Immunity, Vol. 23, 165–175, August, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.immuni.2005.06.008
Suppression of RNA Recognition by Toll-like
Receptors: The Impact of Nucleoside Modification
and the Evolutionary Origin of RNA
Katalin Karikó,1,
* Michael Buckstein,2
Houping Ni,2
thetic antiviral compound R-848 (Jurk et al., 2002), but
a natural ligand has not been identified.and Drew Weissman2
1
Department of Neurosurgery It has been known for decades that selected DNA
and RNA molecules have the unique property to acti-2
Department of Medicine
University of Pennsylvania School of Medicine vate the immune system. It was discovered only recently
that secretion of interferon in response to DNAPhiladelphia, Pennsylvania 19104
is mediated by unmethylated CpG motifs acting upon
TLR9 present on immune cells (Hemmi et al., 2000). For
Summary years, bacterial and mammalian DNA were portrayed
as having the same chemical structure, which hamDNA
and RNA stimulate the mammalian innate im- pered the understanding of why only bacterial, but not
mune system through activation of Toll-like receptors mammalian, DNA is immunogenic. Recently, however,
(TLRs). DNA containing methylated CpG motifs, how- the sequence and structural microheterogeneity of
ever, is not stimulatory. Selected nucleosides in natu- DNA has come to be appreciated. For example, methylrally
occurring RNA are also methylated or otherwise ated cytidine in CpG motifs of DNA has proven to be
modified, but the immunomodulatory effects of these the structural basis of recognition for the innate imalterations
remain untested. We show that RNA sig- mune system. In light of this finding and given that mulnals
through human TLR3, TLR7, and TLR8, but incor- tiple TLRs respond to RNA, a question emerges as to
poration of modified nucleosides m5C, m6A, m5U, whether the immunogenicity of RNA is under the cons2U,
or pseudouridine ablates activity. Dendritic cells trol of similar types of modification. This possibility is
(DCs) exposed to such modified RNA express signifi- not unreasonable given that RNA undergoes nearly one
cantly less cytokines and activation markers than hundred different nucleoside modifications (Rozenski
those treated with unmodified RNA. DCs and TLR- et al., 1999). Importantly, the extent and quality of RNA
expressing cells are potently activated by bacterial modifications depend on the RNA subtype and correand
mitochondrial RNA, but not by mammalian total late directly with the evolutionary level of the organism
RNA, which is abundant in modified nucleosides. We from which the RNA is isolated. Ribosomal RNA, the
conclude that nucleoside modifications suppress the major constituent (w80%) of cellular RNA, contains sigpotential
of RNA to activate DCs. The innate immune nificantly more nucleoside modifications when obsystem
may therefore detect RNA lacking nucleoside tained from mammalian cells versus bacteria. Human
modification as a means of selectively responding to rRNA, for example, has ten times more pseudouridine
bacteria or necrotic tissue. (Ψ) and 25 times more 2#-O-methylated nucleosides
than bacterial rRNA, whereas rRNA from mitochondria,
Introduction an organelle that is a remnant of eubacteria (Margulis
and Chapman, 1998), has very few modifications (BachThe
innate immune system is the first line of defense ellerie and Cavaille, 1998). Transfer RNA is the most heavagainst
invading pathogens (Medzhitov, 2001). This ily modified subgroup of RNA. In mammalian tRNAs, up
system utilizes TLRs to recognize conserved pathogen- to 25% of the nucleosides are modified, whereas there
associated molecular patterns and orchestrate the initi- are significantly less modifications in prokaryotic tRNAs.
ation of immune responses. TLRs are germ line-encoded Bacterial mRNA contains no nucleoside modifications,
signaling receptors with extracellular leucine-rich re- whereas mammalian mRNAs have modified nucleopeats
and intracellular signaling domains. In humans, sides such as 5-methylcytidine (m5C), N6-methyladenten
distinct TLR family members have been identified, osine (m6A), inosine and many 2#-O-methylated nucleand
corresponding microbial ligands for most have osides in addition to N7-methylguanosine (m7G), which
been identified. Several TLRs recognize and respond to is part of the 5#-terminal cap (Bokar and Rottman,
nucleic acids. DNA containing unmethylated CpG mo- 1998). The presence of modified nucleosides was also
tifs, characteristic of bacterial and viral DNA, activate demonstrated in the internal regions of many viral
TLR9 (Hemmi et al., 2000). Double-stranded (ds)RNA, a RNAs including influenza, adeno, and herpes simplex;
frequent viral constituent, has been shown to activate surprisingly, modified nucleosides were more frequent
TLR3 (Alexopoulou et al., 2001; Wang et al., 2004), sin- in viral than in cellular mRNAs (Bokar and Rottman,
gle-stranded (ss)RNA activates mouse TLR7 (Diebold 1998). A substantial number of nucleoside modificaet
al., 2004), and RNA oligonucleotides with phos- tions are uniquely present in either bacterial or mamphorothioate
internucleotide linkages are ligands of hu- malian RNA, thus providing an additional molecular feaman
TLR8 (Heil et al., 2004). Based on structural and ture for immune cells to discriminate between microbial
sequence similarities, TLR7, TLR8, and TLR9 form a and host RNA. Considering that cells usually contain
subfamily. Activation of these receptors depends upon five to ten times more RNA than DNA, presence of such
endosomal acidification and leads to interferon produc- distinctive characteristics on RNA could make them a
tion. Human TLR7 and TLR8 are stimulated by the syn- rich molecular source for sampling by the immune system,
a notion becoming evident by the identification of
multiple TLRs signaling in response to RNA. The role*Correspondence: kariko@mail.med.upenn.edu
Immunity
166
of nucleoside modifications on the immunostimulatory
potential of RNA, however, is not known.
In recent years, we have investigated the immunomodulatory
effect of RNA on human DCs. These studies
demonstrated that in vitro-transcribed RNA activates/
matures DCs (Weissman et al., 2000) partially by a
mechanism in which double-stranded regions of the
RNA signal through TLR3 (Kariko et al., 2004). Recently,
it was noted that in vitro transcripts, or total RNA derived
from bacteria, but not from eukaryotic cells, can
prime DCs for high-level IL-12 secretion (Koski et al.,
2004). The molecular basis for the discrimination between
these various RNAs is not fully understood. In
this report, we sought to determine whether naturally
occurring nucleoside modifications modulate the immunostimulatory
potential of RNA and the role TLRs
might play in this process.
Results
Naturally Occurring RNAs Are Not Equally Potent
Activators of DCs
We recently demonstrated that RNA transcribed in vitro
or released from necrotic mammalian cells activates
Figure 1. Production of TNF-α by MDDCs Transfected with Natu-human DCs (Kariko et al., 2004). In an independent
ral RNA
study, we also determined that although in vitro-tranHuman
MDDCs were incubated with lipofectin alone, or complexed
scribed RNAs are effective, eukaryotic mRNA and tRNA
with R-848 (1 ␮g/ml), or RNA (5 ␮g/ml) from 293 cells (total, nuclear,
did not stimulate cultured human DCs (Koski et al., and cytoplasmic RNAs), mouse heart (polyA+
mRNA), human plate-
2004). This finding prompted us to investigate the im- let mitochondrial RNA, bovine tRNA, bacterial tRNA, and total RNA
munostimulatory potential of different cellular RNA sub- (E. coli) with or without RNase digestion. After 8 hr, TNF-α was
measured in the supernatants by ELISA. Mean values ± SEM aretypes. The objective was to identify the likely RNA comshown.
The results are representative of three independent experi-ponents from necrotic cells that activated DCs in the
ments.
original experimental setting. We first isolated RNA
from different subcellular compartments (cytoplasm,
nucleus, and mitochondria). These RNA isolates as well
rally occurring RNAs are equal. The activation potenas
total RNA, tRNA, and polyA-tail-selected mRNA, all
tials of RNAs seem to have an inverse correlation with
from mammalian sources, were complexed to lipofectin
the extent of their nucleoside modification, because
and added to monocyte-derived DCs (MDDCs) generbacterial
RNA and mtRNA that contain only a few modiated
with GM-CSF and IL-4. We determined that mamfications
are the most potent activators of DCs,
malian total, nuclear, and cytoplasmic RNA and mRNA
whereas extremely modified tRNAs have little to no acall
induced TNF-α secretion, although at very low levels
tivity.
relative to RNA synthesized in vitro by T7 RNA polymerase
(RNAP) (Figure 1). Interestingly, mammalian tRNA
did not induce any detectable level of TNF-α, whereas In Vitro-Transcribed RNA Stimulates Human TLR3,
TLR7, and TLR8, but Most of the Nucleoside-mitochondrial (mt)RNA was the most potent RNA type
to stimulate MDDCs. Considering that mtRNA shares Modified RNAs Are Not Stimulatory
Naturally, all RNA is synthesized from four basic ribonu-more characteristics with bacterial RNA than with other
mammalian RNA types, it was not surprising to find that cleotides, ATP, CTP, UTP and GTP, but some of the incorporated
nucleosides are modified posttranscription-bacterial total RNA was also a very potent activator of
MDDCs (Figure 1). Bacterial tRNA, which is modified ally in almost all types of RNA. The extent and nature
of modifications vary and depend on the RNA type asbut to a lesser extent than mammalian tRNA, induced a
low level of TNF-α, whereas tRNAs from other sources well as the evolutionary level of the organism from
where the RNA is derived. Because findings with natu-(yeast, wheat germ, and bovine) were nonstimulatory
(Figure 1 and data not shown). Similar results were ob- ral RNA suggested that nucleoside modifications might
influence the ability of RNA to activate DCs, we set outserved when RNAs from other mammalian sources
were tested as well as when the RNA repertoires were to further investigate this possibility. First, to obtain
RNA with selected modifications, we performed in vitroanalyzed by using GM-CSF + IFN-α-generated MDDCs
(data not shown). When RNA samples were digested transcription reactions in which one or two of the four
nucleotide triphosphates (NTPs) were substituted withwith Benzonase, capable of cleaving both ssRNA and
dsRNA, RNA signaling was abolished in MDDCs, verify- a corresponding nucleoside-modified NTP. Several sets
of RNA with different primary sequences ranging ining that RNA is the active component that triggers
TNF-α secretion (Figure 1). These findings demonstrate length between 0.7 and 1.9 kb and containing either
none, one, or two types of modified nucleosides werethat from the view of immunostimulation not all natu-
Modified Nucleosides Block RNA from Activating TLR
167
Figure 2. TLR-Dependent Activation by RNA
(A) Aliquots (1 ␮g) of in vitro-transcribed
RNA-1571 without (none) or with m5C, m6A,
Ψ, m5U, or s2U nucleoside modifications
were analyzed on denaturing agarose gel followed
by ethidium bromide-staining and UV
illumination.
(B) 293 cells stably expressing human TLR3,
TLR7, TLR8, and control vectors were
treated with lipofectin alone or complexed
with R-848 (1 ␮g/ml) or the indicated RNA (5
␮g/ml). Modified nucleosides present in
RNA-730 and RNA-1571 are noted. 293ELAM-luc
cells were used as control cells,
but other controls gave similar results.
(C) CpG ODN-2006 (5 ␮g/ml), LPS (1.0 ␮g/
ml), and RNA isolates were obtained from rat
liver, mouse cell line (TUBO) and human
spleen (total), or human platelet mitochondrial
RNA, or from two different E. coli
sources. 293-hTLR9 cells served as control.
After 8 hr, IL-8 was measured in the supernatants
by ELISA.
Mean values ± SEM are shown. Cell lines
transformed to express hTLR3-targeted siRNA
are indicated with an asterisk. The results are
representative of four independent experiments.
N.D., not determined.
transcribed. Modified RNAs analyzed by denaturing gel specific short hairpin (sh)RNA. This newly generated
cell line, which did not respond to poly(I):(C), was usedelectrophoresis were indistinguishable from their nonmodified
counterparts in such that all were intact and for further study (Figure 2B). When the 293-hTLR8 cells
expressing TLR3-targeted shRNA were transfected withmigrated as expected based on their sizes (Figure 2A).
We and others recently demonstrated that in vitro- in vitro-transcribed RNAs, they secreted large amounts
of IL-8; however, transfecting RNA containing any oftranscribed RNA activates human TLR3 (Kariko et al.,
2004) and murine TLR7 (Diebold et al., 2004), whereas the nucleoside modifications did not stimulate these
cells (Figure 2B). In some experiments, we observedchemically synthesized oligoribonucleotides (ORNs)
stimulate murine TLR7 and human TLR8 (Heil et al., that m6A modification in the RNA permitted a limited
amount of IL-8 release. Control cells (293, 293-pUNO2004). Therefore, to determine whether modification of
nucleosides influences the RNA-mediated activation of null, 293-TLR3-sh, and 293-hTLR9) did not respond to
RNA transfection. (Figures 2B and 2C and data notTLRs, we utilized human 293 cell lines stably transformed
to express human TLR3, TLR7, or TLR8 and shown). To rule out that clonal artifacts were responsible
for RNA-induced stimulation, at least three separatemonitored TLR activation through IL-8 release. First,
TLR3-transformed cells were treated with lipofectin- clones for each TLR-expressing cell line were analyzed
and gave similar results.complexed RNA (RNA-1571, RNA-730, or RNA-1866)
and, as expected based on previous studies (Kariko et In a previous study, human TLR8, but not human
TLR7, was shown to signal in response to guanosine-al., 2004), high levels of IL-8 secretion were measured.
RNA containing m6A or s2U modifications did not in- and uridine-rich ssRNA oligomers with phosphorothioate
internucleotide linkages (Heil et al., 2004). Givenduce detectable levels of IL-8 (Figure 2B and data not
shown). The presence of other nucleoside modifica- that TLR7 and TLR8 share R-848 as a ligand, we sought
to determine whether long RNA with natural phospho-tions such as m5C, m5U, Ψ, or m5C/Ψ in the RNA had
a less remarkable suppression or no effect at all on the diester internucleotide linkages was also a shared ligand
for these human TLRs. All of the in vitro-tran-potential of RNA to activate TLR3 (Figure 2B).
We previously noted that the parental 293 cells ex- scribed RNAs induced IL-8 at levels comparable to
R-848 when transfected into 293-hTLR7 cells express-press a low level of endogenous TLR3 (Kariko et al.,
2004). Therefore, the unwanted expression of endoge- ing TLR3-targeted shRNA. However, transfection of
RNA containing modified nucleosides resulted in no in-nous TLR3 was eliminated by stably transfecting the
293-hTLR8 cell line with a plasmid expressing TLR3- duction of IL-8 (Figure 2B). Experiments performed on
Immunity
168
cell lines expressing hTLR7 from different constructs To determine whether primary blood DCs responded
gave similar results. Overall, these experiments demon- to RNA in a manner similar to cytokine-generated DCs,
strate that RNA activates human TLR3, TLR7, and TLR8 we purified primary monocytoid (DC1, BDCA-1+
) and
and that nucleoside modifications limit the capacity of plasmacytoid (DC2, BDCA-4+
) DCs from peripheral
RNA to stimulate these TLRs. Specifically, m6A and s2U blood. Both cell types produced TNF-α when exposed
modifications suppress the ability of RNA to stimulate to R-848, but only DC1 responded to poly(I):(C), though
TLR3, whereas m6A, m5C, m5U, s2U, and Ψ modifica- at a very low level, demonstrating absence of TLR3 actions
block stimulation of TLR7 and TLR8. tivity in DC2. Transfection of in vitro transcripts induced
In the next set of experiments, RNAs isolated from TNF-α secretion in both DC1 and DC2 (Figure 3D). Data
natural sources were tested. First, RNA from different with modified RNA revealed that only transcripts in
mammalian species were transfected into 293 cells sta- which uridine was replaced with m5U, Ψ, or s2U were
bly expressing human TLR3, TLR7, or TLR8 (TLR7 and not stimulatory, whereas RNAs containing m5C and
TLR8 cell lines also expressed TLR3-targeted shRNA). m6A were almost as potent inducers of cytokines as
None of these RNAs induced substantial IL-8 secretion. the corresponding unmodified RNAs. This was unexHowever,
bacterial total RNA, obtained from two dif- pected because DC2s do not express TLR3 or TLR8
ferent E. coli sources, induced robust IL-8 secretion and as such should resemble the response observed in
(Figure 2C). These results and additional experimental 293-hTLR7 cells. To determine whether m5C and m6A
evidence, first that bacterial RNA transfected to 293- exert a dominant stimulatory effect, transcripts with
hTLR9 did not induce IL-8 secretion (Figure 2C) and m6A/Ψ double modification were tested and found to
second that LPS and unmethylated DNA (CpG ODN), be nonstimulatory, whereas the mixture of RNA with
the potential contaminants in bacterial RNA isolates, single type of modification (m6A + Ψ) was a potent cydid
not activate the tested TLRs (Figure 2C), together tokine inducer. This suggested that primary DCs likely
indicate that bacterial RNA is an activator of TLR3, have an additional RNA signaling entity that recognizes
TLR7, and TLR8. Mitochondrial RNA isolated from hu- m5C- and m6A-modified RNA and whose signaling is
man platelets also stimulated human TLR8, but not inhibited by modification of U residues.
TLR3 or TLR7 (Figure 2C). Collectively, these data di- FACS analysis of MDDCs treated with RNA-1571 and
rectly demonstrate that RNA that is scarce in modified its modified versions revealed that modified nucleonucleosides,
such as those isolated from bacteria or sides such as m5C, m6A, Ψ, s2U, and m6A/Ψ decrease
mitochondria, stimulate selected human TLRs, whereas the ability of RNAs to induce cell surface expression of
total mammalian RNA abundant in nucleoside modifi- CD80, CD83, CD86, and MHC class II (Figure 4). Colleccations
are non- or minimally stimulatory. tively, these results demonstrate that the capacity of
RNA to induce DCs to mature and secrete cytokines
Modified Nucleosides Reduce the Capacity of RNA depends on the subtype of DC as well as on the charto
Induce Cytokine Secretion and Activation acteristics of nucleoside modification present in the
Marker Expression by DCs RNA with the general tendency of modifications blockRNAs
containing modified or unmodified nucleosides ing stimulation.
were tested on DCs. A representative data set obtained
with MDDCs and IFN-α-generated MDDCs (Figures 3A
Suppression of RNA-Mediated Immune Stimulation
and 3B) demonstrates that nucleoside modifications diIs
Proportional to the Number of Modified
minish the ability of RNA to induce TNF-α and IL-12
Nucleosides Present in RNA
secretion. Results were similar (data not shown) when
To ideally define the importance of nucleoside modifiother
sets of RNA with the same base modifications but
cations that are components of natural RNA would re-different primary sequences and lengths were tested or
quire the construction of RNAs in vitro that accuratelywhen the RNAs were further modified by adding 5# cap
model the extent and diversity of nucleoside modifica-structure and/or 3# end polyA-tail or by removing the
tions of native RNAs and the ability to selectively re-5# triphosphate moiety, which was previously reported
move modifications present in natural RNA isolates,to promote interferon production (Kim et al., 2004).
both of which are beyond current technology. Most ofRNAs of different length and sequence induced varying
the nucleoside-modified RNA utilized in the presentamounts of TNF-α from DCs, typically less than a
study contained one type of modification amassing2-fold difference (Figure 3C). However, we detected
w25% of the total nucleotides in the RNA. Because themore variability when MDDCs from different donors
ratio of any one particular modified nucleoside, thoughwere used. In most of the experiments, MDDCs revariable,
is much lower than 25% in native RNAs, wesponded to RNA treatment as presented in Figure 3A,
asked what is the minimal frequency of any one partic-but w25% of the time, the presence of m6A reduced
ular modified nucleoside that is sufficient to limit thethe RNA-mediated MDDC activation more potently than
immunostimulatory potential of RNA. To answer thism5C or Ψ did. Under those circumstances, the relative
question, two approaches were used to generate RNAsensitivity of MDDCs to poly(I):(C) and R-848 treatwith
limited numbers of modified nucleosides. First, wements also differed. (Figure S1 available in the Suppletranscribed
RNA in vitro in the presence of decreasingmental Data with this article online). This variability was
amounts of m6A, Ψ, or m5C and increasing amounts ofnot observed for primary DCs described below. By
the corresponding unmodified NTPs. We expected theusing Northern analysis we also confirmed that cellular
incorporation of modified nucleoside phosphates intouptake and stability of the transfected RNAs were not
RNA to be proportional to the ratio contained in theinfluenced by the nucleoside modifications (data not
shown). transcription reaction, because prior RNA yields ob-
Modified Nucleosides Block RNA from Activating TLR
169
Figure 3. Cytokine Production by RNA-Transfected DCs
MDDC (A and C), IFN-α MDDCs (B), and primary DC1 and DC2 (D) were treated for 8–16 hr with lipofectin alone or complexed with R-848 (1
␮g/ml) or the indicated RNA (5 ␮g/ml). Modified nucleosides present in RNA-1571 are noted. TNF-α, IL-12(p70), and IFN-α were measured in
the supernatant by ELISA. Mean values ± SEM are shown. The results are representative of ten (A and C), four (B), and six (D) independent
experiments. N.D., not determined.
tained with T7 RNAP suggested the enzyme utilizes (w2.5%) of modified nucleoside content was required
to inhibit RNA-mediated signaling events (data not shown).NTPs of m6A, Ψ, or m5C almost as efficiently as the
basic NTPs. HPLC analysis confirmed this notion, In the second approach, we utilized chemically synthesized
21-mer ORNs with phosphodiester inter-showing for example, that after digestion of RNA transcribed
in the presence of UTP:ΨTP in a 50:50 ratio, nucleotide linkages and 5# monophosphate and identical
primary sequences but with modified nucleosidesnearly equal amounts of incorporated UMP and ΨMP
were released (Figure 5A). When RNA-1571 with increas- such as m5C, Ψ, or 2#-O-methyl-U (Um) in a single position
(Figure 6A). Results obtained after transfection ofing amounts of modified nucleoside content were
transfected into MDDCs, we detected that the pres- MDDCs with the synthetic ORNs demonstrated that
short unmodified ORNs were capable of inducing sig-ence of an increasing amount of modified nucleosides
proportionally inhibited the capacity of RNA to induce nificant TNF-α secretion, but the presence of a single
nucleoside modification was sufficient to abolish thisTNF-α (Figure 5B). The presence of 0.2%–0.4% m6A,
Ψ, or m5C in the RNA, which corresponds to approxi- effect (Figure 6B). Repeating the experiments on TLRtransformed
293 cells expressing TLR3-targeted siRNA,mately three to six modified nucleosides per one molecule
of the 1571 nt-long RNA, was sufficient to cause we found that control ORN induced 293-hTLR8 cells to
secrete IL-8, whereas those containing modified nucle-detectable inhibition of cytokine secretion (Figure 5B).
When RNAs with modified nucleoside levels of 1.7%– osides did not. When testing ORNs on hTLR3- or hTLR7expressing
cell lines, however, we saw no IL-8 secretion3.2%, which correspond to 14–29 modifications per
molecule, were tested, the RNA could maintain only under any conditions (data not shown). Finally, by using
Northern assay, we tested the 21-mer chemically syn-half of its capacity to induce expression of TNF-α.
When similar transfection experiments were performed thesized ORNs along with 31-mer in vitro transcripts for
their ability to induce TNF-α mRNA in MDDCs. ORN5on TLR-expressing 293 cells, usually a higher percent
Immunity
170
Figure 4. Activation of DCs by RNA
MDDCs were treated for 20 hr with lipofectin
alone or complexed with R-848 (1 ␮g/ml) or
the indicated RNA (5 ␮g/ml). Modified nucleosides
present in RNA-1571 are indicated.
(A) CD83 and HLA-DR staining is shown.
(B) TNF-α was measured in the supernatants
by ELISA (the asterisk represents cells that
were cultured in 30-fold larger than usual
volume of medium for flow cytometry). Mean
fluorescence of CD80 and CD86 was determined
by flow cytometry. Data are representative
of four independent experiments.
and ORN6 (31-mers) caused robust induction, whereas stimulatory effect of RNA on MDDC and 293 cells expressing
individual TLRs.the 21-mer ORN1 control induced less TNF-α mRNA,
although still well detectable, particularly in cells exposed
to the protein synthesis inhibitor cycloheximide, Discussion
which is also known to block degradation of selected
mRNAs. More importantly, ORNs containing a single We demonstrate here that a variety of natural RNAs had
different capacities to activate immune cells. The mostmodified nucleoside induced less TNF-α mRNA, and
consistently, ORN2-Um, the 2#-O-methylated ORN, was potent RNAs were those that had the least number of
modified nucleosides; therefore, we hypothesize thatthe least stimulatory (Figure 6C). Taken together, these
results demonstrate that RNA-mediated immune stimu- nucleoside modification suppresses the immune-stimulatory
effect of RNA. In a quest to prove this, severallation is suppressed proportionally by the number of
modified nucleosides present in RNA. Modification, novel lines of evidence were discovered about RNAmediated
immune activation. Initially, we establishedwhen present in a single or very few positions, depending
on the length of RNA, was sufficient to inhibit the that RNA is a ligand for human TLR7. Next, by using
Modified Nucleosides Block RNA from Activating TLR
171
Figure 5. Analyzing RNA Containing Different
Amounts of Modified Nucleosides
Capped RNA-1571 containing m6A, Ψ, or
m5C was transcribed under conditions in
which the relative ratio of m6ATP, ΨTP, or
m5CTP to the corresponding unmodified
NTP was 0%, 1%, 10%, 50%, 90%, 99%,
and 100%.
(A) All transcripts were digested to monophosphates
and analyzed by reversedphase
HPLC to determine the relative
amount of modified nucleoside incorporation.
For simplicity, only symbols for the nucleosides
are shown. Representative absorbance
profiles obtained by RNA transcribed
in the presence of pseudouridine- and uridine-triphosphates
(Ψ:U) at the indicated ratios
are shown. Elution times are noted for
3#-monophosphates of pseudouridine (Ψ),
cytidine (C), guanosine (G), uridine (U),
7-methylguanosine (m7G), and adenoside (A).
(B) Modified nucleoside content of RNA-
1571. The expected percentage of m6A, Ψ,
or m5C in RNA-1571 was calculated based
on the relative amount of modified NTP in
the transcription reaction and the nucleoside
composition of RNA-1571 (expected percentage).
The values for measured modified nucleoside
content (in percentage) were determined
based on relative values obtained
after quantitation of the HPLC chromatograms.
Based on these measured values and
on the nucleoside content of RNA-1571 (A:
505, U: 451, C: 273, and G: 342), the number
of m6A, Ψ ,or m5C per molecule of RNA-
1571 was calculated. “a” represents values
(%) for m6ATP, ΨTP, and m5CTP relative to ATP UTP and CTP, respectively. “b” represents values for m6A, Ψ, and m5C monophosphates
relative to all NMPs.
(C) MDDCs were transfected with lipofectin-complexed capped RNA-1571 (5 ␮g/ml) containing the indicated amount of m6A, Ψ, or m5C.
After 8 hr, TNF-α was measured in the supernatants by ELISA. Data are expressed as relative inhibition of TNF-α. Mean values ± SEM
obtained in three independent experiments are shown. The number of m6A, Ψ, or m5C per molecule of RNA-1571 was calculated as indicated
in (B).
RNA bearing modified nucleosides such as m5C, m5U, on characteristics of DNA methylation remains an important
component of the immune system. In mamma-s2U, m6A, Ψ, or 2#-O-methyl-U, all constituents of natural
RNA, we showed that modifying U, A, and C nucleo- lian DNA, cytosines in CpG motifs are mostly methylated,
but the lack of such modification in the genomessides, in general, suppresses the capacity of RNA to
activate cytokine-generated DCs, as measured by secre- of microbial pathogens is recognized by TLR9, which
then mediates the induction of the mammalian innatetion of TNF-α and IL-12 and by expression of CD80,
CD83, CD86, and HLA-DR. Interestingly, only uridine immune response (Hemmi et al., 2000).
Despite the fact that the immune stimulatory activitymodifications, such as m5U, s2U, or Ψ, but not m5C
or m6A, could abolish the capacity of RNA to activate of RNA was discovered decades before such was identified
for DNA and that RNA contains numerous modi-primary, blood-derived DCs. Distinct TLRs responded
differently to RNA containing different modified nucleo- fied nucleosides (Rozenski et al., 1999), the effect of
nucleoside modifications on RNA immunity has notsides. RNA with m6A and s2U modifications did not
activate TLR3, and those with m5C, m5U, s2U, m6A, or been explored. From the standpoint of immune activation,
RNA and DNA share many characteristics. WeΨ did not activate TLR7 or TLR8, whereas unmodified
RNA could activate all these human TLRs. Finally, we have shown that RNA, similarly to DNA, is more immunogenic
when derived from bacteria than from mamma-show that RNA-mediated immune stimulation is suppressed
proportionally with the number of modified nu- lian cells (Figures 1 and 2C) (Koski et al., 2004). Similar
to mammalian DNA, mammalian RNA also exerts a lim-cleosides present in RNA and that even a few modifications
are sufficient to exert a suppressive effect. ited but detectable level of immune activation (Figure
1). Others have reported that mammalian RNA inducesNucleoside modification is the foundation of the most
ancient “immune” mechanism. Bacteria methylate se- IFN-α when delivered to immune cells (Diebold et al.,
2004). To explain why mammalian RNAs are immuno-lected nucleosides in their own genome, which enables
them to distinguish and destroy an invader’s unmodi- genic, it was reasoned that in those experiments transfected
RNA entered the endosomal compartments offied DNA with restriction enzymes. During evolution,
the discrimination between host and pathogen based immune cells, therefore the immune system might
Immunity
172
Figure 6. TNF-α Expression by RNA-Transfected DCs
(A) Sequences of oligoribonucleotides (ORNs) synthesized chemically (ORN1-4) or transcribed in vitro (ORN5-6) are shown. Positions of
modified nucleosides Um (2#-O-methyluridine), m5C, and Ψ are highlighted. Human MDDCs were transfected with lipofectin alone (medium),
R-848 (1 ␮g/ml), or with the indicated RNA (5 ␮g/ml) complexed with lipofectin. Where noted, cells were treated with 2.5 ␮g/ml cycloheximide
(CHX). After 8 hr incubation, TNF-α was measured in the supernatant by ELISA (B). Mean values ± SEM are shown. The results are representative
of three independent experiments. RNA isolated from the cells were analyzed by Northern blot (C).
discriminate between self and nonself RNA based on or morphological changes associated with apoptosis
could be detected.cellular location rather than some unique pathogenassociated
molecular pattern (Crozat and Beutler, Both RNA and DNA are central immunogenic determinants
in the autoimmune disease of systemic lupus2004). It has been shown, however, that the human innate
immune system can also discriminate between erythematosus (SLE), which is characterized by production
of autoantibodies directed against DNA, RNA,molecular features of eukaryotic and bacterial mRNA
and recognize mRNA devoid of polyA-tail as stimula- and proteins associated with nucleic acids (Ronnblom
et al., 2003). In the development of SLE, studies nowtory (Koski et al., 2004). In this report, we observed potent
immune stimulation with bacterial, but not with have established the involvement of TLR9 activation by
mammalian DNA that bears hypomethylated CpG mo-mammalian, total RNA and concluded that this was due
to the difference in their modified nucleoside content. tifs (Boule et al., 2004). Another prominent target molecule
in SLE is U1 small nuclear RNA, which has beenThis is supported by the observation that the major
mass of total RNA is rRNA, and modified nucleosides recently shown to activate TLR3 (Hoffman et al., 2004),
suggesting the potential involvement of an RNA-sensi-are abundant in mammalian, but not in bacterial, rRNA:
3% versus 0.8% (Bachellerie and Cavaille, 1998). We tive TLR in the disease process.
Nucleosides in native RNA become modified post-also observed that suppression of RNA-mediated immune
stimulation correlated with the level of this differ- transcriptionally as part of their maturation process. Almost
one hundred different types of modified nucleo-ence in modified nucleoside content (Figure 5C). The
present study now identifies nucleoside modification as sides have been identified in RNA, but the physiological
significance of these alterations is not well understood.a novel feature of RNA recognized by the innate immune
system, specifically by TLR3, TLR7, and TLR8 Most of the modifications occur nonrandomly at positions
conserved across diverse species, implying that(Figure 2). We observed earlier that RNA from necrotic
cells activated DCs, whereas RNA from apoptotic cells they are important. Surprisingly, however, even the
extensively modified tRNA could function without anydid not (Kariko et al., 2004). Based on the results presented
in this report, we now propose that mammalian modification (Sampson and Uhlenbeck, 1988), thus
leaving the role of nucleoside modification very puzzling.RNA, especially the least-modified mtRNA, likely contributed
to the observed effect. The immune potential Pseudouridine is the most abundant modified nucleoside
in RNA. It is generated by isomerization of uri-of mammalian RNA might also explain why degradation
of RNA during apoptosis is so critical. Fragmentation of dines. We have demonstrated here that pseudouridine
along with the other uridine modifications m5U and s2Ugenomic DNA is a well-established process and used
consistently as a technique to define apoptosis itself. uniquely suppress the capacity of RNA to activate primary
DCs (Figure 3D). This finding implies that unmodi-Although less described, a well-orchestrated degradation
of cellular RNA also occurs in apoptotic, but not fied uridine probably contributes to the immune stimulatory
action of RNA. Indeed, several points of evidencenecrotic, cells. Interestingly, the most immunogenic
mtRNA degrades at a very early stage of apoptosis support this suggestion. In an earlier study, poly(U) was
identified as the only homopolymer capable of inducing(Crawford et al., 1997), hours before the breakdown of
cytoplasmic RNA, DNA laddering (Houge et al., 1995), IL-12 in primed DCs (Koski et al., 2004). By using DCs
Modified Nucleosides Block RNA from Activating TLR
173
from TLR7 null mice, Diebold et al. identified TLR7 as RIG-1, and PKR that function in the innate immune system
independently from TLRs (Sen and Sarkar, 2005).the responding receptor for poly(U) treatment (Diebold
et al., 2004). Others have shown that even nucleoside By using TLR7-expressing cell lines, we demonstrate
(Figure 2) that in vitro-transcribed RNA and bacterialmixtures with uridine are sufficient to stimulate PBMCs
to secrete TNF-α (Heil et al., 2004). Of interest, we did RNA, but not dsRNA, are ligands for human TLR7. This
finding is in discordance with results obtained by Heilnot find that poly(U) or any other RNA homopolymer
activated primary DCs or human TLR3, TLR7, or TLR8 et al. (2004), who showed that human TLR7 was nonresponsive
to RNA oligomers with phosphorothioate link-when transformed 293 cells expressing these receptors
were used for testing. Presence of Ψ in RNA promotes age. Differences in the stimulating RNA, such as long
RNA versus short ORNs with phosphorothioate linkage,base stacking, thereby stabilizing RNA duplex regions
(Charette and Gray, 2000), which might explain why likely account for the conflicting result. We observed
that all in vitro-transcribed RNAs, regardless of theirΨ-modified RNA could potently activate TLR3 (Figure 2B).
N6-methyladenosine (m6A) is the only base-modified primary sequence, as well as bacterial RNA activated
TLR7 expressing 293 cells, demonstrating that naturalnucleoside that is present in all RNA types, including
rRNA, tRNA, and snRNA, as well as in mRNAs of cellu- RNA is a ligand for this receptor.
In summary, we demonstrate that selected naturallar and viral origins. The methylation in m6A interferes
with Watson-Crick base pairing, thus, its presence de- RNA isolated from mammalian and bacterial cells and
RNA transcribed in vitro or synthesized chemically acti-stabilizes RNA duplexes (Kierzek and Kierzek, 2003).
This characteristic of m6A might explain why RNA con- vate human DCs and stably transformed 293 cells expressing
human TLR3, TLR7, or TLR8. Such activationtaining m6A did not stimulate TLR3 (Figure 2B). m6A is
present in mRNA of mammalian cells and RNA of was reduced or completely eliminated with RNA containing
naturally occurring modified nucleosides, suchviruses that replicate in the nucleus such as influenza,
adenovirus, HSV, SV40, and RSV (Bokar and Rottman, as m5C, m6A, m5U, pseudouridine, or 2#-O-methyl-U.
Insights gained from this study could advance our un-1998). In general, m6A modifications were found internally,
mostly in coding sequences, and viral mRNA derstanding of autoimmune diseases where nucleic
acids play a prominent role in the pathogenesis, deter-usually contained significantly more m6A than cellular
mRNA (Bokar and Rottman, 1998). Interestingly, Rous mine a role for nucleoside modifications in viral RNA,
and give future directions into the design of therapeu-sarcoma virus replicated similarly with and without m6A
when tested in cell culture (Kane and Beemon, 1987), tic RNAs.
therefore no function could be assigned to m6A in this
viral mRNA. It is tempting to speculate that the pres- Experimental Procedures
ence of m6A in viral RNA might serve the virus by allowPlasmids
and Reagentsing it to avoid immune activation. This suggestion is
Plasmids pTEVluc (D. Gallie, UC Riverside), pT7T3D-MART-1
strengthened by considering that the frequency of m6A
(ATCC, Manassas, VA), pUNO-hTLR3 (InvivoGen, San Diego, CA),
modifications found in viral mRNAs, up to eight per a and pSVren (Kariko et al., 2004) were obtained. Human TLR3-spe-
1.8 kb-long segment of influenza RNA (Narayan et al., cific siRNA, pTLR3-sh was constructed by inserting synthetic ODN-
1987), is sufficient to suppress the capacity of RNA to encoding shRNA with 20 nt-long homology to human TLR3 (nt 703–
722, accession: NM_003265) into plasmid pSilencer 4.1-CMV-neoactivate DCs (Figure 5). Because those early studies
(Ambion, Austin, TX). pCMV-hTLR3 was obtained by first cloningwith viruses were performed in cell culture and not in
hTLR3-specific PCR product (nt 80–2887; accession NM_003265)
animals, the immune suppressive effect of m6A might
into pCRII-TOPO (Invitrogen, Carlsbad, CA), then released with
have been missed. NheI-HindIII cutting and subcloning to the corresponding sites of
RNA containing either m6A or m5C stimulated pri- pcDNA3.1 (Invitrogen). Cells were treated with the following remary
DC1 and DC2 as potently as the corresponding agents: LPS (E. coli 055:B5) (Sigma Chemical Co, St. Louis, MO),
CpG ODN-2006, and R-848 (InvivoGen).nonmodified RNA (Figure 3D). This was an unexpected
finding, because DC2 express only TLR7 (Ito et al.,
2002; Matsumoto et al., 2003) and thus resemble 293- Cells and Cell Culture
Human embryonic kidney 293 cells (ATCC) were propagated inhTLR7 cells that did not respond to any of the modified
DMEM supplemented with glutamine (Invitrogen) and 10% FCSRNA (Figure 2B). Because all tested RNA were deliv(Hyclone,
Ogden, UT) (complete medium). 293-hTLR3 and 293
ered by transfection to the cells, where they could inpUNO
null cell lines were generated by transforming 293 cells with
teract with many different RNA binding proteins, it is pUNO-hTLR3 and pUNO null. Cell lines 293-hTLR7, 293-hTLR8,
possible an RNA receptor is uniquely present in primary and 293-hTLR9 (InvivoGen) were grown in complete medium supplemented
with blasticidin (10 ␮g/ml) (Invivogen). Cell lines 293-DCs, but not in 293 cells or MDDCs. Such an RNA senELAM-luc
and TLR7-293 (M. Lamphier, Eisai Research Institute, An-sor could likely recognize U-rich RNA patterns even in
dover MA) and TLR3-293 cells were cultured as described (Kariko
the presence of the m6A and m5C nucleoside residues,
et al., 2004). Cell lines 293, 293-hTLR7, and 293-hTLR8 were stably
but not when the U residues are masked by modifica- transfected with pTLR3-sh and selected with G-418 (400 ␮g/ml)
tions. Support for this hypothesis was provided by the (Invitrogen). Neo-resistant colonies were screened, and only those
observations that RNA containing both m6A and Ψ that did not express TLR3, determined as lack of IL-8 secretion in
response to poly(I):(C), were used. Cell lines were used as soon asmodifications on the same strand did not activate DC1
possible, because shRNA-mediated suppression of TLR3 becameor DC2, whereas mixtures of RNA containing either
leaky over time. Leukopheresis samples were obtained from HIVm6A
or Ψ modification on separate strands potently acuninfected
volunteers through an IRB-approved protocol. DCs
tivated these cells (Figure 3D). In this regard, there are were produced as described previously and treated with GM-CSF
already examples for single- and double-stranded (50 ng/ml) + IL-4 (100 ng/ml) (Weissman et al., 2000) or IFN-α (1000
U/ml) (Santini et al., 2000) (R&D Systems, Minneapolis, MN) in AIMRNA-responsive cytoplasmic receptors such as FADD,
Immunity
174
V medium (Invitrogen). Primary myeloid and plasmacytoid DCs 96-well plates (5 × 104
cells/well) and cultured without antibiotics.
(DC1 and DC2) were obtained from peripheral blood by using On the subsequent day, the cells were exposed to R-848 or RNA
BDCA-1 and BDCA-4 cell isolation kits (Miltenyi Biotec Auburn, with prior complexing to lipofectin (Invitrogen) as previously deCA),
respectively. scribed (Kariko et al., 1998). The RNA was removed after 1 hr, and
the cells were further incubated in complete medium for 7 hr. SuRNA
pernatants were collected for IL-8 measurement.
In Vitro-Transcribed RNA DCs in 96-well plates (w1.1 × 105
cells/well) were treated with
By using in vitro transcription assays (MessageMachine and Mega- R-848, lipofectin alone, or complexed with RNA for 1 hr when the
Script kits; Ambion), the following long RNAs were generated by medium was replaced by fresh medium. Cells and medium were
T7 RNAP as described (Kariko et al., 1998) (note: the names of harvested at the end of an 8–20 hr incubation; cells were harvested
templates are in parentheses), the number in the name of the RNA for either RNA isolation or flow cytometry, whereas the collected
specifies the length: RNA-1866 (NdeI-linearized pTEVluc) encodes culture medium was subjected to cytokine ELISA. The levels of ILfirefly
luciferase and a 50 nt-long polyA-tail, RNA-1571 (SspI-linear- 12 (p70) (BD Biosciences Pharmingen, San Diego, CA), IFN-α, TNF-α,
ized pSVren) encodes Renilla luciferase, RNA-730 (HindIII-linear- and IL-8 (Biosource International, Camarillo, CA) were measured
ized pT7T3D-MART-1) encodes the human melanoma antigen in supernatants by ELISA. Cultures were performed in triplicate to
MART-1, RNA-713 (EcoRI-linearized pT7T3D-MART-1) corresponds quadruplicate and measured in duplicate.
to antisense sequence of MART-1, and RNA-497 (BglII-linearized
pCMV-hTLR3) encodes a partial 5# fragment of hTLR3. To obtain Analysis of DC Activation
RNA bearing nucleoside modification, the transcription reaction DCs treated as described above were analyzed by flow cytometry
was assembled with the replacement of one (or two) of the basic after 20 hr. DCs were stained with CD83-phycoerythrin mAb (ReNTPs
with the corresponding triphosphate-derivative(s) of the search Diagnostics Inc, Flanders, NJ), HLA-DR-Cy5PE, and CD80
modified nucleotide 5-methylcytidine, 5-methyluridine, 2-thiouri- or CD86-fluorescein isothiocyanate mAb and analyzed on a FACSdine,
N6-methyladenosine, or pseudouridine (TriLink, San Diego, calibur flow cytometer by using CellQuest software (BD BiosciCA).
In each transcription reaction, all four nucleotides or their de- ences).
rivatives were present in equimolar (7.5 mM) concentration. In selected
experiments, 6 mM m7GpppG cap analog (New England Northern Blot Analysis
BioLabs, Beverly, MA) was also included to obtain capped RNA. To RNA was isolated from MDDCs after an 8 hr incubation following
obtain RNA containing increasing amounts of m6A, Ψ, or m5C, the treatment as described above. Where noted, cells were treated
transcription reaction was performed in a reaction mix in which the
with 2.5 ␮g/ml cycloheximide (Sigma) 30 min prior to the stimularatio
of one particular modified NTP relative to the corresponding
tion and throughout the entire length of incubation. RNA samples
unmodified NTP was 0%, 1%, 10%, 50%, 90%, 99%, and 100%. By
were processed and analyzed on Northern blots as described (Karusing
DNA oligodeoxynucleotide templates and T7 RNAP (Silencer
iko et al., 2004) by using human TNF-α and GAPDH probes derived
siRNA construction kit, Ambion), ORN5 and ORN6 were generated.
from plasmids (pE4 and pHcGAP, respectively) obtained from
Natural and Synthetic RNA
ATCC.
Mitochondria were isolated from outdated platelets (obtained from
the University of Pennsylvania Blood Bank under an IRB approved
protocol) by using a fractionation lyses procedure as described by
Supplemental Datathe manufacturer (Mitochondria Isolation Kit; Pierce, Rockford, IL).
Supplemental Data include one figure and are available with thisRNA was isolated from the purified mitochondria, cytoplasmic and
article online at http://www.immunity.com/cgi/content/full/23/2/nuclear fractions of 293 cells, unfractioned 293 cells, rat liver, mouse
165/DC1/.cell line TUBO, and DH5α strain of E. coli by Master Blaster (BioRad,
Hercules, CA). Bovine tRNA, wheat tRNA, yeast tRNA, E. coli
AcknowledgmentstRNA, poly(A)+
mRNA from mouse heart, and poly(I):(C) were purchased
from Sigma; total RNA from human spleen and E. coli RNA
This work was supported by National Institutes of Health grantswere purchased from Ambion. Oligoribonucleotide 5#-monophosAI060505,
AI50484, and DE14825.phates were synthesized chemically (Dharmacon, Lafayette, CO).
Aliquots of RNA samples were incubated in the presence of
Benzonase nuclease (1 U per 5 ␮l of RNA at 1 ␮g/␮l for 1 hr) (Nova- Received: January 14, 2005
gen, Madison, WI). Aliquots of RNA-730 were digested with alkaline Revised: May 26, 2005
phosphatase (New England Biolab). Generally, RNA samples were Accepted: June 15, 2005
analyzed by denaturing agarose or polyacrylamide gel electropho- Published: August 23, 2005
resis for quality assurance. Assays for LPS in RNA preparations
using the Limulus Amebocyte Lysate gel clot assay were negative
References
with a sensitivity of 3 pg/ml (University of Pennsylvania, Core Fa-
cility).
Alexopoulou, L., Holt, A.C., Medzhitov, R., and Flavell, R.A. (2001).
Recognition of double-stranded RNA and activation of NF-kB by
HPLC Analysis
Toll-like receptor 3. Nature 413, 732–738.
Nucleoside monophosphates were separated and visualized via
Bachellerie, J.-P., and Cavaille, J. (1998). Small nucleolar RNAsHPLC (Greenwood and Gentry, 2002). Briefly, first to release free
guide the ribose methylations of eukaryotic rRNAs. In Modificationnucleoside 3#-monophosphates, 5 ␮g aliquots of RNA were diand
Editing of RNA, H. Grosjean and R. Benne, eds. (Washingtongested with 0.1 U RNase T2 (Invitrogen) in 10 ␮l of 50 mM NaOAc
D.C.: ASM Press), pp. 255–272.and 2 mM EDTA buffer (pH 4.5) overnight, then the samples were
injected into an Agilent 1100 HPLC by using a Waters Symmetry Bokar, J.A., and Rottman, F.M. (1998). Biosynthesis and functions
C18 column (Waters, Milford, MA). Buffer A consisted of 30 mM of modified nucleosides in eukaryotic mRNA. In Modification and
KH2PO4 and 10 mM tetraethylammonium phosphate (PicA reagent, Editing of RNA, H. Grosjean and R. Benne, eds. (Washington D.C.:
Waters), pH 6.0. Buffer B was acetonitrile. At a flow rate of 1 mL/ ASM Press), pp. 183–200.
min, a gradient from 100% buffer A to 30% buffer B was run over Boule, M.W., Broughton, C., Mackay, F., Akira, S., Marshak-
60 min. Nucleotides were detected by using a photodiode array at
Rothstein, A., and Rifkin, I.R. (2004). Toll-like receptor 9-dependent
254 nm. Identities were verified by retention times and spectra.
and -independent dendritic cell activation by chromatin-immunoRelative
percentage of modified versus the corresponding unmodiglobulin
G complexes. J. Exp. Med. 199, 1631–1640.
fied nucleosides in RNA was determined for each transcript.
Charette, M., and Gray, M.W. (2000). Pseudouridine in RNA: what,
where, how, and why. IUBMB Life 49, 341–351.Treatment of Cells
Crawford, D.R., Lauzon, R.J., Wang, Y., Mazurkiewicz, J.E.,Parental 293, 293-hTLR7, and 293-hTLR8 cells, all expressing
TLR3-specific siRNA, and 293-hTLR9, TLR3-293 were seeded into Schools, G.P., and Davies, K.J. (1997). 16S mitochondrial ribosomal
Modified Nucleosides Block RNA from Activating TLR
175
RNA degradation is associated with apoptosis. Free Radic. Biol. interferon-alpha producing cells (plasmacytoid dendritic cells) in
autoimmunity. Autoimmunity 36, 463–472.Med. 22, 1295–1300.
Rozenski, J., Crain, P., and McCloskey, J. (1999). The RNA Modifi-Crozat, K., and Beutler, B. (2004). TLR7: a new sensor of viral infeccation
Database: 1999 update. Nucleic Acids Res. 27, 196–197.tion. Proc. Natl. Acad. Sci. USA 101, 6835–6836.
Sampson, J.R., and Uhlenbeck, O.C. (1988). Biochemical and phys-Diebold, S.S., Kaisho, T., Hemmi, H., Akira, S., and Reis e Sousa,
ical characterization of an unmodified yeast phenylalanine transferC. (2004). Innate antiviral responses by means of TLR7-mediated
RNA transcribed in vitro. Proc. Natl. Acad. Sci. USA 85, 1033–1037.recognition of single-stranded RNA. Science 303, 1529–1531.
Santini, S.M., Lapenta, C., Logozzi, M., Parlato, S., Spada, M., DiGreenwood, R.C., and Gentry, D.R. (2002). The effect of antibiotic
Pucchio, T., and Belardelli, F. (2000). Type I interferon as a powerfultreatment on the intracellular nucleotide pools of Staphylococcus
adjuvant for monocyte-derived dendritic cell development and ac-aureus. FEMS Microbiol. Lett. 208, 203–206.
tivity in vitro and in Hu-PBL-SCID mice. J. Exp. Med. 191, 1777–
Heil, F., Hemmi, H., Hochrein, H., Ampenberger, F., Kirschning, C.,
1788.
Akira, S., Lipford, G., Wagner, H., and Bauer, S. (2004). SpeciesSen,
G.C., and Sarkar, S.N. (2005). Transcriptional signaling byspecific recognition of single-stranded RNA via Toll-like receptor 7
double-stranded RNA: role of TLR3. Cytokine Growth Factor Rev.and 8. Science 303, 1526–1529.
16, 1–14.
Hemmi, H., Takeuchi, O., Kawai, T., Kaisho, T., Sato, S., Sanjo, H.,
Wang, T., Town, T., Alexopoulou, L., Anderson, J.F., Fikrig, E., andMatsumoto, M., Hoshino, K., Wagner, H., Takeda, K., and Akira, S.
Flavell, R.A. (2004). Toll-like receptor 3 mediates West Nile virus(2000). A Toll-like receptor recognizes bacterial DNA. Nature 408,
entry into the brain causing lethal encephalitis. Nat. Med. 10,740–745.
1366–1373.
Hoffman, R.W., Gazitt, T., Foecking, M.F., Ortmann, R.A., Misfeldt,
Weissman, D., Ni, H., Scales, D., Dude, A., Capodici, J., McGibney,M., Jorgenson, R., Young, S.L., and Greidinger, E.L. (2004). U1 RNA
K., Abdool, A., Isaacs, S.N., Cannon, G., and Kariko, K. (2000). HIVinduces innate immunity signaling. Arthritis Rheum. 50, 2891–2896.
gag mRNA transfection of dendritic cells (DC) delivers encoded anHouge,
G., Robaye, B., Eikhom, T., Golstein, J., Mellgren, G., Gjerttigen
to MHC class I and II molecules, causes DC maturation, and
sen, B., Lanotte, M., and Doskeland, S. (1995). Fine mapping of 28S
induces a potent human in vitro primary immune response. J. ImrRNA
sites specifically cleaved in cells undergoing apoptosis. Mol. munol. 165, 4710–4717.
Cell. Biol. 15, 2051–2062.
Ito, T., Amakawa, R., Kaisho, T., Hemmi, H., Tajima, K., Uehira, K.,
Ozaki, Y., Tomizawa, H., Akira, S., and Fukuhara, S. (2002). Interferon-a
and Interleukin-12 are induced differentially by Toll-like
receptor 7 ligands in human blood dendritic cell subsets. J. Exp.
Med. 195, 1507–1512.
Jurk, M., Heil, F., Vollmer, J., Schetter, C., Krieg, A.M., Wagner, H.,
Lipford, G., and Bauer, S. (2002). Human TLR7 or TLR8 independently
confer responsiveness to the antiviral compound R-848. Nat.
Immunol. 3, 499.
Kane, S., and Beemon, K. (1987). Inhibition of methylation at two
internal N6-methyladenosine sites caused by GAC to GAU mutations.
J. Biol. Chem. 262, 3422–3427.
Kariko, K., Kuo, A., Barnathan, E.S., and Langer, D.J. (1998). Phosphate-enhanced
transfection of cationic lipid-complexed mRNA
and plasmid DNA. Biochim. Biophys. Acta 1369, 320–334.
Kariko, K., Ni, H., Capodici, J., Lamphier, M., and Weissman, D.
(2004). mRNA is an endogenous ligand for Toll-like receptor 3. J.
Biol. Chem. 279, 12542–12550.
Kierzek, E., and Kierzek, R. (2003). The thermodynamic stability of
RNA duplexes and hairpins containing N6-alkyladenosines and
2-methylthio-N6-alkyladenosines. Nucleic Acids Res. 31, 4472–
4480.
Kim, D.H., Longo, M., Han, Y., Lundberg, P., Cantin, E., and Rossi,
J.J. (2004). Interferon induction by siRNAs and ssRNAs synthesized
by phage polymerase. Nat. Biotechnol. 22, 321–325. Published online:
February 8, 2004. 10.1038/nbt940.
Koski, G.K., Kariko, K., Xu, S., Weissman, D., Cohen, P.A., and
Czerniecki, B.J. (2004). Cutting edge: innate immune system
discriminates between RNA containing bacterial versus eukaryotic
structural features that prime for high-level IL-12 secretion by dendritic
cells. J. Immunol. 172, 3989–3993.
Margulis, L., and Chapman, M.J. (1998). Endosymbioses: cyclical
and permanent in evolution. Trends Microbiol. 6, 342–345.
Matsumoto, M., Funami, K., Tanabe, M., Oshiumi, H., Shingai, M.,
Seto, Y., Yamamoto, A., and Seya, T. (2003). Subcellular localization
of Toll-like receptor 3 in human dendritic cells. J. Immunol. 171,
3154–3162.
Medzhitov, R. (2001). Toll-like receptors and innate immunity. Nat.
Rev. Immunol. 1, 135–145.
Narayan, P., Ayers, D.F., Rottman, F.M., Maroney, P.A., and Nilsen,
T.W. (1987). Unequal distribution of N6-methyladenosine in influenza
virus mRNAs. Mol. Cell. Biol. 7, 1572–1575.
Ronnblom, L., Eloranta, M.L., and Alm, G.V. (2003). Role of natural