- Email: [email protected]
The use of aptamers in large arrays for molecular diagnostics
Molecular Diagnosis Vol. 4 No. 4 1999
The Use of Aptamers in Large Arrays for Molecular Diagnostics E D W A R D N. B R O D Y , M D , PhD, M I C H A E L C. W I L L I S , PhD, J O N A T H A N D. S M I T H , PhD, SUMEDHA JAYASENA, PhD, DOMINIC ZICHI, PhD, LARRY GOLD, PhD Boulder, Colorado
Backgound: Aptamers are single-stranded oligonucleotides derived from an in vitro evolution protocol called systematic evolution of ligands by exponential enrichment (SELEX). They bind tightly and specifically to target molecules; most aptamers to proteins bind with Kds (equilibrium dissociation constant) in the range of 1 pM to 1 nM. Methods and Results: The SELEX protocol has been automated; therefore, hundreds to thousands of aptamers can be made in an economically feasible fashion. Blood and urine can be analyzed on chips that capture and quantitate proteins. SELEX has been adapted to the use of 5-bromo (5-Br) and 5-iodo (5-I) deoxyuridine residues. These halogenated bases can be specifically cross-linked to proteins. Selection pressure during in vitro evolution can be applied for both binding specificity and specific photo-cross-linkability. These are sufficiently independent parameters to allow one reagent, a photo-cross-linkable aptamer, to substitute for two reagents, the capture antibody and the detection antibody, in a typical sandwich array After a cycle of binding, washing, cross-linking, and detergent washing, proteins will be specifically and covalently linked to their cognate aptamers. Conclusions: Because no other proteins are present on the chips, protein-specific stain will now show a meaningful array of pixels on the chip. Learning algorithms and retrospective studies should lead to a robust, simple, diagnostic chip. Key words: systematic evolution of ligands by exponential enrichment (SELEX), photo-cross-linking, universal protein stain, proteomics.
Examination of body fluids has always been an important element of differential diagnosis. From the relatively primitive testing of patients' urine for diabetes mellitus to the relatively sophisticated radioimmune assays for determining protein levels in blood, there has always been one underlying assumption; namely, disease states are manifested in blood, urine, saliva, cerebrospinal fluid (CSF), and so on. We believe the sequencing of the human genome over the next decade will reinforce this idea. Moreover, we propose that a
simple, inexpensive method of analyzing blood levels of proteins coded by the approximately 10r human genes will revolutionize diagnostics.This view is shared by a growing number of medical researchers and pharmaceuticalbiotechnical companies ~ [1 ]. What is needed to critically test this idea is a technique that can rapidly generate large numbers (103 to 10 6) of specific reagents for detecting an equivalent number of proteins in the various body fluids. What is further needed is a robust method ~William Haseltine, chairman and chief executive officer of Human Genome Sciences, lnc (Rockville, MD), is quoted, "The only true equivalent to gene chips is to raise antibodies against all proteins and put these on chips, and do protein profiling the way we do gene profiling. That's ultimately where you want to go, but no one's talking about that yet. A company will pop up to do this one of these days" [1].
From NeXstar Pharmaceuticals; Inc', Boulder, Colorado.
Reprint requests: Edward N. Brody, MD, PhD, NeXstar Pharmaceuticals, Inc, 2860 Wilderness Place, Boulder, CO 80301. Email: [email protected] Copyright © 1999 by Churchill Livingstone ®
1084-8592/99/0404-0014510.00/0
381
382
Molecular Diagnosis Vol. 4 No. 4 December 1999
Fig. 1. A library of approximately 1015 single-stranded nucleic acid molecules are allowed to interact with a desired target. Those molecules that interact are separated from those that do not; this can be done by gel filtration, affinity chromatography, immunoprecipitation, nitrocellulose filter binding, or other appropriate methods. If the library is RNAbased, the selected RNA population is reverse transcribed to DNA by an appropriate reverse transcriptase (RT), and this population is then amplified by polymerase chain reaction (PCR). If the library is DNA-based, it is amplified directly by PCR. Then, transcription by T7 RNA polymerase for RNA libraries or strand separation for DNA libraries generates the single-stranded species that go into the next round of SELEX. dsDNA, doublestranded DNA.
The SELEX Process
~
Single-stranded stable *oligonucleotide library ~- Target
Transcn t,on ~ ~ Free oligos °~P" I Oligo-Target Complex StrandS e p a r a t i o ~
one
SELEX round
~- AmplRdl if:edIPNC?
for displaying such data so that correlations between patterns and disease states can be made. The ease and cost of these techniques must be compatible with largescale production and use. We present just such a series of techniques.
Results and Discussion Systematic evolution of ligands by exponential enrichment ( S E L E X ) is a technique allowing repetitive cycles of selection and amplification of single-stranded nucleic acids in vitro [2-4]. The starting material is synthetic D N A containing fixed sequences for primer annealing in a polymerase chain reaction and, if the single-stranded nucleic acid is an R N A analogue, for allowing initiation of R N A synthesis by T7 R N A polymerase. In between these fixed sequences are regions of r a n d o m sequences, usually 30 or 40 nucleotides long. R N A synthesis or strand separation gives 4 `30( - 1 0 ~s) or 440 (~1024) different single-stranded sequences. A convenient protocol starts with approximately 1015 distinct molecules. The winnowing process (Fig. 1) results in the selection of nucleic acids that correspond to the optimal three-dimensional shape for the chosen parameter. Ordinarily, the parameter is tight binding. When the target is a protein, this binding can be extraordinarily tight (Kas usually between 10 pM and i nM) and very specific (Fig. 2; Tables 1 and 2). A plentiful literature (reviewed in
Clone, sequence, screen (aptamers)
[4,5]) has shown the S E L E X process to be extremely robust.The winning nucleic acids (called aptamers) rival antibodies for specificity and can surpass them for binding to proteins. A key advantage of aptamers over antibodies is that they are synthesized in vitro and are thus readily amenable to controlled, specific, chemical modification. A wide variety of chemical functional groups can be incorporated into S E L E X libraries [6]. These modifications can modify the physical chemistry of the library or confer novel binding, signaling, or catalytic functions [7-10]. Chemical modification of nucleotides both for substrates of polymerases or for chemical synthesis gives aptamers an important advantage over antibodies. Proteins that are poorly immunogenic are frequently encountered (for example, self-proteins) in attempts to elicit antibodies. In contrast, chemical modification of nucleotides should allow evolution of an aptamer to any protein. Furthermore, chemical synthesis of winning aptamers allows attachment of functional groups to be directed toward any desired residue, facilitating both attachment of aptamers to any of a variety of biochips and optimizing for the placement of such reporter groups as fluorescent adducts directly on the aptamers [6,11]. S E L E X can evolve reagents that bind tightly and specifically to proteins in body fluids. Can S E L E X compare with antibodies for the production of diagnostic reagents, particularly in the 5 to 10 years in which the entire human genome will be sequenced? Expression of
AptamerArrays in Diagnostics
•
Brody et al.
383
Affinities [KdS] for Protein Targets First 100 Aptamers Fig. 2. The equilibrium constants (Kds) for the first 100 aptamers to proteins performed either at NeXstar Pharmaceuticals or the University of Colorado are presented here. More than three fourths of the total have KdS of 1 nM or less. Most antibodies and many other pharmaceuticals bind their targets with KdS of 1 to 10 nM.
~1'
c o m p l e m e n t a r y D N A clones from the approximately 105 h u m a n genes will require at least 105 specific reagents for diagnostic detection. Is any technique available for processing 104 such reagents per year for the next 10 years? Although this is probably outside the capacity for antibody production, it is feasible for SELEX. The simple enzymatic nature of the amplification steps in S E L E X has permitted robots to perform rounds of S E L E X [12] (R.D. Jenison, D. Smith, T. Fitzwater, D.
Table 1. Binding Affinity and Specificity of NX1838 Protein Human VEGF~65 Mouse VEGF164 Human VEGFI65 (reduced) Human VEGF~2~ Human PIGFt29 Human VEGF16flPIGF~2,) PDGF-AB TGF-[31 bFGF
Kd (pM) 49 _+ 6 90 _+ 22 >1,000,000 >1,000,000 > 1,000,000 432 +_ 154 800,000 >1,000,000 220,000
The Kd of aptamer NX1838 that was evolved to bind to human vascular endothelial growth factor 165 (VEGF165). The aptamer binds to this 165-amino acid form of the protein with a Kd of 49 pM. It binds almost as strongly to the homologous mouse protein. When the homodimer VEGF is reduced to monomers with sulfhydryl reagents, binding is destroyed. Binding is to the heparin-binding domain of VEGE The VEGFI21 isoform of this protein, which does not contain this heparin-binding site, does not bind the aptamer. A related heparin-binding protein, PIGF129, does not bind the aptamer, nor do three other heparin-binding proteins listed at the bottom of the table. A VEGF~6flPIGF~29 heterodimer binds NX1838 about a tenth as well as does the VEGF165 homodimer. PIGE placenta growth factor; PDGF-AB, platelet-derived growth factor (AB heterodimer);TGF-[31,transforminggrowth factor-betav Data from Ruckman et al. [24].
1 antibodies most or other drugs
Zichi, manuscript in preparation). A new mathematical approach to S E L E X optimization leads to a standard protocol [13]; using this protocol, tight-binding aptamer families have b e e n evolved in 5 to 7 rounds or less. The robot is based on a 96-well plate format and should be able to perform 2.5 rounds per day. Thus, just 1 robot has the potential to evolve 96 × 2.5 × 7 = 1,680 rounds per week; this is 87,360 rounds per year, or 12,480 to 17,472 high-affinity pools per year. Therefore, aptamers theoretically can be produced for diagnostics at a pace that will match protein production coming from the sequencing of the h u m a n genome. Can one aptamer per protein lead to a viable, robust diagnostic format? To know this, we must first review how proteins are detected in body fluids. We start by looking at what might be considered a model for a diagnostic chip. Affinity chromatography uses a specific cap-
Table 2. Aptamers Bind Selectively to the PDGF B-chain Kd (nM)
Ligand PDGF-AA 20t 36t 41t
47 _+ 4 72 _+ 12 49 _+ 8
PDGF-AB
PDGF-BB
0.147 _+ 0.011 0.094 +_ 0.011 0.138 + 0.009
0.127 _+ 0.031 0.093 + 0.009 0.129 _+ 0.011
Three different DNA sequences from a winning aptamer pool against platelet-derived growth factor (PDGF), an AB heterodimer, have been tested for specificity.Data show that all 3 bind to the AB heterodimer and BB homodimer forms of PDGF, but not to the AA form. Binding is therefore B-chain specific. Data from Green et al. [25].
384
Molecular Diagnosis Vol. 4 No. 4 December 1999
ture reagent on a surface. The large excess of capture reagent (usually an antibody or aptamer) binds a particular target and purifies it away from a complex mixture by a combination of affinity and avidity [14,15]. Whether the capture reagent is an aptamer or antibody, the purification of a protein from blood achieved by one passage through this procedure is on the order of 65 % to 85 %. Although this is a remarkable purification (1,000- to 5,000-fold from a complex mixture [14]), it is not good enough for a diagnostic assay. Backgrounds must be much less than this level of purification affords to give accurate measurements of individual proteins in blood (or urine, CSF, and so on). This is why a second, independent reagent is often used in diagnostic tests; these are usually antibodies to a second epitope on the protein (although two aptamers to different epitopes of thrombin have also been used in sandwich assays [Y. Lin, J. Heil, S. Jayasena, manuscript in preparation[). The second reagent gives discrimination that is multiplicative to the first discrimination step. For antibody-based sandwich assays to keep pace with new proteins coming from the Human Genome Project, not only will thousands of antibodies be needed over the next 10 years, but two per protein (each reacting at different epitopes) will be required for multiplexed diagnostic assays on a chip. How do we circumvent the same necessity for two reagents in an aptamer-based assay? Single-stranded nucleic acids can be synthesized using 5-Br deoxyuridine triphosphate (dUTP) or 5-I dUTP as substitutes for deoxythymidine triphosphate in the DNA polymerase reaction. Similarly, 5-Br, 2'F uridine triphosphate (UTP) or 5-I 2'F UTP can be used as replacements for UTP as substrates for T7 RNA polymerase [7]. The singlestranded nucleic acids containing halogenated residues at the 5 position of pyrimidines can be incorporated into a modification of SELEX that is called photo-SELEX [7]. Photo-SELEX incorporates brief laser irradiation, either at 308 nM (for Br) or 325 nM (for I), that crosslinks the single-stranded nucleic acid to an electron-rich amino acid in close proximity on the protein target [7,16-18]. Early in the development of photo-SELEX, it was shown that such covalently linked protein-nucleic acid complexes could be sufficiently proteolysed to allow either Taq DNA polymerase or reverse transcriptase to make a faithful DNA complement to the selected nucleic acid [7]. Thus, photo-SELEX has developed into a protocol for evolving protein-nucleic acid interactions that, on irradiation, give a regiospecific, conformationdependent, cross-link (Fig. 3, see color plate). The specificity of the cross-linking residues is quite remarkable. The 5-bromodeoxyuridine (5-BrdU)-containing aptamer evolved to bind to basic fibroblast growth factor (bFGF) shows a cross-link between only one of six 5-BrdU residues in the optimal aptamer and methionine 143 of the protein (this despite the other potential cross-linking residues in bFGF). Examples of
other equally specific cross-linking have been shown. Because of the specificity of cross-linking that results from the selection pressure maintained during the photo-SELEX protocol, cross-linking is so restrained that it can constitute a second dimension in a diagnostic assay. Thus, proteins in a mixture would first sort out on their cognate aptamers on a chip according to the affinity-avidity distribution analogous to affinity chromatography. There will be several to many noncognate proteins in the biological fluid bound at this stage (up to one third of the total bound, following the affinity chromatography analogy). The aptamer-containing chip is washed to reduce this nonspecific background. To what extent this wash decreases the specific cognate signal depends on the off-rate component of the equilibrium Kd for the specific protein-aptamer interaction. Fortunately, the SELEX process can be optimized for slow off-rate components of the overall Kd. After this short washing step, photo-cross-linking irradiation covalently links aptamers to their cognate proteins. Because of the specificity of this step, any remaining noncognate proteins still bound to aptamers will have a reduced probability of having the correct amino acid in just the right proximity to a BrdU or iododeoxyuridine (IdU) residue to generate a cross-link. After cross-linking, the chip is washed with detergents or other denaturing reagents to remove bound, but not cross-linked, proteins at each pixel and give a specific protein signal for every aptamer on the chip. Before discussing the detection of proteins bound to the chips, let us briefly review evidence supporting the aforementioned scheme. First, the specificity of crosslinking has been shown in some BrdU-containing aptamers. A specific b F G F aptamer binds to bFGF with a Ka of 20 pM. Photo-cross-linking has been performed both with pure protein and protein in the presence of 10% serum (M.C. Golden, B.D. Collins, M.C. Willis, T.C. Koch, manuscript in preparation). In these assays, for which no intermediary washing step has been introduced before photo-cross-linking, there is no other significant cross-linked band visible in the serum sample, and b F G F can be detected down to a concentration of 3.2 pM. More importantly, no competition of material in the specific aptamer b F G F band is seen in the serum sample, suggesting that no blood protein competes for aptamer photo-cross-linking to bFGF. It seems reasonable to expect that with a washing step in between binding and cross-linking, similar specificity would be seen in more concentrated serum (the standard assay protocol). By modeling the combined binding plus photo crosslinking steps as an overall kinetic pathway in which an equilibrium-binding reaction is followed by an irreversible photo-cross-linking step, an overall discrimination rate can be calculated. The simplest model of this process is analogous to a scheme in which an enzyme is challenged with two competing substrates:
Aptamer Arrays in Diagnostics
*
Brody et al.
385
Photo-SELEX Procedure I
I
IIIII
I
fixed A
I
random sequence
gg~ggacgatgcgg ~ 3' ccc~ccbgcbacggg
DNA
I
fixed B
]
I cagaggaegagcggg~.
5'
g t c t c c t g c t c g c c c ~ Biotin
polymemsewith5-BrdUI
Isolate Sense Stand 5 'ggg~ggacgatgcgg
cagaggacgagcggga
3'
ORe
Photo-SELEX Round
Clone Sequence Fig. 3. The modifications to the SELEX procedure needed for photo-SELEX. Single-stranded DNA libraries containing 5-BrdU are allowed to interact with a target protein. Photo-cross-linking is performed as described in text, then SDS-polyacrylamide gel electrophoresis (PAGE) dissociates noncovalent protein-nucleic acid interactions. Unbound single-stranded DNA migrates at its habitual rate in these gels. Covalent protein-DNA complexes migrate more slowly than DNA on these gels. Such slower migrating bands are excised, protein is digested with proteinase K, and the single-stranded DNA is amplified and recycled through rounds of photo-SELEX.
1.hv 1. Biologicalfluid 2. Mild wash
2. Harsh wash
__q I Proteln-speclfic
T Fig. 6. An array of aptamers on a chip is schematized. Each pixel is represented by the teal-colored surface. Aptamers are represented in red and proteins in green. The steps of binding, mild washing, cross-linking, harsh washing, and protein staining are described in text.
386
Molecular Diagnosis Vol. 4 No. 4 December 1999
[A:T]
=
+T/K/KD
A-T
kxT
T
l
Aptamer
+B
hv
"% [A:B]
kxB =
A-B
Where T and B are target and background proteins, respectively, KD is the equilibrium binding constant between aptamer and protein, and kx is the elementary rate constant for irreversible photo-cross-linking. In this scheme, the overall specificity is given by: Fig. 4. Relative reactivity of D N A and proteins with a fluoroA - T
[T](kx/KM) r
A- B
[B] ( k x / K M ) B
where KM is the usual Michaelis constant [19,20]. The implication of this scheme is that specificity in the crosslinking step (kx) will multiplicatively increase the specificity of the equilibrium-binding step (KD). One great advantage of an aptamer-based diagnostic chip is the ease with which proteins covalently linked to aptamers can be detected. Unlike antibody-based reaction schemes, the aptamer-based chip will have no proteins bound to the chip other than the desired analytes. Thus, any protein-specific staining reagent will give an immediate snapshot of a meaningful distribution of blood proteins. Such chemistries have been worked on for more than 70 years, and a huge literature compiled. Specific reagents for tyrosine, cysteine, arginine, and other amino acids are known [21]. Perhaps the most specific reagents are those that react with lysine. Because of its strong nucleophilic character, the primary amine of lysine constitutes a unique target for reactivity. The N7 position of guanine, the strongest nucleophilic group in an aptamer, either DNA- or RNA-based, is much weaker than the primary amine of lysine at p H 7.5. In a series of experiments shown in Figs. 5 and 6, we show that a vicinal dicarbonyl constitutes a universal protein stain. One lysine-reactive reagent will react with all lysines on the chip. Because lysine constitutes 5.7% of human proteins (GenBank Release 108, Aug 15, 1998, Bethesda, MD), on average, such a reagent should be sufficient when combined with a fluorescent reporter adduct to illuminate an entire chip. In the event that some rare proteins have no available lysines, other amino acidspecific reactants could provide additional universal protein stains.
genic dicarbonyl reagent. Human serum albumin (hSa; 1 pmol) and a 42-mer DNA were mixed in the molar ratios indicated at the tops of the gel lanes and reacted with the fluorogenic dicarbonyl reagent CBQCA [23] according to the protocol supplied by the vendor, Molecular Probes (Eugene, OR). The image was obtained by scanning the gel with a Molecular Systems FluorImager (Sunnyvale, CA), M, dye marker lane; N, no DNA or protein; D, a fluorescein-labeled DNA marker.
An aptamer-based chip can thus be stained with one, or at most two, universal protein stains. In such a simple system, fluorescence can be analyzed in arrays on detectors already in use (Fig. 7, see color plate). One automated reader could analyze, in real time, approximately 3,000 to 10,000 chips/d, allowing quantitative data to be garnered from very large numbers of blood samples, probably at first in retrospective studies, to correlate blood protein levels and disease states. With a wide choice of learning algorithms to process these data, it is quite possible to perform ever finer analyses as new markers are added. Because the goal is to simultaneously quantitate thousands of proteins, an algorithm such as the perceptron [22] should be able to transform patterns into yes-no answers. With enough data points, the probability of making a mistake on such a binary decision (Do I have cancer? Has it metastasized? Is it in the bones?) will be vanishingly small. Finally, one imagines that this algorithm will be applicable to each individual in a prospective manner. We anticipate that a drop of blood put onto such a chip once or twice a year will give an immediate readout of a person's health status. Early indicators of disease states could in all probability signal early treatment and eliminate the need for expensive ancillary tests. It is not too soon to imagine that this application of the sequencing of the human genome, the quantification of all important proteins in
Aptamer Arrays in Diagnostics
Fig. 5. Relative reactivity of lysine and nucleotides with a fluorogenic dicarbonyl reagent. N-acetyl lysine (10 IxM) or nucleotide monophosphates (20 mM) were reacted with CBQCA [23] according to a protocol supplied by the vendor, Molecular Probes (Eugene, OR). Fluorescence was detected in a CytoFluor II fluorescence microtiter plate reader (PerSeptive Biosystems,Framingham, MA), Data are the mean of four measurements. NAcLys; AMR adenosine monophosphate; GMR guanosine monophosphate.
2000
I
I
I
I
I
? o o e o
-
Brody et al.
387
I
~ 1500 C "~ ~ 1000
•
o
NAcLys
•
AMP
~'
GNP
°
buffer
2
Z o
e o
I U.
o
500
-.
o
0
0
20
blood (or urine or CSF), will be the most important element of a physician's armamentarium for differential diagnosis in the 21st century.
Acknowledgment The authors thank their colleagues at NeXstar and at the University of Colorado in Boulder for their contributions to the many elements in this work. Many experiments were performed by people who have allowed us to quote from their work. In particular, the authors thank Nebojsa Janjic, Judy Ruckman, Louis Green, Dan Drolet, Mace Golden, Tad Koch, Yun Lin, Jim Heil, Brian Collins, Michael Lochrie, Stephen Capaldi, Philippe Bridonneau, and Gary Cook for their contributions, and Kris Cordova for collating a large amount of material and preparing the manuscript.
Received May 20, 1999. Received in revised form July 19, 1999. Accepted July 19, 1999.
References 1. Zipkin I: Proteomics. BioCentury 1998;6:1-4 2. Tuerk C, Gold L: Systematic evolution of ligands by exponential enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science 1990;249: 505-510 3. Ellington AD, Szostak JW: In vitro selection of RNA molecules that bind specific ligands. Nature 1990;346:818-822 4. Gold L, Polisky B, Uhlenbeck O, Yarus M: Diversity of oligonucleotide functions. Annu Rev Biochem 1995;64:763-797
40
60
80 min
100
120
140
5. Osborne SE, Ellington AD: Nucleic acid selection and the challenge of combinatorial chemistry. Chem Rev 1997;97:349-370 6. Eaton BE: The joys of in vitro selection: Chemically dressing oligonucleotides to satiate protein targets. Curr Opin Chem Biol 1997;1:10-16 7. Jensen KB, Atkinson BL, Willis MC, Koch TH, Gold L: Using in vitro selection to direct the covalent attachment of human immunodeficiency virus type 1 Rev protein to high-affinity RNA ligands. Proc Natl Acad Sci U S A 1995;92:12220-12224 8. Smith D, Kirschenheuter GR Charlton J, Guidot DM, Repine JE: In vitro selection of RNA-based irreversible inhibitors of human neutrophil elastase. Curr Biol 1995;2:741-750 9. Wiegand TW, Janssen RC, Eaton BE: Selection of RNA amide synthases. Chem Biol 1997;4:675-683 10. Wecker M, Smith D, Gold L: In vitro selection of a novel catalytic RNA: Characterization of a sulfur alkylation reaction and interaction with a small peptide. RNA 1996;2:982-994 11. Eaton BE, Pieken WA: Ribonucleosides and RNA. Annu Rev Biochem 1995;64:837-863 12. Cox JC, Rudolph R Ellington AD: Automated RNA selection. Biotechnol Prog 1998;14:845-850 13. Vant-Hull B, Gold L, Zichi D: Basic principles of in vitro selection using combinatorial nucleic acid libraries. Curr Prot Nucleic Acid Chem 1999 14. Romig T, Bell C, Drolet D: Aptamer affinity chromatography; Combinatorial chemistry applied to protein purification. J Chromatogr 1999;731:275284 15; Harlow E, Lane D: Antibodies: A laboratory manual. ColdSpring Harbor Laboratory, Cold Spring Harbor, NY, 1988 16. Hicke B J, Willis MC, Koch TH, Cech TR: Telomeric protein-DNA point contacts identified by photocross-linking using 5-bromodeoxyuridine [published
388
17,
18.
19. 20. 21. 22.
Molecular Diagnosis Vol. 4 No. 4 December 1999
erratum, Biochemistry 1994;33:7744]. Biochemistry 1994;33:3364-3373 WillisMC, LeCuyer KA, Meisenheimer KM, Uhlenbeck OC, Koch TH: An RNA-protein contact determined by 5-bromouridine substitution, photo-crosslinking and sequencing. Nucleic Acids Res 1994;22: 4947-4952 Meisenheimer KM, Meisenheimel' PL, Willis MC, Koch TH: High-yield photocrosslinking of a 5iodocytidine (IC) substituted RNA to its associated protein. Nucleic Acids Res 1996;24:981-982 Fersht A: Enzyme structure and mechanism. New York, NY, W.H. Freeman, 1985 Yarus M: The accuracy of translation. Prog Nucleic Acid Res Mol Biol 1979;23:195-225 Wong S: Chemistry of protein conjugation and crosslinking. Boca Raton, FL, CRC Press, 1991 Schneider TD, Stormo GD, Gold L, Ehrenfeucht A:
Information content of binding sites on nucleotide sequences. J Mol Biol 1986;188:415-431 23. Liu JR Hsieh YZ, Wiesler D, Novotny M: Design of 3-(4-carboxybenzoyl)-2- quinolinecarboxaldehyde as a reagent for ultrasensitive determination of primary amines by capillary electrophoresis using laser fluorescence detection. Anal Chem 1991;63:408-412 24. Ruckman J, Green LS, Beeson J, et al.: 2'-Fluoropyrimidine RNA-based aptamers to the 165-amino acid form of vascular endothelial growth factor (VEGF165). Inhibition of receptor binding and VEGF-induced vascular permeability through interactions requiring the exon 7-encoded domain. J Biol Chem 1998;273:20556-205567 25. Green LS, Jellinek D, Jenison R, Ostman A, Heldin CH, Janjic N: Inhibitory DNA ligands to plateletderived growth factor B-chain. Biochemistry 1996; 35:14413-14424
The Use of Aptamers in Large Arrays for Molecular Diagnostics E D W A R D N. B R O D Y , M D , PhD, M I C H A E L C. W I L L I S , PhD, J O N A T H A N D. S M I T H , PhD, SUMEDHA JAYASENA, PhD, DOMINIC ZICHI, PhD, LARRY GOLD, PhD Boulder, Colorado
Backgound: Aptamers are single-stranded oligonucleotides derived from an in vitro evolution protocol called systematic evolution of ligands by exponential enrichment (SELEX). They bind tightly and specifically to target molecules; most aptamers to proteins bind with Kds (equilibrium dissociation constant) in the range of 1 pM to 1 nM. Methods and Results: The SELEX protocol has been automated; therefore, hundreds to thousands of aptamers can be made in an economically feasible fashion. Blood and urine can be analyzed on chips that capture and quantitate proteins. SELEX has been adapted to the use of 5-bromo (5-Br) and 5-iodo (5-I) deoxyuridine residues. These halogenated bases can be specifically cross-linked to proteins. Selection pressure during in vitro evolution can be applied for both binding specificity and specific photo-cross-linkability. These are sufficiently independent parameters to allow one reagent, a photo-cross-linkable aptamer, to substitute for two reagents, the capture antibody and the detection antibody, in a typical sandwich array After a cycle of binding, washing, cross-linking, and detergent washing, proteins will be specifically and covalently linked to their cognate aptamers. Conclusions: Because no other proteins are present on the chips, protein-specific stain will now show a meaningful array of pixels on the chip. Learning algorithms and retrospective studies should lead to a robust, simple, diagnostic chip. Key words: systematic evolution of ligands by exponential enrichment (SELEX), photo-cross-linking, universal protein stain, proteomics.
Examination of body fluids has always been an important element of differential diagnosis. From the relatively primitive testing of patients' urine for diabetes mellitus to the relatively sophisticated radioimmune assays for determining protein levels in blood, there has always been one underlying assumption; namely, disease states are manifested in blood, urine, saliva, cerebrospinal fluid (CSF), and so on. We believe the sequencing of the human genome over the next decade will reinforce this idea. Moreover, we propose that a
simple, inexpensive method of analyzing blood levels of proteins coded by the approximately 10r human genes will revolutionize diagnostics.This view is shared by a growing number of medical researchers and pharmaceuticalbiotechnical companies ~ [1 ]. What is needed to critically test this idea is a technique that can rapidly generate large numbers (103 to 10 6) of specific reagents for detecting an equivalent number of proteins in the various body fluids. What is further needed is a robust method ~William Haseltine, chairman and chief executive officer of Human Genome Sciences, lnc (Rockville, MD), is quoted, "The only true equivalent to gene chips is to raise antibodies against all proteins and put these on chips, and do protein profiling the way we do gene profiling. That's ultimately where you want to go, but no one's talking about that yet. A company will pop up to do this one of these days" [1].
From NeXstar Pharmaceuticals; Inc', Boulder, Colorado.
Reprint requests: Edward N. Brody, MD, PhD, NeXstar Pharmaceuticals, Inc, 2860 Wilderness Place, Boulder, CO 80301. Email: [email protected] Copyright © 1999 by Churchill Livingstone ®
1084-8592/99/0404-0014510.00/0
381
382
Molecular Diagnosis Vol. 4 No. 4 December 1999
Fig. 1. A library of approximately 1015 single-stranded nucleic acid molecules are allowed to interact with a desired target. Those molecules that interact are separated from those that do not; this can be done by gel filtration, affinity chromatography, immunoprecipitation, nitrocellulose filter binding, or other appropriate methods. If the library is RNAbased, the selected RNA population is reverse transcribed to DNA by an appropriate reverse transcriptase (RT), and this population is then amplified by polymerase chain reaction (PCR). If the library is DNA-based, it is amplified directly by PCR. Then, transcription by T7 RNA polymerase for RNA libraries or strand separation for DNA libraries generates the single-stranded species that go into the next round of SELEX. dsDNA, doublestranded DNA.
The SELEX Process
~
Single-stranded stable *oligonucleotide library ~- Target
Transcn t,on ~ ~ Free oligos °~P" I Oligo-Target Complex StrandS e p a r a t i o ~
one
SELEX round
~- AmplRdl if:edIPNC?
for displaying such data so that correlations between patterns and disease states can be made. The ease and cost of these techniques must be compatible with largescale production and use. We present just such a series of techniques.
Results and Discussion Systematic evolution of ligands by exponential enrichment ( S E L E X ) is a technique allowing repetitive cycles of selection and amplification of single-stranded nucleic acids in vitro [2-4]. The starting material is synthetic D N A containing fixed sequences for primer annealing in a polymerase chain reaction and, if the single-stranded nucleic acid is an R N A analogue, for allowing initiation of R N A synthesis by T7 R N A polymerase. In between these fixed sequences are regions of r a n d o m sequences, usually 30 or 40 nucleotides long. R N A synthesis or strand separation gives 4 `30( - 1 0 ~s) or 440 (~1024) different single-stranded sequences. A convenient protocol starts with approximately 1015 distinct molecules. The winnowing process (Fig. 1) results in the selection of nucleic acids that correspond to the optimal three-dimensional shape for the chosen parameter. Ordinarily, the parameter is tight binding. When the target is a protein, this binding can be extraordinarily tight (Kas usually between 10 pM and i nM) and very specific (Fig. 2; Tables 1 and 2). A plentiful literature (reviewed in
Clone, sequence, screen (aptamers)
[4,5]) has shown the S E L E X process to be extremely robust.The winning nucleic acids (called aptamers) rival antibodies for specificity and can surpass them for binding to proteins. A key advantage of aptamers over antibodies is that they are synthesized in vitro and are thus readily amenable to controlled, specific, chemical modification. A wide variety of chemical functional groups can be incorporated into S E L E X libraries [6]. These modifications can modify the physical chemistry of the library or confer novel binding, signaling, or catalytic functions [7-10]. Chemical modification of nucleotides both for substrates of polymerases or for chemical synthesis gives aptamers an important advantage over antibodies. Proteins that are poorly immunogenic are frequently encountered (for example, self-proteins) in attempts to elicit antibodies. In contrast, chemical modification of nucleotides should allow evolution of an aptamer to any protein. Furthermore, chemical synthesis of winning aptamers allows attachment of functional groups to be directed toward any desired residue, facilitating both attachment of aptamers to any of a variety of biochips and optimizing for the placement of such reporter groups as fluorescent adducts directly on the aptamers [6,11]. S E L E X can evolve reagents that bind tightly and specifically to proteins in body fluids. Can S E L E X compare with antibodies for the production of diagnostic reagents, particularly in the 5 to 10 years in which the entire human genome will be sequenced? Expression of
AptamerArrays in Diagnostics
•
Brody et al.
383
Affinities [KdS] for Protein Targets First 100 Aptamers Fig. 2. The equilibrium constants (Kds) for the first 100 aptamers to proteins performed either at NeXstar Pharmaceuticals or the University of Colorado are presented here. More than three fourths of the total have KdS of 1 nM or less. Most antibodies and many other pharmaceuticals bind their targets with KdS of 1 to 10 nM.
~1'
c o m p l e m e n t a r y D N A clones from the approximately 105 h u m a n genes will require at least 105 specific reagents for diagnostic detection. Is any technique available for processing 104 such reagents per year for the next 10 years? Although this is probably outside the capacity for antibody production, it is feasible for SELEX. The simple enzymatic nature of the amplification steps in S E L E X has permitted robots to perform rounds of S E L E X [12] (R.D. Jenison, D. Smith, T. Fitzwater, D.
Table 1. Binding Affinity and Specificity of NX1838 Protein Human VEGF~65 Mouse VEGF164 Human VEGFI65 (reduced) Human VEGF~2~ Human PIGFt29 Human VEGF16flPIGF~2,) PDGF-AB TGF-[31 bFGF
Kd (pM) 49 _+ 6 90 _+ 22 >1,000,000 >1,000,000 > 1,000,000 432 +_ 154 800,000 >1,000,000 220,000
The Kd of aptamer NX1838 that was evolved to bind to human vascular endothelial growth factor 165 (VEGF165). The aptamer binds to this 165-amino acid form of the protein with a Kd of 49 pM. It binds almost as strongly to the homologous mouse protein. When the homodimer VEGF is reduced to monomers with sulfhydryl reagents, binding is destroyed. Binding is to the heparin-binding domain of VEGE The VEGFI21 isoform of this protein, which does not contain this heparin-binding site, does not bind the aptamer. A related heparin-binding protein, PIGF129, does not bind the aptamer, nor do three other heparin-binding proteins listed at the bottom of the table. A VEGF~6flPIGF~29 heterodimer binds NX1838 about a tenth as well as does the VEGF165 homodimer. PIGE placenta growth factor; PDGF-AB, platelet-derived growth factor (AB heterodimer);TGF-[31,transforminggrowth factor-betav Data from Ruckman et al. [24].
1 antibodies most or other drugs
Zichi, manuscript in preparation). A new mathematical approach to S E L E X optimization leads to a standard protocol [13]; using this protocol, tight-binding aptamer families have b e e n evolved in 5 to 7 rounds or less. The robot is based on a 96-well plate format and should be able to perform 2.5 rounds per day. Thus, just 1 robot has the potential to evolve 96 × 2.5 × 7 = 1,680 rounds per week; this is 87,360 rounds per year, or 12,480 to 17,472 high-affinity pools per year. Therefore, aptamers theoretically can be produced for diagnostics at a pace that will match protein production coming from the sequencing of the h u m a n genome. Can one aptamer per protein lead to a viable, robust diagnostic format? To know this, we must first review how proteins are detected in body fluids. We start by looking at what might be considered a model for a diagnostic chip. Affinity chromatography uses a specific cap-
Table 2. Aptamers Bind Selectively to the PDGF B-chain Kd (nM)
Ligand PDGF-AA 20t 36t 41t
47 _+ 4 72 _+ 12 49 _+ 8
PDGF-AB
PDGF-BB
0.147 _+ 0.011 0.094 +_ 0.011 0.138 + 0.009
0.127 _+ 0.031 0.093 + 0.009 0.129 _+ 0.011
Three different DNA sequences from a winning aptamer pool against platelet-derived growth factor (PDGF), an AB heterodimer, have been tested for specificity.Data show that all 3 bind to the AB heterodimer and BB homodimer forms of PDGF, but not to the AA form. Binding is therefore B-chain specific. Data from Green et al. [25].
384
Molecular Diagnosis Vol. 4 No. 4 December 1999
ture reagent on a surface. The large excess of capture reagent (usually an antibody or aptamer) binds a particular target and purifies it away from a complex mixture by a combination of affinity and avidity [14,15]. Whether the capture reagent is an aptamer or antibody, the purification of a protein from blood achieved by one passage through this procedure is on the order of 65 % to 85 %. Although this is a remarkable purification (1,000- to 5,000-fold from a complex mixture [14]), it is not good enough for a diagnostic assay. Backgrounds must be much less than this level of purification affords to give accurate measurements of individual proteins in blood (or urine, CSF, and so on). This is why a second, independent reagent is often used in diagnostic tests; these are usually antibodies to a second epitope on the protein (although two aptamers to different epitopes of thrombin have also been used in sandwich assays [Y. Lin, J. Heil, S. Jayasena, manuscript in preparation[). The second reagent gives discrimination that is multiplicative to the first discrimination step. For antibody-based sandwich assays to keep pace with new proteins coming from the Human Genome Project, not only will thousands of antibodies be needed over the next 10 years, but two per protein (each reacting at different epitopes) will be required for multiplexed diagnostic assays on a chip. How do we circumvent the same necessity for two reagents in an aptamer-based assay? Single-stranded nucleic acids can be synthesized using 5-Br deoxyuridine triphosphate (dUTP) or 5-I dUTP as substitutes for deoxythymidine triphosphate in the DNA polymerase reaction. Similarly, 5-Br, 2'F uridine triphosphate (UTP) or 5-I 2'F UTP can be used as replacements for UTP as substrates for T7 RNA polymerase [7]. The singlestranded nucleic acids containing halogenated residues at the 5 position of pyrimidines can be incorporated into a modification of SELEX that is called photo-SELEX [7]. Photo-SELEX incorporates brief laser irradiation, either at 308 nM (for Br) or 325 nM (for I), that crosslinks the single-stranded nucleic acid to an electron-rich amino acid in close proximity on the protein target [7,16-18]. Early in the development of photo-SELEX, it was shown that such covalently linked protein-nucleic acid complexes could be sufficiently proteolysed to allow either Taq DNA polymerase or reverse transcriptase to make a faithful DNA complement to the selected nucleic acid [7]. Thus, photo-SELEX has developed into a protocol for evolving protein-nucleic acid interactions that, on irradiation, give a regiospecific, conformationdependent, cross-link (Fig. 3, see color plate). The specificity of the cross-linking residues is quite remarkable. The 5-bromodeoxyuridine (5-BrdU)-containing aptamer evolved to bind to basic fibroblast growth factor (bFGF) shows a cross-link between only one of six 5-BrdU residues in the optimal aptamer and methionine 143 of the protein (this despite the other potential cross-linking residues in bFGF). Examples of
other equally specific cross-linking have been shown. Because of the specificity of cross-linking that results from the selection pressure maintained during the photo-SELEX protocol, cross-linking is so restrained that it can constitute a second dimension in a diagnostic assay. Thus, proteins in a mixture would first sort out on their cognate aptamers on a chip according to the affinity-avidity distribution analogous to affinity chromatography. There will be several to many noncognate proteins in the biological fluid bound at this stage (up to one third of the total bound, following the affinity chromatography analogy). The aptamer-containing chip is washed to reduce this nonspecific background. To what extent this wash decreases the specific cognate signal depends on the off-rate component of the equilibrium Kd for the specific protein-aptamer interaction. Fortunately, the SELEX process can be optimized for slow off-rate components of the overall Kd. After this short washing step, photo-cross-linking irradiation covalently links aptamers to their cognate proteins. Because of the specificity of this step, any remaining noncognate proteins still bound to aptamers will have a reduced probability of having the correct amino acid in just the right proximity to a BrdU or iododeoxyuridine (IdU) residue to generate a cross-link. After cross-linking, the chip is washed with detergents or other denaturing reagents to remove bound, but not cross-linked, proteins at each pixel and give a specific protein signal for every aptamer on the chip. Before discussing the detection of proteins bound to the chips, let us briefly review evidence supporting the aforementioned scheme. First, the specificity of crosslinking has been shown in some BrdU-containing aptamers. A specific b F G F aptamer binds to bFGF with a Ka of 20 pM. Photo-cross-linking has been performed both with pure protein and protein in the presence of 10% serum (M.C. Golden, B.D. Collins, M.C. Willis, T.C. Koch, manuscript in preparation). In these assays, for which no intermediary washing step has been introduced before photo-cross-linking, there is no other significant cross-linked band visible in the serum sample, and b F G F can be detected down to a concentration of 3.2 pM. More importantly, no competition of material in the specific aptamer b F G F band is seen in the serum sample, suggesting that no blood protein competes for aptamer photo-cross-linking to bFGF. It seems reasonable to expect that with a washing step in between binding and cross-linking, similar specificity would be seen in more concentrated serum (the standard assay protocol). By modeling the combined binding plus photo crosslinking steps as an overall kinetic pathway in which an equilibrium-binding reaction is followed by an irreversible photo-cross-linking step, an overall discrimination rate can be calculated. The simplest model of this process is analogous to a scheme in which an enzyme is challenged with two competing substrates:
Aptamer Arrays in Diagnostics
*
Brody et al.
385
Photo-SELEX Procedure I
I
IIIII
I
fixed A
I
random sequence
gg~ggacgatgcgg ~ 3' ccc~ccbgcbacggg
DNA
I
fixed B
]
I cagaggaegagcggg~.
5'
g t c t c c t g c t c g c c c ~ Biotin
polymemsewith5-BrdUI
Isolate Sense Stand 5 'ggg~ggacgatgcgg
cagaggacgagcggga
3'
ORe
Photo-SELEX Round
Clone Sequence Fig. 3. The modifications to the SELEX procedure needed for photo-SELEX. Single-stranded DNA libraries containing 5-BrdU are allowed to interact with a target protein. Photo-cross-linking is performed as described in text, then SDS-polyacrylamide gel electrophoresis (PAGE) dissociates noncovalent protein-nucleic acid interactions. Unbound single-stranded DNA migrates at its habitual rate in these gels. Covalent protein-DNA complexes migrate more slowly than DNA on these gels. Such slower migrating bands are excised, protein is digested with proteinase K, and the single-stranded DNA is amplified and recycled through rounds of photo-SELEX.
1.hv 1. Biologicalfluid 2. Mild wash
2. Harsh wash
__q I Proteln-speclfic
T Fig. 6. An array of aptamers on a chip is schematized. Each pixel is represented by the teal-colored surface. Aptamers are represented in red and proteins in green. The steps of binding, mild washing, cross-linking, harsh washing, and protein staining are described in text.
386
Molecular Diagnosis Vol. 4 No. 4 December 1999
[A:T]
=
+T/K/KD
A-T
kxT
T
l
Aptamer
+B
hv
"% [A:B]
kxB =
A-B
Where T and B are target and background proteins, respectively, KD is the equilibrium binding constant between aptamer and protein, and kx is the elementary rate constant for irreversible photo-cross-linking. In this scheme, the overall specificity is given by: Fig. 4. Relative reactivity of D N A and proteins with a fluoroA - T
[T](kx/KM) r
A- B
[B] ( k x / K M ) B
where KM is the usual Michaelis constant [19,20]. The implication of this scheme is that specificity in the crosslinking step (kx) will multiplicatively increase the specificity of the equilibrium-binding step (KD). One great advantage of an aptamer-based diagnostic chip is the ease with which proteins covalently linked to aptamers can be detected. Unlike antibody-based reaction schemes, the aptamer-based chip will have no proteins bound to the chip other than the desired analytes. Thus, any protein-specific staining reagent will give an immediate snapshot of a meaningful distribution of blood proteins. Such chemistries have been worked on for more than 70 years, and a huge literature compiled. Specific reagents for tyrosine, cysteine, arginine, and other amino acids are known [21]. Perhaps the most specific reagents are those that react with lysine. Because of its strong nucleophilic character, the primary amine of lysine constitutes a unique target for reactivity. The N7 position of guanine, the strongest nucleophilic group in an aptamer, either DNA- or RNA-based, is much weaker than the primary amine of lysine at p H 7.5. In a series of experiments shown in Figs. 5 and 6, we show that a vicinal dicarbonyl constitutes a universal protein stain. One lysine-reactive reagent will react with all lysines on the chip. Because lysine constitutes 5.7% of human proteins (GenBank Release 108, Aug 15, 1998, Bethesda, MD), on average, such a reagent should be sufficient when combined with a fluorescent reporter adduct to illuminate an entire chip. In the event that some rare proteins have no available lysines, other amino acidspecific reactants could provide additional universal protein stains.
genic dicarbonyl reagent. Human serum albumin (hSa; 1 pmol) and a 42-mer DNA were mixed in the molar ratios indicated at the tops of the gel lanes and reacted with the fluorogenic dicarbonyl reagent CBQCA [23] according to the protocol supplied by the vendor, Molecular Probes (Eugene, OR). The image was obtained by scanning the gel with a Molecular Systems FluorImager (Sunnyvale, CA), M, dye marker lane; N, no DNA or protein; D, a fluorescein-labeled DNA marker.
An aptamer-based chip can thus be stained with one, or at most two, universal protein stains. In such a simple system, fluorescence can be analyzed in arrays on detectors already in use (Fig. 7, see color plate). One automated reader could analyze, in real time, approximately 3,000 to 10,000 chips/d, allowing quantitative data to be garnered from very large numbers of blood samples, probably at first in retrospective studies, to correlate blood protein levels and disease states. With a wide choice of learning algorithms to process these data, it is quite possible to perform ever finer analyses as new markers are added. Because the goal is to simultaneously quantitate thousands of proteins, an algorithm such as the perceptron [22] should be able to transform patterns into yes-no answers. With enough data points, the probability of making a mistake on such a binary decision (Do I have cancer? Has it metastasized? Is it in the bones?) will be vanishingly small. Finally, one imagines that this algorithm will be applicable to each individual in a prospective manner. We anticipate that a drop of blood put onto such a chip once or twice a year will give an immediate readout of a person's health status. Early indicators of disease states could in all probability signal early treatment and eliminate the need for expensive ancillary tests. It is not too soon to imagine that this application of the sequencing of the human genome, the quantification of all important proteins in
Aptamer Arrays in Diagnostics
Fig. 5. Relative reactivity of lysine and nucleotides with a fluorogenic dicarbonyl reagent. N-acetyl lysine (10 IxM) or nucleotide monophosphates (20 mM) were reacted with CBQCA [23] according to a protocol supplied by the vendor, Molecular Probes (Eugene, OR). Fluorescence was detected in a CytoFluor II fluorescence microtiter plate reader (PerSeptive Biosystems,Framingham, MA), Data are the mean of four measurements. NAcLys; AMR adenosine monophosphate; GMR guanosine monophosphate.
2000
I
I
I
I
I
? o o e o
-
Brody et al.
387
I
~ 1500 C "~ ~ 1000
•
o
NAcLys
•
AMP
~'
GNP
°
buffer
2
Z o
e o
I U.
o
500
-.
o
0
0
20
blood (or urine or CSF), will be the most important element of a physician's armamentarium for differential diagnosis in the 21st century.
Acknowledgment The authors thank their colleagues at NeXstar and at the University of Colorado in Boulder for their contributions to the many elements in this work. Many experiments were performed by people who have allowed us to quote from their work. In particular, the authors thank Nebojsa Janjic, Judy Ruckman, Louis Green, Dan Drolet, Mace Golden, Tad Koch, Yun Lin, Jim Heil, Brian Collins, Michael Lochrie, Stephen Capaldi, Philippe Bridonneau, and Gary Cook for their contributions, and Kris Cordova for collating a large amount of material and preparing the manuscript.
Received May 20, 1999. Received in revised form July 19, 1999. Accepted July 19, 1999.
References 1. Zipkin I: Proteomics. BioCentury 1998;6:1-4 2. Tuerk C, Gold L: Systematic evolution of ligands by exponential enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science 1990;249: 505-510 3. Ellington AD, Szostak JW: In vitro selection of RNA molecules that bind specific ligands. Nature 1990;346:818-822 4. Gold L, Polisky B, Uhlenbeck O, Yarus M: Diversity of oligonucleotide functions. Annu Rev Biochem 1995;64:763-797
40
60
80 min
100
120
140
5. Osborne SE, Ellington AD: Nucleic acid selection and the challenge of combinatorial chemistry. Chem Rev 1997;97:349-370 6. Eaton BE: The joys of in vitro selection: Chemically dressing oligonucleotides to satiate protein targets. Curr Opin Chem Biol 1997;1:10-16 7. Jensen KB, Atkinson BL, Willis MC, Koch TH, Gold L: Using in vitro selection to direct the covalent attachment of human immunodeficiency virus type 1 Rev protein to high-affinity RNA ligands. Proc Natl Acad Sci U S A 1995;92:12220-12224 8. Smith D, Kirschenheuter GR Charlton J, Guidot DM, Repine JE: In vitro selection of RNA-based irreversible inhibitors of human neutrophil elastase. Curr Biol 1995;2:741-750 9. Wiegand TW, Janssen RC, Eaton BE: Selection of RNA amide synthases. Chem Biol 1997;4:675-683 10. Wecker M, Smith D, Gold L: In vitro selection of a novel catalytic RNA: Characterization of a sulfur alkylation reaction and interaction with a small peptide. RNA 1996;2:982-994 11. Eaton BE, Pieken WA: Ribonucleosides and RNA. Annu Rev Biochem 1995;64:837-863 12. Cox JC, Rudolph R Ellington AD: Automated RNA selection. Biotechnol Prog 1998;14:845-850 13. Vant-Hull B, Gold L, Zichi D: Basic principles of in vitro selection using combinatorial nucleic acid libraries. Curr Prot Nucleic Acid Chem 1999 14. Romig T, Bell C, Drolet D: Aptamer affinity chromatography; Combinatorial chemistry applied to protein purification. J Chromatogr 1999;731:275284 15; Harlow E, Lane D: Antibodies: A laboratory manual. ColdSpring Harbor Laboratory, Cold Spring Harbor, NY, 1988 16. Hicke B J, Willis MC, Koch TH, Cech TR: Telomeric protein-DNA point contacts identified by photocross-linking using 5-bromodeoxyuridine [published
388
17,
18.
19. 20. 21. 22.
Molecular Diagnosis Vol. 4 No. 4 December 1999
erratum, Biochemistry 1994;33:7744]. Biochemistry 1994;33:3364-3373 WillisMC, LeCuyer KA, Meisenheimer KM, Uhlenbeck OC, Koch TH: An RNA-protein contact determined by 5-bromouridine substitution, photo-crosslinking and sequencing. Nucleic Acids Res 1994;22: 4947-4952 Meisenheimer KM, Meisenheimel' PL, Willis MC, Koch TH: High-yield photocrosslinking of a 5iodocytidine (IC) substituted RNA to its associated protein. Nucleic Acids Res 1996;24:981-982 Fersht A: Enzyme structure and mechanism. New York, NY, W.H. Freeman, 1985 Yarus M: The accuracy of translation. Prog Nucleic Acid Res Mol Biol 1979;23:195-225 Wong S: Chemistry of protein conjugation and crosslinking. Boca Raton, FL, CRC Press, 1991 Schneider TD, Stormo GD, Gold L, Ehrenfeucht A:
Information content of binding sites on nucleotide sequences. J Mol Biol 1986;188:415-431 23. Liu JR Hsieh YZ, Wiesler D, Novotny M: Design of 3-(4-carboxybenzoyl)-2- quinolinecarboxaldehyde as a reagent for ultrasensitive determination of primary amines by capillary electrophoresis using laser fluorescence detection. Anal Chem 1991;63:408-412 24. Ruckman J, Green LS, Beeson J, et al.: 2'-Fluoropyrimidine RNA-based aptamers to the 165-amino acid form of vascular endothelial growth factor (VEGF165). Inhibition of receptor binding and VEGF-induced vascular permeability through interactions requiring the exon 7-encoded domain. J Biol Chem 1998;273:20556-205567 25. Green LS, Jellinek D, Jenison R, Ostman A, Heldin CH, Janjic N: Inhibitory DNA ligands to plateletderived growth factor B-chain. Biochemistry 1996; 35:14413-14424