ARTICLE
`
`doi:10.1016/j.ymthe.2005.11.002
`
`Design of Noninflammatory Synthetic siRNA Mediating
`Potent Gene Silencing in Vivo
`
`Adam D. Judge, Gurneet Bola, Amy C.H. Lee, Ian MacLachlan,*
`
`Protiva Biotherapeutics, 100-3480 Gilmore Way, Burnaby, BC, Canada V5G 4Y1
`
`*To whom correspondence and reprint requests should be addressed. Fax: +1 604 630 5103. E-mail: ian@protivabio.com.
`
`Available online 15 December 2005
`
`Targeted silencing of disease-associated genes by synthetic short interfering RNA (siRNA) holds
`considerable promise as a novel therapeutic strategy. However, unmodified siRNA can be potent
`triggers of the innate immune response, particularly when associated with delivery vehicles that
`facilitate intracellular uptake. This represents a significant barrier to the therapeutic development
`of siRNA due to toxicity and off-target gene effects associated with this inflammatory response.
`Here we show that immune stimulation by synthetic siRNA can be completely abrogated by
`selective incorporation of 2V-O-methyl (2VOMe) uridine or guanosine nucleosides into one strand
`of the siRNA duplex. These noninflammatory siRNA, containing less than 20% modified
`nucleotides, can be readily generated without disrupting their gene-silencing activity. We show
`that, coupled with an effective systemic delivery vehicle, 2VOMe-modified siRNA targeting
`apolipoprotein B (apoB) can mediate potent silencing of its target mRNA, causing significant
`decreases in serum apoB and cholesterol. This is achieved at therapeutically viable siRNA doses
`without cytokine induction, toxicity, or off-target effects associated with the use of unmodified
`siRNA. This approach to siRNA design and delivery should prove widely applicable and
`represents an important step in advancing synthetic siRNA into a broad range of therapeutic
`areas.
`
`Key Words: RNA interference, nucleic acid therapeutics, synthetic siRNA, chemical modification,
`inflammatory response, apolipoprotein B, interferons, liposomes
`
`INTRODUCTION
`
`Short interfering RNA (siRNA)1 are double-stranded RNA
`duplexes that silence target gene expression through the
`process of RNA interference (RNAi) [1]. Gene silencing by
`this evolutionarily conserved mechanism results from
`the sequence-specific degradation of mRNA when bound
`within the RNA-induced silencing complex (RISC) by its
`complementary siRNA strand [2,3]. Naturally occurring
`siRNA are derived from the cleavage of long double-
`stranded RNA (dsRNA) by the endonuclease Dicer with
`subsequent unwinding and loading of single strands into
`the RISC. RNAi can also be mediated by the transfection
`of chemically synthesized siRNA [4] and there is currently
`great interest in developing synthetic siRNA that target
`disease-associated genes as therapeutic agents [1]. In this
`
`1Abbreviations used: apoB, apolipoprotein B, TLR, Toll-like receptors,
`sRNAsingle-stranded RNA, dsRNA, double-stranded RNA, siRNA, short
`interfering RNA, IFN-a,
`interferon-a, IL-6,
`interleukin-6, TNF-a, tumor
`necrosis factor-a, RNAi, RNA interference, h-gal, h-galactosidase, PBMC,
`peripheral blood mononuclear cells.
`
`respect, several studies have now reported in vivo silenc-
`ing of endogenous [5] and viral genes [6–9] using
`synthetic siRNA administered via a clinically viable route.
`As part of the innate defense mechanism against
`invading pathogens, the mammalian immune system is
`activated by a number of exogenous RNA [10–12] and
`DNA species [13], resulting in the release of inflamma-
`tory cytokines and interferons (IFN) such as IFN-a.
`Initial studies on the off-target effects of siRNA focused
`on the induction of IFN responses from cell lines, most
`likely through activation of
`the dsRNA-dependent
`protein kinase (PKR)
`[14–16]. Our group [17] and
`others [18,19] have recently demonstrated that syn-
`thetic siRNA can also be potent activators of both the
`murine and the human innate immune system, partic-
`ularly when associated with lipidic or polycation-based
`vehicles that facilitate intracellular delivery via endo-
`somes. Although still poorly defined, immune recog-
`nition of siRNA is sequence dependent [17–19] and
`likely involves signaling through the endosomal Toll-
`like receptor-7 (TLR7) pathway [18]. The consequences
`
`494
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`1525-0016/$30.00
`
`Alnylam Exh. 1079
`
`

`

`doi:10.1016/j.ymthe.2005.11.002
`
`ARTICLE
`
`of activating the innate immune response can be
`severe, particularly in more sensitive species, including
`humans [20,21], due to local and systemic inflamma-
`tory reactions that can be triggered by TLR ligands at
`very low doses. Toxicities associated with the admin-
`istration of siRNA in vivo have been attributed to such
`a response [9,17].
`Stabilization of synthetic siRNA against rapid nuclease
`degradation is often regarded as a prerequisite for in vivo
`and therapeutic applications. This is achieved using
`stabilization chemistries that were previously developed
`for ribozymes and antisense oligonucleotide drugs [22].
`These include chemical modifications to the 2V-OH group
`in the ribose sugar backbone, such as 2V-O-methyl
`(2VOMe) and 2V-fluoro (2VF) substitutions, that are readily
`introduced as modified nucleotides during siRNA syn-
`thesis. A number of reports have shown that siRNA
`containing 2VOMe [23–25], 2VF [24–27], 2V-deoxy [26], or
`blocked nucleic acidQ (LNA) [18,28] modifications can
`retain functional RNAi activity,
`indicating that these
`chemistries can be compatible with the RNAi machinery.
`However, these modifications to siRNA appear to be
`tolerated only in certain ill-defined positional or
`sequence-related contexts as, in many cases, their intro-
`duction has a negative impact on RNAi activity
`[18,23,25,26,28]. Therefore, the design of chemically
`modified siRNA has required a stochastic screening
`approach to identify duplexes that retain potent gene-
`silencing activity.
`Poor uptake of exogenous nucleic acids by cells
`represents an additional barrier to the development of
`siRNA-based drugs. To develop a delivery vehicle that
`enhances intracellular uptake of nucleic acids and is
`suitable for systemic administration, we have encapsu-
`lated siRNA within liposomes termed stable nucleic acid
`lipid particles (SNALP). These systems are effective at
`mediating RNAi in vitro [9,17] and have been shown to
`inhibit viral replication at therapeutically viable siRNA
`doses in a murine model of hepatitis B [9]. These studies
`led to the discovery that chemical modification of the
`siRNA (greater than 90% modified nucleotides), origi-
`nally designed to maximize serum stability and nuclease
`resistance, also abrogated the immunostimulatory prop-
`erties of the unmodified duplex.
`In this report, we demonstrate that only minimal
`2VOMe modifications within one strand of the duplex are
`sufficient to abrogate fully the immunostimulatory
`activity of siRNA, irrespective of its sequence. By restrict-
`ing the degree of chemical modification, noninflamma-
`tory siRNA that retain full RNAi activity can be readily
`generated. Using apolipoprotein B (apoB) as an endoge-
`nous gene target, we show that potent gene silencing can
`be achieved in vivo using these modified siRNA, without
`evidence of cytokine induction, immunotoxicity, or off-
`target effects associated with immune activation trig-
`gered by the unmodified siRNA.
`
`RESULTS
`2V-O-Methyl Modifications within ssRNA Abrogate
`Immune Stimulation
`We have recently shown that synthetic siRNA can
`activate innate immune responses in a sequence-depend-
`ent manner [17] and that this deleterious side effect is
`abrogated in stabilized siRNA with extensive chemical
`modifications (N90% of nucleotides) throughout the
`ribose backbone [9]. To examine the extent and type of
`chemical modification required to inhibit immune cell
`activation by RNA, we selectively introduced 2VOMe
`nucleotides into the GU-rich immunostimulatory motif
`of a single-stranded RNA oligonucleotide (ssRNA) derived
`from a h-galactosidase control (h-gal) siRNA [17]. Oligo-
`nucleotide sequences used in these studies are provided
`in Table 1. 2VOMe modification of the 5 nucleotides
`comprising the immunostimulatory 5V-UGUGU-3V motif
`(2VOMe GU) in the h-gal sense ssRNA completely abro-
`gated IFN-a induction when we treated human periph-
`eral blood mononuclear cell (PBMC) cultures with lipid
`encapsulated ssRNA (Fig. 1A). Inhibition of the interferon
`response was also achieved by selectively modifying
`either the two guanosine (2VOMe 2  G) or the three
`uridine nucleotides (2VOMe 3  U) within the motif. The
`
`TABLE 1: RNA oligonucleotides
`Sequence 5V-3V
`
`Strand
`
`Native (S)
`UUGAUGUGUUUAGUCGCUAUU
`2VOMe GU(S)
`UUGAUGUGUUUAGUCGCUAUU
`2VOMe 3  U(S)
`UUGAUGUGUUUAGUCGCUAUU
`2VOMe 2  G(S)
`UUGAUGUGUUUAGUCGCUAUU
`2VOMe 2  G 3V(S) UUGAUGUGUUUAGUCGCUAUU
`Native (AS)
`*UAGCGACUAAACACAUCAAUU
`2VOMe AC(AS)
`*UAGCGACUAAACACAUCAAUU
`Native (S)
`GUCAUCACACUGAAUACCAAU
`2VOMe U(S)
`GUCAUCACACUGAAUACCAAU
`2VOMe G(S)
`GUCAUCACACUGAAUACCAAU
`2VOMe C(S)
`GUCAUCACACUGAAUACCAAU
`2VOMe A(S)
`GUCAUCACACUGAAUACCAAU
`Native (AS)
`*AUUGGUAUUCAGUGUGAUGACAC
`2VOMe GU(AS)
`*AUUGGUAUUCAGUGUGAUGACAC
`2VOMe U(AS)
`*AUUGGUAUUCAGUGUGAUGACAC
`2VOMe G(AS)
`*AUUGGUAUUCAGUGUGAUGACAC
`Native (S)
`GUGAUCAGACUCAAUACGAAU
`2VOMe U(S)
`GUGAUCAGACUCAAUACGAAU
`Native (AS)
`*AUUCGUAUUGAGUCUGAUCACAC
`2VOMe GU(AS)
`*AUUCGUAUUGAGUCUGAUCACAC
`Native (S)
`GUGGUAUUGUUCCUCCUAAdTdT
`2VOMe GU(S)
`GUGGUAUUGUUCCUCCUAAdTdT
`2VOMe U(S)
`GUGGUAUUGUUCCUCCUAAdTdT
`Native (AS)
`*UUAGGAGGAACAAUACCACdTdT
`2VOMe U(AS)
`*UUAGGAGGAACAAUACCACdTdT
`2VOMe C(AS)
`*UUAGGAGGAACAAUACCACdTdT
`Unmodified (native) and 2V-O-methyl modified (2VOMe) RNA oligonucleotides correspond-
`ing to the sense (S) and antisense (AS) strands of h-gal control, apoB-1, apoB mismatch,
`and vFLIP siRNA are shown. 2VOMe-modified nucleotides are underlined, asterisks represent
`5V phosphates.
`
`Name
`h-Gal
`
`apoB-1
`
`apoB
`mismatch
`
`vFLIP
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`495
`
`

`

`ARTICLE
`
`doi:10.1016/j.ymthe.2005.11.002
`
`inhibitory effect of 2V-O-methylation, however, did not
`appear to require the direct modification of the nucleo-
`tides within the immunostimulatory GU-rich motif since
`selective modification of the two guanosine residues 3V of
`the UGUGU motif, toward the end of the h-gal ssRNA
`(2VOMe 2  G 3V), also resulted in complete abrogation of
`the interferon response in PBMC cultures (Fig. 1A). As
`described previously, the unmodified complementary
`antisense (AS) ssRNA sequence was inherently nonim-
`munostimulatory in these assays [17]. We achieved
`similar results using the transfection reagent polyethyle-
`nimine (PEI) to deliver the RNA to PBMC (Supplementary
`Fig. S1). We applied a similar approach to the modifica-
`
`FIG. 1. 2VOMe modification abrogates ssRNA-mediated interferon induction in
`human PBMC. Liposome-encapsulated, unmodified (native) or 2VOMe U-, G-,
`or GU-modified ssRNA representing the sense (S) and antisense (AS) strands of
`(A) h-gal and (B) apoB-1 siRNA were cultured with PBMC at increasing
`concentrations (5–135 nM). Sequences are detailed in Table 1. IFN-a was
`assayed in culture supernatants at 24 h. Values are means + SD of triplicate
`cultures.
`
`tion of the constituent 21- and 23-base strands of an
`siRNA duplex targeting human and mouse apoB (apoB-1
`siRNA) [5]. As predicted by its GU-rich nucleotide
`sequence [11,17], unmodified apoB-1(AS) ssRNA stimu-
`lated a strong IFN-a response in PBMC cultures, even at
`low concentrations (Fig. 1B). This response was fully
`inhibited by 2VOMe modification of either the 5 nucleo-
`tides comprising the 5V-GUGUG-3V motif (2VOMe GU) or
`the 6 guanosine (2VOMe G) or 7 uridine (2VOMe U)
`residues in apoB-1(AS) ssRNA (Fig. 1B). The unmodified,
`complementary apoB-1 sense oligonucleotide (apoB-1(S))
`encapsulated in lipid particles did not induce IFN-a in
`PBMC (Fig. 1B), although high doses of this oligonucleo-
`tide delivered as PEI polyplexes were found to activate a
`cytokine response (not shown). This weak response to
`PEI-complexed apoB-1(S) ssRNA was also inhibited by
`2VOMe-uridine modification. These findings demonstrate
`that the selective incorporation of 2VOMe-modified
`nucleotides within ssRNA is sufficient to prevent stim-
`ulation of the interferon response from innate immune
`cells.
`
`Selective Nucleotide Modifications within siRNA
`Abrogate Immune Stimulation
`To examine whether selective 2VOMe modifications
`within siRNA duplexes also inhibited immune stimula-
`tion, we generated a series of h-gal and apoB-1 siRNA
`comprising 2VOMe-modified sense or AS strands
`annealed to their complementary unmodified oligonu-
`cleotides (see Table 1). As predicted by results for the
`constituent ssRNA (see Fig. 1A), double-stranded h-gal
`siRNA comprising the 2VOMe-modified UGUGU, 2  G,
`or 3  U sense strand annealed with the unmodified
`(nonimmunostimulatory) AS strand induced no detect-
`able interferon response from human PBMC (Fig. 2A).
`selective 2VOMe modification of
`Interestingly,
`the
`complementary 5V-ACACA-3V motif in the AS strand,
`juxtaposed to the unmodified 5V-UGUGU-3V motif in
`the sense strand, also diminished the level of IFN-a
`induction despite the annealed duplex containing the
`unmodified (immunostimulatory) sense strand (Fig.
`2A). We found similar results with modified apoB-1
`duplex siRNA (Fig. 2B). Unmodified apoB-1 siRNA
`induced a strong IFN-a response in PBMC and this
`reaction was completely abrogated when 2VOMe GU-,
`U-, or G-modified AS strands were incorporated in the
`apoB-1 duplex. Strikingly, modified apoB-1 siRNA
`containing 2VOMe G- or U-modified sense strands
`annealed to the unmodified,
`immunostimulatory, AS
`strand were also rendered nonimmunostimulatory (Fig.
`2B). Abrogation of cytokine induction by 2VOMe G or
`U modifications to the sense strands of modified apoB-
`1 siRNA appeared absolute as even high concentrations
`(675 nM, ~9 Ag/ml) of modified siRNA failed to induce
`IFN-a or inflammatory cytokines such as TNF in PBMC
`cultures (Figs. 2B and 2C).
`
`496
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`

`

`doi:10.1016/j.ymthe.2005.11.002
`
`ARTICLE
`
`FIG. 2. Selective 2VOMe modifications to siRNA
`duplexes abrogate cytokine induction in human PBMC.
`(A, B) Interferon-a and (C) TNF induction from human
`PBMC cultured with increasing concentrations (8–675
`nM) of encapsulated (A) h-gal or (B, C) apoB-1 or apoB
`mismatch siRNA. Cytokine responses to unmodified
`(native) siRNA were compared to duplexes containing
`2VOMe-modified U, G, C, or A residues in either the
`sense (S) or the antisense (AS) strands as indicated (see
`Table 1 for siRNA sequences). Secreted cytokines were
`assayed after 24 h culture. Values are means + SD of
`triplicate cultures.
`
`We did not observe the inhibitory effect of 2V-O-
`methylation on immune stimulation by siRNA with all
`patterns of modification, however, as apoB-1 siRNA
`containing 2VOMe-modified cytidine residues induced
`levels of cytokines similar to those induced by the native
`
`duplex (Fig. 2B). The incorporation of 2VOMe adenosine
`resulted in significant, but not absolute, inhibition of the
`cytokine response. These differences did not simply reflect
`the extent of chemical modification, as the 2VOMe G-, U-,
`C-, and A-modified apoB-1 contain 2, 5, 6, and 8 modified
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`497
`
`

`

`ARTICLE
`
`doi:10.1016/j.ymthe.2005.11.002
`
`FIG. 3. Selective 2VOMe modifications to siRNA duplexes abrogates cytokine induction in vivo. (A, C, E, F) Serum interferon-a and (B, D) TNF and IL-6 levels 6 h
`after intravenous administration of encapsulated (A, B) h-gal, (C, D) apoB-1, (E) apoB mismatch, and (F) vFLIP siRNA. Responses to unmodified (native) siRNA
`were compared to duplexes containing 2VOMe-modified U, G, or C residues in either the sense (S) or the antisense (AS) strands as indicated (see Table 1 for siRNA
`sequences). All mice received 40 Ag encapsulated siRNA. Values are means + SD from three or four animals. Lower levels of quantitation are 75 pg/ml for IFN-a,
`30 pg/ml for TNF, and 60 pg/ml for IL-6.
`
`498
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`

`

`doi:10.1016/j.ymthe.2005.11.002
`
`ARTICLE
`
`nucleotides in the sense strand, respectively. This suggests
`that unmodified U and/or G residues may play a key role in
`immune recognition of the duplex siRNA.
`To confirm that this approach to siRNA design would
`successfully inhibit inflammatory responses to siRNA in
`vivo, we assessed the immunostimulatory activity of the
`2VOMe-modified h-gal and apoB-1 siRNA in mice. Intra-
`venous administration of lipid-encapsulated h-gal (Figs.
`3A and 3B) or apoB-1 (Figs. 3C and 3D) siRNA containing
`2VOMe-modified guanosine or uridine residues in either
`sense or AS strands caused no detectable increase in
`serum IFN-a or inflammatory cytokines such as TNF. This
`was in marked contrast to the unmodified or cytosine-
`modified siRNA that induced substantial elevations in the
`level of these cytokines (Figs. 3C and 3D). We confirmed
`these striking effects of selective 2VOMe modification by
`applying a similar approach to modifying apoB mismatch
`[5] and vFLIP [29] siRNA sequences (see Table 1). For the
`apoB mismatch (Fig. 3E) and vFLIP siRNA duplexes (Fig.
`3F), modifying either the GU-rich regions or only the
`uridine residues in either one of the RNA strands
`completely abrogated cytokine induction by the siRNA
`duplex. Inhibition of the cytokine response to modified
`apoB mismatch siRNA was also confirmed in human
`PBMC cultures (Figs. 2B and 2C). As with apoB-1,
`selective incorporation of 2VOMe cytosine residues into
`vFLIP siRNA did not substantially reduce the IFN-a
`response (Fig. 3F). Thus far, we have consistently
`observed, for each siRNA sequence tested, that the
`introduction of 2VOMe uridine or guanosine residues
`has generated a noninflammatory duplex. Five additional
`examples are provided in Supplementary Figs. S2–S4.
`Taken together, these findings support the conclusion
`that the underlying mechanism for immune recognition
`of short RNA duplexes appears to be largely conserved
`between mouse and humans [17,18]. Our results indicate
`that this mechanism can be profoundly disrupted in
`either species by the incorporation of as few as 2 2VOMe
`
`modified nucleotides within either strand of the RNA
`duplex.
`
`Restricting Modifications to siRNA Sense Strand can
`Help Maintain RNAi Activity
`We assessed the gene silencing activity of native and
`2VOMe-modified apoB-1 siRNA in vitro. Unmodified
`apoB-1 encapsulated within liposomes caused potent,
`dose-dependent inhibition of apoB protein in HepG2
`cell culture supernatants (Fig. 4). Estimated IC50 values
`(~1.5 nM) were in agreement with those established for
`this siRNA sequence using Oligofectamine transfection
`in a similar in vitro model [5]. Modified apoB-1 duplexes
`in which 2VOMe modifications were restricted to the
`nontargeting sense, or passenger, strand displayed apoB
`silencing activity broadly similar to that of the native
`siRNA (Fig. 4). In contrast, modifications to the target-
`ing AS, or guide, strand severely impacted the RNAi
`activity of this duplex. Incorporation of 2VOMe uridine
`or guanosine residues in the AS strand abrogated apoB
`gene silencing, whereas the duplex containing the 5V-
`GUGUG-3V modified AS strand displayed substantially
`reduced activity compared to the native or sense
`modified duplexes (estimated IC50 ~15 nM). Unmodi-
`fied or modified apoB mismatch control siRNA yielded
`no significant inhibition of apoB protein expression
`(Fig. 4). A similar strategy of restricting 2VOMe mod-
`ifications to the sense strands of h-gal 728 and
`luciferase siRNA also proved successful
`in generating
`noninflammatory siRNA that retained full RNAi activity
`(Supplementary Figs. S3 and S4). Although frequently
`reported to occur, it is not yet possible to predict how a
`particular pattern of chemical modifications may neg-
`atively impact RNAi activity [18,23,25,26,28]. Consis-
`tent with our findings on apoB-1, 2VOMe modification
`to the AS strand of an siRNA duplex, particularly at the
`5V end, has been shown to reduce its RNAi activity [25].
`However, other groups have identified siRNA sequences
`
`FIG. 4. In vitro silencing of apoB expression by 2VOMe-
`modified siRNA. HepG2 cells were treated with
`encapsulated apoB-1 or mismatch siRNA at the
`indicated concentrations (0.6–45 nM). Unmodified
`(native) apoB-1 siRNA was compared to apoB-1
`duplexes containing 2VOMe-modified U, G, or C
`residues in the sense (S) strand or GU motif, U, or G
`residues in the antisense (AS) strand as indicated (see
`Table 1 for modified siRNA sequences). Unmodified
`and 2VOMe U(S) apoB mismatch siRNA served as
`control duplexes. ApoB protein in culture superna-
`tants was measured by ELISA after 48 h. ApoB levels
`are expressed as % of PBS-treated control cultures.
`Each value is derived from means of duplicate cultures
`and is representative of three separate experiments.
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`499
`
`

`

`ARTICLE
`
`doi:10.1016/j.ymthe.2005.11.002
`
`that can tolerate extensive 2VOMe modifications to the
`AS strand [8,23]. Our results suggest that selective 2VOMe
`modifications, restricted to the sense strand of siRNA,
`offer a robust approach to overcoming the problem of
`immune activation by siRNA while reducing the chance
`of negatively impacting RNAi activity. It is envisaged that
`this approach can be applied to many, if not all, siRNA
`sequences with inherent capacity to stimulate the innate
`immune response. In our experience, this would encom-
`pass the majority of conventionally designed synthetic
`siRNA.
`
`Potent RNAi Activity without Immune Stimulation
`in vivo
`We assessed 2VOMe-modified apoB-1 siRNA for their
`ability to silence gene expression in vivo. We selected
`2VOMe U(S)- and GU(AS)-modified apoB-1 as noninflam-
`matory duplexes (see Figs. 2 and 3). This also afforded the
`opportunity to assess the impact of chemical modifica-
`tions that reduced in vitro RNAi activity of the AS
`modified siRNA (see Fig. 4). We formulated native or
`2VOMe-modified apoB-1 and mismatch siRNA in SNALP,
`shown previously to deliver siRNA to the liver [9]. For use
`in systemic applications, nucleic acid-based drugs require
`stabilization or protection from nuclease degradation.
`Encapsulation inside the lipid bilayer protected unmodi-
`fied and otherwise labile siRNA from serum nuclease
`degradation for greater than 24 h at 378C in vitro,
`implying that encapsulation offers adequate nuclease
`protection without the need for extensive chemical
`modification to the siRNA. By comparison, naked siRNA
`was fully degraded within 4 h under similar conditions
`(Fig. 5).
`We administered encapsulated apoB siRNA intrave-
`nously to BALB/c mice at 5 mg/kg/day for 3 days. This
`regimen represents a 10-fold reduction in apoB-1 siRNA
`dose originally reported to be efficacious in experiments
`utilizing cholesterol-conjugated, chemically modified
`apoB-1 siRNA [5]. Animals receiving native, immunosti-
`mulatory apoB-1 or mismatch siRNA displayed overt
`symptoms of toxicity as evidenced by a loss of 10.5 and
`9% of initial body weight, respectively, by day 3 (Fig. 6A)
`and mild deterioration in general body condition during
`the course of treatment (not shown). In contrast, treat-
`ment with the 2VOMe-modified siRNA was well tolerated
`with minimal (less than 1%) or no body weight loss (Fig.
`6A). We confirmed abrogation of the innate cytokine
`response in these efficacy studies by in-life serum IFN-a
`analysis (Fig. 6B), and accordingly we attribute the
`toxicity associated with administration of the unmodi-
`fied siRNA to the systemic cytokine response. Of note,
`cytokine levels and body weight loss induced by unmodi-
`fied mismatch siRNA were lower than for the correspond-
`ing active apoB-1 duplex. The mismatch control in this
`case was generated by four G/C substitutions within the
`apoB-1 sequence [5], providing further evidence for the
`
`FIG. 5. Encapsulation of siRNA in lipid particles protects against serum
`nuclease degradation. Unmodified naked (top) or SNALP-encapsulated
`(middle) apoB-1 siRNA were incubated in 50% mouse serum at 378C. Duplex
`integrity was assessed at indicated time points by nondenaturing PAGE
`analysis. Addition of Triton X to disrupt lipid particle integrity (bottom)
`restored siRNA nuclease sensitivity.
`
`sequence-dependent effects on immune stimulation by
`RNA duplexes.
`As a direct measure of RNAi-mediated knockdown, we
`determined apoB mRNA in the liver 2 days after final
`siRNA treatment (Fig. 6C). In both the native and the
`2VOMe U(S)-modified apoB-1-treated groups, apoB mRNA
`levels were significantly reduced compared to PBS-treated
`animals (18 F 2 and 18 F 5% of PBS controls, respec-
`tively). By comparison, mice treated with 2VOMe GU(AS)-
`modified apoB-1 siRNA displayed less pronounced silenc-
`ing of apoB mRNA (44 F 4% of controls), which
`correlated with reduced in vitro RNAi activity of this
`modified siRNA (see Fig. 4). ApoB mRNA levels in the
`modified mismatch group were equivalent to those in
`PBS controls (Fig. 6C), while the native mismatch siRNA
`caused a modest reduction in apoB mRNA levels (79 F
`12% of PBS controls). The modest reduction in liver apoB
`mRNA observed with the native mismatch siRNA was
`apparent in three separate experiments (not shown) and
`correlated with interferon release and symptoms of
`toxicity associated with systemic administration and
`delivery of the unmodified siRNA.
`Silencing of apoB mRNA in the liver resulted in
`proportional, sequence-specific reductions in serum
`apoB protein. Mice treated with native, 2VOMe U(S)- or
`GU(AS)-modified apoB-1 siRNA had serum apoB protein
`levels that were 26, 28, and 47% those of the PBS-treated
`animals, respectively (Fig. 6D). Functional silencing of
`apoB expression was reflected in significant reductions
`in serum cholesterol that also correlated with the
`relative potency of mRNA and protein knockdown. Mice
`treated with native or 2VOMe U(S)- or GU(AS)-modified
`apoB-1 siRNA displayed serum cholesterol
`levels that
`were 48, 51, and 69% of cholesterol levels in the PBS
`control group (Fig. 6E). Neither mismatch siRNA had
`any effect on serum cholesterol (Fig. 6E). In separate
`experiments, the noninflammatory 2VOMe G(S)-modi-
`fied apoB-1 mediated similar reductions in apoB mRNA,
`
`500
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`

`

`doi:10.1016/j.ymthe.2005.11.002
`
`ARTICLE
`
`FIG. 6. Silencing of apoB expression in vivo without activation of the innate immune response. (A–E) In vivo effects following intravenous administration of
`encapsulated apoB-1 or mismatch siRNA in mice. Mice were treated on days 0, 1, and 2 with encapsulated unmodified (native) or 2VOMe U(S)- or GU(AS)-
`modified apoB-1 and native or 2VOMe U(S)-modified mismatch siRNA at 5 mg/kg per day. (A) Daily changes in body weight (% of day 0 weight) of apoB-1 (solid
`symbols) and mismatch (open symbols) siRNA-treated mice over the 4-day study period. (B) Serum IFN-a from test bleeds 6 h after initial treatment; lower level
`of quantitation 75 pg/ml. (C) apoB mRNA levels in liver, (D) apoB protein in serum, and (E) serum cholesterol (mM) 2 days after final siRNA treatment. C and D
`are expressed as % of apoB mRNA and apoB protein compared to PBS-treated animals. All values are means + SD of five animals. All data are representative of
`two separate experiments. Statistical significance versus PBS-treated animals: *P b 0.05, **P b 0.005, ***P b 0.00005; unpaired two-tailed t test on group values
`prior to normalization to PBS.
`
`MOLECULAR THERAPY Vol. 13, No. 3, March 2006
`Copyright C The American Society of Gene Therapy
`
`501
`
`

`

`ARTICLE
`
`doi:10.1016/j.ymthe.2005.11.002
`
`protein, and serum cholesterol, in the absence of IFN
`induction (not shown).
`Results from these studies demonstrate that lipid
`encapsulation of siRNA provides adequate serum stabil-
`ity for systemic applications and negates the need for
`extensive chemical modifications to the RNA. This,
`coupled with the effective delivery of the siRNA pay-
`load to the target organ, in this case the liver, facilitates
`the silencing of endogenous genes. This is exemplified
`in these studies by apoB, a protein that represents a
`potential therapeutic target for hypercholesterolemia.
`Importantly, the 2VOMe-modified siRNA, designed to be
`noninflammatory, displayed potency in vivo that
`is
`equivalent to that of the unmodified siRNA but with-
`out the immunotoxicity and other off-target effects
`associated with the systemic administration of unmodi-
`fied siRNA. We believe that
`this approach will be
`generally applicable to a wide range of gene targets
`and shows great promise for use in a therapeutic
`setting.
`
`DISCUSSION
`Although naked siRNA can trigger RNA interference in
`vivo without apparent IFN induction [30], recent reports
`have shown that, when formulated with lipidic or
`polycation-based delivery systems, unmodified siRNA
`can be potent activators of the innate immune response
`[17–19]. TLR7, located within the endosomal compart-
`ment, has been implicated as the primary immune
`recognition pathway for siRNA [18] and this may, in
`part, explain why delivery vehicles that enter cells
`through the endosomal pathway can potentiate this
`response. Based on the finding that immune activation
`by siRNA is sequence dependent, we have shown
`previously that it is possible to design active siRNA with
`negligible immunostimulatory activity by selecting
`sequences that lack GU-rich motifs [17]. Although this
`strategy has proven successful, it significantly limits the
`number of novel siRNA sequences that can be designed
`against a given target. Furthermore, it currently requires
`some degree of screening due to the relatively ill-defined
`nature of putative RNA immunostimulatory motifs. In
`this report, we highlight a novel and robust approach to
`abrogating synthetic siRNA-mediated immune stimula-
`tion by selective incorporation of 2VOMe-modified
`nucleotides into the siRNA duplex. Remarkably, incorpo-
`ration of as few as two 2VOMe guanosine or uridine
`residues in highly immunostimulatory siRNA molecules
`completely abrogated siRNA-mediated interferon and
`inflammatory cytokine induction in human PBMC and
`in mice in vivo. This degree of chemical modification
`represents ~5% of the native 2V-OH positions in the siRNA
`duplex. Since complete abrogation of the immune
`response required only one of the RNA strands to be
`selectively modified, 2VOMe modifications could be
`
`restricted to the sense strand of the duplex, thereby
`minimizing the potential for attenuating the potency of
`the siRNA. These findings have provided a simple ration-
`ale for the synthesis of nonimmunostimulatory siRNA
`based on native sequences with proven RNAi activity. We
`demonstrate that by combining selectively modified
`siRNA with an effective systemic delivery vehicle, potent
`silencing of an endogenous gene target can be achieved
`in vivo at therapeutically viable doses without the
`deleterious side effects associated with systemic activa-
`tion of the innate immune response.
`It is unclear at present how the introduction of
`2VOMe nucleotides into siRNA prevents recognition of
`the duplex RNA by the immune system. Although there
`is evidence from gene-deficient mice that TLR7 is
`required for siRNA-mediated immune stimulation [18],
`the nature of the physical
`interaction between the
`receptor and the putative ligand is not known. This is
`also true for the defined TLR7/8 ligands that include
`guanosine analogues [31] and ssRNA containing GU or
`poly(U) motifs [11,12]. The mammalian immune system
`comprises a number of alternative signaling pathways
`that can also be activated by RNA species [10,32–35],
`and studies in cell lines suggest that siRNA may be a
`ligand for TLR3 [36] or dsRNA-dependent PKR [14,16]
`under certain conditions. Further work is therefore
`required to confirm the precise mechanism of siRNA-
`mediated immune activation before the molecular basis
`underlying the inhibitory effects of 2VOMe modification
`can be fully elucidated. Since the 2V-OH in the ribose
`backbone is a distinguishing feature of RNA, extensive
`chemical substitutions at this position may be antici-
`pated to disrupt recognition of the modified nucleic acid
`by an RNA binding receptor pathway. However, our
`studies show that 2VOMe-modified siRNA are rendered
`nonimmunostimulatory despite retaining up to 95% of
`their native ribonucleotides, including those comprising
`defined immunostimulatory regions of the RNA. 2VOMe
`is considered to be a relatively bulky chemical group at
`the 2V position that sits within the minor groove of an
`RNA duplex without significantly distorting its A-form
`helical structure [26,37]. This may be s