IL-4 inhibits TGF-b-induced Foxp3+ T cells and,together with TGF-b, generates IL-9+ IL-10+ Foxp3–

effector T cells

Valerie Dardalhon1,2,6, Amit Awasthi1,2,6, Hyoung Kwon1, George Galileos1, Wenda Gao3, Raymond A Sobel4,Meike Mitsdoerffer2, Terry B Strom3, Wassim Elyaman2, I-Cheng Ho5, Samia Khoury2, Mohamed Oukka1 &Vijay K Kuchroo2

Transcription factor Foxp3 is critical for generating regulatory T cells (Treg cells). Transforming growth factor-b (TGF-b) induces

Foxp3 and suppressive Treg cells from naive T cells, whereas interleukin 6 (IL-6) inhibits the generation of inducible Treg cells.

Here we show that IL-4 blocked the generation of TGF-b-induced Foxp3+ Treg cells and instead induced a population of T helper

cells that produced IL-9 and IL-10. The IL-9+IL-10+ T cells demonstrated no regulatory properties despite producing abundant

IL-10. Adoptive transfer of IL-9+IL-10+ T cells into recombination-activating gene 1–deficient mice induced colitis and

peripheral neuritis, the severity of which was aggravated if the IL-9+IL-10+ T cells were transferred with CD45RBhi CD4+ effector

T cells. Thus IL-9+IL-10+ T cells lack suppressive function and constitute a distinct population of helper-effector T cells that

promote tissue inflammation.

After encountering specific antigen, naive CD4+ T cells becomeactivated, expand their populations and differentiate into variouseffector T cell subsets, such as T helper type 1 (TH1), TH2 orinterleukin 17 (IL-17)-producing T helper (TH-17), characterized bydistinct patterns of cytokine secretion. These subsets of T cells havespecific effector functions and recruit different cell types at the site ofinflammation by virtue of the cytokines they produce. T-bet, GATA-3and RORgt are the transcription factors required for the generation ofTH1, TH2 and TH-17 cells, respectively1–3. Whereas TH1 cells clearintracellular pathogens and mediate autoimmune tissue inflamma-tion, TH2 cells are required for the clearance of extracellular patho-gens, and an exaggerated TH2 response induces asthma, allergy andatopy. In contrast, TH-17 cells seem to be involved in controlling bothintracellular and extracellular pathogens and in orchestrating auto-immune tissue inflammation4,5. As exaggerated responses by TH1,TH2 and TH-17 cells can induce tissue inflammation, the maintenanceof immune homeostasis and prevention of immunopathology requirestight regulatory mechanisms to control these effector T cells by subsetsof regulatory T cells (Treg cells). Various types of T cells have beendescribed that mediate these regulatory functions in vivo and preventautoimmunity and tissue inflammation. The CD4+CD25+Foxp3+ andIL-10-producing Treg type 1 (Tr1) cells are crucial in regulating effectorT cell functions6,7. Naturally occurring Treg cells, which are generated

in the thymus and express the transcription factor Foxp3 (A002750),inhibit effector T cells and are crucial in the maintenance of peripheraltolerance8. In addition, emerging data suggest that Foxp3+ T cells canalso be generated in the peripheral immune compartment9–11. How-ever, the in vivo factors that generate inducible Foxp3+ Treg cells in theperiphery are still unclear.In vitro, transforming growth factor-b (TGF-b; A002271), together

with the T cell antigen receptor and costimulatory signals, can induceFoxp3 and generate Treg cells from naive T cells12. TGF-b seems to be acrucial factor in this process, as mice deficient in TGF-b1 have a defectin the maintenance Foxp3+ Treg cells and develop a lethal lympho-proliferative disorder characterized by enhanced TH1 and TH2 effectorresponses. In addition, Foxp3+ T cells from mutant ‘FILIG’ mice,which have attenuated expression of Foxp3, produce more TH2cytokines13; this indicates that Foxp3 negatively regulates the TH2developmental program. How Foxp3 regulates TH2 responses, whetherit is cell intrinsic (in which Foxp3 suppresses induction of TH2-specifictranscription factors) or whether Treg cells suppress TH2 cells indirectlyhas not been tested directly. Initial reports suggest that Foxp3 can bindto the TH2 lineage–specific transcription factor GATA-3 (refs. 14,15),but the consequences of this interaction are not well understood. Oneoutcome of this interaction, however, is that Foxp3 can directly inhibitthe development of TH2 cells.

Received 14 July; accepted 14 October; published online 9 November 2008; doi:10.1038/ni.1677

1Center for Neurologic Diseases, Brigham and Women’s Hospital, Harvard Medical School, Cambridge, Massachusetts 02139, USA. 2Center for Neurologic Diseases,Brigham and Women’s Hospital and 3Transplant Research Center, Beth Israel Hospital, Harvard Medical School, Boston, Massachusetts 02115, USA. 4Palo Alto Veteran’sAdministration Health Care System and Department of Pathology, Stanford University School of Medicine, Palo Alto, California 94304, USA. 5Division of PediatricHematology-Oncology, Children’s Hospital, Department of Pediatric Oncology, Harvard Medical School, Boston, Massachusetts 02115, USA. 6These authors contributedequally to this work. Correspondence should be addressed to M.O. (moukka@rics.bwh.harvard.edu) or V.K.K. (vkuchroo@rics.bwh.harvard.edu).

NATURE IMMUNOLOGY VOLUME 9 NUMBER 12 DECEMBER 2008 1347

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

In addition to Foxp3+ Treg cells, Tr1 cells can suppress effector T cellfunction and tissue inflammation independently of Foxp3 expression.Tr1 cells produce mainly IL-10 (A001243) and suppress immuneresponses in a contact-independent way through the production ofIL-10. Although IL-10 was initially shown to induce Tr1 differentiation,the Tr1 cells thus generated do not proliferate because of the suppres-sive nature of IL-10. That observation suggests that factors other thanIL-10 might be important for the differentiation and populationexpansion of IL-10-secreting Tr1 cells. Several papers have demon-strated that IL-27 together with TGF-b induces the differentiation ofTr1 cells16,17. The IL-10-producing Tr1 cells thus generated have potentimmunosuppressive properties in that they not only inhibit autoim-mune disease and tissue inflammation but also contribute to thesuccessful engraftment of HLA-mismatched bone marrow transplants6.However, IL-10-producing T cells are not always anti-inflammatory,and in some circumstances they have been shown to promoteautoimmunity18,19. It has not yet been elucidated whether theseIL-10-producing cells also produce other cytokines that negate the anti-inflammatory function of IL-10, nor have the differences between thesepro- and anti-inflammatory IL-10-producing cells been determined.

In this study we show that IL-4, a TH2-differentiating cytokine,inhibited TGF-b-induced Foxp3 expression and thus suppressed thenew generation of Foxp3+ Treg cells. Unexpectedly, the two cytokinestogether (TGF-b plus IL-4) induced the generation of mainly Foxp3–

IL-9+IL-10+ T cells that, unlike Tr1 cells, did not suppress T cellresponses even though they produced IL-10. More unexpectedly, theIL-9+IL-10+ T cells not only induced the development of colitis butalso induced peripheral neuritis when transferred together withCD45RBhiCD4+ effector T cells into mice deficient in recombination-activating gene 1 (RAG-1). IL-9 was initially cloned as a T cell growthfactor whose receptor shares the common g-chain with IL-2, IL-4,IL-7, IL-15 and IL-2120, and it has shown to be involved in TH2-typeimmunity against helminth infection21. Our data suggest that T cellsdifferentiating in the presence of IL-4 and TGF-b undergo a uniquedevelopmental program characterized by the production of IL-10 andIL-9, which may not only subvert the immunosuppressive function ofIL-10-producing cells but also promote tissue inflammation, probablyby regulating the population expansion and/or function of othereffector T cells.

RESULTS

IL-4 inhibits TGF-b-induced Foxp3+ T cells

TGF-b induces the new generation of Foxp3+ inducible Treg cells fromnaive CD4+ T cells. Both IL-6 and IL-27 inhibit TGF-b-mediatedFoxp3 induction12,16. To test the effect of IL-4 on the induction ofFoxp3 by TGF-b, we used green fluorescent protein–negative (GFP–)naive CD4+ T cells from mice with a bicistronic enhanced GFPreporter cloned into the endogenous Foxp3 locus (Foxp3-GFPmice)12. As expected, stimulation of the T cell antigen receptor inthe presence of TGF-b induced Foxp3-GFP expression in up to 80% ofT cells, and this was inhibited (to 22%) in the presence of IL-4(Fig. 1a); thus, IL-4 actively inhibits the induction of Foxp3 in thepresence of TGF-b. To understand the mechanism by which IL-4mediates the inhibition of Foxp3 expression, we assessed the cytokinesproduced by T cells activated in the presence of TGF-b plus IL-4.Naive CD4+ T cells stimulated in the presence of IL-4 differentiatedinto TH2 cells and produced IL-10, IL-5, IL-13 and very little IL-9. Thecombination of TGF-b plus IL-4 inhibited the production of IL-4,IL-13 and IL-5 while sparing IL-10 production by these T cells.Furthermore, activation in the presence of TGF-b plus IL-4 alsoinduced the secretion of large amounts of IL-9 (Fig. 1b,c and

Supplementary Fig. 1 online). Notably, T cell populations expandedin the presence of TGF-b plus IL-4 produced as much IL-10 as TH2-differentiated T cells did, but instead of IL-4 but they produced IL-9together with IL-10 (Fig. 1b and Supplementary Fig. 2a online). Todetermine whether antigen-presenting cells (APCs) contributed to thegeneration of this unique subset of IL-9+IL-10+ cells or whether IL-9was being produced by APCs, we also activated naive CD4+ T cellswith plate-bound antibody to CD3 (anti-CD3) and anti-CD28 (with-out any APCs) in the presence of TGF-b together with IL-4. Even inthe absence of APCs, IL-4 inhibited the generation of TGF-b-inducedFoxp3+ T cells and induced T cells that co-produced IL-9 and IL-10(Supplementary Fig. 2a).

To characterize the source of IL-10 production at the cellular level,we did intracellular cytokine staining and compared T cells activatedin the presence of TGF-b, IL-4 or both. Activation of naive T cells inthe presence of IL-4 did not induce Foxp3-GFP expression but didinduce the production of intracellular IL-4 and IL-10, as expected, as apart of TH2 development. We found no production of either inter-feron-g (IFN-g) or IL-17 by responding T cells. However, the additionof IL-4 to TGF-b inhibited Foxp3 expression and induced mainlyproduction of IL-10 from these cells with much less production ofTH2 cytokines, including IL-4 (Fig. 1c). We could not stain for IL-9 atthe single-cell level because of the lack of appropriate antibodyreagents. Therefore, we could not establish the coexpression of IL-10and IL-9 at the cellular level in the T cells differentiated by TGF-b plusIL-4. However, these T cells produced both IL-9 and IL-10, as assessedby enzyme-linked immunosorbent assay (ELISA) and quantitativereal-time PCR (Fig. 1b and Supplementary Fig. 1b–d).

To determine whether IL-9 production was dependent on IL-10 inT cells differentiated with TGF-b plus IL-4, we obtained naive T cellsfrom mice with a GFP reporter cloned into the Il10 locus (IL-10–GFPreporter mice)22 and differentiated the cells in the presence of TGF-bplus IL-4. Then we sorted IL-10–GFP+ and IL-10–GFP– T cells by flowcytometry on the basis of GFP expression. We detected IL-9 mRNAexpression in IL-10–GFP+ and IL-10–GFP– cells (SupplementaryFig. 2a), although we found maximum expression only whenIL-10–GFP+ and IL-10–GFP– cells were together. These T cellsdifferentiated with TGF-b plus IL-4 produced both IL-9 and IL-10and we did not detect any double-producing IL-10+IL-17+ orIL-10+IFN-g+ cells (Supplementary Fig. 2b). However, a few double-producing IL-4+IL-10+ cells (about 8%) were present among the cellsactivated in the presence of TGF-b plus IL-4, especially whenT cells were activated in presence of small amounts of TGF-b.‘Titration’ of TGF-b had an important effect on IL-4 production, asthe percentage of IL-4+IL-10+ cells decreased with the increasingamounts of TGF-b (from 1 to 5 ng/ml). The double-producingIL-4+IL-10+ cells probably represented a transitional state, as IL-4production was extinguished with increasing amounts of TGF-b(Supplementary Fig. 2b). These results show that like IL-6 andIL-27, IL-4 is also able to inhibit Foxp3 induction and induceproduction of mainly IL-9 and IL-10 from the responding T cells.

Studies have shown that IL-10-producing Tr1 cells do not prolifer-ate well and seem to acquire an anergic phenotype. However, IL-9+

IL-10+ T cells induced in the presence of TGF-b plus IL-4 were notanergic and vigorously proliferated when activated by anti-CD3(Fig. 1d). We did not detect expression of T-bet, GATA-3, Foxp3 orRORgt (lineage-specific transcription factors necessary for the gen-eration of TH1, TH2, Treg and TH-17 cells, respectively) in the T cellsdifferentiated with TGF-b plus IL-4 (Fig. 1e). TGF-b may act togetherwith other cytokines to induce new T cell subsets23, as exemplified bythe finding that TGF-b in combination with IL-6 and IL-21 or with

1348 VOLUME 9 NUMBER 12 DECEMBER 2008 NATURE IMMUNOLOGY

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

IL-27 induced TH-17 and Tr1 cells, respectively. Similar to TGF-b andIL-6, which cross-regulate each other’s function and give rise to aunique TH-17 subset, TGF-b in combination with IL-4 inhibited theexpression of transcription factors specific for TH1, TH2, TH-17 andTreg cells (Fig. 1e).

IL-4 inhibition of Foxp3 induction depends on STAT6

To understand the function of IL-4 in the active inhibition of TGF-b-induced Foxp3, we cultured naive T cells from IL-4-transgenic mice inthe presence of various concentrations of TGF-b. Unexpectedly,TGF-b, even at higher doses, did not induce the generation ofFoxp3+ Treg cells (Fig. 2a), which suggested that IL-4 may be domi-nant in inhibiting the induction of Foxp3 expression. After binding toits receptor, IL-4 induces several signaling pathways, including those ofinsulin receptor substrate 2 and STAT6 (Signal Transducer and

Activator of Transcription 6), which form a signaling complex inassociation with the cytoplasmic tail of the IL-4 receptor24. However,STAT6 is the dominant pathway of TH2 differentiation because itinduces GATA-3, the ‘master’ transcription factor for TH2 differentia-tion. To assess whether STAT6 was involved in inducing IL-4-mediatedFoxp3 inhibition, we crossed STAT6-deficient (STAT6-knockout(STAT6-KO)) mice with the Foxp3-GFP ‘knock-in’ mice to generatea ‘STAT6-KO.Foxp3-GFP’ strain.

We purified naive CD4+ T cells from Foxp3-GFP and STAT6-KO.Foxp3-GFP mice and cultured the cells in the presence ofTGF-b or IL-4 or a combination of these two cytokines. UnlikeFoxp3-GFP T cells, in which IL-4 inhibited the induction of Foxp3,STAT6-KO.Foxp3-GFP T cells had higher expression of Foxp3 (80%)even in the presence of TGF-b plus IL-4 (Fig. 2b). Furthermore, whenwe analyzed the cytokines produced by the STAT6-deficient cells

0

20

40

60

80

100

120

Fox

p3-G

FP

+

cells

(%

)

a

80.1 22.7 0.25

No cytokine TGF-β TGF-β + IL-4 IL-4

Foxp3-GFP

Foxp3-GFP

3.23

02468

10121416

IL-1

0 (n

g/m

l)

0200400600800

1,0001,200

IL-5

(pg

/ml)

0

1

2

3

4

5

IL-1

3 (n

g/m

l)

00.51.01.52.02.53.0

IL-9

(ng

/ml)

02468

10121416

Cyt

okin

e (n

g/m

l )

IL-1

0IL

-5 IL-6

IL-1

3IF

N-γTNF

b

17.4 0.92

0.2481.4

3.92 2.13

2667.9

46.9 0.27

0.09152.7

20.9 0.21

0.4378.5

6.37 0.77

0.3692.5

4.2 1.75

27.466.6

1.88 0.11

0.1197.9

2.12 0.11

0.4397.3

19.9 0.24

0.1179.7

3.3 0.64

24.671.5

24.2 0.068

0.01975.8

9.8 0.04

0.2589.9

40.9 0.15

0.2658.7

10.3 7.77

18.363.7

6.47 0.044

0.2993.2

3.51 0.043

0.06196.4

No cytokine TGF-β TGF-β + IL-4 IL-4

IL-4

IL-17

IL-10

IFN-γ

c

0

5

10

15

20

25

Pro

lifer

atio

n (1

04 c.p

.m.)

No cy

tokin

e

IL-4

TGF-β

TGF-β + IL-4

d

0

1

2

3

4

5

6

7

0

1

2

3

4

5

6

7

8

0

50

100

150

200

250

0

1

2

3

4

5

6

7

8

T-be

t exp

ress

ion

(rel

ativ

e)

Fox

p3 e

xpre

ssio

n (r

elat

ive

× 10

3 )

GAT

A-3

exp

ress

ion

RO

Rγ t

exp

ress

ion

No cyto

kine

IL-4

TGF-β

TGF-β + IL-4

No cyto

kine

IL-4

TGF-β

TGF-β + IL-4

No cytokin

eIL-

4

TGF-β

TGF-β + IL-4

No cy

tokin

e

TGF-β

TGF-β + IL-4

T H1

No cy

tokin

e

IL-4

TGF-

βTG

F-β +

IL-4

No cy

tokin

eIL

-4

TGF-β

TGF-β + IL-4

No cy

tokin

eIL

-4

TGF-β

TGF-β + IL-4

No cy

tokin

e

IL-4

TGF-β

TGF-β + IL-4

IL-4

TGF-β

TGF-β + IL-4

T H-1

7

e

TGF-β + IL-4

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

104

104

103

103

102

102

101

101

100

100

Figure 1 IL-4 inhibits TGF-b-induced Foxp3+ T cells. Analysis of naive CD4+Foxp3–CD62L+ T cells from Foxp3-GFP mice activated with or without cytokines

in the presence of irradiated syngenic APCs and anti-CD3 (1 mg/ml). (a) Flow cytometry of Foxp3-GFP expression in CD4+ T cells 72 h after activation in the

presence of TGF-b, IL-4, neither or both. Numbers above bracketed lines (right) indicate percent GFP+ cells. (b) ELISA or bead array analysis of cytokines in

48-hour culture supernatants. TNF, tumor necrosis factor. (c) Flow cytometry of the intracellular expression of IFN-g, IL-4, IL-17 and IL-10 in CD4+ T cells at

day 4 after activation. Numbers in quadrants indicate percent cells in each. (d) Proliferation of cytokine-treated wild-type T cells, assessed as incorporation of

[3H-thymidine]. (e) Quantitative RT-PCR analysis of transcription factors, presented as mRNA expression relative to that of Gapdh (encoding glyceraldehyde

phosphate dehydrogenase). TH1 T cells were generated in the presence of IL-12 and anti-IL-4. TH-17 T cells were differentiated in the presence of TGF-b and

IL-6. Data represent one of three (a,c,d) or two (b,e) experiments (mean and s.d. of triplicate wells, d).

NATURE IMMUNOLOGY VOLUME 9 NUMBER 12 DECEMBER 2008 1349

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

stimulated in the presence of TGF-b plus IL-4, we found that IL-10and IL-9 were not induced, but these cells continued to express Foxp3in the presence of TGF-b plus IL-4 (Fig. 2c,d). As GATA-3 is ‘down-stream’ of STAT6, we also analyzed the induction of IL-9 and IL-10 byTGF-b plus IL-4 in GATA-3-deficient T cells and found that, similarlyto the STAT6-deficient cells, GATA-3-deficient T cells could notproduce IL-9 or IL-10 upon differentiation in the presence of TGF-b plus IL-4 (Fig. 2e). The effect of STAT6 and GATA-3 on theinduction of IL-9+IL-10+ T cells may be indirect rather than direct,as in the absence of STAT6–GATA-3 signaling, TGF-b-induced Foxp3could not be antagonized by IL-4, resulting in the induction of mainlyFoxp3, which may not allow the differentiation of T cells to produceIL-9 and IL-10. These data collectively suggest that the inhibition ofFoxp3 expression by IL-4 is mediated through the IL-4–STAT6 path-way. Thus, the new generation of inducible Foxp3+ Treg cells isprobably modulated by the IL-4 through the activation of theSTAT6–GATA-3 pathway. We next explored whether the IL-4–STAT6pathway affected the function of already established, naturally occur-ring Foxp3+ Treg cells. When we cultured sorted Foxp3– T effector cellswith either Foxp3-GFP+ or STAT6-KO.Foxp3-GFP+ Treg cells, Foxp3-GFP+ Treg cells lost much of their suppressive ability in the presence ofIL-4, whereas STAT6-KO.Foxp3-GFP+ Treg cells still remained partly

suppressive despite the presence of IL-4 (40% versus 70% suppression;Fig. 2f). Thus, the IL-4–STAT6 pathway not only affects the genera-tion of induced Treg cells but also regulates the suppressive capacity ofnaturally occurring Treg cells by a mechanism intrinsic to Treg cells.

Relationship between and Foxp3 and GATA-3

To assess whether GATA-3 affects Foxp3 expression, we isolated CD4+

T cells from GATA-3 conditional knockout mice and activated the cellsin the presence of TGF-b. On day 4 after activation, we collected T cellsand extracted mRNA for analysis of the expression of GATA-3 andFoxp3 by quantitative PCR. TGF-b-induced Foxp3 mRNA expressionwas threefold higher in GATA-3-deficient CD4+ T cells than in wild-type CD4+ T cells. However, the addition of TGF-b resulted in lessGATA-3 expression in wild-type CD4+ T cells (Fig. 3a), which sugges-ted possible inter-regulation between these two transcription factors.

Because of the correlation between the expression of Foxp3 andGATA-3, we sought to determine whether the two transcriptionfactors could physically interact and associate with each other. Weused two different approaches to test those possibilities. In the firstapproach, we transfected human embryonic kidney cells (HEK293cells) with Flag-tagged GATA-3 and Renilla luciferase–tagged Foxp3cDNA. After lysing the cells, we immunoprecipitated GATA-3 (or a

WTIL-4-Tg

α-CD3 +α-CD28

0.5 1.0 5.00

10

20

30

40

TGF-β (ng/ml)

Fox

p3 m

RN

A (

rela

tive

to G

apdh

)

a

1.74 3.9680.81.379.789.3

No cytokine TGF-β TGF-β + IL-4 TGF-β + IL-4IL-4 IL-4

b

104

104

103

103

102

102

101

101

100

100

104

103

102

101

100

1.01 0.2

2.2196.6

2.87 4.21

59.233.7

16.5 1.43

5.5876.5

31.3 0.22

0.08568.4

0.73 0.64

41

2.14 0.26

Foxp3-GFP

Foxp3-GFP

No cytokine TGF-β TGF-β + IL-4 TGF-β + IL-4IL-4 IL-4

IL-1

0

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

104

103

102

101

100

STAT6-KO. Foxp3-GFPFoxp3-GFPc

0

250

500

1,000

1,500

2,000

2,500

STAT6-

KO.

Fox3

-GFP

IL-9

(pg

/ml)

0123456789

10

103Fo

xp3-

GFP

Foxp

3-GFP

STAT6-

KO.

Fox3

-GFPd

STAT6-KO. Foxp3-GFPFoxp3-GFP

500

1,000

1,500

2,000

2,500

0 0

1

2

3

4

5

IL-9

(ng

/ml)

I19

mR

NA

(rel

ativ

e to

Hpr

t × 1

02 )e

0

500

1,000

1,500

2,000

2,500

Il10

mR

NA

(rel

ativ

e to

Hpr

t × 1

02 )

Foxp

3-GFP

STAT6-

KO.

Fox3

-GFP

Foxp

3-GFP

STAT6-

KO.

Fox3

-GFP

Foxp

3-GFP

STAT6-

KO.

Fox3

-GFP

0100200300400500600700800

0102030405060708090

100

STAT6-KO.Foxp3-

GFP Treg

T eff

T eff + T reg

T eff + T reg

T eff +

T reg + IL

-4

T eff +

T reg + IL

-4

T eff + T reg

+ IL-4

T eff + T reg

+ IL-4

T eff +

T reg

T eff +

T reg

Foxp3-GFP Treg

Foxp3-GFP Treg

STAT6-KO.Foxp3-

GFP Treg

Pro

lifer

atio

n (1

02 c.p

.m)

Sup

pres

sion

(%

)

f

I19

mR

NA

(rel

ativ

e to

Hpr

t × 1

02 )

No cy

tokin

e

TGF-β + IL-4

No cy

tokin

e

No cy

tokin

e

No cy

tokin

e

TGF-β + IL-4

TGF-β + IL-4

TGF-β + IL-4

IL-4

IL-4

2.639557.6

Figure 2 The inhibition of Foxp3 induction by IL-4 is STAT6 dependent. (a) RT-PCR analysis of

the induction of Foxp3 mRNA, relative to Gapdh mRNA, in wild-type T cells (WT) and IL-4-

transgenic T cells (IL-4-Tg) treated with anti-CD3 and anti-CD28 (a-CD3 + a-CD28) alone or

with various doses (horizontal axis) of TGF-b. (b,c) Flow cytometry of Foxp3-GFP expression at

72 h (b) and of intracellular Foxp3-GFP and IL-10 expression at day 4 (c) by naive Foxp3-GFP

and STAT6-KO.Foxp3-GFP CD4+Foxp3–CD62L+ T cells activated with or without cytokines in

the presence of irradiated syngenic APCs and anti-CD3 (1 mg/ml). Numbers above bracketed

lines indicate percent GFP+ cells (b) and numbers in quadrants indicate percent cells in

each (c). (d) RT-PCR analysis of the expression of Il9 mRNA, relative to Hprt mRNA, at day 4

(left) and ELISA of the production of IL-9 protein at day 2 (right) by the cells in b,c. (e) RT-PCR

analysis of the expression of Il9 mRNA (left) and Il10 mRNA (right), relative to Hprt mRNA, innaive wild-type (WT) and GATA-3-deficient, Cre+ (GATA-3-KO) CD4+CD62L+ T cells activated with

or without cytokines in the presence of irradiated syngenic APCs and anti-CD3 (1 mg/ml),

presented relative to the expression of Hprt1 (encoding hypoxanthine guanine phosphoribosyl

transferase). Middle, ELISA of IL-9 production. (f) Proliferation (left) and suppression assay

(right) of naive STAT6-deficient responder cells cultured in presence of various cytokines

(horizontal axis) with Treg cells purified from wild-type or STAT6-deficient mice (left). Proliferation is presented as mean and s.d. of triplicate wells;

suppression is presented relative to the proliferation of effector T cells (Teff) alone. Data represent one of two experiments.

1350 VOLUME 9 NUMBER 12 DECEMBER 2008 NATURE IMMUNOLOGY

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

control protein) with anti-Flag and assessed Foxp3 in the immuno-precipitates by monitoring the expression of Renilla luciferase byluminescence. Precipitation of GATA-3 resulted in much more luci-ferase activity than did negative control (empty vector) or anothertranscription factor, such as STAT6 (Fig. 3b), which suggested thatGATA-3 precipitated luciferase-tagged Foxp3. We further confirmedthose results in the second approach, using a coimmunoprecipitationassay. For this, we transfected cells with Flag-tagged GATA-3 togetherwith Myc-tagged Foxp3. After lysing the cells, we immunoprecipitatedthe lysates with beads coated with anti-FLAG. To detect whetherFoxp3 precipitated together with GATA-3, we separated the samplesby SDS-PAGE and, after transfer, blotted the membrane with anti-Myc. GATA-3 immunoprecipitated together with Foxp3 (data notshown), which demonstrated direct association of the two proteins.These assays show that GATA-3 and Foxp3 physically associate witheach other.

Functional interaction between Foxp3 and GATA-3

Next we sought to determine the functional importance of theinteraction reported above by assessing whether Foxp3 affected TH2differentiation. We cloned Foxp3 into a retroviral vector (mouse stemcell virus) and transduced this vector into activated CD4+ T cellsdifferentiated in TH2-polarization conditions (10 ng/ml IL-4 and10 mg/ml anti-IL-12). We then sorted T cells expressing Foxp3 byflow cytometry, extracted RNA and analyzed the expression of variouscytokines. In parallel, we maintained a cohort of cells from eachtransduction condition in culture for 2–3 d, restimulated them withplate-bound anti-CD3 and anti-CD28 and collected the supernatantsfor cytokine analysis by ELISA.

Whereas T cells transduced with empty retroviral vector had highexpression of Il10, Il13 and Il4 mRNAs and modest expression of Il5,consistent with T cells differentiated into TH2 effector cells,T cells transduced with Foxp3 had much less expression of Il4, Il15and Il13 but retained substantial expression of Il10. Similarly, at theprotein level, T cells transduced with control vector produced IL-4 andsmall amounts of IL-2, whereas Foxp3-transduced T cells producedless IL-4 and more IL-10 (Fig. 3c). This was consistent with our initialobservation that activation of T cells in the presence of IL-4 plus

TGF-b resulted in the inhibition of production of IL-4, IL-5, IL-13 butspared IL-10 production by these cells. This would suggest that Foxp3interferes with the expression of genes controlled by GATA-3, pre-sumably by binding or interfering with its ability to transactivatespecific TH2 genes. In particular, these results suggest that theinteraction of Foxp3 with GATA-3 inhibits its ability to transactivatepromoters that are under the control of GATA-3, such as those ofgenes encoding IL-4, IL-5 and IL-13 but not IL-10, which is mostlikely controlled by a different transcription factor.

To address our hypothesis, we analyzed the ability of Foxp3 toinhibit GATA-3-mediated transactivation of an Il5 promoter–lucifer-ase construct. We transfected Jurkat cells with GATA-3, Foxp3 or bothGATA-3 and Foxp3 in the presence of the Il5 promoter–luciferaseconstruct and measured luciferase activity after activating the cellswith the phorbol ester PMA plus ionomycin. As expected, GATA-3strongly transactivated activity of the Il5 promoter–luciferase con-struct, whereas this effect was 50% lower when Foxp3 was coexpressed(Fig. 3d). These results collectively show that Foxp3 can directlyinteract with GATA-3 and that this association inhibits GATA-3-mediated transactivation of Il5, which is one of its target genes.

Function of IL-9+IL-10+Foxp3– T cells

As IL-10-producing Tr1 cells have been shown to suppress T cellproliferation in vitro and to inhibit tissue inflammation in vivo25, wewanted to determine whether naive T cells differentiated in thepresence of TGF-b plus IL-4 that produced IL-9 and IL-10 were alsosuppressive. Although Foxp3+ Treg cells were anergic and inhibited theproliferation of effector T cells in vitro, T cells generated in thepresence of TGF-b plus IL-4 were not anergic or suppressive, asthey vigorously proliferated on their own and, together with effectorT cells, further enhanced T cell proliferation (Fig. 4a). The addition ofneutralizing antibodies to IL-9 or IL-10 to the cultures of effectorT cells with IL-10–GFP+ T cells generated by TGF-b plus IL-4 resultedin less proliferation but did not completely abolish it (Fig. 4a). Thesedata suggest that unlike Foxp3+ Treg cells and IL-10-producing Tr1cells, T cells induced by TGF-b plus IL-4 do not have any suppressiveeffects in vitro but readily proliferate and thus act like effector T cells.

It is apparent not only that Tr1 cells suppress proliferation in vitrobut also that IL-10-producing cells can suppress the induction ofcolitis in vivo26,27. Therefore, to further characterize the IL-9+IL-10+

T cells generated in the presence of both TGF-b and IL-4, we assessedtheir function in vivo in a T cell transfer model of colitis. We purifiednaive T cells from IL-10–GFP reporter mice and activated the cells in

0

50

100

150

200

250

TGF-βCre – – +

GAT

A-3

(rel

ativ

e ex

pres

sion

× 1

03 )

a

– + +

02468

101214161820

TGF-βCre –

+ ++

Fox

p3

(rel

ativ

e ex

pres

sion

× 1

03 )

0

5

10

15

20

25

30

35

40

Luci

fera

se

(nor

mal

ized

exp

ress

ion)

GATA-3 STAT6 pCDNA

b

02468

101214161820

IL-10 IL-13 IL-4 IL-5

Control

Foxp3

mR

NA

(rel

ativ

e ex

pres

sion

× 1

02 )

c

Foxp3 +– –

0

2

4

6

8

10

12

IL-5–LucGATA-3

+ + ++ +–

Luci

fera

se

(nor

mal

ized

exp

ress

ion)

d

0

1

2

3

4

5

6

7

IL-4 IL-10 IL-2

Pro

tein

(ng

/ml)

Control

Foxp3

Figure 3 Relationship between and Foxp3 and GATA-3. (a) RT-PCR analysis

of the expression of Foxp3 and GATA-3 in CD4+ T cells expressing or lacking

GATA3, treated for 48 h with TGF-b (3 ng/ml) to induce Foxp3 expression,

presented relative to Hprt1 expression. (b) Luciferase assay of HEK293 cells

transfected with Renilla luciferase–tagged Foxp3 and plasmids expressing

Flag-tagged GATA-3 or STAT6 or control vector (pCDNA) and

immunoprecipitated with beads coated with anti-Flag, normalized for

transfection efficiency. (c) RT-PCR analysis of the expression of cytokinemRNA in CD4+ T cells activated in TH2-polarizing condition and transduced

with control retroviral vector or Foxp3-GFP vector, assessed in Foxp3-GFP+

cells sorted 48 h later and presented relative to Hprt1 expression (left), and

ELISA of cytokines in supernatants of the Foxp3-GFP-transduced T cells

described above, cultured for 3 more days in presence of IL-2 and then

restimulated for 48 h with plate-bound anti-CD3 and anti-CD28 (right).

(d) Luciferase assay of Jurkat cells transfected for 24 h with an Il5

promoter–luciferase construct (IL-5–Luc) and plasmid(s) expressing Foxp3

or GATA-3 or both, then activated for 3–4 h with PMA plus ionomycin,

normalized for transfection efficiency. Data represent one of two experiments

(a) or are representative of three experiments (b–d).

NATURE IMMUNOLOGY VOLUME 9 NUMBER 12 DECEMBER 2008 1351

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

the presence of TGF-b and IL-4. At day 4, we used flow cytometry tosort the IL-10–GFP+ T cells treated with TGF-b plus IL-4 (T4–IL-10–GFP+ T cells) on the basis of GFP expression. We reconstituted RAG-1-deficient mice with effector T cells (Foxp3–CD45RBhi T cells), witheffector T cells plus T4–IL-10–GFP+ T cells, with only T4–IL-10–GFP+

T cells, or with effector T cells plus naturally occurring Treg cells, andthen assessed the development of colitis by monitoring the bodyweight of the mice and assigning scores for colitis as describedbefore28. We also did histopathological analysis to evaluate theinduction and severity of the disease. All mice reconstituted witheffector T cells did not lose much weight but developed severe colitis,as noted by histopathology, with an average colitis score of 7.5 ± 3.5(Fig. 4b). As a control to show that effector T cells can indeed beregulated, we demonstrated that mice that received both effectorT cells and naturally occurring Treg cells were protected from bothclinical and histological disease (average colitis score, 3.7 ± 1.5).Notably, mice that received T cells cultured in the presence of TGF-b plus IL-4 lost weight and developed a more moderate colitis (averagecolitis score, 5.7 ± 2.6). Furthermore, when we transferred effectorT cells together with IL-9+IL-10–GFP+ T cells generated in thepresence of TGF-b and IL-4, the mice lost the most weight andhistopathology showed the development of a more severe colitis thanthat of mice given effector T cells alone (average colitis score, 8.8 ± 2.3;Fig. 4c). Unexpectedly, however, mice receiving either solely T4–IL-10–GFP+ T cells or both effector T cells (CD45RBhi) and T4–IL-10–GFP+ T cells also showed signs of clinical neurological diseasecharacterized by impaired motor coordination with flaccid monopar-esis or asymmetric paraparesis of the hind limbs. Muscle strength andtone in forelimbs and tail did not seem to be affected. These clinicalsigns differed from the typical manifestations of experimental auto-immune encephalitis (characterized by a limp tail and ascending

paralysis) and suggested instead a peripheral neuropathy. Consistentwith the clinical appearance, histopathological analysis showed evi-dence of peripheral neuritis in mice that received both effector T cellsand T4–IL-10–GFP+ T cells or only T4–IL-10–GFP+ T cells. However,consistent with the clinical results, we found no inflammation in thebrain or spinal cord (Fig. 4c). At weeks 8–9 after transfer and after thedevelopment of colitis, we killed mice and retrieved cells from lymphnodes to assess cytokine production by intracellular cytokine staining.T cells from the RAG-1-deficient mice that received only effector(CD45RBhi) T cells produced mostly IFN-g and very little IL-17(Supplementary Fig. 3a online). However, T cells from the micereceiving both effector (CD45RBhi) T cells and T4–IL-10–GFP+ T cellsproduced both IL-17 and IFN-g, but the production of IL-17predominated over that of IFN-g in these cells. And although wenoted expression of both IL-17 and IFN-g in IL-10–GFP+ and IL-10–GFP– cells, over a third of the IL-10–GFP+ cells also produced bothIL-17 and IFN-g (Supplementary Fig. 3b). These data suggest thatT cells generated in the presence of TGF-b and IL-4 that produce bothIL-9 and IL-10 do not act as Treg cells but instead potentiate thefunctions of effector T cells, which results in additional tissueinflammation not induced by the transfer of effector T cells alone.

DISCUSSION

Here we have shown that when naive T cells were activated in thepresence of both IL-4 and TGF-b, Foxp3 expression was inhibited anda subset of T cells was generated that selectively produced IL-10 andIL-9. Although IL-10-producing cells are known to have regulatoryproperties, the IL-9–IL-10–producing cells described here did notregulate but instead induced tissue inflammation. IL-4 is involved inTH2 differentiation, and the observation that IL-4 receptor signals aretransduced by the transcription factors STAT6 and GATA-3 is well

Figure 4 Function of the IL-9+IL-10+ Foxp3–

T cells generated in presence of TGF-b and IL-4.

(a) Suppression assay with naive CD4+CD62L+

T cells from IL-10–GFP reporter mice left

unactivated or activated with IL-4 plus TGF-band untreated IL-10–GFP+ T cells; naive wild-

type Foxp3– responder T cells (Teff) were cultured

alone or together with naturally occurring Treg

cells (Treg) or T4–IL-10.GFP+ T cells, with or

without anti-IL-9 or anti-IL-10 (below graph), and

proliferation was assessed as [3H]thymidine

incorporation. Data represent one of three

experiments (mean and s.d. of triplicate wells).

(b) Weight loss of RAG-1-deficient mice

reconstituted with sorted CD4+Foxp3-GFP–

CD45RBhi T cells alone (Group 1; n ¼ 8 mice) or

in combination with T4–IL-10–GFP+ T cells

(Group 2; n ¼ 8 mice) or CD4+Foxp3+ naturally

occurring Treg cells (Group 4; n ¼ 6 mice), or

with only T4–IL-10–GFP+ T cells (Group 3; n ¼ 8

mice), presented as the average of the calculated

weight of all mice in each group. Data pooled

from two independent experiments. Pathology

scores: group 1, 7.5 ± 3.5; group 2, 8.8 ± 2.3;

group 3, 5.7 ± 2.6; and group 4, 3.4 ± 1.5.

P ¼ 0.01, group 2 versus group 4, and P ¼ 0.03,

group 3 versus group 4 (Mann-Whitney one-tailedU-test). (c) Histopathology of colitis and peripheral neuritis in mice in the groups described in b. Group 1, mononuclear cell inflammation in the lamina

propria, muscularis and adventitial tissues (arrow). Group 2, inflammation in the lamina propria and submuscularis (left; arrow) and blue staining of myelin

and spinal nerve root with inflammatory cell infiltration (middle) and myelin breakdown product in vacuoles (right; arrows). Group 3, inflammation in the

lamina propria, muscularis mucosa and adventitial tissues (left; arrow) and spinal nerve root with vacuolation with myelin breakdown products (right; arrows).

Group 4, intact mucosal architecture and muscularis (left), intact spinal cord white matter and nerve root (middle), and intact myelin in cauda equina nerve

root (right). Staining, Luxol fast blue plus hematoxylin and eosin. Scale bars, 200 mm (Group 1) or 100 mm (Group 2).

0

10

20

30

40

50

60a

b

c

T4–IL-10.GFP

+

85

90

95

100

105

0 1 2 3 4 5 6 7 8

Time (weeks)

CD45RBhi (Group 1)

CD45RBhi + Treg (Group 4)

CD45RBhi + T4–IL-10.GFP+ (Group 2)T4–IL-10.GFP+ (Group 3)

Wei

ght l

oss

(%)

T eff

T eff + T4–IL-10.G

FP+

T eff + T4–IL-10.G

FP+

+ α-IL-9

T eff + T4–IL-10.G

FP+

+ α-IL-10

T reg

T eff + T reg

Colon

Spinal nerve root

CD

45R

Bhi

(Gro

up 1

)

CD

45R

Bhi

+T

4–IL

-10

–G

FP

+

(Gro

up 2

)C

D45

RB

hi +

Tre

g(G

roup

4)

T4

–IL-

10–

GF

P+

(Gro

up 3

)

Pro

lifer

atio

n (1

04 c.p

.m.)

1352 VOLUME 9 NUMBER 12 DECEMBER 2008 NATURE IMMUNOLOGY

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

described. So far, no lineage-specific transcription factors have beenidentified for IL-10-producing Tr1 cells. It has been suggested thatGATA-3 may participate in the transcriptional regulation of IL-10.However, our data have demonstrated that IL-9–IL-10–producing cellsinduced in presence of TGF-b and IL-4 did not express GATA-3,which suggests that the presence of TGF-b could antagonize theinduction of GATA-3, yet these cells continued to produce IL-10.

STAT proteins are important in T cell differentiation29. STAT6 isinvolved in TH2 differentiation, and a STAT1-binding site has beenfound in the Il10 promoter29,30. In this study, we have shown that theinhibition of TGF-b-induced Foxp3+ T cells by IL-4 and the genera-tion of T cells IL-9+IL-10+ was partially reversed in STAT6-deficientand GATA-3-deficient mice, which suggests that STAT6 and GATA-3may be important in the generation of cells producing both IL-9 andIL-10. However, an alternative interpretation of our data is that in theabsence of IL-4 signaling in STAT6-deficient and GATA-3-deficientmice, there is mainly unopposed induction of Foxp3, which preventsthe differentiation of T cells to produce IL-9 and IL-10. The lack ofGATA-3 expression in IL-9–IL-10–producing cells induced by TGF-bplus IL-4 would partly support that interpretation.

In addition to IL-10, IL-9 was produced by T cells differentiated inthe presence of TGF-b plus IL-4, which suggests that TGF-b was notsimply ‘scaling back’ TH2 responses but instead that TGF-b and IL-4together were inducing a new gene program that induced IL-9production. We did not address whether the combined effects ofTGF-b and IL-4 induced another transcription factor not associatedwith either with TH2 or Treg cells. Notably, Foxp3 and GATA-3physically associated with each other; this finding is perhaps reminis-cent of the interaction between Foxp3 and RORgt, which reciprocallyregulates the induction of Treg or TH-17 cells12,31. Binding of Foxp3 toGATA-3 may regulate the molecular mechanism by which TGF-b andIL-4 mutually inhibit the generation of Treg and TH2 cells. However,this mutual regulation resulted in the production of IL-9 by thedifferentiating T cells. During the course of our studies, two othergroups reported that Foxp3 and GATA-3 physically associate with eachother14,15. Our results have confirmed those findings and have shownthat Foxp3 not only interacted with GATA-3 but also inhibitedthe ability of GATA-3 to transactivate the promoter of a TH2 targetgene, Il5.

TGF-b together with IL-27 can induce and expand populations ofIL-10-producing cells that also produce IFN-g and have strongimmunoregulatory properties16. IL-10 is a potent anti-inflammatorycytokine that inhibits TH1-, TH-17- and TH2-mediated immuneresponses32–34. The importance of IL-10 in regulating immuneresponses has been demonstrated in human infectious diseases suchas tuberculosis, malaria, leishmaniasis and schistosomiasis35–37. Inaddition to being produced by T cells, IL-10 can be produced bymany other cell types, including macrophages, dendritic cells, B cellsand natural killer cells. IL-10 production by T cells is important in thedownregulation of T cell population expansion during antigen-specificimmune responses. Although Tr1 cells produce considerable IL-10,they can also produce variable amounts of IFN-g and are able tosuppress in vitro T cell responses in a Foxp3-independent way. Suchfindings have led to the view that IL-10 produced by T cells isimportant in suppressing inflammation. Here we have shown thatIL-10-producing cells could be generated in large numbers when naiveT cells were cultured in the presence of IL-4 plus TGF-b; however,these T cells also produced IL-9, lacked regulatory properties andfailed to control inflammation when transferred in vivo. The produc-tion of IL-10 and TGF-b is associated with the maintenance ofperipheral tolerance by Treg cells25. In vivo, the absence of IL-10

exacerbates airway inflammation38. Such data indicate that the pro-duction of IL-10 by T cells may not always have suppressive functionsand that the context in which IL-10 is produced by T cells may beequally important in determining whether IL-10-producing T cells areanti-inflammatory or proinflammatory.

It is known that IL-10 is important in regulating tissue inflamma-tion in the gut. More specifically, it has been shown that IL-10produced by Treg cells is essential for this in vivo suppression, asIL-10-deficient Treg cells cannot ‘cure’ T cell transfer–inducedcolitis27,39. In our study, the IL-10-producing T cells generated inthe presence of TGF-b and IL-4 induced colitis and, when transferredtogether with effector T cells, not only failed to prevent the develop-ment of disease but also induced peripheral neuritis, a disease notinduced by effector T cells (CD4+CD45RBhi) alone. These data suggestthat the IL-9+IL-10+Foxp3– T cells generated in the presence of TGF-band IL-4 did not have suppressive functions and therefore represent aunique population of IL-10-producing T cells that promote inflam-mation. It is not clear at this stage whether IL-9 or IL-10 or bothcytokines is (are) involved in inducing tissue inflammation, but webelieve that IL-9 produced by these T cells may be a key factor inpromoting their proinflammatory activity. This is consistent with theobservation that IL-9 in the lung promotes TH2-mediated allergicpulmonary disease characterized by abundant TH2 cytokines40,41. Wealso have found that IL-9 receptors are expressed on effector T cells(data not shown) and that these cells therefore have the potential torespond to IL-9 and differentiate in the presence of IL-9+IL-10+ cells.As IL-9 belongs to the cytokine family of IL-2, IL-4, IL-7, IL-15 andIL-21 and uses the common g-chain to function, we believe that IL-9produced by T cells stimulated with TGF-b and IL-4 may function as agrowth and differentiation factor for other effector T cells. This, webelieve, is one of the explanations why IL-9+IL-10+Foxp3– T cells werenot able to control inflammatory responses but instead acted insynergy with effector cells to potentiate their proinflammatory activity.Consistent with that view were the data obtained with RAG-1-deficient recipients of a mixture of effector T cells and IL-9+IL-10+

T cells, which showed that the effector T cells produced IL-17 notnormally found when effector cells alone were transferred into RAG-1-deficient mice.

In addition to expression of the IL-9 receptor on effector T cells, ourpreliminary data have indicated that the IL-9 receptor is also expressedon naturally occurring Treg cells. Therefore, IL-9 could exert its effectson both effector T cells and Treg cells. Whether IL-9 inhibits orinterferes with the suppressive functions of regulatory cells has notbeen analyzed. However, IL-9 is also produced by Treg cells them-selves42; this IL-9 recruits mast cells to the tissue niche and regulateseffector T cells during graft rejection. Neutralization of IL-9 in vivo inthis setting abrogates Treg cell function and promotes graft rejection. Inour experiments, we tested the effects of IL-9–IL-10–producing T cellsin vivo in RAG-1-deficient mice in the absence of Treg cells; in theseexperimental conditions, the IL-9 produced by the IL-9–IL-10–produ-cing T cells may have acted only on the effector T cells and potentiatedtheir proinflammatory functions and thus induced more severeand unusual tissue inflammation (such as peripheral neuritis),not normally noted with the transfer of effector T cells alone.In conclusion, here we have identified a previously unknownsubset of T cells generated in presence of TGF-b and IL-4 thatproduced both IL-9 and IL-10, did not suppress T cell responsesand promoted inflammation. Thus, depending on the contextin which IL-10 is being produced by T cells, the T cells caninhibit inflammation and protect against autoimmunity, or they canpromote inflammation and mediate tissue destruction.

NATURE IMMUNOLOGY VOLUME 9 NUMBER 12 DECEMBER 2008 1353

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

METHODSMice. C57BL/6 mice, BALB/c mice, RAG-1-deficient (Rag–/–) mice and STAT6-

deficient (Stat6–/–) mice were from Jackson Laboratories. Foxp3-GFP mice12,

IL-4-transgenic mice43 and IL-10–GFP reporter mice22 have been described.

Mice with conditional GATA-3 deficiency (Gata3flox/flox) and IL-10–GFP mice

were provided by I.-C. Ho and R. Flavell, respectively. Mice were housed in

conventional, pathogen-free facilities at the Harvard Institute of Medicine. All

experiments were done in accordance with guidelines of the standing commit-

tee on animals at Harvard Medical School.

In vitro T cell differentiation, proliferation and suppression assays. For T cell

differentiation and proliferation, sorted naive CD4+CD62L+Foxp3-GFP–,

CD4+CD62L+GATA-3-KO (Cre+) or CD4+CD62L+STAT6-KO.Foxp3-GFP–

T cells were cultured with irradiated syngenic APCs in the presence of TGF-

b (3 ng/ml), IL-4 (20 ng/ml) or both. For suppression assays, 2.0 � 104 sorted

Foxp3+ Treg cells from wild-type or STAT6-deficient mice were cultured

together in triplicate wells at a ratio of 1:1 with naive wild-type CD4+ T cell

or naive STAT6-deficient CD4+ T cells (T cell responders) in the presence of

soluble anti-CD3 (1 mg/ml; BioXCell, clone 145-2C11) and irradiated syngenic

APCs. In some experiments, anti-IL-9 (10 mg/ml; BD Biosciences, clone

D9302C12) or anti-IL-10 (10 mg/ml; BD Biosciences, clone JES-16E3) was

added to the culture. Cells were pulsed with [3H]thymidine

(1 mCi/well) for the final 16 h of incubation and incorporation of [3H]thymi-

dine was measured with a microbeta liquid scintillation counter (PerkinElmer).

Intracellular cytokine staining. Cells previously stimulated with either soluble

anti-CD3 together with irradiated APCs or with plate-bound anti-CD3 and

anti-CD28 were restimulated for 4 h at 37 1C in 10% CO2 with PMA (phorbol

12-myristate 13-acetate; 50 ng/ml; Sigma), ionomycin (1 mg/ml; Sigma) and

GolgiStop (1 ml per 1.5 ml; BD Bioscience), followed by surface and intracel-

lular staining of various cytokines according to the manufacturer’s instructions

(BD Bioscience). Cells were analyzed on a flow cytometer (BD Biosciences).

In vivo T cell transfer–induced colitis. C57BL/6 RAG-1-deficient mice were

injected with sorted CD4+Foxp3-GFP–CD45RBhi T cell subpopulations in PBS

as described44,45. Mice received CD4+Foxp3-GFP–CD45RBhi T cells alone or in

combination with CD4+CD62L+ T cells from IL-10–GFP mice stimulated with

TGF-b and IL-4, or with naturally occurring CD4+Foxp3+ Treg cells. Weight loss

was monitored for 9 weeks. Mice were then killed and samples of colonic

lymphoid tissues as well as brain and spinal cord were fixed in neutral

buffered formalin.

Histopathology. Routinely processed, paraffin-embedded samples of small and

large intestine and other tissues were stained with hematoxylin and eosin.

Sections of brain and spinal cord were stained with Luxol fast blue plus

hematoxylin and eosin. The presence and severity of colitis and the presence of

brain and spinal cord inflammation was evaluated by researchers ‘blinded’ to

sample identity.

The severity of colitis was graded semiquantitatively for the following four

criteria on a scale of 0–3: degree of epithelial hyperplasia and goblet cell

depletion; leukocyte infiltration in the lamina propria; area of tissue affected;

and the presence of markers of severe inflammation, such as crypt abscesses,

submucosal inflammation and ulcers. Scores for each criterion were added to

give an overall inflammation score of 0–12 for each sample. The total colonic

score was calculated as the average of the individual scores of sections of

proximal colon, mid-colon and distal colon28.

Measurement of cytokines by ELISA and quantitative RT-PCR. Culture

supernatants were collected at 48 h to assess cytokine production by ELISA

or cytometric bead array (BD Bioscience). RNA was extracted with RNAeasy

columns (Qiagen) after 48–96 h of in vitro stimulation and was analyzed

by quantitative RT-PCR according to the manufacturer’s instructions

(Applied Biosystem).

Retroviral transduction of T cells with conditional deletion of GATA-3 and

TH2-polarized T cells. Total CD4+ T cells from wild-type mice or mice with

loxP-flanked Gata3 were transduced with a mouse stem cell virus–based

retroviral vector expressing Foxp3 or Cre recombinase, respectively. This

retroviral vector is a bicistronic vector encoding enhanced GFP, whose expres-

sion ‘reports’ the expression of the transgene. The retroviral particles were

produced by cotransfection of plasmids encoding group-associated antigen–

polymerase and the ecotropic envelope plus the retroviral vector into the

HEK293 cell line with FuGENE (Roche). Retroviral supernatants were collected

48–72 h after transfection and were centrifuged before use. Freshly isolated

CD4+ T cells purified by magnetic-activated cell sorting were activated with

plate-bound anti-CD3 (BioXCell, clone 145-2C11) and anti-CD28 (BioXCell,

clone PV-1) in the presence of mouse IL-4 (10 ng/ml; R&D Systems) and anti–

mouse IL-12 (10 mg/ml; BD Biosciences, clone C17.8) for TH2 differentiation.

After 24 h of activation, retroviral supernatant was added to the cells together

with polybrene (8 mg/ml). Plates were centrifuged for 45 min at 720g and were

then returned to 37 1C. TGF-b (3 ng/ml) was added to some samples to induce

Foxp3 expression. After 48 h of incubation at 37 1C, GFP+ cells were purified by

flow cytometry and their populations were expanded for an additional 48–72 h

in the presence of IL-2.

Luciferase assays. HEK293 cell lines were transiently cotransfected with Renilla

luciferase–tagged Foxp3 and Flag-tagged GATA-3, Foxp1 or STAT6. FuGENE 6

was used for transfection according to the manufacturer’s instructions (Roche).

Cell extracts were prepared (Promega) and samples were immunoprecipitated

with beads coated with anti-Flag (Sigma, clone M2). Renilla luciferase expres-

sion was assessed by detection of luminescence.

Jurkat cells were transfected with a BioRad electroporator (260 V); 5 � 106

cells in 0.4 ml RPMI medium were transfected with 5 mg reporter plasmid (Il5

promoter–luciferase) and 20 mg expression plasmid. Luciferase assays were

done after activation of the cells for 3 h with PMA and ionomycin.

Accession codes. UCSD-Nature Signaling Gateway (http://www.signaling-gate

way.org): A002750, A002271 and A001243.

Note: Supplementary information is available on the Nature Immunology website.

ACKNOWLEDGMENTSWe thank D. Kozoriz, R. Chandwaskar and D. Lee for cell sorting and technicalassistance; and R. Flavell (Yale University School of Medicine) for IL-10–GFPmice. Supported by the US National Institutes of Health (R01AI073542-01 toM.O., and 1R01NS045937-01, 2R01NS35685-06, 2R37NS30843-11,1R01A144880-03, 2P01A139671-07, 1P01NS38037-04, 1R01NS046414 and aJavits Neuroscience Investigator Award to V.K.K.), the National MultipleSclerosis Society (RG-2571-D-9 to V.K.K. and RG-3882-A-1 to M.O.), theJuvenile Diabetes Research Foundation Center for Immunological Tolerance atHarvard Medical School, and the National Multiple Sclerosis Society, New York(A.A. and V.D.).

AUTHOR CONTRIBUTIONSV.D. and A.A. performed and designed the experiments, collected the data andcontributed to the writing of the manuscript; H.K. and G.G. helped with theimmunoprecipitation assays; W.G. and T.B.S. checked Foxp3 expression in IL-4-transgenic T cells; R.A.S. did the pathology analysis; M.M. diagnosed theneurological signs found in RAG-1–reconstituted mice; W.E. and S.K. helped inthe studies using anti-IL-9; I.-C.H. generated and provided conditional GATA-3-deficient mice; M.O. designed experiments; V.K.K. designed the experiments,supervised the study and edited the manuscript.

Published online at http://www.nature.com/natureimmunology/

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. Szabo, S.J. et al. A novel transcription factor, T-bet, directs Th1 lineage commitment.Cell 100, 655–669 (2000).

2. Zheng, W. & Flavell, R.A. The transcription factor GATA-3 is necessary and sufficient forTh2 cytokine gene expression in CD4 T cells. Cell 89, 587–596 (1997).

3. Ivanov, I.I. et al. The orphan nuclear receptor RORgt directs the differentiation programof proinflammatory IL-17+ T helper cells. Cell 126, 1121–1133 (2006).

4. Ouyang, W., Kolls, J.K. & Zheng, Y. The biological functions of T helper 17 cell effectorcytokines in inflammation. Immunity 28, 454–467 (2008).

5. Korn, T., Oukka, M., Kuchroo, V. & Bettelli, E. Th17 cells: effector T cells withinflammatory properties. Semin. Immunol. 19, 362–371 (2007).

6. Battaglia, M., Gregori, S., Bacchetta, R. & Roncarolo, M.G. Tr1 cells: from discovery totheir clinical application. Semin. Immunol. 18, 120–127 (2006).

1354 VOLUME 9 NUMBER 12 DECEMBER 2008 NATURE IMMUNOLOGY

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

7. Hori, S., Takahashi, T. & Sakaguchi, S. Control of autoimmunity by naturally arisingregulatory CD4+ T cells. Adv. Immunol. 81, 331–371 (2003).

8. Hori, S., Nomura, T. & Sakaguchi, S. Control of regulatory T cell development by thetranscription factor Foxp3. Science 299, 1057–1061 (2003).

9. Kretschmer, K. et al. Inducing and expanding regulatory T cell populations by foreignantigen. Nat. Immunol. 6, 1219–1227 (2005).

10. Apostolou, I. & von Boehmer, H. In vivo instruction of suppressor commitment in naiveT cells. J. Exp. Med. 199, 1401–1408 (2004).

11. Chen, W. et al. Conversion of peripheral CD4+CD25– naive T cells to CD4+CD25+

regulatory T cells by TGF-b induction of transcription factor Foxp3. J. Exp. Med. 198,1875–1886 (2003).

12. Bettelli, E. et al. Reciprocal developmental pathways for the generation of pathogeniceffector TH17 and regulatory T cells. Nature 441, 235–238 (2006).

13. Wan, Y.Y. & Flavell, R.A. Regulatory T-cell functions are subverted and converted owingto attenuated Foxp3 expression. Nature 445, 766–770 (2007).

14. Wei, J. et al. Antagonistic nature of T helper 1/2 developmental programs in opposingperipheral induction of Foxp3+ regulatory T cells. Proc. Natl. Acad. Sci. USA 104,18169–18174 (2007).

15. Mantel, P.Y. et al. GATA3-driven Th2 responses inhibit TGF-b1-induced FOXP3expression and the formation of regulatory T cells. PLoS Biol. 5, e329 (2007).

16. Awasthi, A. et al. A dominant function for interleukin 27 in generating interleukin 10–producing anti-inflammatory T cells. Nat. Immunol. 8, 1380–1389 (2007).

17. Stumhofer, J.S. et al. Interleukins 27 and 6 induce STAT3-mediated T cell productionof interleukin 10. Nat. Immunol. 8, 1363–1371 (2007).

18. Ishida, H. et al. Continuous administration of anti-interleukin 10 antibodies delaysonset of autoimmunity in NZB/W F1 mice. J. Exp. Med. 179, 305–310 (1994).

19. Llorente, L. et al. Role of interleukin 10 in the B lymphocyte hyperactivity andautoantibody production of human systemic lupus erythematosus. J. Exp. Med. 181,839–844 (1995).

20. Van Snick, J. et al. Cloning and characterization of a cDNA for a new mouse T cellgrowth factor (P40). J. Exp. Med. 169, 363–368 (1989).

21. Faulkner, H., Humphreys, N., Renauld, J.C., Van Snick, J. & Grencis, R. Interleukin-9is involved in host protective immunity to intestinal nematode infection. Eur. J.Immunol. 27, 2536–2540 (1997).

22. Kamanaka, M. et al. Expression of interleukin-10 in intestinal lymphocytes detectedby an interleukin-10 reporter knockin tiger mouse. Immunity 25, 941–952 (2006).

23. Bettelli, E., Korn, T., Oukka, M. & Kuchroo, V.K. Induction and effector functions ofTH17 cells. Nature 453, 1051–1057 (2008).

24. Wurster, A.L., Withers, D.J., Uchida, T., White, M.F. & Grusby, M.J. Stat6 and IRS-2cooperate in interleukin 4 (IL-4)-induced proliferation and differentiation but aredispensable for IL-4-dependent rescue from apoptosis. Mol. Cell. Biol. 22, 117–126(2002).

25. Li, M.O. & Flavell, R.A. Contextual regulation of inflammation: a duet by transforminggrowth factor-b and interleukin-10. Immunity 28, 468–476 (2008).

26. Izcue, A., Coombes, J.L. & Powrie, F. Regulatory T cells suppress systemic and mucosalimmune activation to control intestinal inflammation. Immunol. Rev. 212, 256–271(2006).

27. Uhlig, H.H. et al. Characterization of Foxp3+CD4+CD25+ and IL-10-secretingCD4+CD25+ T cells during cure of colitis. J. Immunol. 177, 5852–5860 (2006).

28. Read, S., Malmstrom, V. & Powrie, F. Cytotoxic T lymphocyte-associated antigen 4plays an essential role in the function of CD25+CD4+ regulatory cells that controlintestinal inflammation. J. Exp. Med. 192, 295–302 (2000).

29. Kaplan, M.H. & Grusby, M.J. Regulation of T helper cell differentiation by STATmolecules. J. Leukoc. Biol. 64, 2–5 (1998).

30. Ziegler-Heitbrock, L. et al. IFN-a induces the human IL-10 gene by recruiting both IFNregulatory factor 1 and Stat3. J. Immunol. 171, 285–290 (2003).

31. Zhou, L. et al. TGF-b-induced Foxp3 inhibits TH17 cell differentiation by antagonizingRORgt function. Nature 453, 236–240 (2008).

32. Hoffmann, K.F., Cheever, A.W. & Wynn, T.A. IL-10 and the dangers of immunepolarization: excessive type 1 and type 2 cytokine responses induce distinct forms oflethal immunopathology in murine schistosomiasis. J. Immunol. 164, 6406–6416(2000).

33. Li, C., Corraliza, I. & Langhorne, J. A defect in interleukin-10 leads to enhancedmalarial disease in Plasmodium chabaudi chabaudi infection in mice. Infect. Immun.67, 4435–4442 (1999).

34. Gazzinelli, R.T., Oswald, I.P., James, S.L. & Sher, A. IL-10 inhibits parasite killingand nitrogen oxide production by IFN-g-activated macrophages. J. Immunol. 148,1792–1796 (1992).

35. Boussiotis, V.A. et al. IL-10-producing T cells suppress immune responses in anergictuberculosis patients. J. Clin. Invest. 105, 1317–1325 (2000).

36. Carvalho, E.M. et al. Restoration of IFN-g production and lymphocyte proliferation invisceral leishmaniasis. J. Immunol. 152, 5949–5956 (1994).

37. King, C.L. et al. Cytokine control of parasite-specific anergy in human urinaryschistosomiasis. IL-10 modulates lymphocyte reactivity. J. Immunol. 156, 4715–4721 (1996).

38. Makela, M.J. et al. IL-10 is necessary for the expression of airway hyperresponsivenessbut not pulmonary inflammation after allergic sensitization. Proc. Natl. Acad. Sci. USA97, 6007–6012 (2000).

39. Rubtsov, Y.P. et al. Regulatory T cell-derived interleukin-10 limits inflammation atenvironmental interfaces. Immunity 28, 546–558 (2008).

40. Temann, U.A., Geba, G.P., Rankin, J.A. & Flavell, R.A. Expression of interleukin 9 inthe lungs of transgenic mice causes airway inflammation, mast cell hyperplasia, andbronchial hyperresponsiveness. J. Exp. Med. 188, 1307–1320 (1998).

41. Temann, U.A., Ray, P. & Flavell, R.A. Pulmonary overexpression of IL-9 induces Th2cytokine expression, leading to immune pathology. J. Clin. Invest. 109, 29–39(2002).

42. Lu, L.F. et al. Mast cells are essential intermediaries in regulatory T-cell tolerance.Nature 442, 997–1002 (2006).

43. Tepper, R.I. et al. IL-4 induces allergic-like inflammatory disease and alters T celldevelopment in transgenic mice. Cell 62, 457–467 (1990).

44. Annacker, O. et al. Essential role for CD103 in the T cell-mediated regulation ofexperimental colitis. J. Exp. Med. 202, 1051–1061 (2005).

45. Fahlen, L. et al. T cells that cannot respond to TGF-b escape control by CD4+CD25+

regulatory T cells. J. Exp. Med. 201, 737–746 (2005).

NATURE IMMUNOLOGY VOLUME 9 NUMBER 12 DECEMBER 2008 1355

A R T I C L E S©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y