of June 28, 2018.This information is current as

Nonobese Diabetic Mice Regulatory T Cells in Islets of+of Foxp3

ICOS-Dependent Homeostasis and Function

Mara Kornete, Evridiki Sgouroudis and Ciriaco A. Piccirillo

ol.1101303http://www.jimmunol.org/content/early/2012/01/06/jimmun

published online 6 January 2012J Immunol 

        average*  

4 weeks from acceptance to publicationFast Publication! •    

Every submission reviewed by practicing scientistsNo Triage! •    

from submission to initial decisionRapid Reviews! 30 days* •    

Submit online. ?The JIWhy

Subscriptionhttp://jimmunol.org/subscription

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/About/Publications/JI/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/alertsReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists, Inc. All rights reserved.Copyright © 2012 by The American Association of1451 Rockville Pike, Suite 650, Rockville, MD 20852The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

by guest on June 28, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

The Journal of Immunology

ICOS-Dependent Homeostasis and Function of Foxp3+

Regulatory T Cells in Islets of Nonobese Diabetic Mice

Mara Kornete,*,1 Evridiki Sgouroudis,*,1 and Ciriaco A. Piccirillo*,†

A progressive waning in Foxp3+ regulatory T cell (Treg) functions is thought to provoke autoimmunity in the NOD model of type 1

diabetes (T1D). A deficiency in IL-2 is one of the main triggers for the defective function of Tregs in islets. Notably, abrogation of

the ICOS pathway in NOD neonates or BDC2.5-NOD (BDC2.5) mice exacerbates T1D, suggesting an important role for this

costimulatory pathway in tolerance to islet Ags. Thus, we hypothesize that ICOS selectively promotes Foxp3+ Treg functions in

BDC2.5 mice. We show that ICOS expression discriminates effector Foxp32 T cells from Foxp3+ Tregs and specifically designates

a dominant subset of intra-islet Tregs, endowed with an increased potential to expand, secrete IL-10, and mediate suppressive

activity in vitro and in vivo. Consistently, Ab-mediated blockade or genetic deficiency of ICOS selectively abrogates Treg-mediated

functions and T1D protection and exacerbates disease in BDC2.5 mice. Moreover, T1D progression in BDC2.5 mice is associated

with a decline in ICOS expression in and expansion and suppression by intra-islet Foxp3+ Tregs. We further show that the ICOS+

Tregs, in contrast to their ICOS2 counterparts, are more sensitive to IL-2, a critical signal for their survival and functional

stability. Lastly, the temporal loss in ICOS+ Tregs is readily corrected by IL-2 therapy or protective Il2 gene variation. Overall,

ICOS is critical for the homeostasis and functional stability of Foxp3+ Tregs in prediabetic islets and maintenance of T1D

protection. The Journal of Immunology, 2012, 188: 000–000.

Type 1 diabetes (T1D) is a chronic autoimmune diseaseresulting from a T cell-dependent destruction of theinsulin-producing b-islets of Langerhans (1, 2). The NOD

mouse model develops spontaneous T1D and shares many featureswith human T1D, such as development of autoantibodies andhyperglycemia (1). NOD mice exhibit profound dysregulated im-mune responses, and a progressive loss in immunoregulatorymechanisms is thought to underlie the pathogenesis of T1D (3).Regulatory T cells (Tregs), which constitutively express CD4,Foxp3, and CD25 (4–6), are a major mechanism of peripheraltolerance and have been implicated as a central point in T1Dprogression, as depletion of CD25-expressing cells or genetic ab-lation of Foxp3 results in accelerated T1D (7, 8).Studies point to a progressive waning in Treg functions as a

trigger of T1D onset and progression (9–12). Recently, we showedthat T1D progression is associated with a temporal loss in thespecific capacity of Foxp3+ Tregs to expand/survive in b-islets,which in turn perturbs the Treg/effector T cell (Teff) balance and

unleashes anti-islet immune responses (11, 13). Moreover, a defi-ciency in IL-2, a cytokine essential for the function and fitness ofTregs within islets, was shown to trigger this defective function ofTregs, in turn provoking a Treg/Teff imbalance in islets (13, 14).These results are also reminiscent of earlier studies showing thatT cells from prediabetic NOD mice become hypoproliferative andpoor IL-2 producers at the onset of insulitis, a time point coin-ciding with the waning of Treg functions (15). Consistently, pro-phylactic treatment with low doses of IL-2 in NOD mice canprevent and even reverse established disease in NOD mice by re-storing the survival of intra-islet Tregs (14).Foxp3+ Treg function in NOD mice is costimulation dependent

as disruption of CD28–B7 pathway abrogates Treg developmentand homeostasis and leads to acceleration of T1D in NOD mice (16,17). The contribution of other costimulatory pathways is currentlynot well understood. Notably, ICOS is a CD28 superfamily-relatedmolecule, which plays an important role in T cell activation/survival(18). ICOS enhances IL-4 and IL-10 secretion in Th2 cell responses(18, 19), and blockade of the ICOS–ICOS-ligand pathway duringthe induction of Th1-driven experimental autoimmune encephalo-myelitis exacerbated disease by enhancing IFN-g production (20).Notably, ICOS mRNA is abundantly expressed in pancreatic lymphnodes (pLNs) of T1D-protected BDC2.5-NOD (BDC2.5) TCRtransgenic mice (21). In contrast to T1D-resistant ICOS 2/2 NODmice, ICOS blockade in BDC2.5 mice and NOD neonates or ICOSdeficiency in BDC2.5 mice exacerbated T1D. However, in the re-ported studies, the expression or function of ICOS was solely ex-amined on CD25-expressing (CD25+CD692 T cells), not purifiedFoxp3+, Tregs enriched from inflamed pancreatic sites (21–23).Notably, IL-2 therapy augments Treg numbers in the pancreas

and induces the expression of various Treg-associated proteins suchas Foxp3, CD25, Bcl-2, and ICOS (24). IL-2 is known to enhanceICOS expression on activated T cells, suggesting that a positivefeedback loop exists between IL-2 and ICOS signaling pathways(25). Collectively, these lines of evidence suggest that ICOS mayselectively control immunoregulatory mechanisms of T1D, in-cluding Foxp3+ Treg functions.

*Department of Microbiology and Immunology, McGill University, Montreal, Quebec,Canada H3A 2B4; and †Federation of Clinical Immunology Societies Center of Ex-cellence, Research Institute of the McGill University Health Centre, Montreal, Quebec,Canada H3G 1A4

1M.K. and E.S. contributed equally to this work.

Received for publication May 4, 2011. Accepted for publication November 30, 2011.

This work was supported by grants from the Canadian Institutes for Health Research(MOP67211). C.A.P. holds the Canada Research Chair. M.K. and E.S. have receivedfellowships from the Canadian Institutes for Health Research Training Grant inNeuroinflammation. M.K. has received fellowships from the McGill UniversityHealth Centre Research Institute and Fonds de la recherche en sante du Quebec.

Address correspondence and reprint requests to Dr. Ciriaco A. Piccirillo, ResearchInstitute of the McGill University Health Centre, Montreal General Hospital, 1650Cedar Avenue, Room L11.132, Montreal, QC, Canada H3G 1A4. E-mail address:ciro.piccirillo@mcgill.ca

Abbreviations used in this article: BDC2.5, BDC2.5-NOD; BMDC, bone marrow-derived dendritic cell; LN, lymph node; MFI, mean fluorescence intensity; PI, propi-dium iodide; pLN, pancreatic lymph node; T1D, type 1 diabetes; Teff, effector T cell;Treg, regulatory T cell; WT, wild-type.

Copyright� 2012 by The American Association of Immunologists, Inc. 0022-1767/12/$16.00

www.jimmunol.org/cgi/doi/10.4049/jimmunol.1101303

Published January 6, 2012, doi:10.4049/jimmunol.1101303 by guest on June 28, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

In this study, we hypothesized that ICOS represents a criticalfactor in the stabilization of Foxp3+ Treg functions in the NODmouse model of T1D. We identify ICOS to be preferentially ex-pressed by a subset of Foxp3+ Tregs, in contrast to Foxp32 Teffs,in islets of prediabetic BDC2.5 mice. We also demonstrate thatT1D is associated with a progressive loss in ICOS expression andexpansion of Foxp3+ Tregs in islets of prediabetic mice. ThisICOS+ Treg subset, in contrast to its ICOS2 counterpart, is en-dowed with prominent proliferative and suppressive capacities insitu and readily secretes IL-10 upon islet-Ag stimulation. More-over, icos deficiency or Ab blockade in BDC2.5 mice dampensthe competitive fitness of Tregs in islets and cripples their capacityto protect against T1D. We also show that IL-2 therapy or pro-tective Il2 gene variation readily restores the frequency of ICOS+

Tregs in islets. Strikingly, IL-2 preferentially activates the STAT5pathway in the ICOS+ Treg subset, in contrast to its ICOS2

counterpart, in turn favoring ICOS+Treg cell survival and func-tion. Overall, ICOS costimulation sustains IL-2–driven stability ofFoxp3+ Treg functions in prediabetic islets and maintains pro-tection from T1D.

Materials and MethodsMice

NOD and NOD ICOS2/2 mice were obtained from The Jackson Labora-tory (Bar Harbor, ME). NOD.B6 Idd3 congenic mice were obtained fromTaconic Farms. NOD.TCRa2/2 and BDC2.5 and Thy1.1 BDC2.5 CD4+

transgenic mice were a generous gift from C. Benoist (Harvard Univer-sity). NOD.Foxp3GFP mice were generated by back-crossing them toB6.Foxp3GFP reporter mice for 12 generations (26). BDC.Idd3, BDC2.5ICOS2/2, and Thy1.1 BDC2.5.Foxp3GFP mice were generated in-house.All mice strains were maintained in specific pathogen-free conditions atMcGill University and McGill University Health Centre.

Abs and flow cytometry

Stainings for surface phenotyping were done with the followingfluorochrome-conjugated or biotinylated mAbs: anti-CD3 (145-2c11), anti-CD4 (RM5), anti-CD25 (PC61), anti-Vb4 (CTVB4), anti-ICOS (7E.17G9),and anti-Thy1.1 (HIS51). The expression of Foxp3 (FJK-16s), annexin V,propidium iodide (PI) (eBioscience, San Diego, CA), Ki-67 (B56), p-STAT5 (pY694), and Bcl2 (3F11) (BD Bioscience, Mississauga, ON,Canada) was determined by intracellular staining performed according tothe manufacturer’s protocol. To determine the cytokine production, T cellswere restimulated for 4 h at 37˚C with PMA (20 ng/ml), ionomycin (1 nM)(Sigma-Aldrich, Oakville, ON, Canada), and BD GolgiStop (1:1000 di-lution), and then stained intracellularly with anti–IL-10 (JES5-16E3), anti–IFN-g (XMG1.2), and anti–IL-2 (JES6-5H4). In some instances, T cellswere stimulated in an Ag-specific manner with NOD bone marrow-derived dendritic cells (BMDCs) (4:1 cell ratio) and BDC mimetope(RVRPLWVRME) (100 ng/ml) (Sigma-Aldrich). Data were acquiredon a FACSCalibur or FACSCanto (BD Biosciences) and analyzed withFlowJo software (Tree Star).

Cell purification

Various CD4+ T cell subsets were purified from lymph node (LN) or spleenusing the autoMACS Cell Sorter (Miltenyi Biotech, San Diego, CA) (withpurity ranging from 85 to 95%) or FACSAria II Cell Sorter (BD Bio-sciences) (with a purity.98%) as described previously in Ref. 27. In somecases, T cells were CFSE labeled (5 mM for in vitro and 10 mM for in vivo)(Invitrogen, Burlington, ON, Canada), and expansion of donor T cells wasevaluated by CFSE dilution. Purified T cells were either used for variousin vitro assays or were transferred i.v. into various recipient mice.

In vitro T cell functional assays

Proliferation assays were performed by cultivating CD4+ T cells (5 3 104)in 96-well flat-bottom microtiter plates with irradiated spleen cells (2 3105) and BDC2.5 mimetope (10–100 ng/ml) in the presence or absence ofIL-2 (25–100 IU/ml) for 72 h at 37˚C in 5% CO2. Suppression assays wereperformed by cultivating FACS-sorted CD4+CD25– Teffs or CD4+ Foxp3.GFP2 Teffs (53 104) with titrated numbers of FACS-purified CD4+CD25+

or CD4+Foxp3.GFP+ Tregs, irradiated APCs (2 3 105), and BDC2.5mimetope (2 ng/ml) for 72 h. Teff proliferation was assessed by CFSE di-

lution or cell cultures were pulsed with 1 mCi tritiated thymidine for thelast 12–16 h of culture. To perform apoptosis and STAT5 phosphorylationassays, FACS-sorted Tregs or Teffs (1 3 105) were cultivated with APCs(4 3 105) and BDC2.5 mimetope (10 ng/ml) in the presence or absence ofIL-2 (25–100 u/ml) for 36 and/or 72 h of culture.

In vivo IL-2 treatment

Mice were treated daily with i.p. injections of 25,000 IU recombinant hu-man IL-2 (gift from the Surgery Branch, National Cancer Institute) for 5consecutive days as described in Ref. 14. Control mice were treated withPBS. The experiment was terminated 2 d after the last injection.

Anti-ICOS mAb administration

Anti-ICOS mAb (7E.17G9.G; Bio X Cell, West Lebanon, NH) was ad-ministered i.p. either at low (50 mg) or high (100 mg) dose on days 0, 3, and5 after T cell transfer in NOD.TCRa2/2 recipients.

Diagnosis of diabetes

Blood glycemia levels were determined every 2–3 d with Hemoglukotestkits (Roche Diagnostics, Montreal, QC, Canada), and T1D was diagnosedat values .300 mg/dl.

Statistical analysis

Results are expressed as means 6 SD. Analyses were performed witha Student t test, except for diabetes incidence, where the Kaplan–Meiersurvival test was used. Values of p , 0.05 were considered significant.

ResultsICOS expression designates a dominant Foxp3+ Treg subset inprediabetic islets

ICOS blockade in NOD mice results in T1D exacerbation, sug-gesting that ICOS may play a role in Treg-mediated self-tolerance(21). We first sought to characterize ICOS expression within theFoxp3+ and Foxp32 T cell subsets in nondraining LN (data notshown), draining pLN, and pancreas at the early stage of insulitisin 3- to 4-wk-old BDC2.5 mice, whose TCR is specific fora b-islet–specific autoantigen. We observed a substantial increasein ICOS expression specifically on intrapancreatic Foxp3+ Tregs(Fig. 1A, top row). The frequency of Foxp3+ICOS+ Tregs amongCD4+ T cells was significantly higher in the pancreas comparedwith that in draining pLN (8.8 6 1.8% versus 1.5 6 0.7%, p #

0.01) (Fig. 1A, bottom row, left panel) and represents a dominantfraction of total Foxp3+ Treg in the pancreas compared with thatin draining pLN (57 6 10.2% and 18.9 6 4.5%, p # 0.002) (Fig.1A, bottom row, middle panel). Similarly, the mean fluorescenceintensity (MFI) of ICOS was also significantly greater on Tregs inthe pancreas relative to the pLN (144.36 11.9 versus 38.56 15.9,p # 0.001) (Fig. 1A, bottom row, right panel). Only a smallproportion of ICOS+ Teffs in pLN and pancreas was detectedamong the CD4+ T cell pool (0.96 6 0.4% versus 5.3 6 1.5%)(Fig. 1A, bottom row, left panel) and particularly among Foxp32

T cells (1 6 0.4% versus 6 6 1.6%) (Fig. 1A, bottom row, middlepanel), suggesting that ICOS costimulation may not be requiredfor the optimal function of autoreactive CD4+ T cells.

The striking difference in ICOS expression on intrapancreaticTregs relative to their counterparts in pLN suggested that Tregsacquire distinctive functional and phenotypic properties in situ. InpLN, approximately half of the Foxp3+ Treg pool expressed CD25(49.7 6 4.3%), and only a small proportion of Foxp3+ Tregscoexpressed both CD25 and ICOS (7.56 0.8%) (Fig. 1B). In starkcontrast, a significant frequency of Foxp3+ Tregs coexpressedCD25 and ICOS in the pancreas of BDC2.5 mice (53.8 6 5.2%versus 7.5 6 0.8%, p # 0.0007) (Fig. 1B). We also examined howICOS expression is related to the proliferation of the Foxp3+ Tregpool within the pancreatic lesion. A significant proportion of cy-cling Tregs (based on Ki-67 expression) expressed ICOS in thepancreas relative to pLN (24.4 6 4.5% versus 9.2 6 3.3%, p #

2 ICOS PROMOTES Foxp3+ Treg CELL FITNESS IN T1D

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

0.001) (Fig. 1C, right panel). Notably, the ICOS+ subset exhibiteda much greater proliferative potential compared with the ICOS2

subset of Foxp3+ Tregs (53.1 6 5.7% versus 16.9 6 4.2%, p #0.001) (Fig. 1C, left panel). Remarkably, pancreatic ICOS+ andICOS2 Foxp3+ Treg subsets did not differ in their activation statusas both fractions were found within the CD62LlowCD44high Teffmemory pool, irrespective of their proliferative potential (data notshown), suggesting that ICOS+Foxp3+ Tregs represent tissue-resident memory Tregs in islets. Thus, ICOS expression discrim-inates effector Foxp32 Teffs from Foxp3+ Tregs and specificallydesignates a dominant subset of intra-islet Tregs, endowed with anincreased potential to expand in islets of prediabetic BDC2.5 mice.

ICOS costimulation promotes the competitive fitness of Tregsin vivo

To determine whether ICOS expression conferred a competitiveadvantage to Ag-specific Foxp3+ Tregs in vivo, we examinedFoxp3+ Treg responses in ICOS-deficient (ICOS2/2) BDC2.5mice compared to BDC2.5 (WT) mice. To this end, T cell-deficient NOD.TCRa2/2 mice were adoptively transferred withCD4+ T cells from congenic (Thy1.1+) WT or (Thy1.2+) ICOS2/2

BDC2.5 mice, and the relative occupancy of transferred T cells inpancreatic sites was evaluated 8 d after adoptive transfer. Our

results show an increase in the proportion of WT (Thy1.1+)Foxp3+ Tregs compared with ICOS2/2 (Thy1.2+) Tregs indraining LN (14.0 6 0.8 versus 7.4 6 0.7, p # 0.007) and thepancreas (25.9 6 2.3 versus 10.7 6 2.2, p # 0.0032) (Fig. 2A, leftpanel). The proportion of Foxp32 T cells was slightly increasedfrom ICOS2/2 (Thy1.2+) donors (data not shown). We then ex-amined whether the decreased frequency of ICOS2/2 Foxp3+

Tregs resulted from a reduced cycling in situ. Our analysis of Ki-67 expression showed that ICOS2/2 (Thy1.2+) Tregs expandedless efficiently compared with WT (Thy1.1+) Foxp3+ Tregs indraining LN (65.7 6 2.8 versus 78.0 6 1.2, p # 0.007) (Fig. 2A,right panel) and particularly in the pancreas (77.7 6 4.2 versus95.9 6 1.2, p # 0.0061) (Fig. 2A, right panel). This decrease inICOS2/2 (Thy1.2+) Treg expansion also coincided with a reduc-tion in CD25 expression in ICOS2/2 Tregs compared with WTTregs in pancreas (73.9 6 2.4 versus 49.2 6 4.0, p # 0.0001),suggesting a possible IL-2–ICOS cross-talk in Treg homeostasis inislets (data not shown).We then examined the competitive fitness of Foxp3+ Tregs from

WT or ICOS2/2 BDC2.5 mice in non-lymphopenic hosts. To thisend, we performed a crisscross adoptive transfer whereby WT(Thy1.1+) or ICOS2/2 (Thy1.2+) CD4+ T cells were transferredin either WT (Thy1.1+) or ICOS2/2 (Thy1.2+) recipient mice, and

FIGURE 1. ICOS expression designates a dominant

Foxp3+ Treg subset in prediabetic islets. Cell suspen-

sions of draining pLN and pancreas of 3- to 4-wk old

BDC2.5 mice were prepared and examined for (A)

ICOS frequency and MFI, (B) CD25 expression, and

(C) proportion of cycling cells (based on Ki-67 ex-

pression) within the ICOS+ and ICOS– subsets of

Foxp3+/2 CD4+ T cells and analyzed by FACS. Data

are representative of at least three separate experiments

(n = three mice/group). Results represent the mean 6SD. **p # 0.02, ***p # 0.002.

The Journal of Immunology 3

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

the expansion and cellular frequency of transferred T cells inpancreatic sites were assessed. In either setting, WT (Thy1.1+)Tregs (of donor or host origin) expanded and accumulated in pLN(58.0 6 0.85 versus 15.85 6 1.95, p # 0.0025) (Fig. 2B, leftpanel) or pancreas (66.85 6 13.85 versus 34.4 6 2.0, p # 0.057)(Fig. 2B, left panel) more efficiently compared with ICOS2/2

(Thy1.2+) Tregs (Fig. 2B, left and right panels).

As an alternate approach to assess the proliferation of WT andICOS2/2 T cell subsets, CFSE-labeled CD4+ T cells from WT orICOS2/2 BDC2.5 mice were adoptively transferred into NODrecipients, and the extent of proliferation was determined indraining LN. The proportion of ICOS2/2 Foxp3+ Treg expansionwas significantly lower relative to WT Tregs in pLN (65.1 6 3.6versus 81.4 6 0.35, p # 0.0126) (Fig. 2C, upper panel). Nosignificant difference was observed between Teffs from eithergenotype, indicating that Teff priming and expansion is not im-paired in the absence of ICOS in vivo (Fig. 2C, lower panel).Overall, our results show that ICOS+ Tregs have a greater pro-liferative capacity than Teffs and that a decreased expansion po-tential underlies the defective competitiveness of Ag-specificICOS2/2 Tregs in inflamed sites.

ICOS-deficient Foxp3+ Tregs are less suppressive in vitro thanWT Tregs

We then asked whether ICOS affected the functional potency ofTregs. To address this question, we first characterized the functionof Tregs with differential ICOS expression levels by suppres-sion assays in vitro. CD4+CD25+ICOS+ Tregs were more potentthan CD4+CD25+ICOS2 Tregs at suppressing CD4+ Teffs at allratios examined (Fig. 3A, right and left panels). To ensure thatCD25+ICOS+ Tregs were not contaminated with activated Teffsexpressing these markers, ICOS+Foxp3+ Tregs from BDC2.5.Foxp3GFP reporter mice were assessed for their ability to sup-press the proliferation of Foxp32 (GFP2) Teffs in vitro. ICOS-expressing Foxp3+ (GFP+) Tregs exhibited a greater suppressivecapacity relative to their ICOS2 counterparts (Fig. 3B), consistentwith the increased ability of WT BDC2.5 Tregs, compared withICOS2/2 Tregs, to suppress the proliferation of WT or ICOS2/2

Teffs in vitro (Fig. 3C). Blocking anti-ICOS Abs did not abolishsuppression, suggesting that ICOS expression is not involved inthe effector mechanism of suppression in vitro (data not shown).Moreover, whereas ICOS2 Tregs upregulate ICOS upon TCR stim-ulation in vitro, they do not suppress Teffs as efficiently as ex vivoICOS+ (data not shown). This suggests that ICOS may have animportant role in imprinting this potential in Tregs during theirdevelopment, and possibly designates a unique Foxp3+ Treg subsetoperating in the periphery, similar to an observation made by Ito etal. (28), in human studies. Thus, ICOS-expressing Foxp3+ Tregs areendowed with a more suppressive phenotype compared with theirICOS2 subset and indicate a role for ICOS in potentiating Tregsuppressive functions.

ICOS-deficient Foxp3+ Tregs are unable to inhibit T1Dinduction in vivo

We then assessed how WT and ICOS2/2 Foxp3+ Tregs comparedin their capacity to inhibit pathogenic T cells during the patho-genesis of T1D. WT or ICOS2/2 BDC2.5 CD4+ Teffs weretransferred into NOD.TCRa2/2 hosts either alone or with WTor ICOS2/2 BDC2.5 Tregs (1:8 Treg/Teff ratio), and the onsetof T1D was monitored. Recipient mice transferred with WTor ICOS2/2 Teffs alone developed T1D with a similar onset (11–12 d posttransfer) and similar incidence (80%) indicating thatabsence of ICOS costimulation is not essential for the pathogenicfunctions of Teffs (Fig. 4A). Expectedly, WT CD4+ Treg effi-ciently suppressed the pathogenic potential of WT Teffs, as NOD.TCRa2/2 recipient mice remained 90% T1D-free for up to 23d posttransfer (Fig. 4A). Strikingly, recipient mice receiving WTTeffs and ICOS2/2 Tregs rapidly developed T1D by day 13–15posttransfer, demonstrating that ICOS2/2 Tregs are unable tomaintain self-tolerance compared with WT Tregs (Fig. 4A). Thesein vivo data are consistent with our observations that ICOS2 or

FIGURE 2. ICOS costimulation promotes Treg competitive fitness in

vivo. A, NOD.TCRa2/2 mice received CD4+ T cells (1 3 106) isolated

from pooled LN and spleen of 3- to 4-wk-old donor congenic WT (Thy1.1+)

and ICOS2/2 (Thy1.2+) BDC2.5 mice. Eight days posttransfer, CD4+

donor T cells were examined for Foxp3 expression (left panel) and pro-

liferation of Foxp3+ Tregs (right panel). B, Donor WT (Thy1.1+) or

ICOS2/2 (Thy1.2+) BDC2.5 CD4+ T cells (8 3 106), isolated from pooled

LN and spleen, were adoptively transferred into WT (Thy1.1+) or ICOS2/2

(Thy1.2+) BDC2.5 recipient mice. Representative FACS dot plots and bar

graphs show cycling of donor or resident Foxp3+ Tregs in the target organ

21 d posttransfer. C, MACS-purified, CFSE-labeled CD4+ T cells (53 106)

isolated from pooled LN and spleen of WT and ICOS2/2 BDC2.5 mice

were transferred i.v. into NOD recipient mice. The pLN of recipient mice

were harvested on day 21 posttransfer, and the proliferative capacity of do-

nor CD4+Foxp3+/2 T cells was determined in recipient mice (n = 3 mice/

group). Similar results were obtained in three independent experiments.

Results represent the mean 6 SD.

4 ICOS PROMOTES Foxp3+ Treg CELL FITNESS IN T1D

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

ICOS2/2 Foxp3+ Tregs are unable to suppress Teffs in vitro (Fig.3). These findings confirm that ICOS deficiency in Foxp3+ Tregscripples their capacity to control autoreactive Teffs and mediateT1D protection in contrast to ICOS2/2 Teffs, which fully preservetheir functions in vivo.Given the inability of ICOS2/2 Tregs to suppress induction of

T1D by Teffs, we then asked whether this was due to decreasedexpansion and accumulation of Foxp3+ Tregs in the pLN and

pancreas. To examine this possibility, BDC2.5 CD4+ Teffs weretransferred into NOD.TCRa2/2 mice either alone or with WT andICOS2/2 BDC2.5 Tregs, and the cellular frequency of cyclingFoxp3+ cells (Ki-67 expression) was analyzed on days 15 and 23posttransfer. The frequency of ICOS2/2 Foxp3+ Tregs in the pLNand pancreas was significantly lower compared with recipients ofWT Tregs (22.8 6 4.2 versus 11.13 6 3.02, p # 0.004) (Fig. 4B),and particularly after the onset of T1D (22.8 6 4.2 versus 5.6 6

FIGURE 3. ICOS-deficient Foxp3+ Tregs are less

suppressive in vitro. A, FACS-purified BDC2.5

CD4+CD252ICOS2 responder T cells (5 3 104)

isolated from pooled LN and spleen of 3- to 4-wk-

old donors were CFSE-labeled and stimulated with

mimetope (RVRPLWVRME) (2 ng/ml) and ir-

radiated spleen cells (2 3 105) in the presence or

absence of titrated numbers of FACS-purified

ICOS+ (filled squares) or ICOS2 (open squares)

CD4+CD25+ Tregs. CFSE dilution profiles (left

panel) and cell division analysis (right panel) are

shown at multiple Teff/Treg ratios. B, To exclude

contaminating CD25+ICOS+ Teffs, FACS-purified

Foxp3GFP2ICOS2 responder T cells were cocul-

tured with titrated numbers of ICOS+ or ICOS2

Foxp3GFP+ Tregs from BDC2.5.Foxp3GFP reporter

mice. Responder T cell proliferation was measured

by [3H]thymidine incorporation. C, CFSE-labeled,

FACS-purified CD4+CD252ICOS2 responder T cells

(5 3 104) from WT BDC2.5 mice were activated

in the presence or absence of titrated numbers

of ICOS+/2 CD4+CD25+ Tregs from WT BDC2.5

mice, or CD4+CD25+ Tregs from BDC2.5 ICOS2/2

mice. Similar results were obtained in three inde-

pendent experiments. Results represent the mean 6SD.

FIGURE 4. ICOS genetic deficiency or Ab block-

ade selectively abrogates Foxp3+ Treg-mediated pro-

tection from T1D. A, NOD.TCRa2/2 recipient mice

were adoptively transferred with CD4+CD25– Teffs

(1 3 106) from WT or ICOS2/2 BDC2.5 mice in the

presence or absence of WT or ICOS2/2 CD4+CD25+

Tregs (0.125 3 106) from 4-wk-old BDC2.5 mice.

Blood glucose levels in recipient mice (n = 10/

group) were monitored for T1D incidence every 48 h

posttransfer. B, Twenty-one days after T cell transfer,

mice from each group were sacrificed, pLN and

pancreas were harvested, and the cellular frequency

of cycling (as per Ki-67 expression), islet-specific

Foxp3+ Tregs was examined relative to cycling

Foxp32 Teffs. C, NOD.TCRa2/2 recipient mice

were adoptively transferred with CD4+CD25– Teffs

(1 3 106) from WT in the presence or absence of

WT or ICOS2/2 CD4+CD25+ Tregs (0.125 3 106)

from 4-wk-old BDC2.5 mice. Simultaneously, each

respective group received three subsequent i.p. in-

jections of anti-ICOS mAb (50 or 100 mg). Blood

glucose levels in recipient mice (n = 10/group) were

monitored for T1D incidence every 48 h posttransfer.

D, Nineteen days after T cell transfer, mice from

groups receiving WT or ICOS2/2 Tregs were sacri-

ficed, pancreas was harvested, and the cellular fre-

quency of islet-specific Foxp3+ Tregs was examined

in each group. Data are representative of three sep-

arate experiments. Results represent the means 6SD.

The Journal of Immunology 5

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

0.7, p # 0.0001) (Fig. 4B). Moreover, T1D protection correlateswith the increased proportion of WT Foxp3+ Tregs in the pancreasand with decline in Foxp32 Teff cycling (24.8 6 2.1 versus 33.86 4.6, NS) (Fig. 4B), an observation not seen with ICOS2/2 Treg,particularly after the onset of T1D (24.8 6 2.1 versus 57.6 6 4.7,p # 0.0002) (Fig. 4B). This suggests that the proliferative po-tential of Tregs correlates directly with their functional potency insitu and strongly indicates that Tregs actively suppress autore-active Foxp32 Teffs in the pancreas. Thus, our data show thatICOS2/2 Tregs are unable to prevent T1D induction and correlatewith their decreased accumulation in the pancreatic lesion.

In vivo ICOS Ab blockade selectively targets Foxp3+ Tregs inthe target organ and exacerbates T1D

We then performed an anti-ICOS mAb blockade experiment todemonstrate the dependence and specificity of the ICOS signalin Foxp3+ Treg-mediated T1D protection in vivo. To this end,we administered a blocking anti-ICOS mAb in our Teff/Treg co-transfer system and then assessed the onset of T1D relative to thefrequency of Foxp3+ Tregs in pancreatic sites. Ab-mediated ICOSblockade did not inhibit the diabetogenic potential of Teffs com-pared with control mice (Fig. 4C). The onset of T1D occurredsimilarly in both groups (by day 10–13 posttransfer), suggestingthat ICOS signals are not required for induction of disease, anobservation consistent with results obtained with ICOS2/2 Teffs(Fig. 4A). Expectedly, WT Treg efficiently suppressed the patho-genic potential of WT Teffs and remained T1D-free for the entireduration of the experiment (up to day 19 posttransfer, endpoint ofexperiment) (Fig. 4C). Notably, ICOS blockade completely pre-vented the protection conferred by WT Tregs and rapidly exac-erbated T1D development (Fig. 4C). The rapid onset of T1Dcorrelated with a substantial loss in Foxp3+ Tregs in anti-ICOSmAb-treated mice compared with the control group (9.4 6 7.9versus 33.5 6 7.3, p # 0.015) (Fig. 4D).To demonstrate further that anti-ICOS mAb specifically targets

Foxp3+ Tregs in our system, we cotransferred ICOS2/2 Tregs withWT Teffs in the presence or absence of anti-ICOS mAb treatment.Anti-ICOS mAb treatment failed to change the course of diseaseby 10–16 d posttransfer (Fig. 4C) or frequency of Foxp3+ Tregs inthe pancreas (19.96 6 7.4 versus 16.3 6 3.3, NS) (Fig. 4D), anoutcome consistent with the lack of effect of Ab treatment ondisease progression in mice receiving Teffs alone (Fig. 4C). Thus,our results show that anti-ICOS mAb blockade targets Foxp3+

Tregs specifically and does not affect the pathogenic potential ofdiabetogenic Teffs in T1D. Moreover, we observe that Ab treat-ment specifically abolishes Foxp3+ Tregs in the pancreas but notin the pLN (data not shown). This correlates with our previousobservations that ICOS expression on Foxp3+ Tregs is predomi-nantly expressed in the pancreas and correlates with a functionalpotency of these cells directly in the target organ. Overall, our datapoint to an ICOS-dependent role in favoring the expansion andfunction of Foxp3+ Tregs within the pancreatic environment, inturn increasing the Treg/Teff ratio and tipping the balance to self-tolerance. Whether ICOS imprints such function during Treg de-velopment or specific signals received in the periphery are re-sponsible for such a phenotype is currently unclear.

Intra-islet ICOS+ Foxp3+ Tregs preferentially secrete IL-10

ICOS expression on activated murine and human T cells enablesthem preferentially to produce IL-10 (29–31). ICOS-deficient pa-tients show a severe reduction in IL-10 production, reinforcing therole of ICOS in IL-10 secretion (32). IL-10 is an important im-munomodulatory cytokine, which has been shown to be critical forthe control of T1D in BDC2.5 mice (33, 34). The link between

ICOS and IL-10 led to the hypothesis that ICOS+Foxp3+ Treg-mediated T1D protection is related to this production of IL-10. Tothis end, ICOS+ and ICOS2 Treg subsets were FACS purified fromBDC2.5.Foxp3GFP mice, activated with NOD BMDCs and mim-etope in vitro, and the capacity to produce IL-10 was assessedby flow cytometry. Our results show that only ICOS+ Tregs wereable to secrete IL-10 after Ag-specific stimulation compared withICOS2 Treg (19 6 7.1% versus 1.91 6 4.1%, p # 0.0001) (Fig.5A). Similarly, ICOS expression correlated with IL-10 productionin recently in vitro-activated CD4+ T cells from pLN and pancreasof prediabetic NOD.Foxp3GFP mice. More specifically, a substan-tial proportion of ICOS-expressing CD4+ T cells produced IL-10(4.8 6 1.8%) (Fig. 5B, left panel), and a striking proportion ofFoxp3gfp+ Tregs secreted IL-10 upon in vitro polyclonal stimula-tion (47.2 6 6.4%) (Fig. 5B, right panel). The frequency of IL-10–secreting Foxp3+ Tregs is significantly greater in the pancreasthan in pLN (47.2 6 6.35 versus 3.6 6 0.5, p # 0.0001), wherelittle IL-10 production is detected (Fig. 5B, right panel). Sur-prisingly, the in vitro suppressive activity of ICOS+ or ICOS2

Treg subsets was not affected by IL-10 neutralization suggestingthat redundant effector mechanisms of Treg suppression operatein vitro (data not shown).To assess directly the requirement for ICOS expression in the

production of IL-10 by Foxp3+ Tregs, we examined the poten-tial for IL-10 secretion in BDC2.5 ICOS2/2 mice. To this end,WT or ICOS2/2 BDC2.5 CD4+ T cells were transferred intoNOD.TCRa2/2 recipients and the frequency of IL-10–producingFoxp3+ Treg assessed in pLN and pancreas 8 d posttransfer. Weobserved that the most significant IL-10 production by Tregs oc-curs directly in the pancreas compared with that in the drainingLN (Fig. 5C). Furthermore, we observe that the frequency of IL-10–producing Foxp3+ Tregs is dramatically reduced in the pLN,and particularly in the pancreas, of ICOS2/2 mice compared withthat in WT mice (16.7 6 7.7 versus 3.0 6 0.7, p # 0.002) (Fig.5C). Overall, our results show that ICOS expression is required forIL-10 production by Tregs and likely drives this effector differ-entiation in the pancreas.

Loss in ICOS expression and IL-10 production in Foxp3+ Tregscoincides with T1D progression

We and others have previously shown that Foxp3+ Treg functionwanes with T1D progression, in turn enabling the diabetogenicT cells to escape regulation and promote destructive infiltration ofislets (8, 9, 11–13, 35, 36). We then explored the possibility thatbreakdown in self-tolerance may be attributed to an age-relateddecline in ICOS expression on Tregs, at checkpoints known todemarcate the early (#4 wk) from late ($12 wk) stages of insu-litis in the BDC2.5 model. We observed a remarkable decline inthe cellular frequency of ICOS-expressing Foxp3+ Tregs withinthe pancreas (68.6 6 1.4 versus 47.4 6 1.8, p # 0.0008) (Fig. 6A,right panel), in stark contrast to the pLN. Similarly, ICOS ex-pression on intrapancreatic Foxp3+ Tregs is significantly reducedwith age (144.3 6 11.9 versus 50.7 6 17.6, p # 0.003) (Fig. 6A,left panel), suggesting that the suppressive potential of Tregscorrelates with ICOS-mediated costimulation in Tregs. Moreover,this reduction in the cellular frequency of ICOS+Foxp3+ Tregs wasdirectly related to the loss in expansion (Ki-67 expression) ofICOS-expressing Foxp3+ Tregs directly within the pancreas ofprediabetic mice (50.2 6 3.8 versus 25.0 6 2.6, p # 0.003) (Fig.6B) with time, a trend not observed in pLN (10.4 6 0.3 versus7.9 6 0.7, NS) (Fig. 6B).We and others have previously shown that T1D progression in

NOD mice correlates with a drastic reduction of intrapancreaticTreg numbers (11, 13). We then wondered whether T1D onset

6 ICOS PROMOTES Foxp3+ Treg CELL FITNESS IN T1D

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

could be attributed to a loss of ICOS expression on Foxp3+ Tregs.To this end, CD4+ T cells from BDC2.5.Foxp3GFP mice weretransferred into NOD.TCRa2/2 recipients, and the cellular fre-quency of ICOS+Foxp3+ Tregs in the pLN and pancreas was an-alyzed relative to T1D onset. Notably, the expression level ofICOS as well as the frequency of ICOS+Foxp3+ Tregs sustaineda drastic reduction as recipient mice progressed from prediabetesto overt T1D (10606 92.8 versus 362.36 68.6, p# 0.0001) (Fig.6C). Thus, we show that T1D progression correlates with a loss ofICOS expression on Foxp3+ Tregs, particularly in the pancreas,indicating a loss of Treg suppression throughout T1D progression.Herman et al. (21) showed that IL-10 and ICOS mRNA were

highly expressed by pancreatic CD25+CD692 Tregs relative totheir draining LN counterparts. As pancreatic ICOS+ Foxp3+

Tregs possess the capacity to secrete high amounts of IL-10, wethen assessed whether the loss in ICOS expression with T1D

progression dampened the capacity of Tregs to produce IL-10 inour BDC2.5 transfer model (34). To this end, CD4+ T cells frompLN and pancreas of prediabetic and diabetic recipient mice werereactivated ex vivo, and the secretion of IL-10 was assessed by

FIGURE 5. Intra-islet ICOS+ Foxp3+ Tregs preferentially secrete IL-10.

A, FACS-purified ICOS+ and ICOS2 Foxp3GFP+ T cells were isolated from

pooled LN and spleen of 3- to 4-wk-old BDC2.5GFP reporter mice and then

activated in the presence of NOD BMDCs at a 4:1 ratio and BDC2.5

mimetope (100 ng/ml) for 72 h. The frequency of IL-10–producing Foxp3+

Tregs, relative to ICOS expression, was assessed by flow cytometry (left

panel). The proportion of IL-10–producing Foxp3+ Tregs among ICOS+

and ICOS2 subsets are shown (right panel). B, Cell suspensions from pLN

and pancreas of 4-wk-old NOD.Foxp3GFP reporter mice were reactivated

in vitro for 4 h with PMA/ionomycin, and the frequency of IL-10–pro-

ducing Foxp3+ Tregs, relative to ICOS expression, was determined by

FACS (left panel). The proportion of IL-10–producing ICOS+Foxp3+ Tregs

in pLN and pancreas are shown (right panel). C, NOD.TCRa2/2 mice

were transferred with CD4+ T cells (13 106) isolated from pooled LN and

spleen of 3- to 4-wk-old WT (Thy1.1+) or ICOS2/2 (Thy1.2+) BDC2.5

mice. Eight days posttransfer, the frequency of IL-10–secreting Foxp3+

Tregs in each donor group was determined by FACS in pancreas and pLN,

as in B. Data are representative of three separate experiments. Results

represent the means 6 SD.

FIGURE 6. Temporal loss in ICOS expression and IL-10 production in

Foxp3+ Tregs coincides with T1D onset and progression. A, Shown are the

levels of ICOS expression on Foxp3+ Tregs (left panel) and proportions of

ICOS+Foxp3+ Tregs (right panel) in pLN and pancreas of 3- to 4-wk-old

and 12-wk-old BDC2.5 mice as analyzed by FACS. B, The extent (left

panel) and proportion (right panel) of proliferating (based on Ki-67 ex-

pression) ICOS+Foxp3+ Tregs in pancreas of 3- to 4-wk-old and 12-wk-old

BDC2.5 mice are shown. C, NOD.TCRa2/2 recipient mice were trans-

ferred i.v. with CD4+ T cells (13 106) isolated from pooled LN and spleen

of 3- to 4-wk-old BDC2.5.Foxp3GFP reporter mice, and T1D onset was

monitored as in Fig. 3. Fourteen to twenty-one days posttransfer, Foxp3+

and Foxp32 CD4+ T cells from NOD.TCRa2/2 recipient were restimu-

lated ex vivo with PMA/ionomycin, and the frequency of IL-10–producing

Foxp3+ Tregs, relative to ICOS expression, was assessed in pLN and

pancreas of prediabetic and diabetic mice (top panel). A representative dot

plot of IL-10 production in ICOS+CD4+ T cells in the pancreas of predi-

abetic and diabetic mice is shown (top panel). ICOS expression (MFI, left

y axis: :, pLN; ,, pancreas) relative to the frequency of IL-10+ ICOS+

Tregs (right y axis: n, pLN; N, pancreas) in prediabetic and diabetic mice

(bottom panel). Results are representative of three independent experi-

ments, (n = 7 mice/group). Results represent the means 6 SD.

The Journal of Immunology 7

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

flow cytometry. Notably, the marked drop in the expression levelof ICOS and frequency of ICOS-expressing Foxp3+ Tregs withinthe pancreas of diabetic mice correlated with a 3-fold decline intheir capacity to produce IL-10 (21.5 6 4.7 versus 5.8 6 1.3, p #0.03) (Fig. 6C). Taken together, these data show that ICOS drivesthe differentiation of IL-10–producing Tregs particularly in thepancreas of prediabetic mice and that T1D progression correlateswith a loss of Foxp3+ICOS+ Tregs within the pancreas.

IL-2 preferentially supports the fitness and survival ofICOS-expressing Foxp3+ Tregs

IL-2 is essential for the survival and function of Foxp3+ Tregswithin islets, and a local deficiency in IL-2 was shown to renderTregs unfit and functionally defective, in turn provoking a Treg/Teff imbalance in situ (13). This is similar to earlier studiesshowing that T cells from prediabetic NOD mice become hypo-proliferative and poor IL-2 producers at the onset of insulitis (3).Therefore, we examined whether the decline in the ICOS+ Tregpool, coincident with a lack of suppression of these cells and T1Dprogression, is associated with altered IL-2 levels. We first as-sessed IL-2Ra (CD25) expression in ICOS+ and ICOS2 Tregsubsets in response to IL-2 in vitro. To this end, ICOS+ and ICOS2

Tregs from BDC2.5.Foxp3GFP reporter mice were activatedin vitro with APCs and mimetope in the presence or absence ofIL-2. We observe that ICOS+ Tregs are more responsive to IL-2compared with ICOS2 Tregs, as this resulted in a 28-fold increasein CD25 expression in ICOS+ Tregs and only a 10-fold increase inICOS2 Tregs in response to IL-2 compared with nontreatedconditions (Fig. 7A, left panel). To confirm that ICOS+ Tregspreferentially respond to IL-2, STAT5 phosphorylation (p-STAT5)was assayed as an indicator of activation of the IL-2 signalingpathway. Our results show that administration of IL-2 leads toa significantly higher proportion (83.3 6 2.6% versus 44.0 61.8%, p # 0.002) and MFI (1443 6 19.1 versus 779 6 4.2, p #

0.008) of p-STAT5–positive ICOS+ Tregs compared with ICOS2

Tregs (Fig. 7A, middle and right panels).IL-2–mediated STAT5 phosphorylation is essential for survival

of Foxp3+ Tregs (37–39). As ICOS+ Tregs are more sensitive todifferential IL-2 levels, we then examined whether they also dif-fered in their survival potential. We observe that ICOS+ Tregs,compared with ICOS2 Tregs, are more susceptible to death byIL-2 withdrawal as indicated by the increased frequency of apopto-tic cells in the absence of IL-2 compared with untreated cells (%annexin V+ PI+ cells: 90.3 6 4.6 versus 62.3 6 1.9, p # 0.0003)(Fig. 7B, left panel). IL-2 treatment rescues ICOS+ Tregs fromapoptosis, whereas ICOS2 Tregs remain insensitive to IL-2 (45.461.8 versus 39.0 6 0.9, p # 0.0002) (Fig. 7B, left panel). The IL-2–driven survival of ICOS+ Tregs was also confirmed by the upreg-ulation of Bcl-2 expression in ICOS+ Tregs (1447 6 7.5 versus657 6 4.6, p # 0.008) in response to IL-2 in contrast to ICOS2

Tregs whose Bcl-2 expression remained unchanged irrespective ofIL-2 treatment (605 6 13.2 versus 702 6 6.8, NS) (Fig. 7B,middle and right panels). Overall, ICOS+ Tregs, in contrast toICOS2 Tregs, are more dependent on IL-2 signals, which bolsterCD25 expression and prevent apoptosis in ICOS+ Tregs. Notably,WT and ICOS2/2 BDC2.5 Tregs do not differ in their intrinsicIL-2 response after activation as measured by p-STAT5 levels(data not shown); rather, the ICOS+ Tregs, compared with ICOS2

Tregs, have different IL-2 requirements to maintain their func-tional fitness. Therefore, ICOS expression is not required formediating IL-2 responses in ICOS+ Tregs, but rather delineates anIL-2–responsive Foxp3+ Treg subset in the pancreatic microen-vironment.

Il2 protective allelic variants or prophylactic IL-2 therapyrestores ICOS expression in Tregs in BDC2.5 mice

A local IL-2 deficiency in islets is one of the underlying mecha-nisms in the age-related waning of Treg functions and loss of self-tolerance in NOD mice (13, 14, 24). Consistently, we have pre-viously shown that Tregs within the pancreatic lesion of WTBDC2.5 mice exhibited reduced expansion, survival, and function,defects readily corrected by prophylactic IL-2 therapy or Il2allelic variation in NOD mice (13). Thus, we hypothesized thatIL-2 promotes ICOS expression in vivo and sought to deter-mine whether Il2 allelic variants of T1D-protected congenicBDC2.5Idd3 mice could restore ICOS expression on Tregs and theenhanced frequency of ICOS+Foxp3+ Tregs in situ. We observedan increase in the proportion of ICOS+Foxp3+ Tregs (86.4 6 1.7versus 68.6 6 1.4, p # 0.001) (Fig. 7C, right panel) and extent ofICOS expression (MFI) (282.3 6 54.0 versus 144.3 6 11.9, p #0.04) (Fig. 7C, left and middle panels) on Foxp3+ Tregs in thepancreas of 4-wk-old BDC2.5Idd3 mice relative to age-matchedWT mice. Despite the substantial drop in the frequency of ICOS-expressing Foxp3+ Tregs by 12 wk of age in prediabeticBDC2.5Idd3 mice, the frequency was nonetheless significantlygreater than that in WT mice (86.37 6 1.7 versus 72.3 6 0.4, p #

0.005) (Fig. 7C, right panel). Moreover, the enhanced proportionof ICOS+Foxp3+ Tregs coincided with a 2-fold enhancement inthe fraction of actively proliferating ICOS-expressing Foxp3+

Tregs in 4-wk-old BDC2.5Idd3 mice relative to WT mice (39.1 65.7% versus 24.4 6 4.5%, p # 0.01) (data not shown). Thissignificant difference is maintained at later time points with dis-ease progression (22.3 6 2.8% versus 14.7 6 3.4%, p # 0.04)(data not shown). Overall, Il2 allelic variation, known to increaseIL-2 production in Teffs, also increases ICOS expression onFoxp3+ Tregs within the target organ, in turn, promoting theirfunctions and T1D protection.To confirm whether IL-2 deficiency is an underlying cause of the

temporal loss of ICOS expression observed in Tregs within thepancreas (Fig. 5), a prophylactic IL-2 therapeutic regimen, knownto promote T1D protection, was initiated in prediabetic BDC.2.5mice. In stark contrast to draining pLN, we observed a significantincrease in the proportion of Foxp3+ Tregs within the pancreaticinfiltrate of mice receiving IL-2 relative to PBS controls (13.3 63.1% versus 9.4 6 1.9%, p # 0.03) (Fig. 7D, left panel). Theproportion of ICOS+Foxp3+ Tregs within the CD4+ T cell infiltrateof the pancreas was also enhanced in IL-2–treated BDC2.5 micerelative to PBS-treated controls (87.2 6 4.3 versus 66.4 6 3.8,p # 0.0095), comparable with levels observed in BDC2.5Idd3mice (87.2 6 4.3 versus 84.25 6 1.2) (Fig. 7D, right panel). IL-2therapy also drives the expression of CD25 on Foxp3+ Tregs as theproportion of Foxp3+CD25+ICOS+ was enhanced to levels ob-served in BDC2.5Idd3 mice and in IL-2–treated mice versus PBScontrols (8.86 2.6% versus 56 0.7%, p# 0.05) (Fig. 7D, middlepanel). In addition, our results show that the inability of ICOS2/2

Tregs to suppress T1D in our T cell transfer model could not bereversed by prophylactic administration of IL-2, a condition thatonly partially bolsters Foxp3+ Tregs of ICOS2/2 origin in contrastto WT Tregs (data not shown). Our data show that protective Il2allelic variants or low-dose IL-2 treatment restored ICOS andCD25 expression in intrapancreatic Foxp3+ Tregs, both conditionspromoting T1D protection (13, 14). Overall, this suggests thatICOS expression by Foxp3+ Tregs plays an important role in IL-2–mediated rescue of Foxp3+ Tregs and T1D protection. More-over, our results are reminiscent of those of Grinberg-Bleyer et al.(24), who demonstrated that IL-2 therapy can even reverseestablished T1D in NOD mice via a local effect on pancreatic

8 ICOS PROMOTES Foxp3+ Treg CELL FITNESS IN T1D

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

FIGURE 7. IL-2 is essential for ICOS expression in Foxp3+ Tregs and drives their fitness and survival. A, FACS-purified ICOS+ or ICOS2 Foxp3+ Tregs

(13 105) from pooled LN and spleen of 3- to 4-wk-old BDC2.5.Foxp3GFP reporter mice were stimulated with BDC2.5 mimetope (40 ng/ml) and irradiated

spleen cells (4 3 105) in the presence or absence of IL-2 (5 mg/ml) for 36–72 h. The frequency of IL-2–responsive ICOS+ and ICOS2 Foxp3+ Tregs was

assessed by measuring CD25 expression in ICOS+ and ICOS2 Treg subsets in unstimulated and IL-2–treated conditions demonstrated in histogram plots

(left panel). The frequency of IL-2–responsive ICOS+ and ICOS2 Foxp3+ Tregs was assessed by measuring p-STAT5 by flow cytometry. A representative

histogram plot (middle panel) and MFI values (right panel) for p-STAT5 are shown. MFI values are expressed as d MFI of p-STAT5 in presence or absence

of IL-2. B, The frequency of apoptotic cells (annexin V+/PI+) in ICOS+ and ICOS2 Tregs in the presence or absence of IL-2 was assessed by flow cytometry

(left panel). The frequency of surviving ICOS+ and ICOS2 Tregs in the presence or absence of IL-2 (indicated by Bcl-2 expression) was assessed by flow

cytometry. Representative histogram plots of Bcl-2 expression in ICOS+ and ICOS2 Tregs (middle panel) and MFI values (right panel) are shown. MFI

values are expressed as d MFI of Bcl-2 expression in the presence or absence of IL-2. Each experiment was repeated three times with 10–15 mice per

experiment. Results represent the mean 6 SD. C, The level of ICOS expression (MFI) (left and middle panels) on Foxp3+ Tregs and the frequency (right

panel) of ICOS+Foxp3+ Tregs in pancreas of 3- to 4-wk-old and 12-wk-old BDC2.5 and BDC.Idd3mice was analyzed by FACS. D, Three- to four-week-old

female BDC2.5 mice were injected daily with IL-2 (25,000 IU) for 5 consecutive days. On day 7 after the initial injection, the proportion of Foxp3+ Tregs

within CD4+ T cells (left panel), the proportion of CD25-expressing Foxp3+ Tregs (middle panel), and the proportion of ICOS+ within Foxp3+ Tregs (right

panel) in pLN and pancreas was compared between PBS-treated and IL-2–treated BDC2.5 mice, as well as to BDC2.5Idd3 mice. Data are representative of

at least three separate experiments (n = 3 mice/group). Results represent the mean 6 SD.

The Journal of Immunology 9

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

Tregs by upregulating the expression of genes involved in Tregfunction and survival such as Bcl-2, CD25, Foxp3, and ICOS.

DiscussionA progressive loss in Foxp3+ Treg function in pancreatic sitesunleashes the pathogenic potential of islet-reactive Teffs, conse-quently leading to overt T1D in NOD mice (8, 11, 33, 40–42).Foxp3+ Treg development and function are heavily dependent oncostimulation, as shown by the exacerbated T1D onset in CD282/2

or B72/2 NOD mice (16, 17). Recently, ICOS has emerged asanother critical costimulatory molecule in various autoimmunedisorders (20, 43, 44). Genetic ablation or blockade of ICOS atvarious time points during disease progression yields divergentclinical outcomes, suggesting that ICOS may operate via cell type-or context-dependent mechanisms (20, 45). ICOS mRNA is highlyexpressed in pLN of T1D-protected BDC2.5 mice, and icos genedeficiency or ICOS blockade in neonatal BDC2.5 mice precip-itates T1D (21, 22). Whereas ICOS2/2 NOD mice are T1D re-sistant, ICOS blockade in NOD neonates resulted in T1D exac-erbation (21, 23). However, it is unknown if ICOS differentiallyregulates Foxp32 Teff and Foxp3+ Treg subsets or operates indifferent locations or phases of an anti-islet immune response (21,23). These findings prompted us to evaluate the functional impactof differential ICOS activity in Teff and Treg subsets throughoutT1D progression.In this study, we make the formal demonstration that ICOS is

a key costimulatory pathway that restrains islet autoimmunity byspecifically affecting the homeostasis and function of Foxp3+ Tregsin prediabetic islets of NOD mice. We show that ICOS is a markerthat discriminates effector Foxp32 Teffs from Foxp3+ Tregs andspecifically designates a subset of Tregs occupying islets of pre-diabetic mice. We also show that the intra-islet ICOS-expressingFoxp3+ Treg subset of neonatal BDC2.5 mice, in contrast to con-ventional Foxp3+ Tregs, is endowed with an increased potential toexpand in situ, acquire an IL-10–secreting phenotype, and mediategreater suppressive activity in vitro and in vivo. We further showthat the ICOS+ Treg subset, in contrast to its ICOS2 counterparts,is more sensitive to IL-2–mediated survival, and the temporal lossin ICOS+ Tregs is readily corrected by IL-2 therapy or protectiveIl2 gene variation. ICOS sustains the stability of Foxp3+ Tregfunctions in prediabetic islets, and alterations in the ICOS path-way by ICOS deficiency or mAb-mediated blockade of ICOS inBDC2.5 mice specifically cripple Foxp3+ Treg fitness and functionand abrogate T1D protection. Thus, ICOS maintains the homeo-stasis and functional stability of Foxp3+ Tregs in prediabetic islets.A local IL-2 deficiency was shown to compromise Foxp3+ Treg

survival and function within islets and exacerbate T1D progres-sion, a defect that is readily restored by the protective Il2 allelicvariants or IL-2 therapy in prediabetic or overtly diabetic NODmice (11–14, 24). Several lines of evidence support a modelwhereby a positive feedback IL-2/ICOS loop stabilizes Treg sur-vival and function in the inflamed pancreatic milieu (25). First, Il2allelic variation in T1D-resistant BDC2.5Idd3 mice produces morepotent Tregs expressing higher ICOS levels than those of WTTregs. Second, low-dose IL-2 therapy in BDC2.5 mice at the timeof insulitis restored ICOS levels on intra-islet Foxp3+ Tregs com-parable with those of BDC2.5Idd3 mice. Third, we show thatICOS+ Tregs, in contrast to ICOS2 Tregs, are more responsive toIL-2 in terms of STAT5 phosphorylation and activation, a signalingpathway essential for their survival, fitness, and functional stabil-ity. IL-2–mediated activation of the STAT5 pathway results in theupregulation of ICOS surface expression and increased foxp3 andcd25 gene expression (data not shown), which in turn stabilizesthe suppressive Treg phenotype and prevents T1D. Moreover,

when IL-2 is limiting, ICOS+ Tregs are highly susceptible to ap-optosis, and IL-2 readily rescues them from apoptosis by upreg-ulating Bcl-2, features not observed in the ICOS2 Treg subset.These data show that ICOS+ Tregs are more dependent on IL-2 tomaintain their functional fitness. Thus, a limited bioavailability ofIL-2 in pancreatic environments may compromise ICOS-mediatedstabilization Foxp3+ Treg function in islet autoimmunity.The role of ICOS to instruct T cells to produce IL-10 has been

documented (18). In a mouse model of asthma, the development ofAg-induced IL-10–producing Tregs and consequential suppressionof allergen-induced airway hyperreactivity was dependent on anintact ICOS–ICOS-ligand pathway (46). Moreover, IL-10 hasbeen shown to be immunoprotective in T1D in some contexts (47).However, the potential for IL-10 production by Tregs has not beenas extensively examined as in other models where such productionplays a key modulatory role in disease. In this study, we report thatIL-10 production by Foxp3+ Tregs is restricted within the ICOS+

Foxp3+ Treg subset in contrast to ICOS2Foxp3+ Tregs or Foxp32

Teffs. The greater propensity for ICOS+Foxp3+ Tregs to cycle ininflamed pancreatic sites, in conjunction with their dependence onIL-2, may promote the differentiation of the IL-10–secretingphenotype in islet-reactive Foxp3+ Tregs from recently activatedFoxp3+ Tregs in pancreatic sites. Moreover, the loss of ICOSexpression and consequential IL-10 production by Ag-specificFoxp3+ Tregs with T1D progression indicates that ICOS mighttrigger IL-10 production by Foxp3+ Tregs. More importantly, thedifferentiation of IL-10–producing Foxp3+ Tregs in islets is im-paired in ICOS2/2 mice where Tregs fail to produce significantlevels of IL-10 in the pancreatic sites compared with WT Tregs.Diminished expression of ICOS in Tregs of recently diagnosedhuman T1D patients supports the link between ICOS costimula-tion and the potency of the Foxp3+ Treg compartment (32).Our data do not exclude the possibility that the defective Treg

function observed in ICOS2/2 mice may reflect a role for ICOSin imprinting such function during Treg development. WhetherICOS+ Tregs are generated in the thymus or from ICOS2 Tregs inthe periphery is unknown. We show that ICOS2 Tregs are readilyable to upregulate ICOS upon TCR activation in vitro, or in re-sponse to lymphopenia in vivo. Notably, mere induction of ICOSexpression does not enable these cells to acquire the same func-tional features of ex vivo ICOS+ Tregs. This suggests that thethymus might not only play a critical role in the selection ofFoxp3+ Tregs but also has the ability to imprint different Tregsubsets with different functional potentials in the periphery. Al-ternatively, the important accumulation of ICOS+ Tregs in thepancreas may indicate that the inflammatory signals derived fromthe target organ can influence the expansion and recruitment ofthese cells. Consistently, Herman et al. (21) demonstrated that thegene expression profile of islet-infiltrating Tregs was drasticallydifferent from that of Tregs in pLN, suggesting that the inflam-matory environment of the target tissue directs unique transcrip-tional programs in Tregs that may be related to their mechanismof action in situ.Overall, our studies point to ICOS as an essential signal for the

control of the pool size and instruction of the Treg suppressivephenotype within the islet microenvironment. Our findings extendour current understanding of the role of ICOS in T1D pathogenesisand suggest that IL-2 insufficiency may provoke a temporal loss inICOS expression, which correlates with a functional waning ofTregs and breakdown in self-tolerance. As the icos gene may befunctionally polymorphic in human T1D and contribute to T1Drisk, it is tempting to speculate that such genetic variation maybe affecting Treg-dominant mechanisms of islet tolerance (48).The elucidation of the mechanisms involved in maintaining

10 ICOS PROMOTES Foxp3+ Treg CELL FITNESS IN T1D

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

self-tolerance will add to our current understanding and therapy ofautoimmune disorders such as T1D.

AcknowledgmentsWe thank Marie-Helene Lacombe from the McGill University Health Cen-

ter Research Institute Immunophenotyping Platform for cell sorting.

DisclosuresThe authors have no financial conflicts of interest.

References1. Anderson, M. S., and J. A. Bluestone. 2005. The NOD mouse: a model of im-

mune dysregulation. Annu. Rev. Immunol. 23: 447–485.2. Bach, J. F., and L. Chatenoud. 2001. Tolerance to islet autoantigens in type 1

diabetes. Annu. Rev. Immunol. 19: 131–161.3. Delovitch, T. L., and B. Singh. 1997. The nonobese diabetic mouse as a model of

autoimmune diabetes: immune dysregulation gets the NOD. Immunity 7: 727–738.4. Hori, S., T. Nomura, and S. Sakaguchi. 2003. Control of regulatory T cell de-

velopment by the transcription factor Foxp3. Science 299: 1057–1061.5. Khattri, R., T. Cox, S.-A. Yasayko, and F. Ramsdell. 2003. An essential role for

Scurfin in CD4+CD25+ T regulatory cells. Nat. Immunol. 4: 337–342.6. Fontenot, J. D., M. A. Gavin, and A. Y. Rudensky. 2003. Foxp3 programs the

development and function of CD4+CD25+ regulatory T cells. Nat. Immunol. 4:330–336.

7. Billiard, F., E. Litvinova, D. Saadoun, F. Djelti, D. Klatzmann, J. L. Cohen,G. Marodon, and B. L. Salomon. 2006. Regulatory and effector T cell activation levelsare prime determinants of in vivo immune regulation. J. Immunol. 177: 2167–2174.

8. Chen, Z., A. E. Herman, M. Matos, D. Mathis, and C. Benoist. 2005. Where CD4+CD25+ T reg cells impinge on autoimmune diabetes. J. Exp. Med. 202: 1387–1397.

9. Pop, S. M., C. P. Wong, D. A. Culton, S. H. Clarke, and R. Tisch. 2005. Singlecell analysis shows decreasing FoxP3 and TGFbeta1 coexpressing CD4+CD25+regulatory T cells during autoimmune diabetes. J. Exp. Med. 201: 1333–1346.

10. Gregori, S., N. Giarratana, S. Smiroldo, and L. Adorini. 2003. Dynamics ofpathogenic and suppressor T cells in autoimmune diabetes development. J.Immunol. 171: 4040–4047.

11. Tritt, M., E. Sgouroudis, E. d’Hennezel, A. Albanese, and C. A. Piccirillo. 2008.Functional waning of naturally occurring CD4+ regulatory T-cells contributes tothe onset of autoimmune diabetes. Diabetes 57: 113–123.

12. You, S., M. Belghith, S. Cobbold, M.-A. Alyanakian, C. Gouarin, S. Barriot,C. Garcia, H. Waldmann, J. F. Bach, and L. Chatenoud. 2005. Autoimmunediabetes onset results from qualitative rather than quantitative age-dependentchanges in pathogenic T-cells. Diabetes 54: 1415–1422.

13. Sgouroudis, E., A. Albanese, and C. A. Piccirillo. 2008. Impact of protectiveIL-2 allelic variants on CD4+ Foxp3+ regulatory T cell function in situ andresistance to autoimmune diabetes in NOD mice. J. Immunol. 181: 6283–6292.

14. Tang, Q., J. Y. Adams, C. Penaranda, K. Melli, E. Piaggio, E. Sgouroudis,C. A. Piccirillo, B. L. Salomon, and J. A. Bluestone. 2008. Central role of de-fective interleukin-2 production in the triggering of islet autoimmune destruc-tion. Immunity 28: 687–697.

15. Jaramillo, A., B. M. Gill, and T. L. Delovitch. 1994. Insulin dependent diabetesmellitus in the non-obese diabetic mouse: a disease mediated by T cell anergy?Life Sci. 55: 1163–1177.

16. Tang, Q., K. J. Henriksen, E. K. Boden, A. J. Tooley, J. Ye, S. K. Subudhi,X. X. Zheng, T. B. Strom, and J. A. Bluestone. 2003. Cutting edge: CD28controls peripheral homeostasis of CD4+CD25+ regulatory T cells. J. Immunol.171: 3348–3352.

17. Salomon, B., D. J. Lenschow, L. Rhee, N. Ashourian, B. Singh, A. Sharpe, andJ. A. Bluestone. 2000. B7/CD28 costimulation is essential for the homeostasis ofthe CD4+CD25+ immunoregulatory T cells that control autoimmune diabetes.Immunity 12: 431–440.

18. Dong, C., R. I. Nurieva, and D. V. Prasad. 2003. Immune regulation by novelcostimulatory molecules. Immunol. Res. 28: 39–48.

19. Nurieva, R. I., J. Duong, H. Kishikawa, U. Dianzani, J. M. Rojo, I. Ho,R. A. Flavell, and C. Dong. 2003. Transcriptional regulation of th2 differentia-tion by inducible costimulator. Immunity 18: 801–811.

20. Rottman, J. B., T. Smith, J. R. Tonra, K. Ganley, T. Bloom, R. Silva, B. Pierce,J.-C. Gutierrez-Ramos, E. Ozkaynak, and A. J. Coyle. 2001. The costimulatorymolecule ICOS plays an important role in the immunopathogenesis of EAE. Nat.Immunol. 2: 605–611.

21. Herman, A. E., G. J. Freeman, D. Mathis, and C. Benoist. 2004. CD4+CD25+ Tregulatory cells dependent on ICOS promote regulation of effector cells in theprediabetic lesion. J. Exp. Med. 199: 1479–1489.

22. Prevot, N., C. Briet, H. Lassmann, I. Tardivel, E. Roy, J. Morin, T. W. Mak,A. Tafuri, and C. Boitard. 2010. Abrogation of ICOS/ICOS ligand costimulationin NOD mice results in autoimmune deviation toward the neuromuscular system.Eur. J. Immunol. 40: 2267–2276.

23. Hawiger, D., E. Tran, W. Du, C. J. Booth, L. Wen, C. Dong, and R. A. Flavell.2008. ICOS mediates the development of insulin-dependent diabetes mellitus innonobese diabetic mice. J. Immunol. 180: 3140–3147.

24. Grinberg-Bleyer, Y., A. Baeyens, S. You, R. Elhage, G. Fourcade, S. Gregoire,N. Cagnard, W. Carpentier, Q. Tang, J. Bluestone, et al. 2010. IL-2 reversesestablished type 1 diabetes in NOD mice by a local effect on pancreatic regu-latory T cells. J. Exp. Med. 207: 1871–1878.

25. Yagi, J., Y. Arimura, U. Dianzani, T. Uede, T. Okamoto, and T. Uchiyama. 2003.Regulatory roles of IL-2 and IL-4 in H4/inducible costimulator expression onactivated CD4+ T cells during Th cell development. J. Immunol. 171: 783–794.

26. Fontenot, J. D., J. P. Rasmussen, L. M. Williams, J. L. Dooley, A. G. Farr, andA. Y. Rudensky. 2005. Regulatory T cell lineage specification by the forkheadtranscription factor foxp3. Immunity 22: 329–341.

27. Thornton, A. M., and E. M. Shevach. 1998. CD4+CD25+ immunoregulatoryT cells suppress polyclonal T cell activation in vitro by inhibiting interleukin 2production. J. Exp. Med. 188: 287–296.

28. Ito, T., S. Hanabuchi, Y.-H. Wang, W. R. Park, K. Arima, L. Bover, F. X.-F. Qin,M. Gilliet, and Y.-J. Liu. 2008. Two functional subsets of FOXP3+ regulatoryT cells in human thymus and periphery. Immunity 28: 870–880.

29. Kohyama, M., D. Sugahara, S. Sugiyama, H. Yagita, K. Okumura, and N. Hozumi.2004. Inducible costimulator-dependent IL-10 production by regulatory T cellsspecific for self-antigen. Proc. Natl. Acad. Sci. USA 101: 4192–4197.

30. McAdam, A. J., T. T. Chang, A. E. Lumelsky, E. A. Greenfield, V. A. Boussiotis,J. S. Duke-Cohan, T. Chernova, N. Malenkovich, C. Jabs, V. K. Kuchroo, et al.2000. Mouse inducible costimulatory molecule (ICOS) expression is enhancedby CD28 costimulation and regulates differentiation of CD4+ T cells. J.Immunol. 165: 5035–5040.

31. Lohning, M., A. Hutloff, T. Kallinich, H. W. Mages, K. Bonhagen, A. Radbruch,E. Hamelmann, and R. A. Kroczek. 2003. Expression of ICOS in vivo definesCD4+ effector T cells with high inflammatory potential and a strong bias forsecretion of interleukin 10. J. Exp. Med. 197: 181–193.

32. Warnatz, K., L. Bossaller, U. Salzer, A. Skrabl-Baumgartner, W. Schwinger,M. van der Burg, J. J. M. van Dongen, M. Orlowska-Volk, R. Knoth, A. Durandy,et al. 2006. Human ICOS deficiency abrogates the germinal center reaction andprovides a monogenic model for common variable immunodeficiency. Blood107: 3045–3052.

33. Nurieva, R. I., X. M. Mai, K. Forbush, M. J. Bevan, and C. Dong. 2003. B7h isrequired for T cell activation, differentiation, and effector function. Proc. Natl.Acad. Sci. USA 100: 14163–14168.

34. van Berkel, M. E. A. T., and M. A. Oosterwegel. 2006. CD28 and ICOS: similaror separate costimulators of T cells? Immunol. Lett. 105: 115–122.

35. Brusko, T. M., C. H. Wasserfall, M. J. Clare-Salzler, D. A. Schatz, andM. A. Atkinson. 2005. Functional defects and the influence of age on the fre-quency of CD4+ CD25+ T-cells in type 1 diabetes. Diabetes 54: 1407–1414.

36. D’Alise, A. M., V. Auyeung, M. Feuerer, J. Nishio, J. Fontenot, C. Benoist, andD. Mathis. 2008. The defect in T-cell regulation in NOD mice is an effect on theT-cell effectors. Proc. Natl. Acad. Sci. USA 105: 19857–19862.

37. Burchill, M. A., C. A. Goetz, M. Prlic, J. J. O’Neil, I. R. Harmon, S. J. Bensinger,L. A. Turka, P. Brennan, S. C. Jameson, and M. A. Farrar. 2003. Distinct effectsof STAT5 activation on CD4+ and CD8+ T cell homeostasis: development ofCD4+CD25+ regulatory T cells versus CD8+ memory T cells. J. Immunol. 171:5853–5864.

38. Antov, A., L. Yang, M. Vig, D. Baltimore, and L. Van Parijs. 2003. Essential rolefor STAT5 signaling in CD25+CD4+ regulatory T cell homeostasis and themaintenance of self-tolerance. J. Immunol. 171: 3435–3441.

39. Yao, Z., Y. Kanno, M. Kerenyi, G. Stephens, L. Durant, W. T. Watford,A. Laurence, G. W. Robinson, E. M. Shevach, R. Moriggl, et al. 2007. Nonre-dundant roles for Stat5a/b in directly regulating Foxp3. Blood 109: 4368–4375.

40. Fontenot, J. D., J. P. Rasmussen, M. A. Gavin, and A. Y. Rudensky. 2005. Afunction for interleukin 2 in Foxp3-expressing regulatory T cells. Nat. Immunol.6: 1142–1151.

41. Tang, Q., J. Y. Adams, A. J. Tooley, M. Bi, B. T. Fife, P. Serra, P. Santamaria,R. M. Locksley, M. F. Krummel, and J. A. Bluestone. 2006. Visualizing regu-latory T cell control of autoimmune responses in nonobese diabetic mice. Nat.Immunol. 7: 83–92.

42. Yamazaki, S., T. Iyoda, K. Tarbell, K. Olson, K. Velinzon, K. Inaba, andR. M. Steinman. 2003. Direct expansion of functional CD25+ CD4+ regulatoryT cells by antigen-processing dendritic cells. J. Exp. Med. 198: 235–247.

43. Dong, C., A. E. Juedes, U.-A. Temann, S. Shresta, J. P. Allison, N. H. Ruddle,and R. A. Flavell. 2001. ICOS co-stimulatory receptor is essential for T-cellactivation and function. Nature 409: 97–101.

44. Greenwald, R. J., G. J. Freeman, and A. H. Sharpe. 2005. The B7 familyrevisited. Annu. Rev. Immunol. 23: 515–548.

45. Sporici, R. A., R. L. Beswick, C. von Allmen, C. A. Rumbley, M. Hayden-Ledbetter, J. A. Ledbetter, and P. J. Perrin. 2001. ICOS ligand costimulation isrequired for T-cell encephalitogenicity. Clin. Immunol. 100: 277–288.

46. Akbari, O., G. J. Freeman, E. H. Meyer, E. A. Greenfield, T. T. Chang,A. H. Sharpe, G. Berry, R. H. DeKruyff, and D. T. Umetsu. 2002. Antigen-specific regulatory T cells develop via the ICOS-ICOS-ligand pathway and in-hibit allergen-induced airway hyperreactivity. Nat. Med. 8: 1024–1032.

47. Hussain, S., and T. L. Delovitch. 2007. Intravenous transfusion of BCR-activatedB cells protects NOD mice from type 1 diabetes in an IL-10-dependent manner.J. Immunol. 179: 7225–7232.

48. Ihara, K., S. Ahmed, F. Nakao, N. Kinukawa, R. Kuromaru, N. Matsuura,I. Iwata, S. Nagafuchi, H. Kohno, K. Miyako, and T. Hara. 2001. Associationstudies of CTLA-4, CD28, and ICOS gene polymorphisms with type 1 diabetesin the Japanese population. Immunogenetics 53: 447–454.

The Journal of Immunology 11

by guest on June 28, 2018http://w

ww

.jimm

unol.org/D

ownloaded from