nature immunology  VOLUME 11 NUMBER 12 DECEMBER 2010 1093

A rt i c l e s

Regulatory T cells (Treg cells) are a unique subset of CD4+ T cells that are essential for controlling autoinflammatory diseases. However, they also prevent beneficial antitumor responses and sterilizing immunity against certain infections. Consequently, the modulation of Treg cell activity and generation of Treg cells ex vivo are important goals of immu-notherapy. Naturally occurring, thymus-derived CD4+ Treg cells (nTreg cells) express the lineage-specific transcription factor Foxp3, which is required for their development, homeostasis and function1–3. Despite their limited numbers (5–10% of CD4+ T cells), Treg cells have a pivotal role in immune homeostasis. Indeed, it has been suggested this small population of nTreg cells is able to maintain immune balance through the in vivo conversion of non-Treg cells into suppressive cells, a proc-ess called ‘infectious tolerance’. This ‘contagious’ spread of suppression might be a principal mechanism underlying transplantation tolerance4 and the modulation of autoimmune and inflammatory diseases, such as experimental allergic encephalomyelitis (EAE)5 and asthma6. Although the mechanisms that mediate infectious tolerance remain unclear, the immunomodulatory cytokines transforming growth factor-β (TGF-β) and interleukin 10 (IL-10) have both been linked to this process.

Induced regulatory T cells (iTR cells) can be generated in the periphery or in vitro from conventional CD4+Foxp3− T cells (Tconv cells)7–9. There is substantial interest in the therapeutic potential of iTR cells because antigen-specific regulatory populations can be gen-erated that are potently inhibitory in vivo10,11. Two types of iTR cells have been described on the basis of the cytokines that induce them:

TGF-β–iTR cells and IL-10–iTR cells. TGF-β–iTR cells are generated after the activation of T cells in the presence of TGF-β with or without retinoic acid and IL-2. Both types of iTR cells are potently suppres-sive both in vitro and in vivo10,12,13, but they have distinct molecular signatures. Although TGF-β–iTR cells express Foxp3 and secrete mainly TGF-β, IL-10–iTR cells remain Foxp3− after conversion and are defined by abundant IL-10 secretion.

Treg cell–based approaches to treating inflammatory conditions such as allergy, autoimmune diseases and graft-versus-host responses have great potential but also have limitations11. The therapeutic potential of human Treg cells is limited by their polyclonal specifi-city, poorly defined markers for enrichment and lower proliferative capacity, which limits ex vivo population expansion. Antigen-specific iTR cells can be generated ex vivo, but their utility is restricted by technical complexities in their generation, their limited potency and/or ambiguity regarding stability, and their longevity in vivo. Thus, the identification of a well-defined population of Treg cells that can be readily generated ex vivo and are stable and potently inhibitory in vivo is a critical goal for effective cell-based immunotherapy. IL-35 is a Treg cell–specific cytokine that is required for the maximum regulatory activity of mouse Treg cells in vitro and in vivo14. In this study, we show that IL-35, like IL-10 and TGF-β, generated human and mouse iTR cells, and we address many issues relating to their in vivo efficacy and stability, generation by nTreg cells and physiological contribution to the regulatory milieu.

1Department of Immunology, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA. 2Department of Microbiology and Department of Pathobiology, University of Pennsylvania, Philadelphia, Pennsylvania, USA. 3Department of Bioinformatics, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA. 4Department of Pathology, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA. 5Shenandoah Biotechnology, Monroe, Ohio, USA. 6Department of Antibody Applications, R&D Systems, Minneapolis, Minnesota, USA. 7Department of Microbiology and Immunology, Dartmouth Medical School, Lebanon, New Hampshire, USA. 8Present address: Department of Research, Des Moines University, Des Moines, Iowa, USA. Correspondence should be addressed to D.A.A.V. (vignali.lab@stjude.org).

Received 2 September; accepted 27 September; published online 17 October 2010; doi:10.1038/ni.1952

IL-35-mediated induction of a potent regulatory T cell populationLauren W Collison1, Vandana Chaturvedi1, Abigail L Henderson1,8, Paul R Giacomin2, Cliff Guy1, Jaishree Bankoti1, David Finkelstein3, Karen Forbes1, Creg J Workman1, Scott A Brown1, Jerold E Rehg4, Michael L Jones5, Hsiao-Tzu Ni6, David Artis2, Mary Jo Turk7 & Dario A A Vignali1

Regulatory T cells (Treg cells) have a critical role in the maintenance of immunological self-tolerance. Here we show that treatment of naive human or mouse T cells with IL-35 induced a regulatory population, which we call ‘iTR35 cells’, that mediated suppression via IL-35 but not via the inhibitory cytokines IL-10 or transforming growth factor-b (TGF-b). We found that iTR35 cells did not express or require the transcription factor Foxp3, and were strongly suppressive and stable in vivo. Treg cells induced the generation of iTR35 cells in an IL-35- and IL-10-dependent manner in vitro and induced their generation in vivo under inflammatory conditions in intestines infected with Trichuris muris and within the tumor microenvironment (B16 melanoma and MC38 colorectal adenocarcinoma), where they contributed to the regulatory milieu. Thus, iTR35 cells constitute a key mediator of infectious tolerance and contribute to Treg cell–mediated tumor progression. Furthermore, iTR35 cells generated ex vivo might have therapeutic utility.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

1094  VOLUME 11 NUMBER 12 DECEMBER 2010 nature immunology

A rt i c l e s

RESULTSSuppressive effects of human IL-35-treated Tconv cellsHuman IL-35 suppressed the proliferation of umbilical cord–derived human CD4+ Tconv cells to a degree similar to that pro-duced by activated Treg cells (Supplementary Figs. 1 and 2). Tconv cells activated with beads coated with antibody to CD3 (anti-CD3) and anti-CD28 in the presence of IL-35 substantially upregulated EBI3 and IL12A mRNA, which encode the two constituents of IL-35 (Ebi3 and p35, respectively; Fig. 1a), but did not upregulate IL-10 or TGF-β (Supplementary Fig. 3). Single-cell analysis by both intracellular cytokine staining (Fig. 1b) and confocal micro-scopy (Fig. 1c) suggested that treatment with IL-35 induced expres-sion of IL-35 in essentially all IL-35-treated human CD4+ Tconv cells, but mock treatment did not. Similarly, CD4+CD45RA+CD25− Tconv cells from adult peripheral blood expressed EBI3 and IL12A mRNA but not TGFB or IL10 mRNA after activation in the pres-ence of IL-35 but not in its absence (Supplementary Fig. 3i,j and data not shown).

We next assessed whether IL-35-treated cells assumed the functional phenotype of iTR cells. Tconv cells activated in the presence of IL-35 were hyporesponsive to secondary restimulation, but Tconv cells acti-vated in the absence of IL-35 (called ‘control-treated cells’ here) were not (Fig. 1d). To determine whether IL-35-pretreated Tconv cells had acquired regulatory capacity, we cocultured those cells, as potential sup-pressors, together with freshly purified responder Tconv cells. Although control-treated cells lacked any suppressive capacity, IL-35-treated cells were strongly suppressive (Fig. 1e). Tconv cells treated with human IL-35 also suppressed the proliferation of responder Tconv cells across a permeable membrane, in the absence of direct cell contact, but control-treated cells did not (Fig. 1f), which supported the idea of a role for cytokine-mediated suppression. Moreover, neutralizing anti-IL-35 blocked their suppressive capacity, but anti-IL-10 and anti-TGF-β did not (Fig. 1g and Supplementary Fig. 3). Together these data suggest that IL-35 converts human Tconv cells into a homogeneous population of iTR cells that suppress via IL-35.

Suppression by IL-35-treated mouse Tconv cells in vitroGiven that human IL-35 mediated the generation of iTR cells, we sought to determine if mouse IL-35 had a similar ability (Supplementary Fig. 4). Tconv cells activated in the presence of mouse IL-35 upregulated both Ebi3 and Il12a mRNA but not Il10 or Tgfb mRNA (Fig. 2a and Supplementary Fig. 5). IL-35-treated cells secreted IL-35, which was equivalent to the amount of IL-35 produced by natural Treg cells. Neither control-treated Tconv cells nor IL-35-treated Ebi3−/− Tconv cells secreted IL-35 (Fig. 2b). We next assessed whether IL-35-treated mouse cells, like their human counterparts, assumed an iTR phenotype. As a com-parison, we included treatment with other suppressive cytokines (IL-10 and TGF-β), as well as another Ebi3-containing cytokine (IL-27), in the analysis. Consistent with published reports15, previously activated Tconv cells proliferated well in response to secondary restimulation (Fig. 2c). Tconv cells pretreated with IL-10 and IL-27 also proliferated in response to restimulation (Supplementary Fig. 6). However, Tconv cells pretreated with either IL-35 or TGF-β were hyporesponsive to restimulation, albeit to a lesser degree than were freshly purified nTreg cells. To determine whether these cytokine-pretreated Tconv cells had acquired a regula-tory ability, we cultured them, as potential suppressor cells, together with freshly purified responder Tconv cells (Fig. 2d and Supplementary Fig. 6). Whereas control-treated, IL-10-treated and IL-27-treated Tconv cells had no effect on the proliferation of responder cells, TGF-β-treated Tconv cells suppressed the proliferation of responder T cells12. Mouse Tconv cells pretreated with IL-35 strongly suppressed the proliferation of responder T cells. Furthermore, IL-35-treated Tconv cells suppressed T cell proliferation across a permeable membrane, but control-treated Tconv cells did not (Fig. 2e), which indicated the involvement of soluble suppressive mediators. A concentration of approximately 500–700 pg/ml of IL-35 was required to mediate induction of the expression of Ebi3 and Il12a and of the suppressive phenotype (Supplementary Fig. 4).

To determine the mechanism of suppression, we first showed that IL-35-pretreated Il10−/− Tconv cells (which cannot make IL-10) suppressed responder T cells, but Ebi3−/− Tconv cells (which cannot make IL-35) did not (Fig. 2f). In addition, CD4-dnTGFβRII Tconv

30 100

100

Control + isotype Control + anti-p35

Con

trol

Isotype Anti-p35

IL-3

5

IL-35 + isotype IL-35 + anti-p35

80

60

40

20

100

p35

101 102 103 1040

75

50

25

0

** *

* * **

*

*** **

**

*

20

10

Contro

l

Contro

l

IL-3

5IL

-35

EBI3

exp

ress

ion

(rel

ativ

e)

Pro

lifer

atio

n (1

03 c.p

.m.)

Pro

lifer

atio

n (1

03 c.p

.m.)

Sup

pres

sion

(%

)

Sup

pres

sion

(%

)

IL12A

exp

ress

ion

(rel

ativ

e)

Eve

nts

(% o

f max

)

0

5035

5070 NS NS

60

50

40

30

20

10

0

40

30

20

10

–10

–20

0

Control IL-35 IL-1

0IL

-35

TGF-β

30

25

20

15

10

64:1 32:1 16:1 8:1 4:1 2:1

+ +

5

0

NS

Contro

l

IL-3

5

40

30

20

10

+–

+ +Tconv:

Control Tconv IL-35-treated Tconv

Tconv/treated Tconv

Tconv:

SN:NT Ab: –

SN:

0

a d e

f g

b c

of the cells in a, assessed by incorporation of [3H]thymidine. (e) Suppression of freshly purified responder Tconv cells by control or IL-35-treated Tconv cells, mixed at various ratios (Tconv/treated Tconv) and cultured together for 9 d with activation as in a. (f) Suppressive effect of control-treated or IL-35-treated Tconv cells (in the top chamber of Transwell plate) toward responder Tconv cells (in the bottom chamber), both activated as in a, with [3H]thymidine added directly to the responder cells for the final 8 h of the 9-day assay. (g) Suppression of responder Tconv cells by IL-35-treated Tconv cells, as in f, supplemented with neutralizing (NT Ab) anti-IL-10, anti-TGF-β or anti-IL-35. NS, not significant; *P < 0.05, **P < 0.005 and ***P < 0.001 (unpaired t-test). Data represent ten (a,d–g), four (b) or three (c) independent experiments (mean and s.e.m.).

Figure 1 Treatment of Tconv cells with human IL-35 confers a regulatory phenotype. (a) Quantitative PCR analysis of the expression of EBI3 mRNA (left) and IL12A mRNA (right) generated from RNA extracted from Tconv cells purified from cord blood and treated for 9 d with supernatants of HEK293T human embryonic kidney cells transfected with IL-35-expressing vector or control (empty) vector during activation with human IL-2 plus beads coated with anti-CD3 and anti-CD28; results are presented relative to those of control-treated Tconv cells. (b) Flow cytometry quantification of IL-35 in cells repurified from those in a and activated for 4 h with the phorbol ester PMA and ionomycin, then stained with anti-p35 or isotype-matched control antibody (isotype). (c) Microscopy of p35 expression, assessed as in b: yellow, anti-p35 or isotype-matched control antibody; grey, phalloidin; blue, DNA-intercalating dye (DAPI). Original magnification, ×63. (d) Proliferation

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunology  VOLUME 11 NUMBER 12 DECEMBER 2010 1095

A rt i c l e s

cells, which are unable to respond to TGF-β, were fully suppressed by IL-35-treated Tconv cells (Supplementary Fig. 7). Analysis with cytokine-neutralizing antibodies showed that IL-35-pretreated Tconv cells mediated suppression via IL-35 but not via IL-10 or TGF-β (Fig. 2g,h; the anti-Ebi3 used neutralized IL-35 but not IL-27 (Supplementary Fig. 8)). Collectively, these data suggest that mouse IL-35 converts Tconv cells into an iTR population that mediates sup-pression exclusively via IL-35. Given that IL-35 was central to both the generation and the suppressive function of IL-35–iTR cells, we refer to this iTR population as ‘iTR35 cells’. We refer to the control Tconv cells activated in the absence of IL-35, which did not acquire a suppressive phenotype, as ‘iTRcon cells’.

Highly restricted genetic signature of iTR35 cellsGiven our finding that IL-35 converted proliferative Foxp3− Tconv cells into hyporesponsive, strongly suppressive iTR35 cells, we next sought to define their genetic signature. Foxp3 was neither induced by nor required for the generation of iTR35 cells. Whereas nTreg cells and TGF-β–iTR cells expressed Foxp3, neither iTR35 cells nor iTRcon cells expressed Foxp3 (Supplementary Fig. 9). Moreover, Foxp3−/− Tconv cells could be converted into iTR35 cells that expressed IL-35 and mediated suppression in a manner indistinguishable from that of their wild-type counterparts (Supplementary Fig. 9c). In addition, iTR35 cells did not express Foxp3 after in vivo inoculation, as demonstrated through the use of iTR35 cells generated from Foxp3gfp mice (which express a green fluorescent protein–Foxp3 fusion protein) in an in vivo model of homeostatic expansion. We transfer iTRcon cells, wild-type iTR35 cells and Ebi3−/− iTR35 cells into mice deficient in recombination-activating gene 1 (Rag1−/− mice) and, 7d later, analyzed the cells on the basis of

congenic Thy1 markers and Foxp3 expression. We found no induction of Foxp3 in any of the transferred iTR cells (Supplementary Fig. 9d).

We next compared the global gene expression of iTR35 cells and iTRcon cells by Affymetrix GeneChip microarray. Before these analyses, we veri-fied that the iTR35 cells used for analysis expressed Ebi3 and Il12a mRNA, secreted IL-35 and suppressed responder Tconv cells (Supplementary Fig. 10). Although we observed differences between nTreg cells and Tconv cells, no genes were substantially upregulated or downregulated (over threefold) in iTR35 cells versus iTRcon cells (Supplementary Fig. 11). Nevertheless, the iTR35 cells generated in five independent experiments expressed Ebi3 and Il12a mRNA, secreted IL-35 and mediated potent in vitro suppression. However, we cannot exclude the possibility that other genes with poor probe sets may be expressed by iTR35 cells but were not identified in this analysis. We also compared the Treg cell genetic signa-ture with that of iTR35 cells in a meta-analysis. Although iTR35 cells and iTRcon cells shared expression of some genes with both Treg cells and Tconv cells, their expression patterns were nearly identical, which suggested that modest transcriptional changes were responsible for the phenotype observed in iTR35 cells.

The minimal difference between iTR35 cells and iTRcon cells in their gene-expression profile was further supported by analysis of T cell acti-vation and costimulatory molecule expression and cytokine production. Although the expression of most of the proteins and cytokines examined was indistinguishable in iTR35 cells versus iTRcon cells, we observed less secretion of granulocyte-monocyte colony-stimulatory factor, interferon-γ and IL-4, although the differences were not statistically significant (Supplementary Fig. 12). In addition, surface molecules such as CTLA-4 and CD25, which have been described as mediators of nTreg cell suppression, were similarly upregulated in both iTR35 cells

30a

b

c d

g h

e f80

20 50 30

20

10

10

0

20

30

40

50NS

0

–10 –10

40

30

20

10

0

NS60

50

40

30

Sup

pres

sion

(%

)

Sup

pres

sion

(%

)

20

10

0

60

50

40

30

20

10

–10

0

NT Ab: –IL

-10

IL-3

5 64:1 32:1 16:1 8:1 4:1 2:1TGF-β

NS ***

***

*** ***

Control Tconv

Tconv

Ebi3

TconvWT WTcon IL-35 IL-35 SNTreg

Control Tconv

Tconv/treated Tconv

Tconv/Treg

Tconv + IL-35 Tconv + IL-35 + isotype

Tconv + IL-35 + NT IL-35

Tconv: Tconv:SN:

WTW

T

Ebi3–/

–

II10–/

–WT

Control IL-35

24:1 12:1 6:1 2:1 16:1 8:1 4:1 2:1

T cell/bead

* *** *

IL-35-treated Tconv IL-35-treated Tconv

15

10

5

0

70

60

50

40

30

20

10

0

* *

20

Ebi3

exp

ress

ion

(rel

ativ

e)

Pro

lifer

atio

n (1

03 c.p

.m.)

Pro

lifer

atio

n (1

03 c.p

.m.)

Sup

pres

sion

(%

)

Sup

pres

sion

(%

)

ll12a

exp

ress

ion

(rel

ativ

e)

10

0Control IL-35 Control IL-35

*

****

Ebi3–/–

(IL-35 SN) or control vector (Con SN), then repurified and cultured for an additional 24 h, followed by immunoprecipitation with monoclonal anti-p35; blots were probed with monoclonal anti-Ebi3. Far left two lanes, untreated Tconv cells and Treg cells (controls). (c) Proliferative capacity of control-treated or IL-35-treated Tconv cells, prepared as in a, stimulated with beads coated with anti-CD3- and anti-CD28. (d) Proliferation of responder Tconv cells mixed at various ratios with control-treated or IL-35-treated Tconv cells (Tconv/treated Tconv) and cultured for 72 h in the presence of beads coated with anti-CD3 and anti-CD28. (e) Proliferation of responder Tconv cells (activated with beads coated with anti-CD3 and anti-CD28) in the bottom chamber of a Transwell plate, in the presence of control- or IL-35-treated Tconv cells in the top chamber stimulated with beads coated with anti-CD3- and anti-CD28. Results are presented relative to those of responder Tconv cells stimulated in the absence of cells in the top chamber. (f) Suppression of the proliferation of responder Tconv cells by wild-type, Ebi3−/− or Il10−/− IL-35-treated Tconv cells, prepared as in a, at a ratio of 4:1 (responder Tconv cells/treated Tconv cells). (g) Suppression of the proliferation of responder Tconv cells as in f, in the presence of neutralizing anti-IL-10, anti-TGF-β or anti-IL-35. (h) Suppressive capacity of IL-35-treated Tconv cells (Tconv + IL-35), assessed as in g in the presense of neutralizating anti-IL-35 (NT IL-35) or isotype-matched control antibody (isotype). *P < 0.05, **P < 0.005 and ***P < 0.001 (unpaired t-test). Data represent four to eight independent experiments (mean and s.e.m.).

Figure 2 Treatment of Tconv cells with mouse IL-35 converts them into an IL-35-producing suppressive population. (a) Expression of Ebi3 and Il12a mRNA in control-treated or IL-35-treated Tconv cells activated for 72 h with beads coated with anti-CD3 and anti-CD28; results are presented relative to those of control-treated Tconv cells. (b) Immunoblot analysis of supernatants of wild-type (WT) or Ebi3−/− Tconv cells activated as in a in the presence of supernatants of HEK293T cells transfected with IL-35-expressing vector

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

1096  VOLUME 11 NUMBER 12 DECEMBER 2010 nature immunology

A rt i c l e s

and iTRcon cells, which challenges the idea of their involvement in the suppressive function of iTR35 cells. Furthermore, the frequency of CTLA-4+ iTR35 was relatively low (<15%). Although our data suggest that iTR35 cells have a highly restricted genetic signature that results in a CD4+Foxp3−Ebi3+p35+IL-10−TGF-β− phenotype, we cannot rule out the possibility that there are some molecular differences that might distinguish iTR35 cells from iTRcon cells that were not identified by this analysis or that this signature also applies to an undefined nonregulatory population. Together these results suggest that IL-35 treatment mediates limited rather than global changes in gene expression.

Potent suppression by iTR35 cells in vivoWe tested the regulatory capacity of iTR35 cells in five different in vivo models. We first assessed whether iTR35 cells could restore immune homeostasis and prevent lethal autoimmunity in Foxp3−/− mice16,17.

We transferred iTR35 cells and various control populations into new-born (2- to 3-day-old) Foxp3−/− mice. Approximately 25 d later, we determined external clinical scores, splenic and lymph node CD4+ T cell numbers and histological scores (lungs, liver and skin). As expected, nTreg cells and TGF-β–iTR cells were able to restore immune homeostasis and prevent autoimmunity, but iTRcon cells were not (Fig. 3a–d and Supplementary Fig. 13). We found that iTR35 cells were as effective as nTreg cells at restoring immune homeostasis and prevent-ing autoimmunity in Foxp3−/− mice. Notably, neither Ebi3−/− iTR35 cells nor Il12a−/− iTR35 cells were able to restore immune homeostasis, which demonstrated the necessity for IL-35 production in vivo by iTR35 cells.

Treg cells can control the homeostatic population expansion of Tconv cells in lymphopenic Rag1−/− mice14,18,19. We adoptively transferred purified wild-type Thy-1.1+ Tconv cells, either alone or in the presence of control-treated or IL-35-treated Thy-1.2+ T cells, into Rag1−/− mice

6

a d

e

h i j

f g

b

c

NS NS NS

NS

NS

NS

NS NS

NS

NS

NS NS EAE

Saline control

160 100 3.0

2.5

2.0

1.5

1.0

0.5

80

60

40

20

0

Homeostasis,regulatory cell

Homeostasis,target cell

*** ** **

**** ****

*

Reg

ulat

ory

Thy

-1.2

cel

ls (

×105 )

Tum

or s

ize

(mm

3 )

Wei

ght c

hang

e (%

)

His

tolo

gica

l sco

re

Tar

get

Thy

-1.1

cel

ls (

×105 )

Clin

ical

sco

re14012010080604020

400

7 9 11Time after inoculation (d) Time after transfer (weeks)

1

7.5 4

3

2

1

0

5.02.5

–2.5–5.0–7.5–10

0

2 3 413 15

300

200

100

0

0

NS NS

Foxp3–/–recovery

Foxp3–/–WT

**

***

**

** ***

*** ***

– –Treg or iTR: Treg TGF-βiTR

iTRcon

iTRcon

WTiTR35

Mouse: Foxp3–/–

Treg

Treg

Foxp3–/–

Foxp3–/– Foxp3–/–

Ebi3–/–

iTR35ll12a–/–

iTR35

Foxp3–/– Foxp3–/– Foxp3–/–

Foxp3–/– + TGF-β–iTR

Foxp3–/– + no Treg Foxp3–/– + nTreg

Foxp3–/– + iTR35Foxp3–/– + iTRcon

WT + no Treg

*** *** *****

Clin

ical

sco

reT

otal

CD

4+ s

plen

ocyt

es (

×106 )

His

tolo

gica

l sco

re

5

4

3

2

20

15

10

5

30

20

10

0

0Mouse:

1

0

Treg or iTR:

Mouse:

Treg or iTR:

Foxp3–/–WT

WT

– –

– –

TGF-βiTR

TGF-βiTR

Foxp3–/–

iTRcon

Foxp3–/–

WTiTR35

iT Rco

nThy-1.1 target:

Thy-1.2 regulatory:

No cell transferCD4 + CD8 + nTregCD4 + CD8CD4 + CD8 + iTR35

iTRcon

iTRcon

IBD IBD

iTR35

iTR35

B16 1° tumor

iT Rco

n

iTRcon

iTR35

iT R35

iT R35

++

–

+ + + + + +

WTiTR35

Foxp3–/–

Ebi3–/–

iTR35

Ebi3–/–

iTR35

Foxp3–/–

ll12a–/–

iTR35

Ebi3–/

–

iT R35

Ebi3–/

–

iT R35

ll12a–/–

iTR35

Foxp3–/–

Foxp3–/–Foxp3–/–Foxp3–/–Foxp3–/–Foxp3–/–Foxp3–/–Foxp3–/–

–

10 12 14 16 18 20Time after induction (d)

22 24 26 28

Ebi3–/– iTR35WT Treg

Figure 3 Suppressive effects of iTR35 cells in vivo. (a–d) Prevention of disease in 3-week-old Foxp3−/− mice given intraperitoneal injection (at 2 or 3 d of age) of nTreg cells, iTRcon cells or wild-type, Ebi3−/− or Il12a−/− iTR35 cells or TGF-β–iTR cells, monitored by clinical score (a), splenic CD4+ T cell numbers (b) and histology (c). Far left (WT −), untreated, age-matched, wild-type littermates (control). (d) Histopathology of ear pinna of Foxp3−/− mice treated as in a–c or their untreated wild-type littermates. Original magnification, ×10. (e,f) Homeostatic population expansion of Thy-1.1+ Tconv cells (as target cells; f) injected intravenously into Rag1−/− mice alone or with Thy-1.2+ iTRcon cells, wild-type iTR35 cells or Ebi3−/− iTR35 cells (as regulatory cells; e). (g) EAE scores of C57BL/6 mice injected intravenously with wild-type nTreg, iTRcon, wild-type or Ebi3−/− iTR35 cells 12–18 h before disease induction via immunization with a peptide of amino acids 35–55 of myelin oligodendrocyte glycoprotein in complete Freund’s adjuvant and pertussis toxin. P = 0.002, PBS versus iTR35 (Wilcoxon matched-pairs test). (h) Tumor diameter in Rag1−/− mice injected intravenously with various combinations of cells (key) and then injected intradermally with B16 cells. P = 0.0025, CD4 + CD8 versus CD4 + CD8 + nTreg, and P = 0.0048, CD4 + CD8 versus CD4 + CD8 + iTR35 (Wilcoxon matched-pairs test). (i,j) Weight (i) and colonic histology scores (j) of Rag1−/− mice injected intravenously with Tconv cells; after 3–4 weeks, mice that developed signs of IBD were given iTRcon cells or iTR35 cells. *P < 0.05, **P < 0.005 and ***P < 0.001 (unpaired t-test). Data represent at least two independent experiments with eight to twelve mice per group (mean and s.e.m.).

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunology  VOLUME 11 NUMBER 12 DECEMBER 2010 1097

A rt i c l e s

and counted splenic responder (Thy-1.1+) and suppressor (Thy-1.2+) T cells 7 d later17. Thy-1.2+ iTRcon cell populations expanded sig-nificantly and failed to block the population expansion of Thy 1.1+ responder Tconv cells (Fig. 3e,f). In contrast, Thy-1.2+ iTR35 cells, but not Ebi3−/− iTR35 cells, had low proliferative capacity and substantially limited the proliferation of Thy-1.1+ responder Tconv cells.

EAE is a model of multiple sclerosis and can be induced in C57BL/6 mice by immunization with a peptide of amino acids 35–55 of myelin oligodendrocyte glycoprotein. Adoptive transfer of nTreg cells results in lower EAE disease severity5,20,21. To determine whether iTR35 cells could diminish or prevent EAE, we transferred 1 × 106 nTreg cells, iTRcon cells or iTR35 cells into mice before inducing EAE5,17. Consistent with published reports, clinical scores were lower in recipi-ents of nTreg cells, whereas the disease course in mice that received iTRcon cells or saline alone was unaffected (Fig. 3g). Notably, mice treated with iTR35 cells were completely protected from EAE, whereas mice that received Ebi3−/− iTR35 cells were indistinguishable from control mice that received saline, which suggested that IL-35 produc-tion by iTR35 cells in vivo is required for this protection.

Treg cells can prevent antitumor CD8+ T cell responses against the poorly immunogenic B16 melanoma22,23. We adoptively transferred wild-type naive CD4+CD25− T cells and CD8+ T cells into Rag1−/− mice, alone or in combination with nTreg or iTR35 cells, followed by intrader-mal injection of B16 melanoma cells, and monitored tumor size daily17. Tumors were smaller in recipients of CD4+ and CD8+ T cells that lacked Treg cells than in untreated Rag1−/− mice (Fig. 3h). In contrast, transfer of either nTreg cells or iTR35 cells with CD4+ and CD8+ T cells completely blocked the antitumor response, which resulted in more aggressive tumor growth similar to that of untreated Rag1−/− mice. Surgical excision of the primary tumor and subsequent secondary tumor challenge at a distal site demonstrated that concomitant tumor immunity was also prevented by both nTreg cells and iTR35 cells (Supplementary Fig. 13).

We initiated inflammatory bowel disease (IBD) by adoptive trans-fer of naive CD4+CD45RBhiCD25− T cells into Rag1−/− recipient mice and monitored disease by assessing weight loss and histopathol-ogy of the colon24. After mice developed clinical symptoms of IBD (~4 weeks after T cell transfer), they received iTRcon cells or iTR35 cells and were monitored daily17. Mice that received iTR35 cells dem-onstrated recovery from disease, indicated by weight gain (Fig. 3i) and less histopathology (Fig. 3j and Supplementary Fig. 13), but mice that received iTRcon cells did not. We showed that iTR35 cells were also able to cure IBD induced by CD4-dnTGFβRII Tconv cells to further demonstrate that TGF-β is not required for the in vivo suppressive ability of iTR35 (Supplementary Fig. 14).

Finally, we assessed the importance of IL-35 production by iTR35 cells in vivo with a unique Ebi3-specific antibody that neutralized

IL-35 but not IL-27 (Supplementary Fig. 8). Administration of this antibody blocked the suppressive ability of iTR35 cells in an in vivo model of homeostatic expansion, but administration of an isotype-matched control antibody did not (Supplementary Fig. 13g), consist-ent with our observations obtained with Ebi3−/− or Il12a−/− iTR35 cells in the three in vivo models (Fig. 3). Together these data demonstrate that iTR35 cells have a potent suppressive ability in a wide variety of in vivo models and that this activity is dependent on IL-35 produc-tion in vivo.

Stability of iTR35 cells in vivoIt has been suggested that iTR cells generated ex vivo are unstable in vivo. Although our five in vivo transfer experiments suggested that iTR35 cells have some degree of stability, we used two approaches to assess this directly. First, we generated CD45.2+ iTR35 or TGF-β–iTR cells in vitro and adoptively transferred them into CD45.1+ C57BL/6 mice to moni-tor cell recovery and function over time. We recovered both iTR35 cells and TGF-β–iTR cells from the spleen after transfer and found that both retained expression of their signature genes (Ebi3 and Il12a (iTR35) or Foxp3 and Tgfb (TGF-β–iTR); Supplementary Fig. 15). As observed in vitro, inoculation for 3 weeks with iTR35 cells in vivo failed to induce Foxp3 expression, which suggested that this critical nTreg cell transcrip-tion factor is not required for the maintenance and function of iTR35 cells. Although we recovered 33% of the initial iTR35 cell inoculum from the spleen 3 weeks after transfer, we recovered only 12% of TGF-β–iTR cells (Fig. 4a). In addition, purified iTR35 cells still retained considerable suppressive ability, whereas the function of TGF-β–iTR cells was approxi-mately 50% lower (Fig. 4b). Although this suggested that iTR35 cells might be more stable in vivo, it did not exclude the possibility that iTR35 cells and TGF-β–iTR cells might home to different anatomical locations in the mouse, which could affect their recovery from the spleen. Second, we transferred nTreg cells, iTR35 cells or TGF-β–iTR cells into 2- to 3-day-old Foxp3−/− mice and determined how long they could prevent the onset of a moribund state (clinical score ≥4). By 5 weeks after transfer, all recipi-ents of TGF-β–iTR cells were moribund, compared with 40% or 33% of recipients of nTreg cells or iTR35 cells, respectively (Fig. 4c). Furthermore, the survival of the remaining recipients of nTreg cells or iTR35 cells was longer, with 100% morbidity not being reached until 6.5 weeks or 8 weeks, respectively. Although additional experiments are needed to fully evaluate the long-term stability of iTR35 cells in homeostatic and inflammatory environments, these data suggest that they are functionally stable in vivo and are apparently at least as stable as nTreg cells.

Treg cell–mediated suppression generates iTR35 cellsIt has been suggested that Treg cells can amplify their suppressive ability by converting nonregulatory populations into suppressive

Figure 4 Stability of iTR35 cells and TGF-β–iTR cells in vivo. (a) Recovery of splenic iTR cells 25 d after adoptive transfer of iTR35 or TGF-β–iTR cells (generated in vitro from CD45.2+ Tconv cells) into CD45.1+ C57BL/6 mice, assessed by flow cytometry of CD45.2+ cells and presented as percentage of total cells injected. (b) Supression of Tconv cell proliferation by either freshly generated iTR cells (Before transfer) or iTR cells recovered after a period of in vivo ‘resting’ (After transfer), mixed at various ratios (Tconv/Treg) and stimulated for 72 h with beads coated with anti-CD3 and anti-CD28, assessed by [3H]thymidine incorporation. (c) Disease development in Foxp3−/− mice injected with natural Treg cells or iTR cells at 2–3 d of age; mice with a clinical score of 4 were considered moribund. *P < 0.05 and **P < 0.001 (unpaired t-test). Data represent at least two independent experiments with five to twelve mice per group (mean and s.e.m.).

40

4040

50

60

60

Before transferAfter transfer

iTR35iTR35

iTR35

**

30

30

iT R35

TGF-β–iT R

TGF-β iTRTGF-β iTR

TGF-β iTR

2020

20

*

10 10

100

80Mor

ibun

d (%

)

–108:1 4:1 2:1 0 1 2 3 4 5 6 7 8 9

Time after transfer (weeks)Tconv/Treg

Treg

0

0

0

Rec

over

y (%

)

Sup

pres

sion

(%

)

a b c

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

1098  VOLUME 11 NUMBER 12 DECEMBER 2010 nature immunology

A rt i c l e s

cells, consistent with the idea of infectious tolerance, and that this process might be cytokine mediated25–27. It is known that nTreg cells are a natural source of IL-35, which increases five- to tenfold after contact with target Tconv cells14,28. Thus, we sought to determine whether nTreg cell–derived IL-35 could mediate conversion to iTR35 cells. We first purified Tconv cells that had been cultured with and suppressed by nTreg cells for 3 d (called ‘suppressed Tconv cells’ here) and found that the expres-sion of both Ebi3 and Il12a mRNA was significantly upregulated after coculture to an amount similar to that of nTreg cells (Fig. 5a,b). Furthermore, suppressed Tconv cells generated by culture together with wild-type cells secreted IL-35, but those generated by cul-ture with Ebi3−/− nTreg cells did not (Fig. 5c). This demonstrates that IL-35 secretion by nTreg cells is required for the induction of IL-35 secretion by cocultured, suppressed Tconv cells. To deter-mine whether suppressed Tconv cells express Foxp3, a prerequisite for mediating the regulatory activity of nTreg cells and TGF-β–iTR cells, we activated Thy-1.2+ Foxp3gfp Tconv cells alone or in combina-tion with Thy-1.1+ nTreg cells. Unlike TGF-β–iTR cells, but similar to iTR35 cells, suppressed Tconv cells remain Foxp3− after activation in the presence of Treg cells (Supplementary Fig. 16), which sug-gested that TGF-β does not mediate this conversion. These data raise the possibility that iTR35 cells are generated from the suppressed Tconv population.

We next assessed whether suppressed Tconv cells gained the phenotypic characteristics of a regulatory population. Suppressed Tconv cells were profoundly unresponsive to stimulation with anti-CD3 and were potently suppressive in vitro (Fig. 5d,e). Treg cells can secrete IL-10, TGF-β and

IL-35, which might influence their ability to convert Tconv cells into suppressed Tconv cells. Likewise, the same cytokines could be secreted by suppressed Tconv cells and contribute in an autocrine way to their conver-sion and/or suppressive activity. To address these issues we first cocul-tured Tconv cells and nTreg cells that were wild-type or lacked the ability to produce IL-35 (Ebi3−/− or Il12a−/−) or IL-10 (Il10−/−), or were unable to respond to TGF-β (CD4-dnTGFβRII). Although the generation of sup-pressed Tconv cells that were hyporesponsive and had a regulatory ability did not require TGF-β-mediated signaling, the absence of both IL-35 and IL-10 from the nTreg cell–Tconv cell coculture blocked their development and/or function (Fig. 5d,e). Further analysis of nTreg cell–Tconv cell coc-ultures in which only one population was mutant showed that both IL-10 and IL-35 derived from nTreg cells was required for the generation of the regulatory suppressed Tconv population. IL-35 derived from suppressed Tconv cell was also required for conversion, as suggested by analysis of signature genes by quantitative PCR (Supplementary Fig. 15d).

We included neutralizing antibodies during the conversion proc-ess or in the secondary suppression assay to further assess the roles of IL-35, IL-10 and TGF-β (Fig. 5f). Anti-TGF-β had no effect at either stage, and IL-10 neutralization partially blocked conversion but not the regulatory capacity of suppressed Tconv cells, which sug-gested that IL-10 is required for optimal conversion of suppressed Tconv cells into a regulatory population. In contrast, neutralization

30a

d f

e

g h

b c25 NS

24h

IP: anti-p35Blot: anti-Ebi3

Conversion

Function

Suppressed TconvSaline control

EAE4

3

2

1

010 15 20

Time after induction (d)

Clin

ical

sco

re

25 30

Treg

NS

NS NS

NS

72 h

Sort

Ebi3

IgL

20

15

10

5

0

** ** ** *NS

Tconv

Tconv: WT WT

WT WT

––

WT WT WT

WT

WT

WT

WT

WT WT

NT Ab: None IL-10 IL-35

WT CD4-dnTGF-βRII

CD4-dnTGF-βRII

CD4-dnTGF-βRII

CD4-dnTGF-βRII

WTTreg:

Tconv:

Treg:

Treg:Suppressed

Tconv:

Tconv:

+ + + +++– –

Tconv Tconv

NS

NS NS NS NS NS NS

– –

NS***

*** *** * *

** *

**

***

***NS NS NS NS

Resortedafter culture

Resortedafter culture

Treg Tconv Tconv Tconv TregCell: Cell:Anti-CD3 +anti-CD28:

Anti-CD3 +anti-CD28:

20

Ebi3

exp

ress

ion

(rel

ativ

e)P

rolif

erat

ion

(103 c

.p.m

.)S

uppr

essi

on (

%)

T c

ells

(×1

04 )

II12a

exp

ress

ion

(rel

ativ

e)

10

0

2580

60

40

20

0

80

60

40

20

–20

0

20

15

10

5

80

60

40

20

0

–20

40

30

20

10

0

–40

0

Tconv

Tconv +

WT Treg

Tconv +

Ebi3–/– Treg

Tconv TconvTreg

Ebi3–/–

Ebi3–/–

Ebi3–/–

Ebi3–/–

II12a–/–

II12a–/–

II12a–/–

II10–/–

II10–/–

WT

WT WTWT

WT

WT WT

Homeostasis, target cellNS**

Ebi3–/–

Ebi3–/–

Ebi3–/–

Ebi3–/–

II12a–/– II10–/– II10–/–

II10–/–

II10–/–

II10–/–II10–/–Ebi3–/–

Sup

pres

sion

(%

)S

uppr

essi

on (

%)

TGF-β

– – –

– –

Figure 5 Treg cells generate iTR35 cells in an IL-35- and IL-10-dependent manner. (a,b) Expression of Ebi3 (a) and Il12a (b) by Tconv cells activated for 72 h with Treg cells at a ratio of 4:1 in the presence (+) or absence (–) of beads coated with anti-CD3 and anti-CD28. (c) Immunoprecipitation (IP) and immunoblot (IB) analysis of the secretion of IL-35 into supernatants of Tconv cells suppressed by coculture as in a, then repurified and cultured for an additional 24 h. (d) Proliferation of suppressed Tconv cells cocultured as in a, then repurified and activated with anti-CD3 and anti-CD28. (e) Suppression of responder Tconv cell proliferation by suppressed Tconv cells generated as in a. (f) Suppression of responder Tconv cell proliferation by suppressed Tconv cells generated as in a, in the presence of neutralizing anti-IL-10, anti-TGF-β or anti-IL-35 during either culture with nTreg cells (Conversion; top) or suppression of responder Tconv cells (Function; bottom). (g) Splenic T cell numbers in Rag1−/− mice injected with Tconv cells (target cell) alone or together with suppressed Tconv cells (as regulatory cells), assessed 7 d after transfer. (h) EAE scores of C57BL/6 mice injected intravenously with suppressed Tconv cells, nTreg cells or saline control 12–18 h before disease induction (via immunization as in Fig. 3g). *P < 0.05, **P < 0.005 and ***P < 0.001 (a–g; unpaired t-test). Data represent at least two independent experiments with eight to twelve mice per group (mean and s.e.m.).

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunology  VOLUME 11 NUMBER 12 DECEMBER 2010 1099

A rt i c l e s

of IL-35 prevented both the conversion and regulatory function of suppressed Tconv cells. Collectively, these data suggest that some or all of the suppressed Tconv cells were iTR35 cells. The precise contribu-tion of IL-10 to Treg cell–mediated conversion to iTR35 cells remains to be fully elucidated, because IL-10 alone did not induce the expres-sion of Ebi3 or Il12a mRNA (Supplementary Fig. 5). However, the addition of a low dose of IL-10 seemed to augment the IL-35- mediated conversion, which might help offset the delayed produc-tion of IL-35 by nTreg cells (Supplementary Fig. 16c). Together these data suggest that IL-35, either from a natural source (nTreg cells) or supplemented exogenously, mediates iTR35 conversion.

Next we assessed the regulatory capacity of nTreg cell–suppressed Tconv cells (nTreg cell–induced iTR35 cells) in vivo. First we found that they were able to significantly suppress the homeostatic popu-lation expansion of cotransferred naive Tconv cells in Rag1−/− mice in a manner similar to that of nTreg cells and iTR35 cells (Fig. 5g). However, suppressed Tconv cells generated from Ebi3−/− Tconv cells cultured with wild-type nTreg cells failed to suppress the population expansion of cotransferred Tconv cells. Next we found that in the EAE model, peak clinical disease scores were decreased by sup-pressed Tconv cells to a degree similar to that obtained with nTreg cells (Fig. 5h). However, suppressed Tconv cells did not ameliorate EAE as effectively as iTR35 cells did, which suggested that only a propor-tion of this suppressed Tconv cell population was iTR35 cells or that conversion in vitro was suboptimal because of the time required for the potentiation of IL-35 production by nTreg cells14,28. Nevertheless, these data support the idea that iTR35 cells are generated from Tconv cells to some degree by nTreg cells during suppression. In con-trast, there was no evidence of the generation of IL-10–iTR cells or TGF-β–iTR cells in this setting.

Induction of iTR35 cells in vivo by nTreg cellsWe reasoned that the generation of iTR35 cells in vivo would occur predominantly in inflammatory or disease environments in which

optimally stimulated nTreg cells might be secreting large amounts of IL-35. Infection with Trichuris muris, an intestinal nematode, promotes Treg cell responses in the large intestine29. Thus, using the CD4+Foxp3−Ebi3+IL-12a+ iTR35 cell signature, we assessed whether iTR35 cells were detectable after T. muris infection. We infected Foxp3gfp mice with T. muris and purified Foxp3+ and Foxp3− T cells from spleens, small intestines and large intestines 14 d after infection. The expression of both Ebi3 and Il12a was much higher in Foxp3+ Treg cells in both the small and large intestines than it was in splenic Treg cells (Fig. 6a,b), consistent with published observa-tions that nTreg cells result in tenfold higher IL-35 expression in the presence of Tconv cells14. Although we observed essentially no expression of Ebi3 or Il12a in splenic Foxp3−CD4+ T cells, their expression was substantial in similar isolates from the small intes-tines and especially the large intestines (the main site of infection). Indeed, the expression of Ebi3 and Il12a was statistically indistin-guishable in Foxp3+ Treg cells and Foxp3−CD4+ T cells from the large intestines. IL-35 expression by naive, activated or memory CD4+ T cells has not been reported14, and Foxp3−CD4+ T cells from the intestines or mesentaeric lymph nodes of uninfected mice did not express Ebi3 or Il12a (data not shown), which raised the possibility that iTR35 cells were being generated by Treg cells in this inflammatory microenvironment.

The inflammation induced by solid tumors attracts Treg cells30–35. Using B16 melanoma and MC38 colorectal adenocarcinoma as model systems22,36,37, we inoculated Foxp3gfp mice with tumor cells, resected solid tumors 15–17 d after transfer (B16) or 12 d after transfer (MC38) and purified Foxp3+ and Foxp3− T cells from spleens and tumors. Tumor-infiltrating Foxp3+ Treg cells had much higher expression of both Ebi3 and Il12a (Fig. 6c–f). Tumor-infiltrating Foxp3− T cells also substantially upregulated their expression of Ebi3 and Il12a. We made similar observations in the two distinct tumor types. We fur-ther analyzed IL-35 secretion and its physiological relevance in the B16 melanoma system. Although we observed moderate secretion of

a c eb

hg

d f400

300

200

Ebi3

exp

ress

ion

(rel

ativ

e)

IL12a

expr

essi

on (

rela

tive)

100

320

220120806040200

IL12a

expr

essi

on (

rela

tive)

IL12a

expr

essi

on (

rela

tive) 680

4802808030

20

10

00

600 80300

200

100

0

60

40

20

0

400200150

Ebi3

exp

ress

ion

(rel

ativ

e)

Ebi3

exp

ress

ion

(rel

ativ

e)

100

50

0

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

Spleen

Foxp3

+

WT Foxp3gfp mice Ebi3–/– Foxp3gfp mice

Spleen Tumor

Spleen Tumor Spleen Tumor

Smallintestine

Largeintestine

IP: anti-p35Blot: anti-Ebi3

Ebi3

IgL

Ebi3

–/– F

oxp3

–

Ebi3

–/– F

oxp3

+

WT

Fox

p3+

WT

Fox

p3–

0

20

40

60

Sup

pres

sion

(%

)

80 NS

NS

*

** ****

**** *

*

*

**

***

*** **** **

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

Spleen Smallintestine

Largeintestine

Spleen Tumor

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

Spleen Tumor

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

CD

4+ F

oxp3

+

CD

4+ F

oxp3

–

Spleen Tumor

Foxp3

–

Foxp3

+

Foxp3

–

Foxp3

+

Foxp3

–

Foxp3

+

Foxp3

–

relative to those of CD4+Foxp3− cells purified from the spleen. (g) Immunoprecipitation (with monoclonal anti-p35) and immunoblot analysis (with monoclonal anti-Ebi3) of the secretion of IL-35 into supernatants of CD4+Foxp3− and CD4+Foxp3+ cells purified from Foxp3gfp or Ebi3−/−Foxp3gfp mice as in c,d and cultured for 24 h. (h) Proliferation of cells purified as in g and mixed for 72 h at a ratio of 4:1 with fresh responder Tconv cells, assessed by [3H]thymidine incorporation. *P < 0.05, **P < 0.005 and ***P < 0.001 (unpaired t-test). Data represent two to three independent experiments (B16) or one experiment (MC38) with eight to ten mice per group (mean and s.e.m.).

Figure 6 IL-35-producing Foxp3− iTR35 cells develop in vivo. (a,b) Expression of Ebi3 (a) and Il12a (b) in CD4+Foxp3− and CD4+Foxp3+ cells purified from Foxp3gfp mice infected with T. muris. (c,d) Expression of Ebi3 (c) and Il12a (d) in CD4+Foxp3− and CD4+Foxp3+ cells purified from tumors and spleens excised from Foxp3gfp mice 15–17 d after intradermal injection of B16 cells (1.2 × 105). (e,f) Expression of Ebi3 (e) and Il12a (f) in CD4+Foxp3− and CD4+Foxp3+ cells purified from tumors and spleens excised from Foxp3gfp mice 12 d after subcutaneous injection of MC38 cells (2 × 106). Results in a–f are presented

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

1100  VOLUME 11 NUMBER 12 DECEMBER 2010 nature immunology

A rt i c l e s

IL-35 by splenic Foxp3+ Treg cells ex vivo, both Foxp3+ Treg cells and Foxp3−CD4+ tumor-infiltrating T cells had substantial and similar secretion of IL-35 (Fig. 6g). Finally, we assessed whether tumor-infiltrating CD4+Foxp3−Ebi3+Il12a+ T cells were able to suppress the proliferation of fresh responder Tconv cells in vitro. Although their sup-pressive ability was not as potent as that of tumor-infiltrating Foxp3+ T cells, our results demonstrated that tumor-derived Foxp3−CD4+ T cells mediated effective suppression in vitro in an IL-35-dependent manner (Fig. 6h).

Contribution of iTR35 cells to the regulatory milieu in vivoFinally, we assessed the physiological contribution of iTR35 cells to the Treg cell–induced regulatory milieu. We reasoned that if the development of iTR35 cells in tumors had an important role in block-ing antitumor immunity, then mice reconstituted with Tconv cells that lacked the ability to be converted into iTR35 cells would develop smaller tumors. Therefore, we reconstituted Rag1−/− mice with wild-type CD8+ T cells, with or without wild-type Treg cells, plus either wild-type or Ebi3−/− CD4+ Tconv cells. As expected, tumors were smaller (50–90 mm3) in recipients of CD4+ and CD8+ T cells that lacked Treg cells regardless of whether wild-type or Ebi3−/− CD4+ Tconv cells were transferred. Cotransfer of nTreg cells with wild-type CD4+ and CD8+ T cells blocked antitumor immunity and resulted in aggressive tumor growth (470 mm3; Fig. 7). Analysis of congenically marked tumor-infiltrating CD4+ T cells confirmed high expression of Ebi3 and Il12a similar to that of Treg cells (Supplementary Fig. 17). Thus, in this example, both IL-35-producing nTreg cells and iTR35 cells contributed to the suppressive milieu (Figs. 6 and 7 and Supplementary Fig. 17). Cotransfer of nTreg cells and CD8+ T cells with Ebi3−/− CD4+ Tconv cells (which are unable to be converted to IL-35-producing iTR35 cells) only partially blocked antitumor immunity, resulting in intermediate tumor growth (220 mm3). These results suggest that Treg cell–mediated induction of the develop-ment of iTR35 cells has a substantial effect on tumor burden and is responsible for approximately half the regulatory milieu in the tumor microenvironment (as the tumor burden was diminished from 470 mm3 to 220 mm3). Furthermore, these data suggest that nTreg cells mediate their suppressive activity in part by inducing iTR35 cells. Together these data suggest a model in which nTreg cells mediate the conversion of suppressed CD4+ T cells into iTR35 cells in an IL-35- and IL-10-dependent manner (Supplementary Fig. 18).

DISCUSSIONWe have shown here that iTR35 cells represent unique members of the Treg cell family that were generated by IL-35 and mediated their suppression exclusively via IL-35. They did not express Foxp3, did not require the other key suppressive cytokines (IL-10 or TGF-β) for

conversion and were distinct from the known induced regulatory populations of TGF-β–iTR cells and IL-10–iTR cells. We found that iTR35 cells were generated by a single, short-term stimulation of the T cell antigen receptor in the presence of IL-35 (mouse,3 d; humans, 6 d), unlike the other iTR populations described before, TGFβ iTR and IL-10–iTR cells, which require longer conversion protocols, multiple cell types and/or additional molecules for optimal generation13,38,39.

Infectious tolerance, whereby Treg cells confer a suppressive phe-notype on Tconv cells, has been described in both mouse and human systems25,27. It has been suggested that TGF-β might have a critical role in mediating infectious tolerance40. As IL-35-secreting nTreg cells converted Tconv cells into iTR35 cells, a suppressed Tconv cell popula-tion with regulatory potential, it is possible that IL-35 and iTR35 cells represent additional, important mediators of infectious toler-ance. Indeed, our data suggest that approximately half the regulatory microenvironment in the tumor was mediated by nTreg cell–induced iTR35 cells. This also suggested that iTR35 cells contributed to tumor progression. The generation of iTR cells from suppressed Tconv cells required IL-35 and, to a lesser extent, IL-10. IL-10 may directly potentiate the generation of iTR35 cells by IL-35-producing nTreg cells or it may simply slow the activation and/or proliferation of Tconv cells, thus indirectly facilitating the conversion to iTR35 cells. The generation of IL-35 in a purified, recombinant form has remained elusive, perhaps because of its apparent instability, and it is possible that this instability is physiologically important in limiting its effects to the local tissue environment. Single-cell analysis of human iTR35 cells by intracellular cytokine staining and confocal microscopy sug-gested that IL-35 induces homogeneous expression of IL-35 in human CD4+ Tconv cells. Whether this expression remains homogeneous in different in vivo inflammatory situations or with mouse iTR35 cells remains to be determined. Although iTR35 cells seemed to provide a substantial physiological contribution to the regulatory milieu estab-lished by nTreg cells in B16 melanoma, further studies are needed to fully elucidate their contribution in diverse disease settings.

Whether a similar system operates in humans remains to be determined. Studies suggest that human Treg cells do not make IL-35 (refs. 41,42); however, our unpublished data challenge those observations and emphasize the importance of further elucidation. Nevertheless, human iTR35 clls were generated and suppressed the proliferation of primary human T cells in an IL-35-dependent manner. Furthermore, mouse iTR35 cells were potently suppressive in five distinct in vivo models. The potential therapeutic applica-tion of ex vivo–generated IL-10–iTR cells and TGF-β–iTR cells is complicated by complexities in their generation, their short half-life and reversal of their suppressive capacity over time or by IL-2 (refs. 12,43,44). Although additional experiments are needed to fully assess the clinical potential of iTR35 cells, our data suggest that they represent a highly potent and stable iTR population that might have considerable therapeutic utility.

METHODSMethods and any associated references are available in the online version of the paper at http://www.nature.com/natureimmunology/.

Accession codes. GEO: microarray data, GSE24210 (series) and GSM595497–GSM595512 (specific array files).

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

ACKNoWLEDGMENTSWe thank R. Blumberg and T. Kuo (Brigham and Women’s Hospital) for Ebi3−/− mice; A. Rudensky (Memorial Sloan-Kettering Cancer Center) for Foxp3gfp

Figure 7 The suppressive T cell milieu in the tumor microenvironment is largely due to iTR35 cells. Tumor diameter of Rag1−/− mice reconstituted with wild-type C57BL/6 CD8+ T cells and wild-type or Ebi3−/− CD4+ Tconv cells with or without wild-type Treg cells, then injected intradermally the next day on the right flank with B16 cells (1.2 × 105), assessed on day 15 after inoculation. *P < 0.05 and **P < 0.001 (unpaired t-test). Data represent at least two independent experiments with six to twelve mice per group (mean and s.e.m.).

0CD8+: WT

WT Ebi3–/– Ebi3–/–WTWT–– WT

WT WT WTCD4+:

Treg:

100

200

Tum

or s

ize

(mm

3 )

300

400

500

NS ** **600

700

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunology  VOLUME 11 NUMBER 12 DECEMBER 2010 1101

A rt i c l e s

mice; J. Ihle (St. Jude Children’s Research Hospital; with permission from A. Rudensky) for Foxp3−/− mice; T. Geiger (St. Jude Children’s Research Hospital) for Il10−/− mice; B. Triplett, D. Regan, M. Howard and M. McKenna (St. Louis Cord Blood Bank) for cord blood samples; D. Campana (St. Jude Children’s Research Hospital) for the proprietary permeabilization buffer; A. Korman and M. Selby (Medarex–Bristol Myers Squibb) for MC38 colorectal adenocarcinoma cells; and S. Burns, H. Chi, R. Cross, K. Forbes, D. Green, G. Lennon, L. Jones, A. Krause, T. Moore, S. Morgan, A. Szymczak-Workman and K. Vignali for discussions and assistance. Supported by the National Institutes of Health (R01 AI39480 to D.A.A.V.; R01 AI61570 and R01 AI74878 to D.A.; and F32 AI072816 to L.W.C.), the Australian National Health and Medical Research Council Overseas Biomedical Fellowship Program (P.R.G.), the National Cancer Institute Comprehensive Cancer Center (CA21765 subaward to D.A.A.V.) and the American Lebanese Syrian Associated Charities (D.A.A.V.).

AUTHoR CoNTRIBUTIoNSL.W.C. designed (with help from D.A.A.V.) and did all mouse experiments, analyzed data and wrote the manuscript; V.C. did human experiments; A.L.H. (with L.W.C.) did the B16 tumor experiments; J.B. did the MC38 tumor experiments; P.R.G. infected mice with T. muris; C.G. did confocal microscopy; D.F. analyzed Affymetrix data; K.F. and S.A.B (with C.J.W.) generated and screened monoclonal antibodies to IL-35; C.J.W. coordinated the development of monoclonal anti-IL-35 and aided in figure preparation; M.L.J. generated and purified mouse Ebi3 protein for immunization and the development of monoclonal antibodies; H.-T.N. provided reagents and information; J.E.R. created and did histological analyses of Foxp3−/− mice; D.A. designed T. muris experiments and provided input on their interpretation; M.J.T. provided training for the B16 tumor model and provided input to research design and interpretation; and D.A.A.V. conceived of the research, directed the study and edited the manuscript.

CoMPETING FINANCIAL INTERESTSThe authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/natureimmunology/.

Published online at http://www.nature.com/natureimmunology/. reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/.

1. Fontenot, J.D., Rasmussen, J.P., Gavin, M.A. & Rudensky, A.Y. A function for interleukin 2 in Foxp3-expressing regulatory T cells. Nat. Immunol. 6, 1142–1151 (2005).

2. Gavin, M.A. et al. Foxp3-dependent programme of regulatory T-cell differentiation. Nature 445, 771–775 (2007).

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

4. Waldmann, H., Adams, E., Fairchild, P. & Cobbold, S. Infectious tolerance and the long-term acceptance of transplanted tissue. Immunol. Rev. 212, 301–313 (2006).

5. Selvaraj, R.K. & Geiger, T.L. Mitigation of experimental allergic encephalomyelitis by TGF-B induced Foxp3+ regulatory T lymphocytes through the induction of anergy and infectious tolerance. J. Immunol. 180, 2830–2838 (2008).

6. Nguyen, K.D., Vanichsarn, C. & Nadeau, K.C. Impaired IL-10-dependent induction of tolerogenic dendritic cells by CD4+CD25hiCD127lo/− natural regulatory T cells in human allergic asthma. Am. J. Respir. Crit. Care Med. 180, 823–833 (2009).

7. Bluestone, J.A. & Abbas, A.K. Natural versus adaptive regulatory T cells. Nat. Rev. Immunol. 3, 253–257 (2003).

8. Shevach, E.M. From vanilla to 28 flavors: multiple varieties of T regulatory cells. Immunity 25, 195–201 (2006).

9. Workman, C.J., Szymczak-Workman, A.L., Collison, L.W., Pillai, M.R. & Vignali, D.A. The development and function of regulatory T cells. Cell. Mol. Life. Sci. 6, 2603–2622 (2009).

10. Roncarolo, M.G. et al. Interleukin-10-secreting type 1 regulatory T cells in rodents and humans. Immunol. Rev. 212, 28–50 (2006).

11. Verbsky, J.W. Therapeutic use of T regulatory cells. Curr. Opin. Rheumatol. 19, 252–258 (2007).

12. Chen, W. et al. Conversion of peripheral CD4+CD25− naive T cells to CD4+CD25+ regulatory T cells by TGF-β induction of transcription factor Foxp3. J. Exp. Med. 198, 1875–1886 (2003).

13. Groux, H. et al. A CD4+ T-cell subset inhibits antigen-specific T-cell responses and prevents colitis. Nature 389, 737–742 (1997).

14. Collison, L.W. et al. The inhibitory cytokine IL-35 contributes to regulatory T-cell function. Nature 450, 566–569 (2007).

15. Sakaguchi, S. Regulatory T cells: key controllers of immunologic self-tolerance. Cell 101, 455–458 (2000).

16. Huter, E.N. et al. TGF-β-induced Foxp3+ regulatory T cells rescue scurfy mice. Eur. J. Immunol. 38, 1814–1821 (2008).

17. Workman, C.J. et al. In vivo Treg suppression assays. in Regulatory T cells: Methods and Protocols (Springer-Humana, New York, in the press).

18. Annacker, O., Pimenta-Araujo, R., Burlen-Defranoux, O. & Bandeira, A. On the ontogeny and physiology of regulatory T cells. Immunol. Rev. 182, 5–17 (2001).

19. Workman, C.J. et al. Lymphocyte activation gene-3 (CD223) regulates the size of the expanding T cell population following antigen activation in vivo. J. Immunol. 172, 5450–5455 (2004).

20. Kohm, A.P., Carpentier, P.A., Anger, H.A. & Miller, S.D. Cutting edge: CD4+CD25+ regulatory T cells suppress antigen-specific autoreactive immune responses and central nervous system inflammation during active experimental autoimmune encephalomyelitis. J. Immunol. 169, 4712–4716 (2002).

21. McGeachy, M.J., Stephens, L.A. & Anderton, S.M. Natural recovery and protection from autoimmune encephalomyelitis: contribution of CD4+CD25+ regulatory cells within the central nervous system. J. Immunol. 175, 3025–3032 (2005).

22. Turk, M.J. et al. Concomitant tumor immunity to a poorly immunogenic melanoma is prevented by regulatory T cells. J. Exp. Med. 200, 771–782 (2004).

23. Zhang, P., Cote, A.L., de Vries, V.C., Usherwood, E.J. & Turk, M.J. Induction of postsurgical tumor immunity and T-cell memory by a poorly immunogenic tumor. Cancer Res. 67, 6468–6476 (2007).

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

25. Jonuleit, H. et al. Infectious tolerance: human CD25+ regulatory T cells convey suppressor activity to conventional CD4+ T helper cells. J. Exp. Med. 196, 255–260 (2002).

26. Milojevic, D., Nguyen, K.D., Wara, D. & Mellins, E.D. Regulatory T cells and their role in rheumatic diseases: a potential target for novel therapeutic development. Pediatr Rheumatol Online J 6, 20 (2008).

27. Waldmann, H. Tolerance can be infectious. Nat. Immunol. 9, 1001–1003 (2008).

28. Collison, L.W., Pillai, M.R., Chaturvedi, V. & Vignali, D.A. Regulatory T cell suppression is potentiated by target T cells in a cell contact, IL-35- and IL-10-dependent manner. J. Immunol. 182, 6121–6128 (2009).

29. D’Elia, R., Behnke, J.M., Bradley, J.E. & Else, K.J. Regulatory T cells: a role in the control of helminth-driven intestinal pathology and worm survival. J. Immunol. 182, 2340–2348 (2009).

30. Di Stasi, A. et al. T lymphocytes coexpressing CCR4 and a chimeric antigen receptor targeting CD30 have improved homing and antitumor activity in a Hodgkin tumor model. Blood 113, 6392–6402 (2009).

31. Liyanage, U.K. et al. Prevalence of regulatory T cells is increased in peripheral blood and tumor microenvironment of patients with pancreas or breast adenocarcinoma. J. Immunol. 169, 2756–2761 (2002).

32. North, R.J. & Bursuker, I. Generation and decay of the immune response to a progressive fibrosarcoma. I. Ly-1+2− suppressor T cells down-regulate the generation of Ly-1−2+ effector T cells. J. Exp. Med. 159, 1295–1311 (1984).

33. Olkhanud, P.B. et al. Breast cancer lung metastasis requires expression of chemokine receptor CCR4 and regulatory T cells. Cancer Res. 69, 5996–6004 (2009).

34. Zou, W. Immunosuppressive networks in the tumour environment and their therapeutic relevance. Nat. Rev. Cancer 5, 263–274 (2005).

35. Zou, W. Regulatory T cells, tumour immunity and immunotherapy. Nat. Rev. Immunol. 6, 295–307 (2006).

36. Kocak, E. et al. Combination therapy with anti-CTL antigen-4 and anti-4–1BB antibodies enhances cancer immunity and reduces autoimmunity. Cancer Res. 66, 7276–7284 (2006).

37. Peggs, K.S., Quezada, S.A., Chambers, C.A., Korman, A.J. & Allison, J.P. Blockade of CTLA-4 on both effector and regulatory T cell compartments contributes to the antitumor activity of anti-CTLA-4 antibodies. J. Exp. Med. 206, 1717–1725 (2009).

38. Barrat, F.J. et al. In vitro generation of interleukin 10-producing regulatory CD4+ T cells is induced by immunosuppressive drugs and inhibited by T helper type 1 (Th1)- and Th2-inducing cytokines. J. Exp. Med. 195, 603–616 (2002).

39. Kemper, C. et al. Activation of human CD4+ cells with CD3 and CD46 induces a T-regulatory cell 1 phenotype. Nature 421, 388–392 (2003).

40. Andersson, J. et al. CD4+ FoxP3+ regulatory T cells confer infectious tolerance in a TGF-β-dependent manner. J. Exp. Med. 205, 1975–1981 (2008).

41. Allan, S.E., Song-Zhao, G.X., Abraham, T., McMurchy, A.N. & Levings, M.K. Inducible reprogramming of human T cells into Treg cells by a conditionally active form of FOXP3. Eur. J. Immunol. 38, 3282–3289 (2008).

42. Bardel, E., Larousserie, F., Charlot-Rabiega, P., Coulomb-L′Hermine, A. & Devergne, O. Human CD4+ CD25+ Foxp3+ regulatory T cells do not constitutively express IL-35. J. Immunol. 181, 6898–6905 (2008).

43. Horwitz, D.A. et al. Regulatory T cells generated ex vivo as an approach for the therapy of autoimmune disease. Semin. Immunol. 16, 135–143 (2004).

44. Schwartz, R.H. Models of T cell anergy: is there a common molecular mechanism? J. Exp. Med. 184, 1–8 (1996).

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunology doi:10.1038/ni.1952

ONLINE METHODSMice. Ebi3−/− mice were initially provided by R. Blumberg and T. Kuo. Foxp3gfp mice were provided by A. Rudensky. Foxp3−/− were provided by J. Ihle with permission from A. Rudensky. Il10−/− mice were provided by T. Geiger. CD4-dnTGFβRII, Il12a−/−, Rag1−/−, C57BL/6 and B6.PL mice were from the Jackson Laboratories. Animal experiments were done in specific pathogen–free facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, at St. Jude Children’s Research Hospital or the University of Pennsylvania (T. muris experiments) as approved by their respec-tive Institutional Animal Care and Use Committees according to national, state and institutional guidelines.

Human umbilical cord blood and peripheral blood. Human umbilical cord blood was obtained from the umbilical vein immediately after delivery (with informed consent provided the mother and approved by St. Louis Cord Blood Bank Institutional Review Board). Samples were provided by B. Triplett, D. Regan, M. Howard and M. McKenna at St. Louis Cord Blood Bank. Human peripheral blood cells were obtained from apheresis rings obtained from the St. Jude Blood Donor Center. All use at St. Jude Children’s Research Hospital was approved by the Institutional Review Board.

Generation of monoclonal antibody to IL-35. Recombinant mouse Ebi3 was cloned and expressed in a proprietary Escherichia coli expression system by M.L.J. and was used for immunization of Ebi3−/− mice. Clones V1.4F5.29, V1.4H6.25 and V1.4C4.22 were subsequently chosen for their utility in immunoprecipitation and immunoblot analysis and specific neutralization of IL-35 bioactivity.

IL-27 bioassay. Tconv cells were activated for 4 d in the presence of recom-binant IL-27 and IL-12 (50 ng/ml) and neutralizing monoclonal anti-IL-2. IL-27 activity was assessed by interferon-γ secretion in the presence of the appropriate monoclonal antibodies.

Enzyme-linked immunoassay of IL-35. Microtiter plates (96-well; Dynatech Laboratories) were coated with monoclonal anti–mouse Ebi3 (355028; R&D systems). Plates were blocked with 1% BSA and then washed, and standard or 293T cell culture medium was added. Plates were probed with biotinylated rat anti–mouse p35 (45806; R&D Systems) and streptavidin–horseradish peroxi-dase secondary antibody for detection. The concentration of IL-35 was deter-mined with a fusion protein of purified mouse IL-35 and the Fc fragment.

Generation and confirmation of recombinant IL-35. Because of its poor production and instability, recombinant IL-35 was generated by transient transfection of HEK293T cells as described14,45. A construct encoding Ebi3 and p35 linked by a glycine-serine linker or by a ‘self-cleaving’ 2A peptide was used for the generation of IL-35, and an empty vector was used as a control. HEK293T cells were transfected with TransIT transfection reagent (Mirus), the medium was changed, and cells were cultured for an additional 36 h to facili-tate protein secretion. Dialyzed, filtered supernatant from cells was used at a volume of 25% of total culture medium to induce conversion to iTR35 cells (full details of IL-35 confirmation available on request). Confirmation was achieved by neutralization of bioactivity with specific antibodies; demonstration that bioactivity was removed by antibody precipitation; and use of beads to present bioactive IL-35. Beads were generated that presented IL-35 in a suppressive or nonsuppressive format. Neutralizing monoclonal antibody (for nonsuppres-sive beads; 27537; R&D Systems), non-neutralizing monoclonal antibody (for suppressive beads; 25806, R&D Systems) or the appropriate isotype-matched control antibody (human immunoglobulin G1 (1171; R&D Systems) or rat immunoglobulin G2 (5447; R&D Systems)) was added for 4 h to supernatants of HEK293T cells transfected with IL-35-expressing vector or control vector, and protein G beads were added for an additional 12–18 h. Beads were cultured for 3–6 d with Tconv cells activated with anti-CD3 and anti-CD28 for analysis of suppressive capacity.

Beads coated with anti-CD3 and anti-CD28. Sulfate beads 4 μm in diameter (Molecular Probes) were incubated with 13.3 μg/ml of anti-CD3 (mouse, 145-2C11; human, OKT3; Biolegend) and 26.6 μg/ml of anti-CD28 (mouse, 37.51; human, CD28.2; Biolegend) to facilitate conjugation.

Flow cytometry, intracellular staining and cell sorting. Tconv cells (CD4+CD25−CD45RBhi) and Treg cells (CD4+CD25+CD45RBlo) were puri-fied by flow cytometry. Cell surface molecules were stained with the following fluorescence-conjugated monoclonal antibodies: anti-CD4 (GK1.5 or RM4-5), anti-CD69 (H1.2F3), anti-CD25 (PC61) and anti-CD73 (TY/11.8; all from Biolegend); and anti-LAG-3 (C9B7W), anti-CTLA-4 (UC10-4F10-11) and anti-CD28 (37.51; all from BD Pharmingen). For intracellular staining, human cord Tconv cells were activated for 9 d with beads coated with anti-CD3 and anti-CD28 in the presence of supernatants of HEK293T cells transfected with IL-35-expressing vector or control vector14. A proprietary permeabilization buffer developed in the laboratory of D. Campana was used for fixation and permeabilization. Cells were stained with monoclonal anti-p35 (27537; R&D Systems) or immunoglobulin G1 isotype-matched control antibody (1171; R&D Systems) and were analyzed by flow cytometry.

Mouse iTR35 cells, suppressed Tconv cells, suppressive Treg cells and TGF-β–iTR cells. Tconv cells were activated with beads coated with anti-CD3 and anti-CD28 and 293T cell–generated IL-35 at a volume 25% of the total culture volume for the generation of mouse iTR35 cells14. Where indicated, recom-binant IL-10, TGF-β or IL-27 was added at a concentration of 100 ng/ml. For the generation of suppressed Tconv cells and suppressive Treg cells, puri-fied Tconv cells were activated for 72 h with beads coated with anti-CD3 and anti-CD28 and Treg cells. Suppressed Tconv and suppressive Treg cells from the coculture were resorted on the basis of congenic markers or labeling with CFSE (carboxyfluorescein diacetate succinimidyl ester). For the conversion of TGF-β–iTR cells, TGF-β (5 ng/ml) was added for 5 d to cultures containing Tconv cells activated with anti-CD3 and anti-CD28. Where indicated, neu-tralizing anti-IL-10 (10 μg/ml; JES5-2A5; BD Bioscience), anti-TGF-β (10 μg/ml; 1D11; R&D Systems) or anti-IL35 (10 μg/ml; V1.4H6.25; produced in-house as described above) was added.

Suppression of human Tconv cells and conversion of iTR35 cells by human IL-35. Mononuclear cells from human cord blood were separated on a Ficoll gradient and Tconv cells and Treg cells were purified by flow cytometry on the basis of their expression of CD4 (OKT4; Biolegend) and CD25 (BC96; Biolegend). Anti-CD45RA (HI100; Biolegend) was added as an additional marker for the purification of Tconv cells and Treg cells from peripheral blood. Purity was verified with an intracellular Foxp3 staining kit (88-8999-40; eBioscience). Tconv cells were activated in X-VIVO medium containing beads coated with anti-CD3 and anti-CD28, 20% (vol/vol) human sera (Lonza) and 100 units/ml of human IL-2. Human IL-35 was generated and tested as described for mouse IL-35, with a few exceptions14. Human cells were cultured for 9 d and were resorted for proliferation and suppression assays to assess the activity of iTR35 cells. In addition, suppression assays were carried out for 6 d. Where indicated, neutralizing anti-IL-10 (10 g/ml; JES3D97; Biolegend), anti-TGF-β (10 g/ml; 1D11; R&D Systems) or anti-p35 (10 g/ml; 27537; R&D Systems).

Confocal microscopy. Purified iTRcon cells and iTR35 cells were allowed to adhere to chambered coverglass (Nunc), were fixed with 4% (vol/vol) formal-dehyde (Polysciences) and were made permeable with 0.2% (vol/vol) Triton X-100 in PBS. After being blocked, cells were stained with anti–human IL12A (p35; 27537; R&D Systems) or isotype-matched control antibody (1171; R&D Systems). The actin cytoskeleton was detected with phalloidin (Molecular Probes) and nuclei were labeled with DAPI (4,6-diamidino-2-phenylindole; Molecular Probes). Cells were imaged with a Zeiss inverted spinning-disc confocal microscope and Slidebook acquisition and analysis software (Intelligent Imaging Innovations).

Cytokine analysis. A Multiplex bead–based analysis (Millipore) was used for simultaneous quantification of the secretion of 26 cytokines and chemokines after 72 h of activation of iTRcon cells or iTR35 cells.

RNA, cDNA and quantitative real-time PCR. RNA was isolated with the Qiagen microRNA extraction kit and cDNA was reverse-transcribed with the cDNA Archive kit (Applied Biosystems). The cDNA samples were subjected to 40 cycles of amplification with TaqMan primers and probes in an ABI Prism 7900 Sequence Detection System instrument and were quantified by the com-parative cycling threshold method.

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature immunologydoi:10.1038/ni.1952

Immunoprecipitation and immunoblot analysis. These were done as described33,34. All cells were cultured for 24 h and culture supernatants were collected for analysis. Supernatants were immunoprecipitated with anti–mouse IL12a (p35; 45806; R&D Systems) and protein G–Sepharose beads. Immunoprecipitates were resolved by SDS-PAGE (Invitrogen Life Technologies) and blots were probed with biotinylated monoclonal anti–mouse Ebi3 (V1.4F5.29; produced in-house as described above).

In vitro proliferation and suppression assays. Cells (2.5 × 104) were acti-vated for 72 h with beads coated with anti-CD3 and anti-CD28. Cultures were pulsed with 1 mCi [3H]thymidine for the final 8 h of the 72-hour assay and cells were collected for analysis of incorporation (by measurement of c.p.m.). Suppression assays were done as described with some modifications46. Cells monitored for suppressive capacity were mixed with Tconv cells and beads coated with anti-CD3 and anti-CD28 at a responder/suppressor ratio of 4:1. Transwell experiments used 96-well plates with a pore size of 0.4 μm (Millipore). Responder Tconv cells (5 × 104) in the bottom chamber were acti-vated with beads coated with anti-CD3 and anti-CD28, and iTR35 or iTRcon cells were incubated in the top chamber. After 64 h of culture, [3H]thymidine was added directly to the responder Tconv cells in the bottom chamber for the final 8 h of the 72-hour assay.

Affymetrix analysis of genetic signature. The generation of iTRcon and iTR35 cells was verified by quantitative PCR analysis of the expression of Ebi3 and Il12a as well as IL-35 secretion. RNA was extracted and quality was confirmed by ultraviolet spectrophotometry and by analysis on an Agilent 2100 Bioanalyzer (Agilent Technologies). Total RNA (100 ng) was processed in the Hartwell Center for Biotechnology & Bioinformatics according to the Affymetrix eukaryote two-cycle target labeling protocol. Biotin-labeled cRNA (20 μg) was hybridized overnight to the mouse 430v2 GeneChip array, then the arrays were scanned and expression values were summarized with the default parameters of the MAS5 algorithm as implemented in the GCOS 1.4 software (Affymetrix). Signals were normalized for each array by scaling to a 2% trimmed mean of 500. Nine Treg cell, three activated Treg cell and six resting Treg cell samples were compared with nine Tconv cell, three activated Tconv cell and six resting Tconv cell samples. These samples were examined with Mouse Genome 430 2.0 Arrays, and MAS 5.0 signal data werecollected and transformed by log start to stabilize the variance. For analysis of significance, a t-test was applied to each probe set. Probe sets with a P value of less than 0.0001, an absolute value log ratio of Treg cells to Tconv cells of at least 3 (log2) and a defined gene name were selected. Scores were calculated for each gene and rescaled from 0 to 1 according to the following formula: gene score (g) = (observed mean g – minimum mean g) / (maximum mean g – minimum mean g). A score of 0 is blue, 0.5 is black and 1 is yellow; intermediate values are shaded on this scale in the heat map of scores (Supplementary Fig. 11a,b,e). Magnitude and abundance plots and volcano plots were generated with Stata/SE 11.0 data analysis and statistical software (special edition for analaysis of large data sets; STATA). Statistical tests and batch-effect removal were done with the Partek Genomics Suite. Expression data from the Affymetrix Mouse Genome 430 2.0 arrays were analyzed as MAS 5.0 signals transformed by log start according to the following formula: log signal = ln (signal + 20) (ref. 47). The log2 ratios were calculated in STATA according to the following formula:

log ratio of A to B = log(exp(mean log signal A) / exp(mean log signal b)) / log(2). The average log signal in the magnitude and abundance plots is the average of the two means in the log ratio: A = (mean log signal A) + (mean log signal b) / 2. The Treg and Tconv comparison and IL-35-treated versus control were corrected for batch effects. The partial P value for the comparison of interest was then transformed to generate the significance score: score = −log10 (P value).

Assessment of iTR cell stability in vivo. The iTR35 or TGF-β–iTR cells gener-ated in vitro with CD45.2+ Tconv cells were adoptively transferred intravenously into CD45.1+ C57BL/6 mice. At 7, 15 or 25 d after transfer, splenic iTR cells were counted by flow cytometry of CD45.2+ cells. The frequency of recovered cells relative to total cells injected was used to calculate percent recovery. For analysis of suppressive capacity, Tconv cells were mixed for 72 h at various ratios with recovered iTR cells and beads coated with anti-CD3 and anti-CD28.

In vivo animal models for analysis of regulatory cell function. The Foxp3−/− rescue model17,48, homeostasis model14,17,19, IBD model17,49 and EAE disease model5,17 were done as described. For IL-35 expression analysis in the B16 and MC38 tumor models, Foxp3gfp or Ebi3−/−Foxp3gfp mice were inoculated with B16 melanoma or MC38 colorectal adenocarcinoma cells. The B16 melanoma model was done as described17,22. MC38 colorectal adenocarcinoma cells (a gift from A. Korman and M. Selby)36,37 were cultured in DMEM and 10% (vol/vol) FBS, then were washed and resuspended in PBS before subcutaneous injection of 2 × 106 cells into the flank. MC38 tumors were excised at 12 d, when tumors were 5–15 mm in diameter. In some homeostasis model experi-ments, mice were treated with anti-IL-35 (V1.4C4.22) or isotype-matched control antibody (MPC11; BioXcell) according to the following dosing regi-men: day 0, 500 g; day 3, 250 g; day 6, 250 g.

T. muris infection. T.muris was maintained in genetically susceptible mice (such as AKR/J and B6.CB17-Prkdcscid/SzJ mice) and T. muris eggs were isolated as described50. Mice were infected orally with 30 embryonated T. muris eggs and killed on day 14 after infection. CD4+Foxp3+ and CD4+Foxp3−fractions were enriched with a FACSDiva, then mRNA was isolated and real-time PCR analysis was done as described above.

Statistical analysis. Statistical significance was determined with an unpaired t-test with GraphPad Prism software except where noted otherwise (Fig. 3g,h, Wilcoxon matched-pairs test).

45. Vignali, D.A. & Vignali, K.M. Profound enhancement of T cell activation mediated by the interaction between the TCR and the D3 domain of CD4. J. Immunol. 162, 1431–1439 (1999).

46. Huang, C.T. et al. Role of LAG-3 in regulatory T cells. Immunity 21, 503–513 (2004).

47. Rocke, D.M. & Durbin, B. Approximate variance-stabilizing transformations for gene-expression microarray data. Bioinformatics 19, 966–972 (2003).

48. Koch, M.A. et al. The transcription factor T-bet controls regulatory T cell homeostasis and function during type 1 inflammation. Nat. Immunol. 10, 595–602 (2009).

49. Mottet, C., Uhlig, H.H. & Powrie, F. Cutting edge: cure of colitis by CD4+CD25+ regulatory T cells. J. Immunol. 170, 3939–3943 (2003).

50. Artis, D. et al. The IL-27 receptor (WSX-1) is an inhibitor of innate and adaptive elements of type 2 immunity. J. Immunol. 173, 5626–5634 (2004).

© 2

010

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.