Nucleoside optimization for RNAi: A high-throughput platform

Article abstract of DOI:10.1021/ja2068774

The RNA induced silencing complex (RISC) contains at its core the endonuclease Argonaute (Ago) that allows for guide strand (GS)-mediated sequence-specific cleavage of the target mRNA. Functionalization of the sugar/phosphodiester backbone of the GS, which is in direct contact with Ago, presents a logical opportunity to affect RISCs activity. A systematic evaluation of modified nucleosides requires the synthesis of phosphoramidites corresponding to all four canonical bases (A, U, C, and G) and their sequential evaluation at each position along the 21-nucleotide-long GS. With the use of a platform approach, the sequential replacement of canonical bases with inosine greatly simplifies the problem and defines a new activity baseline toward which the corresponding sugar-modified inosines are compared. This approach was validated using 2′-O-benzyl modification, which demonstrated that positions 5, 8, 15, and 19 can accommodate this large group. Application of this high-throughput methodology now allows for hypothesis-driven rational design of highly potent, immunologically silent and stable siRNAs suitable for therapeutic applications.

Full text of DOI:10.1021/ja2068774

COMMUNICATION
pubs.acs.org/JACS
Nucleoside Optimization for RNAi: A High-Throughput Platform
Gabor Butora,*,† Denise M. Kenski,‡ Abby J. Cooper,‡ Wenlang Fu,† Ning Qi,† Jenny J. Li,‡
W. Michael Flanagan,‡ and Ian W. Davies†
†Department of Process Chemistry, Merck & Co., Rahway, New Jersey 07065, United States
‡Department of RNA Sciences, Sirna (a wholly owned subsidiary of Merck and Co.), 1700 Owens Street, Fourth Floor, San Francisco,
California 94158, United States
S Supporting Information
b
along the sugar/phosphodiester backbone. In contrast, the static,
ABSTRACT: The RNA induced silencing complex (RISC)
contains at its core the endonuclease Argonaute (Ago) that
allows for guide strand (GS)-mediated sequence-specific
cleavage of the target mRNA. Functionalization of the
sugar/phosphodiester backbone of the GS, which is in direct
contact with Ago, presents a logical opportunity to affect
RISC’s activity. A systematic evaluation of modified nucleo-
sides requires the synthesis of phosphoramidites corre-
sponding to all four canonical bases (A, U, C, and G) and
their sequential evaluation at each position along the 21-
nucleotide-long GS. With the use of a platform approach,
the sequential replacement of canonical bases with inosine
greatly simplifies the problem and defines a new activity
baseline toward which the corresponding sugar-modified
inosines are compared. This approach was validated using
20-O-benzyl modification, which demonstrated that posi-
tions 5, 8, 15, and 19 can accommodate this large group.
Application of this high-throughput methodology now
allows for hypothesis-driven rational design of highly po-
tent, immunologically silent and stable siRNAs suitable for
therapeutic applications.
cofactor-like interaction of unsubstituted RNA with the potential
binding pockets on the Ago2 protein surface is not optimal and
could account for RISC’s overall slow performance.20,21
Chemical modifications of the sugar/phosphodiester back-
bone present a logical opportunity to optimize the performance
of this complex. Improved siRNAꢀAgo2 interaction should
result in increased binding selectivity; more effective strand
selection; and improved catalytic turnover, siRNAꢀAgo2 com-
plex stability, and product release. Moreover, chemically mod-
ified siRNA duplexes are expected to be more resistant to
RNAase-mediated cleavage and have decreased affinity to toll-
like receptors and dsRNA-dependent protein kinase, resulting in
decreased immunogenicity.22,23
Most of the chemically modified nucleosides used today in
RNAi were synthesized to convey enzymatic stability to RNA
oligomers, and their design was not guided by siRNAꢀAgo2
binding considerations.24ꢀ26 Although the development of
structureꢀactivity relationships (SARs) for novel nucleosides
at distinct positions on the GS would be highly beneficial, it is
hindered by the complexity of the associated chemistry. Such an
effort would require independent SAR data collection for each
nucleotide along the siRNA GS, as the local Ago2 protein surface
adjacent to each particular position is different. Furthermore,
access to sugar-modified nucleosides containing all four canoni-
cal bases would be necessary, since a systematic investigation
would utilize a test siRNA with a predetermined sequence of
adenosine (A), guanosine (G), cytidine (C), and uridine (U) bases.
In theory, synthesis of 21-siRNA oligomers containing one
instance of the modified nucleoside (positions 1ꢀ21, “walkthrough”)
would be necessary. Evaluation of a single modification in a 96-
well-plate format (500nM loading/well) would typically require
∼1 g of each of the four canonical sugar-modified nucleoside
phosphoramidites. Depending on the character of the modifica-
tion, an estimated 40 synthetic steps would be needed to secure
these building blocks. This logistical complexity renders the
interrogation of such a huge chemical space intractable.
NA interference (RNAi) is a mechanism of post-transla-
R
tional gene silencing where double-stranded RNAs (dsRNAs),
such as short interfering RNAs (siRNAs), induce degradation of
sequence-specific homologous mRNA.1ꢀ3 siRNAs are RNA
duplexes containing 19ꢀ21 nucleotides that are widely used
for functional genetic or cellular screens.4 RNAi-based therapy
presents an attractive opportunity to engage targets not acces-
sible through conventional small molecules.5ꢀ7 To realize the
immense therapeutic potential of RNAi, many properties of
siRNAs, including potency,8 stability,9 immunogenicity,10,11
and delivery,12ꢀ14 must be optimized.
The RNA-induced silencing complex (RISC) is associated
with an RNAase H-type endonuclease, Argonaute (Ago), that
allows for guide strand (GS)-mediated sequence-specific clea-
vage of the target mRNA.15 A crystal structure of Thermus
thermophilus Argonaute 2 (Ago2) containing a chemically mod-
ified GS revealed that nucleotides 2ꢀ8 of the GS are preas-
sembled in an A-form helix and that the GS makes contact with
the surface of Ago2 through its sugar/phosphodiester backbone.16,17
This repetitive structure of the RNA scaffold bodes well for the
highly conserved primary roles of RNA, transcription18 and
translation,19 both of which require rapid ratcheting movement
In this communication, we report a platform solution for high-
throughput evaluation of chemically modified nucleosides. Ac-
cording to this, the natural bases present in a canonical test
oligomer (ApoB 10162, Figure 1A, X = cytosine, “Unmodified”)
are sequentially replaced by a suitable universal base.27ꢀ29 We
selected hypoxanthine30 for its nearly isoenergetic WatsonꢀCrick
Received: July 22, 2011
Published: September 23, 2011
r
2011 American Chemical Society
16766
dx.doi.org/10.1021/ja2068774 J. Am. Chem. Soc. 2011, 133, 16766–16769
|

Journal of the American Chemical Society
COMMUNICATION
Scheme 1a
t
a Conditions: (a) Bu2Si(OTf)2, imidazole, DMF; (b) Ac2O, TEA,
DMAP, DMF; (c)33 ArSO2Cl, TEA, DMAP, DCM; (d) trimethylsilyl
ethanol, DABCO, DBU, dioxane; (e) MeONa, 0 °C, MeOH.
Scheme 2a
a Conditions: (a) benzyl bromide, BEMP, MeCN; (b) TBAF, KF, 55 °C,
THF; (c) DMTCl, pyridine; (d) 2-cyanoethyl diisopropylchloropho-
sphoramidite, 1-methyl imidazole, DIEA, DCM. DMTr = 4,40-di-
methoxytrityl; CE = cyanoethyl.
Figure 1. Position-dependent effect of (A) the ribose-unmodified
universal base (e.g., inosine), (B) the ribose-modified universal base
(e.g., substituted inosine), and (C) the ribose-modified canonical base
on mRNA degradation. The numbers along the x axis indicate the
position along the guide strand (GS) of the siRNA duplex, with position
1 being the 50 end. Walkthrough connotes that the ribose-modified
nucleoside was systematically substituted at each position along the GS
and analyzed for mRNA target degradation (y axis). “Unmodified”
represents the level of target mRNA inhibition determined for the all-
ribose duplex.
performance of the RISC complex. Substituents placed at posi-
tion 20-O, 10-C, or 40-C are all capable of affecting this interaction,
and the design of more sophisticated nonribose alternatives is
definitely possible. Many of the chemical modifications that are
tolerated in siRNAs reside at the 20 position of the ribose
ring,34,35 and we demonstrate the utility of our high-throughput
approach by evaluating a simple 20-O-benzyl modification.
Replacement of the four canonical heterocycles, correspond-
ing to A, U, C, and G, with a universal base allows for the design
of a downstream relay such as 6, greatly simplifying the associated
chemistry (Scheme 1). With this intermediate in hand, the final
20-O-benzyl-modified phosphoramidites became accessible in as
few as four chemical steps (Scheme 2). To facilitate high-
throughput parallel synthesis of analogous phosphoramidites
containing modified benzyl groups, the first two and last two
steps in Scheme 2 could be combined into single operations.
As a test sequence, we used a synthetic double-stranded 21-
mer corresponding to positions 10162 to 10182 of the mRNA
from the human ApoB gene. The siRNAs were tested at a
subsaturating concentration of 1 nM that is capable of target-
specific mRNA degradation and can be detected by quantitative
polymerase chain reaction in cell-based assays (Figure 1A, X =
cytosine, “Unmodified”). First, the ability of siRNAs containing
sugar-unmodified universal nucleotides to degrade target mRNA
was assessed (Figure 2A1, black bar, first three positions shown).
This established a new baseline toward which the sugar-modified
inosine-containing siRNAs were compared (Figure 2A1, blue
bar). To account for variations between measurements, the raw
interaction with cytosine, uracil, adenine,31 and to a lesser extent
with guanine.32 The cell-based performance of this universal-
nucleoside-containing RNA set establishes a new activity base-
line (Figure 1A, X = hypoxanthine) toward which the analogous
sugar-modified oligomers are compared (Figure 1B, X = hypox-
anthine, modified ribose, i.e., 20-O-benzyl). If the activity of the
sugar-modified inosine surpasses that of the unmodified inosine
placed at the same position along the siRNA oligomer, this
relationship should be also reflected in the improved perfor-
mance of the sugar-modified full canonical nucleoside. Similarly,
if the effect of the sugar-modified inosine is detrimental to
activity, we would expect this also to be paralleled by the modified
canonical nucleoside. The four chemically modified canonical
nucleosides (Figure 1C, N = canonical base, modified ribose, i.e.
20-O-benzyl) are then synthesized only when the modification has
demonstrated a benefit using this simplified platform.
Any sugar modification capable of projecting a group toward
the Ago2 surface can in principle be used to optimize the
16767
dx.doi.org/10.1021/ja2068774 |J. Am. Chem. Soc. 2011, 133, 16766–16769

Journal of the American Chemical Society
COMMUNICATION
Figure 4. Position-dependent effect of 20-O-(4-oxazol-2-yl)benzyl
modification on target mRNA degradation evaluated on the inosine
(blue) and canonical (black)siRNA (r2 = 0.50, p < 0.0007; ApoB10162).
Figure 2. Schematic representation of the workflow used to evaluate 20-
O-benzyl modification: (A) inosine and (B) canonical background. Only
positions 1ꢀ3 are shown.
To verify these predictions, 20-O-benzyl-modified canonical
nucleoside amidites (corresponding to A, U, C, and G) were
synthesized following proprietary or published procedures,36
incorporated into the GS of the test siRNA (Figure 1C, modified
ribose, N = canonical base), and evaluated (Figure 2B1ꢀ3). As
predicted by the inosine platform, positions 5, 8, 15, and 19 are
the most accommodating of 20-O-benzyl modifications, and
these siRNAs have the same or better activity as the unmodified
oligomers (Figure 3, black). Potency data obtained for canonical
sequences carrying the 20-O-benzyl modifications confirmed
these observations (Figure S2 and Table S3 in the Supporting
Information). To the best of our knowledge, this finding
represents the first report of relatively large benzyl group being
tolerated along the GS, including in the seed region.
The mRNA target selection of an siRNA-primed RISC com-
plex is controlled by the base sequence of the GS anchored by the
Ago2. As a consequence, sugar modification is largely sequence-
independent, enabling a fragment screening approach to opti-
mization of the siRNAꢀprotein surface interaction. Evaluation
of the 20-O-benzylribose using an siRNA targeting the human
propyl hydroxylase domain 2 (PHD2) gene, which lacks homol-
ogy with the ApoB sequence, confirmed this assertion and
highlights the potential of the platform approach to siRNA
optimization (Figure S4).
To exemplify the potential for SAR data collection, the
spatially more demanding 20-O-(p-oxazol-2-yl)benzyl group
was examined. The results suggested marginal tolerability at
positions 5, 15, and 16 (Figure 4, blue) and intolerability at
positions 2 and 11ꢀ14. These observations lead to the prediction
that the modification would not be beneficial. Synthesis of the
canonical 20-O-(p-oxazol-2-yl)benzyl phosphoramidites and the
corresponding siRNA (Figure 4, black) verified this prediction
experimentally.
Figure 3. Position-dependent effect of 20-O-benzyl modification on
target mRNA degradation evaluated on the inosine (blue) and canonical
(black) siRNA (r2 = 0.64, p < 0.0001; ApoB10162).
data were normalized to unmodified canonical siRNA (Figure 2A2)
and compared with the inosine baseline (Figure 2A3; Figure 3, blue).
The canonical siRNA (Figure 1A, X = cytosine, “Un-
modified”) at 1 nM concentration exhibited 60% mRNA levels.
With the exception of positions 11 (opposing the mRNA
cleavage site) and 18, replacement of any canonical heterocycle
with hypoxanthine (Figure 1A, X = hypoxanthine) resulted in the
loss of no more than 50% of its activity. The sugar-modified
siRNAs containing 20-O-benzylinosines at positions 1ꢀ4, 12, 14,
and 18 further decreased the mRNA degradation in comparison
with the sugar-unmodified inosine baseline. Therefore, we con-
cluded that a 20-O-benzyl modification would be detrimental if
placed here as part of a canonical nucleoside. On the other hand,
the performance of 20-O-benzyl-modified inosine at positions 5,
8, 15, and 19 was similar or better than the sugar-unmodified
inosine baseline, so we concluded that a 20-O-benzyl-modified
nucleoside would be well-tolerated with respect to activity when
placed at positions 5, 8, 15, or 19 of the siRNA (Figure 3, blue).
A platform comparison of the 20-O-benzyl group in its natural
(ribo) and unnatural (arabino) configurations suggested that the
RNAi performance of a modifying group is not simply a con-
sequence of its size. The 20-O-ara-benzyl group was less tolerated,
with notable difference at positions 7 and 15 (Figure S5).
As expected, sequential inclusion of inosine caused a position-
dependent decrease in the melting temperature (Tm) that was
not substantially changed by 20-O-benzylation(FigureS6). Similarly,
16768
dx.doi.org/10.1021/ja2068774 |J. Am. Chem. Soc. 2011, 133, 16766–16769

Journal of the American Chemical Society
COMMUNICATION
(8) Koller, E.; Propp, S.; Murray, H.; Lima, W.; Bhat, B.; Prakash,
T. P.; Allerson, C. R.; Swayze, E. E.; Marcusson, E. G.; Dean, N. M.
Nucleic Acids Res. 2006, 34, 4467.
(9) Watts, J. K.; Deleavey, G. F.; Damha, M. J. Drug Discovery Today
2008, 13, 842.
(10) Sioud, M.; Furset, G. J. Biomed. Biotechnol. 2006, 1.
(11) Sioud, M. Trends Mol. Med. 2006, 12, 167.
(12) White, P. J. Clin. Exp. Pharmacol. Physiol. 2008, 35, 1371.
(13) Tseng, Y. C.; Mozumdar, S.; Huang, L. Adv. Drug Delivery Rev.
2009, 61, 721.
(14) Howard, K. A. Adv. Drug Delivery Rev. 2009, 61, 710.
(15) Hall, T. M. T. Structure 2005, 13, 1403.
(16) Wang, Y. L.; Sheng, G.; Juranek, S.; Tuschl, T.; Patel, D. J.
Nature 2008, 456, 209.
(17) Wang, Y. L.; Juranek, S.; Li, H. T.; Sheng, G.; Tuschl, T.; Patel,
Figure 5. Position-dependent effect of 20-O-benzyl modification on the
melting temperature (Tm) of the duplex: modified siRNA (blue) and
unmodified siRNA (black) (ApoB10162).
D. J. Nature 2008, 456, 921.
a walkthrough of 20-O-benzyl-modified canonical nucleosides did
not cause a substantial deviation of Tm values either, and no
apparent correlation to the observed positional effect on target
mRNA levels was observed (Figure 5).
(18) Kornberg, R. Angew. Chem., Int. Ed. 2007, 46, 6957.
(19) Simonovic, M.; Steitz, T. A. Biochim. Biophys. Acta 2009,
1789, 612.
(20) Rivas, F. V.; Tolia, N. H.; Song, J. J.; Aragon, J. P.; Liu, J. D.;
Hannon, G. J.; Joshua-Tor, L. Nat. Struct. Mol. Biol. 2005, 12, 340.
(21) Haley, B.; Zamore, P. D. Nat. Struct. Mol. Biol. 2004, 11, 599.
(22) Kariko, K.; Buckstein, M.; Ni, H. P.; Weissman, D. Immunity
2005, 23, 165.
(23) Kariko, K.; Weissman, D. Curr. Opin. Drug Discovery Dev. 2007,
10, 523.
(24) Manoharan, M. Curr. Opin. Chem. Biol. 2004, 8, 570.
(25) Bramsen, J. B.; et al. Nucleic Acids Res. 2009, 37, 2867.
(26) Zhang, H. Y.; Du, Q.; Wahlestedt, C.; Liang, Z. C. Curr. Top.
Med. Chem. 2006, 6, 893.
(27) Loakes, D. Nucleic Acids Res. 2001, 29, 2437.
(28) Kool, E. T. Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 1.
(29) Sherer, E. C.; York, D. M.; Cramer, C. J. J. Comput. Chem. 2003,
24, 57.
(30) Chizhikov, V.; Wagner, M.; Ivshina, A.; Hoshino, Y.; Kapikian,
A. Z.; Chumakov, K. J. Clin. Microbiol 2002, 40, 2398.
(31) Janke, E. M. B.; Riechert-Krause, F.; Weisz, K. J. Phys. Chem. B
2011, 115, 8569.
(32) Burkard, M. E.; Turner, D. H. Biochemistry 2000, 39, 11748.
(33) Zlatev, I.; Vasseur, J. J.; Morvan, F. Tetrahedron 2007, 63,
11174.
(34) Jackson, A. L.; Burchard, J.; Leake, D.; Reynolds, A.; Schelter, J.;
Guo, J.; Johnson, J. M.; Lim, L.; Karpilow, J.; Nichols, K.; Marshall, W.;
Khvorova, A.; Linsley, P. S. RNA 2006, 12, 1197.
(35) Robbins, M.; Judge, A.; Liang, L.; McClintock, K.; Yaworski, E.;
MacLachlan, I. Mol. Ther. 2007, 15, 1663.
In this communication, we have described a novel high-
throughput approach for evaluation of chemically modified nucle-
osides for RNAi. The method was validated using inosine as a
universal nucleoside and exemplified using 20-O-benzyl and 20-O-(p-
oxazol-2-yl)benzyl modifications in multiple sequences. The 20-O-
benzyl-modified nucleosides exhibited an excellent correla-
tion with the SAR data collected using the universal base.
Application of this high-throughput methodology now en-
ables hypothesis-driven design in analogy to small-molecule
fragment screening. This platform approach enhances the
value of innovative chemistry in the discovery of highly
potent, immunologically silent, and stable siRNAs suitable
for therapeutic applications.
’ ASSOCIATED CONTENT
S
Supporting Information. Experimental procedures, four
b
figures and one table mentioned in the text, and complete ref 25.
This material is available free of charge via the Internet at http://
pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
gabor_butora@merck.com
(36) Compositions and Methods for Detecting and Modulating
RNA Activity and Gene Expression. WO 91/10671, July 25, 1991.
’ ACKNOWLEDGMENT
The authors thank the RNA Oligosynthesis Group (Merck
Process Chemistry, Rahway, NJ), especially Derek von Langen,
John Limanto, Mark Levorse, and Rick Sidler, for synthesis and
physical characterization of the RNA oligomers and Ed Sherer
(Merck, Rahway, NJ) for support with computer modeling.
’ REFERENCES
(1) Fire, A. Z. Angew. Chem., Int. Ed. 2007, 46, 6967.
(2) Mello, C. C. Angew. Chem., Int. Ed. 2007, 46, 6985.
(3) Grimm, D. Adv. Drug Delivery Rev. 2009, 61, 672.
(4) Elbashir, S. M.; Harborth, J.; Lendeckel, W.; Yalcin, A.; Weber,
K.; Tuschl, T. Nature 2001, 411, 494.
(5) de Fougerolles, A.; Vornlocher, H. P.; Maraganore, J.; Lieberman
J. Nat. Rev. Drug Discovery 2007, 6, 443.
(6) Castanotto, D.; Rossi, J. J. Nature 2009, 457, 426.
(7) Siomi, M. C. Adv. Drug Delivery Rev. 2009, 61, 668.
16769
dx.doi.org/10.1021/ja2068774 |J. Am. Chem. Soc. 2011, 133, 16766–16769