values for ka and kd from Table 2, is in the range of 1 order of magnitude. This difference shows the extent of the influence of surface effects, especially rebinding, upon the interaction between the ON. The cholesteryl-modified PTO shows an increased affinity constant and association rate constant. A 26% increase in the strength of hybridization was found for the association, while kinetics were accelerated to 132% as compared with the unmodified molecule. According to thermodynamic and kinetic properties, the 3′ cholesteryl-modified multidrug resistance antisense PTO can therefore be selected as the most promising candidate for further biological testing concerning suppression of the mdr1b gene. Mismatch Effects. Both the affinity and the association rate constants show systematic changes dependent on the number and relative position of mismatches in the investigated sequences. As can be seen in Figure 9, the affinity constant of the interaction between fully complementary 20mer phosphorothioate amdr-S200MM and smdr-O20-0MM ((3.2 ( 0.7) × 108 M-1) is reduced to (7 ( 0.2) × 107 M-1 after introducing one mismatch. Introduction 2856 Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

of a second mismatch results in a further reduction to (3.3 ( 0.15) × 107 M-1. An affinity constant of (2.5 ( 0.12) × 107 M-1 was obtained with three mismatches (not shown in Figure 9). In correspondence with the changes in affinity, we found a lowering of the association rate constant from (3.1 ( 0.2) × 104 M-1 s-1 (fully complementary sequence) to (2.4 ( 0.18) × 104 M-1 s-1 (one mismatch), (1.95 ( 0.08) × 104 M-1 s-1 (two mismatches), and finally (1.07 ( 0.07) × 104 M-1 s-1 (three mismatches). The influence of the relative position of sequence misalignments in an oligonucleotide upon the affinity for the complementary target sequence was further investigated with RIfS. A mismatch in direct proximity to another misalignment (amdr-S203MM-Var1) showed an affinity constant of 7.4 × 105 M-1, which was about 1.5 orders of magnitudes lower than the affinity constant of amdr-S20-3MM-Var2 (2.5 × 107), a variant with the same sequence but three evenly distributed mismatches. Obviously a single misalignment in the terminal region of the 20mer phosphorothioates is more easily compensated. These findings are in agreement with results of systematic experiments with 8mer phosphodiester published by Persson et al.26 The experiments with different sequence misalignments allow a comparison of the effect of terminal or backbone modification on binding strength with the effect of base mismatches. CONCLUSION AND OUTLOOK Herein, we have investigated the kinetic and thermodynamic properties of oligonucleotides (ON) structurally modified to improve metabolic stability and pharmacokinetic properties and to increase the affinity of the antisense for the sense strand. An affinity constant higher, or at least comparable, to the hybridization strength in the natural interaction between unmodified DNA and mRNA strands is a basic requirement for the selectivity and biological antisense activity of ON. We have demonstrated that a label-free optical transducer technique based on thin film interference (RIfS) gives access to a comprehensive kinetic and thermodynamic characterization of antisense-sense interaction. One approach to defeating this multidrug resistance (mdr) is to inhibit translation of the mdr1b-mRNA through binding of an antisense ON to functional regions of the mRNA strand. Affinity constants of interaction of 15mer mdr antisense ON, with various backbone and terminal modifications, were measured in homogeneous phase. The homogeneous titration assay used gives access to affinity constants not influenced by binding events on the heterogeneous liquid-solid phase of the transducer. The affinity constants are therefore obtained under homogeneous phase conditions comparable to the interaction under intracellular or in vivo conditions. Binding constants varied from 1.9 ( 0.1 × 108 M-1 (cholesteryl modification) to 5((0.4) × 107 M-1 (tocopherol modification). Phosphorothioate (PTO) ON showed a reduction in binding strength of more than 1 order of magnitude as compared with the phosphodiester analogue. The binding kinetics of the different mdr antisense molecules to the target sequence were investigated by evaluation of the association part of binding curves obtained at transducers with low binding capacity for various concentrations of antisense (26) Persson, B.; Stenhag, K.; Nilsson, P.; Larsson, A.; Uhlen, M.; Nygren, P. A. Anal. Biochem. 1997, 246, 34.

molecules. Association rate constants varied from 3.1((0.2) × 104 M-1 s-1 (3′ poly(ethylene glycol)-modified 15mer ON) to 4.3((0.1) × 104 M-1 s-1 (3′ hexadecyl-modified 15mer ON). According to the thermodynamic and kinetic data presented, the 3′ cholesteryl-modified mdr PTO showed an increase in the affinity and the association rate constants of 25 and 32%, respectively, compared with the unmodified PTO. This result allows a preselection of this modification for further biological tests characterizing uptake efficiency, metabolic stability, and biological antisense effect. Other modifications, for example a 3′ tocopherol variant, showed clear decreases in affinity and therefore poorer prospects with regard to biological antisense effect. The level of performance achieved with RIfS in characterizing kinetic and thermodynamic aspects of ON target strand interaction opens the door for investigation of other structural classes as potential antisense molecules. Aside from the design of new

antisense molecules, a systematic study of antisense-sense interaction using a general screening approach based on RIfS could generate binding data in the context of a systematic analysis of the targeting problem. A multiple parallel RIfS setup, established in our group, opens up the possibility of following a highthroughput screening approach. Further studies with the investigated antisense ONs are in progress which are aimed at measuring the antisense effects and rate of cellular uptake in biological systems. This will allow a comparison of the thermodynamic and kinetic results with the biological antisense efficiency.

Received for review September 23, 1998. Accepted March 21, 1999. AC981057V

Analytical Chemistry, Vol. 71, No. 14, July 15, 1999

2857