The roles of hydrogen bonding and sterics in RNA interference

Article abstract of DOI:10.1002/anie.200601311

(Formula Presented) Better than normal: RNA-interference studies with mismatched target RNA have demonstrated sequence selectivity (at the single-nucleotide level) at many different positions of the RNA strand. The use of rF in place of rU at position 7 a

Full text of DOI:10.1002/anie.200601311

Communications
RNA Structures
DOI: 10.1002/anie.200601311
The Roles of Hydrogen Bonding and Sterics in
RNA Interference**
Alvaro Somoza, Jijumon Chelliserrykattil, and
Eric T. Kool*
RNA interference (RNAi) has become one of the most
important new tools for biological research in the past
decade.[1–5] This methodology is used widely for “knocking
down” (regulating) expression of specific genes in cell
cultures, and it can be carried out conveniently by using
synthetic RNA oligonucleotides (short interfering RNAs or
siRNAs) that are complementary to a segment of a desired
messenger RNA target. When the appropriate double-
stranded 21mer RNAs (the sense and antisense strands) are
added to a cell culture, they are taken up by the cellular RNA-
induced silencing complex (RISC), which presents the
separated antisense (“guide”) strand for binding and subse-
quent cleavage of the target mRNA.[1–5,6] One of the most
useful features of this approach is that the resulting mRNA
cleavage occurs with sequence selectivity[7,8] so that one gene
can often be knocked down to low levels of activity with little
effect on the rest of cellular gene expression.[9,10]
Recent RNA-interference studies with mismatched target
RNAs have demonstrated sequence selectivity (at the single-
nucleotide level) at many positions on the standard 21-
nucleotide probe length.[8] The origins of this selectivity are
not known; selectivity may arise from base-pair hydrogen
bonding, which contributes to selective hybridization,[11] or
from steric complementarity of nucleobases, which is impor-
tant in replication by DNA polymerases.[12] Furthermore,
selectivity could come chiefly from the RNA itself, or could
be modulated by the RISC complex. Beyond this, it is not
clear why responsiveness to mismatches varies along the
siRNA strand.[8] Although recent structural studies of Piwi/
Argonaute/Zwille (PAZ) domains[13] and Argonaute pro-
teins[14] (parts of the RISC complex) have led to models of
RNA cleavage, no structures of the RISC complex bound to
siRNA and mRNA are yet available.
Figure 1. Modified RNA structures and sequences used in this study.
A) 2,4-Difluorobenzene ribonucleoside rF, a uridine nonpolar isostere,
is shown next to uridine (rU). B) Sequences of siRNA duplexes used in
RNAi experiments. Sites of rF substitutions in various experiments are
marked with “Fn”. The “guide”-strand sequence (antisense to the
mRNA) is above; “passenger” strands (corresponding to the target
mRNA and mutants) are below. The shown sequence corresponds to
nucleotides 501–519 in Renilla luciferase mRNA.
rated into a siRNA guide strand in place of natural uridine.
We found that this analogue can maintain near-wild-type
activity in human cells at a number of positions in the strand
and importantly, we observe that it can retain and even
enhance sequence selectivity.
The analogue rF contains difluorobenzene, a uracil
isostere, in place of the earlier-used difluorotoluene deoxy-
riboside (dF) as a thymidine mimic in DNA.[17] This nonpolar
structure serves as a probe for the importance of hydrogen
bonding and electrostatics in the RNA context.[15,16,18] RNA
hybridization studies have shown that rF pairs have little or
no inherent selectivity and are destabilizing to the RNA
duplex[15] consistent with its nonpolar properties; this is also
consistent with the behavior of dF in DNA.[19] In separate
experiments, we incorporated rF into eleven different posi-
tions in place of natural rU along an RNA guide strand that
was complementary to a luciferase reporter gene in an A-rich
site (Figure 1).
Thermal denaturation studies of synthetic RNA duplexes
showed that rF is destabilizing to 21mer double-stranded
siRNAs when it is placed near the center (see the Supporting
Information). However, this destabilization lessened near the
duplex ends (positions 1, 3, and 19), and virtually no
destabilization was seen at unpaired positions 20 and 21.
This is similar to the positional effects of mismatches on the
helix stability of RNA duplexes,[20] and is reminiscent of the
behavior of dF in DNA.[11] To measure selectivity, studies
were also carried out with RNAs mismatched at a central
position (position 7), where destabilization by rF is high.
Results confirmed (see the Supporting Information) that this
nonpolar nucleoside gives no pairing selectivity for ade-
nine.[15,16,18,19] By comparison, natural uridine showed sub-
stantial pairing selectivity for adenine.
Herein, we describe experiments that give insights into
the origins of RNAi activity and selectivity. We have
evaluated this with a nonpolar, non-hydrogen-bonding ribo-
nucleoside isostere (rF; Figure 1)[15,16] that we have incorpo-
[*] Dr. A. Somoza,Dr. J. Chelliserrykattil,Prof. Dr. E. T. Kool
Department of Chemistry,Stanford University
Stanford,CA 94305-5080 (USA)
Fax: (+1)650-725-0259
E-mail: kool@stanford.edu
[**] This work was supported by the US National Institutes of Health
(GM072705),and by a Marie Curie OIF Fellowship (6th Framework
Program,UE) to A.S.
We then carried out separate RNA-interference studies in
HeLa cells with the eleven modified guide-RNA strands
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
4994
ꢀ 2006 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2006, 45, 4994 –4997

Angewandte
Chemie
containing single rF substitutions. Experiments were carried
out in triplicate. The cells were first transfected with dual
reporter plasmids that express Renilla luciferase and firefly
luciferase, the activities of which could be separately quanti-
tated by their luminescence.[21] The synthetic RNAs were
complementary to positions 501–519 of Renilla luciferase
mRNA[22] and not complementary to the firefly mRNA. The
effects of the different RNAs on luciferase expression were
evaluated after dosing with 0.21–21 ng of RNA in the cell
media, and measuring relative luminescence responses after
22 h. The firefly luminescence is an internal control that rules
out possible indirect effects of the different siRNAs, such as
selective cell toxicity, variable cellular uptake, or differential
enzymatic degradation.
Results showed that the rF substitution disrupted Renilla
luciferase-specific RNA interference activity at two central
positions of the guide RNA; surprisingly, however, a number
of other positional substitutions retained near-wild-type
activity (Figure 2). Substitutions of rF at positions 10 and 11
Figure 3. Dual plot comparing helix stability (Tm) and RNA gene
knockdown activity (% suppression) with rF substitution along the
guide strand of siRNA duplexes at the positions shown. Data at the far
right (“wt”) are for the natural rU-substituted siRNA. See the Support-
ing Information for the original Tm data. Suppression data are from
2.1-ng siRNA amounts (see the Supporting Information).
unique opportunity to test the chemical origins of sequence
selectivity in RNA interference. The corresponding position
in the Renilla luciferase mRNA is universally variable in the
gly171 codon. We therefore prepared three new mutant
plasmids encoding Renilla luciferase with singly mismatched
mRNAs. Measurement of activity with these four targets
showed that the naturally substituted guide RNA did indeed
distinguish single mismatches (Figure 4); the mismatches led
to as much as a 2-fold drop in activity at a 2.1-ng RNA dose.
This is consistent with a report for a different RNA target in
which target mutations at position 7 led to significant
selectivity.[8] In our experiments, a U–U mismatch was well
tolerated, whereas a U–C mismatch was most disruptive to
activity.
Remarkably, experiments with rF (position 7) showed
that it also displayed selectivity, and more surprisingly, the
level of selectivity was higher than that of the natural base. A
selectivity of nearly threefold greater that the F–C mismatch
was seen (Figure 4), and the rF-containing strand also showed
significant selectivity for cleavage of the G mutant, which rU
discriminated poorly. Thus, we conclude that, at least at
position 7 of the RISC complex, mRNA-target-sequence
selectivity does not require canonical Watson–Crick hydrogen
bonding.
These experiments give new insight into the mechanism of
RNA interference. We have shown that a nonpolar nucleo-
base analogue can maintain substantial cellular RNAi activity
at nine of eleven sites tested, establishing clearly that
canonical Watson–Crick hydrogen bonding is not crucial at
all positions. We do note a general correspondence of Tm with
mRNA-suppression activity, however (Figure 3), suggesting
that not many simultaneous substitutions in guide RNA
would be well tolerated. Notably, two positions tested (7 and
10) deviate markedly from this correlation, suggesting con-
siderable changes in RNA–protein interactions across this
Figure 2. Histogram of gene-specific RNA interference activity for rF-
substituted siRNAs at one site in the Renilla luciferase mRNA
expressed in HeLa cells. The numbers refer to the position of rF in the
guide-RNA strand. Data were normalized by an internal control of a
noncomplementary firely luciferase gene. Data are for the 2.1-ng RNA
amount,and were measured in triplicate; standard deviations are as
indicated. See the Supporting Information for the full data. wt=wild
type.
caused loss of most of the interference activity even with a
higher 21-ng dose of siRNA (see the Supporting Informa-
tion); these positions flank the expected cleavage site in the
target mRNA.[3] In contrast to this, all other substitutions
retained substantial activity within 2–4-fold of the natural
uracil-substituted strand. Especially remarkable is substitu-
tion at position 7, which showed near-wild-type activity
despite the strong destabilization that this substitution
causes (Figure 3). At the higher (21 ng) dose, guide RNAs
containing rF at positions 1, 7, 19, 20, and 21 all showed near-
wild-type activity (see the Supporting Information). Inter-
mediate effects were found at positions 3, 5, 13, and 14. These
results suggest that canonical hydrogen bonding is not
necessary at several positions for the RISC complex to
maintain high levels of RNA interference activity with the
intended target.
The retention of strong activity at position 7 with the rF
analogue, despite its central location in the siRNA, offers a
Angew. Chem. Int. Ed. 2006, 45, 4994 –4997
ꢀ 2006 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 4995

Communications
the two studies, carried out with different targets, suggests
that these effects are general for the RISC complex. Sequence
selectivity with mismatched targets was not examined in that
study.
The current results strongly suggest that at position 7, the
sequence selectivity of RNA interference arises not from the
selectivity of hydrogen bonding, nor from selectivity caused
by the RNA backbone alone. A possible hypothesis is that the
selectivity arises instead from an enforced steric selection by
the RISC complex (and Argonaute in particular[14]). Experi-
ments with base analogues with varied size or shape could test
this steric hypothesis explicitly. At this point, it is unknown
whether selectivity at other positions depends on hydrogen
bonding or not.
Finally, we note with interest the observation that use of
rF in place of rU at position 7 appears to enhance sequence
selectivity beyond that of the natural base. We suggest that
this may be due to the general destabilization of the guide–
target duplex by the nonpolar base analogue; it is known that
a nearby mismatch can increase specificity of DNA hybrid-
ization,[23] perhaps by a similar destabilization mechanism. In
any case, the result suggests a more-general practical utility of
nonpolar isosteres in RNA interference studies.
Experimental Section
Nucleoside phosphoramidite synthesis: The rF ribonucleoside was
prepared as described.[15,16] For automated RNA synthesis, the
previously unknown 5-dimethoxytrityl, 2’-O-TOM (TOM = [(triiso-
propylsilyl)oxy]methyl), 3’-O-phosphoramidite derivative was pre-
pared. Details are given in the Supporting Information.
Modified oligoribonucleotides: RNA sequences were made
following standard protocols for 2’-O-TOM-protected phosphorami-
dites[24] and were purified by polyacrylamide gel electrophoresis and
characterized by MALDI-TOF mass spectrometry. Details are
supplied in the Supporting Information.
RNA interference studies and mismatched luciferase vectors:
The dual reporter Renilla/firefly assay was carried out by using HeLa
cells transfected with reporter expression plasmids purchased from
Promega (San Luis Obispo, CA). Sequence mutant plasmids were
prepared following standard methods and were characterized by
sequencing. Details are given in the Supporting Information.
Figure 4. Selectivity of modified siRNA guide strands for singly mis-
matched RNA targets. The mismatch is opposite position 7 of the
guide RNA and involves mutating the corresponding nucleotide in the
complementary RNA. A) Histogram of RNAi activity showing rF-
enhanced selectivity for adenine relative to mutations at position 513
in the luciferase target mRNA (position 513 is opposite position 7 on
the guide RNA). B) Thermal stability histogram showing inherent
hybridization selectivity of rU,and lack of selectivity of rF,as measured
by thermal denaturation of the guide/passenger duplexes correspond-
ing to the mutations in part (A). Raw data are given in the Supporting
Information. Tm (melting temperature) represents the RNA duplex
stability.
Received: April 4, 2006
Revised: May 16, 2006
Published online: June 27, 2006
Keywords: base pairs · difluorobenzene · isosteres ·
.
protein structures · RNA structures
localized part of the RNA. The overall profile of rF-
substitution responses suggests that guide-RNA positions 10
and 11, in particular, may require hydrogen bonding or base-
pairing stability with the target (or both) for efficient RNA
cleavage activity.
After submission of this manuscript, a report appeared
testing difluorotoluene riboside, closely related to the current
rF, in siRNA activity.[18] This study found similar positional
effects on RNA interference, with the base analogue causing
strong disruption of activity at positions 10 and 11, and little
or no effect at position 7. The consistency of findings between
[1] A. Fire, S. Xu, M. K. Montgomery, S. A. Kostas, S. E. Driver,
C. C. Mello, Nature 1998, 391, 806 – 811.
[2] S. M. Elbashir, J. Martinez, A. Patkaniowska, W. Lendeckel, T.
Tuschl, EMBOJ. 2001, 20, 6877 – 6888.
[3] S. M. Elbashir, W. Lendeckel, T. Tuschl, Genes Dev. 2001, 15,
188 – 200.
[4] G. Hutvagner, P. D. Zamore, Curr. Opin. Genet. Dev. 2002, 12,
225 – 232.
[5] T. Tuschl, G. Meister, Nature 2004, 431, 343 – 349.
[6] M. Manoharan, Curr. Opin. Chem. Biol. 2004, 8, 570 – 579.
[7] B. L. Bass, Cell 2000, 101, 235 – 238.
4996 www.angewandte.org
ꢀ 2006 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2006, 45, 4994 –4997

Angewandte
Chemie
[8] Q. Du, H. Thonberg, J. Wang, C. Wahlestedt, Z. Liang, Nucleic
Acids Res. 2005, 33, 1671 – 1677.
[9] J.-T. Chi, H. Y. Chang, N. N. Wang, D. S. Chang, N. Dunphy, P. O.
Brown, Proc. Natl. Acad. Sci. USA 2003, 100, 6343 – 6346.
[10] D. Semizarov, L. Frost, A. Sarthy, P. Kroeger, D. N. Halbert,
S. W. Fesik, Proc. Natl. Acad. Sci. USA 2003, 100, 6347 – 6352.
[11] a) B. A. Schweitzer, E. T. Kool, J. Am. Chem. Soc. 1995, 117,
1863 – 1872; b) E. T. Kool, J. C. Morales, K. M. Guckian, Angew.
Chem. 2000, 112, 1046 – 1068; Angew. Chem. Int. Ed. 2000, 39,
990 – 1009.
[12] S. Moran, R. X.-F. Ren, E. T. Kool, Proc. Natl. Acad. Sci. USA
1997, 94, 10506 – 10511.
[13] a) A. Lingel, B. Simon, E. Izaurralde, M. Sattler, Nat. Struct.
Mol. Biol. 2004, 11, 576 – 577; b) J. B. Ma, K. Ye, D. J. Patel,
Nature 2004, 429, 318 – 322; c) A. Lingel, M. Sattler, Curr. Opin.
Struct. Biol. 2005, 15, 107 – 115.
[14] a) J. J. Song, S. K. Smith, G. J. Hannon, L. Joshua-Tor, Science
2004, 305, 1434 – 1437; b) J. J. Song, L. Joshua-Tor, Curr. Opin.
Struct. Biol. 2006, 16, 5 – 11.
[15] J. Parsch, J. W. Engels, Helv. Chim. Acta 2000, 83, 1791 – 1808.
[16] K. M. Guckian, Ph.D. Thesis, University of Rochester (NY),
1998.
[17] K. M. Guckian, E. T. Kool, Angew. Chem. 1997, 109, 2942 – 2945;
Angew. Chem. Int. Ed. 1997, 36, 2825 – 2828.
[18] J. Xia, A. Noronha, I. Toudjarska, F. Li, A. Akinc, R. Braich, M.
Frank-Kamenetsky, K. G. Rajeev, M. Egli, M. Manoharan, ACS
Chem. Biol. 2006, 1, 176 – 182.
[19] S. Moran, R. X.-F. Ren, S. Rumney, E. T. Kool, J. Am. Chem.
Soc. 1997, 119, 2056 – 2057.
[20] Y. Xu, H.-Y. Zhang, D. Thormeyer, O. Larsson, Q. Du, J. ElmØn,
C. Wahlestedt, Z. Liang, Biochem. Biophys. Res. Commun. 2003,
306, 712 – 717.
[21] R. Kierzek, M. E. Burkard, D. H. Turner, Biochemistry 1999, 38,
14214 – 14223.
[22] G. Grentzmann, J. A. Ingram, P. J. Kelly, R. F. Gesteland, J. F.
Atkins, RNA 1998, 4, 479 – 486.
[23] Z. Guo, Q. Liu, L. M. Smith, Nat. Biotechnol. 1997, 15, 331 – 335.
[24] a) S. Pitsch, P. A. Weiss, L. Jenny, A. Stutz, X. Wu, Helv. Chim.
Acta 2001, 84, 3773 – 3795; b) S. Jin, C. V. Miduturu, D. C.
McKinney, S. K. Silverman, J. Org. Chem. 2005, 70, 4284 – 4299.
Angew. Chem. Int. Ed. 2006, 45, 4994 –4997
ꢀ 2006 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 4997