Nielsen H. RNA: Methods and Protocols

Подождите немного. Документ загружается.

174 Sioud

post-transcriptional gene silencing (PTGS) had been described

years ago in plants, and now it is believed to function as a surveil-

lance system for blocking the function of har mful RNAs such

as viral RNAs (2). Remarkably, RNAi is systemic in both plants

and nematodes, spreading from cell to cell. In C. elegans,RNAi

is also heritable: silencing can be transferred to the progeny of

the worm originally injected with the trigger dsRNA. Viral infec-

tion, inverted repeat transgenes, or aberrant transcription prod-

ucts all lead to the production of dsRNA that can be converted to

siRNAs.

It should be noted that the work of Hamilton and Baulcombe

has provided the ﬁrst decisive ﬁndings showing that a distinct

population of 21- to 24-nucleotide (nt)-long RNAs with anti-

sense sequence of silenced genes invariably accumulated in co-

suppressed plants (3). Their work also suggests that 21–24 nt

RNA duplexes could provide sequence speciﬁcity to the machin-

ery that degrades homologous mRNAs in RNAi/PTGS (3).

Unlike invertebrates, vertebrates react to long dsRNA by activat-

ing the interferon pathway (4). However, it has been shown that

chemically made small RNAs, known as siRNAs, with features

of Dicer cleavage products were sufﬁcient to mediate RNAi in

human cells without activating the IFN r esponse (5). Subsequent

to this landmark ﬁnding by Tuschl and colleagues in 2001, siRNA

technology is now used as a standard tool in most laboratories for

gene function analysis and drug-target discovery and validation.

Furthermore, siRNAs have emerged as a powerful tool for ther-

apeutic gene silencing (6, 7). In principle, the mRNA encoding

any protein that is associated with a disease can be cleaved selec-

tively by siRNAs. However, recent data make it clear that siRNA

faces some major hurdles before it can used as a drug.

2. RNAi Pathway

In general the RNAi pathway is initiated by the enzyme Dicer,

which cleaves long dsRNA into double-stranded siRNA. Dicer

is a dsRNA-speciﬁc RNAase III family ribonuclease, which gen-

erates siRNA duplexes containing of 20–25 nt in length. Dicer

leaves 2-nt 3



-overhangs and 5



-phosphate groups in each strand.

These siRNA duplexes are then incorporated into a multiprotein

complex, the RNA-induced silencing complex (RISC). Notably,

synthetic siRNAs enter directly into the RNAi pathway (see

Fig. 12.1). Subsequent to strand separation, the antisense (guide)

strand guides the RISC to recognize and cleave target mRNA

sequences. The catalytic activity of RISC is mediated by Arg-

onaute 2 (Ago2) protein, the only Ago family member that is

Promises and Challenges in Developing RNAi as a Research Tool and Therapy 175

Sense strand-mediated

mRNA recognition

mRNA cleavage

Synthetic

siRNA duplexes

5’p

3’

5’

mRNA

5’ (A)n 3’

(A)n 3’

5’

Degradation by

cellular nucleases

Interferon

response

pathway

3’

5’

3’

Ago2

TRBP

5’p

3’

5’p

3’

Low stability

(AU-rich)

High stability

(GC-rich)

RISC Assembly

RISC

recycling

Degradation of

the sense strand

Fig. 12.1. Schematic representation of gene silencing by siRNAs. In contrast to long double-stranded RNAs, siRNAs are

directly loaded into a multi-protein complex termed RNA-induced silencing complex (RISC), where the sense (passenger)

strand with high 5



-stability) is cleaved by the nuclease AGO2. This will lead to strand separation. Subsequently, the

RISC containing the antisense (guide) strand seeks out and binds to complementary mRNA sequences. Bound mRNA

molecules are then cleaved by AGO2 and cleaved mRNA fragments are rapidly degraded by cellular nucleases. Following

dissociation, the active RISC is able to recycle and cleave additional mRNA molecules.

cleavage competent (8, 9). The protein members of the arg-

onaute family are highly basic proteins that contain two com-

mon domains, PAZ and PIWI domains. While the PA2 domain

is responsible for RNA binding, the PIWI domain mediates the

interaction with Dicer and contains the nuclease activity that

cleaves of target mRNAs. Ago2 is also responsible for cleavage

of the passenger siRNA strand, thus facilitating the formation

of functional RISC complexes (10, 11). Analysis of the crystal

structures of a siRNA guide strand associated with Ago2 PIWI

domain revealed that nucleotides 2–8 form a seed sequence that

directs target mRNA recognition by RISC (12). While mammals

and C. elegans each have a single Dicer that makes both miR-

NAs and siRNAs, Drosophila has two Dicers (13): Dicer-1 makes

miRNAs, whereas Dicer-2 is specialized for siRNA production.

Although the discovery of RNAi has provided a new tool to study

gene function and drug target validation, recent studies have

shown that siRNA duplexes can activate innate immunity and

silence a variety of genes in addition to the intended target gene

(14–17). Therefore, the development of strategies that block

siRNA unwanted effects is crucial to their therapeutic use. Also,

176 Sioud

there remain other important obstacles for effective therapeutic

use of siRNA, including stability and delivery.

3. Design Rules

Gene silencing by siRNA varies markedly in mammalian cells,

where the gene-silencing effectiveness depends very much on the

target sequence positions (sites) selected from the target gene

(18). Since individual siRNAs Vary extremely in their effeciency

to induce down-regulation of the target, several siRNAs have to

be designed for each target gene and evaluated for their efﬁcacy

and lack of side effects (see below). The earliest guidelines for

siRNA design were proposed by Elbashir and colleagues (5). They

suggested that synthesizing siRNA duplexes of 19 base-paired

nucleotides with 2-nt 3



-overhangs at either end (referred to as

21-nt siRNA duplexes) mediates efﬁcient cleavage of the target

mRNA. Preferably, the 19-nt target sequence should be ﬂanked

in the mRNA with two adenosines at the 5



-end. Regions at the

mRNA to select the target site are preferably in the coding region,

100 bp from the AUG start codon. Despite these early rules, the

siRNA efﬁcacy is highly dependent on target sequence and siRNA

efﬁcacy cannot be predicted from the primary sequence. There-

fore, a successful siRNA research project may require the design

of various distinct siRNAs at a high cost. More recently, some

studies showed that the 5



-end of the antisense strand might be

incorporated into the RNA-induced silencing complex and strand

incorporation may depend on weaker base pairing, indicating an

A-U terminus may lead to more strand incorporation than a G-C

terminus (19, 20). Therefore, it was concluded that the relative

thermodynamic stability of the 5



-ends of the two siRNA strands

in the duplex determines the identity of the selected strand,

either the antisense (guide) or the sense (passenger) strand. Other

factors reported to be involved in gene-silencing efﬁcacy are GC

content, point-speciﬁc nucleotides, speciﬁc motif sequences, and

secondary structures of mRNAs ( 21). Based on the published

data, several siRNA design rules/guidelines using efﬁcacy-related

factors have been r eported. Although there are common pre-

ferred and unpreferred nucleotides at both position 1 and 19

in some guidelines, there is the problem of inconsistencies for

nucleotide frequency of each position suggesting that some rules

from guidelines are sequence-dependent.

Despite these design limitations, the experimental data indi-

cated four apparent features of the siRNA sequence to possibly

serve to discriminate active siRNAs from those that are non-active

(21). First, the 5



-end of the antisense strand ( AS) of active siR-

Promises and Challenges in Developing RNAi as a Research Tool and Therapy 177

NAs may always be A or U, with the counterpart of non-active

siRNAs may always be C or G. Second, the 5



-end of the sense

strand (SS) of active siRNAs are preferably G or C, with the coun-

terpart of non-active siRNAs being A or U. Third, in the case of

active siRNAs, the 5



-terminal AS are A/U-rich whereas the cor-

responding region of non-active siRNAs are G/C rich. Fourth,

position 10 of the target site should be U.

In addition to the siRNA sequence, mRNA secondary struc-

ture is also considered to be an important factor in predicting

siRNA efﬁcacy. However, there are conﬂicting results concern-

ing the effects of secondary structures on siRNA functionality.

On one hand, some studies suggested that the secondary struc-

ture of the mRNA plays a role in determining the efﬁcacy of gene

silencing. On the other hand, other studies did not ﬁnd any cor-

relation between functionality of the siRNA and second structure

of the target mRNA. Therefore, this issue still requires further

investigation.

4. Detection of

Exogenous RNAs

by Innate Immune

Receptors

Notably, the immune system has evolved cellular and molecular

strategies to discriminate between foreign and self nucleic acids.

Among the cytoplasmic sensors of long dsRNA is the dsRNA-

dependent protein kinase (PKR) that phosphorylates serine and

threonine residues of target proteins (22). Most human cells con-

stitutively express a low level of PKR that remain inactive. How-

ever, upon binding to dsRNA, PKR forms a homodimer result-

ing in its autophosphorylation and activation. Activated PKR

phosphorylates various substrates including the protein synthe-

sis initiation factor elF-2α and blocks the translation of viral and

cellular proteins, an essential step in antiviral resistance. It should

be noted that PKR’s binding to dsRNAs is sequence-independent

and the pr esence of interferon upregulates its expression.

Although PKR is implicated in antiviral immunity, it is mainly

IFN effector and not absolutely required for IFN production.

More recently two additional intracellular helicases, retinoic-

acid-inducible gene I (RIG-I) and/or melanoma differentiation-

associated gene 5 (MDA-5) were identiﬁed (23). RIG-I encodes

a caspase recruitment domain (CARD) at the N terminus, in

addition to an RNA helicase domain. The RNA helicase domain

requires ATPase activity and is responsible for viral dsRNA r ecog-

nition and induction of conformational changes leading to the

interaction of the RIG-I CARD domain with another CARD-

containing adaptor protein, known as IPS-1, MAVS, Cardif, or

VISA (24). IPS-1 is an outer mitochondrial membrane binding

178 Sioud

protein, which activates IRF3 and IFR-7 through TBK1/IKKi,

resulting in the production IFN-β production. Mitochondrial

retention of IPS-1 is essential for IRF3, IRF7, and NF-κB acti-

vation by RIG-1 (24).

Although RIG-I seems to be an essential sensor of viral RNAs,

microbial nucleic acids are also recognized by toll-like recep-

tors (TLRs), especially in immune cells (25, 26). Whereas most

TLRs ar e expressed in the plasma membrane for detecting bacte-

rial components, TLR3, TLR7, TLR8, and TLR9, are expressed

in intracellular compartments (endosomes, lysosomes) (25). The

immune function of this cellular localization is to sense viral

RNAs. TLR3 is also expressed on the cell surface and it is believed

to recognize extracellular viral dsRNAs (27). TLR7 and TLR8

recognize viral ssRNA and small synthetic antiviral compounds

referred to as imidazoquinolines. TLR9 recognizes unmethylated

CpG-DNA motifs that exist in both viral and bacterial DNA, but

are suppressed or methylated in the vertebrate genomes (28, 29).

It should be noted that intracellular NOD-like receptors detect

bacteria, whereas viruses are mainly detected by TLRs and RIG-

like receptors. The virus-detecting TLRs operate mainly in plas-

macytoid dendritic cells by responding to viral nucleic acids that

enter the cell via e ndocytosis. In these cells, the major immune

response is the production of type 1 interferon (30).

Although siRNAs were initially thought to bypass the IFN

response because they are too short to be recognized by dsRNA

sensors (5), we and others have shown that they could acti-

vate innate immunity in mammalian cells (14–17). Early stud-

ies indicated sequence-independent activation of PKR and TLR3

signaling pathways by siRNAs (17, 31). However, recently it

was demonstrated that PKR and TLR3 do not represent the

major pathways by which chemically synthesized siRNAs activate

immunity in immune cells (32–35). Indeed, a group of siRNA

sequences stimulated monocytes or dendritic cells to produce

proinﬂammatory cytokines and type I interferons. This response is

mainly mediated through TLR7 in mice and TLR7/8 in humans.

Under our experimental conditions, ss siRNAs were more effec-

tive than ds siRNAs in activating TLR7/8 in human monocytes

and peripheral blood mononuclear cells (PBMCs) (32, 35). The

extent of siRNA unwanted effects has r ecently been conﬁrmed by

expression proﬁling using microarrays, which identiﬁed over 400

siRNA-affected genes in PBMCs. Genes encoding for proinﬂam-

matory cytokines, interferons, and Mx proteins are among the

genes that are signiﬁcantly induced (36). Mx proteins are IFN-

induced GTPases that form complexes with dynamin disrupting

trafﬁcking or activity of viral polymerases, thereby interfering with

viral replication.

Promises and Challenges in Developing RNAi as a Research Tool and Therapy 179

5. What Is the

Nature of

IFN-Inducing

Motif Present in

One Sequence But

Absent in

Another?

TLR7 and TLR8 recognize certain siRNA sequences, provided

they are delivered to the endosomes via cationic liposomes. Ini-

tial experiments indicate that some types of secondary structures

and/or speciﬁc nucleotides are responsible for the activation of

NF-κB signaling pathway by siRNAs in human monocytes (14).

Judge and colleagues found that the 5



-UGUGU motif was indis-

pensable for the immune activation by a siRNA in human blood

cells (34). However, Hornung and colleagues identiﬁed a 9-

nt motif RNA motif (5



-GUCCUUCAA) that is recognized by

TLR7 in the context of siRNA duplexes and the activity does

not depend on GU content (33). Collectively, our data indicated

that interferon induction by siRNAs can not be easily suppressed

by selecting siRNA sequences without the GU dinucleotides.

Indeed, several siRNA sequences without GU induced TNF-α

production in human PBMC and monocytes (32). Although the

precise nature of the RNA motifs responsible of innate immune

activation is not known, we showed that the ability of ssRNA or

dsRNAs to activate TNF-α production is largely depend on the

uridine content (35). Indeed, their replacement with adenosines

abrogated immune activation by either ss or ds siRNAs. A recent

study by Goodchild and colleagues (37), where they analyzed

the effects of 250 siRNA sequences, conﬁrms the importance of

uridines for siRNA stimulation. Also, their data supported our

proposed model in which ds siRNAs are dissociated in the endo-

somes, leading to the activation of TLR7/8 by ss siRNAs (32).

6. The Molecular

Basis of RNA

Sensing by RIG-I

As indicated above, recent studies on the immune response

to chemically made siRNAs have highlighted the involvement

of the endosomes. Indeed, cytoplasmic delivery of synthetic

siRNA by electroporation into human blood cells did not induce

either inﬂammatory cytokines or interferons, whereas the same

sequences when delivered by lipid did (32). Thus, synthetic siR-

NAs are not detected by cytoplasmic sensors for viral RNAs in

immune cells. However, the data do not explain why ds siR-

NAs were not sensed by RIG-1, a major viral RNA sensor. It has

been demonstrated that chemically made siRNA duplexes harbor-

ing 2-nt 3



-overhangs cannot engage innate immunity activation,

whereas the same siRNA sequence with blunt ends did (38). The

authors showed that RIG-1 can bind to siRNAs with or without

180 Sioud

2-nt 3



-overhangs, but only siRNAs with blunt ends could acti-

vate RIG-1. These ﬁndings imply that endogenous shRNAs or

microRNA harboring the Dicer signature, 2-nt 3



-overhang, are

not an ideal stimulator of RIG-I. Thus, the structures of the 5



ends between shRNAs (substrate for Dicer) and non-self dsRNAs

such as viral RNAs are critical for self and non-self discrimination.

It should be noted that the predominant form of naturally

occurring dsRNAs in mammalian cells is derived from endoge-

nously expressed miRNAs that constitute a large class of noncod-

ing small RNAs involved in gene regulation in a variety of organ-

isms ranging from plants to mammalians (39). Presently, more

than 1,000 potential human miRNAs have been identiﬁed and

numerous have been experimentally validated. Usually miRNAs

are transcribed from endogenous genes by RNA polymerase II

as long RNA precursor called a primary miRNA (pri-miRNA),

containing one or more distinct miRNAs. In the nucleus the

RNA pr e cursors are processed by Drosha to 60–80 nt RNA hair-

pin intermediate, bearing 2-nt 3



-overhang, called a pre-miRNA.

Interestingly, the Drosha cleavage site was shown to be 11 base

pairs from the stem single-stranded RNA junction (40). Pro-

cessed pre-miRNAs are then transported from the nucleus to

the cytoplasm by exportin-5, where its 2-nt 3



-overhang is rec-

ognized by Dicer, which generates the mature miRNA that can

evade the detection by innate immune receptors such as RIG-I

(see Fig. 12.2).

During our studies, we have also found that synthetic ss

siRNAs (21 nt) do not activate innate immunity when delivered

to the cytoplasm via electroporation (32). These used RNAs do

not contain 2-nt 3



-overhang because they are single-stranded.

To further examine the contribution of RIG-I in sensing exoge-

nous RNAs, we have transfected monocytes with either T7 RNA

polymerase (RNAP) transcribed siRNAs or chemically made

siRNAs. The inhibition of endosome maturation by chloroquine

abrogated the immuostimulatory activity of chemically made

siRNAs, but not the T7 RNAP-made siRNAs (41). In addition,

the immunostimulatory effect of the T7 RNAP-made siRNAs was

not inhibited with 2-aminopurine, a speciﬁc inhibitor of PKR.

Therefore, which cytoplasmic factors are able to sense in vitro

transcribed RNA? Additional studies from other investigators

showed that RIG-I senses ssRNA-bearing 5



-triphosphate, a

speciﬁc signature of viral and in vitro transcribed RNAs (42).

Interestingly, artiﬁcial capping or base modiﬁcations of the



-triphosphate bearing RNA abolished immune response. In

general, self-RNA undergo several modiﬁcations to eliminate or

mask the 5



-triphosphate group. However, the reported data do

not explain why certain endogenous RNA with 5



-triphosphates

(e.g., 7SL RNA, an abundant cytoplasmic RNA) escape RIG-I

recognition. Natural 2



ribose-modiﬁcations might protect

Promises and Challenges in Developing RNAi as a Research Tool and Therapy 181

Dicer

Pre-miRNA

RNPs

miRNA-mediated

target recognition

mRNA cleavage

5’

3’

5’

(A)n 3’

5’

3’

Drosha

Nucleus

Translation arrest

Cytoplasm

AAAAA..3’

5’

Pri-miRNAs

Pre-miRNA

Exportin 5

5’

3’

5’

(60-80 nt)

DGCR8

miRNA:miRNA*

duplex

Ago-2

Fig. 12.2. Gene regulation by miRNAs. MiRNAs are derived from genome transcribed

primary transcripts (pri-miRNAs) that are predicted to form multiple stems and hairpin

structures. Pri-miRNAs are processed by the ribonuclease Drosha to 70–80 nucleotide

pre-miRNAs that are transported to the cytoplasm by exportin 5. Subsequently, they are

processed by Dicer into mature 22–24 nucleotide miRNAs, which are then incorporated

into a RNP complex that can direct either RNA cleavage (perfect complementarity with

mRNA) or translation arrest (mismatches with target mRNA).

endogenous RNA-bearing 5



-triphosphate from being detected

by RIG-I and other RNA sensors. MDA5, the most closely

related protein of RIG-I, is also an IFN-inducible protein.

However, the exact mechanisms of RNA sensing by MDA5 has

yet to be deﬁned but seems to sense dsRNA structures from

certain viruses.

7. Separation of

siRNA Unwanted

Effects from Gene

Silencing

Much of the recent interest in the mechanisms involved in RNA

sensing or tolerance by the immune system was generated by

the observation that siRNA can activate innate immunity (14,

17). Considering the high frequency of uridines in messen-

ger RNAs it is more likely that a high proportion of self and

non-self chemically made siRNA sequences will activate innate

182 Sioud

immunity. Therefore, it would be desirable to develop strategies

that evade immune activation. At least three distinct ways to avoid

immune activation by siRNAs can be used. The ﬁrst would be to

use delivery agents that avoid the delivery and/or retention of

siRNA within the endosomes. The second way relies on the use

of modiﬁed nucleotides. In this respect, Morrissey and colleagues

showed that the incorporation of various 2



-modiﬁed nucleotides

in siRNA sequence abrogated their immunostimulatory potency

(43). However, the chemical modiﬁcations that block immune

activation must be chosen carefully so as not to inhibit siRNA

silencing potency. Thus, ﬁnding the appropriate chemical mod-

iﬁcations for interfering with siRNA immune activation will be

important for exploring their therapeutic applications.

Fortunately, we have shown that replacement of only uridines

with their 2



-ﬂuoro, 2



-deoxy,or2



-O-methyl modiﬁed coun-

terparts can abrogate immune recognition of ss siRNA or ds

siRNAs by TLRs without reducing their silencing potency (35,

see Fig. 12.3). In accordance with our data, Judge and col-

leagues demonstrated that the incorporation of 2



-O-methyl-

uridine or 2



-O-mehtyl guanosine residues into siRNAs abrogates

their immunostimulatory potency (44). Collectively, the pub-

lished data offer the possibility of choosing the appropriate mod-

iﬁcations that evade immune activation without reducing siRNA-

silencing. I recommend the use of 2



-deoxy uridine or thymidine

CH3

U-2'-OH

U-2'-F

U-2'-O-methyl

U-2'-H

2500

5000

7500

10000

TNF-α (pg/ml)

5’-UGCUAUUGGUGAUUGCCUCTT-3’

Fig. 12.3. A representative example of TNF-α production in human monocytes in response to unmodiﬁed or 2



-uridine

modiﬁed RNAs. Cells were transfected with either unmodiﬁed or 2



-uridine modiﬁed single-stranded siRNA. Subsequent

to 18 h transfection time, secreted TNF-α in culture supernat ants was measured by ELISA.

Promises and Challenges in Developing RNAi as a Research Tool and Therapy 183

TLR7 TLR8 TLR7 TLR8

TLR7 TLR8

2’-hydroxy

uridines

2’-O-methyl

uridines

2’-fluoro

uridines

2’-deoxy

Uridines or

Thymidines

Recognition

Gene

silencing

Gene

silencing

Gene

silencing

Gene

silencing

Immune

Activation

No activation No activation

No activation

Fig. 12.4. An overview of the effects of chemical modiﬁcations on gene silencing and activation of endosomal TLR7/8.

For details see the text.

modiﬁed siRNAs because they showed no signiﬁcant immunos-

timulatory effects and binding to TLR7/8 ( see Fig. 12.4).

The ﬁnding that 2



-modiﬁed RNAs can evade immune acti-

vation suggests that naturally modiﬁed RNAs are not sensed by

TLR7/8. Support of this view has been provided by Karikó and

colleagues, who demonstrated that modiﬁcations that are fre-

quently found in mammalian RNA (such as pseudouridine, 2



-O-

methyl) can interfere with the capacity of RNA to activate TLR-7

in dendritic cells (45). Thus, unmodiﬁed RNA corresponding to

mammalian sequences would be expected to activate TLR7 more

effectively than native RNAs provided they are delivered to the

endosomes (32).

The ﬁnding that unmodiﬁed, but not 2



-modiﬁed RNA, are

potent triggers of innate immunity also raised questions about the

differences in their structures that might be relevant to binding

to TLR7/8. Which step is affected by 2



-modiﬁcations, and why

can’t 2



-modiﬁed RNAs trigger immune activation?. One way

to address the ﬁrst question is to assess whether 2



-modiﬁed

RNA antagonize with immunostimulatory RNAs to trigger

TLR7/8 signaling. Studies of transfected human monocytes

show that 2



-O-methyl modiﬁed RNAs abrogates the activation

of TLR7 by immunostimulatory RNAs (46). Of considerable

interest, is that 2



-O-methyl modiﬁed RNAs suppressed immune

activation at very low concentrations (47). In addition, we have

shown that they can effectively inhibit immune activation by a

variety of immunostimulatory sequences including bacterial and

mitochondrial RNAs. Also, chemically modiﬁed RNA can