Bioconjugate Chem. 1997, 8, 885−890
885
Libraries of Multifunctional RNA Conjugates for the Selection of New RNA Catalysts Felix Hausch and Andres Ja¨schke* Institut fu¨r Biochemie, Freie Universita¨t Berlin, Thielallee 63, 14195 Berlin, Germany. Received June 20, 1997X
An in vitro selection system was developed for the selection of RNA molecules catalyzing bimolecular reactions between small reactants. The system is based on the direct selection protocol and involves libraries of multifunctional RNA conjugates rather than unmodified RNA transcripts. For the preparation of RNA conjugate libraries, a dinucleotide analog has been designed and synthesized containing a poly(ethylene glycol) linker with an embedded photocleavage site and a terminal attachment site for coupling potential reactants. Reactants are first coupled to the dinucleotide analog by activated ester chemistry and then ligated to the 3′-ends of enzymatically prepared RNA pool molecules, giving libraries of complex conjugates. Species that become attached to biotin on incubation with a biotinylated partner are isolated using streptavidin-derivatized matrices and then subjected to a photocleavage step. Selective cleavage of the linker releases only those RNA species in which reaction has taken place at the linker-coupled reactant, while products with the biotin attached to internal positions of the RNA part remain immobilized. Efficient photocleavage is achieved by laser irradiation at 355 nm, and the released RNAs are intact and amplifiable by reverse transcription. All steps are shown to be compatible with the overall selection procedure, as was shown by performing a model selection cycle. Besides allowing a broader scope of reaction types to be selected for, the strategy relieves the RNA from the requirement to possess substrate properties as well as catalytic activity, and the use of a cleavable linker will suppress the selection of catalysts for side reactions.
INTRODUCTION
Combinatorial RNA1 libraries have found increasing use for the identification of new ligands and catalysts (Gold et al., 1995; Lorsch and Szostak, 1996). In the most basic versions of in vitro selection or SELEX (for Systematic Evolution of Ligands by EXponential enrichment) (Ellington and Szostak, 1990; Tuerk and Gold, 1990), a pool of RNA sequence variants is subjected to a selection (or partitioning) step, and the selected molecules are copied and enzymatically amplified. The ability to iterate the selection amplification cycle ideally allows the isolation of the most active sequences, even when they are exceedingly rare in the initial pool or when they have only a small advantage over competitors. This iterative procedure is possible only because nucleic acids can be copied enzymatically and is not available for organic small-molecule libraries and chemically synthesized peptide or carbohydrate libraries. Due to this property, recursive deconvolution, tagging, or binary coding is unnecessary, and the complexity of libraries used in SELEX experiments is great compared to other classes of compounds. * Author to whom correspondence should be addressed [telephone (49) 30 838 6023; fax (49) 30 838 6413; e-mail
[email protected]]. X Abstract published in Advance ACS Abstracts, October 15, 1997. 1 Abbreviations: BSA, bovine serum albumine; cDNA, copyDNA; CPG, controlled pore glass; DNA, deoxyribonucleic acid; DMF, N,N-dimethylformamide; DMSO, dimethyl sulfoxide; DMT, 4,4′-dimethoxytrityl; dNTPs, deoxyribonucleoside 5′-triphosphates; DTT, dithiothreitol; HEPES, N-(2-hydroxyethyl)piperazine-N′-(2-ethanesulfonic acid); MALDI-TOF, matrixassisted laser desorption/ionization time-of-flight (mass spectrometry); NTPs, (ribo)nucleoside 5′-triphosphates; PCR, polymerase chain reaction; PEG, poly(ethylene glycol); RNA, ribonucleic acid; RT-PCR, reverse transcription-polymerase chain reaction; SELEX, Systematic Evolution of Ligands by EXponential enrichment; TBDMS, tert-butyldimethylsilyl; Tris, tris(hydroxymethyl)aminomethane.
S1043-1802(97)00115-8 CCC: $14.00
The most successful strategy for the isolation of RNA molecules with catalytic properties is direct selection. This process requires that the desired reaction be configured so that the catalytic molecules are themselves changed in a way that provides a basis for their enrichment during the selection step (Williams and Bartel, 1996). Selection criterion is ideally the catalysis of a chemical reaction. However, as a catalyst is per definition required to leave the reaction unchanged, catalysis must be coupled with some other principle that leads to a selectable change in the catalyst molecule, and the way this is achieved is self-modification, letting one and the same molecule act as both catalyst and substrate. The principle is illustrated in Figure 1a. A synthetic combinatorial RNA pool (typical complexity 1013-1016 different species) is incubated with a potential reactant that carries an anchor group, which is either a specific functional group (e.g., thiol, thiophosphate, primary aliphatic amine) or an affinity tag (e.g., biotin). Some RNA molecules react and are then covalently linked to that anchor group. These species can be specifically separated from the unreacted excess RNA using suitably derivatized solid supports (e.g., activated thiopropyl agarose or streptavidin agarose). The isolated species are then enzymatically amplified, and the scheme is repeated several times. This strategy has been successfully applied to the identification of RNA catalysts for ligation, alkylation, acylation, transesterification, and phosphorylation reactions (Bartel and Szostak, 1993; Illangasekare et al., 1995; Lorsch and Szostak, 1994; Wilson and Szostak, 1995). Despite the success of direct selection, however, there are several drawbacks that severely limit its application: • Since one of the reactants is always the RNA itself, only catalysts for self-modifying reactions of RNA can be identified. If modified monomers were incorporated into RNA pools (Jensen et al., 1995; Wecker et al., 1996), their different chemistry could be exploited, too, but alignment of the substrates has always been controlled by base© 1997 American Chemical Society
886 Bioconjugate Chem., Vol. 8, No. 6, 1997
Hausch and Ja¨schke
to acquire a selectable change in a characteristic property as a result of a successful catalysis. This is achieved by coupling one of the reactants (A) to each individual RNA sequence. To allow optimal orientation of the reactant and the RNA, both are tethered by a flexible polymeric linker. By high dilution during the reaction step, selfmodification (i.e., reaction of the RNA’s “own” reactant) is favored over true intermolecular catalysis. After reaction, the products AB would then be linked to both the anchor group (allowing isolation) and the RNA (allowing amplification). The incorporation of a cleavage site within the linker allows discrimination of side products at the RNA backbone in favor of the desired reaction products at the reactant A. EXPERIMENTAL PROCEDURES
Figure 1. (a) Direct selection of RNA catalysts. The transcribed RNA is incubated with the reaction partner (X) carrying an anchor group (e.g., biotin). The reacted species (including reaction products at internal positions) are separated from the excess of inactive species, e.g., by affinity chromatography. Amplification by RT-PCR yields an enriched DNA pool that is subjected to additional rounds of selection and amplification. (b) Direct selection with linker-coupled reactants. Reactant A attached to a flexible PEG linker is coupled to the RNA transcript by T4 RNA ligase. The RNA conjugates are incubated with reactant B carrying a biotin moiety, which allows subsequent immobilization of the reaction products on streptavidinagarose. The correct reaction products (reactions at reactant A) can be specifically released by photocleavage of the linker, whereas the products of side reactions with the biotin attached to internal positions of the RNA remain immobilized and are not further amplified.
pairing interactions. Catalysts for reactions between two small reactants, being the basis of biochemistry, prebiotic chemistry, and organic synthesis, could so far not be isolated using direct selection. • To be selected, the catalytically active RNA molecule must also possess substrate properties. RNA molecules, being excellent catalysts but poor substrates, are eliminated in the selection step. • Selection criterion is the attachment of the anchor group, while the position of attachment is irrelevant. Therefore, catalysts for side reactions at different positions of the RNA (Illangasekare et al., 1995; Lorsch and Szostak, 1994) or for different chemistries (Wilson and Szostak, 1995) are enriched and finally isolated along with the desired species, as long as these modifications do not interfere with reverse transcription. Here we decribe an approach developed to overcome these limitations involving multifunctional RNA conjugates (Figure 1b). To select for catalysts of a general bimolecular reaction A + B f AB, the active RNAs have
General. UV-MALDI-TOF mass spectra were obtained on a Bruker Reflex, and IR-MALDI-TOF mass spectra were obtained on a Vision 2000, Model SEQ 1-23, Thermo Analysis, Schwartz, Hempstead, U.K. Laser experiments were performed with a Nd-YAG laser, Spektron, SL800 G. Synthesis of the Dinucleotide Analogs 1a and 1b. The photocleavable building block 2 was synthesized as described (Ordoukhanian and Taylor, 1996) and incorporated into the dinucleotide analog 1a by solid phase synthesis on an Applied Biosystems 391 synthesizer using a 1 µmol RNA cycle. 3′-Aminomodifier-C6-CPG support (3), DMT-hexaethylene glycol phosphoramidite (4), N4-benzoyl-5′-O-DMT-2′-TBDMS-cytidine-3′-O-(2-cyanoethyl-N,N′-diisopropylphosphoramidite) (5), and bis(cyanoethyl)-N,N-diisopropylphosphoramidite (6) were from Chemgenes, Waltham, MA (Scheme 2). Cleavage from the solid support and deprotection were performed in a 1 mL mixture of 33% aqueous ammonia and ethanol (3:1) at 55 °C for 24 h and overnight incubation of the lyophilized mixture with 0.4 mL 1 M tetrabutylammonium fluoride in DMF (Aldrich) at room temperature. This solution was diluted with 5 mL of 60 mM NH4OAc and passed over a 2 mL DEAE-Sephadex A-25 column equilibrated with 0.05 M triethylammonium acetate, pH 7, followed by a 15 mL wash with 0.1 M triethylammonium acetate, pH 7, and elution of the product with 2 M triethylammonium acetate, pH 7. The product fractions were lyophilized and purified by reversed phase chromatography on a HP 1028B liquid chromatograph with a Beckman Ultrasphere C18 column (4.6 × 250 mm, 80 Å pore) using a 0.1 M triethylammonium acetate gradient, pH 7, containing 0-80% acetonitrile with compound 1a eluting at ∼25% acetonitrile with a total yield of 22%. Fourteen nanomoles of this compound was biotinylated with 0.7 mg of sulfosuccinimidyl 6-(biotinamido)hexanoate (Sigma) in 100 µL of 0.1 M K2HPO4 buffer, pH 8, for 12 h at 25 °C. After quenching with 10 µL of 1 M NH4OAc, purification by HPLC yielded 5.3 nmol of 1b. Transcription. For the preparation of pool RNA, a randomized DNA pool was synthesized as described (Famulok, 1994). Thirty picomoles of this doublestranded DNA template were transcribed in a 100 µL reaction mix containing 80 mM HEPES, pH 7.5, 22 mM MgCl2, 1 mM spermidine, 10 mM DTT, 0.12 mg/mL BSA, 4 mM of each ribonucleoside triphosphate (Boehringer Mannheim), 10 µCi [R-32P]CTP, 100 units of T7 RNA polymerase (Stratagene), and 60 units of RNasin (MBI Fermentas) for 4 h at 37 °C. After DNA digestion with 20 units of DNase I (Boehringer Mannheim) for 30 min at 37 °C, RNA transcripts were purified by electrophoresis on a denaturing 8% polyacrylamide gel, elution, and ethanol precipitation.
Multifunctional RNA Conjugates
For the preparation of 25-mer RNA, a double-stranded DNA template was obtained by hybridization of two oligonucleotides synthesized by standard phosphoramidite chemistry [sequences: 3′-d(AG ATT ATG CTG AGT GAT ATC CTC GAG TCG GAA GTG ACG AGG TGG)-5′ and 5′-d(TC TAA TAC GAC TCA CTA TAG GAG CTC AGC CTT CAC TGC)-3′]. This template was transcribed under the same conditions as for the RNA pool and purified by electrophoresis on a 15% denaturing polyacrylamide gel. Ligation. Two hundred picomoles of the 25-mer RNA transcript were incubated with 400 pmol of dinucleotide analog in a 50 µL reaction volume containing 50 mM HEPES, pH 7.8, 20 mM MgCl2, 50 µg/mL BSA, 3.5 mM DTT, 10% DMSO, 0.1 mM ATP, 30 units of RNasin, and 120 units of T4 RNA ligase (MBI Fermentas) at 4 °C for 4 h, and the ligation product was purified as described for the transcript. For the ligation of the RNA-pool, the RNA molecules were first hybridized with a 3-fold excess of a 20-mer deoxyoligonucleotide [sequence: 3′-d(CCG TGG TGC CAG CCT AGG TG)-5′] complementary to the 3′-end of the pool RNA by heating to 95 °C for 1 min and slowly cooling down to room temperature. Ligation conditions were similar to those for the 25-mer, except for a higher (5-fold) excess of 1a and a longer ligation time (16 h). Ligated pool RNA was purified on a 8% denaturing polyacrylamide gel. Photolysis. Fifty picomoles of the ligated 25-mer dissolved in 80 µL of 0.1 M Tris-HCl, pH 7, and 10 mM EDTA was irradiated in a quartz cuvette with a NdYAG laser (5 mJ/pulse, 10 ns/pulse, 8 Hz, λ ) 355 nm). After 0, 8, 32, and 480 pulses, 10 µL aliquots were withdrawn and analyzed by electrophoresis on a 15% denaturing polyacrylamide gel. Immobilization, Photorelease, and RT-PCR. Pool RNA (10 pmol) was immobilized by incubation for 30 min with 20 µL of streptavidin-agarose on a microcon spin filter prewashed three times with 100 µL of 3 mg/mL tRNA solution (RNase free, Boehringer Mannheim). Unbound RNA was washed away by subsequent incubation and centrifugation of 200 µL of tRNA solution (3 mg/ mL), 2 × 200 µL of 8 M guanidinium hydrochloride, 0.1 M Tris-HCl, pH 7.5, 10 mM EDTA, and 2 × 200 µL of tRNA solution (3 µg/mL). Bound RNA was released by resuspending the agarose in 200 µL of 3 µg/mL tRNA solution, irradiation on the filter for 3 min with a NdYAG laser (30 mJ/pulse), and washing the agarose three times with 200 µL of tRNA solution (3 µg/mL). A 1% aliquot of the obtained RNA was annealed with 200 pmol of primer and subjected to reverse transcription in a 20 µL reaction mix with 50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl2, 10 mM DTT, 0.5 mM dNTPs (MBI Fermentas), and 200 units of Superscript II (Gibco BRL) for 1 h at 37 °C. The reaction mix was then diluted to a 100 µL volume with the final concentration of 3 µM of each primer, 20 mM Tris-HCl, pH 8.5, 65 mM KCl, 2 mM MgCl2, 2 mM DTT, 0.3 mM dNTPs, and 5 units of Taq polymerase (AGS). After 10 cycles of PCR amplification, the resulting DNA was analyzed on a 2% agarose gel and the bands were visualized by ethidium bromide staining. RESULTS
The most essential requirement for the practical realization of the strategy illustrated in Figure 1b was to develop an efficient method to synthesize multifunctional RNA conjugates from RNA transcripts. While numerous strategies exist for the site-specific incorporation of both linkers and non-nucleotide modifiers during chemical synthesis of oligonucleotides (Beaucage and
Bioconjugate Chem., Vol. 8, No. 6, 1997 887 Scheme 1. Design of Dinucleotide Derivative 1
Scheme 2. Reagents Used in the Synthesis of 1a
Iyer, 1993; Goodchild, 1990), there is currently no method available that would be suitable for direct use in in vitro selection schemes. The RNA is enzymatically synthesized in each round of selection, so modification has to occur either during or after enzymatic synthesis. Ideally, linker, cleavage site, and reactant should be combined in one block by chemical synthesis, which then could be introduced in a single step in each selection cycle. As the general method, we chose to use enzymatic ligation of a suitably derivatized substrate with T4 RNA ligase (England and Uhlenbeck, 1978; Igloi, 1996). Poly(ethylene glycol) was chosen as a synthetic linker because of its high flexibility, solubility, and previously described compatibility with oligonucleotide chemistry (Ja¨schke, 1997; Ja¨schke et al., 1993, 1996). We decided to use a photochemical cleavage reaction (Ordoukhanian and Taylor, 1996) for its high specifity and for its independence of other chemical parameters (e.g., pH, redox potential, solvent). To combine these features in one molecule, we designed compound 1a containing the following elements (Scheme 1): • a 5′-phosphorylated dinucleotide (pCpC) for enzymatic ligation with T4 RNA ligase, • two hexaethylene glycol units as flexible linkers, • an o-nitrobenzyl derivative as a photolytical cleavage site, and • a primary aliphatic amino group for coupling of potential reactants by activated ester chemistry. Compound 1a was synthesized by automated synthesis introducing the photocleavage site by the use of phosphoramidite 2, which was prepared as described (Ordoukhanian and Taylor, 1996). Starting with an aminoalkyl-functionalized glass support 3, hexaethylene glycol phosphoramidite (4), compound 2, hexaethylene glycol phosphoramidite (4), and then cytidine phosphoramidite (5) (twice) were added sequentially, all employing standard protecting groups (Scheme 2). Finally, chemical phosphorylation was carried out on the synthesizer using
888 Bioconjugate Chem., Vol. 8, No. 6, 1997
Figure 2. Time course of the ligation of compound 1a to a 25mer RNA transcript containing 32P-labeled cytidine residues and photocleavage of the resulting RNA conjugate. (a) Aliquots taken from the T4 RNA ligation mix at indicated times were analyzed on a 15% denaturing polyacrylamide gel and visualized by autoradiography. (b) Purified RNA conjugates obtained by ligation were irradiated in solution at 355 nm with a Nd-YAG laser. Aliquots were taken after exposure to the indicated number of laser pulses and examined by electrophoresis on a 15% polyacrylamide gel. Band index: 1, ligation product; 2, primary photoproduct with attached o-nitrosophenylketone; 3, secondary product after β-elimination of vinyl o-nitrosophenyl ketone; 4, unmodified 25-mer transcript.
bis(cyanoethyl) N,N-diisopropyl phosphoramidite (6). After standard deprotection and HPLC purification, compound 1a was characterized by IR-MALDI mass spectrometry, confirming the correct product composition (calculated [MH]+ m/z 1787.2; found [MH]+ m/z 1787.3). To check the desired properties, the dinucleotide analog 1a was first reacted with an N-hydroxysulfosuccinimidylactivated biotin as a model for the attachment of potential reactants to the primary amino group by activated ester chemistry. Ninety-five percent conversion of the dinucleotide analog was observed by HPLC with the biotinylated product eluting at significantly higher retention times (not shown). The identity of the resulting biotin conjugate 1b was corroborated by UV-MALDI mass spectrometry (calculated [M - H]- m/z 2125.7; found [M - H]- m/z 2125.5). Due to UV excitation (λ ) 377 nm), additional peaks of the predicted photofragments were detected (calculated m/z 911.9, 1212.8; found m/z 912.2, 1214.2), indicating a photocleavage reaction proceeding in agreement with the accepted mechanism (Pillai, 1980). The incorporation of the synthetic linker into RNA transcriptssa key step in the modified selection schemeswas studied by ligation with T4 RNA ligase. As rigorous analytics is complicated with randomized pools, first a 32P-labeled RNA 25-mer [sequence: 5′-r(G GAG CUC AGC CUU CAC UGC UCC ACC)-3′] obtained by in vitro T7 transcription was incubated with a 2-fold excess of 1a, and the ligation was followed by electrophoretic analysis of aliquots withdrawn from the reaction mixture at appropriate times. The time course in Figure 2a shows the disappearance of the original 25-mer and the simultaneous formation of the RNA conjugate with up to 95% conversion. No additional bands with lower electrophoretic mobility, indicating circularization or oligomerization products, were detected. The ligation product was eluted, purified by precipitation, redissolved, and irradiated by an increasing number of Nd-YAG laser pulses (λ ) 355 nm). The rapid and quantitative cleavage of the RNA conjugate after 32 pulses (Figure 2b) correlates with the formation of the photoproducts with higher electrophoretic mobility, which can be resolved into two distinct bands by high-resolution electrophoresis (not shown). The upper band can be assigned to the primary photoproduct, an o-nitrosophenyl
Hausch and Ja¨schke
ketone, as a result of the photoinduced intramolecular redox reaction, which partially reacts further by β-elimination to release the photoactive moiety leaving a PEG phosphate attached to the 3′-end of the RNA (Ordoukhanian and Taylor, 1996). This secondary product may be assigned to the lower band in Figure 2b. The two photoproducts were unaffected by further irradiation. A 15-fold overexposure did not lead to any additional bands that would indicate photomodification, crosslinking, or dimerization. To allow the use of the proposed strategy in direct selection, the methods described above have to be compatible with the standard selection scheme without interfering with any other step in the selection cycle. To check the feasibility of the new steps under selection conditions, we performed a selection cycle according to the procedure shown in Figure 1b. Because in the first rounds of a selection project reaction yields are very low and products can often hardly be detected, we substituted the reaction step by artificially creating a situation as it would be after a completely successful reaction. This was achieved by ligating an RNA pool containing two constant sequences for amplification purposes and a randomized domain of 74 nucleotides [sequence: 5′-r(G GAG CUC AGC CUU CAC UGC-N74-GGC ACC ACG GUC GGA UCC AC)-3′, complexity of 1013 different species (Famulok, 1994)] with biotinylated dinucleotide 1b leading to species that correspond to the desired reaction products: RNA pool molecules are connected to biotin as an anchor group via a cleavable linker. Sufficient ligation of the pool required modification of the protocol. A higher excess of the dinucleotide analog and longer incubation times were necessary, and addition of an oligonucleotide complementary to the 3′-end of the pool molecules significantly improved ligation yields, probably by making “buried” 3′-ends accessible to the enzyme. Seventy-five percent yields were obtained in overnight incubation with a 5-fold excess of 1b. Purified ligation product was incubated with streptavidin-agarose on a spin filter for 30 min. Unbound RNA was washed away under denaturing conditions until no radioactivity could be detected in the filtrate. Eighty percent binding was observed (Figure 3a), while unspecific binding of unmodified RNA transcript was determined to be
