The amino acid sensor GCN2 inhibits inflammatoryresponses to apoptotic cells promoting toleranceand suppressing systemic autoimmunityBuvana Ravishankara,1,2, Haiyun Liua,1,3, Rahul Shindea, Kapil Chaudharya, Wei Xiaoa, Jillian Bradleya,Marianne Koritzinskyb,c,d, Michael P. Madaioe, and Tracy L. McGahaa,e,4

aCancer Immunology, Inflammation, and Tolerance Program, Georgia Regents University Cancer Center, Augusta, GA 30912; bPrincess Margaret CancerCenter, University Health Network, Toronto, ON, Canada M5G 2M9; cDepartment of Radiation Oncology, Faculty of Medicine, University ofToronto, Toronto, ON, Canada M5T 1P5; dInstitute of Medical Sciences, Faculty of Medicine, University of Toronto, Toronto, ON, Canada M5S 1A8;and eSection of Nephrology, Department of Medicine, Medical College of Georgia, Georgia Regents University, Augusta, GA 30912

Edited by Ira Tabas, Columbia University, New York, NY, and accepted by the Editorial Board July 14, 2015 (received for review March 2, 2015)

Efficient apoptotic cell clearance and induction of immunologictolerance is a critical mechanism preventing autoimmunity andassociated pathology. Our laboratory has reported that apoptoticcells induce tolerance by a mechanism dependent on the trypto-phan catabolizing enzyme indoleamine 2,3 dioxygenase 1 (IDO1)in splenic macrophages (MΦ). The metabolic-stress sensing proteinkinase GCN2 is a primary downstream effector of IDO1; thus, wetested its role in apoptotic cell-driven immune suppression. In vitro,expression of IDO1 in MΦs significantly enhanced apoptotic cell-driven IL-10 and suppressed IL-12 production in a GCN2-dependentmechanism. Suppression of IL-12 protein production was dueto attenuation of IL-12 mRNA association with polyribosomesinhibiting translation while IL-10 mRNA association with polyrib-osomes was not affected. In vivo, apoptotic cell challenge drove arapid, GCN2-dependent stress response in splenic MΦs with in-creased IL-10 and TGF-β production, whereas myeloid-specific de-letion of GCN2 abrogated regulatory cytokine production withprovocation of inflammatory T-cell responses to apoptotic cell anti-gens and failure of long-tolerance induction. Consistent with a rolein prevention of apoptotic cell driven autoreactivity, myeloid de-letion of GCN2 in lupus-prone mice resulted in increased immunecell activation, humoral autoimmunity, renal pathology, andmortality.In contrast, activation of GCN2 with an agonist significantly reducedanti-DNA autoantibodies and protected mice from disease. Thus, thisstudy implicates a key role for GCN2 signals in regulating the tolero-genic response to apoptotic cells and limiting autoimmunity.

immunometabolism | stress | tolerance | autoimmunity | apoptosis

Multiple lines of evidence suggest that apoptotic cells are amajor source of autoantigens in the autoimmune disease

systemic lupus erythematosus (SLE). This evidence includes thefact that the dominant autoantigens targeted in SLE are largelynuclear components that are exposed on apoptotic blebs (1, 2).Moreover, the body of evidence suggests that increased celldeath and defective clearance are primary drivers of SLE de-velopment and progression (3).In homeostatic conditions, apoptotic cells drive suppressive

innate and adaptive tolerogenic immune responses that protectagainst initiation of inflammatory autoimmunity. Macrophages(MΦ) are key drivers of the early innate response to apoptoticcells (4, 5), and recent work has shown that tissue-resident MΦsare responsible for initial IL-10 production, promoting regula-tory adaptive immunity to apoptotic cell antigens and preventionof inflammatory autoimmunity (4, 6, 7).On a molecular level, the mechanisms driving the initial tol-

erogenic response of myeloid cells to apoptotic material in vivoare poorly defined, although studies suggest a direct role forinhibitory signaling via the MAPK pathway (8), and a recent re-port indicated that responses to cytoplasmic DNA may be a keymechanism of tolerance to self (9). Likewise, we have reportedthat apoptotic cells drive rapid expression of the tryptophan

(Trp)-metabolizing enzyme indoleamine 2,3 dioxygenase 1 (IDO1)in MARCO+ marginal zone (MZ) MΦs that is critical for pre-vention of apoptotic cell-driven autoimmunity (6). Moreover,studies by our group (6) and Sharma et al. (9) have shown alink between a loss of IDO1 activity and exaggeration of auto-immune pathology in MRLlpr/lpr mice.The best-described IDO1 effector mechanism is driven by

consumption of intracellular and microenvironmental Trp, withsubsequent elicitation of the so-called integrated stress response(10, 11). In the cell, amino acid deficiencies are detected by theser/thr kinase general control nonderepressible 2 (GCN2). Ac-tivated GCN2 phosphorylates the α subunit of eukaryotic initi-ation factor 2 (eIF2α), modulating ribosome assembly and alteringprotein translation (12).We have previously reported that GCN2 modulates cytokine

production in LPS-stimulated MΦs by mechanisms dependenton transcriptional and translational modification (11). At thetranscriptional level, GCN2 activation drives induction of tran-scription factors that are responsible for manifestation of the stressresponse, including a nodal driver of stress-induced transcription,activating transcription factor 4 (ATF4) (13). Moreover, throughinduction of ATF4 and downstream targets, including C/EBP

Significance

Metabolic stress potently modifies immunity. Recently our lab-oratory identified the stress kinase GCN2 as a key modulator ofmacrophage responses to toll-like receptor ligands; however, therole of myeloid GCN2 signals in sterile inflammation and ho-meostatic tolerance is not known. In this study, we tested therequirement of GCN2 for tolerance to apoptotic cells and pre-vention of autoimmunity in a model of lupus. Our results showthat GCN2 in myeloid cells is a critical effector of apoptotic cell-driven tolerance required for regulatory cytokine production andprevention of inflammatory immunity. Moreover, the data in-dicate that targeting GCN2 is an effective approach to prevent-ing autoimmunity, providing a rationale for developing tools tomanipulate GCN2 function in inflammatory immune disease.

Author contributions: B.R., H.L., and T.L.M. designed research; B.R., H.L., R.S., K.C., W.X.,J.B., M.K., and T.L.M. performed research; M.K., M.P.M., and T.L.M. analyzed data; andT.L.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. I.T. is a guest editor invited by the Editorial Board.1B.R. and H.L. contributed equally to this work.2Present address: Department of Cancer Immunology, Genentech, San Francisco, CA94080.

3Present address: Department of Dermatology, Johns Hopkins University School of Med-icine, Baltimore, MD 21218.

4To whom correspondence should be addressed. Email: tmcgaha@gru.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1504276112/-/DCSupplemental.

10774–10779 | PNAS | August 25, 2015 | vol. 112 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1504276112

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

homologous protein (CHOP), GCN2 controls autophagy (14)and is critical for cross-presentation of antigens in dendritic cells(DCs) (15).As a whole, the data indicates GCN2 signaling is an important

mechanism regulating innate immunity; however the role of GCN2in sterile inflammatory conditions is not known. We previouslyshowed that apoptotic cell challenge induced rapid expression ofCHOP in MZ MΦs in an IDO1-dependent mechanism (6). Giventhat (i) IDO1 is dependent on GCN2 for manifestation of stresssignals and (ii) inhibition of IDO1 converts the immune responseto apoptotic cells from tolerogenic to inflammatory, we hypothe-sized that GCN2 would be required for apoptotic cell-driventolerogenic immunity.In this report, we present evidence that IDO1 modulates MΦ

cytokine production when exposed to apoptotic cells in a mech-anism dependent on GCN2. Moreover, when GCN2 is deleted,there is a profound defect in tolerance to apoptotic cells. Thus,our results highlight a previously unidentified mechanism of GCN2-dependent, metabolism-driven suppression of innate responses toapoptotic cells.

ResultsIDO1 Suppresses IL-12p40 Association with Polyribosomes by a MechanismDependent on GCN2. The biological relevance of metabolic stresssignals to immune homeostasis or tolerogenic responses to apoptoticcells is not known. To explore this, we transduced bone marrow-derived MΦs (BM MΦs) with an IDO1-GFP fusion lentiviral con-struct (11), generating MΦ populations with constitutive IDO1 ac-tivity. In IDO1+ MΦs, production of kynurenine, the rate-limitingstep in IDO1-mediated Trp consumption, was increased 13-foldcompared with controls (Fig. 1A). Moreover, Trp in IDO1+ cultureswas undetectable by HPLC, demonstrating that IDO1 completelyconsumed all available Trp (Fig. 1A). There was comparablekynurenine production and Trp consumption in GCN2KO MΦs,suggesting that GCN2 had no impact on IDO1 activity (Fig. 1A);however, unlike C57BL/6 (B6) MΦs, GCN2KOMΦs failed to showmRNA induction for stress-response genes (CHOP and GADD45b),suggesting that whereas IDO1 was active in the cultures, transmissionof a stress response by Trp depletion required GCN2 (Fig. 1B).To test the impact of IDO1 activity and GCN2 deletion on

MΦ responses to apoptotic cells, we cultured IDO1+ GCN2 WTand KO MΦs with apoptotic thymocytes at a 1:10 MΦ/apoptotic

cell ratio for 12 h and then measured IL-10 and IL-12p40 pro-teins by ELISA. Exposure to apoptotic cells drove a regulatorycytokine response with a 32-fold increase in IL-10 (Fig. 1C) anda threefold decrease in IL-12p40 relative to baseline in controlMΦs (Fig. 1C). In IDO1+ MΦs, baseline expression of IL-10 wasincreased fivefold compared with GFP+ controls (10 ± 11 pg/mLfor GFP+ vs. 52 ± 15 pg/mL for IDO1-GFP+) (Fig. 1C). More-over, apoptotic cell coculture with IDO1+ MΦs induced a morethan twofold increase in IL-10 and an almost complete loss of IL-12p40 protein compared with control MΦs (Fig. 1C). Addition ofthe IDO1 inhibitor 1-methyl-D-tryptophan (D1MT) reversedapoptotic cell-driven suppression of IL-12 and induction of IL-10in IDO1+ MΦs, demonstrating that enzymatic activity is requiredfor promotion of suppressive cytokine production by IDO1.GCN2KO IDO1+ MΦs exhibited an elevated basal IL-12p40protein level that was only partially reduced by coculture withapoptotic cells (Fig. 1C). Moreover, although apoptotic cellsinduced IL-10 in IDO1+GCN2KOMΦs, production was only 50%of that observed in control MΦs (Fig. 1C). These results suggestthat IDO1 enhances apoptotic cell-driven regulatory responses by aGCN2-dependent mechanism.As expected, expression of IDO1 increased phosphorylation of

eIF2α (p-eIF2α) and induction of ATF4 protein (Fig. 1D andFig. S1 A and B). Coculture of MΦs with apoptotic cells en-hanced this effect, suggesting that apoptotic cells potentiatestress response signal transduction in MΦs. In contrast, in theabsence of GCN2, IDO1 expression resulted in a prominentreduction in p-eIF2α detectable by Western blot analysis (Fig.1D and Fig. S1A). Furthermore, coculture with apoptotic cellsfailed to elicit either p-eIF2α or ATF4, indicating that in theabsence of GCN2, neither IDO1 nor apoptotic cells were suffi-cient to activate integrated stress response signals.IDO1 expression increased basal IL-12p40 mRNA abundance

threefold compared with controls in a GCN2-dependent manner.However, the addition of apoptotic cells increased IL-10 mRNA10-fold and reduced IL-12p40 mRNA by 10-fold (Fig. 1E).Nevertheless, IL-12p40 message expression was consistentlyhigher in the presence of IDO1, an effect that was dependent onGCN2 (Fig. 1E). GCN2 alters mRNA association with thepolyribosome (polysome) modulating translational activity (13);thus, we hypothesized that GCN2 may restrict IL-12p40 trans-lation, which would result in the observed reduction in protein.Therefore, we examined ATF4, IL-10, and IL-12p40 mRNA

Fig. 1. GCN2 activation promotes apoptotic cell-driven IL-10 production in MΦs. (A and B) BM MΦswere transduced with IDO1-GFP+ or control lentivi-rus as described in Materials and Methods. Then,24 h later, kynurenine and Trp in the culture super-natant mRNA for the indicated genes were measuredby sqPCR. (C) At 24 h after lentiviral transduction asdescribed in A, BM MΦs were incubated with apo-ptotic thymocytes at a 10:1 apoptotic cell:MΦ ratio.Then, 18 h later, cell supernatants were collected,and cytokines were measured by ELISA. (D) Westernblot analysis for the proteins indicated was done onwhole-cell lysates from BM MΦ cultures as describedin C. (E) mRNA from BMMΦ/apoptotic cell cultures asin C was collected at the indicated time points, andcytokine message expression was assessed by sqPCR.In some groups, 20 μM D1MT was added. (F) Sucrosegradient fractionation of cellular lysates was fol-lowed by sqPCR analysis for the indicated mRNAspecies from 8-h BM MΦ/apoptotic cultures treatedas in C. In A, B, C, and E, bars and line points rep-resent mean ± SD values for triplicate samples. InD, Western blot images are representative of threeseparate experiments. In F, pooled material fromtriplicate samples was analyzed. *P < 0.05; **P <0.01, Student’s t test. Experiments were repeatedthree times, with similar results.

Ravishankar et al. PNAS | August 25, 2015 | vol. 112 | no. 34 | 10775

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

association with ribosomes as indicators of translation. For this,we subjected lysates of apoptotic cell-stimulated MΦs to frac-tionation on sucrose gradient, and determined the distribution ofmRNA by semiquantitative PCR (sqPCR) as described previously(11). Overall, IDO1 expression reduced mRNA association withpolysomes in a GCN2-dependent manner in MΦs (Fig. S2A).In contrast, ATF4 exhibited a shift in mRNA association tohigh molecular weight polysomes (Fig. 1F). This agrees with thefindings of Harding et al. (13), who reported increased ATF4association with polysomes during amino acid starvation stress.Moreover ATF4 association with polysomes was dependent onGCN2, given that GCN2KO MΦs exhibited an identical mRNAassociation profile regardless of the presence or absence of IDO1activity (Fig. 1F). IDO1 had the opposite effect on IL-12p40mRNA, which showed a decrease in polysome association in aGCN2-driven fashion (Fig. 1F), whereas in contrast, IL-10 mRNAassociated with the high molecular weight polysomes in a mannerinsensitive to either IDO1 or GCN2 (Fig. 1F). This result sug-gested that metabolic stress induced by consumption of Trp didnot impact the translation of IL-10. Taken as a whole, the ex-periments illustrated in Fig. 1 indicate that IDO1-driven meta-bolic stress may generate intracellular and microenvironmentalconditions conducive to regulatory cytokine translation after ap-optotic cell exposure.

GCN2 Is Required for Apoptotic Cell-Driven Tolerogenic Immunity inVivo. Apoptotic cells induce prominent CHOP expression insplenic MZ MΦs and CD8α+ DCs that is abrogated by the IDO1antagonist D1MT (6). This suggests that IDO1 activity is re-quired for apoptotic cell-driven stress responses in MΦs. Simi-larly, we found that apoptotic cells induced rapid expression ofthe stress-response genes Gadd34 and Gadd45b in FACS-sortedMZ MΦs and CD8α+ DCs (Fig. S2B). In contrast, GCN2KOmice showed an attenuation of stress gene induction in a mannercomparable to IDO1-inhibited groups, indicating a potential linkbetween GCN2 and IDO1 in apoptotic cell-driven stress re-sponses in phagocytes (Fig. S2B).Apoptotic cell exposure induces TGF-β in CD8α+ DCs and

IL-10 in MZMΦs in vivo (6). To test the requirement of myeloidGCN2 for suppressive cytokine expression after apoptotic cellchallenge, we crossed B6.Eif2ak4flox/flox (i.e., GCN2flox/flox) mice

with B6.Lysmcre/+ mice to generate myeloid-specific GCN2KOmice(Fig. S3). Apoptotic cell challenge in vivo induced expression ofIL-10 mRNA predominately in MZ MΦs (FACS-sorted viaSignR1) and TGF-β1 mRNA in CD8α+ DCs; however, total ormyeloid-specific GCN2 disruption abrogated regulatory cytokineinduction, and the phagocytes instead showed induction of IL-12p40 mRNA (Fig. 2A). Accordingly, examination of splenic ly-sates after apoptotic cell injection revealed a significant increasein both IL-10 and TGF-β protein, an effect that was blocked bymyeloid GCN2 deletion (Fig. 2B), suggesting that GCN2 signalsare mechanistically required for apoptotic cell-driven innateregulatory responses.We previously reported that apoptotic cell-associated antigens

fail to induce adaptive T-cell responses (4, 6, 7). This effect isdependent on IDO1 expression and MZ MΦs, given that Ido1−/−mice or mice depleted of MZ MΦs showed vigorous T-cell re-sponses to apoptotic cells (6). Because we hypothesized thatIDO1 inhibits T-cell responses via activation of GCN2, we rea-soned that GCN2KO mice would show increased inflammatoryT-cell responses to apoptotic cell antigens. To test this, we adop-tively transferred carboxyfluorescein succinimidyl ester (CFSE)-labeled transgenic OTII T cells into GCN2flLysMcre mice, fol-lowed 24 h later by challenge with apoptotic OVA+ thymocytes,and assessed proliferation (CFSE dye dilution), activation (CD44induction), cytokine production, and response to restimulationin vitro.In B6 mice, exposure to OVA+ apoptotic cells did not activate

OTII T cells, as demonstrated by the lack of change in splenicOTII T-cell numbers, and no observable change in CFSE fluores-cence intensity or CD44 surface expression (Fig. 2C). In contrast,when total GCN2KO or GCN2flLysMcre mice were challengedwith OVA+ apoptotic cells, the OTII T cells underwent prolif-eration and activation, with the majority of T cells exhibiting aCFSElowCD44high phenotype (Fig. 2C). A sizeable portion of thenaïve OTII T-cell population expressed IL-10, as determined by FACSanalysis (Fig. 2D). IL-10+ OTII T-cell numbers were slightlyincreased at 5 d after apoptotic cell challenge; however, inGCN2flLysMcre mice, apoptotic cell challenge resulted in a lossof IL-10+ T cells and significant increases in IFN-γ+ and IL-17a+OTII T cells (Fig. 2D), suggesting that myeloid GCN2 signalssuppress inflammatory T-cell responses to apoptotic cell antigens.

Fig. 2. Myeloid GCN2 signals are required to sup-press adaptive immunity to apoptotic cells in vivo. (A)At 4 h after apoptotic cell challenge (107 in vivo),SignR1+ MZ MΦs and CD11c+CD8α+ DCs were sortedby FACS, and cytokine message expression for theindicated species was determined by sqPCR. (B) Miceof the indicated phenotype were challenged with 107

apoptotic cells; 18 h later, spleens were collected, andcytokine protein concentrations were determined onwhole spleen lysate by ELISA. (C) Mice received 2.5 ×106 MACS-purified CFSE-labeled Thy1.1+CD4+ OTII Tcells in vivo, followed by challenge with 107 OVA+

apoptotic cells. At 3 d later, spleens were collected,and CFSE and CD44 intensity in the CD4+Thy1.1+ Tcells was determined by flow cytometry. Plots aregated on Thy1.1+CD4+ cells. (D) Splenocytes frommice treated as described in C were restimulatedin vitro for 5 h with PMA/ionomycin as described in SIMaterials and Methods, after which intracellular FACSanalysis for the indicated cytokines in Thy1.1+OTII+

T cells was done. (E) FACS-purified Thy1.1+OTII+ T cellswere stimulated with peptide-pulsed CD11c+ DCs,and proliferation was determined by thymidine in-corporation. cpm, count per minute. (F) Male B6thymocytes (107) were adoptively transferred in vivointo mice of the indicated genotype, followed bymale skin engraftment 7 d later (as described in SIMaterials and Methods), and then monitored forgraft rejection for 50 d. n = 10 mice/group. In A, bars represent the value for pooled samples from four mice. In B and D, bars represent mean ± SD values for fivemice. In C and D, plots are representative of five mice per group. *P < 0.05, **P < 0.01, Student’s t test. Experiments were repeated four times, with similar results.

10776 | www.pnas.org/cgi/doi/10.1073/pnas.1504276112 Ravishankar et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

Although we have shown that apoptotic cells do not driveadaptive T-cell responses, whether the lack of responsiveness isrelated to immunologic ignorance or to anergy is unclear. To testwhether apoptotic cells induce active suppression of T cells, werestimulated MACS-enriched OTII T cells with OVA-peptidepulsed CD11c+ DCs in vitro and determined proliferation via3H-thymidine incorporation. Naïve OTII T cells proliferated ro-bustly on stimulation in vitro, whereas OTII T cells purified fromOVA+ apoptotic cell-challenged mice failed to proliferate (Fig.2E). In contrast, if the T cells had seen apoptotic antigens in theabsence of myeloid GCN2, suppression failed and there was atwofold increase in thymidine incorporation compared withnonexposed controls (i.e., NIL group) (Fig. 2E), indicating that(i) apoptotic cells drive suppression of OTII T cells and (ii)suppression requires GCN2 signals in myeloid cells.Systemic apoptotic cell exposure results in long-term tolerance

to associated antigens and allografts (7). Given that GCN2flLysMcre

mice showed abrogated suppression of inflammatory immunity toapoptotic cells, we reasoned that long-term tolerance would becompromised as well. To test this, we used a model of femalerecognition to male H-Y antigen, in which exposure to male apo-ptotic cells leads to long-term tolerance to male skin grafts (7, 16).In agreement with our previous studies, control groups of femalemice rejected male skin, with a mean graft survival time of 26 d;in contrast, female mice receiving a single injection of maleapoptotic cells before skin engraftment showed tolerance withno graft rejection (Fig. 2F). Tolerance was dependent on bothIDO1 and GCN2, as demonstrated by the fact that treatmentwith D1MT at the time of apoptotic cell challenge abrogatedallograft acceptance (Fig. S4), and that GCN2flLysMcre micereceiving male apoptotic cells exhibited graft rejection in a pat-tern comparable to that seen in control B6 mice (Fig. 2F). Thus,apoptotic cell-driven tolerance manifests through a mechanism thatdepends on IDO1 induction with subsequent GCN2-dependentmetabolic stress signals.

Myeloid GCN2 Signals Restrict Spontaneous Autoimmune DiseaseProgression. In diseases of chronic inflammation such as SLE,IDO1 activity is often elevated, acting as a regulatory mechanismto limit disease pathology (17–21). In accordance with this, we

recently identified MZ MΦs and IDO1-driven regulation as keyfactors limiting SLE manifestation and progression (4, 6). Be-cause the data suggest that GCN2 is the principle downstreammolecular effector of the IDO1 response to apoptotic cells inphagocytes, we predicted that GCN2 deletion would accelerateautoimmunity and pathology in lupus. To test this, we fixedmyeloid GCN2 deficiency on the B6.FcγR2b−/− background andanalyzed the mice for autoimmune disease development andprogression.Female B6.FcγR2b− /− mice (hereinafter referred as R2B)

develop fulminant pathology with high-titer αdsDNA IgG,chronic B-cell, MΦ, and DC activation, and 50% mortality at age9–12 mo due to severe glomerulonephritis (22). R2B mice alsodevelop significant splenomegaly by age 6 mo. Deletion of GCN2amplified this phenotype, with splenocyte numbers doubled inR2B GCN2flLysMcre mice compared with controls (Fig. 3A).There was constitutive p-eIF2α detectable by Western blotanalysis in B6 spleen lysate, which was not increased in R2B mice(Fig. 3B and Fig. S1C); however, when GCN2 was deleted, therewas a reduction of p-eIF2α, suggestive of reduced stress signaling(Fig. 3B and Fig. S1C). IDO1 was not detected in B6 spleen;however, IDO1 protein was induced 10-fold in splenic lysatesfrom lupus-prone R2B mice, an effect that was increased byGCN2 deletion (Fig. 3B and Fig. S1D). Given that IDO1 is in-duced by IFNs and inflammatory reactions (10), this suggeststhat increased inflammatory activity may be driving the observedincreases in IDO1.FACS analysis of splenocytes from 6-mo-old R2B GCN2flLysMcre

mice revealed an expansion in splenic CD11c+ DCs comparedwith R2B mice, with specific increases in plasmacytoid DCs andmyeloid (i.e., CD11b+) DCs (Fig. 3C). In contrast, CD8α+ DCswere decreased by more than threefold in R2B GCN2flLysMcre

mice compared with either R2B or B6 mice (Fig. 3C), suggesting adifferential effect of GCN2 deficiency on splenic DC populations.MZ B-cell expansion in R2B mice was not affected by GCN2disruption; however, CD24low follicular B cells were increasedtwofold compared with control R2B mice, indicative of increasedchronic B-cell activation (Fig. 2D).F4/80+ MΦs showed increased surface CD86 and MHCII

staining in R2B mice compared with B6 controls (Fig. 4E).Deletion of GCN2 in MΦs did not affect MHCII, but CD86+ cellpercentages were increased by 50% in the R2B GCN2flLysMcre

mice. The effect of GCN2 deletion was more exaggerated in MZMΦs, which exhibited a significant increase in CD86 and MHCIIexpression compared with R2B MZ MΦs (Fig. 3E).CD8α+ DCs were primarily MHCIIhigh and showed higher

CD86 expression compared with MΦs. In R2B mice, MHCIIstaining for CD8α+ DCs was relatively unchanged, but there wasan increase in overall CD86high cells (Fig. 3E). In contrast, themajority of CD8α+ DCs had significantly down-regulated MHCIIin R2B GCN2flLysMcre mice (Fig. 3E) suggestive of a reductionin antigen presentation as well as overall CD8α+ DC numbers inthe absence of GCN2.When MZ MΦs from R2B mice were FACS-sorted and ex-

amined for cytokine message, we found increased mRNA for theinflammatory cytokines IL-12p40, IL-6, and IFN-β1, as well as a25-fold increase in IL-10 message expression, compared withage-matched B6 MZ MΦs (Fig. 4A). Deletion of GCN2 in MZMΦs was associated with a twofold increase in IL-12p40 andfourfold increases in IL-6 and IFNβ1 compared with R2B MZMΦs, indicating increased inflammatory potential. In agreementwith Fig. 2A, deletion of GCN2KO was associated with a loss ofIL-10 mRNA relative to R2B controls in the MZ MΦs, sug-gesting compromised regulatory cytokine production, which mayexacerbate inflammation (Fig. 4A). Cytokine measurementsfrom whole spleen lysates agreed with the mRNA analysis, withan elevated inflammatory phenotype in R2B GCN2flLysMcre spleencompared with R2B mice (Fig. 4B).Total serum IgG was increased in R2B mice compared with

the B6 mice, but was not affected by GCN2 deletion (Fig. 4C). Incontrast, there was a significant increase in serum αdsDNA IgG

Fig. 3. MZ MΦ and B-cell activation are increased in lupus-prone micelacking myeloid GCN2 function. (A) Splenocyte numbers from 6-mo-old miceof the indicated genotype. (B) Representative Western blot of whole spleenlysate from two mice per group described in A. m, mouse. (C–E) Flowcytometry analysis of groups described in A. In D, B-cell plots shown aregated on a B220+ population. Graphs in E are gated on the markers in-dicated above. Bars represent mean ± SD values for eight mice. *P < 0.05,**P < 0.01, Student’s t test. ns, not significant. Experiments were repeatedfour times, with similar results.

Ravishankar et al. PNAS | August 25, 2015 | vol. 112 | no. 34 | 10777

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

in R2B GCN2flLysMcre mice beginning at 4 mo of age (Fig. 4D).When kidneys from 6- to 7-mo-old mice were examined for pa-thology, both R2B and R2B GCN2flLysMcre mice showed globalpathology with enlarged and sclerotic glomeruli, tubulointer-stitial and perivascular infiltrate, and tubular atrophy with castformation (Fig. S5A).Whereas both R2B and R2B GCN2flLysMcre mice had similar

glomerular pathology scores, R2B GCN2flLysMcre mice exhibitedincreased tubular injury (Fig. S5B). R2B and R2B GCN2flLysMcre

mice had prominent diffuse immune complex deposition in theglomeruli consisting primarily of IgG associated with complementC3 fixation (Fig. S5C). F4/80+ MΦs were conspicuous throughoutthe tubules, and staining for MΦs was similar in R2B and R2BGCN2flLysMcre mice (Fig. S5C). In agreement with tubular pa-thology scores, proteinuria was greater in R2B GCN2flLysMcre

mice compared with R2B mice, indicative of increased kidneydisease (Fig. 4E).Renal failure is the primary driver of death in murine lupus,

and we found increased mortality in the R2B GCN2flLysMcre

mice, with 50% mortality at 7.5 mo and 100% mortality at 11 mo,compared with 52% mortality at 12 mo in R2B mice (Fig. 4F),suggesting that myeloid deficiencies in GCN2 enhanced in-flammation, ultimately driving pathology and mortality.FcγRIIb-deficient mice respond abnormally to apoptotic cell

challenge in vivo, with increased plasma cell formation and DCresponses (23). Based on the foregoing data, we hypothesized thatGCN2 deletion would increase the dysfunctional responses to ap-optotic cells driving elevated autoantibody production after expo-sure. To test this, we challenged R2B and R2B GCN2flLysMcre

mice four times (once weekly) with 107 apoptotic thymocytes and

collected serum 1 d before each apoptotic cell challenge, with afinal collection at day 28 (Fig. S6), and measured αdsDNA IgGproduction by ELISA. By the final collection (day 28), repeatedapoptotic cell exposure had increased the αdsDNA IgG con-centration 4.5-fold in R2B mice relative to sham-injected controls(Fig. S7A). R2B GCN2flLysMcre mice challenged with apoptoticcells showed a similar fold increase in αdsDNA IgG at day 28 ofthe experiment; however, the increased baseline autoreactivity(which was twofold higher at day 28 relative to the R2B group)translated to a significant increase in autoreactivity in GCN2KOmice challenged with apoptotic cells (Fig. S7A). These resultssuggest that the lack of GCN2 signals in myeloid cells potentiatesthe abnormal apoptotic cell response in R2B mice, promoting amore vigorous autoimmune response to self.Halofuginone (HF) acts as a GCN2 agonist by inhibiting

prolyl-tRNA synthetase (24). Because GCN2 stress signals are akey effector of tolerogenic immunity, we predicted that GCN2activation with HF would reduce autoimmunity in R2B mice. Totest this, we treated 4-mo-old female mice with HF for 2 mo andmeasured serum αdsDNA IgG at the experimental endpoint. HFreduced αDNA IgG by fivefold, suggesting that GCN2 activationcan suppress systemic autoimmunity (Fig. S7B). IDO1 inhibitionwith D1MT significantly increased αdsDNA serum concentrationsin R2B mice (Fig. S7B). In these mice, treatment with HF had aneven more pronounced effect, reducing autoreactivity by 10-foldcompared with the control D1MT-treated animals. Finally, theeffect was dependent on GCN2, given that R2B GCN2flLysMcre

mice did not show reductions in αdsDNA IgG concentrations afterHF treatment (Fig. S7B). Thus, our data support the prospect oftargeted GCN2 manipulation as a therapy for systemic autoim-mune disease.

DiscussionThe ability to sense the available nutrient supply is an essentialcomponent of cellular physiology. As a consequence, multiplemechanisms have evolved to allow cells to respond to fluctuatingavailability of glucose and ATP, amino acids, and oxygen re-quired for aerobic respiration and biosynthesis. Several of thesemolecular “sensors” play key roles in immunity, driving cellularexpansion, differentiation, and the biosynthesis profile.The primary sensors of amino acid availability are mTOR and

GCN2. mTOR (as a component of mTORC1) interacts withbranched-chain amino acids to promote translation and ribosomalassembly (25), whereas GCN2 binds directly to any unchargedtRNA, driving its kinase activity and inhibiting ribosomal assemblyon nascent transcripts (10). GCN2 has been identified as a mo-lecular target for IDO1-induced proliferative arrest and anergy innaïve T cells (26). Subsequently, GCN2 was found to be requiredfor IDO1-driven Treg activation (27, 28) indicating a key role asan effector for acquired T-cell tolerance.We recently reported that IDO1 is required for apoptotic cell-

induced tolerance (6). Specifically, IDO1 was rapidly inducedin MZ MΦ after apoptotic cell challenge in vivo, suppressingMΦ and DC production of inflammatory cytokines, inhibitingimmunity to apoptotic antigens, and limiting autoimmunity inMRLlpr/lpr mice. In apoptotic cell-injected mice, expression of thestress gene CHOP corresponded to the pattern of IDO1 ex-pression, and inhibition of IDO1 abrogated CHOP mRNA in-duction, suggesting that an IDO1-dependent stress signal wasinduced in apoptotic cell phagocytes. Because GCN2 is theprimary effector of IDO1 stress, this result raises the possibilitythat GCN2 is required for myeloid responses to apoptotic cells;however, the role of GCN2 in MΦ and DC cell function has notbeen examined until recently.We previously found that GCN2 signals have a significant effect

on MΦ cytokine production after activation by toll-like receptorligands, enhancing transcription of IL-6 but providing a trans-lational block by virtue of eIF2α phosphorylation (11). The datafrom the present study suggest a similarly dominant role for GCN2in the apoptotic cell response, with the translational block man-ifested by GCN2 activation serving to prevent inflammatory

Fig. 4. GCN2 limits inflammatory cytokine production and systemic auto-immune disease in lupus-prone mice. (A) SignR1+ MZ MΦs were sorted from6-mo-old mice of the indicated genotype by FACS and cytokine message forthe indicated species was determined by sqPCR. (B) Whole spleen lysatesfrom 6-mo-old mice of the indicated genotype were collected, and cytokineprotein concentrations were determined by ELISA. (C) Total serum IgG from6-mo-old mice was measured by ELISA. (D) Serum was collected monthlyfrom mice at the indicated ages, and IgG reactivity against dsDNA wasmeasured by ELISA. (E) Urine was collected from 6- to 7- mo-old mice, andprotein concentrations were measured as described inMaterials and Methods.(F) Survival for mice of the indicated genotype over a 12-mo period wasmonitored. In A, bars represent the value for pooled samples from four mice.In B and C, bars represent mean ± SD values for eight mice. In D and E,graphs show mean ± SD values for eight mice. In F, n = 10 mice per group,and significance was determined as described in Materials and Methods.*P < 0.05, **P <0.01, ***P < 0.0001, Student’s t test. ns, not significant.Experiments were repeated three times, with similar results, with the ex-ception of the experiment shown in F, which was repeated twice.

10778 | www.pnas.org/cgi/doi/10.1073/pnas.1504276112 Ravishankar et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

cytokine production creating a permissive translational environ-ment for IL-10 synthesis and downstream immune suppression.GCN2 deletion in myeloid cells had a significant effect on

systemic autoimmunity, accelerating immune dysfunction andexacerbating pathology and mortality. This is consistent with arole for GCN2 in regulating myeloid inflammatory activity andsuggests a function in innate responses to apoptotic cells andrestriction of autoimmunity. Sundrud et al. (29) reported thatHF inhibits Th17 T-cell differentiation, protecting mice in anexperimental model of multiple sclerosis by inducing a GCN2-dependent stress response. A later study found that the mecha-nism of action was prolyl-tRNA synthetase inhibition mimickingproline starvation (24). We found that treatment of lupus-pronemice with HF significantly reduced autoimmunity. Importantly,when IDO1 was inhibited, HF still had the capacity to potentlyreduce αdsDNA IgG concentrations in serum, suggesting a by-pass of the need for IDO1 to restrict autoimmunity. Moreover,the finding that HF had no affect on humoral autoimmunity inGcn2flLysMcre mice suggests that agonizing GCN2 activity inmyeloid cells can restrict autoimmunity. Nevertheless, this resultmust be interpreted with caution, because GCN2 is expressed inall cell types. Therefore, we cannot not rule out the possibility ofa MΦ-independent role for GCN2 in prevention of inflammatoryautoimmunity after HF treatment.In conclusion, this report provides evidence of a mechanism by

which metabolic stress signals delivered by GCN2 inhibit innate

inflammatory responses to apoptotic cells, restricting adaptiveimmunity and preventing autoimmunity. Importantly, the dem-onstration that GCN2 activation by HF potently reduced auto-immune reactivity suggests GCN2 represents a therapeutic targetin autoimmune disease therapy.

Materials and MethodsAnimals. C57BL/6J (B6), B6.Act-mOVA-II (Act-mOVA), B6.LysM-Cre, OTII Thy1.1+,B6.Eif2ak4tm1.2Dron (GCN2−/−), B6.Eif2ak4tm1.1Dron (GCN2flox/flox), and B6.Fcgr2b−/−

mice were obtained from a colony maintained at Georgia Regents University, inaccordance with Institutional Animal Care and Use Committee guidelines.

Lentiviral Transduction. BMMΦs were generated as previously described (23).For further description see SI Materials and Methods.

Halofuginone Treatment. Halofuginone hydrobromide (HF) (Sigma) wasresuspended in phosphate buffered saline (PBS) at a concentration of 1 mg/mL.For in vivo administration, mice were injected with HF (1 μg/gm) by in-traperitoneal administration twice weekly for a period of 8 wk.

Other Materials and Methods. All other methods are described in SI Materialsand Methods.

ACKNOWLEDGMENTS. We thank Dr. Brad Wouters for assistance in thisresearch report. This work was supported by National Institutes of HealthGrants AI092213 and AI105500 (to T.L.M.).

1. Casciola-Rosen LA, Anhalt G, Rosen A (1994) Autoantigens targeted in systemic lupuserythematosus are clustered in two populations of surface structures on apoptotickeratinocytes. J Exp Med 179(4):1317–1330.

2. Rosen A, Casciola-Rosen L (1999) Autoantigens as substrates for apoptotic proteases: Im-plications for the pathogenesis of systemic autoimmune disease. Cell Death Differ 6(1):6–12.

3. Colonna L, Lood C, Elkon KB (2014) Beyond apoptosis in lupus. Curr Opin Rheumatol26(5):459–466.

4. McGaha TL, Chen Y, Ravishankar B, van Rooijen N, Karlsson MC (2011) Marginal zonemacrophages suppress innate and adaptive immunity to apoptotic cells in the spleen.Blood 117(20):5403–5412.

5. Ravishankar B, McGaha TL (2013) O death where is thy sting? Immunologic toleranceto apoptotic self. Cell Mol Life Sci 70(19):3571–3589.

6. Ravishankar B, et al. (2012) Tolerance to apoptotic cells is regulated by indoleamine2,3-dioxygenase. Proc Natl Acad Sci USA 109(10):3909–3914.

7. Ravishankar B, et al. (2014) Marginal zone CD169+ macrophages coordinate apoptotic cell-driven cellular recruitment and tolerance. Proc Natl Acad Sci USA 111(11):4215–4220.

8. Grönwall C, et al. (2012) MAPK phosphatase-1 is required for regulatory naturalautoantibody-mediated inhibition of TLR responses. Proc Natl Acad Sci USA 109(48):19745–19750.

9. Sharma S, et al. (2015) Suppression of systemic autoimmunity by the innate immuneadaptor STING. Proc Natl Acad Sci USA 112(7):E710–E717.

10. McGaha TL, et al. (2012) Amino acid catabolism: A pivotal regulator of innate andadaptive immunity. Immunol Rev 249(1):135–157.

11. Liu H, et al. (2014) GCN2-dependent metabolic stress is essential for endotoxemiccytokine induction and pathology. Mol Cell Biol 34(3):428–438.

12. Baird TD, Wek RC (2012) Eukaryotic initiation factor 2 phosphorylation and trans-lational control in metabolism. Adv Nutr 3(3):307–321.

13. Harding HP, et al. (2000) Regulated translation initiation controls stress-induced geneexpression in mammalian cells. Mol Cell 6(5):1099–1108.

14. B’chir W, et al. (2013) The eIF2α/ATF4 pathway is essential for stress-induced auto-phagy gene expression. Nucleic Acids Res 41(16):7683–7699.

15. Ravindran R, et al. (2014) Vaccine activation of the nutrient sensor GCN2 in dendriticcells enhances antigen presentation. Science 343(6168):313–317.

16. Simpson E, Scott D, Chandler P (1997) The male-specific histocompatibility antigen,H-Y: A history of transplantation, immune response genes, sex determination andexpression cloning. Annu Rev Immunol 15:39–61.

17. Kwidzinski E, et al. (2005) Indolamine 2,3-dioxygenase is expressed in the CNS anddown-regulates autoimmune inflammation. FASEB J 19(10):1347–1349.

18. Anderson G, Rodriguez M (2010) Multiple sclerosis, seizures, and antiepileptics: Roleof IL-18, IDO, and melatonin. Eur J Neurol 18(5):680–685.

19. Kurz K, et al. (2011) Effects of adalimumab therapy on disease activity and interferon-γ-mediated biochemical pathways in patients with rheumatoid arthritis. Autoimmunity44(3):235–242.

20. Gurtner GJ, Newberry RD, Schloemann SR, McDonald KG, Stenson WF (2003) In-hibition of indoleamine 2,3-dioxygenase augments trinitrobenzene sulfonic acid co-litis in mice. Gastroenterology 125(6):1762–1773.

21. Jasperson LK, et al. (2008) Indoleamine 2,3-dioxygenase is a critical regulator of acutegraft-versus-host disease lethality. Blood 111(6):3257–3265.

22. McGaha TL, Sorrentino B, Ravetch JV (2005) Restoration of tolerance in lupus bytargeted inhibitory receptor expression. Science 307(5709):590–593.

23. McGaha TL, Karlsson MC, Ravetch JV (2008) FcgammaRIIB deficiency leads to auto-immunity and a defective response to apoptosis in Mrl-MpJ mice. J Immunol 180(8):5670–5679.

24. Keller TL, et al. (2012) Halofuginone and other febrifugine derivatives inhibit prolyl-tRNA synthetase. Nat Chem Biol 8(3):311–317.

25. Han JM, et al. (2012) Leucyl-tRNA synthetase is an intracellular leucine sensor for themTORC1-signaling pathway. Cell 149(2):410–424.

26. Munn DH, et al. (2005) GCN2 kinase in T cells mediates proliferative arrest and anergyinduction in response to indoleamine 2,3-dioxygenase. Immunity 22(5):633–642.

27. Baban B, et al. (2009) IDO activates regulatory T cells and blocks their conversion intoTh17-like T cells. J Immunol 183(4):2475–2483.

28. Sharma MD, et al. (2009) Indoleamine 2,3-dioxygenase controls conversion of Foxp3+

Tregs to TH17-like cells in tumor-draining lymph nodes. Blood 113(24):6102–6111.29. Sundrud MS, et al. (2009) Halofuginone inhibits TH17 cell differentiation by activating

the amino acid starvation response. Science 324(5932):1334–1338.30. Feldman S, et al. (1974) In vitro assessment of in vivo absorption of drug complexes.

J Pharm Sci 63(3):454–456.31. Palmiter RD (1974) Magnesium precipitation of ribonucleoprotein complexes: Expe-

dient techniques for the isolation of undergraded polysomes and messenger ribo-nucleic acid. Biochemistry 13(17):3606–3615.

32. McGaha TL, Phelps RG, Spiera H, Bona C (2002) Halofuginone, an inhibitor of type-Icollagen synthesis and skin sclerosis, blocks transforming growth factor-beta–mediatedSmad3 activation in fibroblasts. J Invest Dermatol 118(3):461–470.

33. Huang L, et al. (2012) Engineering DNA nanoparticles as immunomodulatory re-agents that activate regulatory T cells. J Immunol 188(10):4913–4920.

34. Sharma MD, et al. (2007) Plasmacytoid dendritic cells from mouse tumor-draininglymph nodes directly activate mature Tregs via indoleamine 2,3-dioxygenase. J ClinInvest 117(9):2570–2582.

35. Jha V, et al. (2014) Lymphocyte Activation Gene-3 (LAG-3) negatively regulatesenvironmentally-induced autoimmunity. PLoS One 9(8):e104484.

Ravishankar et al. PNAS | August 25, 2015 | vol. 112 | no. 34 | 10779

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020