Engineering a selective small-molecule substrate binding site into a deoxyribozyme

Article abstract of DOI:10.1002/anie.200702217

A small goal: An RNA ligase deoxyribozyme is engineered to accept a small-molecule NTP substrate in a multiple-turnover fashion. Selective binding is enforced by hydrogen bonding, and structural preorganization within the NTP itself is important for its efficient utilization as a substrate. This study points the way toward a broader use of small-molecule substrates with nucleic acid enzymes. (Figure Presented)

Full text of DOI:10.1002/anie.200702217

Communications
DOI: 10.1002/anie.200702217
Catalytic DNA
Engineering a Selective Small-Molecule Substrate Binding Site into a
Deoxyribozyme**
Claudia Höbartner and Scott K. Silverman*
Binding of small molecules by nucleic acids contributes to
numerous biological processes. For example, riboswitches
bind a variety of metabolites to control gene expression,[1] and
ribozymes such as the group I intron use an obligatory
guanosine cofactor as the nucleophile in the first step of
splicing.[2] In several cases, artificial nucleic acid aptamers[3]
and enzymes[4] (collectively, functional nucleic acids
(FNAs)[5]) have been identified that interact with small
molecules.[6,7] In general, understanding the interactions
between nucleic acids and small molecules is important for
identifying useful FNAs and for rationally extending the
scope of their functions.
We recently reported the 7S11 deoxyribozyme, which
creates 2’,5’-branched RNA by mediating the attack of an
RNA 2’-hydroxy-group nucleophile at a 5’-triphosphate
electrophile.[8,9] Both 7S11 and a more general variant,
10DM24,[10] have
a three-helix-junction architecture in
which the two RNA oligonucleotide substrates and the
deoxyribozyme interact through extensive Watson–Crick
base pairing (Figure 1a). The 5’-triphosphorylated guanosine
electrophile is presented to the branch-site-adenosine nucle-
ophile, which is held at the terminus of the P4 (paired region
P4) RNA:DNA helix by Watson–Crick hydrogen bonds. The
ligation product is a 2’,5’-branched RNA. Because many
natural ribozymes have multi-helix-junction architectures,[11]
we are interested to determine the generality of helix-junction
platforms for rational deoxyribozyme development. Towards
this goal, herein we successfully engineered the 10DM24
deoxyribozyme to mediate the multiple-turnover ligation
reaction of a small-molecule nucleoside triphosphate (NTP)
rather than a 5’-triphosphorylated oligonucleotide as an
electrophilic substrate. We also examined in detail the
requirements for productive substrate binding. Hydrogen
bonding contributes substantially to selective binding of the
NTP, and structural preorganization within the substrate is
important for its efficient utilization.
Figure 1. The 10DM24 deoxyribozyme and the use of a small-molecule
substrate. a) Secondary structure and schematic three-helix-junction
tertiary structure of 10DM24. The detailed three-dimensional structure
of the deoxyribozyme-substrate complex is not known. b) Engineering
into 10DM24 of a well-defined binding site for GTP as a small-
molecule substrate.
We began our experiments with the original 10DM24
arrangement shown in Figure 1a. Conceptually breaking the
right-hand (R) oligonucleotide substrate shown in blue
immediately to the 3’ side of the first nucleotide leads, in
principle, to a deoxyribozyme–substrate complex in which
guanosine 5’-triphosphate (GTP) can bind as a discrete
electrophile in the location corresponding to the 5’-terminal
position of the P4 helix (Figure 1b). We experimentally tested
the ability of 10DM24 to catalyze ligation according to this
design by using free GTP as a substrate, thereby transferring
guanosine 5’-monophosphate (GMP) to the branch-site-
adenosine 2’-hydroxy group. When RD, an oligoribonucleotide
cofactor that corresponds to all of the remaining nucleotides
[*] Dr. C. Höbartner, Prof. S. K. Silverman
Department of Chemistry
University of Illinois at Urbana-Champaign
600 South Mathews Avenue, Urbana, IL 61801 (USA)
Fax: (+1)217-244-8024
E-mail: scott@scs.uiuc.edu
Homepage: http://www.scs.uiuc.edu/skslab.html
[**] This research was supported by the National Institutes of Health
(GM-65966 to S.K.S.) and the Austrian Science Fund (Erwin
Schrödinger fellowship to C.H.). S.K.S. is a Fellow of the David and
Lucile Packard Foundation. We thank Chandra Miduturu and Gerald
Charleston for discussions.
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
7420
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 7420 –7424

Angewandte
Chemie
of R, was added to 10DM24 along with GTP, ligation was
efficient (Figure 2; 94% yield in 5 h and kobs = 0.034 minÀ1
under the standard incubation conditions of 1 mm GTP and
40 mm MgCl2 at pH 9.0 and 378C). When RD was omitted, no
ATP) and their derivatives. When 1 mm ATP was provided as
a small-molecule substrate in place of GTP by using the
original 10DM24 sequence and RD, no reaction was observed
(< 1% in 5 h). However, when the corresponding deoxyribo-
zyme nucleotide was changed from C to T, substantial ligation
was observed with ATP (33% yield in 5 h and kobs
=
0.0008 minÀ1 under standard conditions; Figure 2) but no
longer with GTP (< 1% in 5 h). Furthermore, the kobs value
increased 16-fold to 0.013 minÀ1 under the enhanced condi-
tions with 10 mm ATP (82% yield in 3 h).[13] These data are as
expected for a Watson–Crick base pair between the deoxy-
ribozyme and the NTP substrate.
We varied the number of hydrogen bonds between the
NTP substrate and the deoxyribozyme, anticipating that this
could influence the efficiency of the ligation reaction. Indeed,
the ligation yield and rate increased when 2,6-diaminopurine
ribonucleoside triphosphate (DTP) rather than ATP was
paired with T in the DNA (Figure 2; 68% in 5 h and kobs
=
0.0032min À1 under standard conditions; 90% in 3 h and kobs
=
0.027 minÀ1 under enhanced conditions).[13] In contrast, when
the original C in the deoxyribozyme was retained and inosine
triphosphate (ITP) rather than GTP was provided as the
substrate, a decrease in activity was observed (25% in 5 h and
k
k
obs = 0.0007 minÀ1 under standard conditions; 84% in 3 h and
obs = 0.014 minÀ1 under enhanced conditions).[13] Therefore,
three hydrogen bonds (GTP, DTP) rather than two hydrogen
bonds (ITP, ATP) lead to better activity (solid versus dashed
lines in Figure 2c). Perhaps surprisingly, replacing the ade-
nine nucleobase of the substrate with 2-aminopurine led to a
tenfold decrease in the kobs value,[13] even though both purine
derivatives can form two hydrogen bonds with T in the
deoxyribozyme.[15] Therefore, in terms of contribution to NTP
substrate binding, the hydrogen bond facing the major groove
is more important than the hydrogen bond facing the minor
groove.
Figure 2. Reaction of a small-molecule NTP substrate catalyzed by the
10DM24 deoxyribozyme. Successful ligation is observed only when the
NTP substrate has Watson–Crick complementarity to the terminal P4
DNA nucleotide of 10DM24. a) Watson–Crick interactions between the
NTP substrate (top) and the terminal P4 DNA nucleotide of 10DM24
(bottom). b) PAGE images showing the 5-h time point after the
ligation reactions were performed with 1 mm of the indicated NTP
(– indicates no NTP), 40 mm MgCl2, pH 9.0, and 378C. The ligation
products are marked with an arrowhead. c) Kinetic plots for Watson–
In addition to the Watson–Crick base pair involving the
NTP substrate, the structural model shown in Figure 1b also
depicts a base pair at the second position of the P4 helix. Our
previous data strongly suggested the presence of this second
base pair when a full-length R substrate is used (as in
Figure 1a). This was confirmed here with more comprehen-
sive data.[13] We then established that this second base pair is
required when using GTP as the substrate with the RD
cofactor; furthermore, alteration of the base-pair identity is
tolerated with only moderate changes in activity.[13,16] These
data, along with the Watson–Crick covariation involving the
NTP substrate itself, provide compelling support for the
binding model depicted in Figure 1b.[17] To place this Watson–
Crick binding mode in context, the other artificial aptamers
and nucleic acid enzymes that interact with NTP substrates
generally do so through non-Watson–Crick interactions
(where the interaction mode is known) with mm to mm
binding constants.[6,7] In contrast, the natural purine-binding
riboswitches bind their cognate nucleobase through Watson–
Crick interactions.[18] In the latter cases, the nucleobase
ligands are completely engulfed by the RNA, which enables
quite low (nm) dissociation constants. A Watson–Crick bind-
ing mode is also observed for the preQ1 riboswitch, which has
nm affinity for its ligand.[19] It should be noted that not all
!
*
Crick combinations. The points are (top to bottom) GTP , DTP
,
~
&
ATP , and ITP . The solid lines denote reactions of NTPs that form
three Watson–Crick hydrogen bonds with the deoxyribozyme, whereas
the dashed lines denote reactions of NTPs that form only two Watson–
Crick hydrogen bonds.
reaction was observed.[12,13] The identity of the PAGE-
purified product was confirmed by partial alkaline hydrolysis
and MALDI mass spectrometry.[13] From a plot of the kobs
value versus the GTP concentration, the Kd,app value (appar-
ent dissociation constant) for GTP was found to be greater
than 1 mm. The kobs value increased eightfold to 0.26 minÀ1
under the enhanced incubation conditions of 10 mm GTP and
150 mm MgCl2 at pH 9.0 and 378C (94% yield in 3 h).
We then examined the generality of the ligation reaction
by using other NTP substrates in place of GTP. The analogous
reaction with the full-length R oligonucleotide as the
substrate proceeds well when a 5’-terminal G is present and
with only a fivefold lower kobs value than with 5’-A, whereas
5’-C supports greatly reduced activity and 5’-U leads to almost
no product (in all cases, the corresponding deoxyribozyme
nucleotide is changed to maintain Watson–Crick complemen-
tarity).[14] We therefore focused on the purine NTPs (GTP and
Angew. Chem. Int. Ed. 2007, 46, 7420 –7424
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7421

Communications
biologically relevant interactions between RNA and sub-
had the expected connectivity, as confirmed by partial
alkaline hydrolysis (Figure 3c). By design, the nucleobase
and triphosphate (i.e., recognition and reactive) moieties of
GclvTP and GacvTP are not structurally constrained by the five-
membered ribose ring that is present within GTP itself.
Therefore, the poor reactivities of these two modified
substrates demonstrate that the preorganization enforced by
the ribose ring of GTP contributes substantially to the
efficiency of the deoxyribozyme-catalyzed ligation reaction.
All of our previous deoxyribozyme-mediated ligation
reactions with two RNA oligonucleotide substrates had
displayed only single-turnover ligation behavior, which was
attributed to product inhibition (similar to natural protein
enzymes that ligate nucleic acids).[23] In contrast, upon the
10DM24-catalyzed reaction of the oligoribonucleotide 2’-
hydroxy group with the NTP substrate, the binding affinity of
the RNA for the deoxyribozyme is not expected to increase
substantially. On this basis, we anticipated that multiple
turnover should be observable in the engineered deoxyribo-
zyme–NTP system. Indeed, with GTP as the substrate, we
were for the first time able to observe unambiguous multiple-
turnover behavior by using an RNA ligase deoxyribozyme
(five turnovers were observed in 5 h).[13]
Finally, we showed that a binding site for an NTP cofactor
can be located adjacent to the substrate binding site. This was
achieved by removing an additional nucleotide from the RD
cofactor (blue in Figure 1b), forming the shorter RDD cofactor
and requiring two added nucleotides to reconstitute the
complete P4 region (Figure 4a). When GTP was incubated
with the parent 10DM24 deoxyribozyme (which has 3’-CCTT-
5’ in its P4 region) along with the RDD cofactor and the
oligoribonucleotide that provides the 2’-hydroxy group, a
product band matching the branched standard was observed.
Additionally, a small amount of a more slowly migrating
product was also observed (Figure 4b, top section, lane 3).
Partial alkaline hydrolysis of both products revealed that they
are branched with respective connectivities A–G and A–GG
(where A is the branch-site adenosine).[13] We hypothesized
that the new A–GG product is formed by initial templated but
otherwise uncatalyzed synthesis of a GG dinucleotide (i.e.,
pppGpG) from two GTP molecules followed by 10DM24-
catalyzed branch formation by using this dinucleotide.[24]
Consistent with this hypothesis, the purified A–G product
was unreactive with GTP when resubjected to the reaction
conditions,[13] which suggests that the two G nucleotides are
not attached successively. The pppGpG dinucleotide was
synthesized independently by using T7 RNA polymerase[6,13]
and, as expected, led solely to the A–GG product (Figure 4b,
top section, lane 4; also as expected, the RDD cofactor was
required to observe this product[13]). Although the pppGpG
strates are high affinity; for example, the glmS riboswitch
binds glucosamine 6-phosphate (GlcN6P) with a Kd,app value
of merely 0.2m m.[20]
We evaluated changes to the small-molecule NTP sub-
strates that probe the role of the ribose ring, including
potential effects of structural preorganization.[13] Both 2’-
deoxyGTP (dGTP) and 2’,3’-dideoxyGTP (ddGTP) are
tolerated well, with no diminution of yield and at most a
threefold decrease in the kobs value relative to GTP. Similarly,
arabino-ATP (which has the opposite 2’-configuration rela-
tive to ATP), dATP, and ddATP all have a kobs value within
twofold of ATP itself. From these data, we conclude that the
deoxyribozyme does not require the 2’- or 3’-hydroxy groups,
nor does it directly contact either the 2’- or 3’-hydrogens of the
ribose ring. We additionally considered how perturbations in
the structural preorganization of the substrate impact the
ligation activity by using two substrate analogues (Figure 3a).
clv
À
First, in place of GTP we used C2 C3-cleaved GTP (G TP),
À
which lacks the C2 C3 bond of the ribose ring but has the
same number of heavy (non-hydrogen) atoms.[21] Second, in
place of GTP we used acyclovir triphosphate (GacvTP), where
acyclovir is the guanosine analogue that lacks both the C2and
C3 carbons and hydroxy groups of the ribose ring.[22] For both
GclvTP and GacvTP, only a very small amount of ligation
activity was observed (Figure 3b); the kobs value was dimin-
ished relative to GTP by approximately 1000-fold (GclvTP) or
300-fold (GacvTP).[13] The products were isolated by PAGE; all
substrate had a Kd,app value of greater than 1 mm with RDD
,
similar to the Kd,app value for GTP with RD,[13] the ligation
reaction with pppGpG and RDD had a kobs value that was
sixfold higher than for the analogous reaction with GTP and
RD (Figure 4c).
When similar experiments were performed by using the
mutant 10DM24 deoxyribozyme that has CTTT rather than
CCTT in the P4 region, the A–GG product was not observed
when using GTP (Figure 4b, bottom section, lane 3). How-
Figure 3. Evaluation of the role of structural preorganization within the
substrate. a) Structures of GTP, cleaved GTP (GclvTP), and acyclovir
triphosphate (GacvTP). b) Ligation assays with GTP and the two
analogues. Data were obtained with 10 mm NTP (or analogue),
150 mm MgCl2, pH 9.0, and 378C. The ligation products are marked
with an arrowhead. c) Partial alkaline hydrolysis demonstrates the
expected connectivity for each product. The branch-site adenosine is
underlined. The partial alkaline hydrolysis (HOÀ) ladder was assigned
by comparison with the RNase T1 cleavage ladder (T1).
7422
www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 7420 –7424

Angewandte
Chemie
labeling of the RNA 3’ terminus.[25] Therefore, deoxyribo-
zyme-catalyzed internal attachment to RNA of a small-
molecule NTP substrate will likely form the basis of a succinct
and nonperturbing site-specific RNA modification strategy.[27]
It should also be possible to use covalently modified NTPs
directly as ligation substrates. This may require selection in
vitro of new deoxyribozymes that accept such substrates.
Received: May 20, 2007
Published online: August 10, 2007
Keywords: deoxyribozymes · hydrogen bonds · nucleic acids ·
.
receptors · RNA recognition
[1] M. Mandal, R. R. Breaker, Nat. Rev. Mol. Cell Biol. 2004, 5,
451 – 463.
[2] P. L. Adams, M. R. Stahley, A. B. Kosek, J. Wang, S. A. Strobel,
Nature 2004, 430, 45 – 50.
[3] T. Hermann, D. J. Patel, Science 2000, 287, 820 – 825.
[4] G. F. Joyce, Annu. Rev. Biochem. 2004, 73, 791 – 836.
[5] N. K. Navani, Y. Li, Curr. Opin. Chem. Biol. 2006, 10, 272 – 281.
[6] F. Huang, Z. Yang, M. Yarus, Chem. Biol. 1998, 5, 669 – 678.
[7] a) Y. Li, Y. Liu, R. R. Breaker, Biochemistry 2000, 39, 3106 –
3114; b) Y. Li, R. R. Breaker, Proc. Natl. Acad. Sci. USA 1999,
96, 2746 – 2751; c) W. Wang, L. P. Billen, Y. Li, Chem. Biol. 2002,
9, 507 – 517.
Figure 4. Use of a second NTP as a cofactor for the ligation reaction.
a) Schematic depiction; compare with Figure 1b. b) The shortened P4-
region sequence is depicted adjacent to each set of gel lanes (the
asterisk in the bottom section denotes the single altered DNA
nucleotide). The standard (lane 1) is the branched product made by
using GTP as the substrate with RD (see Figure 2b, G–C lane). The 5-h
time points were obtained at pH 9.0 and 378C. In lane 3, 20 mm GTP
and 150 mm MgCl2 were also included. In lane 4, 1 mm pppGpG and
40 mm MgCl2 were also included. The branched products are assigned
as A–G (filled arrowhead) and A–GG (open arrowhead) and the
branch-site adenosine is underlined. c) Comparison of kobs values for
GTP or pppGpG as the substrate when aligned in several combina-
tions.
[8] a) R. L. Coppins, S. K. Silverman, Nat. Struct. Mol. Biol. 2004,
11, 270 – 274; b) R. L. Coppins, S. K. Silverman, Biochemistry
2005, 44, 13439 – 13446.
[9] R. L. Coppins, S. K. Silverman, J. Am. Chem. Soc. 2005, 127,
2900 – 2907.
[10] E. Zelin, Y. Wang, S. K. Silverman, Biochemistry 2006, 45, 2767 –
2771.
[11] D. M. J. Lilley, Q. Rev. Biophys. 2000, 33, 109 – 159.
[12] The phosphorylation state of the 5’ terminus of the RD cofactor
could be varied (5’-triphosphate, 5’-monophosphate, or 5’-OH)
without changing the kobs value by more than fourfold. All
experiments in this manuscript used the 5’-monophosphorylated
RD cofactor.
[13] See the Supporting Information for all data that is not shown in
the figures.
[14] This trend was reported for 7S11 by using 5’-adenylated
substrates,[9] and a similar trend is observed with 10DM24 (see
the Supporting Information).
ever, the separately prepared pppGpG dinucleotide still led
to the A–GG product (Figure 4b, bottom section, lane 4), but
with a 140-fold lower efficiency.[13] Although a G–C base pair
is preferred, G can occupy the second position of the P4 helix
regardless of the deoxyribozyme nucleotide across from it.[13]
The data with the mutant deoxyribozyme suggest that binding
at the second P4 position of GTP in the G–Twobble geometry
strongly disfavors templated GG dinucleotide synthesis, but
still permits slow reaction of the branch-site adenosine 2’-
hydroxy group with the GTP molecule bound as the substrate
at the first P4 position.
In summary, we have shown that the three-helix-junction
architecture of the 10DM24 deoxyribozyme fosters rational
engineering of a selective binding site for a small-molecule
NTP substrate that reacts in multiple-turnover fashion. In
analogy to the natural purine-binding riboswitches, the
selectivity of 10DM24 for its NTP substrate is enforced by
Watson–Crick hydrogen bonding. Structural preorganization
within the small-molecule substrate is important for catalytic
activity of the deoxyribozyme, and a cofactor binding site can
be introduced adjacent to the substrate binding site. These
results establish a detailed baseline for further rational
approaches to identify and improve the functions of nucleic
acid enzymes. Utilization of an NTP substrate by the 10DM24
deoxyribozyme may also have practical value. For example,
the nucleotide that is attached to the RNA substrate by the
10DM24 deoxyribozyme has a 2’,3’-diol moiety that is
susceptible to periodate oxidation and subsequent derivati-
zation, which is analogous to long-established approaches for
[15] For practical reasons, these experiments were performed with
the 2’-deoxy-NTPs (i.e., dATP and d2AP-TP, where 2AP is 2-
aminopurine). Because dATP is almost as efficient a substrate as
ATP (see below), the 2’-deoxy modification of d2AP-TP is not
responsible for its poor reactivity as a substrate relative to dATP.
[16] The identity of the second base pair could in principle influence
the NTP binding affinity through stacking or other effects.
[17] A modest amount of ligation activity was observed when GTP or
ITP was used along with A in the deoxyribozyme (see the
Supporting Information). This indicates that non-Watson–Crick
interactions between the NTP substrate and deoxyribozyme are
possible, albeit much less productive than Watson–Crick inter-
actions.
[18] a) R. T. Batey, S. D. Gilbert, R. K. Montange, Nature 2004, 432,
411 – 415; b) A. Serganov, Y. R. Yuan, O. Pikovskaya, A.
Polonskaia, L. Malinina, A. T. Phan, C. Hꢀbartner, R. Micura,
R. R. Breaker, D. J. Patel, Chem. Biol. 2004, 11, 1729 – 1741.
[19] A. Roth, W. C. Winkler, E. E. Regulski, B. W. Lee, J. Lim, I.
Jona, J. E. Barrick, A. Ritwik, J. N. Kim, R. Welz, D. Iwata-
Reuyl, R. R. Breaker, Nat. Struct. Mol. Biol. 2007, 14, 308 – 317.
Angew. Chem. Int. Ed. 2007, 46, 7420 –7424
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7423

Communications
[20] W. C. Winkler, A. Nahvi, A. Roth, J. A. Collins, R. R. Breaker,
templated synthesis of a 3’–5’-linked GG dinucleotide. These two
possibilities cannot be distinguished on the basis of the available
data.
Nature 2004, 428, 281 – 286.
[21] GclvTP was synthesized from periodate-oxidized GTP by reduc-
tion with NaBH4 as described in the Supporting Information.
[22] GacvTP was synthesized from guanine in four steps as described
in the Supporting Information.
[25] O. W. Odom, Jr., D. J. Robbins, J. Lynch, D. Dottavio-Martin, G.
Kramer, B. Hardesty, Biochemistry 1980, 19, 5947–5954. Before
oxidation of the 2’,3’-diol of the attached nucleotide, the 3’-
terminal diol must first be rendered unreactive. This could be
achieved by creating a 2’,3’-cyclic phosphate; for example, as
formed by an RNA-cleaving deoxyribozyme.[26]
[23] A. Flynn-Charlebois, Y. Wang, T. K. Prior, I. Rashid, K. A.
Hoadley, R. L. Coppins, A. C. Wolf, S. K. Silverman, J. Am.
Chem. Soc. 2003, 125, 2444 – 2454.
[24] Use of dGTP in place of GTP led to a similar amount of the A–G
product but no detectable A–GG product. One interpretation is
that the template-synthesized GG dinucleotide is 2’–5’-linked.
Alternatively, the GTP 2’-OH group could be required for
[26] S. K. Silverman, Nucleic Acids Res. 2005, 33, 6151 – 6163.
[27] For a related strategy, see: D. A. Baum, S. K. Silverman, Angew.
Chem. 2007, 119, 3572– 3574; Angew. Chem. Int. Ed. 2007, 46,
3502– 3504.
7424 www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 7420 –7424