Toll-like receptors 9 and 3 as essential complonents of innate immune defense against mouse cytomegalovirus infection

Toll-like receptors 9 and 3 as essential componentsof innate immune defense against mousecytomegalovirus infectionKoichi Tabeta*†, Philippe Georgel*†, Edith Janssen‡, Xin Du*, Kasper Hoebe*, Karine Crozat*, Suzanne Mudd*,Louis Shamel*, Sosathya Sovath*, Jason Goode*, Lena Alexopoulou§, Richard A. Flavell§¶�, and Bruce Beutler*�

*Department of Immunology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037; ‡Division of Cellular Immunology, La JollaInstitute for Allergy and Immunology, 10355 Science Center Drive, San Diego, CA 92121; and §Section of Immunobiology and ¶The Howard Hughes MedicalInstitute, Yale University, New Haven, CT 06520

Contributed by Richard A. Flavell, January 22, 2004

Several subsets of dendritic cells have been shown to produce type IIFN in response to viral infections, thereby assisting the natural killercell-dependent response that eliminates the pathogen. Type I IFNproduction can be induced both by unmethylated CpG-oligode-oxynucleotide and by double-stranded RNA. Here, we describe acodominant CpG-ODN unresponsive phenotype that results from anN-ethyl-N-nitrosourea-induced missense mutation in the Tlr9 gene(Tlr9CpG1). Mice homozygous for the Tlr9CpG1 allele are highly suscep-tible to mouse cytomegalovirus infection and show impaired infec-tion-induced secretion of IFN-��� and natural killer cell activation. Wealso demonstrate that both the Toll-like receptor (TLR) 9 3 MyD88and TLR3 3 Trif signaling pathways are activated in vivo on viralinoculation, and that each pathway contributes to innate defenseagainst systemic viral infection. Whereas both pathways lead to typeI IFN production, neither pathway offers full protection against mousecytomegalovirus infection in the absence of the other. The Tlr9CpG1

mutation alters a leucine-rich repeat motif and lies within a receptordomain that is conserved within the evolutionary cluster encompass-ing TLRs 7, 8, and 9. In other TLRs, including three mouse-specific TLRsdescribed in this paper, the affected region is not represented. Thephenotypic effect of the Tlr9CpG1 allele thus points to a critical role forTLR9 in viral sensing and identifies a vulnerable amino acid within theectodomain of three TLR proteins, essential for a ligand response.

Natural killer cells (NK) make an essential contribution tomammalian innate immune defense against viral infections

(1). NK can directly sense viral pathogens through a specific andcomplex set of inhibitory and activating receptors (2–4). Themolecular interactions that are required for viral sensing andresponse have been best documented in the case of mouse cyto-megalovirus (MCMV) infection. The MCMV-encoded proteinm157 interacts directly with the NK membrane protein LY49H,which contains an immunoreceptor tyrosine-based activation motif(ITAM) (5, 6). On ligand binding, LY49H causes NK activation,manifested in part by IFN-� and perforin production.

However, an effective NK response also depends on extrinsicsignals. Once MCMV infection is established, NK are activated bycytokines and chemokines, including type I IFN and IL-12, whichare secreted by dendritic cells (DC). These molecules are known toinfluence the outcome of MCMV infection (7–9). Indeed, duringMCMV infection, bidirectional communication between DC andNK is established. Both DC-derived cytokines and direct pathogenrecognition are essential for NK activation. In turn, NK participatein the maintenance of the DC population (10). However, little isknown about the earliest events that lead to type I IFN synthesis,which is known to occur during viral infection.

The host sensors that initially detect viral pathogens and causecytokine production by myeloid cells have been investigated byseveral groups, some of which have pointed to possible involve-ment of Toll-like receptors (TLRs) (11, 12). The TLRs constitutea family of ligand-binding molecules that engage a variety ofmicrobial products (13). On binding, they activate signaling

cascades leading to the synthesis of proinflammatory molecules.TLR3, which recognizes double-stranded RNA (dsRNA) (14),and TLR9, which recognizes double-stranded DNA unmethyl-ated at CpG motifs (15, 16), are likely to play a key role in viralinfections (17). Indeed, two groups have reported that herpessimplex virus (HSV)-1- and HSV-2-induced IFN-� production isabrogated in cells harvested from TLR9 knockout mice (11, 12).A role for TLR3 in viral detection has been suggested by in vivoand ex vivo analyses that showed that this receptor is involved inthe recognition of poly I:C and dsRNA from Lang reovirus (14).However, none of these studies has demonstrated that theseTLRs are crucial for an efficient antiviral response in vivo.

We previously described the phenotype of Lps2 mice, whichcarry a nonsense mutation in the Trif�Ticam-1 gene (18, 19). TheTrif gene encodes an essential adaptor molecule for MyD88-independent signaling downstream from TLRs 3 and 4 (18, 20).In macrophages harvested from homozygous Lps2 mutant mice,neither LPS nor dsRNA is capable of triggering phosphorylationor dimerization of IFN response factor (IRF)-3. Moreover, theanimals display impaired responses to MCMV, fail to producetype I IFN when infected, and exhibit an abnormally elevatedsplenic viral titer during infection.

In the present study, we have identified a codominant N-ethyl-N-nitrosourea (ENU)-induced phenotype in which homozygousanimals fail to respond to CpG-oligodeoxynucleotide (ODN)and also show severely impaired responses to MCMV infectionas evidenced by viral titers measurement and survival analysis.This immunodeficiency phenotype is caused by a hypomorphicTlr9 allele (Tlr9CpG1), encoding a structurally aberrant receptor.Tlr9CpG1 homozygotes display a low level of cytokine inductionand NK activation on viral infection. Mice homozygous for a nullallele of MyD88 show a similar phenotype, indicating that theTLR9 3 MyD88 axis is essential for effective innate antiviraldefense. We also tested the response to MCMV infection in micehomozygous for a null allele of Tlr3. Our results indicate thatTLR3 plays a significant role, although a less crucial role thanTLR9, in defense against systemic MCMV infection.

Materials and MethodsMice. C57BL�6, BALB�c, Tlr3�/� mice (14), MyD88�/� mice (pro-vided by S. Akira, Osaka University, Osaka), and TLR9CpG1/CpG1

Abbreviations: DC, dendritic cells; dsRNA, double-stranded RNA; ENU, N-ethyl-N-nitrosourea; LRR, leucine-rich repeat; MCMV, mouse cytomegalovirus; NK, natural killercells; ODN, oligodeoxynucleotide; pfu, plaque-forming units; TLR, Toll-like receptor; HSV,herpes simplex virus; TNF, tumor necrosis factor; TCR, T cell antigen receptor.

Data deposition: The sequences reported in this paper have been deposited in the GenBankdatabase (accession nos. AY510704–AY510706).

†K.T. and P.G. contributed equally to this work.

�To whom correspondence should be addressed. E-mail: [email protected] [email protected].

© 2004 by The National Academy of Sciences of the USA

3516–3521 � PNAS � March 9, 2004 � vol. 101 � no. 10 www.pnas.org�cgi�doi�10.1073�pnas.0400525101

mice were maintained under pathogen-free conditions in TheScripps Research Institute animal care facility. All mice used in theexperiments were 6–9 weeks in age. All experimental procedureswere conducted in accordance with institutional guidelines foranimal care and use.

ENU Mutagenesis and in Vitro Stimulation of Thioglycolate-ElicitedMacrophages. Germ-line mutagenesis was performed in mice byusing ENU (19). Ex vivo stimulation of peritoneal macrophageswith multiple microbial inducers and measurement of secretedtumor necrosis factor (TNF)-� activity were conducted as de-scribed in refs. 18 and 19. For analysis of the Tlr9CpG1/CpG1

phenotype, CpG-ODN (5�-tccatgacgttcctgatgct-3�) was used atthe stated concentrations. Secreted TNF-� activity was mea-sured 4 hours after induction by using the L-929 cell bioassay.

Virus Culture and Assay. MCMV (Smith strain) was prepared by invivo propagation in 3-week-old BALB�c mice infected with 1 �104 plaque-forming units (pfu) by an i.p. route. Salivary glandhomogenates were prepared at day 14 postinfection, and viraltiter was determined by plaque assay (21). Splenic viral titerswere determined by plaque assay 4 days after MCMV infection.

Cytokine Measurements. IL-12 and IFN-� were measured in serumharvested 36 h after MCMV infection by using an ELISA (R &D Systems). Type I IFN titers were determined with a bioassay,

measuring the protection conferred by serially diluted serumagainst the cytopathic effect of the vesicular stomatitis virus inL929 cells (22). Purified IFN-� (R & D Systems) was used as astandard.

Fluorescence-Activated Cell Sorting Assay. Splenocytes were har-vested and passed through a cell strainer 36 h postinfection, intoDulbecco’s modified Eagle’s medium containing 10% FBS and2% penicillin�streptomycin solution (GIBCO). After fixationand permeabilization, cells were stained for the surface markersNK1.1 and T cell antigen receptor (TCR) � as well as intracel-lular IFN-� by using antibodies from BD Bioscience. Flowcytometry was performed by using a FACSCalibur cell sorterand analyzed with CELLQUEST software.

Transfection of HEK 293 Cells and Assay of TLR9 Signaling Activity.Full-length muTLR9 cDNAs (amplified from C57BL6�J andTlr9CpG1/CpG1 mutant peritoneal macrophage mRNA) werecloned into pFlagCMV3 mammalian expression vector (Sigma).Plasmids were isolated by using the Endo-free Maxi prep kit(Qiagen, Valencia, CA). HEK293 cells (2 � 105) were trans-fected by using Lipofectamine (Invitrogen) with 0.4 �g of eachexpression vector and endothelial leukocyte adhesion molecule1 (ELAM-1) luciferase reporter vector (23). After 24 h, cellswere stimulated with 0.1 and 1.0 �M CpG-ODN for 5 h.

Fig. 1. Isolation and identification of CpG1, an ENU-induced point mutation in TLR9. (A) Trace file of amplified genomic DNA from homozygous mutant mice(upper chromatogram) and normal animals (lower chromatogram). (Left) Location of the mutation within the 11th LRR in the TLR9 ectodomain. (Lower) Multiplealignment of all human and mouse TLR protein sequences reveals that the mutation resides in a region shared only by TLRs 7, 8, and 9. At left, the identity ofeach sequence is indicated. H, human paralog; M, mouse paralog. Note that TLRs 11 and 12 are nearest homologs, and TLR13 is most closely related to TLR3. (B)Transfection-based assay of TLR9 signaling activity. Error bars indicate the SD of duplicate transfections.

Tabeta et al. PNAS � March 9, 2004 � vol. 101 � no. 10 � 3517

IMM

UN

OLO

GY

Luciferase assays were performed by using the Steady-Gloluciferase assay system (Promega).

Identification and cDNA Cloning of TLRs 11, 12, and 13. All murineand human ESTs were downloaded from the National Center forBiotechnology Information database, translated in all six readingframes by using the program PEPDATA, and searched for Toll�IL-1R homology (TIR) domain homology (HMMERSEARCH, E �10) by using a hidden Markov model based on representatives ofall known animal TIR domains. Three hits were acquired amongthe mouse ESTs. The corresponding cDNAs were amplifiedfrom macrophage mRNA by using primers based on genomicsequence to capture complete coding regions. Alignment of allhuman and mouse TLR protein sequences was performed byusing the CLUSTALW program (gap opening penalty � 10; gapextension penalty � 0.20; PAM series as the protein weightmatrix). The TLR coding sequences were submitted to GenBank(accession nos. AY510704–AY510706).

ResultsIdentification of an ENU-Induced CpG-ODN Non-Responder Phenotypeand Determination of the Causative Mutation Within the Tlr9 Locus.Among �11,000 mice examined to date, we identified an F3mutant male in which thioglycolate-elicited peritoneal macro-phages were markedly hyporesponsive to stimulation with CpG-ODN. Normal TNF-� production was observed when macro-phages were stimulated with all other TLR agonists (data notshown). This finding suggested that the mutation, termed CpG1,specifically affected TLR9 signaling.

On sequencing the Tlr9 coding region, a T3 C base transitionwas observed at position 1496 (gi:13626029), predicting the aminoacid substitution L499P (Fig. 1A). To examine the affected residuewith reference to all other mammalian TLRs, a HMMERSEARCH ofall human and mouse ESTs was performed to generated an updatedset of TIR domain proteins for alignment. Three mouse TLRs wereidentified (TLRs 11, 12, and 13). They have not previously beenreported and lack human orthologs. The total complement ofmouse TLR proteins now numbers 12 because the mouse orthologof human TLR10 is a degenerate pseudogene.

All 10 human TLRs, and all 12 mouse TLRs, were optimallyaligned. Residue 499 of the mouse TLR9 protein lies within apart of the receptor ectodomain that, on comparison with allother TLRs in the multiple alignment plot, is represented onlyin TLRs 7, 8, and 9 [known to represent an evolutionary cluster(24)]. The leucine in question is located near the beginning of theeleventh leucine-rich repeat (LRR) motif predicted by simplemodular research tool (SMART) analysis. In TLR8, an isoleucineis represented at this position.

When reconstructed and overexpressed in transfected HEK293 cells, the mutant TLR9 protein was insensitive to stimulationwith CpG-ODN whereas the wild-type protein, expressed in thesame system, was strongly CpG-ODN responsive (Fig. 1B). Inaddition, homozygous mutant mice were bred to normalC57BL�6 mice and then backcrossed to the mutant stock so thatTlr9CpG1 homozygotes and heterozygotes were obtained. Cellsfrom Tlr9CpG1/CpG1 mice were uniformly unresponsive to CpG-ODN (0.1 �M concentration) stimulation whereas cells fromTlr9�/� mice were uniformly responsive. HeterozygotesTlr9CpG1/� exhibited an intermediate phenotype. These dataestablish that the CpG1 phenotype is linked to the observedmutation in Tlr9 (P � 0.0001) and also reveal that the mutantallele is codominant, or alternatively, that the wild-type allele ishaploinsufficient (Fig. 2A).

The effect of heterozygosity for the Tlr9CpG1 allele on TNF-�and IL-12 secretion was further examined by varying the con-centration of CpG-ODN used as a stimulus. A high CpG-ODNconcentration (1 �M) allows a clear-cut distinction betweenTlr9CpG1/� and Tlr9CpG1/CpG1 genotypes but fails to resolve

Tlr9CpG1/� and Tlr9�/� genotypes (Fig. 2 B and C). Hence, thecodominant character of the Tlr9CpG1 allele is observed only atlow CpG-ODN concentrations (�0.1 �M).

The Tlr9CpG1/CpG1 genotype was found to protect D-galactosamine-sensitized animals against the lethal effect ofCpG-ODN-induced TNF-� in vivo. Homozygous mutants andwild-type controls were sensitized with D-galactosamine (20 mgper mouse) and then injected with CpG-ODN (20 nmol permouse). One hundred percent of the homozygous mutantssurvived the injection whereas 83% of wild-type C57BL�6 micesuccumbed within 10 h (Fig. 2D; P � 0.0025).

The TLR9 3 MyD88 Signaling Axis Is Critical to Control MCMVInfection. Tlr9CpG1/CpG1 mice were examined in a second screen,designed to identify mutations that impair MCMV resistance. Inthis screen, a systemic infection is induced by i.p. injection ofMCMV, by using an inoculum (5 � 105 pfu) that is well toleratedby normal C57BL�6 mice. This protocol has been established todistinguish between sensitive strains (such as BALB�c), whichsuccumb 4–5 days after the inoculation, and resistant strains(such as C57BL�6), which survive even 8 days after the infection.

As shown in Fig. 3A, Tlr9CpG1/CpG1 mice accumulate exagger-ated viral loads 4 days after infection, with titers approachingthose observed in the BALB�c strain. Because TLR9 signalsthrough MyD88 (25), we determined the viral load in MyD88�/�

mice and observed that these animals also show high viral titers.We have previously reported that the Lps2 allele of the

Trif�Ticam-1 gene impairs macrophage IFN-� production inresponse to dsRNA. As a result, these mice are also permissivefor MCMV infection (18). Because Trif is necessary for TLR3signaling (18, 20), which is MyD88-independent (18), we rea-soned that either component of the Trif 3 TLR3 axis mightimpair the immune response to MCMV. Accordingly, Tlr3�/�

mice were examined for MCMV susceptibility. The average viraltiter in the spleen of infected Tlr3�/� mice never reached thatobserved in Tlr9CpG1/CpG1 or MyD88�/� mice but showed a highlysignificant increase compared with wild-type controls (Fig. 3A),

Fig. 2. In vivo effects of the Tlr9CpG1 mutation. (A) TNF-� secretion by macro-phages from controls C57BL�6 (Tlr9�/�, filled dots), heterozygotes (Tlr9CpG1/�,gray dots) and homozygotes (Tlr9CpG1/CpG1, open dots) animals after CpG-ODNinduction (0.1 �M). Each dot represents the result of a duplicate induction assayperformed on cells from a single animal. (B) TNF-� production or (C) IL-12p40production by peritoneal macrophages at low (0.1 �M) CpG ODN concentration,as influenced by Tlr9 genotype. Error bars indicate SD; n � 5 mice. (D) Kaplan–Meier survival curves for Tlr9CpG1/CpG1 mice and Tlr9�/� mice, after sensitizationwith D-galactosamine and challenge with CpG-ODN. Mice were monitored for a3-day period, at which time all survivors seemed healthy.

3518 � www.pnas.org�cgi�doi�10.1073�pnas.0400525101 Tabeta et al.

comparable to that previously reported in TrifLps2/Lps2 mutantmice (19). In the course of our infection studies, we repeatedlyobserved that, as in BALB�c mice, MyD88�/� and Tlr9CpG1/CpG1

mice showed obvious signs of sickness 4 days after MCMVinoculation. As shown in Fig. 3B, death follows viral infection ofTlr9CpG1/CpG1 and MyD88�/� mice almost as rapidly as it does inBALB�c animals. Tlr3�/� mutants showed no significant sur-vival difference compared with C57BL�6 controls although, at asubjective level, the Tlr3�/� animals seemed sicker than controls,and a trend toward higher mortality was observed.

Cytokine Induction and NK Activation Are Impaired in Tlr9CpG1�CpG1

and MyD88�/� Mice. Because the secretion of certain cytokines is awell known and essential response for the clearance of viralinfections, we infected C57BL�6 control mice and mutant mice withan inoculum of virus that was sublethal during the term of theexperiment (5 � 104 pfu per mouse) and monitored their cytokinelevels in the serum at a time (36 h postinfection) when this responseis at its maximum (26). As shown in Fig. 4A, type I IFN secretionis dramatically reduced in MyD88�/� mice as well as in Tlr9CpG1/CpG1

mice. IL-12 p40 levels (Fig. 4B) and IFN-� levels (Fig. 4C) are alsodiminished in the homozygous mutants. We also note that IFN-�production is strongly inhibited in MyD88�/� animals, which mightbe attributable to additional signaling defects in these mice (seeDiscussion). The decrease in serum cytokine levels is less pro-nounced in Tlr3�/� mice but is statistically significant.

A clear link has been established between cytokine induction andNK activation during MCMV infection (8–10). We analyzed splenicNK and NKT cell populations 36 h postinfection by gating onNK1.1��TCR�� (for NK) and NK1.1��TCR�� (for NKT cells)and then measuring intracellular IFN-�. Fig. 5 shows the results ofthis analysis. The values (mean � SD, n � 4 mice) indicate theproportion of IFN-�� NK or NKT cells among the total population.These data illustrate the pronounced effect (a 4-fold reduction) ofboth the Tlr9CpG1/CpG1 and the MyD88�/� genotypes on NK andNKT cell activation that occurs in the course of MCMV infection.The Tlr3�/� genotype was associated with a smaller, but stillsignificant, decrease in the proportion of activated NK and NKTcells.

DiscussionThe L499P substitution specified by Tlr9CpG1 falls within a centrallyplaced LRR motif: 1 of 19 such motifs found in the TLR9

Fig. 3. Tlr9CpG1/CpG1, MyD88�/�, and Tlr3�/� mice are hypersusceptible to viralinfections. (A) Viral titers, expressed as log pfu per spleen, were determined inmice 4 days after i.p. inoculation with 5 � 105 pfu of MCMV. (B) Mice of theindicated genotype were infected i.p. with 5 � 105 pfu of MCMV, and survivalwas monitored for a period of 7 days. P values indicate comparisons with thesurvival curve of C57BL�6 control animals.

Fig. 4. In vivo impairment of cytokine production after MCMV infection in micelacking the TLR93MyD88 signaling pathway. (A) IFN-��� activity was measuredin the serum of noninfected (ni) controls and MCMV-infected animals 36 hpostinoculation. Values are expressed as international units (IU)�ml of serum.IL-12p40(B)and IFN-� (C) concentrationsweremeasured intheserumofthesameanimals as those shown in A by ELISA and are expressed in ng�ml. Data representmean values with SD (n � 4 mice). Statistical analysis was performed by using theANOVA test with PRISM software (GraphPad, San Diego) (*, P � 0.05, **, P � 0.001,with respect to the C57BL�6 control values).

Tabeta et al. PNAS � March 9, 2004 � vol. 101 � no. 10 � 3519

IMM

UN

OLO

GY

ectodomain. Within this LRR, the mutation would be expected todisrupt an alpha helix that normally contributes to the single loopformed by all LRRs (27). The region of the receptor within whichthe mutation occurs is represented only in a subset of TLR proteins.Neither the previously published TLRs (TLRs 1–6 and TLR10) northe three mouse TLRs reported herein (TLRs 11, 12, and 13)exhibit a homologous region. It might therefore be imagined thatthe residue in question has a specialized function, related to thetypes of ligands that are engaged by TLRs 7, 8, and 9 (nucleotide-based molecules).

The Tlr9CpG1 allele was identified in two independent screen-ing procedures: an ex vivo assay designed to identify newcomponents of the TLR signaling pathways, and an in vivoprotocol used to recover mutations that affect the innate anti-viral response. The mutation markedly impairs the antiviralresponse, and the magnitude of the impairment is very similar tothat associated with the MyD88�/� genotype, and indeed, to thatassociated with the BALB�c genotype. Hence, absent input fromthe TLR93MyD88 signaling axis, the normal m1573 LY49H

sensing apparatus within the NK of the C57BL�6 mouse isinadequate to contain an MCMV infection.

Although MyD88 serves most of the TIR domain receptors(excluding TLR3), it is logical to suppose that most of theprotection that MyD88 affords during MCMV infection resultsfrom its interaction with TLR9. Consistent with this conclusion,TLR2 and TLR4 deficiencies have no influence on the course ofMCMV infection (P.G., unpublished results). On the other hand,the viral resistance phenocopy observed in MyD88�/� andTlr9CpG1/CpG1 is imperfect. The exceptionally low level of serumIFN-� observed in infected MyD88�/� animals (Fig. 4C) mightreflect a requirement for MyD88 in certain TLR-independentsignaling pathways for NK activation. IL-18, which signalsthrough MyD88 (28) but not Trif (P.G., E.J., and K.H., unpub-lished results) and is essential for NK expansion after mCMVinfection (10), may account for the discrepancy.

Because the DNA of herpesviruses such as MCMV is G�C-richand has immunostimulatory activities (29), the MCMV-sensitivityphenotype that we have observed is likely due to a lack of

Fig. 5. Impairment of NK and NKT cell activation after MCMV inoculation in TLR9 and MyD88 mutant mice. Purified splenic cells from uninfected orMCMV-infected animals were recovered 36 h postinoculation and gated against NK1.1 and TCR� surface markers. (Left) IFN-� intracellular staining (and isotypeantibody as control) for NK (NK1.1�TCR��). (Right) Results for NKT cells (NK1.1�TCR��). Mean values � SD are indicated (n � 4 mice).

3520 � www.pnas.org�cgi�doi�10.1073�pnas.0400525101 Tabeta et al.

TLR9-mediated viral DNA recognition and resulting impairmentof IFN-��� secretion. Two independent studies have recentlyidentified plasmacytoid dendritic cells (pDC) and�or IFN produc-ing cells (IPC) as the source of type I IFN production during HSV-1and HSV-2 infections (11, 12) and have shown that IFN productionis TLR9-dependent. Our data do not permit discrimination be-tween these two categories of DCs as principal sources of type IIFNs in MCMV infection. We have examined the splenic pDC(CD8��CD11c��CD11b�) population in MCMV-infectedMyD88�/� mice and observed that they undergo a reduction innumbers at day 1.5 after infection, and an expansion at day 4 afterinfection: a pattern also observed in wild-type control animals (datanot shown). This finding, however, does not rule out their involve-ment in viral DNA recognition because this pattern of response isitself dependent on NK activation (10). It has also been suggestedthat IFN-��� produced by IPC on viral challenge can activateimmature DC (30), which in turn secrete type I IFN. Alternatively,both types of DC, as well as macrophages [which also express TLR9(31)], may be activated by MCMV in vivo and synergize in resistingthe viral pathogen.

Our data also implicate TLR3 as a key participant in the antiviralresponse. Approximately 1,000-fold augmentation of viral load inthe spleen was observed in Tlr3�/� mice inoculated with MCMV,a finding consistent with earlier studies that showed a comparableincrease in TrifLps2/Lps2 mice (18). TLR3 is known to serve as aligand for dsRNA, and we presume that dsRNA may be producedas a consequence of bidirectional transcription from the MCMVgenome in the course of infection (32).

The fact that impairment of either TLR3 or TLR9 signalingpathways has a dramatic effect on the course of disease issomewhat surprising. The TLR33 Trif signaling axis is believedto be independent of the TLR9 3 MyD88 signaling axis. Yetboth signals lead to the induction of type I IFN (12, 18, 20).Abrogation of TLR3 signaling causes a �60% decrement, andabrogation of TLR9 signaling causes a �90% decrement in theamount of type I IFN that is measured in serum after infection.Hence, the two pathways seem to elicit the production of type IIFN in a superadditive or codependent manner.

It is possible that this relatively modest superadditivity resultsfrom signal transducer and activator of transcription (STAT)-1-mediated induction of additional type I IFN in response to IFN thatis produced (33). However, when an infectious endpoint is exam-ined, the effect of mutational inactivation of either the TLR3 orTLR9 pathway is even more pronounced, suggesting that numerousfunctional defects (rather than the relatively modest observeddefect of type I IFN production) contribute to immunocompro-mise. MCMV titer is elevated by three orders of magnitude as theresult of Tlr3 mutation, and by four orders of magnitude as theresult of Tlr9 mutation. Whereas it might have been supposed thateither pathway would complement the loss of the other, it seems,to the contrary, that both pathways are essential for containment ofinfection. Are these the only two pathways that allow DC to senseMCMV infection? It is possible that dsRNA sensing may also occurby means of protein kinase R (PKR), which may prompt theproduction of type I IFN (34). The relative importance of thispathway in DC remains to be established. Compound homozygousmutants (Tlr3�/�;Tlr9CpG1/CpG1) are currently being established andshould clarify whether residual awareness of infection exists, absentboth receptors for viral sensing.

In the past few years, a large volume of data has demonstratedthe involvement of TLRs in the defense against bacterial infections,but their role in viral infection has remained unclear. In contrast toprevious reports (11, 12), our in vivo infectious model establishes aprominent and essential function for TLR9 in MCMV resistancebecause mice devoid of CpG-mediated signaling quickly die afterviral inoculation. Furthermore, our data suggest that both MyD88-dependent and -independent signals are essential for cytokineresponses. TLR3- and TLR9-dependent type I IFN productionactivates NK, which in turn confine the viral pathogen and preventits rapid spread during the critical interval before activation of anadaptive immune response.

We thank Marc Dalod (Centre d’Immunologie de Marseille-Luminy,Marseille, France) for helpful advice. This work was supported byNational Institutes of Health Grant U54A154523 and by the FondationPhilippe (to P.G.).

1. French, A. R. & Yokoyama, W. M. (2003) Curr. Opin. Immunol. 15, 45–51.2. Brown, M. G., Dokun, A. O., Heusel, J. W., Smith, H. R., Beckman, D. L.,

Blattenberger, E. A., Dubbelde, C. E., Stone, L. R., Scalzo, A. A. & Yokoyama,W. M. (2001) Science 292, 934–937.

3. Daniels, K. A., Devora, G., Lai, W. C., O’Donnell, C. L., Bennett, M. & Welsh,R. M. (2001) J. Exp. Med. 194, 29–44.

4. Lee, S. H., Girard, S., Macina, D., Busa, M., Zafer, A., Belouchi, A., Gros, P.& Vidal, S. M. (2001) Nat. Genet. 28, 42–45.

5. Arase, H., Mocarski, E. S., Campbell, A. E., Hill, A. B. & Lanier, L. L. (2002)Science 296, 1323–1326.

6. Smith, H. R., Heusel, J. W., Mehta, I. K., Kim, S., Dorner, B. G., Naidenko,O. V., Iizuka, K., Furukawa, H., Beckman, D. L., Pingel, J. T., et al. (2002) Proc.Natl. Acad. Sci. USA 99, 8826–8831.

7. Asselin-Paturel, C., Boonstra, A., Dalod, M., Durand, I., Yessaad, N., Dezutter-Dambuyant, C., Vicari, A., O’Garra, A., Biron, C., Briere, F. & Trinchieri, G.(2001) Nat. Immunol. 2, 1144–1150.

8. Dalod, M., Salazar-Mather, T. P., Malmgaard, L., Lewis, C., Asselin-Paturel,C., Briere, F., Trinchieri, G. & Biron, C. A. (2002) J. Exp. Med. 195, 517–528.

9. Dalod, M., Hamilton, T., Salomon, R., Salazar-Mather, T. P., Henry, S. C.,Hamilton, J. D. & Biron, C. A. (2003) J. Exp. Med. 197, 885–898.

10. Andrews, D. M., Scalzo, A. A., Yokoyama, W. M., Smyth, M. J. & Degli-Esposti, M. A. (2003) Nat. Immunol. 4, 175–181.

11. Krug, A., Luker, G. D., Barchet, W., Leib, D. A., Akira, S. & Colonna, M.(2003) Blood 103, 1433–1437.

12. Lund, J., Sato, A., Akira, S., Medzhitov, R. & Iwasaki, A. (2003) J. Exp. Med.198, 513–520.

13. Takeda, K., Kaisho, T. & Akira, S. (2003) Annu. Rev. Immunol. 21, 335–376.14. Alexopoulou, L., Holt, A. C., Medzhitov, R. & Flavell, R. A. (2001) Nature 413,

732–738.15. Hemmi, H., Takeuchi, O., Kawai, T., Kaisho, T., Sato, S., Sanjo, H., Matsu-

moto, M., Hoshino, K., Wagner, H., Takeda, K. & Akira, S. (2000) Nature 408,740–745.

16. Ahmad-Nejad, P., Hacker, H., Rutz, M., Bauer, S., Vabulas, R. M. & Wagner,H. (2002) Eur. J. Immunol. 32, 1958–1968.

17. Vaidya, S. A. & Cheng, G. (2003) Curr. Opin. Immunol. 15, 402–407.18. Hoebe, K., Du, X., Georgel, P., Janssen, E., Tabeta, K., Kim, S. O., Goode, J.,

Lin, P., Mann, N., Mudd, S., et al. (2003) Nature 424, 743–748.19. Hoebe, K., Du, X., Goode, J., Mann, N. & Beutler, B. (2003) J. Endotoxin Res.

9, 250–255.20. Yamamoto, M., Sato, S., Hemmi, H., Hoshino, K., Kaisho, T., Sanjo, H.,

Takeuchi, O., Sugiyama, M., Okabe, M., Takeda, K. & Akira, S. (2003) Science301, 640–643.

21. Orange, J. S., Wang, B., Terhorst, C. & Biron, C. A. (1995) J. Exp. Med. 182,1045–1056.

22. Orange, J. S. & Biron, C. A. (1996) J. Immunol. 156, 4746–4756.23. Schindler, U. & Baichwal, V. R. (1994) Mol. Cell. Biol. 14, 5820–5831.24. Du, X., Poltorak, A., Wei, Y. & Beutler, B. (2000) Eur. Cytokine Netw. 11, 362–371.25. Schnare, M., Holt, A. C., Takeda, K., Akira, S. & Medzhitov, R. (2000) Curr.

Biol. 10, 1139–1142.26. Ruzek, M. C., Miller, A. H., Opal, S. M., Pearce, B. D. & Biron, C. A. (1997)

J. Exp. Med. 185, 1185–1192.27. Kobe, B. & Deisenhofer, J. (1995) Curr. Opin. Struct. Biol. 5, 409–416.28. Adachi, O., Kawai, T., Takeda, K., Matsumoto, M., Tsutsui, H., Sakagami, M.,

Nakanishi, K. & Akira, S. (1998) Immunity 9, 143–150.29. Lundberg, P., Welander, P., Han, X. & Cantin, E. (2003) J. Virol. 77,

11158–11169.30. Le Bon, A. & Tough, D. F. (2002) Curr. Opin. Immunol. 14, 432–436.31. Gao, J. J., Diesl, V., Wittmann, T., Morrison, D. C., Ryan, J. L., Vogel, S. N.

& Follettie, M. T. (2002) J. Leukocyte Biol. 72, 1234–1245.32. Jacobs, B. L. & Langland, J. O. (1996) Virology 219, 339–349.33. Marie, I., Durbin, J. E. & Levy, D. E. (1998) EMBO J. 17, 6660–6669.34. Diebold, S. S., Montoya, M., Unger, H., Alexopoulou, L., Roy, P., Haswell,

L. E., Al-Shamkhani, A., Flavell, R., Borrow, P. & Reis e Sousa, C. (2003)Nature 424, 324–328.

Tabeta et al. PNAS � March 9, 2004 � vol. 101 � no. 10 � 3521

IMM

UN

OLO

GY