MINIREVIEW
Staying onmessage:designprinciplesfor controlling
nonspecific responsesto siRNA
Shirley Samuel-Abraham
1
and Joshua N. Leonard
1,2
1 Department of Chemical and Biological Engineering, Northwestern University, Evanston, IL, USA
2 Member, Robert H. Lurie Comprehensive Cancer Center, Northwestern University, Chicago, IL, USA
Introduction
In the decade since RNA interference (RNAi) was
initially discovered in Caenorhabditis elegans [1] and
shown to be inducible in mammalian cells [2,3],
technologies for harnessing this mechanism to induce
targeted gene silencing have become routine laboratory
tools and, increasingly, are making their way into
clinical trials (reviewed in Castanotto & Rossi [4]).
Over this same period, however, it has become clear
that the short interfering RNA (siRNA) commonly
delivered to induce RNAi can also induce multiple
nonspecific effects. A poignant example comes from the
first system for which clinical trials of RNAi were
Keywords
innate immunity; OAS1; off-target; RIG-I;
RNA interference; RNAi; short interfering
RNA; siRNA; TLR; Toll-like receptors
Correspondence
J. N. Leonard, Department of Chemical and
Biological Engineering, Northwestern
University, 2145 Sheridan Rd, Room E-136,
Evanston, IL 60208 USA
Fax: +1 847 491 3728
Tel: +1 847 491 7455
E-mail: j-leonard@northwestern.edu
(Received 7 July 2010, accepted 26 August
2010)
doi:10.1111/j.1742-4658.2010.07905.x
Short interfering RNAs (siRNA) are routinely used in the laboratory to
induce targeted gene silencing by RNA interference, and increasingly, this
technology is being translated to the clinic. However, there are multiple
mechanisms by which siRNA may be recognized by receptors of the innate
immune system, including both endosomal Toll-like receptors and cytoplas-
mic receptors. Signaling through these receptors may induce multiple non-
specific effects, including general reductions in gene expression and the
production of type I interferons and inflammatory cytokines, which can
lead to systemic inflammation in vivo. The pattern of immune activation
varies depending upon the types of cells and receptors that are stimulated
by a particular siRNA. Although we are still discovering the mechanisms
by which these recognition events occur, our current understanding pro-
vides useful guidelines for avoiding immune activation. In this minireview,
we present a design-based approach for developing siRNA-based experi-
ments and therapies that evade innate immune recognition and control
nonspecific effects. We describe strategies and trade-offs related to siRNA
design considerations including the choice of siRNA target sequence, chem-
ical modifications to the RNA backbone and the influence of the delivery
method on immune activation. Finally, we provide suggestions for conduct-
ing appropriate controls forsiRNA experiments, because some commonly
employed strategies do not adequately account for known nonspecific
effects and can lead to misinterpretation of the data. By incorporating
these principles into siRNA design, it is generally possible to control
nonspecific effects, and doing so will help to best utilize this powerful
technology for both basic science and therapeutics.
Abbreviations
dsRNA, double-stranded RNA; GFP, greem fluorescent protein; IFN, interferon; IL, interleukin; OAS1, 2¢-5¢-oligoadenylate synthetase;
PKR, protein kinase R; RIG-I, retinoic acid-inducible gene I; RISC, RNA-induced silencing complex; RNAi, RNA interference; siRNA,
short interfering RNA; ssRNA, single-stranded RNA; TLR, Toll-like receptor.
4828 FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS
initiated – intravitreous injection of siRNA against
vascular endothelial growth factor to block angiogenesis
in patients with blinding choroidal neovascularization
[4]. Recent data from animal models of choroidal neo-
vascularization indicate that the therapeutic benefits of
this treatment are mediated in large part by nonspecific
mechanisms involving recognition of siRNA by the
innate immune system [5,6]. Nonetheless, definitive
proof that siRNA can also induce RNAi-mediated spe-
cific gene silencing in human patients was recently
reported in a clinical trial for nanoparticle-mediated
siRNA delivery for melanoma treatment [7]. In addi-
tion, some clinical strategies are now being designed
to harness both the specific and nonspecific effects of
siRNA therapeutics [8,9], although the relative contri-
butions of each mechanism remain somewhat unclear.
Given the complexity and potential subtlety of these
nonspecific effects, siRNA-based experiments and
pre-clinical studies should incorporate our growing
knowledge of the molecular features that give rise to
innate immune recognition. This review presents a
design-oriented approach forcontrolling innate
immune system-mediated interactions when developing
siRNA-based therapeutics.
In higher animals, RNAi constitutes one arm of an
arsenal of innate defenses against viral infections.
Consequently, the same molecules that induce targeted
gene silencing through RNAi [including double-
stranded RNA (dsRNA) and siRNA] also induce
nonspecific antiviral responses through these overlap-
ping mechanisms. Two cytoplasmic receptors that
have long been known to recognize long dsRNA
include protein kinase R (PKR) and 2¢-5¢-oligoadeny-
late synthetase (OAS1). Upon binding to dsRNA,
PKR catalyzes the phosphorylation of eIF2a and I jB,
which induces a general inhibition of translation and
drives the production of type I interferons (e.g. IFN-a
and IFN-b) through NF-jB [10,11]. Most siRNA are
shorter than the 30 bp minimum dsRNA length
required to potently activate PKR [12], and although
some reports indicate that detectable PKR activation
can be induced by siRNA, it is not yet clear whether
this low-level of activation induces biologically rele-
vant responses [13,14]. OAS1 is activated by binding
to dsRNA and induces sequence-independent degrada-
tion of viral and cellular single-stranded RNA
(ssRNA) by activating RNaseL [15]. OAS1 also plays
an important role in the amplification of innate
immune responses, because OAS1 expression is upreg-
ulated by type I interferon, and the small dsRNA
products of RNaseL-digested cellular or viral mRNA
can activate innate immune receptors in neighboring
cells [16]. Some evidence indicates that certain dsRNA
of only 19 bp in length can activate OAS1 directly
[17].
Our current understanding is that the most potent
nonspecific siRNA-induced effects are mediated by
more recently characterized receptors located in dis-
tinct subcellular compartments. The nucleic acid-
responsive Toll-like receptors (TLRs) interact with
pathogen-associated molecules in endosomal vesicles,
and TLR3 [5,6,18], TLR7 [19–21] and TLR8 [19,22]
have each been implicated in the response to siRNA.
Of these, TLR7 and TLR8 are thought to mediate the
dominant immune response tosiRNA in vivo, and
each responds even more robustly to the single-
stranded RNA constituents of an siRNA duplex [23].
TLR7 and TLR8 signal through the MyD88 pathway
and induce the production of type I interferons and
inflammatory cytokines [24]. However, the overall
immune responses induced through these receptors dif-
fer because of their unique patterns of expression –
TLR7 is expressed by plasmacytoid dendritic cells and
B cells and mediates interferon-dominated responses,
whereas TLR8 is expressed on myeloid dendritic cells,
monocytes and macrophages, and mediates inflamma-
tory cytokine-dominated responses [25]. TLR3 signals
through a unique adapter called TRIF and is an espe-
cially potent inducer of IFN-b and inflammatory cyto-
kines such as interleukin (IL)-6 and tumor necrosis
factor-a [24,26].
In the cytosol, retinoic acid-inducible gene I (RIG-I)
and melanoma differentiation-associated gene 5 recog-
nize dsRNA and play central but distinct roles in
antiviral defense [27]. However, only RIG-I is known
to be activated by siRNA [28], and this mechanism is
thought to explain many observations of nonspecific
changes in gene expression and interferon production
induced by cytoplasmically localized siRNA. Each
of these innate immune receptors recognizes defined
siRNA molecular features, and some features are rec-
ognized by multiple receptors. The following sections
summarize our current understanding of these recogni-
tion events and how siRNA might be designed to
control immune recognition (Fig. 1).
siRNA design considerations
siRNA sequence
When selecting an siRNA sequence, potency is often
the first consideration, and strategies for selecting a
potent siRNA sequence for a given target mRNA are
discussed in a companion minireview in this issue by
Walton et al. [29]. However, the siRNA sequence also
plays an important role in determining whether a given
S. Samuel-Abraham and J. N. Leonard Controllingnonspecificresponsesto siRNA
FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS 4829
siRNA duplex will induce an innate immune response.
The rules for predicting this property are not entirely
known, but they clearly vary between innate immune
receptors. Activation of TLR7 and TLR8 is particu-
larly dependent onsiRNA sequence. Characteristic
features of RNA such as the presence of uridine resi-
dues and the ribose sugar backbone are necessary for
recognition of RNA by TLR7, and short single-
stranded RNA need only contain several uridines
in close proximity to effectively activate TLR7 [30].
Certain sequence motifs, such as GUCCUUCAA, may
be particularly immunostimulatory [20]. This property
also depends on the overall length of the ligand,
because 19-bp siRNAs containing this motif were more
potent inducers of cytokine production in plasmacy-
toid dendritic cells than were 12- or 16-bp siRNAs that
contain the same motif. Modification of immunostimu-
latory sequences modulates immune stimulation in a
context-dependent manner. In one such example, sub-
stitution of U with A abrogated tumor necrosis factor-
a and IL-6 induction in periperhal blood mononuclear
cells, whereas substituting G with A abrogated only
the induction of IFN-a in plasmacytoid dendritic cells
without affecting the induction of tumor necrosis fac-
tor-a, IL-6 and IL-12 in peripheral blood mononuclear
cells [19]. The overall sequence composition of an
siRNA can also influence its immunostimulatory prop-
erties. Single-stranded RNAs that are GU-rich are
potent ligands for human TLR7 and TLR8, whereas
AU-rich motifs preferentially activate TLR8 [19,22].
However, these features are not necessarily required,
because some siRNAs also activate TLR7 independent
of GU content [20], and sequences that lack G and U
nucleotides can still trigger an immune response [31].
Other innate receptors, such as OAS1, also exhibit
dsRNA sequence-dependent activation [17]. In this
study, synthetic dsRNAs of 19 bp in length were eval-
uated for their capacity to activate OAS1. All OAS1-
stimulating dsRNAs contained the consensus motif,
NNWW(N
9
)WGN (W indicates an A or U), and
mutational analysis confirmed that this motif is
required for activation. The consensus is only 16 nucle-
otides long, because it occurred at various positions
along the 19-nucleotide sequences tested. Interestingly,
these 19-bp dsRNAs are substantially shorter than the
oligonucleotides typically thought to activate OAS1,
which suggests that siRNA might also directly activate
OAS1 by a similar mechanism.
The length of an siRNA is generally an impor-
tant determinant of innate immune activation. Initial
Backbone chemistry
siRNA sequence
End features
Key
Fig. 1. Design considerations forcontrollingnonspecificresponsesto siRNA. This figure summarizes known recognition interactions
between innate immune receptors (green ovals) and siRNA molecular features (blue rectangles), grouped by category of design consider-
ation, and strategies that can be employed to overcome such recognition (red octagons, with interruption of a recognition interaction indi-
cated by red lines). Backbone chemistry modifications at the 2¢ position of ribose moities include deoxy (-H), fluoro (-F) and O-methyl
(-O-Me) substitutions. These substitutions can generally be limited to a subset of sites within the sense strand to balance suppression of im-
munostimulation with retention of capacity to induce RNAi. To some extent, one can select siRNA target sequences that avoid known im-
munostimulatory motifs (the list shown here is representative but not exhaustive). Choosing end features such as 3¢ overhangs and avoiding
5¢ triphosphates reduce immune stimulation by both RIG-I and unknown receptors (i.e. ‘???’). The siRNA image is modified from PDB struc-
ture 2F8S.
Controlling nonspecificresponsestosiRNA S. Samuel-Abraham and J. N. Leonard
4830 FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS
studies indicated that siRNAs shorter than 30 bp could
evade the immune system and thus avoid any off-
target activity [3]. However, subsequent studies indicated
that dsRNA molecules longer than 21 bp can lead to a
sequence-independent interferon response [32]. In some
studies, even 19-bp molecules provided the minimal
length required for immune stimulation [20]. The
length of dsRNA required to activate TLR3 also
remains somewhat uncertain. Our in vitro studies using
TLR3 reporter cells and biophysical measurements
using recombinant TLR3 protein indicated that ligands
shorter than 30 bp neither bind nor activate human
TLR3 [33]. However, shorter siRNAs have been shown
to induce TLR3-dependent inflammation [5,6]. In each
of these cases, it is likely that recognition of siRNA by
multiple receptors may explain some apparent conflicts
between these observations, although this can only be
resolved by elucidating the molecular mechanisms of
siRNA recognition by each receptor.
To date, the rules governing the relationship
between siRNA sequence and the capacity to stimulate
an innate immune response are not yet clear. There-
fore, in practice, controlling innate immune responses
to siRNA still requires a systematic characterization of
the immunostimulatory properties of multiple alterna-
tive siRNA sequences for each given target, or the
implementation of additional strategies for suppressing
immune stimulation.
Backbone chemistry
Naturally occurring nucleoside modifications in mam-
malian RNA appear to provide a mechanism by which
the innate immune system discriminates self-oligonu-
cleotides from those of viral origin [34]. Similarly,
some immune recognition of siRNA may be abrogated
by altering the chemistry of the RNA backbone. To
implement this strategy, one must decide whether to
modify every base in a strand or only selected bases,
and whether to modify just one strand or both strands
in a duplex.
Backbone modification choices are guided in large
part by the mechanism through which an siRNA par-
ticipates in RNAi via the RNA-induced silencing com-
plex (RISC). Only one strand of the siRNA duplex is
incorporated into RISC, and in order to direct RISC
to cleave a target mRNA sequence, the antisense
strand must be incorporated to serve as a template.
Modifications to the backbone chemistry of a
strand may impair its incorporation into RISC, so
siRNA-induced silencing is best maintained if modifi-
cations are confined to the sense strand [35,36].
However, modifications at position 9 of the sense
strand (immediately upstream of the cleavage site) may
inhibit sense strand cleavage, which reduces the
efficiency of RISC assembly and therefore gene silenc-
ing [37]. Although some alterations to the antisense
strand abrogate gene silencing, certain antisense modi-
fications seem to preserve functionality [8,38]. At this
point, the rules for predicting which site and type of
modifications one should use on the antisense strand
to preserve its functionality are not clear [20,36,38].
Furthermore, some immunostimulatory antisense
strands can be made nonstimulatory by modifying the
backbone chemistry of the cognate sense strand (and
only the sense strand) in a duplex [39,40]. This
trans-inhibition of immune activation may indicate
that the receptor involved recognizes the duplex rather
than the component single strands. An additional
advantage to modifying the backbone chemistry of the
sense strand is that by impairing the incorporation of
this strand into RISC, one avoids off-target gene
silencing of mRNAs that are complementary to the
siRNA sense strand.
A variety of siRNA backbone modification chemis-
tries have been investigated for their capacity to sup-
press immune activation while maintaining gene
silencing activity. Because of the requirement of
ribose-containing nucleotides for many types of
immune stimulation [30], one common strategy is to
replace the 2¢-hydroxyl group of the ribose backbone
with 2¢-fluoro, 2¢-deoxy or 2¢-O-methyl groups [23]. In
particular, making such substitutions at uridine resi-
dues often reduces the immunostimulatory capacity of
siRNA [23]. Although strand-wide modifications have
also been investigated for their capacity to block
immune activation [8], such extensive changes are
probably not required. For example, incorporation of
only two 2¢-O-methyl guanosine or uridine residues in
the sense strands of highly immunostimulatory siRNA
molecules was sufficient to abrogate siRNA-mediated
interferon and inflammatory cytokine induction in
human peripheral blood mononuclear cells and in mice
in vivo [39]. In this example, such modifications repre-
sented 5% of the native 2¢-hydroxyl positions in the
siRNA duplex, and no other modifications were
required. Notably, 2¢-O-methyl modification of cyti-
dines was not as effective as the other substitutions in
abrogating the immune response. For dsRNAs that
activate OAS1, 2¢-O-methyl substitution of residues in
the stimulatory motif of the sense strand abolished
OAS1 activation (these positions are presumed to inter-
act with OAS1), whereas similar substitutions on the
opposite strand preserved stimulation of OAS1 [17].
Certain backbone modifications necessitate consider-
ations unique to their particular chemistry. Making
S. Samuel-Abraham and J. N. Leonard Controllingnonspecificresponsesto siRNA
FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS 4831
2¢-deoxy substitutions is equivalent to including DNA
bases in siRNA molecules (except in the case of
2¢-deoxy uridine, which remains distinct from thymi-
dine), and this substitution has been reported to
increase silencing activity [41]. However, it is also pos-
sible that these ligands might activate TLR9, especially
if they contain CpG motifs [42]. A distinct type of
modification is the use of locked nucleic acids, wherein
the ribose contains a 2¢-O, 4¢-C methylene bridge. This
modification renders oligonucleotides resistant to nuc-
leases and may also reduce the immunostimulatory
activity of siRNA [20]. Locked nucleic acid modifica-
tions at the 3¢-termini or both the 3¢- and 5¢-termini of
the sense strand of an siRNA duplex block immune
stimulation but have very little effect on the capacity
of the siRNAto induce RNAi. Conversely, locked
nucleic acid modifications at the termini of the anti-
sense strand do not affect immune stimulation, but
RNAi induction may be impaired or even abrogated
(in the case in which both 5¢- and 3¢-termini of the
antisense strand are modified).
Overall, these findings suggest several general strate-
gies that reduce immune stimulation and preserve
functionality, such as modifying the sense strand of an
siRNA duplex (only) at the 2¢ positions of several
ribose moieties. However, no one strategy is yet uni-
versally applicable. For example, 2¢-O-methyl substitu-
tion of uridines did not prevent siRNA-mediated
activation of TLR3 [5]. For now, some systematic
investigation of possible backbone modifications (or at
least sites to be modified) is required to find the opti-
mal balance between maintaining siRNA efficacy and
preventing nonspecific effects.
End features
The termini (ends) of an siRNA are major determi-
nants of immune recognition. In the context of viral
infections, RIG-I detects viral RNA by binding to its
uncapped 5¢ triphosphate terminus [28]. Maximal acti-
vation of RIG-I requires that the 5¢ triphosphate end
of the dsRNA be blunt [43]. Not suprisingly, siRNAs
that share either or both of these features are also im-
munostimulatory. For this reason, siRNAs transcribed
in vitro from phage polymerases are particularly immu-
nostimulatory unless they are processed to remove 5¢
triphosphates (or the initially transcribed nucleotides
to which these moieties are attached) [44]. Mimicking
the 3 ¢ overhangs that result when DICER processes
long dsRNA into siRNA seems to improve siRNA
properties in several ways. When compared with blunt-
ended oligonucleotides, siRNA with 3¢ overhangs more
efficiently induce gene silencing in vivo [37], and by
adding 3¢ overhangs, otherwise immunostimulatory 27-
bp siRNA can evade immune recognition [13]. Given
our current understanding of these features, end chem-
istry-mediated immune stimulation can generally be
avoided.
Delivery vehicles and strategies
The use of siRNA delivery vehicles is essential for
practical siRNA-mediated silencing because naked
siRNA face rapid degradation in the extracellular
environment and are not efficiently internalized into
cells [45,46]. Various strategies for efficiently delivering
siRNA are discussed in the companion minireview in
this issue by Shim & Kwon [47]. The choice of delivery
strategy also impacts whether an siRNA will induce
innate immune activation.
In trafficking from the extracellular environment,
through endosomal compartments and to the cyto-
plasm, there exist multiple points at which recognition
of siRNA by the innate immune system may occur.
TLR-mediated recognition of siRNA takes place in en-
dosomes. Receptor–ligand interactions are thought to
require this acidic milieu because inhibitors of endoso-
mal maturation, such as bafilomycin, block immune
activation by siRNA via TLR7 and TLR8 [48]. Conju-
gation of siRNAs to cholesterol may enhance cytoplas-
mic delivery, and to some extent, such complexes may
bypass the endosomes without activating endosomal
receptors [46]. Experimentally, direct delivery of
siRNA to the cytoplasm by electroporation may also
suppress an immune response [48]. However, because
siRNA must be released into the cytoplasm in order
for them to be incorporated into RISC, any siRNA
motifs that activate cytoplasmic receptors would still
induce immune activation regardless of the choice of
delivery vehicle. When siRNA are systemically admin-
istered, targeting these molecules to specific cellular
subsets may also reduce stimulation of the innate
immune response in nontargeted cells. For example, a
protamine–antibody fusion protein was designed to
deliver siRNA specifically to tumor cells expressing the
ErbB2 antigen [49]. Although no interferon-induced
gene expression was observed when delivering an anti-
green fluorescent protein (GFP) control siRNAto cells
via a protamine–antibody fusion, is it not possible to
conclude from these experiments that targeting facili-
tated immune evasion.
An siRNA may be immunologically inert when deliv-
ered as a naked siRNA but will stimulate immunity
when complexed with a delivery vehicle. Such effects
have been observed using cationic lipids, such as
N-[1-(2,3-Dioleoyloxy)propyl]-N,N,N-trimethylammonium
Controlling nonspecificresponsestosiRNA S. Samuel-Abraham and J. N. Leonard
4832 FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS
methylsulfate [48] and lipofectamine [35], cationic
polymers such as poly(ethyleneimine) or poly(l-lysine)
[35] and stable nucleic acid–lipid particles [39]. When
interpreting these results, it is important to remember
that complexing siRNA with a delivery vehicle may
have several effects. Because naked siRNAs are not
efficiently taken into cells, some of the enhanced
immune stimulation observed may be due to enhanced
trafficking of siRNAto endosomes and the cytoplasm,
and therefore to enhanced interaction with endosomal
and cytoplasmic receptors. It is also possible that
presentation of siRNA in a large polyvalent complex
makes these ligands more immunostimulatory (to either
endosomal or cytoplasmic receptors) than the free
siRNA would be alone.
An additional consideration is that the subcellular
location at which immune activation occurs determines
the type of immune response that is induced. For
example, ligand-mediated activation of TLR7 (or
TLR9) in endosomal compartments induces type I
interferon production via IRF-7, whereas activation of
TLR7 or TLR9 in lysosomal compartments may
induce inflammatory cytokine secretion via IRF-5
[50,51]. This may be related to the observation that
siRNA complexed with lipofectamine or poly(l-lysine)
(which form large complexes) induces a response domi-
nated by inflammatory cytokines, whereas siRNA
complexed with poly(ethyleneimine) or stable nucleic
acid lipid particles induces a response dominated by
interferon production [35]. It is possible that differ-
ences in intracellular trafficking might explain the dis-
tinct biological effects conferred by these vectors.
Overall, no delivery vehicle is sufficient to confer full
and general protection against siRNA-induced immune
activation, particularly that which is mediated by cyto-
plasmic receptors. It is likely that any delivery vehicle
will need to be paired with other strategies for evading
immune activation.
Concluding remarks
For most mRNA targets, it should be possible to gen-
erate multiple siRNA that induces specific gene knock-
down without inducing nonspecific inhibition of
nontargeted genes. Some strategies can be employed
generally, such as avoiding terminal 5¢ triphosphates
and including 3¢ overhangs [13,28,37,43,44]. Choices of
siRNA sequence are specific to the mRNA targeted,
and although it may be prudent to avoid potent immu-
nostimulatory motifs (such as those known to activate
TLR7, TLR8, and OAS1 [17,20,30]), it may also be
possible to overcome this activation through judicious
modifications to the siRNA backbone [20,23,39]. In
particular, making 2¢-hydroxy substitutions in several
ribose moieties in the siRNA sense strand (such as 2¢-
O-methyl and 2¢-fluoro) may suffice to block recogni-
tion of potentially stimulatory motifs by innate
immune receptors while retaining the capacity to func-
tionally induce RNAi. In practice, selecting these sites
currently requires both avoiding known trouble spots
(i.e. position 9, immediately upstream of the RISC
cleavage site [37]) and experimentally evaluating possi-
ble combinations of backbone modifications. Many
delivery vehicles may enhance immune stimulation by
siRNA [35,39,48], and although others may suppress
some mechanisms of immune activation [46], one can-
not rely upon vehicle choice alone, particularly because
free siRNAs are eventually released into the cyto-
plasm, where they may interact with cytoplasmic
receptors. Strategies that seek to intentionally induce
specific types of immune activation are more challeng-
ing, because in many cases the precise nature of the
immune recognition event is unknown. In general, this
gap in knowledge underlies current limitations on our
ability to predict the immunostimulatory capacity of a
given siRNA design.
In particular, we know little about the mechanisms
by which siRNAs are recognized by the TLRs known
to play central roles in nonspecificresponsesto siRNA
in vivo. TLR7 and TLR8 are thought to mediate the
majority of both inflammatory cytokine and inter-
feron-dominated immune responsestosiRNA in vivo
[25,52], yet we do not know how these receptors recog-
nize siRNA, ssRNA, dsRNA or their commonly used
nucleoside analog ligands. Similarly, TLR3 plays an
important role in siRNA-mediated nonspecific immune
activation [5,6], yet the mechanism by which recogni-
tion of siRNA occurs is also unclear. For example,
some evidence suggests that siRNA-mediated activa-
tion of TLR3 occurs at the cell surface [5,6], yet it is
not clear how the receptor would interact with dsRNA
in this neutral pH milieu. When recognizing longer
dsRNA, an acidic milieu is required so that histidine
residues on TLR3 become positively charged and
interact electrostatically with the negatively charged
backbone of the dsRNA ligand [53–55]. Thus siRNA-
mediated activation might occur via a coreceptor or
via a mechanism distinct from that by which longer
dsRNA is recognized. It is also possible that, at least
in some cases, TLR3-dependent immune activation by
siRNA occurs by indirect mechanisms. For example,
activation of RNaseL by OAS1 (which may itself be
induced by siRNA-mediated production of interferon
through other receptors) produces a pool of self-
derived dsRNA ligands, some of which fall into size
ranges that may activate TLR3 on neighboring cells
S. Samuel-Abraham and J. N. Leonard Controllingnonspecificresponsesto siRNA
FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS 4833
[16,33]. In this way, TLR3 would be an important part
of an siRNA-induced feedback loop even if TLR3 did
not recognize siRNA directly. Another complication is
that cell lines transfected to overexpress TLR3 exhibit
generally enhanced interferon-induced responses [18],
so overexpression of TLR3 may also enhance cytoplas-
mic receptor-mediated responsesto siRNA. Further
investigations are required to elucidate the mechanisms
of these recognition events in order to enhance our
ability to predict, a priori, whether a given siRNA will
activate these potent immune responses.
Given our understanding of the various mechanisms
by which siRNAs induce nonspecific immune responses,
it is essential that appropriate experimental controls be
designed accordingly. Traditionally, control siRNAs
have included target sequences derived from GFP or
luciferase, a random sequence, or a scrambled form of
the test siRNA target sequence. Failing to account for
the nonspecific effects of either the control or the test
siRNA can lead to misinterpretation of experimental
results. This was recently demonstrated in a murine
model of influenza, in which an anti-influenza siRNA
conferred greater antiviral protection than did an anti-
GFP control siRNA [56]. However, this protection was
conferred by nonspecific immune activation, which
appeared to be specific only because the anti-GFP con-
trol siRNA was particularly nonimmunostimulatory.
For these reasons, it is necessary to include experimental
controls that make it possible to differentiate between
the specific and nonspecific effects of a given test
siRNA. For example, in an in vivo model for hepatitis B
virus infection, an unmodified inverted siRNA control
was found to nonspecifically inhibit viral replication [8].
Therefore, both unmodified (potentially immunostimu-
latory) and chemically modified (nonimmunostimulato-
ry) versions of both anti-hepatitis B virus and control
siRNA were tested to evaluate the relative contributions
of specific and nonspecific antiviral effects. Finally,
experiments evaluating whether a particular siRNA (or
siRNA-delivery technology, for that matter) is immuno-
stimulatory must be designed considering that expres-
sion patterns of the innate immune receptors that
recognize siRNA vary between cell types (especially
between immune and non-immune cells), and that rec-
ognition by different receptors and different cells results
in distinct patterns of innate immune responses (e.g.
interferon vs. inflammatory cytokine production). Con-
sidering all known mechanisms of innate immune stimu-
lation by siRNA and working to further advance our
understanding of these recognition events are each of
paramount importance as we continue todesign and
interpret siRNA-based experiments and tap the enor-
mous potential of siRNA-based therapeutics.
Acknowledgements
This work was supported with funding from North-
western University and the Robert R. McCormick
School of Engineering and Applied Science.
References
1 Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE
& Mello CC (1998) Potent and specific genetic
interference by double-stranded RNA in Caenorhabd-
itis elegans. Nature 391, 806–811.
2 Wianny F & Zernicka-Goetz M (2000) Specific interfer-
ence with gene function by double-stranded RNA in
early mouse development. Nat Cell Biol 2, 70–75.
3 Elbashir SM, Harborth J, Lendeckel W, Yalcin A,
Weber K & Tuschl T (2001) Duplexes of 21-nucleotide
RNAs mediate RNA interference in cultured mamma-
lian cells. Nature 411, 494–498.
4 Castanotto D & Rossi JJ (2009) The promises and pit-
falls of RNA-interference-based therapeutics. Nature
457, 426–433.
5 Kleinman ME, Yamada K, Takeda A, Chandrasekaran
V, Nozaki M, Baffi JZ, Albuquerque RJ, Yamasaki S,
Itaya M, Pan Y et al. (2008) Sequence- and target-inde-
pendent angiogenesis suppression by siRNA via TLR3.
Nature 452, 591–597.
6 Cho WG, Albuquerque RJ, Kleinman ME, Tarallo V,
Greco A, Nozaki M, Green MG, Baffi JZ, Ambati BK,
De Falco M et al. (2009) Small interfering RNA-induced
TLR3 activation inhibits blood and lymphatic vessel
growth. Proc Natl Acad Sci USA 106, 7137–7142.
7 Davis ME, Zuckerman JE, Choi CH, Seligson D, Tol-
cher A, Alabi CA, Yen Y, Heidel JD & Ribas A (2010)
Evidence of RNAi in humans from systemically admin-
istered siRNA via targeted nanoparticles. Nature 464,
1067–1070.
8 Morrissey DV, Lockridge JA, Shaw L, Blanchard K,
Jensen K, Breen W, Hartsough K, Machemer L, Radka
S, Jadhav V et al. (2005) Potent and persistent in vivo
anti-HBV activity of chemically modified siRNAs. Nat
Biotechnol 23, 1002–1007.
9 Poeck H, Besch R, Maihoefer C, Renn M, Tormo D,
Morskaya SS, Kirschnek S, Gaffal E, Landsberg J,
Hellmuth J et al. (2008) 5¢-Triphosphate-siRNA:
turning gene silencing and Rig-I activation against
melanoma. Nat Med 14, 1256–1263.
10 Lampson GP, Tytell AA, Field AK, Nemes MM &
Hilleman MR (1967) Inducers of interferon and host
resistance. I. Double-stranded RNA from extracts of
Penicillium funiculosum. Proc Natl Acad Sci USA 58,
782–789.
11 Williams BR (2001) Signal integration via PKR. Sci
STKE 2001, re2.
Controlling nonspecificresponsestosiRNA S. Samuel-Abraham and J. N. Leonard
4834 FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS
12 Manche L, Green SR, Schmedt C & Mathews MB
(1992) Interactions between double-stranded RNA
regulators and the protein kinase DAI. Mol Cell Biol
12, 5238–5248.
13 Marques JT, Devosse T, Wang D, Zamanian-
Daryoush M, Serbinowski P, Hartmann R, Fujita T,
Behlke MA & Williams BR (2006) A structural
basis for discriminating between self and nonself
double-stranded RNAs in mammalian cells.
Nat Biotechnol 24, 559–565.
14 Zhang Z, Weinschenk T, Guo K & Schluesener HJ
(2006) siRNA binding proteins of microglial cells: PKR
is an unanticipated ligand. J Cell Biochem 97, 1217–
1229.
15 Minks MA, Benvin S, Maroney PA & Baglioni C
(1979) Metabolic stability of 2¢ 5 ¢oligo (A) and activity
of 2¢ 5¢oligo (A)-dependent endonuclease in extracts of
interferon-treated and control HeLa cells. Nucleic Acids
Res 6, 767–780.
16 Malathi K, Dong B, Gale M Jr & Silverman RH (2007)
Small self-RNA generated by RNase L amplifies antivi-
ral innate immunity. Nature 448, 816–819.
17 Kodym R, Kodym E & Story MD (2009) 2¢-5¢-Oligoa-
denylate synthetase is activated by a specific RNA
sequence motif. Biochem Biophys Res Commun 388 ,
317–322.
18 Kariko K, Bhuyan P, Capodici J & Weissman D (2004)
Small interfering RNAs mediate sequence-independent
gene suppression and induce immune activation by sig-
naling through toll-like receptor 3. J Immunol 172,
6545–6549.
19 Heil F, Hemmi H, Hochrein H, Ampenberger F,
Kirschning C, Akira S, Lipford G, Wagner H & Bauer S
(2004) Species-specific recognition of single-stranded
RNA via toll-like receptor 7 and 8. Science 303, 1526–
1529.
20 Hornung V, Guenthner-Biller M, Bourquin C, Ablasser
A, Schlee M, Uematsu S, Noronha A, Manoharan M,
Akira S, de Fougerolles A et al. (2005) Sequence-
specific potent induction of IFN-alpha by short
interfering RNA in plasmacytoid dendritic cells through
TLR7. Nat Med 11, 263–270.
21 Gantier MP, Tong S, Behlke MA, Xu D, Phipps S,
Foster PS & Williams BR (2008) TLR7 is involved in
sequence-specific sensing of single-stranded RNAs in
human macrophages. J Immunol 180, 2117–2124.
22 Forsbach A, Nemorin JG, Montino C, Muller C,
Samulowitz U, Vicari AP, Jurk M, Mutwiri GK,
Krieg AM, Lipford GB et al. (2008) Identification of
RNA sequence motifs stimulating sequence-specific
TLR8-dependent immune responses. J Immunol 180,
3729–3738.
23 Sioud M (2006) Single-stranded small interfering RNA
are more immunostimulatory than their double-
stranded counterparts: a central role for 2¢-hydroxyl
uridines in immune responses. Eur J Immunol 36,
1222–1230.
24 Kumar H, Kawai T & Akira S (2009) Pathogen recog-
nition in the innate immune response. Biochem J 420,
1–16.
25 Gorden KB, Gorski KS, Gibson SJ, Kedl RM,
Kieper WC, Qiu X, Tomai MA, Alkan SS & Vasilakos
JP (2005) Synthetic TLR agonists reveal functional
differences between human TLR7 and TLR8. J Immunol
174, 1259–1268.
26 Yamamoto M, Sato S, Hemmi H, Hoshino K,
Kaisho T, Sanjo H, Takeuchi O, Sugiyama M, Okabe
M, Takeda K et al. (2003) Role of adaptor TRIF in the
MyD88-independent toll-like receptor signaling
pathway. Science 301, 640–643.
27 Kato H, Takeuchi O, Sato S, Yoneyama M, Yamamoto
M, Matsui K, Uematsu S, Jung A, Kawai T, Ishii KJ
et al. (2006) Differential roles of MDA5 and RIG-I
helicases in the recognition of RNA viruses. Nature
441, 101–105.
28 Hornung V, Ellegast J, Kim S, Brzozka K, Jung A,
Kato H, Poeck H, Akira S, Conzelmann KK, Schlee M
et al. (2006) 5¢-Triphosphate RNA is the ligand for
RIG-I. Science 314, 994–997.
29 Walton SP, Wu M, Gredell JA & Chan C (2010)
Designing highly active siRNAs for therapeutic applica-
tions. FEBS J 277 , 4806–4813.
30 Diebold SS, Massacrier C, Akira S, Paturel C, Morel Y
& Reis e Sousa C (2006) Nucleic acid agonists for
Toll-like receptor 7 are defined by the presence of
uridine ribonucleotides. Eur J Immunol 36, 3256–3267.
31 Layzer JM, McCaffrey AP, Tanner AK, Huang Z, Kay
MA & Sullenger BA (2004) In vivo activity of nuclease-
resistant siRNAs. RNA 10, 766–771.
32 Sledz CA, Holko M, de Veer MJ, Silverman RH &
Williams BR (2003) Activation of the interferon system
by short-interfering RNAs. Nat Cell Biol 5, 834–
839.
33 Leonard JN, Ghirlando R, Askins J, Bell JK, Margulies
DH, Davies DR & Segal DM (2008) The TLR3
signaling complex forms by cooperative receptor
dimerization. Proc Natl Acad Sci USA 105, 258–263.
34 Kariko K, Buckstein M, Ni H & Weissman D (2005)
Suppression of RNA recognition by Toll-like receptors:
the impact of nucleoside modification and the evolu-
tionary origin of RNA. Immunity 23, 165–175.
35 Judge AD, Sood V, Shaw JR, Fang D, McClintock K
& MacLachlan I (2005) Sequence-dependent stimulation
of the mammalian innate immune response by synthetic
siRNA. Nat Biotechnol 23, 457–462.
36 Prakash TP, Allerson CR, Dande P, Vickers TA,
Sioufi N, Jarres R, Baker BF, Swayze EE, Griffey RH
& Bhat B (2005) Positional effect of chemical
modifications on short interference RNA activity in
mammalian cells. J Med Chem 48, 4247–4253.
S. Samuel-Abraham and J. N. Leonard Controllingnonspecificresponsesto siRNA
FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS 4835
37 Leuschner PJ, Ameres SL, Kueng S & Martinez J
(2006) Cleavage of the siRNA passenger strand
during RISC assembly in human cells. EMBO Rep 7,
314–320.
38 Czauderna F, Fechtner M, Dames S, Aygun H,
Klippel A, Pronk GJ, Giese K & Kaufmann J (2003)
Structural variations and stabilising modifications of
synthetic siRNAs in mammalian cells. Nucleic Acids
Res 31, 2705–2716.
39 Judge AD, Bola G, Lee AC & MacLachlan I (2006)
Design of noninflammatory synthetic siRNA mediat-
ing potent gene silencing in vivo. Mol Ther 13, 494–505.
40 Robbins M, Judge A, Liang L, McClintock K,
Yaworski E & MacLachlan I (2007) 2¢-O-methyl-
modified RNAs act as TLR7 antagonists. Mol Ther 15,
1663–1669.
41 Chiu YL & Rana TM (2003) siRNA function in RNAi:
a chemical modification analysis. RNA 9, 1034–1048.
42 Sen G, Flora M, Chattopadhyay G, Klinman DM,
Lees A, Mond JJ & Snapper CM (2004) The critical
DNA flanking sequences of a CpG oligodeoxynucleo-
tide, but not the 6 base CpG motif, can be replaced
with RNA without quantitative or qualitative changes
in Toll-like receptor 9-mediated activity. Cell Immunol
232, 64–74.
43 Schlee M, Roth A, Hornung V, Hagmann CA,
Wimmenauer V, Barchet W, Coch C, Janke M,
Mihailovic A, Wardle G et al. (2009) Recognition of
5¢ triphosphate by RIG-I helicase requires short blunt
double-stranded RNA as contained in panhandle of
negative-strand virus. Immunity 31, 25–34.
44 Kim DH, Longo M, Han Y, Lundberg P, Cantin E &
Rossi JJ (2004) Interferon induction by siRNAs and
ssRNAs synthesized by phage polymerase. Nat
Biotechnol 22, 321–325.
45 Chiu Y-L, Ali A, Chu C-y, Cao H & Rana TM (2004)
Visualizing a correlation between siRNA localization,
cellular uptake, and RNAi in living cells. Chemistry &
Biology 11, 1165–1175.
46 Soutschek J, Akinc A, Bramlage B, Charisse K,
Constien R, Donoghue M, Elbashir S, Geick A,
Hadwiger P, Harborth J et al. (2004) Therapeutic
silencing of an endogenous gene by systemic administra-
tion of modified siRNAs. Nature 432, 173–178.
47 Shim MS & Kwon YJ (2010) Efficient and targeted
delivery of siRNA in vivo. FEBS J 277, 4814–4827.
48 Sioud M (2005) Induction of inflammatory cytokines
and interferon responses by double-stranded and single-
stranded siRNAs is sequence-dependent and requires
endosomal localization. J Mol Biol 348, 1079–1090.
49 Song E, Zhu P, Lee SK, Chowdhury D, Kussman S,
Dykxhoorn DM, Feng Y, Palliser D, Weiner DB,
Shankar P et al. (2005) Antibody mediated in vivo
delivery of small interfering RNAs via cell-surface
receptors. Nat Biotechnol 23, 709–717.
50 Honda K, Ohba Y, Yanai H, Negishi H, Mizutani T,
Takaoka A, Taya C & Taniguchi T (2005) Spatiotem-
poral regulation of MyD88-IRF-7 signalling for robust
type-I interferon induction. Nature 434, 1035–1040.
51 Asselin-Paturel C & Trinchieri G (2005) Production of
type I interferons: plasmacytoid dendritic cells and
beyond. J Exp Med 202, 461–465.
52 Iwasaki A & Medzhitov R (2004) Toll-like receptor
control of the adaptive immune responses. Nat Immunol
5, 987–995.
53 de Bouteiller O, Merck E, Hasan UA, Hubac S, Ben-
guigui B, Trinchieri G, Bates EE & Caux C (2005) Rec-
ognition of double-stranded RNA by human toll-like
receptor 3 and downstream receptor signaling requires
multimerization and an acidic pH. J Biol Chem 280,
38133–38145.
54 Liu L, Botos I, Wang Y, Leonard JN, Shiloach J,
Segal DM & Davies DR (2008) Structural basis of
toll-like receptor 3 signaling with double-stranded
RNA. Science 320, 379–381.
55 Pirher N, Ivicak K, Pohar J, Bencina M & Jerala R
(2008) A second binding site for double-stranded RNA
in TLR3 and consequences for interferon activation.
Nat Struct Mol Biol 15, 761–763.
56 Robbins M, Judge A, Ambegia E, Choi C, Yaworski E,
Palmer L, McClintock K & MacLachlan I (2008)
Misinterpreting the therapeutic effects of small interfer-
ing RNA caused by immune stimulation. Hum Gene
Ther 19, 991–999.
Controlling nonspecificresponsestosiRNA S. Samuel-Abraham and J. N. Leonard
4836 FEBS Journal 277 (2010) 4828–4836 ª 2010 The Authors Journal compilation ª 2010 FEBS
. MINIREVIEW
Staying on message: design principles for controlling
nonspecific responses to siRNA
Shirley Samuel-Abraham
1
and Joshua N. Leonard
1,2
1 Department. immune activation. Initial
Backbone chemistry
siRNA sequence
End features
Key
Fig. 1. Design considerations for controlling nonspecific responses to siRNA. This