Quantitative analysis of receptors for adenosine nucleotides obtained via in vitro selection from a library incorporating a cationic nucleotide analog

Article abstract of DOI:10.1021/ja9816436

5-(3″-Aminopropynyl)-2′-deoxyuridine (dJ), a modified nucleoside with a side chain carrying a cationic functional group, was incorporated into an oligonucleotide library, which was amplified using the Vent DNA polymerase in a polymerase chain reaction (PC

Full text of DOI:10.1021/ja9816436

J. Am. Chem. Soc. 1999, 121, 9781-9789
9781
Quantitative Analysis of Receptors for Adenosine Nucleotides
Obtained via In Vitro Selection from a Library Incorporating a
Cationic Nucleotide Analog
‡,
Thomas R. Battersby,† Darwin N. Ang,‡ Petra Burgstaller,§,| Simona C. Jurczyk,
Michael T. Bowser,| Danielle D. Buchanan,| Robert T. Kennedy,| and Steven A. Benner*, ‡,|
Contribution from the Department of Chemistry, Department of Anatomy and Cell Biology, and the
Florida Center for Heterocyclic Compounds, UniVersity of Florida, GainesVille, Florida 32611, and
EraGen, Inc. 12085 Research DriVe, Alachua, Florida 32615
ReceiVed May 11, 1998. ReVised Manuscript ReceiVed NoVember 16, 1998
Abstract: 5-(3′′-Aminopropynyl)-2′-deoxyuridine (dJ), a modified nucleoside with a side chain carrying a
cationic functional group, was incorporated into an oligonucleotide library, which was amplified using the
Vent DNA polymerase in a polymerase chain reaction (PCR). When coupled to an in vitro selection procedure,
PCR amplification generated receptors that bind ATP. This is the first example of an in vitro selection generating
oligonucleotide receptors where the oligonucleotide library has incorporated a cationic nucleotide functionality.
The selection yielded functionalized receptors having sequences differing from a motif known to arise in a
standard selection experiment using only natural nucleotides. Surprisingly, both the natural and the functionalized
motifs convergently evolved to bind not one, but two ATP molecules cooperatively. Likewise, the affinity of
the receptors for ATP had converged; in both cases, the receptors are half saturated at the 3 mM concentrations
of ATP presented during the selection. The convergence of phenotype suggests that the outcome of this selection
experiment was determined by features of the environment during which selection occurs, in particular, a
highly loaded affinity resin used in the selection step. Further, the convergence of phenotype suggests that the
optimal molecular phenotype has been achieved by both selections for the selection conditions. This interplay
between environmental conditions demanding a function of a biopolymer and the ability of the biopolymer to
deliver that function is strictly analogous to that observed during natural selection, illustrating the nature of
life as a self-sustaining chemical system capable of Darwinian evolution.
Introduction
any polymer built from a limited set of building blocks,
“structure space” in nucleic acids is well defined, with a
countable number of possible sequences. This allows us to
quantify the probability of finding a receptor or catalyst meeting
certain specifications within that space. In many cases, the
probability is low, especially when compared with the structure
space defined by the 20 amino acid building blocks of
polypeptides.3 For example, to have a 50% chance of holding
a single RNA molecule able to catalyze a template-directed
ligation reaction by a modest (by protein standards) factor of
104, a library of RNA molecules 220 nucleotides in length must
contain ∼2 × 1013 random sequences.4
In vitro selection is a combinatorial method that generates
receptors, ligands, and catalysts from nucleic acid libraries
containing as many as 1015 different molecules.1 It exploits a
“selection” procedure to separate RNA or DNA molecules with
specific binding or catalytic properties from those lacking these
properties, the polymerase chain reaction (PCR) to amplify
single selected RNA or DNA molecules to give a large number
of their descendants, and mutation to allow the descendent
molecules to “evolve”, improving their binding or catalytic
activities.
In addition to its technological value,2 in vitro selection offers
the opportunity to ask general questions about the relation
between structure and function in organic molecules. As with
Natural evolution in the biomolecular world is also described
as the search for function within a biopolymeric sequence space.5
As with natural evolutionary experiments, the outcome of an
in vitro selection experiment reflects many parameters. These
include the details of the selection conditions, the number of
selection cycles performed (and hence the amount of structure
space explored), and the “searchability” of structure space, an
abstract concept that depends on the “ruggedness” of the
landscape that relates fitness to structure. Ruggedness is
determined by the structure and structural versatility of the
biopolymer whose sequence space is being searched.
* To whom correspondence should be addressed at the Department of
Chemistry, University of Florida.
† EraGen, Inc.
‡ Department of Anatomy and Cell Biology, University of Florida.
§ Present address: Aventis Research & Technologies GmbH & Co. KG,
65926 Frankfurt-am-Main, Germany.
| Department of Chemistry, University of Florida.
Florida Center for Heterocyclic Compounds, University of Florida.
(1) (a) Szostak, J. W. In Redesigning the Molecules of Life; Benner, S.
A., Ed.; Springer-Verlag: Heidelberg, Germany, 1988; pp 87-114. (b)
Joyce, G. F. Gene 1989, 82, 83-87. (c) Joyce, G. F. UCLA Symp. Mol.
Cell. Biol. New Ser. 1989, 94, 361-371. (d) Irvine, D.; Tuerk, C.; Gold, L.
J. Mol. Biol. 1991, 222, 739-761.
(2) For some examples, see (a) Wiegand, T. W.; Janssen, R. C.; Eaton,
B. E. Chem. Biol. 1997, 4, 675-683. (b) Zhang, B. L.; Cech, T. R. Nature
1997, 390, 96-100.
(3) Johnsson, K.; Allemann, R. K.; Widmer, H.; Benner, S. A. Nature
1993, 365, 530-532.
(4) Bartel, D. P.; Szostak, J. W. Science 1993, 261, 1411-1418.
(5) (a) Benner, S. A.; Ellington, A. D. CRC Crit. ReV. Biochem. 1988,
23, 369-426. (b) Benner, S. A. Chem. ReV. 1989, 89, 789-806.
10.1021/ja9816436 CCC: $18.00 © 1999 American Chemical Society
Published on Web 10/13/1999

9782 J. Am. Chem. Soc., Vol. 121, No. 42, 1999
Battersby et al.
experiments, dJTP was examined as a substrate for a range of
thermostable polymerases. As has been observed with previous
experiments with nonstandard bases,14 different polymerases
behaved quite differently with respect to the unnatural modifica-
tion. Taq polymerase, for example, incorporates dATP opposite
dJ in the template but stops rather than incorporating dJTP
opposite dA in a template. Tth and Tfl polymerases pause when
encountering dJ both in a template and as a triphosphate. Vent
polymerase, however, incorporates both dJTP opposite dA and
dATP opposite dJ with only modest amounts of pausing. Vent
polymerase was therefore chosen as a tool for incorporating dJ
into a PCR experiment.
PCR experiments were then run to show that Vent is able to
amplify oligonucleotides containing the positively charged
nucleobase dJ. Parallel PCR experiments were run with natural
deoxynucleoside triphosphate mixes and mixes having TTP
replaced by dJTP (data not shown). In both cases, amplification
was seen, with the dJTP supporting PCR amplification only
slightly less well than TTP with the Vent polymerase.
With Vent polymerase enabling PCR amplification with dJTP,
two in vitro selection experiments were run in parallel to select
for receptors that bind the adenosine derivatives ATP, ADP,
and AMP, one using a library built from natural nucleotides T,
dA, dC, and dG (and therefore exploring “natural structure
space”) and the other using a library built from dJ, dA, dC, and
dG (therefore exploring “unnatural structure space”). These
experiments followed closely the recipe of Huizenga and
Szostak, who did this selection with natural nucleotides some
time ago.15 Nine rounds of selection, each separated by 10-14
cycles of amplification, were performed by passing successively
enriched libraries through a column with immobilized ATP,
ADP, and AMP. Products were cloned (with T replacing dJ),
and a set of clones was sequenced.
Figure 1. 5-(3′′-Aminopropynyl)-2′-deoxyuridine (“dJ”, R) 2′-deox-
yribose), a modified nucleoside with a side chain carrying a cationic
functional group.
One approach to understand the interaction between these
parameters involves altering the nature of the building blocks
used by the biopolymer, incorporating nonstandard6 or func-
tionalized7 derivatives of nucleic acids in in vitro selection
experiments. Technical obstacles hinder this approach. Even
modest discrimination by a polymerase against an unnatural
functionalized nucleotide may lead to a selective loss of the
functionality during PCR amplification. While polymerases are
known to incorporate nucleotides carrying neutral functionality8
and two in vitro selection experiments have incorporated such
nucleotides,9 it has proven especially difficult to amplify
oligonucleotides containing nucleotide analogues carrying posi-
tively charged functionality, the functionality most notably
missing from standard DNA and RNA.
We report here the first example of incorporation of a
positively charged functional group into an in vitro selection
experiment and the first quantitative comparison of a selection
experiment with standard biopolymers and functionalized
biopolymers of any kind. The results provide a remarkable
example of convergent evolution of physical behavior in two
molecular systems, as well as a view of the landscapes that relate
structure and behavior in macromolecules.
A total of 17 sequences were recovered from the in vitro
selection experiment performed with standard nucleotides
(Figure 2). All appeared to be derived from different ancestors.
A total of 39 sequences were recovered from the dJ library
(Figure 2). These represented 30 ancestral sequences, with
sequences 401 and 409; 4011, 4013, and 4014; 504, 505, 509,
5010, and 5015; 602 and 609; and 6010 and 6015 forming five
individual families.
The randomized sequences obtained from the selection with
natural nucleotides were approximately 69 nucleotides in length,
virtually identical to the length of the initial random pool. In
contrast, the average length of the randomized region resulting
from selection with the dJ library was only 25 nucleotides.
Further, the natural library yielded randomized regions contain-
ing 25% T, while the dJ library yielded randomized regions
containing 14% dJ. These results suggest that Vent, although
satisfactory as a polymerase for performing these experiments,
nevertheless selects against dJ and that further work developing
new polymerases that incorporate functionalized and nonstand-
ard nucleotides is warranted. Work to improve polymerases is
ongoing in these laboratories.16
Results
5-(3′′-Aminopropynyl)-2′-deoxyuridine10 (Figure 1), trivially
designated as “dJ”, was prepared by coupling N-propynyltri-
fluoroacetamide11,12 and 2′-deoxy-5-iodouridine with the aid of
a palladium catalyst and converted to the corresponding
triphosphate following a procedure of Ludwig and Eckstein.13
The side chain ammonium ion, with an expected pKa of ∼9.5,
carries a positive charge at physiological pH. In preliminary
(6) (a) Switzer, C. Y.; Moroney, S. E.; Benner, S. A. J. Am. Chem. Soc.
1989, 111, 8322-8323. (b) Piccirilli, J. A.; Krauch, T.; Moroney, S. E.;
Benner, S. A. Nature 1990, 343, 33-37. (c) Krosigk, U.; von, Benner, S.
A. J. Am. Chem. Soc. 1995, 117, 5361-5362. (d) Vo¨gel, J. J.; Benner, S.
A. J. Am. Chem. Soc. 1994, 116, 6929-6930. (e) Vo¨gel, J. J.; Altorfer, M.
M.; Benner, S. A. HelV. Chim. Acta 1993, 76, 2061-2069. (f) Vo¨gel, J. J.;
von Krosigk, U.; Benner, S. A. J. Org. Chem. 1993, 58, 7542-7547.
(7) (a) Heeb, N. V.; Benner, S. A. Tetrahedron Lett. 1994, 35, 3045-
3048. (b) Robertson, M. P.; Miller, S. L. Science 1995, 268, 702-705. (c)
Dewey, T. M.; Zyzniewski, M. C.; Eaton, B. E. Nucleosides Nucleotides
1996, 15, 1611-1617. (d) Kodra, J.; Benner, S. A. Synth. Lett. 1997, 939-
940.
(8) (a) Langer, P. R.; Waldrop, A. A.; Ward, D. C. Proc. Natl. Acad.
Sci. U.S.A. 1981, 78, 6633-6637. (b) Prober, J. M.; Trainor, G. L.; Dam,
R. J.; Hobbs, F. W.; Robertson, C. W.; Zagursky, R. J.; Cocuzza, A. J.;
Jensen, M. A.; Baumeister, K. Science 1987, 238, 336-341.
(9) (a) Latham, J. A.; Johnson, R.; Toole, J. J. Nucleic Acids Res. 1994,
22, 2817-2822. (b) Tarasow, T. M.; Tarasow, S. L.; Eaton, B. E. Nature
1997, 389, 54-57.
(10) (a) Hobbs, F. W., Jr. J. Org. Chem. 1989, 54, 3420-3422. (b)
Casalnuovo, A. L.; Calabrese, J. C. J. Am. Chem. Soc. 1990, 112, 4324-
4330.
The sequences were then examined to see whether they
resembled the sequences reported by Huizenga and Szostak,
who also selected a DNA molecule that could bind to adenosine
derivatives.15 Their paper does not report a sample of the
(14) (a) Horlacher, J.; Hottiger, M.; Podust, V. N.; Hu¨bscher, U.; Benner,
S. A. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 6329-6333. (b) Lutz, M. J.;
Held, H. A.; Hottiger, M.; Hu¨bscher, U.; Benner, S. A. Nucleic Acids Res.
1996, 24, 1308-1313.
(15) Huizenga, D. E.; Szostak, J. W. Biochemistry 1995, 34, 656-665.
(16) (a) Lutz, M.; Horlacher, J.; Benner, S. A. Bioorg. Med. Chem. Lett.
1998, 8, 499-504. (b) Lutz, M. J.; Horlacher, J.; Benner, S. A. Bioorg.
Med. Chem. Lett. 1998, 8, 1149-1152.
(11) Cruickshank, K. A.; Stockwell, D. L. Tetrahedron Lett. 1988, 29,
5221-5224.
(12) Meyer, R. B., Jr. in Protocols for Oligonucleotide Conjugates:
Synthesis and Analytical Techniques; Agrawal, S., Ed.; Humana Press:
Totowa, New Jersey; pp 77-80.
(13) Ludwig, J.; Eckstein, F. J. Org. Chem. 1989, 54, 631-635.

Receptors for Adenosine Nucleotides
J. Am. Chem. Soc., Vol. 121, No. 42, 1999 9783
Figure 2. Sequences generated from in vitro selection starting from a natural library and from a library containing dJ. Underlined regions are
derived from the primer or primer binding regions.
sequences generated by the selection directly. Huizenga and
Szostak did, however, report a single 42mer sequence derived
from the selection, generated 43 unique sequences from it using
mutation-prone PCR followed by a selection, and analyzed these
for evidence of a conserved motif among those that continued
to bind adenosine derivatives. Lin and Patel examined by NMR
the solution structure of a 25mer designed from the sequences
obtained by mutation-prone PCR, and to which had been added
an additional base pair on the stem;17 the structure is shown
schematically in Figure 3.

9784 J. Am. Chem. Soc., Vol. 121, No. 42, 1999
Battersby et al.
Huizenga-Szostak fold was likely to be adopted by five of these
(Figure 5) (sequences 4011, 4012, 4013, 4014, and 608).
Sequences 4011, 4013, and 4014 are identical and presumably
arose from a single ancestral sequence in the original library.
They are 31 nucleotides long, and share 31 nucleotides in
sequence. As there are ∼4 × 1018 possible 31mers, it is unlikely
that these three sequences were found three times in the initial
population. Sequence 4012 apparently arose independently. All
presumably adopt the Lin-Patel-Huizenga-Szostak fold. They
all contain the two G consensus regions, and all have the
potential to form the inner and outer Watson-Crick base pairs
required by the fold. In addition, sequence 404 may possibly
adopt a Lin-Patel-Huizenga-Szostak fold, even though a part
of the second G-rich region is not canonical and it contains a
non-GT stem mismatch in the outer stem positions closest to
the conserved regions (not observed in any of the Huizenga-
Szostak binders). An inner Watson-Crick pair is possible, as
is an outer stem joined by three external Watson-Crick base
pairs (Figure 5).
Most of the sequences arising from the dJ selection do not
obviously fit the Lin-Patel-Huizenga-Szostak motif. Many
do, however, display features in common, whether they are
obtained for binding to AMP, ADP, or ATP (Figure 6). A new
motif can be defined relative to these features as having (i) a
stretch of six G’s followed by an A, (ii) a preceding GGA, (iii)
the potential for Watson-Crick base pairs in inside stems just
after the GGA and immediately before the GGGGGGA
sequences, and (iv) the potential for Watson-Crick base pairs
in outside stems four nucleotides before the GGA and im-
mediately following the GGGGGGA sequence.
When selected for affinity to ADP, the dJ-library also yielded
a receptor entirely lacking dJ in the randomized region and
containing only three dJ-nucleotides in the segment of the
receptor arising through copying of the 3′-primer binding site.
It is represented by sequences 504, 505, 509, 5010, and 5015.
All five of these sequences appear to have arisen from a single
ancestor, as they are 28 nucleotides long and differ by a single
nucleotide substitution (Figure 7). This sequence does not fit
well the Lin-Patel-Huizenga-Szostak motif. It has three
G-rich segments, not the two found in the Patel-Huizenga-
Szostak motif. Further, the sequence lacks both the standard
internal Watson-Crick potential pair as well as the external
pairing that might generate the outer stem. This may be an as-
yet undescribed motif for a DNA receptor for adenosine
derivatives.
To explore the affinity of the natural and functionalized
receptors, two sequences containing a representative Lin-Patel-
Huizenga-Szostak motif (CTACCTGGGGGAGCATTGGG-
GAGGAAGGTAGCCGTGCGAAAA, Sequence A and GTGCT-
TGGGGGAGTATTGCGGAGGAAAGCGGCCCTGCTGAAG,
Sequence B) and a sequence containing dJ (Sequence 409,
Figure 2) were synthesized with 3′-fluorescent tags and their
binding to ATP examined. Huizenga and Szostak reported that
DNA molecules having the motif contained in Sequences A
and B bound adenosine with an equilibrium disassociation
constant (Kd) of ∼6-8 µM. We re-examined the binding of
ATP to Sequence A and Sequence B using affinity capillary
electrophoresis,18 and found a sigmoidal binding curve char-
acteristic of a 1:2 complex in both cases (Figure 8). No evidence
could be found for a significant contribution of the 1:1 complex
between the receptor and ATP at thermodynamic equilibrium.
A termolecular complex was also observed by Lin and Patel in
the NMR structure of the Huizenga-Szostak receptor with
Figure 3. A schematic illustrating the fold determined by Lin and
Patel17 for a DNA molecule prepared by Huizenga and Szostak,15 and
how sequence 4013, selected from the dJ-functionalized DNA library
for binding to AMP, fits. Conserved Watson-Crick pairing is indicated
by o. Nucleotides present in the Lin-Patel structure are in upper case;
nucleotides different in Sequence 4013 are in lower case.
From this combination of experiments, including both the
structure and the pattern of sequence conservation, we defined
a “Lin-Patel-Huizenga-Szostak motif” in terms of the fol-
lowing features: i. G5, A10, G18, and A23 are absolutely
invariant and were conserved in the Huizenga-Szostak mu-
tagenesis studies for effective receptors. ii. G6, G8, G9, G19,
G21, and G22 are nearly always guanosine. In the Huizenga-
Szostak mutagenesis studies, only 12 of the 264 positions
sampled failed to have a G at these positions, and all of these
substituted G by A. The NMR structure shows that G6-G21
and G8-G19 form two reverse Hoogsteen base pairs. G9 and
G22 bind to the AMP ligand in the Lin-Patel structure. iii.
The bases adjacent to G5 and A23 always form a Watson-
Crick base pair (or, in some cases, a G-T wobble). In the
Huizenga-Szostak mutagenesis studies, only 4 of the 44
examples sampled were these positions not Watson-Crick
complementary and, of these, three were GT pairs. These
features were used to define the Lin-Patel-Huizenga-Szostak
motif. In addition, several sequence features frequently were
found. iv. G7 and A20 are usually purines, with one position G
and the other A. In the Huizenga-Szostak mutagenesis studies,
only 2 of the 88 positions mutated to pyrimidines. v. The bases
at positions A10 and G18, first position of the inner loop, can
usually form a Watson-Crick base pair, or a GT wobble. In
the Huizenga-Szostak mutagenesis studies, 8 of the 44
examples had a mismatch at this position; of these, three were
GT pairs. In the NMR structure, a Watson-Crick base pair was
formed. vi. The outer stem frequently could form two, and
sometimes more, Watson-Crick pairs. In the Huizenga-Szostak
mutagenesis studies, 25 of the 44 sequences had three or more
of these. Of the remainder, five had the potential for forming
two Watson-Crick pairs plus one GT pair, 10 could form
Watson-Crick pairs in two of the three positions, while four
could form one Watson-Crick pair plus one GT wobble.
Huizenga and Szostak note that one base pair was sufficient to
detect ligand interaction on the ATP column, but three potential
base pairs were required for optimal affinity.
Of the 17 sequences obtained from the unfunctionalized
library in this work, sequences 202, 204, 206, 207, 209, 2010,
301, 302, 305, 306, 309, and 3010 (12 in total, 71%) met these
criteria to an acceptable degree (Figure 4). The remaining
sequences (201, 203, 208, 304, and 307) from the standard
selection did not obviously fit the Lin-Patel-Huizenga-
Szostak motif. Sequence 204 has a T at a position requiring a
G; sequence 3010 is missing the A at position 23.
A similar analysis was then applied to the sequences selected
from the dJ library. This analysis suggested that the Lin-Patel-
(18) Chu, Y. H.; Avila, L. Z.; Gao, J. M.; Whitesides, G. M. Acc. Chem.
Res. 1995, 28, 461-468.
(17) Lin, C. H.; Patel, D. J. Chem. Biol. 1997, 4, 817-832.

Receptors for Adenosine Nucleotides
J. Am. Chem. Soc., Vol. 121, No. 42, 1999 9785
Figure 4. Analysis of the sequences obtained from the unfunctionalized library to identify sequences that fit the Lin-Patel-Huizenga-Szostak
motif.15,17 Positions above + are absolutely conserved in the Lin-Patel-Huizenga-Szostak motif. Underlined regions are the G-rich segments that
are largely conserved. Positions above x violate the motif. Segments forming in the inner and outer stem base pairs are indicated by a * below the
sequence. Italicized regions are derived from primer or primer-binding regions.
Figure 5. A comparison of the Lin-Patel-Huizenga-Szostak motif with sequences from the dJ selection that might display some of its features.
Underlined regions are the G-rich motifs. Segments forming in the inner and outer stem base pairs are indicated by a * below the sequence.
Positions above x violate the motif (note especially the violations in sequence 404). Italicized regions are derived from primer or primer-binding
regions. Sequences 4011, 4013, and 4014 are identical and appear to be derived from a single sequence in the pool.
AMP.17 These authors also found no concentration of AMP
where a bimolecular complex could be observed.
respectively, different from the (6-8) × 10-6 M (note the units)
disassociation constant reported for adenosine in the litera-
ture.15,20
Remarkably, the dJ-containing Sequence 409 also forms a
termolecular complex with ATP, as shown by a sigmoidal
binding curve (Figure 8). This implies that the equilibrium
constant must again have units of M2, not M. Again, no
(19) Bowser, M. T.; Chen, D. D. Y. Anal. Chem. 1998, 70, 3261-3270.
This termolecular complex implies that the equilibrium
constants describing the binding of adenosine derivatives to
Sequence A and Sequence B must have units of M2, not M.
Fitting of the binding data using the method of Bowser and
Chen19 gave affinity constants of 9 ( 2 × 10-6 and 15 ( 2 ×
10-6 M2 (note the units) for Sequence A and Sequence B,

9786 J. Am. Chem. Soc., Vol. 121, No. 42, 1999
Battersby et al.
Figure 6. Most of the sequences arising from the dJ selection do not obviously fit the Lin-Patel-Huizenga-Szostak motif. They do, however,
conform to a new motif, defined as having (i) a stretch of six G’s followed by an A, (ii) a preceding GGA, (iii) the potential for Watson-Crick
base pairs in inside stems just after the GGA and immediately before the GGGGGGA sequences, and (iv) the potential for Watson-Crick base
pairs in outside stems four nucleotides before the GGA and immediately following the GGGGGGA sequence. Italicized regions are derived from
primer or primer-binding regions. Segments having the potential for Watson-Crick complementarity are indicated by a * below the sequence. A
G-T wobble is indicated by a w. Sequences presumed to be derived from a single ancestral sequence in the pool are grouped together.
Figure 7. The dJ-poor sequences emerging from the selection with dJ libraries for receptors that bind to ADP. The three G-rich segments are
underlined. Italicized regions are derived from primer or primer-binding regions. These oligonucleotides all appear to have arisen from a single
ancestral sequence in the pool.
significant amounts of 1:1 complex were detected by affinity
capillary electrophoresis.
experiments have been used to quantitatively understand the
relationship between structure and behavior in the large col-
lections of molecules that are the starting point for these
selection experiments. In vitro selection experiments may make
their most important contribution by way of such analyses, as
they concern questions at the center of both chemistry (How
are behaviors distributed in a landscape of structure space?) and
biology (How does Darwinian evolution identify functional
behavior in a random search of a fitness-structure landscape?).
These results show how the introduction of expanded genetic
alphabets containing extra letters6 and new chemical functional-
Discussion
Although in vitro selection has been known as an experi-
mental technique for more than a decade, few in vitro selection
(20) A false perception of the affinity between a receptor and a ligand
can be obtained if single data points are fitted to a model that makes an
incorrect assumption about the molecularity of the complex; see Benner,
S. A.; Burgstaller, P.; Battersby, T. R.; Jurczyk, S. in The RNA World, 2nd
ed.; Cech, T. R., Atkins, J., Eds.; Cold Spring Harbor Press: Plainview,
NY, 1999; 163-181.

Receptors for Adenosine Nucleotides
J. Am. Chem. Soc., Vol. 121, No. 42, 1999 9787
is, we believe, the first time that such convergence has been
observed in a “non-living” system. It is, of course, commonplace
in living systems. Convergence at the level of sequence may
also occur from time to time.21 Thus, certain features of bat
wings and bird wings are strictly analogous (weight-power
ratios, for example), even though the two physiological struc-
tures are not homologous. The analogous phenotypes suggest
that the features of the phenotype are optimal; such a conclusion
would not be possible in the absence of convergence. Thus,
the interplay between environmental conditions demanding a
function of a biopolymer and the ability of the biopolymer to
deliver that function is strictly analogous to that observed during
natural selection, illustrating the nature of life as a self-sustaining
chemical system capable of Darwinian evolution.
Is there any reason to believe that structure space delivered
by the functionalized biopolymer containing dJ offered a richer
density of functionality than the natural biopolymer? More work
needs to be done to say for certain. However, it is interesting
to note that the sequences selected to bind ATP were longer
and, consequently, had a higher number of dJs than the
sequences arising from selection to bind the less negatively
charged adenosine analogues ADP and AMP.
Figure 8. Equilibrium binding curves determined for Sequence A
(diamonds) containing only standard nucleotides, and Sequence 409
(squares), combining the randomized region (in italics) with the 5′-
and 3′-primer binding sites (GGTCGTCTAGAGTATGCGGTAG-
GAACGJCAGJG GGGGGAG CAJAJGGJGJGAJACGCGACCGAA-
GAAGCJJGGCCCAJG), and containing the functionalized nucleotide
dJ.
It is clear, however, that the addition of functionality did not
change dramatically the phenotype-structure landscape. The first
indicator of this is the fact that the dJ library generated a few
oligonucleotides that almost certainly fold in the same way that
the Lin-Patel-Huizenga-Szostak motif folds and presumably
bind adenosine derivatives in an analogous way. Further, we
prepared one of the dJ-containing receptors, Sequence 409, with
T replacing dJ. This receptor was found to bind ATP, although
with a 2-fold weaker affinity (13 ( 4 × 10-6 M2) than the
analogous sequence containing dJ. The fact that the Lin-Patel-
Huizenga-Szostak fold evidently appears only infrequently in
the dJ-selection suggests that the selection procedure is demand-
ing enough to exclude receptors with 2-fold weaker affinity.
Together, these results suggest that the functionalization con-
tributed by dJ influences modestly, rather than dramatically, the
affinity-structure landscape, at least for these receptors selected
under rather nondemanding conditions. When called upon to
provide catalytic activity, however, DNA makes greater use of
functionality (Ang et al., in preparation).
ity7 empowers in vitro selection as a tool for addressing such
general questions.
Several features of the results reported here are worth noting.
First, our experiments with natural DNA libraries independently
generated the same motif as generated by Huizenga and Szostak
under analogous conditions. This convergence is not surprising,
as the Huizenga-Szostak motif is short enough to be present
in multiple copies in any random library of the size used in
these experiments.
Less expected, however, was the convergence of molecular
phenotype starting from two types of library, the natural library
and the functionalized dJ library. Both libraries yielded receptors
that bind two molecules of ATP cooperatively in a termolecular
complex. We can only speculate why forming a ternary complex
is optimal for the DNA and dJ-functionalized DNA systems.
Almost certainly relevant is the fact that the selection recipe
exploits an agarose affinity resin that is highly loaded with ATP
(∼3 mM effective concentration); most of the glucose units in
the agarose strand had one ATP molecule covalently bound.
For this reason, if a receptor binds one polymer-bound ATP
molecule, it would almost always have the opportunity to bind
a second. It appears that sequences exploiting this opportunity
to grasp the solid support during the selection step are present
in greater abundance in both libraries than sequences that have
comparable affinity by binding just one agarose-bound ATP.
The independent searches of natural and unnatural sequence
spaces also converged on similar affinity constants. The
equilibrium constant for ATP binding to Sequence 409 is 6 (
1 × 10-6 M2 (note the units). The convergence of the two
disassociation constants presumably reflects the fact that the
effective concentration of adenosine derivative attached to the
solid support was ∼3 × 10-3 M. At these concentrations, the
receptors were approximately half saturated with ligand. Thus,
the receptor evidently evolved an affinity that is optimal for
the selection conditions. The affinity was not maximized (that
is, increased until the receptor could bind no tighter), presumably
because no pressure selected for receptors with higher affinity,
and higher-affinity receptors are more scarce in the sequence
space than lower-affinity receptors.
Experimental Section
Reagents were purchased from Aldrich or Fisher, unless otherwise
indicated. NMR spectra were recorded on a Varian XL 300 spectrometer
at 300 MHz (1H) referenced to TMS, at 75.4 MHz (13C) referenced to
solvent, and at 121.4 MHz (31P) with H3PO4 as standard in the solvents
as given. UV spectra were measured on a Varian Cary 1 Bio
spectrophotometer. Mass spectrometry was performed by Spectroscopy
Services of the University of Florida Chemistry Department.
Chromatography. Column chromatography was performed with
silica gel (230-425 mesh, Fisher) and the indicated eluting solvents.
Silica gel plates with 254 nm fluorescence (Whatman) were used for
TLC. TLC separations were also visualized by staining with a Ce/Mo
reagent (2.5% phosphormolybdic acid, 1% Ce(IV)(SO4)2‚4H2O, 6% H2-
SO4 in H2O) and heating. AG 1-X8 resin (Bio-Rad) was obtained as
the Cl- form and converted to the HCO3 form by washing with 16
-
volumes of 1 M NH4HCO3 solution, followed by deionized water and
finally with 0.5 M NH4HCO3 solution; following this treatment, no
eluting Cl- was detected. Ion-exchange chromatography was done with
DEAE Sephadex (Sigma) equilibrated in 0.2 M (Et3NH)HCO3 (pH 7.0).
HPLC was performed using a Waters PrepLC 4000 System with a 486
tunable absorbance detector. Reversed-phase HPLC separation was done
Whether it occurs in biology or in nonliving chemical
systems, convergence of phenotype suggests that the phenotype
converged upon is optimal for a particular environment. This
(21) Stewart, C. B.; Schilling, J. W.; Wilson, A. C. Nature 1987, 330,
401-404.

9788 J. Am. Chem. Soc., Vol. 121, No. 42, 1999
Battersby et al.
using a Waters PrepLC 25 mm module containing a single PrepPak
cartridge (Prep Nova-Pak HR C18 6 µm 60 Å, 25 × 100 mm).
5-(3′′-Trifluoroacetamidopropynyl)-2′-deoxyuridine. 5-Iodo-2′-
deoxyuridine (Sigma; 1.41 mmol, 500 mg) was dissolved in dry DMF
(12 mL). Ar was passed through this solution for 10 min. (Ph3P)4Pd
(0.1 equiv, 0.141 mmol, 163 mg) was added, and Ar was passed through
the solution for another 5 min. Triethylamine (2.0 equiv, 2.83 mmol,
285 mg, 0.393 mL) was added via syringe, followed by addition of
N-propynyltrifluoroacetamide (2.5 equiv, 3.53 mmol, 533 mg) and CuI
(0.2 equiv, 0.282 mmol, 53.7 mg). The mixture was stirred at 40 °C
for 5 h, the solvent was evaporated, and the residue was dissolved in
MeOH/methylene chloride 1:1 (10 mL). Ion-exchange resin (AG1-X8
HCO3- form, 1.5 g) was added to remove the Et3N‚HI byproduct, and
the mixture was stirred at room temperature for 30 min. The mixture
was filtered through Celite, the solid was washed with MeOH/methylene
chloride 1:1 (10 mL), and the combined filtrates were evaporated. The
residue was purified by silica column chromatography (chloroform/
MeOH 8.25:1.75) to yield product as a yellow foam containing some
dry MeOH (1 mL). A solution of anhydrous HCl in anhydrous MeOH
(10%, 1 mL) was added, and the reaction was complete after 3 min
(TLC, chloroform/10% MeOH, Rf ) 0.46). It was cooled to 0 °C, and
aqueous NaHCO3 solution was added to adjust pH 7. It was extracted
(ethyl acetate) and dried (Na2SO4) and the solvent evaporated. The
residue was purified by column chromatography (chloroform/10%
1
MeOH) to give 371 mg (68%) of a slightly yellow foam. H NMR
(CD3OD): 2.03 (s, 3H, Ac), 2.18-2.40 (m, 2H, 2′), 3.71 (m, 2H, 5′),
4.02 (m, 1H, 4′), 4.19 (d, 2H, H-9), 5.21 (m, 1H, 3′), 6.14 (t, 1H, 1′),
8.25 (s, 1H, H-6). 13C NMR (CD3OD): 20.9 (Ac), 30.7 (C-9), 39.0
(2′), 62.8 (5′), 76.0 (3′), 76.4 (C-8), 86.9 (1′), 87.1 (C-7), 88.5 (4′),
99.9 (C-5), 111.6, 115.4, 119.2, 123.0 (q, CF3, J ) 285.99 Hz), 145.5
(C-6), 151.1 (C-2), 158.2, 158.7 (q, COCF3, J ) 37.78 Hz), 164.4 (C-
4), 172.2 (Ac). HRMS (FAB+): calculated, 420.1019; found, 420.1026.
5-(3′′-Aminopropynyl)-2′-deoxyuridine 5′-triphosphate. The com-
pound from the previous step (1.21 mmol, 505 mg) was coevaporated
with pyridine and dissolved in 1.75 mL of pyridine/5.25 mL of dioxane
(anhydrous), and 2-chloro-4H-1,3,2-benzodioxaphosphorin-4-one (1.1
equiv, 1.33 mmol, 269 mg), dissolved in 1.4 mL dioxane, was added.
A white precipitate formed immediately, and the mixture was stirred
at room temperature for 10 min. Tributylammonium pyrophosphate (819
mg), dissolved in 4.55 mL of DMF/1.3 mL of tributylamine, was added,
and the precipitate was dissolved within a few seconds. The solution
was stirred for another 10 min, iodine (1.1 equiv, 1.33 mmol, 337 mg),
dissolved in 30 mL of pyridine/0.6 mL of water was added, and the
resulting solution was stirred for 15 min. The reaction was quenched
by addition of 5% aqueous Na2SO3. The solvent was evaporated at
room temperature, and the residue was dissolved in water (25 mL)
and stirred at room temperature for 30 min. Concentrated aqueous
ammonia (100 mL) was added, and it was stirred for another 5 h. The
solvent was again evaporated at room temperature and the residue
purified by ion-exchange chromatography on DEAE Sephadex using
a gradient of 0.1M-0.5M NEt3HCO3. The UV absorbing fractions were
evaporated at 20 °C, and the product was stored at -20 °C. Reversed-
phase HPLC purification of the triphosphate was done at a flow rate
of 5.1 mL/min using 25 mM (Et3NH)OAc (pH 7.0) (solvent A) and
20% (v/v) MeCN in solvent A (solvent B) in the following linear binary
gradient: 100% A (1 min); 100% A to 90% A (39 min); retention
time 33 min. The concentration of dJTP was estimated using an
extinction coefficient (ꢀ290 ) 9000 M-1cm-1) consistent with measured
values for analogously substituted compounds.22 A high-resolution mass
spectrum was obtained showing an exact mass of 519.9885, corre-
sponding to a calculated exact mass of 519.9923. 31P NMR (D2O):
-1.97, -2.13 (d, J ) 19.18 Hz), -6.33, -6.48 (d, J ) 19.18 Hz),
1
DMF. Rf: 0.42 (chloroform/MeOH 8.25:1.75). H NMR (DMSO-d6):
2.13 (m, 2H, 2′), 3.60 (m, 2H, 5′), 3.81 (m, 1H, 4′), 4.24 (m, 3H, H-9,
3′), 5.12 (t, 1H, 5′-OH), 5.27 (d, 1H, 3′-OH), 6.11 (t, 1H, 1′), 8.22 (s,
1H, H-6), 10.09 (t, 1H, NH chain), 11.67 (s, 1H, NH cycl.). 13C NMR
(DMSO-d6): 29.5 (C-9), 40.7 (2′), 61.0 (5′), 70.2 (3′), 75.4 (C-8), 84.8
(1′), 87.5, 87.7 (4′, C-7), 97.7 (C-5), 113.9, 117.8 (q, CF3, J ) 287.95
Hz), 144.2 (C-6), 149.5 (C-2), 155.9, 156.4 (q, COCF3, J ) 37.25 Hz),
161.7 (C-4).
5-(3′′-Trifluoroacetamidopropynyl)-5′-O-dimethoxytrityl-2′-deox-
yuridine. The product from above (1.34 mmol, 506 mg) was coevapo-
rated with pyridine, then dissolved in dry pyridine (10 mL) and cooled
to 0 °C. Et3N (2 equiv, 2.68 mmol, 271 mg, 0.373 mL), DMAP (0.25
equiv, 0.336 mol, 41 mg), and 4,4′-dimethoxytrityl chloride (1.2 equiv,
1.61 mmol, 545 mg) were added, and the mixture was stirred at 0 °C
for 5 min and then at room temperature for 4 h. TLC (chloroform/
10% MeOH, Rf ) 0.47) did not show any starting material. MeOH (2
mL) was added and the mixture evaporated. The residue was extracted
(ethyl acetate/aqueous NaHCO3 solution), the combined organic layers
were washed with water and dried (Na2SO4), and the solvent was
evaporated. The residue was purified by silica column chromatography
(chloroform/10% MeOH) to give 902 mg (99%) of product as a yellow
1
foam. H NMR (CDCl3): 2.44-2.61 (m, 2H, 2′), 3.34 (m, 2H, 5′),
3.73 (s, 6H, MeO), 3.89 (m, 1H, 4′), 4.14 (m, 2H, H-9), 4.59 (m, 1H,
3′), 6.34 (t, 1H, 1′), 6.80 (m, 4H, DMT), 7.14-7.33, 7.61-7.70 (m,
9H, DMT), 8.21 (s, 1H, H-6). 13C NMR (CDCl3): 30.3 (C-9), 41.6
(2′), 55.2 (MeO), 63.5 (5′), 72.0 (3′), 75.3 (C-8), 86.0 (1′), 86.9 (Cq
trityl), 87.0, (C-7), 87.3 (4′), 99.9 (C-5), 113.3 (DMT), 113.8, 117.6
(q, CF3, J ) 286.52 Hz), 126.9, 127.8, 128.0, 129.9, 135.4 (all DMT),
143.6 (C-6), 144.5 (DMT), 149.4 (C-2), 156.4, 156.9 (q, COCF3, J )
37.70 Hz), 158.5 (DMT), 162.6 (C-4).
1
-16.80 (dt). H NMR (D2O): 1.24 (t, 12H, Et3N), 2.40 (m, 2H, 2′),
3.18 (q, 8H, Et3N), 4.06 (m, 2H, 5′), 4.19 (m, 2H, H-9), 4.25 (m, 1H,
4′), 4.61 (m, 1H, 3′), 6.21 (t, 1H, 1′), 8.31 (s, 1H, H-6). 13C NMR
(D2O): 6.8 (Et3N), 28.2 (C-9), 37.9 (2′), 45.0 (Et3N), 63.4 (5′), 68.1
(3′), 76.4 (C-8), 83.9, 84.1, 84.4 (1′, 4′, C-7), 96.5 (C-5), 144.4 (C-6),
148.7 (C-2), 162.5 (C-4). UV (H2O, pH 7) λmax 230, 290 nm. MS (ESI+)
522 [M + H]+, 544 [M + Na]+.
3′-O-Acetyl-5-(3′′-trifluoroacetamidopropynyl)-5′-O-dimethox-
ytrityl-2′-deoxyuridine. The product from above (1.33 mmol, 902 mg)
was coevaporated with pyridine and then dissolved in pyridine (10 mL),
and DMAP (0.25 equiv, 0.332 mmol, 40.5 mg), Et3N (2.5 equiv, 3.32
mmol, 335 mg, 0.462 mL), and Ac2O (1.2 equiv, 1.60 mmol, 163 mg,
0.15 mL) were added. It was stirred at room temperature for 2.5 h
(TLC: chloroform/10% MeOH, Rf ) 0.75); then MeOH (5 mL) was
added to stop the reaction and it was evaporated to dryness. The residue
was extracted (water/ethyl acetate), the organic layer was dried (Na2-
SO4), and the solvent was evaporated. The residue was purified by
silica column chromatography (chloroform/10% MeOH) to give 937
DNA Reagents and Materials. ThermoPol 1× buffer and Vent
DNA polymerase were purchased from New England Biolabs, standard
dNTPs and Taq DNA polymerase from Promega, Microspin S-300 HR
columns from Pharmacia, streptavidin-agarose from Fluka, and AMP-,
ADP-, and ATP-agarose (C-8 attachment) from Sigma. The random-
sequence DNA library was synthesized by Microsynth (Balgach,
Switzerland); all primers, templates, and 3′-fluorescein-tagged Se-
quences A and B were from Integrated DNA Technologies (Coralville,
IA). Oligonucleotides were purified by polyacrylamide gel electro-
phoresis (PAGE).
1
mg (98%) of product as a yellow foam. H NMR (CDCl3): 2.08 (s,
DNA Pools. A library of DNA containing the four natural bases in
equal proportion and having a 70 base random region flanked by two
defined primer binding sites of 22 and 20 bases (Figure 3) was
synthesized (200 nmol scale). The library was purified (denaturing 8%
PAGE), yielding 3.2 nmol of oligonucleotide. Approximately 10% of
the DNA was able to be PCR amplified, giving ∼2 × 1014 different
molecules. All 120 µg of DNA was PCR amplified in a reaction volume
of 50 mL (10 mM KCl; 10 mM (NH4)2SO4; 20 mM Tris‚HCl, pH 8.8;
2 mM MgSO4; 0.1% Triton X-100 (1× ThermoPol reaction buffer);
0.2 mM dNTPs; 0.5 µM primers; 20 units/mL Vent DNA polymerase)
3H, Ac), 2.38-2.64 (m, 2H, 2′), 3.42 (m, 2H, 5′), 3.77 (s, 6H, MeO),
3.96 (m, 1H, 4′), 4.18 (m, 2H, H-9), 5.45 (m, 1H, 3′), 6.34 (t, 1H, 1′),
6.85 (m, 4H, DMT), 7.23-7.48, 7.63-7.71 (m, 9H, DMT), 8.20 (s,
1H, H-6).13C NMR (CDCl3): 20.8 (Ac), 30.2 (C-9), 38.7 (2′), 55.1
(MeO), 63.5 (5′), 74.9 (3′), 75.1 (C-8), 84.4 (1′), 85.3 (4′), 87.1 (Cq
trityl), 87.2 (C-7), 99.4 (C-5), 113.3 (DMT), 113.6, 117.4 (q, CF3, J )
287.42 Hz), 126.9, 127.7, 128.0, 129.9, 135.2 (all DMT), 143.1 (C-6),
144.3 (DMT), 149.4 (C-2), 156.3, 156.8 (q, COCF3, J ) 37.78 Hz),
158.6 (DMT), 162.0 (C-4), 170.3 (Ac).
3′-O-Acetyl-5-(3′-trifluoroacetamidopropynyl)-2′-deoxyuridine.
The compound obtained above (1.30 mmol, 937 mg) was dissolved in
(22) Robins, M. J.; Barr, P. J. J. Org. Chem. 1983, 48, 1854-1862.

Receptors for Adenosine Nucleotides
J. Am. Chem. Soc., Vol. 121, No. 42, 1999 9789
for six cycles (temperature cycle conditions: 94 °C, 4 min; 55 °C, 7
min; 72 °C, 7 min). The PCR product was extracted with phenol/CHCl3,
ethanol precipitated, redissolved in 1 mL of TE (10 mM Tris‚HCl, pH
7.6; 1 mM EDTA), and purified using Microspin S-300 HR columns.
Biotinylated primer to initiate complementary strand synthesis and
functionalized nucleoside dJ were introduced to create a pool of dJ-
containing biotinylated DNA (dJ pool) by PCR amplifying 22 µg of
DNA (0.3 nmol) in a volume of 30 mL (1× ThermoPol reaction buffer;
0.2 mM dATP, dCTP, and dGTP; 0.2 mM dJTP; 0.5 µM biotinylated
primer and standard primer; 20 units/mL Vent DNA polymerase) over
10 cycles (temperature cycle conditions as above). A biotinylated pool
made of the four naturally occurring nucleosides (standard pool) was
amplified with the biotinylated primer and the four standard dNTPs
over six cycles. After phenol/CHCl3 extraction, each pool was redis-
solved in TE (200 µL). Analogous PCR amplifications were done for
each pool (0.5 mL) with 20 µCi R-32P-dATP and combined with the
larger amplification, to have a small portion of radiolabeled DNA.
A streptavidin-agarose column (0.5 mL), which had been equilibrated
in wash buffer (50 mM NaCl; 10 mM Tris‚HCl, pH 7.5; 1 mM EDTA),
was used to make the DNA single-stranded. The DNA was applied to
the column and incubated (5 min, rt), and then the column was washed
with 10 volumes of wash buffer. The streptavidin-agarose was
suspended in 0.5 M NaOH (1.5 mL), removed from column, and
incubated (37 °C, 15 min). The suspension was returned to the column,
and the NaOH solution containing the single-stranded DNA was eluted
and collected. The DNA was ethanol precipitated and redissolved in
water (150 µL).
Selection. The pools were preselected to minimize enrichment of
DNA molecules interacting with column material.23 A preselection
agarose gel was prepared from 6-aminohexanoic acid N-hydroxysuc-
cinimide ester-sepharose 4B. The gel was washed with 1 mM HCl (200
mL/g) for 1 h and then incubated (5 min, rt) in 0.5 M NaCl; 0.1 M
Tris‚HCl, pH 8.0 solution. The gel was washed with several volumes
of 0.1M NaOAc, pH 4.0; 0.5 M Na Cl and then with 0.1 M Tris‚HCl,
pH 8.0; 0.5 M Na Cl. The resulting preselection agarose was stored in
1 mM EDTA, pH 8.0 at 4 °C.
All permutations of selections against AMP-, ADP-, or ATP-agarose
using the standard pool or the dJ pool were done (totaling six separate
selection experiments). The adenosine phosphate immobilized on
agarose was washed with several volumes of water and stored in 1
mM EDTA, pH 8.0, at 4 °C. Preselection agarose (0.2 mL) and
adenosine phosphate agarose (0.5 mL) were equilibrated with 20
volumes of selection buffer (250 mM Na Cl; 50 mM Tris‚HCl, pH
7.6; 5 mM MgCl2). Single-stranded DNA (33 µg, ∼1 nmol), from the
combination of the 30 mL PCR product with the analogous radiolabeled
0.5 mL PCR product, was denatured in selection buffer (95 °C, 3 min)
and allowed to cool (15-20 min). The DNA was applied to the
preselection column, eluted with 2.5 volumes of selection buffer, and
applied directly to the selection (immobilized adenosine phosphate)
column. The selection column was washed with 5 volumes of selection
buffer. Interacting molecules were eluted with 3 volumes of selection
buffer containing the appropriate adenosine phosphate at 1.5× the
concentration indicated by the manufacturer of the AMP-, ADP-, or
ATP-agarose. The amount of adenosine phosphate on the agarose varied
between batches (1.3-4.5 mM).
selection experiment, and for the dJ pool, 15 clones were sequenced.
The randomized region of the dJ-containing sequences was shortened
during the selection. This result suggests biases against longer oligo-
nucleotide receptors and dJ, although these results do not permit a
conclusion as to whether these biases arise in the selection or the
amplification steps.
Fluorescently Labeled Sequence 409. Sequence 409 with T
replacing each dJ was used as a template to produce Sequence 409 in
parallel PCR amplifications (20 × 0.1 mL reaction volumes) with dJTP
and a biotinylated primer. Each 0.1 mL PCR reaction (1× ThermoPol
reaction buffer; 0.2mM dATP, dCTP, and dGTP; 0.2 mM dJTP; 0.5
µM biotinylated primer and standard primer; 20 units/mL Vent DNA
polymerase; 1 nM template) was amplified over 12 cycles (temperature
cycle conditions: 94 °C, 1 min; 55 °C, 2 min; 72 °C, 2 min). The
PCR products were pooled, phenol/CHCl3 extracted twice, the aqueous
phase was ethanol precipitated, and the resulting pellet was redissolved
in TE (250 µL). The redissolved PCR product was made single-stranded
using streptavidin-agarose, as described for the DNA pools. The
oligonucleotide was ethanol precipitated with 3 µg of glycogen as a
carrier.
Sequence 409 was fluorescently labeled at the 3′ end with terminal
transferase (Boehringer Mannheim) and fluorescein-12-2′,3′-dideox-
yuridine 5′-triphoshate (Boehringer Mannheim). Sequence 409 (0.2
nmol) was dissolved in 45 µL of labeling buffer (200 mM potassium
cacodylate; 25 mM Tris‚HCl, pH 6.6; 0.25 mg/mL bovine serum
albumin) containing cobalt chloride (5 mM), fluorescein-12-ddUTP (40
µmol), and terminal transferase (0.6 units/µL). The reaction mixture
was incubated (1 h, 37 °C) and phenol/CHCl3 extracted twice, and the
aqueous phase was ethanol precipitated with glycogen as a carrier.
Kd Determinations. Affinity capillary electrophoresis was performed
using a Beckman P/ACE 2200 automated capillary electrophoresis
system with laser-induced fluorescence (LIF) detection (Beckman-
Coulter, Fullerton, CA). The 488 nm line (3 mW) of the Ar+ laser was
used for excitation, and fluorescence was collected at 520 nm.
Separations were performed using 50 µm i.d. capillaries (Polymicro
Technologies, Phoenix, AZ). All capillaries were coated with poly-
acrylamide using a method similar to that of Doln´ık et al.24 The
acrylamide solution used in preparing the capillary was 3 mL of aqueous
acrylamide (10% w/v; Sigma), 7 mL of water, 100 µL of 10% aqueous
ammonium persulfate (Sigma), and 10 µL of tetramethylethylenedi-
amine (Sigma). Capillaries were stored in tris buffer (50 mM Tris‚
HCl, pH 7.6; 50 mM sodium chloride; 50 mM magnesium chloride)
when not in use.
DNA samples (∼200 nM) were prepared daily in tris buffer. Each
DNA sample was heated at 88-90 °C for 2.5 min and then cooled to
room temperature before beginning experiments. A 30-50 mM ATP
(Sigma) stock was made in tris buffer, and the pH was adjusted to 6.5
with 1 M sodium hydroxide (Fisher). Electrophoresis buffers were made
from this stock in concentrations ranging from 500 µM to 10 mM ATP.
The capillary was rinsed with the electrophoresis buffer for 3 min before
each 5 s hydrodynamic injection. The separation potential was 10 kV
in a 37 cm capillary. The length to the detector was 30 cm. Data were
collected with Beckman P/ACE software version 2.0 using a 386 PC
computer. Binding constants were determined by nonlinear least-squares
regression using the variance-covariance method.
The eluted DNA was ethanol precipitated in the presence of 100 µg
of glycogen and redissolved in water (50 µL). The precipitated DNA
was PCR amplified (0.5 mL reaction volume) in the presence of 20
µCi R-32P-dATP, as described above, to generate a biotinylated pool.
Generally, 10-14 cycles of PCR were necessary to get a sufficient
amount of DNA. The PCR product was extracted with phenol/CHCl3,
ethanol precipitated, and made single-stranded with streptavidin-agarose
as above. Half of this single-stranded DNA was stored at -20 °C, and
the remaining half was used for the next round of selection. The iterative
process was repeated with another selection column for 8 further rounds.
Cloning and Sequencing. After the ninth round of selection, the
eluted DNA was amplified in a 0.3 mL reaction volume (1× Taq
polymerase buffer; 0.2 mM dNTPs; 1.0 µM primers; 6 U Taq DNA
polymerase) for 10 cycles. The resulting DNA with T replacing dJ was
cloned into the vector pCR 2.1-TOPO using the TOPO TA Cloning
Kit (Invitrogen). For the standard pool, 10 clones were sequenced per
Acknowledgment. The authors acknowledge support from
Office of Naval Research (N00025-96-1-0362 to S.A.B.), the
National Institutes of Health (GM 54048 to S.A.B.), and the
National Science Foundation (CHE 9531428).
Supporting Information Available: Additional experimen-
tal procedures and data (PDF). This material is available free
of charge via the Internet at http://pubs.acs.org.
JA9816436
(24) Doln´ık, V.; Xu, D.; Yadav, A.; Bashkin, J.; Marsh, M.; Tu, O.;
Mansfield, E.; Vainer, M.; Madabhushi, R.; Barker, D.; Harris, D. J.
Microcolumn Sep. 1998, 10, 175-184.
(23) Famulok, M. J. Am. Chem. Soc. 1994, 116, 1698-1706.