fj.14-264507.full

The FASEB Journal • Research Communication

CXCR3 expression defines a novel subset of innate CD8+

T cells that enhance immunity against bacterialinfection and cancer upon stimulation with IL-15

Steve Oghumu,*,† Cesar A. Terrazas,* Sanjay Varikuti,* Jennifer Kimble,* Stephen Vadia,‡

Lianbo Yu,§ Stephanie Seveau,‡,{ and Abhay R. Satoskar*,‡,1

*Department of Pathology, The Ohio State University Medical Center, Columbus, Ohio, USA;†Department of Oral Biology, The Ohio State University College of Dentistry, Columbus, Ohio, USA; and‡Department of Microbiology, §Center for Biostatistics, and {Department of Microbial Infection andImmunity, The Ohio State University, Columbus, Ohio, USA

ABSTRACT Innate CD8+ T cells are a heterogeneouspopulation with developmental pathways distinct fromconventional CD8+ T cells. However, their biology, classi-fication, and functions remain incompletely understood.We recently demonstrated the existence of a novel pop-ulation of chemokine (C-X-C motif) receptor 3 (CXCR3)-positive innate CD8+ T cells. Here, we investigated thefunctional properties of this subset and identified effectormolecules and pathways which mediate their function.Adoptive transfer of IL-15 activated CXCR3+ innate CD8+

T cells conferred increased protection against Listeriamonocytogenes infection in susceptible IFN-g2/2 micecompared with similarly activated CXCR32 subset. Thiswas associated with enhanced proliferation and IFN-gproduction in CXCR3+ cells. Further, CXCR3+ innate cellsshowed enhanced cytotoxicity against a tumor cell line invitro. In depth analysis of the CXCR3+ subset showed in-creased gene expression of Ccl5, Klrc1, CtsW, GP49a,IL-2Rb, Atp5e, and Ly6c but reduced IFN-gR2 and Art2b.Ingenuity pathway analysis revealed an up-regulation ofgenes associated with T-cell activation, proliferation, cy-totoxicity, and translational initiation in CXCR3+ pop-ulations. Our results demonstrate that CXCR3 expressionin innate CD8+ T cells defines a subset with enhanced cy-totoxic potential and protective antibacterial immunefunctions. Immunotherapeutic approaches against in-fectious disease and cancer could utilize CXCR3+ innateCD8+ T-cell populations as novel clinical interventionstrategies.—Oghumu, S., Terrazas, C. A., Varikuti, S.,Kimble, J., Vadia, S., Yu, L., Seveau, S., and Satoskar, A. R.CXCR3 expression defines a novel subset of innate CD8+

T cells that enhance immunity against bacterial infectionand cancer upon stimulation with IL-15. FASEB J.29, 000–000 (2015). www.fasebj.org

Key Words: granzyme • interferon • listeria • cytokine •

cytotoxicity

THE DISCOVERY OF INNATE CD8+ T cells has broadened ourunderstanding of T-cell subset classification, development,and function. Phenotypically, these cells are known tohavea comparatively restricted T-cell receptor repertoire andexpress markers for memory T cells, including CD44 andCD122, allowing them togenerate rapid effector responsesupon cytokine stimulation (1, 2). Moreover, unlike acti-vated CD8+ T cells generated during adaptive immunity,innate CD8+ T cells are present in naive, and even germ-free, mice. A number of studies have identified factorswhich contribute to the development, maintenance andeffector functions of innateCD8+T cells. Theirmaturationin the thymus involves nonclassic major histocompatibilitycomplex class I molecules, and unlike conventional CD8+

T cells, they do not require Tec protein tyrosine kinases IL-2 inducible T-cell kinase (Itk) or resting lymphocyte kinase(Rlk) (2). Recently, the signaling lymphocyte activationmolecule (SLAM) family receptors coupled with a down-stream SLAM-associated protein signaling adaptor mole-cule have been shown to regulate the development andmaintenance of this population, which is dependent onthe production of IL-4 (3, 4). Other reports have demon-strated a requirement for IL-15 in the maintenance andmediation of the effector functions of innate CD8+ T cells(5). However, specific mechanisms governing the gener-ation of innate CD8+ T-cell populations, as well as theirfunctions, are still not completely understood.

We have recently demonstrated that a large proportionof innate CD8+ T cells expresses chemokine (C-X-Cmotif)receptor 3 (CXCR3), and this subset expresses increasedlevels of activation markers and responds more rapidly toIL-2 and IL-15 stimulation than CXCR32 innate CD8+

T cells (6). This was accompanied by increased expressionof IFN-g as well as the cytotoxic molecule granzyme B.These results present the possibility that these cells couldbe effective in antitumor immune responses as well as incontributing to immunity against intracellular bacteria.Previous reports have demonstrated a role for class Ib

Abbreviations: CCL5, chemokine (C-X-C motif) ligand 5;CIBER, CXCR3 IRES Bb-cistronic EGFP reporter mouse;CTL, cytotoxic T lymphocyte; CXCR3, chemokine (C-X-Cmotif) receptor 3; eIF, eukaryotic initiation factor; FACS,fluorescence activated cell sorter; IPA, ingenuity pathwayanalysis; qRT-PCR, quantitative RT-PCR

1 Correspondence: Department of Pathology, 1645 Neil Ave.,Columbus, OH 43210, USA. E-mail: [email protected]: 10.1096/fj.14-264507

0892-6638/15/0029-0001 © FASEB 1

The FASEB Journal article fj.14-264507. Published online December 2, 2014.

restricted innate CD8+ T-cell populations in early antibac-terial immune responses before the onset of adaptiveimmunity (7–10).CXCR3-expressing subpopulationsof in-nate CD8+ T cells could potentially provide more potentimmune responses against a bacterial infectious challenge.Moreover, because activatedCD8+Tcells play a vital role inantitumor immunity, strategies aimed at activatingCXCR3expressing innate CD8+ T cells could be a viable approachto cancer immunotherapy.

Given the high importance yet incomplete under-standing of the biology and function of the heterogeneouspopulation of innate CD8+ T cells, we have further char-acterized subsets of this population and identified effectormolecules which mediate their function. We have alsoexamined the relative contributions of these populationsto antibacterial as well as antitumor cell responses. Ourresults indicate that CXCR3 expressing innate CD8+ T-cellpopulations display enhanced cytotoxicity against tumorcells and provide increased protection against primary in-fection by Listeria monocytogenes in vivo. These studiesstrongly suggest that CXCR3 expressing innate CD8+

T cells could represent a possible vaccine target in themanagement of intracellular bacterial infection and neo-plastic disease.

MATERIALS AND METHODS

Mouse strains

C57BL/6 wild-type and C57BL/6 IFN-g knockout mice werepurchased fromThe JacksonLaboratory (BarHarbor,ME,USA).CXCR3 IRES Bicistronic EGFP reporter (CIBER) mice (back-crossed to C57BL/6 background for 13 generations) were gen-erated by our group as described previously (6). All mice usedwere maintained in a pathogen-free animal facility at The OhioState University in accordance with U.S. National Institutes ofHealth and institutional guidelines.

Flow cytometry and cell sorting

Single cell suspensions fromspleensor lymphnodeswerederivedfrom naive CIBER mice, washed with PBS and blocked withnormal mouse serum or anti-CD16/CD32 antibodies. In someexperiments, T cells were enriched by passing splenocytesthroughnylonwool column(Polysciences,Warrington,PA,USA)according to the manufacturer’s instructions. Cells were in-cubated with fluorescently labeled anti-CD8, anti-CD62L, andanti-CD44 antibodies (Biolegend, San Diego, CA, USA). For in-tracellular staining, stimulated cells were stained for extracellularmarkers, fixed with 2% para-formaldehyde, permeabilized, andstainedwith anti–IFN-g antibodies (Biolegend). Cells were eitheracquired on a fluorescence activated cell sorter (FACS) Cantoflow cytometer or sorted on a FACS Aria cell sorter (BD Bio-sciences, San Jose, CA, USA) at the flow cytometry core facility atOhio State University Medical Center. Analysis was performedwith CellQuestPro software (BD Biosciences) or FlowJo software(Tree Star Incorporated, Ashland, OR, USA), and sorted pop-ulations were used for in vivo and in vitro experiments.

Microarray analysis

Total RNAwas isolated from sortedCXCR3+ andCXCR32 innateCD8+ T-cell as well as naive CD8+ T-cell populations from about 3

to 5 CIBER mice using an RNeasy kit (Qiagen, Valencia, CA,USA). RNA quantity, quality, and integrity were confirmed byNanodrop and Agilent Bioanalyzer before inclusion in the array.Microarray processing was performed at the Micro Array SharedResource, The Ohio State University. RNA amplification, frag-mentation, and labeling were carried out according to manu-facturer’s protocols (Affymetrix, SantaClara,CA,USA).The arrays(GeneChip Mouse Gene 2.0ST) were hybridized for 16 h at 45°Cand 60 rpm. Washing and staining of arrays was performed at thefluidics station 450 according to manufacturer’s protocol(Affymetrix). The microarrays were scanned using an AffymetrixGeneChip Scanner 3000 7GwithAffymetrixGeneChipCommandConsole (AGCC) software. Background correction and quantilenormalization was performed to adjust technical bias, and ex-pression levels were summarized over the probe set using the ro-bustmultiarray averagemethod (11). A filteringmethod based onpercentageof arrays abovenoise cutoffwas applied tofilter out low-expression genes. Affymetrix Expression Console software and Rstatistical software (http://www.r-project.org/) was used for theanalysis. Microarray expression data have been submitted to theGene Expression Omnibus (GSE accession no. GSE60068).

Ingenuity pathway analysis of gene expression arrays

Molecular interactions among differentially regulated genes be-tween CXCR3+ and CXCR32 innate CD8+ T cells were exploredusing ingenuity pathway analysis (IPA) (Qiagen). Each mousegene identifier was mapped to its corresponding gene in the In-genuity Pathway Knowledge Base. Families of genes that were up-or down-regulated in CXCR3 expressing innate CD8+ T cellscompared to CXCR32 innate CD8+ T cells were integrated intopredictive network models on the basis of gene interactionswithin a biologic pathwaydefined in the literature as contained inthe Ingenuity Pathway Knowledge Base.

RT-PCR validation

Total RNA was extracted from naive, CXCR3+ and CXCR32 in-nateCD8+T-cell populationsusing theRNeasy kit (Qiagen).RNAwas reverse transcribed to cDNAusing theSuperScriptVilo cDNAsynthesis kit (Invitrogen, Carlsbad, CA, USA). Primer sequencesand cycling conditions for RT-PCR were obtained using thePrimer Bank website (http://pga.mgh.harvard.edu/primer-bank/index.html), Harvard Medical School. PCR amplificationwas performed in a CFX 96 RT-PCR cycler (Bio-Rad, Hercules,CA, USA) using SYBR Green (Bio-Rad) for detection. Data werenormalized to b-actin and presented as fold induction over naivecells using the DDCT method.

Adoptive transfer and cellular analysis

Spleens were removed aseptically from naive CIBER mice, andsingle cell suspensionswereprepared after lysis of red blood cells.Naive CD8+, CXCR3+ innate CD8+, and CXCR32 innate CD8+

T-cell populations were sorted using FACS Aria (BDBiosciences)as described. Sorted cells were preactivated with IL-15 beforeadoptive transfer as previously demonstrated (12). One millioncells from sorted, preactivated populations were labeled withCellTrace Violet Cell Proliferation Kit (Invitrogen) and injectedinto IFN-g knockout mice via tail vein injections. On d 1 afteradoptive transfer, recipient mice were infected with L. mono-cytogenes. On d 4 after adoptive transfer, mice were humanelykilled, and spleens were analyzed for the presence of adoptivelytransferred cells, cell proliferation, and intracellular IFN-g pro-duction (Biolegend) by flow cytometry.

2 Vol. 29 March 2015 OGHUMU ET AL.The FASEB Journal x www.fasebj.org

Bacterial infection and enumeration

Listeria monocytogenes strain 10403S (wild type; strain was a giftfrom Dr. D. Portnoy, University of California, Berkeley, CA,USA) was grown overnight at 37°C in brain–heart infusion.Overnight cultures were diluted 1/20 in brain–heart infusionand grown at 37°C until OD600 = 0.7 to 0.8. Bacteria werewashed 3 times and diluted in PBS. Mice were infected by tailvein injection with 104 bacteria. Livers and spleens were col-lected 72 h after infection. To enumerate colony-formingunits, organs were homogenized in PBS, serial dilutions ofhomogenates were plated on brain–heart infusion agar, andsamples were incubated at 37°C for 24 h.

Cytotoxicity assay

Naive CD8+, CXCR3+ innate CD8+, and CXCR32 innate CD8+

T cells were sorted from naive CIBER mice and then stimu-lated with IL-15 for 48 h. Stimulated CD8+ T-cell subsets werecocultured with P815 mastocytoma cell line (American TypeCulture Collection, Manassas, VA, USA) at E:T ratios of 1:1and 5:1. Cytotoxicity of target cells was measured by lactatedehydrogenase release using the Cytotoxicity Detection Kit(Roche Diagnostics, Indianapolis, IN, USA) with appropriatecontrols.

Statistical analysis

Statistical analyses were performed by Prism 5 software (Graph-Pad Software, San Diego, CA, USA). Student’s unpaired t testwas used to determine statistical significance of values obtained.P values of ,0.05 were considered statistically significant. TheR statistical package was used for the analysis of microarray data.

RESULTS

CXCR3+ innate CD8+ cells confer increasedprotection against L. monocytogenes infection

We recently identified a subset of innate CD8+ T cells thatexpress the chemokine receptor CXCR3 (6). These cellswere distinct from CXCR32 innate CD8+ T cells in theexpression of antiapoptotic factors as well as in the pro-duction of granzyme B and IFN-g after stimulation withIL-2 or IL-15 (6), which suggested an enhanced role forCXCR3+ innate CD8+ T cells in antibacterial immuneresponses. This led us to test the protective effect of thispopulation against an infectious challenge by an intra-cellular bacterium. We used an in vivo bacterial infectionmodel using L. monocytogenes, which causes a food-borne

Figure 1. CXCR3+ innate CD8+ cells confer increased protection against L. monocytogenes infection. Bacterial loads in the (A)spleen and (B) liver of L. monocytogenes infected wild-type C57BL/6 mice or IFN-g knockout mice that received either no cells(IFN-g2/2), CXCR3+ innate CD8+ T cells (CXCR3+ . IFN-g2/2), or CXCR32 innate CD8+ T cells (CXCR32 . IFN-g2/2). C, D)Intracellular IFN-g production by (C) CXCR3+ or (D) CXCR32 innate CD8+ T cells purified from CIBER mice and adoptivelytransferred to IFN-g2/2 mice and subsequently infected with L. monocytogenes. E) Percentage of IFN-g producing, adoptivelytransferred innate CD8+ T cells in IFN-g2/2 mice infected with L. monocytogenes. F, G) Cellular proliferation of sorted andadoptively transferred CXCR3+ and CXCR32 innate CD8+ T-cell populations in IFN-g2/2 mice infected with L. monocytogenes.Percentages of proliferating cells are depicted in (G). Data are presented as mean 6 SEM and are representative of 2 separateexperiments from 3 to 5 individual mice per group. *P , 0.05, **P , 0.01, ***P , 0.001.

CXCR3 CONFERS IMMUNITY TO INFECTION AND CANCER 3

disease called listeriosis in animals and humans. Protectiveimmune response against L. monocytogenes is T-cell medi-ated, and IFN-g is essential for early control of infection.IFN-g–deficient mice are highly susceptible to L. mono-cytogenes infection, and early IFN-g production primarilyby CD8+ T cells is essential for innate immune protec-tion (13). This is therefore a suitable bacterial infectionmodel to test theprotective abilityofour innateCD8+T-cellsubsets. CXCR3+ and CXCR32 innate CD8+ T cells weresorted from naive CIBER mice and activated in vitro withIL-15. Sorted cells were transferred intravenously into IFN-g–deficient C57BL/6 mice 1 d before L. monocytogenes in-fection. Three days after infection, bacterial loads wereevaluated in livers and spleens of recipient mice, and theproliferative potential and effector functions of adoptivelytransferred innate CD8+ T-cell subsets were analyzed. Asexpected, IFN-g–deficient mice had much higher bacterialloads in their livers and spleens compared to wild-type con-trols. IFN-g–deficientmiceadoptively transferredwith innateCD8+ T-cell subsets had significantly reduced bacterial bur-dens in their livers and spleens compared to nontransferredIFN-g–deficientmice(Fig.1A,B).MicethatreceivedCXCR3+

innate CD8+ T cells displayed significantly less bacterialloads in the spleens than those that received CXCR32 in-nate CD8+ T cells (Fig. 1A). Bacterial counts in the liverwere also less inCXCR3+ innateCD8+ cell-transferredmice,although the difference was not significant (Fig. 1B).

Adoptively transferredCXCR3+andCXCR32 innateCD8+

T cells were detected in the spleens of infected mice atd 3 after infection. The protective effect of the CXCR3+

innate CD8+ T-cell subset was accompanied by enhancedIFN-g production compared to CXCR32 innate CD8+

T cells, as demonstrated by intracellular flow cytometricanalysis of adoptively transferred cells in the spleens ofrecipient mice (Fig. 1C–E). Further, we observed thattransferred CXCR3+ innate CD8+ T cells proliferated toa greater extent than the CXCR32 innate CD8+ T-cellpopulation afterL.monocytogenes infection in vivo (Fig. 1F,G). Taken together, these results demonstrate that

CXCR3+ innate CD8+ T cells confer greater protectionagainst infection by L. monocytogenes.

CXCR3+ innate CD8+ T cells show enhanced tumorcytotoxicity in vitro

Immune cytotoxic effector cells are major contributors toantitumor immunity (14). Previous work on innate CD8+

T cells and our current detailed transcriptional analysisof this population suggest that CXCR3 expressing innateCD8+ T cells are potent mediators of antitumor immunity(6). We therefore tested the functional ability of CXCR3expressing and nonexpressing innate CD8+ T cells in me-diating cytotoxicity against P815 target cancer cells in vitro.Sorted innate CD8+ cell subsets from naive mice werepreactivated with IL-15 in vitro and cocultured with P815target cells. CXCR3+ innate CD8+ T cells were significantlymore potent than CXCR32 innate CD8+ T cells and naiveCD8+ T cells at killing target tumor cells in vitro (Fig. 2).Unstimulated cells were not cytotoxic to tumor cells in vitro(Fig. 2). Our data demonstrate that CXCR3+ innate CD8+

cells are potentially more efficient inmediating tumor cellcytotoxicity. These results demonstrate the possibility ofexploiting this subset of innate CD8+ T cells in therapeuticapproaches to cancer immunosuppression and subsequenttumor regression.

Gene expression of innate CD8+ T-cell subsets

To further characterize potential mechanisms of action be-hind the potent antibacterial and antitumor immunitycharacteristic ofCXCR3+ innateCD8+Tcells, weperformedgene expression profiling of sorted naive, CXCR3+, andCXCR32 innate CD8+ T cells from naive mice (Fig. 3A).When compared to innate CD8+ T cells that do not expressCXCR3, CXCR3+ innate CD8+ T cells displayed enhancedexpressionof thechemokine(C-X-Cmotif) ligand5(CCL5)as well as genes associated with cellular activation and pro-liferation including killer cell lectin-like receptor subfamily

Figure 2. CXCR3+ innate CD8+ T cells show enhanced tumor cell cytotoxicity in vitro. A, B) Cytotoxic activity of IL-15–stimulated naive,CXCR3+ innate CD8+ T cells, and CXCR32 innate CD8+ T-cell populations against P815 mastocytoma cell line at effector to target ratiosof (A) 1:1 and (B) 5:1, as determined by lactate dehydrogenase release assay. Unstimulated cells are also shown. Data are presented asmean 6 SEM of 3 individual samples and are representative of 2 separate experiments. **P , 0.01, ***P , 0.001. ND, not detectable.

4 Vol. 29 March 2015 OGHUMU ET AL.The FASEB Journal x www.fasebj.org

C, member 1 (klrc1), glycoprotein 49A (Gp49a), and IL-2receptor b (IL-2Rb). The cysteine protease known to regu-late T-cell cytolytic activity, cathepsin W (ctsW), was also up-regulated inCXCR3expressing innateCD8+Tcells (Fig. 3B,C). This subset also up-regulated genes that have previouslybeen shown to be associated with a T-bet–dependent type 1CD8 cytotoxic T-cell (Tc1) programming, which leads toenhanced antitumor responses, including T-box 21 (tbx21),klrc1, IL-2 receptorb (il2rb), cytotoxicT lymphocyte (CTL)–associated protein 2 a (ctla2a), CTL-associated protein 2 b(ctla2b), and Fas ligand (faslg), which have been shown to beimportant in CD8+ T-cell cytotoxicity (15). Other genes sig-nificantly up-regulated include ATP synthase, H+ trans-porting, mitochondrial F1 complex, e subunit (ATP5e)involved in ATP synthesis and oxidative phosphorylation,and Ly6c important in signal transduction and cytokineproduction duringT-cell activation. In support of this result,previouswork inour laboratorydemonstrated increasedcellsurface expression of Ly6c in CXCR3+ innate CD8+ T cells(6). Genes that were significantly down-regulated inCXCR3+ innate CD8+ T cells compared to CXCR32 innateCD8+ T cells include IFN-g receptor 2 (IFNgR2) andADP-ribosyl transferase 2b (Art2b), which is involved inNAD+-induced cell death (Fig. 3B, C).

To confirm the results of the gene expression micro-array, we performed quantitative RT-PCR (qRT-PCR) val-idation on RNA obtained from naive, CXCR3+, andCXCR32 innate CD8+ T-cell subsets (Fig. 4). The genesanalyzed represent the range of T-cell functions detectedby themicroarray data.The relative expressions of genes inthe T-cell subsets analyzed by qRT-PCR were consistentwith the gene expression changes observed by genemicroarray. Taken together, our results establish a pheno-typic distinction between CXCR3 expressing and non-expressing subsets of innate CD8+ T cells and definea important functional role for CXCR3+ innate CD8+

T cells in antibacterial and antitumor immune responses.

Ingenuity pathway analysis of differentially regulatedgenes in CXCR3+ and CXCR32 innate CD8+ T cells

To further characterize potential signaling pathways pref-erentially utilized by the CXCR3 expressing subset of in-nate CD8+ T cells which play a role in their function, weperformed IPA (Ingenuity Systems, http://www.ingenuity.com/) on genes that were differentially regulated in thissubset compared to CXCR32 innate and naive CD8+

T cells. Major canonical pathways that were significantly

Figure 3. Gene expression profiling of sorted innate CD8+ T-cell subsets. A) Flow cytometric analysis of lymph node cells from naiveCIBER mice showing CXCR3+ and CXCR32 innate CD8+ T-cell populations. Gating strategy used for sorting CXCR3+ and CXCR32

innate CD8+ T-cell populations are also shown. B) Heat map with gene expression patterns of highly dsyregulated genes in naive,CXCR3+, and CXCR32 innate CD8+ T-cell populations as determined by microarray data analysis. Gene expression variations arerepresented by color. C) Fold induction of significant genes in CXCR3+ and CXCR32 innate CD8+ T-cell populations compared tonaive CD8+ T cells, as determined by microarray. Data were obtained from sorted cells from a pool of 5 mice.

CXCR3 CONFERS IMMUNITY TO INFECTION AND CANCER 5

modulated between CXCR3+ and CXCR32 innate CD8+

T cells are shown in Fig. 5A and include eukaryotic initia-tion factor (eIF) 2 and eIF4 signaling. eIF2 signaling reg-ulates translation initiation in response to stressors such asbacterial infection. Up-regulated eIF2 signaling has beenshown to play a vital role in immunity to intracellular bac-terial infection including L. monocytogenes (16). eIF4 andp70S6K also play critical roles in translational regulation.Molecules associated with the significant canonical path-ways inCXCR3+ vs.CXCR32 innateCD8+T cells identifiedusing IPA are listed in Table 1. Significant pathways in theCXCR3+ innate CD8+ vs. naive CD8+ sample set aredepicted in Fig. 5B, and molecules associated with thesepathways are listed in Table 2.

Gene interaction networks analysis of 162 genes differ-entially expressed in CXCR3+ innate CD8+ T cells relativeto CXCR32 innate CD8+ T cells revealed 2 important sig-naling networks in which most of the genes are up-regulated in CXCR3+ innate CD8+ T cells (Fig. 6). Thesesignaling networks are associated with T-cell activation,proliferation, and cytotoxic effector functions (Fig. 6A), aswell as with RNA posttranscriptional modification, trans-lational control, and protein synthesis (Fig. 6B). In-terestingly, IFN-gR2 andT-cell–specificGTPase (Tgtp1/2),which is regulated by IFN-g, are down-regulated in thissubset, suggesting a diminished response to IFN-g activa-tion compared to CXCR32 innate CD8+ T cells.

DISCUSSION

Our results provide compelling evidence for the pheno-typic and functional heterogeneity of innate CD8+ T-cellsubsets and a role for CXCR3 expression in predictingdistinct effector functions within this diverse population.CXCR3 has canonically been associated with type 1 im-mune responses, drivenby themaster regulator, T-bet, anddefined by IFN-g production, resulting in host protectiveresponses against intracellular pathogens and during a tu-mor challenge (14, 15, 17–19).Our present work, coupledwith previous data from our laboratory (6), strongly sug-gests comparable phenotypic characteristics between theCXCR3 expressing subset of innate CD8+ T cells and Tc1effector cells (15, 20, 21), such as the enhanced expressionof Ly6c, granzyme B, T-bet, and IFN-g. Added supportingevidence for the roleofCXCR3+ innateCD8+Tcells in type1 effector responses was the observed significant down-regulationof IFN-gR2 in thispopulation. Ithasbeen shownthatTh1 cells donot respond to IFN-g stimulation (22, 23),and reduced IFN-gR2 expression is essential formediatingthe effector functions of Th1 cells (24). It is therefore notsurprising that CXCR3+ innate CD8+ T cells contribute totype 1 immunity during the early stages of an infectiouschallenge. However, it is noteworthy that the expression ofCD62L in CXCR3+ innate CD8+ T cells, as well as the rel-ative abundance of this population in naive (and even

Figure 4. qRT-PCR validation of select dysregulated genes among innate CD8+ T-cell subsets. A–F) Relative gene expression of(A) Cxcr3, (B) Ccl5, (C) Klrc1, (D) GP49b, (E) Art2b, and (F) CtsW in sorted CXCR3+ and CXCR32 innate CD8+ T-cell populationsfrom naive CIBER mice as determined by qRT-PCR. Data are presented as fold induction relative to naive CD8+ T cells and asmean 6 SEM of duplicates obtained by pooling samples from 3 or 4 individual mice.

6 Vol. 29 March 2015 OGHUMU ET AL.The FASEB Journal x www.fasebj.org

germ-free) mice (6), distinguish them from Tc1 effectorcells, which usually arise after an infectious challenge.Functionally, this subset of innate CD8+ T cells seems to beinvolved in antibacterial and antitumor effector functionsduring the innate phase of the immune response.

Our rationale for using IL-15 to preactivate innateCD8+ T-cell subsets before our assays was to demonstratethe behavior of these cells in a naturally occurring

inflammatory environment, such as during microbialinfection or neoplastic disease. IL-15 is significantly pro-duced in microbial or tumor microenvironments, whichis necessary for the optimal activity of innate lymphocytesincluding innate CD8+ T cells, NK cells, and NKT cells(12, 25, 26). Interestingly, IL-15 has been shown to berequired for themaintenance of the total innate CD8+ T-cell population (1, 5). In previous studies using innate

TABLE 1. Top canonical pathways and associated molecules modulated between CXCR3+ and CXCR32 innate CD8+ T cells

Ingenuity canonical pathway Molecules

eIF2 signaling RPS2, RPS19, RPL30, RPS21, eIF4A2, RPL37A, RPLP0, RPS12, RPS7,RPL14, RPL8, RPL35, RPS16, RPS9, RPL5, RPS3, RPL13, RPL13A,RPSA, RPL38

Regulation of eIF4 and p70S6K signaling ITGB1, RPS7, RPS16, RPS2, RPS9, RPS19, eIF4A2, RPS21, RPS3, RPS12,RPSA

mTOR signaling RPS7, RPS16, RPS2, RPS9, RPS19, eIF4A2, RPS21, RPS3, RPS12, RPSAT helper cell differentiation IFNGR2, TBX21, IL18R1Mitochondrial dysfunction COX7B, NDUFA3, PSENEN, Atp5eGranulocyte adhesion and diapedesis ITGB1, XCL1, CCL5, IL18RAPOxidative phosphorylation COX7B, NDUFA3, Atp5eHepatic fibrosis/hepatic stellate cell activation IFNGR2, CCL5, FASLG, IL18RAPNK cell signaling Klra4, HCST, KLRC1Eumelanin biosynthesis MIFChondroitin and dermatan biosynthesis CHSY1Thioredoxin pathway TXNCell cycle: G1/S checkpoint regulation PA2G4, RPL5IL-15 signaling FASLG, IL2RB

Figure 5. Top gene networks generated from ingenuity pathway analysis and significantly modulated between innate andconventional naive CD8 T cells. A) Top networks significantly modulated (log P value) between CXCR3+ and CXCR32 innateCD8+ T-cell populations. B) Top networks significantly modulated (log P value) between CXCR3+ innate CD8+ T cells and naiveCD8+ T cells. Molecules associated with these pathways are depicted in Tables 1 and 2.

CXCR3 CONFERS IMMUNITY TO INFECTION AND CANCER 7

CD8+ T cells in an in vivo infection model with L. mono-cytogenes, authors preactivated these cells with IL-15 (12).Further, these studies showed that unstimulated and IL-2stimulated innate CD8+ T cells provided minimal pro-tection against Listeria infection. In keeping with thesestudies, we preactivated our cells with IL-15. The noveltyof our study is that, unlike previous work that adoptivelytransferred total innate CD8+ T cells after preactivationwith IL-15, we reveal a specific subset that is more re-sponsive to this cytokine, which has biologic relevance intumor killing and antibacterial activity.

CD8+ T cells play an important role in the control andelimination of L. monocytogenes and other intracellularpathogens (8, 9, 13, 27–30). Our results point out thatamong the subsets of innate CD8+ cells, those expressingCXCR3 played amajor role in controlling the early growthof L. monocytogenes. Enhanced IFN-g production by thesecells correlated with reduced bacterial burden, suggestingthat the protective ability of CXCR3+ innate CD8+ T cells islargely mediated by IFN-g production, which has beenshown to be important for protection against L. mono-cytogenes (13). Although other cells, such as NK cells, havebeen shown to produce IFN-g early after L. monocytogenesinfection, they are less efficient than CD8+ T cells at re-storing protective innate responses in IFN-g knockoutmice (13). This was shown to be due toNK cell localizationat the red pulp of infected mice, whereas CD8+ cells werefound tobe in theT-cell areaof the spleen, colocalizedwithbacteria andmacrophages (13). As a chemokine receptor,CXCR3 does play a role in the migration of effector lym-phocytes to infected sites (19), and this may further facili-tate protection by CXCR3+ innate CD8+ T cells. However,

the ability to produce larger amounts of IFN-g appears tobe major reason behind the enhanced protection byCXCR3+ innate CD8+ T cells compared to the CXCR32

subset, as observed in our study. It is therefore apparentthat CXCR3 defines an innate CD8+ T-cell phenotype thatmediates a potent antibacterial immune response.

Although the lineage development of CXCR3+ andCXCR32 innateCD8+Tcells areparallel but separate fromthe pathway of conventional T-cell development (1, 31,32), the transcriptional profiles of these subsets of innateCD8+ cells are suggestive of distinct pathways of differen-tiation and effector activity. Evidence for enhanced cyto-lytic activity of CXCR3+ innate CD8+ T-cell populationis seen in enhanced CatW expression, a cysteine proteaseexpressed by NK cells and CTLs that has been shown tobe associated with cytotoxicity of target cells (33, 34).GP49a expression was also enhanced in CXCR3+ innateCD8+ T cells. Although related to the inhibitory receptorGP49b, which down-regulates activation signals that leadto cytotoxicity in T cells and NK cells (35), GP49a lacksan intracellular immunoreceptor tyrosine-based inhibitorymotif (ITIM) and appears to exert an opposite effect byeliciting the activation of these cells (36), evidently con-tributing to their cytotoxicity. Other genes observed tobehighly expressed inCXCR3+ innateCD8+Tcells includeklra4 and klrc1, generally expressed by NK cells, and areassociated with cellular cytotoxicity as well as the chemo-kine ccl5, which is highly expressed by cytotoxic NK andCD8+ T cells in response to IL-15 signaling (37, 38).

Gene expression analyses of sorted innate CD8+ T-cellpopulations show that the CXCR3 expressing subset ismore responsive to cytokine stimulationby IL-15and IL-18,

TABLE 2. Top canonical pathways and associated molecules modulated between CXCR3+ innate CD8+ and naive CD8+ T cells

Ingenuity canonical pathway Molecules

Hepatic fibrosis / hepatic stellate cell activation IGFBP4, IGF1R, IFNGR2, CCL5, STAT1, CCR7, FASLG,TIMP2, IL18RAP

T helper cell differentiation IL6ST, IFNGR2, CXCR5, STAT1, TBX21, IL18R1Rac signaling ITGB1, JUN, CFL2, CD44, PIP4K2A, ITGA4Caveolar-mediated endocytosis signaling ITGB1, ITGAE, HLA-A, ITGB7, ITGA4PTEN signaling ITGB1, TGFBR3, IGF1R, CDKN1B, FASLG, ITGA4Cross-talk between dendritic cells and NK cells HLA-A, CD226, CCR7, FASLG, IL2RBProtein ubiquitination pathway USO1, USP28, DNAJC9, PSMD7, HLA-A, PSMB1,

ANAPC10, NEDD4LRegulation of cellular mechanics by calpain protease ITGB1, CDKN1B, ACTN1, ITGA4Cdc42 signaling ITGB1, JUN, H2-T10, CFL2, HLA-A, ITGA4p38 MAPK signaling ATF1, DUSP10, STAT1, FASLG, IL18RAPCTL-mediated apoptosis of target cells HLA-A, CASP8, FASLGGranulocyte adhesion and diapedesis ITGB1, PECAM1, XCL1, CCL5, ITGA4, IL18RAPILK signaling ITGB1, JUN, CFL2, LEF1, ITGB7, ACTN1Agranulocyte adhesion and diapedesis ITGB1, PECAM1, XCL1, CCL5, ITGB7, ITGA4Leukocyte extravasation signaling ITGB1, CD44, PECAM1, ACTN71, TIMP2, ITGA4iNOS signaling JUN, IFNGR2, STAT1Tec kinase signaling ITGB1, GTF2I, STAT1, FASLG, ITGA4Wnt/b-catenin signaling JUN, TGFBR3, HDAC1, CD44, LEF1RhoGDI signaling ITGB1, CFL2, CD44, PIP4K2A, ITGA4Glioma invasiveness signaling CD44, PLAUR, TIMP2Myc mediated apoptosis signaling IGF1R, CASP8, FASLGDNA methylation and transcriptional repression signaling HDAC1, MBD2Activation of IRF by cytosolic pattern recognition receptors JUN, IFNA4, STAT1Cell cycle: G1/S checkpoint regulation HDAC1, RPL5, CDKN1BTumoricidal function of hepatic NK cells CASP8, FASLG

8 Vol. 29 March 2015 OGHUMU ET AL.The FASEB Journal x www.fasebj.org

which further enhances their proliferative and cytotoxicpotential. These cells display enhanced expression of IL-2Rb, IL-18R1, and IL-18R accessory protein (IL-18RAP),indicating enhanced responsiveness to IL-2, IL-15, and IL-18, cytokines known to potentiate the cytotoxic effects ofNK cells and CTLs (5, 29). Previous work by our groupshowed increased IFN-g production and granzyme B ex-pression upon stimulation by IL-2, IL-15, or IL-12/IL-18(6), which confirms that CXCR3+ innate CD8+ T cells arepreferentially activated by these cytokines. Moreover, thegreater cytolytic activity of IL-15 stimulatedCXCR3+ innateCD8+ T cells against P815 target cells, as revealed in thisstudy, corroborates the enhanced sensitivity of this pop-ulation to cytokine stimulation and the potential for anti-tumor immune responses.

A significant finding in our study using IPA was the in-creased expression of genes involved in translational initi-ation. A significant number of ribosomal proteins thatmake up the 40S and 60S ribosomal subunit as well aseIF4A were up-regulated in the CXCR3+ subset of innateCD8+Tcells. In response tovarious stimuli or stressors suchas oxidative stress or viral infection, translation initiationoccurs through a coordinated process involving ribosomalproteins, modification enzymes, and ribosome-associatedtranslation factors (16). This process appears to be en-hanced inCXCR3+ innate CD8+T cells, which is consistentwith theirobservedpreactivated state.Biologic triggers that

induce this activated state in the absence of infectiousstimuli and factors responsible for the maintenance ofthis population of homeostatically active cells are stillnot clearly understood.

In conclusion, we demonstrated that CXCR3 expressinginnate CD8+ T cells display enhanced tumor cytotoxicityand confer protection against L. monocytogenes infection.Our grouphas previously shown thatCXCR3 is essential forimmunity against the intracellular parasiteLeishmaniamajor(19). Furthermore, the importanceofCXCR3 inantitumorresponses was well characterized in a recent breast cancertumor model (18). Consistent with these findings, ourresults suggest that CXCR3 expression in innate CD8+

T cells significantly amplifies its cytotoxic potential andprotective immune ability characterized by the productionof IFN-g. Immunotherapeutic approaches to infectious dis-ease and cancer management that use adoptive transfer ofeffector cells (14) could utilize CXCR3+ innate CD8+ T-cellpopulations as novel clinical intervention strategies.

This work was supported by the U.S. National Institutes ofHealth (NIH) (Grants R03-AI090231, RC4-AI092624, R34-AI100789, R21-AT004160, and R03-CA164399) (to A.R.S.),NIH National Institute of Dental and Craniofacial Research(Training Grant T32DE014320) (to S.O.), and NationalCouncil of Science and Technology, Mexico (CONACYT)(to C.T.).

Figure 6. Ingenuity pathway analysis of differentially regulated genes between CXCR3+ innate, CXCR32 innate, and naive CD8+

T cells subsets. A) Functional pathway analysis of immune related gene interactions associated with cellular activation,differentiation, and cytotoxicity between innate CD8+ T-cell subsets. B) Functional pathway analysis of genes associated with 40Sand 60S ribosomal subunits as well as others involved in translation initiation between innate CD8+ T-cell subsets. Genes in redare significantly up-regulated and genes in green are significantly down-regulated in CXCR3+ compared to CXCR32 innate CD8+

T cells.

CXCR3 CONFERS IMMUNITY TO INFECTION AND CANCER 9

REFERENCES

1. Dubois, S., Waldmann, T. A., and Muller, J. R. (2006) ITK and IL-15 support two distinct subsets of CD8+ T cells. Proc. Natl. Acad.Sci. USA 103, 12075–12080

2. Berg, L. J. (2007) Signalling through TEC kinases regulatesconventional versus innate CD8(+) T-cell development. Nat. Rev.Immunol. 7, 479–485

3. Sintes, J., Cuenca, M., Romero, X., Bastos, R., Terhorst, C., Angulo,A., and Engel, P. (2013) Cutting edge: Ly9 (CD229), a SLAM familyreceptor, negatively regulates the development of thymic innatememory-like CD8+ T and invariant NKT cells. J. Immunol. 190, 21–26

4. Verykokakis, M., Boos, M. D., Bendelac, A., and Kee, B. L. (2010)SAP protein-dependent natural killer T-like cells regulate thedevelopment of CD8(+) T cells with innate lymphocyte charac-teristics. Immunity 33, 203–215

5. Itsumi, M., Yoshikai, Y., and Yamada, H. (2009) IL-15 is criticalfor the maintenance and innate functions of self-specific CD8(+)T cells. Eur. J. Immunol. 39, 1784–1793

6. Oghumu, S., Dong, R., Varikuti, S., Shawler, T., Kampfrath, T.,Terrazas, C. A., Lezama-Davila, C., Ahmer, B. M., Whitacre, C. C.,Rajagopalan, S., Locksley, R., Sharpe, A. H., and Satoskar, A. R.(2013) Distinct populations of innate CD8+ T cells revealed ina CXCR3 reporter mouse. J. Immunol. 190, 2229–2240

7. Cho, H., Choi, H.-J., Xu, H., Felio, K., and Wang, C.-R. (2011)Nonconventional CD8+ T cell responses to Listeria infection inmice lacking MHC class Ia and H2-M3. J. Immunol. 186, 489–498

8. D’Orazio, S. E., Halme, D. G., Ploegh, H. L., and Starnbach, M. N.(2003) Class Ia MHC-deficient BALB/c mice generate CD8+ T cell-mediated protective immunity against Listeria monocytogenes in-fection. J. Immunol. 171, 291–298

9. Seaman, M. S., Wang, C. R., and Forman, J. (2000) MHC class Ib-restricted CTL provide protection against primary and secondaryListeria monocytogenes infection. J. Immunol. 165, 5192–5201

10. Xu, H., Chun, T., Choi, H. J., Wang, B., and Wang, C. R. (2006)Impaired response to Listeria in H2-M3-deficient mice revealsa nonredundant role of MHC class Ib–specific T cells in hostdefense. J. Exp. Med. 203, 449–459

11. Irizarry, R. A., Hobbs, B., Collin, F., Beazer-Barclay, Y. D.,Antonellis, K. J., Scherf, U., and Speed, T. P. (2003) Exploration,normalization, and summaries of high density oligonucleotidearray probe level data. Biostatistics 4, 249–264

12. Dhanji, S., Chow, M. T., and Teh, H. S. (2006) Self-antigenmaintains the innate antibacterial function of self-specific CD8T cells in vivo. J. Immunol. 177, 138–146

13. Berg, R. E., Crossley, E., Murray, S., and Forman, J. (2005)Relative contributions of NK and CD8 T cells to IFN-gammamediated innate immune protection against Listeria mono-cytogenes. J. Immunol. 175, 1751–1757

14. Grange, M., Buferne, M., Verdeil, G., Leserman, L., Schmitt-Verhulst,A.-M., and Auphan-Anezin, N. (2012) Activated STAT5 promoteslong-lived cytotoxic CD8+ T cells that induce regression of autoch-thonous melanoma. Cancer Res. 72, 76–87

15. Grange, M., Verdeil, G., Arnoux, F., Griffon, A., Spicuglia, S., Maurizio,J., Buferne, M., Schmitt-Verhulst, A.-M., and Auphan-Anezin, N.(2013) Active STAT5 regulates T-bet and eomesodermin expressionin CD8 T cells and imprints a T-bet-dependent Tc1 program withrepressed IL-6/TGF-b1 signaling. J. Immunol. 191, 3712–3724

16. Shrestha, N., Bahnan, W., Wiley, D. J., Barber, G., Fields, K. A.,and Schesser, K. (2012) Eukaryotic initiation factor 2 (eIF2)signaling regulates proinflammatory cytokine expression andbacterial invasion. J. Biol. Chem. 287, 28738–28744

17. Shrikant, P. A., Rao, R., Li, Q., Kesterson, J., Eppolito, C., Mischo,A., and Singhal, P. (2010) Regulating functional cell fates in CD8T cells. Immunol. Res. 46, 12–22

18. Oghumu, S., Varikuti, S., Terrazas, C., Kotov, D., Nasser, M. W.,Powell, C. A., Ganju, R. K., and Satoskar, A. R. (2014) CXCR3deficiency enhances tumor progression by promotingmacrophageM2polarization in amurine breast cancer model. Immunology 143, 109–119

19. Rosas, L. E., Barbi, J., Lu, B., Fujiwara, Y., Gerard, C., Sanders,V. M., and Satoskar, A. R. (2005) CXCR32/2 mice mount anefficient Th1 response but fail to control Leishmania majorinfection. Eur. J. Immunol. 35, 515–523

20. Cerwenka, A., Morgan, T. M., Harmsen, A. G., and Dutton, R. W.(1999) Migration kinetics and final destination of type 1 and type

2 CD8 effector cells predict protection against pulmonary virusinfection. J. Exp. Med. 189, 423–434

21. Woodland, D. L., and Dutton, R. W. (2003) Heterogeneity ofCD4(+) and CD8(+) T cells. Curr. Opin. Immunol. 15, 336–342

22. Bach, E. A., Szabo, S. J., Dighe, A. S., Ashkenazi, A., Aguet, M.,Murphy, K. M., and Schreiber, R. D. (1995) Ligand-inducedautoregulation of IFN-gamma receptor beta chain expression inT helper cell subsets. Science 270, 1215–1218

23. Pernis, A., Gupta, S., Gollob, K. J., Garfein, E., Coffman, R. L.,Schindler, C., and Rothman, P. (1995) Lack of interferongamma receptor beta chain and the prevention of interferongamma signaling in TH1 cells. Science 269, 245–247

24. Tau, G. Z., von der Weid, T., Lu, B., Cowan, S., Kvatyuk, M.,Pernis, A., Cattoretti, G., Braunstein, N. S., Coffman, R. L., andRothman, P. B. (2000) Interferon g signaling alters the functionof T helper type 1 cells. J. Exp. Med. 192, 977–986

25. Mitani, A., Nishimura, H., Hirose, K., Washizu, J., Kimura, Y.,Tanaka, S., Yamamoto, G., Noguchi, T., and Yoshikai, Y. (1999)Interleukin-15 production at the early stage after oral infectionwith Listeria monocytogenes in mice. Immunology 97, 92–99

26. Kennedy, M. K., Glaccum, M., Brown, S. N., Butz, E. A., Viney, J. L.,Embers, M., Matsuki, N., Charrier, K., Sedger, L., Willis, C. R., Brasel,K., Morrissey, P. J., Stocking, K., Schuh, J. C., Joyce, S., and Peschon,J. J. (2000) Reversible defects in natural killer andmemory CD8 T celllineages in interleukin 15-deficient mice. J. Exp. Med. 191, 771–780

27. Berg, R. E., Crossley, E., Murray, S., and Forman, J. (2003)Memory CD8+ T cells provide innate immune protection againstListeria monocytogenes in the absence of cognate antigen. J. Exp.Med. 198, 1583–1593

28. D’Orazio, S. E., Troese, M. J., and Starnbach, M. N. (2006)Cytosolic localization of Listeria monocytogenes triggers an earlyIFN-gamma response by CD8+ T cells that correlates with innateresistance to infection. J. Immunol. 177, 7146–7154

29. Hu, J., Sahu, N., Walsh, E., and August, A. (2007) Memoryphenotype CD8+ T cells with innate function selectively developin the absence of active Itk. Eur. J. Immunol. 37, 2892–2899

30. Thompson, L. J., Kolumam, G. A., Thomas, S., and Murali-Krishna,K. (2006) Innate inflammatory signals induced by variouspathogens differentially dictate the IFN-I dependence of CD8T cells for clonal expansion and memory formation. J. Immunol.177, 1746–1754

31. Broussard, C., Fleischacker, C., Horai, R., Chetana, M., Venegas,A. M., Sharp, L. L., Hedrick, S. M., Fowlkes, B. J., and Schwartzberg,P. L. (2006) Altered development of CD8+ T cell lineages in micedeficient for the Tec kinases Itk and Rlk. Immunity 25, 93–104

32. Huang, W., Hu, J., and August, A. (2013) Cutting edge: innatememory CD8+ T cells are distinct from homeostatic expandedCD8+ T cells and rapidly respond to primary antigenic stimuli.J. Immunol. 190, 2490–2494

33. Stoeckle, C., Gouttefangeas, C., Hammer, M., Weber, E., Melms,A., and Tolosa, E. (2009) Cathepsin W expressed exclusively inCD8+ T cells and NK cells, is secreted during target cell killingbut is not essential for cytotoxicity in human CTLs. Exp. Hematol.37, 266–275

34. Wex, T., Wex, H., Hartig, R., Wilhelmsen, S., and Malfertheiner,P. (2003) Functional involvement of cathepsin W in the cytotoxicactivity of NK-92 cells. FEBS Lett. 552, 115–119

35. Katz, H. R., Vivier, E., Castells, M. C., McCormick, M. J., Chambers,J. M., and Austen, K. F. (1996) Mouse mast cell gp49B1 contains twoimmunoreceptor tyrosine-based inhibition motifs and suppressesmast cell activation when coligated with the high-affinity Fc receptorfor IgE. Proc. Natl. Acad. Sci. USA 93, 10809–10814

36. Lee, K. H., Ono, M., Inui, M., Yuasa, T., and Takai, T. (2000)Stimulatory function of gp49A, a murine Ig-like receptor, in ratbasophilic leukemia cells. J. Immunol. 165, 4970–4977

37. Chenoweth, M. J., Mian, M. F., Barra, N. G., Alain, T., Sonenberg,N., Bramson, J., Lichty, B. D., Richards, C. D., Ma, A., and Ashkar,A. A. (2012) IL-15 can signal via IL-15Ra, JNK, and NF-kB to driveRANTES production by myeloid cells. J. Immunol. 188, 4149–4157

38. Perera, L. P., Goldman, C. K., and Waldmann, T. A. (1999) IL-15induces the expression of chemokines and their receptors inT lymphocytes. J. Immunol. 162, 2606–2612

Received for publication September 24, 2014.Accepted for publication October 28, 2014.

10 Vol. 29 March 2015 OGHUMU ET AL.The FASEB Journal x www.fasebj.org