Protein phosphatase 1 subunit Ppp1r15a/GADD34regulates cytokine production in polyinosinic:polycytidylic acid-stimulated dendritic cellsGiovanna Clavarinoa,b,c,1, Nuno Cláudioa,b,c,1, Alexandre Daleta,b,c, Seigo Terawakia,b,c, Thérèse Coudercd,e,Lionel Chassona,b,c, Maurizio Ceppia,b,c, Enrico K. Schmidta,b,c, Till Wengera,b,c, Marc Lecuitd,e,f, Evelina Gattia,b,c,2,and Philippe Pierrea,b,c,2

aCentre d’Immunologie de Marseille-Luminy, Aix-Marseille University, 13288 Marseille, France; bInstitut National de la Santé et de la Recherche Médicale(INSERM), U631, 13288 Marseille, France; cCentre National de la Recherche Scientifique, Unité Mixte de Recherche 6102, 13288 Marseille, France; dMicrobesand Host Barriers Group, Institut Pasteur, 75015 Paris, France; eINSERM, Equipe avenir U604, 75015 Paris, France; and fService des Maladies Infectieuses etTropicales, Université Paris Descartes, Hôpital Necker-Enfants malades, Assistance Publique-Hôpitaux de Paris, 75015 Paris, France

Edited by Douglas T. Fearon, University of Cambridge School of Clinical Medicine, Cambridge, United Kingdom, and approved December 9, 2011 (received forreview March 21, 2011)

In response to inflammatory stimulation, dendritic cells (DCs) havea remarkable pattern of differentiation that exhibits specific mech-anisms to control the immune response. Here we show that in re-sponse to polyriboinosinic:polyribocytidylic acid (pI:C), DCs mounta specific integrated stress response during which the transcriptionfactor ATF4 and the growth arrest and DNA damage-inducible pro-tein 34 (GADD34/Ppp1r15a), a phosphatase 1 (PP1) cofactor, areexpressed. In agreement with increased GADD34 levels, an exten-sive dephosphorylation of the translation initiation factor eIF2αwasobserved during DC activation. Unexpectedly, although DCs displayan unusual resistance to protein synthesis inhibition induced in re-sponse to cytosolic dsRNA, GADD34 expression did not have amajorimpact on protein synthesis. GADD34, however, was shown to berequired for normal cytokine production both in vitro and in vivo.These observations have important implications in linking furtherpathogen detection with the integrated stress response pathways.

type-I interferon | Toll-like receptor | protein kinase RNA activated |puromycin | unfolded protein response

Dendritic cells (DCs) are regulators of the immune response,which, upon stimulation by conserved microbial products

(PAMPs), are most efficient at inducing differentiation of naiveT cells through massive cytokine production and optimization oftheir antigen presentation capacity (1). Double-stranded RNAs(dsRNA), a hallmark of virus replication, is recognized by Toll-like receptor 3 (TLR3) (2) or cytosolic RNA DExD/H-box hel-icases, such as melanoma-associated gene-5 (MDA5) (3) or theDDX1, DDX21, and DHX36 (4). Detection of pI:C, a dsRNAmimic, promotes the nuclear translocation of IRF-3 and IRF-7and antiviral type-I IFN production (5). IFN triggers, amongmany other antiviral genes, the dsRNA-dependent protein ki-nase (PKR) (6, 7). PKR activation by dsRNA promotes trans-lation initiation factor 2 alpha (eIF2α) phosphorylation leadingto protein synthesis shutoff and inhibition of viral replication (8).PKR is also necessary for normal type-I IFN secretion by DCs inresponse to dsRNA stimulation (9, 10).Different stress signals trigger eIF2α phosphorylation, thus

attenuating mRNA translation and promoting a specific tran-scriptional response known as the integrated stress response(ISR) (11). To date four eIF2α kinases have been identified,PKR, PKR-like ER kinase (PERK) (12), general control non-derepressible-2 (GCN2) (13), and heme-regulated inhibitor(HRI) (14). PERK, which is activated by an excess of unfoldedproteins in the ER lumen (12), is necessary to mount part ofa particular ISR, known as the unfolded protein response (UPR).The UPR encompasses a group of signals that cope with

perturbations in ER homeostasis. In addition to PERK, the UPRis initiated by additional ER sensors and associated transcription

factors such as IRE1, XBP1 and ATF6 (15, 16). In addition toinhibiting global protein synthesis, eIF2α phosphorylation byPERK allows the specific translation of the transcription factorATF4 (14). This factor promotes cell adaptation to stress byheightening ER functions through specific mRNA transcription.Immune responses and immune cell development can be pro-foundly affected by abnormalities in the UPR (17, 18) and TLR-dependent activation of XBP1, via the production of reactiveoxygen species (ROS), is required for normal production ofproinflammatory cytokines in macrophages (19).We show here that activated DCs resist the translation arrest,

normally elicited by the PKR-dependent eIF2α phosphorylation inresponse to cytosolic dsRNA delivery. Concomitantly, pI:C de-tection promotes the expression of the inducible phosphatase 1(PP1)-cofactor, GADD34 (20). GADD34 efficiently dephosphor-ylates eIF2α in activated DCs, but has little impact on the specificresistance of DCs to dsRNA-triggered translation arrest. Alter-natively, GADD34 is required for optimal transcription and pro-duction of the cytokines IFN-β and interleukin-6 (IL-6), suggestingthat induction of the ATF4-dependent branch of the UPR in DCsis a critical signaling module of the innate response to dsRNA.

ResultsDC Stimulation by pI:C Induces the ATF4 Transcription Factor. Toidentify signaling pathways involved in dsRNA response, agenomewide expression analysis was performed in pI:C-stimu-lated mouse bone marrow-derived DCs (bmDCs) using Affy-metrix Mouse Genome 430 2.0 arrays. We found that at leastnine transcripts, typically expressed in different cells exposed totunicamycin (11, 21, 22), were also induced in DCs respondingto pI:C (Table S1).Up-regulation of these transcripts was confirmed by quanti-

tative RT-PCR (qPCR) (Fig. 1 and Fig. S1). Among them, thekey UPR transcription factors, ATF4 and CCAAT/enhancerbinding protein homologous protein (CHOP) mRNAs were

Author contributions: G.C., N.C., E.G., and P.P. designed research; G.C., N.C., A.D., S.T.,T.C., L.C., M.C., E.K.S., and T.W. performed research; G.C., N.C., A.D., M.L., E.G., andP.P. analyzed data; and G.C., N.C., E.G., and P.P. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The expression profiling by arrayGEO data have been deposited in theGene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no.GSM452757).1G.C. and N.C. contributed equally to this work.2To whom correspondence may be addressed. E-mail: pierre@ciml.univ-mrs.fr or gatti@ciml.univ-mrs.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1104491109/-/DCSupplemental.

3006–3011 | PNAS | February 21, 2012 | vol. 109 | no. 8 www.pnas.org/cgi/doi/10.1073/pnas.1104491109

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

increased by two- and eightfold, respectively (Fig. 1 A and C).ATF4 mRNA is normally poorly translated in unstressed cells, butupon eIF2α phosphorylation, a rapid synthesis of the ATF4 proteinhas been observed (14). In DCs, ATF4 protein levels were in-creased upon pI:C stimulation and detected in the nuclear extractsafter 8 h of stimulation, similarly to control DCs, in which a bonafide UPR was induced with thapsigargin (Fig. 1B and Fig. S2).Nuclear translocation of ATF4 normally induces CHOP ex-

pression, which can trigger apoptosis, and enhance the transcrip-tion of several ATF4 target genes such as GADD34 and ERO1-α(22). As expected from the array analysis, increasedCHOPmRNAlevels were detected (Fig. 1C), however, we failed to visualizeprotein expression in LPS and pI:C-stimulated DCs extracts. Incontrast, CHOP was detected in control tunicamycin-treated DCas well as in LPS-stimulated mouse leukaemic macrophages(RAW 264.7 cell line) (Fig. 1D). We tested different doses andmodes of pI:C delivery to DCs and in agreement with the lack ofCHOP expression, we found no major apoptotic death inductionin our experimental system (Fig. S3A). Thus, pI:C-activated DCsdisplay an ATF4 transcriptional signature during which, CHOPexpression seems to be specifically down-modulated at the trans-lational level, which presumably contributes to the prevention ofapoptosis in activated DCs (23).

GADD34 Is Up-Regulated in Activated DCs. GADD34 acts togetherwith PP1 to dephosphorylate eIF2α and relieves translation

repression during ER stress (22, 24, 25). In response to solublepI:C, GADD34 mRNA transcription was enhanced at least 14-fold (Fig. 1E), contrasting with the modest 1.5-fold up-regulationof PP1 expression (Fig. 1G) and 2-fold up-regulation of the“constitutive” protein phosphatase 1 regulatory subunit 15B(CReP) mRNA (Fig. 1H) (26). Matching the transcriptionalanalysis, and irrespectively of the mode of pI:C delivery (solubleor lipofected), GADD34 protein levels were induced and con-tinuously increased during activation (Fig. 1F). Addition ofproteasome inhibitor MG132 during the last 2 h of pI:C treat-ment facilitated the detection of short-lived GADD34 by pre-venting its degradation (Fig. 1F) (27) and confirmed GADD34expression during DC maturation.

eIF2α Is Dephosphorylated During DC Activation. Protein synthesiswas monitored using puromycin labeling followed by immunoblot(Fig. 2A) (28). As previously shown for LPS-activated DCs (29),translation was enhanced in the first hours of pI:C stimulationfollowed by a reduction at 16 h. The levels of phosphorylated eIF2α(P-eIF2α) were gradually lost during pI:C stimulation (Fig. 2B),independently of themode of dsRNAdelivery (Fig. S3B). Levels ofP-eIF2α were surprisingly high in nonstimulated DCs comparedwith mouse embryonic fibroblasts (MEFs), and no up-regulationof P-eIF2α could be visualized during maturation (Fig. 2 B and C).We further investigated whether these high levels of P-eIF2αwere physiologically relevant by analyzing purified mouse spleenCD11c+ DCs. Explanted CD11c+ DCs displayed even higher P-eIF2α levels than nonstimulated bmDCs. This observation wasfurther confirmed by immunohistochemistry of spleen sections, inwhich high P-eIF2α was mostly detected in resting CD11c+ DCsand not in neighboring CD3+ T cells in situ (Fig. S4). Spleen DCsand nonstimulated bmDCs, therefore, display naturally high levelsof P-eIF2α, which might explain their low level of translationin vitro and in vivo and might be important for exerting their sen-tinel function (29).

Fig. 1. ATF4 and GADD34 are induced during DC activation with pI:C.Experiments were performed in bmDCs stimulated with 10 μg/mL of pI:C or,when indicated, in RAW macrophages with 100 ng of LPS. mRNA levels ofATF4 (A), CHOP (C), GADD34 (E), PP1 (G), and CReP (H) were measured byqPCR. (B) ATF4 protein levels were detected by immunoblot in nuclearextracts. Histone deacetylase 1 (HDAC1) is shown as loading control. Proteinlevels of CHOP (D) and GADD34 (F) were detected by immunoblot in totalcell lysates. pI:C was added as soluble or lipofected (pI:C-lip). When in-dicated, MG132 was used to inhibit proteasomal degradation of GADD34and was added for 2 h before harvesting the cells. ER stress-inducing drugsthapsigargin (th) or tunicamycin (tun) serve as positive control treatmentsfor the expression of ATF4, CHOP, and GADD34. Data represented in A, C, E,G, and H are mean ± SD of three independent experiments.

0 8 20 0 8 20 0 8 0 8 dmso guana dmso sal

P-eIF2α

pI:C(h)

eIF2α

E WT GADD34ΔC/ΔC

eIF2α

PKR

β-actin(h)

P-eIF2α

Th0 8 20 20 4

lipo sol

pI:C

Th0 8 20 20 4

lipo sol

pI:C

D

splenic CD11c+iDC MEF

eIF2α

P-eIF2α

C

0 1 4 8 16 24

eIF2α

pI:C (h)

P-eIF2α

B

co chx 0 4 8 16 24pI:C (h)

pu

rom

ycin

β-actinA

Fig. 2. eIF2α dephosphorylation is controlled by GADD34 during pI:C-in-duced DC maturation. (A) Protein synthesis was quantified in proteinextracts of pI:C-activated DCs using puromycin labeling followed by immu-noblot. Cells not treated with puromycin (co) or cycloheximide (chx) wereused as controls. (B) Cell extracts were blotted for P-eIF2α and total eIF2α;P-eIF2α decreases with time after pI:C stimulation. (C) Steady-state levels ofP-eIF2α nonstimulated bmDC (iDC), MEF, and CD11c+ splenocytes. (D and E)Immature and pI:C-stimulated DCs were treated with GADD34 inhibitorsguanabenz and salubrinal. Both drugs were added at the same time as pI:Cand for a period of 8 h to nonstimulated cells. Protein extracts were blottedfor P-eIF2α and eIF2α. (E) P-eIF2α levels of WT and GADD34ΔC/ΔC DCs stim-ulated with lipofected (lipo) or soluble pI:C (sol). Nonstimulated cells weretreated with lipofectamine only. PKR blot is shown as control of DC activa-tion. Thapsigargin (th) was used as positive control for P-eIF2α.

Clavarino et al. PNAS | February 21, 2012 | vol. 109 | no. 8 | 3007

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

Because P-eIF2α intensity was inversely correlated withGADD34 expression in activated cells, we tested the function ofPP1-GADD34 and -CReP complexes by inhibiting their activitywith the pharmacological inhibitors salubrinal and guanabenz(30, 31). In the presence of both drugs, P-eIF2α levels wereenhanced by pI:C treatment (Fig. 2D). The same result wasfound when using DCs inactivated for GADD34 (GADD34ΔC/ΔC)(25), in which P-eIF2α was considerably augmented upon acti-vation compared with WT cells (Fig. 2E and quantification inFig. S3C). Thus, GADD34 and PP1 dephosphorylate eIF2α andcounteract efficiently the activation of eIF2α kinases in pI:C-stimulated DCs.

PKR Is Up-Regulated and Phosphorylates eIF2α in pI:C-Activated DCs.PKR acts as a signal transducer in the proinflammatory responseto different PAMPs throughTLR signaling and direct activation bycytosolic dsRNA. PKR activation results in eIF2α phosphoryla-tion and protein synthesis arrest (7, 32). As expected, PKR, beingtype I IFN-inducible, was strongly up-regulated upon pI:C stimu-lation (Figs. 2E and 3 A and B). In nonactivated PKR−/− DCs,P-eIF2α levels were close to normal, indicating that eIF2α kinasesother than PKR are responsible for the high degree of eIF2αphosphorylation in nonstimulated cells (Fig. 3A). Importantly,eIF2α phosphorylation was nearly abolished in activated PKR−/−

DCs, indicating that PKR is mostly responsible for this activityafter pI:C stimulation (Fig. 3A). Cytosolic dsRNA delivery bylipofection also decreased P-eIF2α levels in maturing cells, beingmore evident in PKR−/− cells (Fig. 3B). Independently of themodeof pI:C delivery, PKR activity is therefore efficiently counteractedby GADD34 induction. This observation contrasts with previouswork on MEFs exposed to cytosolic pI:C or Sindbis virus (33).Moreover, although PKRwas shown to promote apoptosis in LPS-stimulatedmacrophages (34), we found that procaspase 3 cleavagewas reduced in response to pI:C, again coinciding with eIF2α de-phosphorylation and contrasting with the UPR-inducing drug,thapsigargin, which enhanced both caspase-3 cleavage and eIF2αphosphorylation in DCs (Fig. S3B).

GADD34 expressionwas not affected by PKRdeletion (Fig. 3C),correlating with the absence of abundant P-eIF2α in PKR−/−-acti-vated cells (Fig. 3A). Recently, PKR and its NOX2-dependent ac-tivation have been implicated in amplifying part of the UPR andlinking it to inflammation (19, 35, 36). To gain further insights intothe signaling pathways controlling GADD34 expression during DCactivation by pI:C, we monitored GADD34 expression in DCsderived from different knockout mice. Conversely to what wasobserved in NOX2

−/−macrophages (35), NOX2 deletion did not

interfere with GADD34 induction upon DC activation (Fig. 3D).However, we found that toll/interleukin 1 receptor domain-con-taining adapter inducing IFN-β (TRIF) was absolutely required toinduce GADD34 expression in response to soluble and lipofectedpI:C. GADD34 remained undetectable in TRIF−/− DCs (Fig. 3E),even using longer kinetics of activation (Fig. S5A), whereas its ex-pression was found normal in activated MDA5−/− cells (Fig. 3F).Thus, the cytosolic helicase MDA5 is dispensable for GADD34-inducible expression by dsRNA even upon lipofection delivery.When we investigated ATF4 expression, we also found that itssynthesis was attenuated upon TRIF inactivation, albeit not com-pletely abolished (Fig. S2). This result further suggests that at leasttwo signaling pathways are working in parallel upon pI:C detectionand that GADD34 and ATF4 expression are mostly TRIF de-pendent in pI:C-activated DCs.

Cytosolic Delivery of pI:C Does Not Inhibit Protein Synthesis in DCs.As cytosolic accumulation of dsRNA normally induces a PKR-dependent translational arrest (32), protein synthesis was mon-itored in DCs and fibroblasts exposed to soluble or lipofectedpI:C. pI:C lipofection of MEFs efficiently caused a PKR-depen-dent translation arrest within 8 h (Fig. 4A). The translation arrestobserved in WT MEFs suggests that soluble dsRNA can effi-ciently access the cytosol of these cells and interact with PKR.In the case of DCs, and as anticipated from the low levels ofP-eIF2α induced by pI:C lipofection (Fig. 3B and Fig. S3B),translation was not inhibited even after 8 h of exposure toequivalent levels of cytosolic pI:C (Fig. 4A and Fig. S5B). We

A B

C D

E

F

Fig. 3. PKR phosphorylates eIF2α in pI:C-activated DCs. WT and PKR−/− DCs were stimulated with pI:C alone (sol) or in combination with lipofectamine (lip) forthe indicated time points (A and C) or for 8 h (B). Protein extracts were blotted for PKR, P-eIF2α (A and B) and GADD34 (C). Sodium arsenite (as) was used asa positive control for P-eIF2α induction. (D) GADD34 immunoblot was performed on lysate of NOX2−/− bmDCs activated with either pI:C or LPS for 8 and 24 h.(E) WT and TRIF−/− bmDCs were stimulated with pI:C as indicated and tested for the presence of GADD34 in protein extracts. (F) WT and MDA5−/− bmDCs werestimulated with lipofected pI:C and immunoblotted for GADD34. Treatment with MG132 (2 h) was used to facilitate GADD34 visualization.

3008 | www.pnas.org/cgi/doi/10.1073/pnas.1104491109 Clavarino et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

also observed by confocal microscopy the efficient access ofsoluble pI:C in the DC cytosol (Fig. S5C), likely to induce con-comitantly the different dsRNA-sensing pathways (e.g., TLR3and MDA5), as suggested by the levels of IFN-β productionobserved in TRIF-inactivated cells upon soluble pI:C stimulationin vitro (Fig. S5D) (4) and in vivo (37). On the basis of theseobservations and for greater consistency in our experiments, wedecided to use only lipofected pI:C, which is more likely to berelevant physiologically (38).WhenGADD34ΔC/ΔCDCs were exposed to cytosolic pI:C, their

translation levels remained similar to those of WT cells (Fig. 4B).GADD34ΔC/ΔCDCswere clearly activated as shown by the patternof S6 ribosomal protein phosphorylation (P-S6) normally associ-ated with TRIF-dependent mTOR activation and protein syn-thesis enhancement (29). Activated DCs are therefore able toresist PKR-dependent translational arrest and rely on the in-duction of GADD34 to shift the biochemical equilibrium towardeIF2α dephosphorylation. GADD34, however, is not important toestablish ormaintain DC resistance to dsRNA-induced translationarrest (Fig. 4B). In contrast, GADD34 was absolutely required foreIF2α dephosphorylation and translation recovery upon thapsi-gargin treatment (Fig. 4C). Thus, GADD34 seems to play a spe-cific role in the innate response toward dsRNA independently ofits previously characterized function in the UPR.

Cytokine Production Is Affected by GADD34 Inactivation. GADD34impact on translational initiation being limited, we examined thephenotype of activated GADD34ΔC/ΔC DCs. Lipofected pI:C-driven maturation of GADD34ΔC/ΔC CD11c+ cells was foundnormal, judging by the increase of surface MHC II and CD86levels (Fig. S6). In contrast with these results, a reduction in IFN-β and IL-6 levels was detected in the cell culture supernatantsof GADD34ΔC/ΔC DCs (Fig. 5A). To define whether GADD34deficiency impacted cytokine production at the transcriptional ortranslational level, we measured the mRNA expression of IFN-βand IL-6 transcripts and observed that, in the absence of func-tional GADD34, the levels of these transcripts were reduced by

half after 8 h of pI:C stimulation (Fig. 5B). GADD34 activity,therefore, enhances the transcription of different cytokinesdownstream of the TRIF adapter.

GADD34 Regulates Levels of IFN-β in Vivo. To evaluate the con-sequences of GADD34 deletion at the whole animal level, weinjected i.v. the Friend leukemia virus B sensitive mouse strain(FVB) mice with pI:C complexed with the liposomal transfectionreagent DOTAP. We could show that under these conditions,GADD34 is induced rapidly in CD11c+ splenocytes (Fig. S7). Wenext analyzed IFN-β blood levels at 3 and 6 h postinjection. In FVBWT animals, production of IFN-β peaked at 3 h postinjection,whereas after 6 h the levels returned to basal levels (Fig. 5C). Aspredicted from the in vitro data and compared with WT mice,a marked reduction of IFN-β was observed in GADD34ΔC/ΔC mice3 h postinjection (Fig. 5C). We tested next the relevance ofGADD34 during viral infection by monitoring type-I IFN pro-duction in Chikungunya virus (CHIKV)-infected 12-d-old mice.We turned to this small RNA enveloped virus, because it is knownto be a strong inducer of type-I IFN in vivo (39), a response key formouse neonates to control the infection (40). At 72 h post-inoculation, GADD34ΔC/ΔC pups displayed significantly less IFN-βin the serum and in the joints than their WT littermates, whereas asexpected, virus titers were increased in GADD34ΔC/ΔC organs(Fig. 5D). GADD34 is therefore an immunologically relevantdsRNA-inducible factor that participates in the optimization ofcytokine production in response to pI:C and viral infection.

DiscussionDCs treated with pI:C display an ATF4 expression signaturesharing common features with an ISR, including GADD34 in-duction. Interestingly, GADD34 induction has also been singledout in a transcriptome analysis of Listeria monocytogenes-infectedmacrophages (41) and during corona virus infection of fibroblasts(42), suggesting its association with pathogen detection. In vitro,the penetration of dsRNA in all cellular compartments can bedirectly detected through TLR3, DExD/H-box helicases, and PKR

+ puromycin

pu

rom

ycin

thapsigargin (h)chx 0 0.5 1 3 6 0 0.5 1 3 6

β-actin

eIF2α

P-eIF2α

GADD34ΔC/ΔCWTC

+ puromycin

P-S6S6

chx 0 4 8 16 24 chx 0 4 8 16 24pI:C pI:C

β-actin

GADD34ΔC/ΔCWT

pu

rom

ycin

B

pu

rom

ycin

pI:C co chx mock lip sol

pI:C co chx mock lip sol

PKR-/-WTMEFs

β-actin

+ puromycinpI:C

co chx mock lip solpI:C

co chxmock lip sol

DCsWT PKR-/-

A

Fig. 4. DCs are protected from translation inhibition induced by cytosolic pI:C detection. (A) Protein synthesis was monitored by immune detection ofpuromycin incorporation in WT and PKR−/− MEFs and DCs treated with soluble pI:C (sol) or lipofected (lip) for 8 h. Controls: no puromycin (co), chx, andlipofectamine-treated (mock) cells. (B) WT and GADD34ΔC/ΔC DCs were lipofected with pI:C for the indicated time and translation measured by antipuromycinimmunoblot. (C) bmDCs from WT and GADD34ΔC/ΔC mice were challenged with thapsigargin for the indicated time. Protein translation, P-eIF2α, and totaleIF2α were detected by immunoblotting.

Clavarino et al. PNAS | February 21, 2012 | vol. 109 | no. 8 | 3009

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

(43), which can also be rapidly activated by TLR ligation to pro-mote p38 and NF-κB signaling (32). This eIF2α kinase is necessaryto achieve functional DC maturation (9); however, its activationshould normally lead to translation inhibition through eIF2αphosphorylation. Surprisingly, and in contrast to fibroblasts, DCsare not affected by dsRNA-induced translation inhibition. Thisspecificity could allow DCs to prioritize the signal transductionpathways governing their innate immunity function over thepathways normally protecting cellular integrity from viral in-fection.When fully activated, these pathways would lead to proteintranslational arrest and/or apoptosis, through CHOP productionand caspase cleavage, thus impairing DC function at a crucial timeof the immune response (29). The lack of translation inhibitiontogether with the absence of CHOP production, could allow DCsto carry their function even when infected with viruses as alreadyevoked for human monocyte-derived DCs submitted to influenzavirus infection (44).GADD34 expression has primarily been shown to operate as

a negative feedback loop during the UPR and allows for trans-lation recovery through eIF2α dephosphorylation. We haveshown that in DCs, GADD34 is required to prevent thapsigargin-induced protein synthesis arrest but GADD34 inactivation doesnot impact significantly on protein synthesis in response todsRNA delivery. Activated DCs are therefore atypical cells inwhich eIF2α phosphorylation levels do not fully correlate withprotein synthesis intensity and confirms that specific translationregulation pathways operate in during microbe detection, as ob-served during the proteasome-dependent cleavage of eIF4GI andDAP5 translation factors at late stages of their activation (29).DCs have been reported to express high levels of XBP1, a

transcription factor essential for ER homeostasis during UPR (15,45, 46) and necessary for normal DC development (17, 47). Inmacrophages, TLR4 andTLR2 stimulation activates theER-stresssensor kinase IRE1α and its downstream target, XBP1, withoutinducing other UPR branches (19). Our observations suggest that

in DCs, the ATF4-dependent pathway is activated upon dsRNAdetection. Interestingly nonactivatedDCs possess abnormally highlevels of P-eIF2α, which could, like XBP1 transcription, reflecta specific activation of the UPR during their differentiation andexplain their low mRNA translation activity (29). GADD34 in-duction by pI:C in DCs requires the adapter TRIF and, contrary toXBP1, is not linked to NOX2 activity. The fact that pI:C entersefficiently the cell cytosol and the recent discovery of the TRIF-dependent DExD/H-box helicases DDX1, DDX21, and DHX36(4), suggests that according to the mode of dsRNA delivery, TRIFcould integrate different signals initiated by TLR3 and/or thesehelicases. The total absence of GADD34 induction by cytosolic pI:C in TRIF−/−DCs suggests that this pathway, leading to GADD34expression, works in parallel with the induction of the MDA-5sensing pathway to produce cytokines such as IFN-β efficiently.Our observations also imply the existence of alternative sig-

naling pathways, which allow ATF4 and GADD34 translation inthe absence of increased P-eIF2α levels (48). Potentially, thehigh levels of P-eIF2α present in iDCs are sufficient to promotetheir synthesis during the early stages of DC activation. CHOPmRNA transcription was also found to be up-regulated in DCs,but no translation product could be detected, suggesting thatCHOP synthesis is also tightly regulated in this cell type, as an-ticipated from the potent protective effect mediated by TLRagonists injection in mice subjected to systemic ER stress (49).dsRNA is probably an extreme example of microbial-associated

stressor because it can also induce PKR through direct recogni-tion in the cytosol. Interestingly, although GADD34 controlseIF2α dephosphorylation, its inactivation does not overly impacttranslational control. However, like for XBP1 deficiency, we couldshow that IFN-β and IL-6 transcriptions were decreased inGADD34ΔC/ΔC DCs in vitro and in deficient mouse in vivo. Inabsence of an obvious effect of GADD34 deletion on proteinsynthesis, this observation suggests that GADD34/PP1 is requiredto dephosphorylate/activate a key factor in the signal transduction

0

1

2

3

4

IFN

-β (

ng

/ml o

r n

g/g

)

D

0

20

40

60

80

0 3 6pI:C (h)

IFN

-β (

pg

/ml)

IFN

-β (

pg

/ml)

GADD34ΔC/ΔCWT GADD34ΔC/ΔC

WTC

0

20

40

60

80

100

0 4 8 pI:C (h)

IL-6

mR

NA

fo

ld in

crea

se

GADD34ΔC/ΔCWT

0

50

100

150

200

250

0 4 8 pI:C (h)IFN

-β m

RN

A f

old

incr

ease

GADD34ΔC/ΔCWT

B0

50100150

200250

300

0 4 8 pI:C (h)

IL-6

(p

g/m

l) GADD34ΔC/ΔCWT

0

10

20

30

40

50

0 4 8 pI:C (h)

GADD34ΔC/ΔCWTA

Muscle

Joint

WT GADD34ΔC/ΔC

Virus titers (/g of tissue):

7,3x104

5,2x105

5,0x105

2,5x106

p=0,11

p=0,01

serum muscle joint

p=0,04

Fig. 5. GADD34 promotes normal secretion of IFN-β and IL-6 in response to pI:C. (A) Culture supernatants of WT and GADD34ΔC/ΔC bmDCs in which pI:C wasdelivered in the cytoplasm by lipofectionwere analyzed by ELISA for IFN-β and IL-6. (B) IFN-β and IL-6mRNA expression levelsweremeasured by qPCR. One of fourindependent experiments with similar results is shown for each panel. (C and D) GADD34 is relevant for normal IFN-β protein levels in vivo. IFN-β levels weremeasured in the serum of mice that were i.v. injected with 20 μg/mL of pI:C for 3 or 6 h. pI:C was delivered conjugated with DOTAP; controls were performedinjecting PBSwithDOTAP. (D) IFN-β dosage in tissues and serumofmice inoculatedwith 106 pfu of CHIKV via the ID route. Fifteen-day-oldmicewere analyzed 72hafter inoculation. Each data point represents the arithmetic mean ±SD for at least four mice. Average virus titers in the different organs are indicated.

3010 | www.pnas.org/cgi/doi/10.1073/pnas.1104491109 Clavarino et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1

pathway downstream of TRIF. Recently, PP1-mediated de-activation of IkB kinase (IKK) in response to TNF-α, has beenlinked to GADD34 recruitment by the CUE domain-containing 2protein (CUEDC2) (50). CUEDC2 silencing in macrophages ledto increased transcription of inflammatory cytokine IL-6, by de-creasing putative GADD34-dependent dephosphorylation of IKK.We could not recapitulate this finding inDCs, because IL-6 mRNAwas less induced in activated GADD34ΔC/ΔC cells. A search foradditional GADD34/PP1 targets will allow better understandingof the regulation pathways controlled by this specific pathogen-induced stress response. Activation of ATF4 and GADD34should therefore be considered as an integral part of the innateimmunity signaling cascades downstream of dsRNA sensors.

Experimental ProceduresMouse bmDCs were treated with pI:C for different time periods. An Affy-metrix Mouse Genome MOE 430 2.0 gene microarray chip was used to

identify genes involved in dsRNA response of pI:C-activated DCs. Potentialtargets were confirmed by qPCR. Protein extracts of DCs fromWT or differentgene-targeted mice were analyzed by immunoblot for the presence ofGADD34, P-eIF2α, or to study protein translation through puromycin in-corporation (28). In vivo injections of pI:C were performed as shown before(3). Cytokine levels were measured by ELISA. Detailed experimental proce-dures can be found in SI Experimental Procedures.

ACKNOWLEDGMENTS. We thank L. Alexopoulou for the TRIF−/− mice;M. Colonna for MDA5−/− mouse bone marrow; C. Reis e Sousa for PKR−/−

mouse bone marrow, PKR−/− MEFs, and for sharing unpublished results;L. Wrabetz for the GADD34ΔC/ΔC mice; and S. Meresse for the NOX2−/− mice.We also thank the Plateforme d’Imagerie Commune du Site de Luminy(PICsL) Ministère de l’Education Nationale et de la Recherche (MENRT) Fun-dação para a Ciência e Tecnologia (FCT) for expert technical assistance. Thiswork is supported by grants (to P.P.) from La Ligue Nationale Contre le Can-cer, the Agence Nationale pour la Recherche 07-MIME-005 “DC-TRANS,” andthe Agence Nationale de Recherches sur le SIDA. G.C. is supported by fellow-ships from la Fondation pour la Recherche Médicale. N.C. is supported byFundação para a Ciência e Tecnologia Grant SFRH/BD/40112/2007 (Portugal).

1. Banchereau J, Steinman RM (1998) Dendritic cells and the control of immunity. Nature392:245–252.

2. Alexopoulou L, Holt AC, Medzhitov R, Flavell RA (2001) Recognition of double-strandedRNA and activation of NF-kappaB by Toll-like receptor 3. Nature 413:732–738.

3. Gitlin L, et al. (2006) Essential role of mda-5 in type I IFN responses to polyriboinosinic:polyribocytidylic acid and encephalomyocarditis picornavirus. Proc Natl Acad Sci USA103:8459–8464.

4. Zhang Z, et al. (2011) DDX1, DDX21, and DHX36 helicases form a complex with theadaptor molecule TRIF to sense dsRNA in dendritic cells. Immunity 34:866–878.

5. Kawai T, Akira S (2006) Innate immune recognition of viral infection. Nat Immunol 7:131–137.

6. Donzé O, et al. (2004) The protein kinase PKR: A molecular clock that sequentiallyactivates survival and death programs. EMBO J 23:564–571.

7. Scheuner D, et al. (2006) Double-stranded RNA-dependent protein kinase phos-phorylation of the alpha-subunit of eukaryotic translation initiation factor 2 mediatesapoptosis. J Biol Chem 281:21458–21468.

8. Williams BR (1999) PKR; a sentinel kinase for cellular stress. Oncogene 18:6112–6120.9. Diebold SS, et al. (2003) Viral infection switches non-plasmacytoid dendritic cells into

high interferon producers. Nature 424:324–328.10. Schulz O, et al. (2010) Protein kinase R contributes to immunity against specific viruses

by regulating interferon mRNA integrity. Cell Host Microbe 7:354–361.11. Harding HP, et al. (2003) An integrated stress response regulates amino acid me-

tabolism and resistance to oxidative stress. Mol Cell 11:619–633.12. Harding HP, Zhang Y, Bertolotti A, Zeng H, Ron D (2000) Perk is essential for translational

regulation and cell survival during the unfolded protein response. Mol Cell 5:897–904.13. Berlanga JJ, et al. (2006) Antiviral effect of the mammalian translation initiation

factor 2alpha kinase GCN2 against RNA viruses. EMBO J 25:1730–1740.14. Lu PD, Harding HP, Ron D (2004) Translation reinitiation at alternative open reading

frames regulates gene expression in an integrated stress response. J Cell Biol 167:27–33.15. Ron D, Walter P (2007) Signal integration in the endoplasmic reticulum unfolded

protein response. Nat Rev Mol Cell Biol 8:519–529.16. Zhang K, Kaufman RJ (2008) From endoplasmic-reticulum stress to the inflammatory

response. Nature 454:455–462.17. Iwakoshi NN, Pypaert M, Glimcher LH (2007) The transcription factor XBP-1 is essential

for the development and survival of dendritic cells. J Exp Med 204:2267–2275.18. Todd DJ, Lee AH, Glimcher LH (2008) The endoplasmic reticulum stress response in

immunity and autoimmunity. Nat Rev Immunol 8:663–674.19. Martinon F, Chen X, Lee AH, Glimcher LH (2010) TLR activation of the transcription

factor XBP1 regulates innate immune responses in macrophages. Nat Immunol 11:411–418.

20. Connor JH, Weiser DC, Li S, Hallenbeck JM, Shenolikar S (2001) Growth arrest andDNA damage-inducible protein GADD34 assembles a novel signaling complex con-taining protein phosphatase 1 and inhibitor 1. Mol Cell Biol 21:6841–6850.

21. Okada T, Yoshida H, Akazawa R, Negishi M, Mori K (2002) Distinct roles of activatingtranscription factor 6 (ATF6) and double-stranded RNA-activated protein kinase-likeendoplasmic reticulum kinase (PERK) in transcription during the mammalian un-folded protein response. Biochem J 366:585–594.

22. Marciniak SJ, et al. (2004) CHOP induces death by promoting protein synthesis andoxidation in the stressed endoplasmic reticulum. Genes Dev 18:3066–3077.

23. Nakayama Y, et al. (2010) Molecular mechanisms of the LPS-induced non-apoptoticER stress-CHOP pathway. J Biochem 147:471–483.

24. Novoa I, Zeng H, Harding HP, Ron D (2001) Feedback inhibition of the unfoldedprotein response by GADD34-mediated dephosphorylation of eIF2alpha. J Cell Biol153:1011–1022.

25. Novoa I, et al. (2003) Stress-induced gene expression requires programmed recoveryfrom translational repression. EMBO J 22:1180–1187.

26. Jousse C, et al. (2003) Inhibition of a constitutive translation initiation factor 2alphaphosphatase, CReP, promotes survival of stressed cells. J Cell Biol 163:767–775.

27. Brush MH, Shenolikar S (2008) Control of cellular GADD34 levels by the 26S protea-some. Mol Cell Biol 28:6989–7000.

28. Schmidt EK, Clavarino G, Ceppi M, Pierre P (2009) SUnSET, a nonradioactive methodto monitor protein synthesis. Nat Methods 6:275–277.

29. Lelouard H, et al. (2007) Regulation of translation is required for dendritic cellfunction and survival during activation. J Cell Biol 179:1427–1439.

30. Boyce M, et al. (2005) A selective inhibitor of eIF2alpha dephosphorylation protectscells from ER stress. Science 307:935–939.

31. Tsaytler P, Harding HP, Ron D, Bertolotti A (2011) Selective inhibition of a regulatorysubunit of protein phosphatase 1 restores proteostasis. Science 332:91–94.

32. Williams BR (2001) Signal integration via PKR. Sci STKE 2001:re2.33. Gorchakov R, Frolova E, Williams BR, Rice CM, Frolov I (2004) PKR-dependent and

-independent mechanisms are involved in translational shutoff during Sindbis virusinfection. J Virol 78:8455–8467.

34. Hsu LC, et al. (2004) The protein kinase PKR is required for macrophage apoptosisafter activation of Toll-like receptor 4. Nature 428:341–345.

35. Li G, Scull C, Ozcan L, Tabas I (2010) NADPH oxidase links endoplasmic reticulumstress, oxidative stress, and PKR activation to induce apoptosis. J Cell Biol 191:1113–1125.

36. Nakamura T, et al. (2010) Double-stranded RNA-dependent protein kinase linkspathogen sensing with stress and metabolic homeostasis. Cell 140:338–348.

37. Miyake T, et al. (2009) Poly I:C-induced activation of NK cells by CD8 alpha+ dendriticcells via the IPS-1 and TRIF-dependent pathways. J Immunol 183:2522–2528.

38. Frleta D, et al. (2009) Influenza virus and poly(I:C) inhibit MHC class I-restricted pre-sentation of cell-associated antigens derived from infected dead cells captured byhuman dendritic cells. J Immunol 182:2766–2776.

39. Schilte C, et al. (2010) Type I IFN controls chikungunya virus via its action on non-hematopoietic cells. J Exp Med 207:429–442.

40. Couderc T, et al. (2008) A mouse model for Chikungunya: Young age and inefficienttype-I interferon signaling are risk factors for severe disease. PLoS Pathog 4:e29.

41. Leber JH, et al. (2008) Distinct TLR- and NLR-mediated transcriptional responses to anintracellular pathogen. PLoS Pathog 4:e6.

42. Wang X, et al. (2009) Inhibition of protein kinase R activation and upregulation ofGADD34 expression play a synergistic role in facilitating coronavirus replication bymaintaining de novo protein synthesis in virus-infected cells. J Virol 83:12462–12472.

43. Jelinek I, et al. (2011) TLR3-specific double-stranded RNA oligonucleotide adjuvantsinduce dendritic cell cross-presentation, CTL responses, and antiviral protection. JImmunol 186:2422–2429.

44. Cella M, et al. (1999) Maturation, activation, and protection of dendritic cells inducedby double-stranded RNA. J Exp Med 189:821–829.

45. Yoshida H, Matsui T, Yamamoto A, Okada T, Mori K (2001) XBP1 mRNA is induced byATF6 and spliced by IRE1 in response to ER stress to produce a highly active tran-scription factor. Cell 107:881–891.

46. Calfon M, et al. (2002) IRE1 couples endoplasmic reticulum load to secretory capacityby processing the XBP-1 mRNA. Nature 415:92–96.

47. Smith JA, et al. (2008) Endoplasmic reticulum stress and the unfolded protein re-sponse are linked to synergistic IFN-beta induction via X-box binding protein 1. Eur JImmunol 38:1194–1203.

48. Harding HP, et al. (2000) Regulated translation initiation controls stress-induced geneexpression in mammalian cells. Mol Cell 6:1099–1108.

49. Woo CW, et al. (2009) Adaptive suppression of the ATF4-CHOP branch of the un-folded protein response by toll-like receptor signalling. Nat Cell Biol 11:1473–1480.

50. Li HY, et al. (2008) Deactivation of the kinase IKK by CUEDC2 through recruitment ofthe phosphatase PP1. Nat Immunol 9:533–541.

Clavarino et al. PNAS | February 21, 2012 | vol. 109 | no. 8 | 3011

IMMUNOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 6,

202

1