nature immunology  VOLUME 17 NUMBER 2 FEBRUARY 2016 187

Self-reactive CD4+ T cells are generated in the normal course of rear-rangement of the T cell antigen receptor (TCR)1. Low-affinity inter-actions between the TCR and ligands composed of self peptide and major histocompatibility (MHC) class II (self peptide–MHC class II) are essential for the positive selection of CD4+ T cells in the thymus1. However, high-affinity interactions between the TCR and self peptide–MHC class II can cause autoimmunity. Therefore, tolerance mechanisms have evolved to limit T cells with TCRs with high affinity for self peptide–MHC class II ligands.

Different tolerance mechanisms may be engaged depending on where a self peptide–MHC class II epitope is displayed. A develop-ing thymocyte may die by apoptosis, or it may differentiate into a regulatory T cell (Treg cell) if its TCR binds strongly to a self peptide– MHC class II epitope displayed on particular antigen-presenting cells1,2. Epitopes in this class may be derived from proteins expressed or taken up by MHC class II–positive (MHCII+) thymic epithelial cells or dendritic cells. However, MHCII+ cells in the thymus might not display epitopes from all host proteins. Thus, T cells expressing TCRs with high affinity for certain self epitopes may not be deleted in the thymus and thus may exit to secondary lymphoid organs2–6. Some T cells in this class may be specific for epitopes from proteins that are not secreted or are expressed in locations with poor lymphatic drain-age. Since these proteins would not reach antigen-presenting cells in

secondary lymphoid organs, specific T cells would be retained in the repertoire in a fully responsive state of ignorance7,8. Other tissue- restricted proteins that are secreted or are present in dying cells may reach secondary lymphoid organs, where they would be processed by antigen-presenting cells to form self peptide–MHC class II epitopes. CD4+ T cells that encounter and recognize these epitopes may differentiate into Treg cells2 or may become anergic due to a lack of costimulation derived from the innate immune system1,9.

Much of the evidence cited above has been provided by studies of mice with transgenic expression of monoclonal TCRs specific for self peptide–MHC class II epitopes7,8,10. Analysis of such mice can reveal tolerance mechanisms because of the large numbers of T cells avail-able for study. However, this approach has several limitations. Because a mouse with transgenic TCR expression contains a single TCR from a larger repertoire, it can reveal only those tolerance mechanisms that apply to TCRs with that affinity for a self peptide–MHC class II epitope11. Further, it is clear that an abundance of T cells with the same TCR can create abnormal outcomes12–14. This issue has come to the fore in several studies showing that clonal deletion is not as prominent a mechanism in polyclonal T cell repertoires, as has been reported for certain monoclonal T cell populations4,5,15. These studies, however, have been limited to epitopes with ubiquitous patterns of expression.

1Department of Microbiology and Immunology, Center for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. 2Department of Laboratory Medicine and Pathology, Center for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. 3Diabetes Center, University of California San Francisco, San Francisco, California, USA. 4Department of Medicine, Center for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. 5Department of Laboratory Medicine and Pathology, Institute for Translational Neuroscience, University of Minnesota Medical School, Minneapolis, Minnesota, USA. 6Present address: Laboratory of Parasitic Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland, USA. Correspondence should be addressed to M.K.J. (jenki002@umn.edu).

Received 21 August 2015; accepted 21 October 2015; published online 4 January 2016; doi:10.1038/ni.3327

Tolerance is established in polyclonal CD4+ T cells by distinct mechanisms, according to self-peptide expression patternsDeepali Malhotra1, Jonathan L Linehan1,6, Thamotharampillai Dileepan1, You Jeong Lee2, Whitney E Purtha3, Jennifer V Lu3, Ryan W Nelson1, Brian T Fife4, Harry T Orr5, Mark S Anderson3, Kristin A Hogquist2 & Marc K Jenkins1

Studies of repertoires of mouse monoclonal CD4+ T cells have revealed several mechanisms of self-tolerance; however, which mechanisms operate in normal repertoires is unclear. Here we studied polyclonal CD4+ T cells specific for green fluorescent protein expressed in various organs, which allowed us to determine the effects of specific expression patterns on the same epitope-specific T cells. Peptides presented uniformly by thymic antigen-presenting cells were tolerated by clonal deletion, whereas peptides excluded from the thymus were ignored. Peptides with limited thymic expression induced partial clonal deletion and impaired effector T cell potential but enhanced regulatory T cell potential. These mechanisms were also active for T cell populations specific for endogenously expressed self antigens. Thus, the immunotolerance of polyclonal CD4+ T cells was maintained by distinct mechanisms, according to self-peptide expression patterns.

A rt i c l e s

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

188  VOLUME 17 NUMBER 2 FEBRUARY 2016 nature immunology

A rt i c l e s

A need therefore exists for a systematic study of tolerance within normal polyclonal CD4+ T cell repertoires specific for self peptides with defined patterns of expression. Here we carried out such a study using cellular enrichment based on peptide–MHC class II tetramers. Intrathymic deletion was the mechanism used for T cells specific for peptides from proteins expressed uniformly by thymic dendritic cells and/or medullary thymic epithelial cells, whereas ‘ignorance’ was the mechanism used for tolerance to peptides from cytosolic or nuclear proteins expressed only outside the thymus. Partial intrathymic deletion and impaired effector T cell potential but enhanced Treg cell potential among remaining clones was the mechanism of tolerance for many peptides from tissue-restricted proteins that were also expressed sparsely in the thymus. Thus, CD4+ T cells tolerate self peptides with different expression patterns via different mechanisms.

RESULTSA transgenic model of CD4+ T cell tolerance to selfWe used enhanced green fluorescent protein (eGFP), enhanced yellow flu-orescent protein (eYFP), or cyan fluorescent protein (CFP) as model self antigens in mouse strains with transgenic expression of these proteins and the MHC class II molecule I-Ab. These fluorescent proteins all contain an I-Ab-binding peptide first identified in eGFP (HDFFKSAMPEGYVQE;

designated eGFPp), which was incorporated into an I-Ab tetramer used for the detection of CD4+ T cells with TCRs specific for this epitope16 (Fig. 1a,b). This experimental system allowed us to determine the effects of different patterns of self-antigen expression on the same epitope-specific T cell repertoire. Another advantage of this approach was that wild-type C57BL/6 (B6) mice, which express the I-Ab molecule but not eGFP, eYFP or CFP, could be used as controls with which to study the relevant T cell repertoire in the absence of the self peptide.

We assessed tolerance in the eGFPp–I-Ab–specific CD4+ T cell repertoire in 14 different strains with transgenic expression of eGFP, eYFP or CFP (Supplementary Table 1) by immunizing the mice with eGFPp in complete Freund’s adjuvant (CFA). We obtained single-cell suspensions of secondary lymphoid organs from the immunized mice and stained these with the eGFPp–I-Ab tetramer labeled with allophycocyanin, then with magnetic beads coated with antibody to allophycocyanin. The suspensions underwent enrichment for tetramer-bound cells on magnetized columns, then were eluted and stained with antibodies designed to identify T cells. We assessed the differentiation of eGFPp–I-Ab–specific CD4+ T cells into the TH1, TH17, follicular helper T cell (TFH cell) and Treg cell subsets by intra-cellular staining of the lineage-defining transcription factors T-bet, RORγt, Bcl-6 and Foxp3, respectively (Fig. 1c).

101

102

103

104

105

106

CD

44hi

Tet

+ C

D4+

T c

ells

a

UBCeG

FP

ACTBeG

FP

ItgaxeY

FP

AireeG

FP

Cdh1CFP

ChateGFP

Pcp2eG

FP

Myh

11eG

FP

Ins2eG

FP

CD207eY

FP

Zfp42

3eG

FP

Foxp3

eGFP

Foxd1

eGFP

Ins1eG

FPW

T

****** ***

NS******

*** ******** NS***

***

d

CD44hi Tet+ cells

T reg

T FH T H1

T H17

1

2

3

0 9619228

838

347

957

567

01,

8383,

0054,

1735,

3406,

5087,

676

WTIns1eGFP

Foxd1eGFP

Ins2eGFP

CD207eYFP

Myh11eGFP

Foxp3eGFP

Zfp423eGFP

Pcp2eGFP

UBCeGFP

ItgaxeYFP

ChateGFP

Cdh1CFP

AireeGFP

ACTBeGFP

b

0 103 104 105

0

103

104

105

102

56

Tet

CD

44

WTWT

0 103 104 105

0

103

104

105

0

103

104

1051.1Treg

13.1TH17

85.8

0 103 104 105–103

57.0TFH

22.0TH1

Foxp3

RO

Rγt

T-b

et

Bcl-6

c

CD

44

4.5

0.06

57.5Treg

1.5TH17

41.0

39.4TFH

11.2TH1

RO

Rγt

Cluster 2:

Ins2eGFP

Cluster 3:

UBCeGFP

64.3Treg

7.1TH17

0TFH

0TH1

28.6

332.3Treg

9.8TH17

87.9

0

103

10–3

104

105

0

104

103

105

0

103

104

105

0 103 104 105 0 103 104 105 0 103 104 105

53.1TFH

18.8TH1

Cluster 1:

Ins1eGFP

Cluster 2:

Ins2eGFP

Cluster 3:

UBCeGFP

Cluster 1:

Ins1eGFP

e

Tet Foxp3 Bcl-6

T-b

et

Figure 1 Three main patterns of tolerance to self antigens. (a) Total CD44hiTet+ (eGFPp–I-Ab-tetramer (Tet)-binding) CD4+ T cells from pooled secondary lymphoid organs of wild-type (WT) and transgenic mice 13–14 d after immunization with 100 µg eGFPp emulsified in CFA. Each symbol represents an individual mouse (colors match those of clusters identified in d); small horizontal lines indicate the geometric mean. (b) Flow cytometry analyzing CD44 and tetramer (Tet) staining in tetramer-enriched CD4+ T cells from pooled secondary lymphoid organs of wild-type (WT) mice primed 14 d earlier with eGFPp in CFA. (c) Expression of RORγt and Foxp3 by Tet+ cells (left) and of T-bet and Bcl-6 by Foxp3−RORγt−Tet+ cells (right) among the CD44hiTet+ CD4+ T cells identified in b. (d) Unsupervised hierarchical clustering of CD44hiTet+ CD4+ T cells (geometric mean cell number and phenotype) in 15 mouse strains (right margin), grouped into three main clusters (far left and right). (e) Flow cytometry analyzing CD44 and tetramer staining of tetramer-enriched CD4+ T cells from pooled secondary lymphoid organs of three eGFP-expressing mouse strains in clusters 1–3 identified in (d) (left), and transcription factor expression as in c by the CD4+CD44hiTet+ cells identified at left (right). Numbers adjacent to outlined areas (b,c,e) indicate percent cells in each. NS, not significant (P > 0.05); *P < 0.05, **P < 0.005 and ***P < 0.0001, versus wild-type B6 mice (one-way analysis of variance (ANOVA) of log10-transformed values). Data are pooled from 12 independent experiments with 3–23 mice per group (a,d) or are representative of 12 experiments (b,c) or eight experiments (e).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 17 NUMBER 2 FEBRUARY 2016 189

A rt i c l e s

Wild-type B6 mice, for which eGFPp is a foreign peptide, had on average 23,000 eGFPp–I-Ab–specific CD44hi CD4+ T cells in their secondary lymphoid organs 14 d after immunization (Fig. 1a). Most of these cells were Foxp3− effector T cells, although a small subpopulation of Foxp3+ Treg cells was present (Fig. 1c). The Foxp3− population consisted of about 10% RORγt+ TH17 cells, 20% T-bet+ TH1 cells, 50% Bcl-6+ TFH cells, and 20% cells that could not be assigned to a known lineage (Fig. 1c).

The mice with transgenic expression of eGFP, eYFP or CFP gen-erated a spectrum of eGFPp–I-Ab–specific CD44hi CD4+ effector T cell numbers in secondary lymphoid organs following immuniza-tion. At one end of the spectrum, mice with transgenic expression of eGFP from the gene encoding insulin 1 (Ins1eGFP) produced the same number of cells as wild-type mice did (Fig. 1a), indicative of a com-plete lack of tolerance. At the other end of the spectrum, mice with transgenic expression of eGFP from the gene encoding ubiquitin C (UBCeGFP) generated fewer than 100 eGFPp–I-Ab–specific CD44hi T cells (Fig. 1a), indicative of a profound state of tolerance. The other 12 transgenic strains produced 200–10,000 eGFPp–I-Ab–specific CD44hi T cells (Fig. 1a).

The eGFPp–I-Ab–specific CD44hi T cells in the transgenic mice also displayed a range of phenotypes. Treg cells, TH1 cells, TFH cells and TH17 cells were generated in each strain by the priming of mice with eGFPp in CFA. We then analyzed the data with an unsupervised hierarchical clustering algorithm17 to search for patterns of tolerance induction

Figure 2 Ins1eGFP mice ‘ignore’ the eGFPp–I-Ab epitope. (a) Flow cytometry analyzing eGFPp–I-Ab–streptavidin– phycoerythrin (Tet-PE) staining versus eGFPp–I-Ab–streptavidin–allophycocyanin (Tet-APC) staining of tetramer-enriched CD4+ T cells from pooled spleens and lymph nodes of unimmunized wild-type or Ins1eGFP mice. Numbers in top right quadrants indicate percent double-tetramer-positive eGFPp–I-Ab–specific (double-Tet+) cells. (b) Quantification of double-Tet+ CD4+ T cells from pooled spleens and lymph nodes of unimmunized wild-type and Ins1eGFP mice. Each symbol represents an individual mouse; small horizontal lines indicate the geometric mean. (c) Frequency of CD44lo cells among double-Tet+ CD4+ T cells from pooled spleens and lymph nodes of unimmunized wild-type or Ins1eGFP mice. P values (NS), unpaired t-test of log10-transformed (b) or linear (c) values. Data are representative of five experiments (a) or are pooled from five experiments with 16–17 mice per group (b, geometric mean; c, mean and s.d.).

(Fig. 1d). We used absolute cell numbers for this clustering to capture information about clonal proliferation and differentiation. This analysis revealed three major clusters. Cluster 1 was typified by Ins1eGFP mice and was characterized by wild-type numbers of TFH cells, TH1 cells and TH17 cells and relatively few Treg cells (Fig. 1d,e). Cluster 2 was exemplified by Ins2eGFP mice and had small numbers of TFH cells, TH1 cells and TH17 cells and relatively large numbers of Treg cells (Fig. 1d,e). Cluster 3 was typified by UBCeGFP mice and was charac-terized by a profound reduction in all subsets (Fig. 1d,e). These results suggested that three main mechanisms accounted for the CD4+ T cell tolerance to self antigens with different patterns of expression.

Ignorance is the tolerance mechanism for cluster 1Cluster 1 included wild-type mice and mice with patterns of very limited eGFP expression: Ins1eGFP mice express eGFP exclusively in pancreatic beta cells18,19, whereas mice with transgenic expres-sion of eGFP from the gene encoding the transcription factor Foxd1 (Foxd1eGFP) express eGFP during kidney and eye development20,21. Because Ins1eGFP and Foxd1eGFP mice responded to immunization with eGFPp like mice that did not express eGFP (Fig. 1), it is likely that their eGFPp–I-Ab–specific T cell repertoires never encountered this epitope and were in a naive state. We tested this hypothesis in Ins1eGFP mice by quantifying eGFPp–I-Ab–specific CD4+ T cells in unimmunized mice (Fig. 2a,b). We stained single-cell suspensions of secondary lymphoid organs with eGFPp–I-Ab tetramers labeled with

0.044

Ins1eGFP

0

0.052

WTa

0102

103 104 105

103

104

105

Tet

-PE

Tet-APC

b102

101

100Dou

ble-

Tet

+ C

D4+

T c

ells

Ins1eGFPWT

NS

CD

44lo

cells

(%

)

c

020406080

100

Ins1eGFPWT

NS

Dou

ble-

Tet

+

CD

4+ T

cel

ls

101

102

WT

Ins2eG

FP

d *

101

100

102

Dou

ble-

Tet

+C

D4+

CD

8–

thym

ocyt

es

*b

WT

Ins2eG

FP

0 102 103 104 105

0102

103

104

105 0.10 0.07

Ins2eGFPWTa

Tet (APC)

Tet

(P

E)

0

WT Ins2eGFP

0102 103 104 105

102

103

104

105 0.15 0.13

c

Tet (APC)

Tet

(P

E)

2080

0 102 103 104 105

0

50K

100K

150K

200K

250K

0100

Foxp3

SS

C-A

e WT Ins2eGFP f

0

1020304050

Mic

e w

ith d

oubl

e-T

et+

Tre

g ce

lls (

%) *

WT

Ins2eGFPFigure 3 eGFPp–I-Ab–specific CD4+ T cells undergo limited clonal deletion in the thymus. (a) Flow cytometry analyzing Tet-PE versus Tet-APC staining (as in Fig. 2a) of tetramer-enriched CD4+CD8− thymocytes from wild-type or Ins2eGFP mice. (b) Quantification of double-Tet+ CD4+CD8− thymocytes from wild-type and Ins2eGFP mice. (c) Flow cytometry analyzing Tet-PE versus Tet-APC staining (as in Fig. 2a) of tetramer-enriched CD4+ T cells from pooled spleens and lymph nodes of unimmunized wild-type or Ins2eGFP mice. (d) Quantification of double- Tet+ CD4+ T cells from pooled spleens and lymph nodes of wild-type or Ins2eGFP mice. (e) Flow cytometry analyzing Foxp3 expression by double-Tet+ CD4+CD8− thymocytes from wild-type and Ins2eGFP mice. SSC, side scatter. (f) Frequency of wild-type and Ins2eGFP mice with double-Tet+ Treg cells detected in the thymus. Each symbol (b,d) represents an individual mouse; small horizontal lines indicate the geometric mean. Numbers in top right quadrants (a,c) indicate percent double-Tet+ cells; numbers above outlined areas (e) indicate percent Foxp3− cells (left) or Foxp3+ cells (right). *P < 0.05 (unpaired t-test of log10-transformed values (b,d) or χ2 test (f)). Data are representative of five experiments (a,e) or eight experiments (c) or are pooled from five experiments with 17–19 mice per group (b), eight experiments with 20–41 total mice per group (d) or five experiments with 23 wild-type mice and 20 Ins2eGFP mice (f).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

190  VOLUME 17 NUMBER 2 FEBRUARY 2016 nature immunology

A rt i c l e s

allophycocyanin or phycoerythrin, then with magnetic beads coated with antibody to phycoerythrin or antibody to allophycocyanin before magnetic enrichment. We stained cells with the same tetramer labeled with two different fluorochromes to maximize the TCR specificity of the assay16. Ins1eGFP mice had an average of 19 eGFPp–I-Ab–specific CD4+ T cells in their secondary lymphoid organs, similar to wild-type mice (Fig. 2b). These populations also had similar frequencies of naive-phenotype (CD44lo) cells (Fig. 2c). Together with the observa-tion that Ins1eGFP and wild-type mice mounted comparable responses following immunization, these results indicated that tolerance in cluster 1 was due to ignorance.

Partial clonal deletion and Treg cell induction shape cluster 2Cluster 2 was characterized by low numbers of effector cells and rela-tively large numbers of Treg cells in mice primed with eGFPp (Fig. 1). It consisted of several mice with tissue-restricted promoters that also had sparse expression in the thymus. Ins2eGFP mice express eGFP fused to the C peptide of pro-insulin in pancreatic beta cells22. They also express eGFP in a small number of medullary thymic epithelial cells19. Mice with transgenic expression of eGFP from the gene encoding the transcription factor Foxp3 (Foxp3eGFP) express eGFP in developing Treg cells in the thymus, whereas mice with transgenic expression of eYFP from the gene encoding langerin (CD207eYFP)23 express eYFP in epidermal Langerhans cells and perhaps thymic dendritic cells. Mice with transgenic expression of eGFP from the gene encoding the contractile protein smooth muscle myosin heavy chain 11 (Myh11eGFP) or the transcription factor zinc finger protein 423 (Zfp423eGFP) are expected to express eGFP at moderate levels in medullary thymic epithelial cells24. Mice with transgenic expression of eGFP from the gene encoding Purkinje cell protein 2 (Pcp2eGFP), which express eGFP in cerebellar Purkinje neurons in the brain25, were also in this cluster. We selected Ins2eGFP mice for further analysis of the tolerance mechanisms operating in cluster 2.

We sought evidence for intrathymic clonal deletion by quantifying eGFPp–I-Ab–specific CD4+CD8− thymocytes. Wild-type mice had an average of 16 eGFPp–I-Ab–specific CD4+CD8− thymocytes and 20 peripheral CD4+ T cells, whereas Ins2eGFP mice had an average of 10 and 13 such cells, respectively (Fig. 3a–d). These results indicated that about 30% of eGFPp–I-Ab–specific T cell clones were deleted in Ins2eGFP mice. Foxp3+ eGFPp–I-Ab–specific CD4+CD8− thymocytes were detectable in wild-type and Ins2eGFP thymuses, although not in all mice, due to the small size of this population (Fig. 3e,f). However, significantly more Ins2eGFP thymuses (40%) than wild-type thymuses contained eGFPp–I-Ab–specific Foxp3+ CD4+CD8− thymocytes (Fig. 3e,f). Thus, Treg cells were spared from deletion or were even increased in the presence of this self antigen.

The clones that were deleted in Ins2eGFP mice might have had TCRs with the highest affinities for the eGFPp–I-Ab epitope. We addressed

this possibility by measuring the mean fluorescence intensity of tetramer binding, which is a direct correlate of TCR affinity for peptide– MHC class II26. Thymocytes could not be used for this purpose because they were too rare for accurate measurement. We therefore used expanded populations of eGFPp–I-Ab–specific cells in immu-nized mice for this analysis. We found no substantial difference in the mean fluorescence intensity of eGFPp–I-Ab tetramer staining on cells from wild-type and Ins2eGFP mice immunized with eGFPp in CFA (data not shown). Because CFA forms a long-lasting depot at the injec-tion site27, persistent presentation of the eGFPp–I-Ab epitope might have given low-affinity T cells an advantage, which would diminish our ability to detect a subtle difference. We therefore repeated the analysis using an eGFP-expressing strain of attenuated (∆actA∆inlB) Listeria monocytogenes (Lm-eGFP) that causes a very transient infec-tion28. Lm-eGFP-induced clonal proliferation of eGFPp–I-Ab–specific T cells was significantly impaired in Ins2eGFP mice compared with that in wild-type mice (Fig. 4a,b). Furthermore, the mean fluorescence intensity of eGFPp–I-Ab tetramer staining on cells from Ins2eGFP mice was lower than that of cells from wild-type mice (Fig. 4c). These results indicated that some clones with TCRs with high affinity for the eGFPp–I-Ab epitope were deleted in Ins2eGFP mice.

The fact that eGFP was probably expressed in the thymuses of Ins2eGFP mice under the control of the transcriptional regulator Aire led us to investigate whether tolerance occurred in this organ in an Aire-dependent fashion. We first addressed this question by trans-planting wild-type or Ins2eGFP thymuses into athymic B6-nude mice29. Then, 8–12 weeks later, we primed mice subcutaneously with eGFPp in CFA. B6-nude recipients of wild-type thymus grafts generated large numbers of CD44hi eGFPp–I-Ab–specific CD4+ T cells, as seen in intact wild-type mice, whereas B6-nude mice that received Ins2eGFP thymus grafts produced a much smaller population (Fig. 5a). B6-nude recipients of Ins2eGFP thymus grafts produced larger numbers and greater frequencies of CD44hi eGFPp–I-Ab–specific Treg cells than did the recipients of wild-type thymus grafts (Fig. 5b,c). Foxp3− effector cell differentiation was also impaired in recipients of Ins2eGFP thymus grafts (Fig. 5d). Thus, much of the tolerance observed in Ins2eGFP mice was induced in the thymus.

We assessed the involvement of Aire through the use of Ins2eGFPAire−/− mice. These mice had numbers of eGFPp–I-Ab–specific CD4+CD8− thymocytes comparable to those of wild-type and Aire−/− mice (Fig. 5e), which demonstrated that clonal deletion in Ins2eGFP mice was Aire dependent. Furthermore, thymic Treg cell numbers in Ins2eGFPAire−/− mice were similar to those in wild-type mice (data not shown), and the clonal proliferation and differentiation of Foxp3− effector cells was very similar in Ins2eGFPAire−/− mice and Aire−/− mice following immunization (Fig. 5f–h). Therefore, much of the tolerance in Ins2eGFP mice was induced in the thymus by an Aire-dependent mechanism. The total number of Foxp3+ Treg cells, however, was

CD

44hi

Tet

+

CD

4+ T

cel

ls

100

101

102

103

104 **b

WT Ins2eGFP

Tet

MF

I

0

0.5

1.0

1.5 *c

WT Ins2eGFP

0.12

Ins2eGFP

0 102 103 104 105

0102

103

104

1051.06

WTa

Tet (APC)

Tet

(P

E)

Figure 4 High-affinity T cells are deleted in Ins2eGFP mice. (a) Flow cytometry analyzing Tet-PE versus Tet-APC staining (as in Fig. 2a) of tetramer-enriched CD4+ T cells from pooled spleens and lymph nodes of wild-type or Ins2eGFP mice 6–7 d after infection with Lm-eGFP. Numbers in top right quadrants indicate percent double-Tet+ cells. (b) Quantification of CD44hi Tet+ CD4+ T cells from pooled spleens and lymph nodes of mice 6–7 d after infection with Lm-eGFP. (c) Mean fluorescence intensity (MFI) of tetramers in the cells in b; results are normalized to those of wild-type controls. Each symbol (b,c) represents an individual mouse; small horizontal lines indicate the geometric mean (b) or the mean (c). *P < 0.005 and **P < 0.0001 (unpaired t-test of log10-transformed (b) or linear (c) values). Data are representative of three experiments (a) or are pooled from three independent experiments with 7–11 mice per group (b,c).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 17 NUMBER 2 FEBRUARY 2016 191

A rt i c l e s

still elevated in some immunized Ins2eGFPAire−/− mice compared with that in Aire−/− mice (Fig. 5g), which suggested that a non- Aire-dependent mechanism might have been involved in this aspect of tolerance. Overall, these studies indicated that tolerance in CD4+ T cell repertoires specific for self epitopes expressed in a limited fashion in the thymus was associated with intrathymic deletion of some clones and impaired generation of effector T cells, but enhanced generation of Treg cells, from the remaining clones.

Clonal deletion regulates self antigens in cluster 3Cluster 3 was characterized by a profound clonal proliferation defect following immunization (Fig. 1). This cluster included mice expressing eGFP under the control of promoters not regulated by Aire in MHCII+

cells in the thymus. UBCeGFP and mice with transgenic expression of eGFP from the gene encoding β-actin (ACTBeGFP) express eGFP in thymic dendritic cells, medullary thymic epithelial cells, and all other nucleated cells30,31. Mice with transgenic expression of eYFP from the gene encoding the integrin CD11c (ItgaxeYFP) express eYFP in thymic dendritic cells32, whereas mice with transgenic expression of eGFP or CFP from the gene encoding choline acetyltransferase (ChateGFP)33,34, cadherin 1 (Cdh1CFP)35 or Aire (AireeGFP)36 express eGFP in medullary thymic epithelial cells. We chose UBCeGFP mice for further analysis as representatives of cluster 3.

We quantified eGFPp–I-Ab–specific CD4+CD8− thymocytes to determine whether intrathymic clonal deletion was responsible for the proliferation defect observed in UBCeGFP mice. Wild-type mice

e

NS NS *102

101

100

Dou

ble-

Tet

+ C

D4+

CD

8– thym

ocyt

es

Aire–/–Ins2eGFPAire−/−WT

Ins2eGFP

105

104

Foxp3+ Foxp3–

b

**

NS

CD

44hi

Tet

+

CD

4+ T

cel

ls

103

102

101

Donorthymus

WT

Ins2eGFP

Donorthymus

WT

Ins2eGFP

0

20

40

60

80

100

c

Fox

p3+ c

ells

(%

)

*

d

Other

T H17T H

1T FH

**** **

CD

44hi

Tet

+

CD

4+ T

cel

ls

NS104

103

102

101

Donorthymus

WT

Ins2eGFPa

105

104

103

102

**

CD

44hi

Tet

+

CD

4+ T

cel

lsDonorthymus

WT

Ins2eGFP

h Aire–/–

Ins2eGFPAire−/−

Other

T H17T H

1T FH

NSNSNS NS

CD

44hi

Tet

+

CD

4+ T

cel

ls

104

103

102

101

f

NS NS ***

103

104C

D44

hi T

et+

CD

4+ T

cel

ls

105

Aire–/–

Ins2eGFPAire−/−WT

Ins2eGFP

Foxp3

+

Foxp3

–

g

*NS

CD

44hi

Tet

+

CD

4+ T

cel

ls

105

104

103

102

101

100

Aire–/–

Ins2eGFPAire−/−

UBCeGFPWTa

0 102 103 104 105

0102

103

104

105 0.1 0.01

Tet (APC)

Tet

(P

E)

0 102 103 104 105

0

102

103

104

105

WT UBCeGFPc0.060.20

Tet (APC)

Tet

(P

E)

b

100

101

102

Dou

ble-

Tet

+

CD

4+C

D8–

thym

ocyt

es

WT

UBCeG

FP

*

100

101

102

Dou

ble-

Tet

+

CD

4+ T

cel

ls

WT

UBCeG

FP

*d

Figure 6 Expression of a self epitope by thymic antigen-presenting cells induces intrathymic clonal deletion. (a) Flow cytometry analyzing Tet-PE versus Tet-APC staining (as in Fig. 2a) of tetramer-enriched CD4+CD8− thymocytes from wild-type or UBCeGFP mice. (b) Quantification of CD4+CD8− double-Tet+ cells in the thymuses of wild-type or UBCeGFP mice. (c) Flow cytometry analyzing Tet-PE versus Tet-APC staining (as in Fig. 2a) of tetramer-enriched CD4+ T cells from pooled spleens and lymph nodes of wild-type or UBCeGFP mice. (d) Quantification of CD4+ double-Tet+ cells in pooled spleens and lymph nodes of wild-type or UBCeGFP mice (results for wild-type mice replotted from Fig. 2b). Numbers in top right quadrants (a,c) indicate percent double-Tet+ cells. Each symbol (b,d) represents an individual mouse; small horizontal lines indicate the geometric mean. *P < 0.0001 (unpaired t-test of log10-transformed values). Data are representative of two experiments (a) or four experiments (c) or are pooled from two independent experiments with five mice per group (b) or four independent experiments with 10–17 mice per group (d).

Figure 5 Aire-mediated thymic expression of eGFP promotes clonal deletion and Treg induction in Ins2eGFP mice. (a) Quantification of CD44hi Tet+ CD4+ T cells in pooled spleens and lymph nodes of B6-nude mice given wild-type or Ins2eGFP thymus grafts 8–12 weeks before immunization with 100 µg of eGFPp in CFA, analyzed 14 d after immunization. (b) Quantification of CD44hi Tet+ CD4+ Foxp3+ Treg cells and Foxp3− effector T cells from mice in a. (c) Frequency of Foxp3+ T cells among cells identified in a. (d) Quantification of each Foxp3− T effector cell subset (horizontal axis) in populations in a. (e) Quantification of double-Tet+ CD4+CD8− thymocytes (double-tetramer stained and analyzed as in Fig. 2a) in pooled thymuses from wild-type, Aire−/−, Ins2eGFPAire−/− and Ins2eGFP mice (results for wild-type and Ins2eGFP mice replotted from Fig. 3b). (f) Quantification of CD44hi Tet+ CD4+ T cells in pooled spleens and lymph nodes of wild-type, Aire−/−, Ins2eGFPAire−/− and Ins2eGFP mice 14 d after immunization with eGFPp in CFA (results for wild-type and Ins2eGFP mice replotted from Fig. 1a). (g) Quantification of CD44hi Tet+ Foxp3+ Treg cells and Foxp3− effector T cells in the mice in f. (h) Quantification of each Foxp3− T effector cell subset (horizontal axis) among cells quantified in f. Each symbol (a,e,f) represents an individual mouse; small horizontal lines indicate the geometric mean. *P < 0.05, **P < 0.005 and ***P < 0.0001 (unpaired t-test of log10-transformed (a,b,d,g,h) or linear (c) values or one-way ANOVA of log10-transformed values (e,f)). Data are from two independent experiments with 5–11 mice per group (a–d; error bars (b–d), s.d.), six independent experiments with 12–21 mice per group (e) or five independent experiments with 6–23 mice per group (f–h; error bars (g,h), s.d.).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

192  VOLUME 17 NUMBER 2 FEBRUARY 2016 nature immunology

A rt i c l e s

had an average of 20 eGFPp–I-Ab–specific CD4+CD8− thymocytes, while UBCeGFP mice had fewer than one (Fig. 6a,b). However, we detected a few eGFPp–I-Ab–specific CD4+ T cells in the secondary lymphoid organs of UBCeGFP mice (Fig. 6c,d). These results sug-gested that most eGFPp–I-Ab–specific CD4+CD8− thymocytes were deleted in the thymus, although a few escaped and accumulated in the secondary lymphoid organs. The profound lack of proliferation following immunization indicated that these few cells were unresponsive. These results indicated that MHC class II–bound peptides from proteins with abundant expression in MHCII+ cells in the thymus induced a profound state of immunotolerance mediated in large part by intrathymic clonal deletion.

Thymic self-antigen expression correlates with toleranceThe specific mechanisms of tolerance associated with self epitopes in clusters 1–3 appeared to depend on self-antigen expression by MHCII+ cells in the thymus. To test this hypothesis, we assessed expression of eGFP and eYFP by EpCAM+ thymic epithelial cells and thymic dendritic cells from representatives of these clusters (Supplementary Fig. 1). Thymuses from wild-type and Ins1eGFP mice (in cluster 1) had fewer than 10 I-Ab+ eGFP+ or eYFP+ EpCAM+ epithelial cells or dendritic cells (Fig. 7a–c), consistent with the ignorant phenotype of eGFPp–I-Ab–specific CD4+ T cells in these mice. At the other end of the spectrum, ItgaxeYFP and ACTBeGFP mice (in cluster 3) had more than 300,000 I-Ab+ eYFP+ or eGFP+ dendritic cells and EpCAM+ epi-thelial cells in their thymuses (Fig. 7a–c), which probably accounted for the substantial clonal deletion observed in these mice. Finally, Ins2eGFP, Pcp2eGFP and CD207eYFP thymuses contained several hundred I-Ab+ eGFP+ or eYFP+ EpCAM+ epithelial cells, and CD207eYFP thy-muses also contained ~3,000 I-Ab+ eYFP+ dendritic cells (Fig. 7a–c). We observed a significant inverse correlation between the number of eGFP+ or eYFP+ thymic antigen-presenting cells and the number of eGFPp–I-Ab–specific CD4+ T cells present 14 d after immunization with eGFPp in CFA (Fig. 7d). These results were consistent with the hypothesis that the number of MHCII+ thymic antigen-presenting cells expressing a particular self antigen determines the responsive-ness of the corresponding CD4+ T cell population.

Tolerance to true self antigens occurs by these mechanismsWe then used information from the fluorescent protein system to identify the tolerance mechanisms for CD4+ T cell repertoires for six genuine self peptide–I-Ab epitopes. We immunized B6 mice with the

self peptides, or with foreign peptides as a control, and determined the compositions of the resulting Foxp3− effector T cell populations and Foxp3+ Treg cell populations. We analyzed the frequency of TFH cells, TH1 cells, TH17 cells and Treg cells with an unsupervised hier-archical clustering algorithm to determine if the responses to the self peptides aligned with the clusters defined in the fluorescent protein system. We used the frequency of cells in each subset, rather than absolute numbers, for this comparison because the sizes of the pre-immunization repertoires for the natural epitopes were unknown. The use of absolute numbers of cells in each subset would have been problematic for epitopes with populations that had the same differ-entiation potential but simply varied in size.

Peptides from the constant region of immunoglobulin M, which is very abundant in the blood, or from the ATP-binding cassette sub-family G, member 1, which has high expression in thymic dendritic cells and medullary thymic epithelial cells24, elicited fewer than 200 CD44hi cells that bound the relevant peptide–I-Ab tetramers (Fig. 8a). These peptides elicited too few CD44hi cells to accurately deter-mine the frequency of each subset and thus were not analyzed by the clustering algorithm. The minimal responses to these self epitopes, however, indicated that these CD4+ T cell repertoires underwent extensive intrathymic clonal deletion, as in the mice with transgenic eGFP, eYFP or CFP expression in cluster 3.

The unsupervised hierarchical clustering analysis produced two clusters. The first (‘foreign’) cluster included responses to foreign peptides from listeriolysin O protein of L. monocytogenes37,38 and

d

0 2 4 60

1

2

3

4

5

R2 = 0.87P = 0.0022

Tet

+ C

D4+

T c

ells

(lo

g 10)

eGFP+ or eYFP+ thymic APCs (log10)

WTIns1eGFP

Ins2eGFP

CD207eYFP

ItgaxeYFP

Pcp2eGFP

ACTBeGFP

ItgaxeY

FP

CD207eY

FP

Ins2eG

FP

Ins1eG

FPW

T100

101

102

103

104

105

eGF

P+ o

r eY

FP

+

EpC

AM

+ e

pith

elia

l cel

ls

b

Pcp2eG

FP

******

******

ACTBeG

FP

Dendritic cellsEpCAM+

epithelial cellsa

I-A

b

Cluster 1:WT

eGFP or eYFP

105

104

103

102

102 103 104 105

0

0

Cluster 1:Ins1eGFP

Cluster 2:Ins2eGFP

Cluster 3:ItgaxeYFP

Cluster 2:CD207eYFP

Cluster 3:ACTBeGFP

Cluster 2:Pcp2eGFP

0.00030.003

9.08

0.0010.015

0.0010.12

0.360.71

0.0071.73

90.9

100

101

102

103

104

105

106

eGF

P+ o

r eY

FP

+

dend

ritic

cel

ls

c

***

***

*

***

ItgaxeY

FP

CD207eY

FP

Ins2eG

FP

Ins1eG

FPW

T

Pcp2eG

FP

ACTBeG

FP

80.4

87.1

NS

NSNS

Figure 7 Frequency of self antigen–positive thymic antigen-presenting cells correlates with the responsiveness of the corresponding CD4+ T cell population. (a) Flow cytometry analyzing the expression of MHC class II (I-Ab) and eGFP or eYFP by EpCAM+ thymic epithelial cells (left) and thymic dendritic cells (right) from various members of clusters 1–3 (left margin). Numbers below outlined areas indicate percent I-Ab+eGFP+ or I-Ab+eYFP+ cells. (b,c) Quantification of eGFP+ or eYFP+ EpCAM+ thymic epithelial cells (b) or eGFP+ or eYFP+ thymic dendritic cells (c) in various mouse strains (horizontal axes; colors match those of clusters in Fig. 1d), identified as in a. *P < 0.05, **P < 0.0005 and ***P < 0.0001 (one-way ANOVA of log10-transformed values). (d) Linear regression analysis of the number of eGFP+ or eYFP+ thymic antigen-presenting cells versus the number of Tet+ CD4+ T cells in various mouse strains 14 d after immunization with eGFPp or eYFPp in CFA (mouse strains in plot: each symbol represents an individual strain; colors match those of clusters in Fig. 1d); results presented as log10(geometric mean from Fig. 1a) versus log10(geometric mean of total I-Ab+ eGFP+ or eYFP+ EpCAM+ thymic epithelial cells and dendritic cells). P value, F-test. Data are from five experiments with three to six mice per group (a–c), (b,c; error bars, s.d.) or 17 experiments (d).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 17 NUMBER 2 FEBRUARY 2016 193

A rt i c l e s

Figure 8 Tolerance mechanisms identified in eGFP-expressing mouse strains govern tolerance to true self antigens. (a) Quantification of CD44hi peptide–I-Ab+ (Tet+) CD4+ T cells from pooled spleens and lymph nodes of mice 13–14 d after immunization with 100 µg peptide (horizontal axis) emulsified in CFA: IgMH, constant region of immunoglobulin M; ABCG1, ATP-binding cassette sub-family G, member 1; PLP, proteolipid protein; PDIA2, protein disulfide–associated isomerase 2; JPH2, junctophilin 2; MOG, myelin oligodendrocyte glycoprotein; CLGN(BD), calnexin from B. dermatitidis; LLO, listeriolysin O protein of L. monocytogenes; 2W, model peptide. Each symbol represents an individual mouse (colors match those of clusters in b); small horizontal lines indicate the geometric mean. (b) Unsupervised hierarchical clustering of 16 self antigen–I-Ab+ and foreign antigen–I-Ab+ (Tet+) CD4+ T cell populations (from a and Fig. 1a), clustered by the frequency of cells in each subset; far right, two main clusters identified. Data are pooled from 17 experiments with 4–23 mice per epitope (a,b).

calnexin from Blastomyces dermatitidis, and the model peptide 2W39, as well as the cluster 1 responses to eGFPp (Fig. 1d) in wild-type, Ins1eGFP and Foxd1eGFP mice (Fig. 8b). The second (‘self ’) cluster consisted of responses to the eGFP epitope in the mice with transgenic eGFP or eYFP expression in cluster 2 (Fig. 1d).

A peptide from junctophilin 2, which is expressed exclusively in skel-etal muscle and the heart40, elicited 5,000 CD44hi cells that bound the relevant peptide–I-Ab tetramer (Fig. 8a). This population produced a composition of Treg cells and effector T cells similar to that of populations specific for foreign peptides, as indicated by its assignment to the ‘foreign’ cluster (Fig. 8b). These results indicated that the CD4+ T cell tolerance to the junctophilin 2 epitope in B6 mice was maintained by ignorance.

Peptides from myelin oligodendrocyte glycoprotein and from pro-teolipid protein, which are expressed in the nervous system, and from protein disulfide–associated isomerase 2, which is expressed in the pancreas and digestive tract41, elicited 1,000–10,000 peptide–I-Ab– specific CD44hi cells (Fig. 8a). These populations contained a lower frequency of Foxp3− effector T cells and a greater frequency of Treg cells than those of populations specific for non-mouse peptides (Fig. 8b). The responses to these self epitopes were placed in the ‘self ’ cluster, which also included responses to the eGFP epitope in the mice with transgenic eGFP or eYFP expression in cluster 2 (Fig. 1d). Thus, tolerance to these endogenously expressed peptides resembled that associated with limited thymic eGFP, eYFP or CFP expression.

DISCUSSIONHere we demonstrated that tolerance in the normally diverse CD4+ T cell repertoire can be attributed to one of three mechanisms, depend-ing on the expression pattern of the self epitope. A profound state of tolerance existed in CD4+ T cell repertoires specific for epitopes from proteins that were easily accessed or were expressed by many MHCII+ cells in the thymus. This tolerance manifested as nearly complete deletion of self epitope–specific T cells in the thymus, which probably resulted from abundant display of self epitope on MHCII+ thymic dendritic cells and/or medullary thymic epithelial cells. Although substantial clonal deletion has been reported for human and mouse T cells specific for self epitopes from the Y chromosome or driven by the ACTB promoter4,5,15, it was not as extensive as that observed here for an epitope expressed from the UBC promoter. The UBC promoter might have led to an extraordinarily large number of self epitopes on thymic antigen-presenting cells, which would cause deletion of nearly

all self epitope–specific T cells. Alternatively, deletion within the small eGFPp–I-Ab–specific CD4+ T cell repertoire might have been more efficient because of limited inter-clonal competition.

A state of immunological ignorance existed in CD4+ T cell rep-ertoires specific for epitopes from proteins encoded by genes not regulated by Aire that were expressed exclusively outside the thymus. In addition, these epitopes were probably not displayed in second-ary lymphoid organs because their parent proteins are cytosolic and are expressed by cells that do not turn over in a way that allows local dendritic cells to produce the epitopes or access draining lymph nodes. The complete lack of presentation of these self epitopes would dictate that the corresponding CD4+ T cells exist in a naive and fully responsive state of ignorance. Self epitopes in this cluster were essentially foreign to the relevant T cell repertoires.

A complex form of tolerance controlled CD4+ T cell repertoires specific for epitopes from proteins with tissue-restricted patterns of expression but that were also expressed by small numbers of MHCII+ thymic antigen-presenting cells. Analysis of the Ins2eGFP member of this group revealed that tolerance was associated with intrathymic deletion of ~30% of the epitope-specific clones. Partial clonal dele-tion has also been observed in studies of a polyclonal CD8+ T cell population specific for a tissue-restricted epitope11. In our studies, the deleted clones had TCRs with the highest affinities for the eGFPp–I-Ab ligand. Thymocytes with high-affinity TCRs might be especially sen-sitive to deletion because they can garner sufficient signal from sin-gle encounters with rare thymic antigen-presenting cells. In contrast, deletion of thymocytes with TCRs of lower affinity might require numerous such encounters42,43, which are unlikely to occur when relevant thymic antigen-presenting cells are rare, as in cluster 2.

The partial deletion of the epitope-specific polyclonal population studied in Ins2eGFP mice depended on Aire, indicative of the involve-ment of medullary thymic epithelial cells. Furthermore, we found that expression of eGFP or eYFP from the Ins2, Pcp2 or CD207 locus occurred in only a small number of thymic epithelial cells and dendritic cells, which probably allowed eGFPp–I-Ab–specific T cells with TCRs of lower affinity to escape deletion. In agreement with that pos-sibility, eGFPp–I-Ab–specific T cells were nearly completely deleted in AireeGFP mice, which express eGFP in all medullary thymic epithe-lial cells36. Although medullary thymic epithelial cells were a likely source of self peptide in these models, antigen transfer to neighboring dendritic cells might have mediated deletion44.

a

CD

44hi

Tet

+ C

D4+

T c

ells

IgM

Hp

ABCG1pPLP

p

PDIA2p

JPH2p

MOGp

LLOp

2W101

102

103

104

105

106

CD44hi Tet+ (%)

Foreign(cluster 1)

Self(cluster 2)

b

1.3

3.4

5.5

7.7

9.811

.914

.116

.224

.031

.739

.547

.255

.062

.8

WT (eGFPp)Ins1eGFP (eGFPp)WT (CLGN(BD)p)WT (2W)WT (JPH2p)

Zfp423eGFP (eGFPp)WT (LLOp)Foxd1eGFP (eGFPp)WT (MOGp)Ins2eGFP (eGFPp)CD207eYFP (eGFPp)Myh11eGFP

(eGFPp)WT (PLPp)

Foxp3eGFP

(eGFPp)Pcp2eGFP (eGFPp)WT (PDIA2p)

T reg

T FH T H1

T H17

eGFPp

CLGN(B

D)p

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

194  VOLUME 17 NUMBER 2 FEBRUARY 2016 nature immunology

A rt i c l e s

The non-deleted T cells specific for cluster 2 epitopes generated Foxp3− effector T cells less well than did the corresponding wild-type T cells. This defect extended to TFH cells but was particularly striking for TH1 cells and TH17 cells. Published work has shown that TH1 clones tend to have TCRs of higher affinity than the affinity of the TCRs on TFH cells45. Thus, it is possible that the TCRs of lower affinity found on epitope-specific T cells in this cluster were incapable of generating enough signals to form TH1 cells or TH17 cells. Alternatively, these cells might have been induced into a partially anergic state due to persist-ent recognition of self epitope in a non-inflammatory environment, as suggested by published studies of mice with transgenic TCR expres-sion46–48. If so, the thymic transplant experiments presented here would suggest that this anergy induction occurs mainly in the thymus.

Enhanced proliferation of Treg cells after immunization was another component of tolerance to the epitopes in cluster 2. One explanation for this feature was that the pre-immunization repertoires of T cells specific for those epitopes showed enrichment for thymus-derived Treg cells, which might occur when cells are shunted into the Treg cell lineage following the recognition of self epitopes with moderately high affinity in the thymus49. The observation that Treg cells were detected more frequently in the pre-immunization eGFPp–I-Ab–specific popu-lation in Ins2eGFP mice than in wild-type mice is consistent with this possibility. This is only a tentative conclusion, however, because the eGFPp–I-Ab–specific populations were very small in both situations. It is also possible that Foxp3− T cells in the eGFPp–I-Ab–specific population had enhanced potential to differentiate into induced Treg cells after immunization. In any case, our results suggest that Treg cells are especially important for tolerance to epitopes from tissue- restricted antigens that also exhibit limited thymic expression.

The mechanisms of tolerance identified for the epitopes studied here all appeared to be induced in the thymus. It is possible that other epitopes are tolerated by mechanisms that operate in the secondary lymphoid organs, such as epitopes from paternal fetal proteins that appear in pregnant females. Peripheral deletion and anergy seem to be the tolerance mechanisms for these antigens50.

METHODSMethods and any associated references are available in the online version of the paper.

Note: Any Supplementary Information and Source Data files are available in the online version of the paper.

AcKNOWLEDgMENTSWe thank D. Mueller for reviewing the manuscript; J. Walter and O. Rainwater for technical assistance; P. Lauer (and Aduro Biotech) for the Lm-eGFP strain PL1113. Supported by the US National Institutes of Health (P01 AI35296 to M.K.J. and K.A.H.; F32 AI114050 to D.M.; T32 GM008244 and F30 DK093242 to R.W.N.; T32 AI07313 to J.L.L.; K99 AI114884 to Y.J.L.), the Wallin Neuroscience Discovery Fund (to H.T.O. and M.K.J.) and the Juvenile Diabetes Research Foundation (2-2011-662 to B.T.F.).

AUTHOR cONTRIBUTIONSD.M. designed the study, did experiments, analyzed data, and wrote the manuscript, J.L.L., T.D., Y.J.L., W.E.P., J.V.L., R.W.N. and M.S.A. did experiments, B.T.F., H.T.O. and K.A.H. provided discussions; and M.K.J. designed the study, analyzed data and wrote the manuscript.

COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.

reprints and permissions information is available online at http://www.nature.com/reprints/index.html.

1. Xing, Y. & Hogquist, K.A. T-cell tolerance: central and peripheral. Cold Spring Harb. Perspect. Biol. 4, a006957 (2012).

2. Wing, K. & Sakaguchi, S. Regulatory T cells exert checks and balances on self tolerance and autoimmunity. Nat. Immunol. 11, 7–13 (2010).

3. Semana, G., Gausling, R., Jackson, R.A. & Hafler, D.A. T cell autoreactivity to proinsulin epitopes in diabetic patients and healthy subjects. J. Autoimmun. 12, 259–267 (1999).

4. Moon, J.J. et al. Quantitative impact of thymic selection on Foxp3+ and Foxp3− subsets of self-peptide/MHC class II-specific CD4+ T cells. Proc. Natl. Acad. Sci. USA 108, 14602–14607 (2011).

5. Yu, W. et al. Clonal deletion prunes but does not eliminate self-specific αβ CD8+ T lymphocytes. Immunity 42, 929–941 (2015).

6. Maeda, Y. et al. Detection of self-reactive CD8+ T cells with an anergic phenotype in healthy individuals. Science 346, 1536–1540 (2014).

7. Ohashi, P.S. et al. Ablation of “tolerance” and induction of diabetes by virus infection in viral antigen transgenic mice. Cell 65, 305–317 (1991).

8. Oldstone, M.B., Nerenberg, M., Southern, P., Price, J. & Lewicki, H. Virus infection triggers insulin-dependent diabetes mellitus in a transgenic model: role of anti-self (virus) immune response. Cell 65, 319–331 (1991).

9. Jenkins, M.K. & Schwartz, R.H. Antigen presentation by chemically modified splenocytes induces antigen-specific T cell unresponsiveness in vitro and in vivo. J. Exp. Med. 165, 302–319 (1987).

10. Akkaraju, S. et al. A range of CD4 T cell tolerance: partial inactivation to organ-specific antigen allows nondestructive thyroiditis or insulitis. Immunity 7, 255–271 (1997).

11. Zehn, D. & Bevan, M.J. T cells with low avidity for a tissue-restricted antigen routinely evade central and peripheral tolerance and cause autoimmunity. Immunity 25, 261–270 (2006).

12. Hataye, J., Moon, J.J., Khoruts, A., Reilly, C. & Jenkins, M.K. Naive and memory CD4+ T cell survival controlled by clonal abundance. Science 312, 114–116 (2006).

13. Marzo, A.L. et al. Initial T cell frequency dictates memory CD8+ T cell lineage commitment. Nat. Immunol. 6, 793–799 (2005).

14. Bautista, J.L. et al. Intraclonal competition limits the fate determination of regulatory T cells in the thymus. Nat. Immunol. 10, 610–617 (2009).

15. Bouneaud, C., Kourilsky, P. & Bousso, P. Impact of negative selection on the T cell repertoire reactive to a self-peptide: a large fraction of T cell clones escapes clonal deletion. Immunity 13, 829–840 (2000).

16. Nelson, R.W. et al. T cell receptor cross-reactivity between similar foreign and self peptides influences naive cell population size and autoimmunity. Immunity 42, 95–107 (2015).

17. Reich, M. et al. GenePattern 2.0. Nat. Genet. 38, 500–501 (2006).18. Hara, M. et al. Transgenic mice with green fluorescent protein-labeled pancreatic

beta-cells. Am. J. Physiol. Endocrinol. Metab. 284, E177–E183 (2003).19. Anderson, M.S. et al. Projection of an immunological self shadow within the thymus

by the aire protein. Science 298, 1395–1401 (2002).20. Carreres, M.I. et al. Transcription factor Foxd1 is required for the specification of

the temporal retina in mammals. J. Neurosci. 31, 5673–5681 (2011).21. Kobayashi, A. et al. Identification of a multipotent self-renewing stromal progenitor

population during mammalian kidney organogenesis. Stem Cell Rep. 3, 650–662 (2014).

22. Ben-Yehudah, A. et al. Specific dynamic and noninvasive labeling of pancreatic beta cells in reporter mice. Genesis 43, 166–174 (2005).

23. Kaplan, D.H. et al. Autocrine/paracrine TGFβ1 is required for the development of epidermal Langerhans cells. J. Exp. Med. 204, 2545–2552 (2007).

24. Heng, T.S., Painter, M.W. & Immunological Genome Project Consortium. The Immunological Genome Project: networks of gene expression in immune cells. Nat. Immunol. 9, 1091–1094 (2008).

25. Gong, S. et al. A gene expression atlas of the central nervous system based on bacterial artificial chromosomes. Nature 425, 917–925 (2003).

26. Crawford, F., Kozono, H., White, J., Marrack, P. & Kappler, J. Detection of antigen-specific T cells with multivalent soluble class II MHC covalent peptide complexes. Immunity 8, 675–682 (1998).

27. Warren, H.S., Vogel, F.R. & Chedid, L.A. Current status of immunological adjuvants. Annu. Rev. Immunol. 4, 369–388 (1986).

28. Portnoy, D.A., Auerbuch, V. & Glomski, I.J. The cell biology of Listeria monocytogenes infection: the intersection of bacterial pathogenesis and cell-mediated immunity. J. Cell Biol. 158, 409–414 (2002).

29. Cordier, A.C. & Haumont, S.M. Development of thymus, parathyroids, and ultimo-branchial bodies in NMRI and nude mice. Am. J. Anat. 157, 227–263 (1980).

30. Schaefer, B.C., Schaefer, M.L., Kappler, J.W., Marrack, P. & Kedl, R.M. Observation of antigen-dependent CD8+ T-cell/ dendritic cell interactions in vivo. Cell. Immunol. 214, 110–122 (2001).

31. Okabe, M., Ikawa, M., Kominami, K., Nakanishi, T. & Nishimune, Y. ‘Green mice’ as a source of ubiquitous green cells. FEBS Lett. 407, 313–319 (1997).

32. Lindquist, R.L. et al. Visualizing dendritic cell networks in vivo. Nat. Immunol. 5, 1243–1250 (2004).

33. Panneck, A.R. et al. Cholinergic epithelial cell with chemosensory traits in murine thymic medulla. Cell Tissue Res. 358, 737–748 (2014).

34. Tallini, Y.N. et al. BAC transgenic mice express enhanced green fluorescent protein in central and peripheral cholinergic neurons. Physiol. Genomics 27, 391–397 (2006).

35. Snippert, H.J. et al. Intestinal crypt homeostasis results from neutral competition between symmetrically dividing Lgr5 stem cells. Cell 143, 134–144 (2010).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 17 NUMBER 2 FEBRUARY 2016 195

A rt i c l e s

36. Gardner, J.M. et al. Deletional tolerance mediated by extrathymic Aire-expressing cells. Science 321, 843–847 (2008).

37. Pepper, M., Pagan, A.J., Igyarto, B.Z., Taylor, J.J. & Jenkins, M.K. Opposing signals from the Bcl6 transcription factor and the interleukin-2 receptor generate T helper 1 central and effector memory cells. Immunity 35, 583–595 (2011).

38. Geginat, G., Schenk, S., Skoberne, M., Goebel, W. & Hof, H. A novel approach of direct ex vivo epitope mapping identifies dominant and subdominant CD4 and CD8 T cell epitopes from Listeria monocytogenes. J. Immunol. 166, 1877–1884 (2001).

39. Rees, W. et al. An inverse relationship between T cell receptor affinity and antigen dose during CD4+ T cell responses in vivo and in vitro. Proc. Natl. Acad. Sci. USA 96, 9781–9786 (1999).

40. Takeshima, H., Komazaki, S., Nishi, M., Iino, M. & Kangawa, K. Junctophilins: a novel family of junctional membrane complex proteins. Mol. Cell 6, 11–22 (2000).

41. Fu, X.M., Dai, X., Ding, J. & Zhu, B.T. Pancreas-specific protein disulfide isomerase has a cell type-specific expression in various mouse tissues and is absent in human pancreatic adenocarcinoma cells: implications for its functions. J. Mol. Histol. 40, 189–199 (2009).

42. Rosette, C. et al. The impact of duration versus extent of TCR occupancy on T cell activation: a revision of the kinetic proofreading model. Immunity 15, 59–70 (2001).

43. Au-Yeung, B.B. et al. A sharp T-cell antigen receptor signaling threshold for T-cell proliferation. Proc. Natl. Acad. Sci. USA 111, E3679–E3688 (2014).

44. Klein, L., Kyewski, B., Allen, P.M. & Hogquist, K.A. Positive and negative selection of the T cell repertoire: what thymocytes see (and don’t see). Nat. Rev. Immunol. 14, 377–391 (2014).

45. Tubo, N.J. et al. Single naive CD4+ T cells from a diverse repertoire produce different effector cell types during infection. Cell 153, 785–796 (2013).

46. Apostolou, I., Sarukhan, A., Klein, L. & von Boehmer, H. Origin of regulatory T cells with known specificity for antigen. Nat. Immunol. 3, 756–763 (2002).

47. Tanchot, C., Barber, D.L., Chiodetti, L. & Schwartz, R.H. Adaptive tolerance of CD4+ T cells in vivo: multiple thresholds in response to a constant level of antigen presentation. J. Immunol. 167, 2030–2039 (2001).

48. Kearney, E.R., Pape, K.A., Loh, D.Y. & Jenkins, M.K. Visualization of peptide-specific T cell immunity and peripheral tolerance induction in vivo. Immunity 1, 327–339 (1994).

49. Lee, H.M., Bautista, J.L., Scott-Browne, J., Mohan, J.F. & Hsieh, C.S. A broad range of self-reactivity drives thymic regulatory T cell selection to limit responses to self. Immunity 37, 475–486 (2012).

50. PrabhuDas, M. et al. Immune mechanisms at the maternal-fetal interface: perspectives and challenges. Nat. Immunol. 16, 328–334 (2015).

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology doi:10.1038/ni.3327

ONLINE METHODSMice. 4- to 8-week old C57BL/6 (B6) mice, B6.Cg-Foxn1nu/J (B6-nude) mice, B6.PL-Thy1a/CyJ (CD90.1+) mice, C57BL/6-Tg(UBC-GFP)30Scha/ J (UBCeGFP) mice, B6.Cg-Tg(Ins1-EGFP)1Hara/J (Ins1eGFP) mice, B6.Cg-Tg (RP23-268L19-EGFP)2Mik/J (ChateGFP) mice, B6;FVB-Tg(Zfp423-EGFP)7Brsp/J (Zfp423eGFP) mice51, B6.Cg-Tg(Myh11-cre,-EGFP)2Mik/ J (Myh11eGFP) mice52, B6.129P2(Cg)-Cdh1tm1Cle/J (Cdh1CFP) mice, B6;129S4-Foxd1tm1(GFP/cre)Amc/J (Foxd1eGFP) mice and C57BL/6-Tg(CAG-EGFP)131Osb/LeySopJ (ACTBeGFP) mice were purchased from the Jackson Laboratory or the National Cancer Institute Mouse Repository. Mice were used at 6–12 weeks of age. Ins2eGFP B6 mice, Ins2eGFP × CD90.1+ mice, Ins1eGFP × B6 mice, Ins2eGFPAire+/− × Aire+/− mice, stock Tg(Pcp2-EGFP)BT153Gsat/Mmmh mice (backcrossed over eight generations onto the B6 background) (Pcp2eGFP mice), B6.Cg-Tg(Itgax-Venus)1Mnz/J (ItgaxeYFP) mice, CD207cre × B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/J (CD207eYFP) mice23 and Foxp3eGFP mice53 were bred and housed under specific pathogen–free conditions according to the guidelines of the University of Minnesota Institutional Animal Care and Use Committee and the National Institutes of Health. AireeGFP mice36 were bred and housed under specific pathogen–free conditions in accordance with the National Institutes of Health and American Association of Laboratory Animal Care standards, and with the animal care and use regulations of the University of California, San Francisco. Analysis of thymus cell numbers was carried out in mice 3–5 weeks of age.

Immunization. Mice were given subcutaneous injection of 100 µl (split over two sites) of CFA emulsion (Sigma-Aldrich) containing 100 µg of peptide (GenScript) in Dulbecco’s PBS (Life Technologies). Mice were immunized with specific peptides contained in the tetramers described below with the following exceptions: MOG peptide (GWYRSPFSRVVHL) and peptide from calnexin from B. dermatitidis (LVVKNPAAHHAI).

Infection. The ∆actA∆inlB strain of Lm-eGFP (PL1113) was provided by P. Lauer. Sequence encoding the eGFP construct (hyperSPO1 promoter-eGFP)54 was cloned into the pPL1 vector55 and was integrated at the comK locus in the actA-inlB-double mutant56. Mice were given intravenous injection of 1 × 107 Lm-eGFP bacteria and were analyzed 6–7 d after infection.

Tetramers. Biotin-labeled I-Ab monomers containing eGFP peptide (HDFFKSAMPEGYVQE), 2W (EAWGALANWAVDSA), peptide from listeriolysin O protein of L. monocytogenes (NEKYAQAYPNVS), peptide from calnexin from B. dermatitidis (LVVKNPAAHHAIS), peptide from junctophilin 2 (VYTWPSGNTFE), peptide from myelin oligodendrocyte glycoprotein (GWYRSPFSRVV), peptide from proteolipid protein (CLVGAPFASLVA), peptide from the constant region of immunoglobulin M (EKYVTSAPMPEPGAPG)57, I-Ab-binding-stabilized variant of peptide from ATP-binding cassette sub-family G, member 1 (YYMTPQPSDVV) or peptide from protein disulfide–associated isomerase 2 (EYSKAAALLAA) covalently linked to the I-Ab β-chain were grown in Drosophila melanogaster S2 cells also expressing the I-Ab α-chain. Monomers were purified and com-bined with streptavidin-phycoerythrin (PE) or streptavidin-allophycocyanin (APC) (Prozyme) to produce fluorescence-labeled I-Ab tetramers as described16,37,58.

Cell enrichment and flow cytometry. Single-cell suspensions were pre-pared from pooled mouse spleens and lymph nodes (axillary, brachial, inguinal, cervical, mesenteric, pancreatic and para-aortic) or thymuses by mechanical disruption. Cells were stained for 1 h at room temperature with allophycocyanin-conjugated tetramers. In some experiments cells were also stained with phycoerythrin-conjugated tetramers. Samples underwent enrich-ment for tetramer-binding cells on magnetized columns as described58. For thymic epithelial cell and dendritic cell analysis, thymuses were harvested and enzymatically digested as described59. Single-cell suspensions were stained with biotinylated antibody to (anti-)CD11c (HL3; BD) and allophycocyanin-labeled anti-EpCAM (G8.8; BioLegend). Cells were enriched using EasySep Biotin Positive Selection (18559; StemCell Technologies) and EasySep Mouse APC Positive Selection (18452; StemCell Technologies) kits according to the manufacturer’s instructions.

Antibodies. Samples enriched for tetramer-binding cells were stained for surface markers for 30 min on ice with antibodies to the following: CD4 (GK1.5; BD), CD8 (53-6.7; BD), CD90.1 (HIS51; eBioscience), CD90.2 (53-2.1; 30-H12; eBioscience), CD3e (145-2C11; BD), CD11b (M1/70; eBioscience), CD11c (N418; eBioscience), F4/80 (BM8; eBioscience), CD44 (IM7; BD), CD45.1 (A20; BioLegend), CD45.2 (104; eBioscience) and/or B220 (RA3–6B2; eBioscience). Cellular viability was confirmed with GhostDye Red 780 (Tonbo Biosciences). For analysis of the expression of transcription factors, stained cells were fixed and permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to the manufacturer’s instructions. Cells were stained overnight at 4 °C with anti-T-bet (4B10; BioLegend), anti-Foxp3 (FJK-16s; eBioscience), anti-RORγt (Q31-378; BD) and anti-Bcl-6 (K112-91; BD). For analysis of thymic dendritic cells and thymic epithelial cells, following enrichment, cells were stained with anti-CD11c (N418; eBioscience), anti-CD19 (6D5; BioLegend), anti-I-A/I-E (M5/114.15.2; BioLegend), anti-CD8 (53-6.7; BD), anti-CD90.2 (53-2.1, 30-H12; eBioscience), anti-CD11b (M1/70; eBioscience) and anti-CD45.2 (104; BD). Cells were analyzed by flow cytometry on a Fortessa (BD). Data were analyzed using FlowJo software (TreeStar).

Thymus transplants. Thymus transplants were performed as described60. Thymus lobes were harvested from prenatal pups (day 18 after conception) generated by crossing of Ins2eGFP mice and CD90.1+ B6 mice. An incision was made on the left abdomen of anesthetized 4- to 6-week-old female B6-nude to expose the kidney, and the donor thymic lobes were grafted beneath the kidney capsule. The muscle layer was sutured and the skin was closed with surgical clips. Thymic engraftment was assessed 6 weeks after transplantation by examination of peripheral blood for the presence of mature αβ T cells by flow cytometry. Donor genotypes were determined after transplantation.

Statistical analysis. Two-tailed, unpaired Student’s t-tests, one-way ANOVA, χ2 test, and linear regression were performed with Prism (GraphPad) software.

Hierarchical clustering. Unsupervised hierarchical clustering and heat-map generation were carried out with the ‘HierarchicalClustering’ and ‘HierarchicalClusteringViewer’ modules of GenePattern software (Broad Institute)17,61,62.

51. Gupta, R.K. et al. Zfp423 expression identifies committed preadipocytes and localizes to adipose endothelial and perivascular cells. Cell Metab. 15, 230–239 (2012).

52. Xin, H.B., Deng, K.Y., Rishniw, M., Ji, G. & Kotlikoff, M.I. Smooth muscle expression of Cre recombinase and eGFP in transgenic mice. Physiol. Genomics 10, 211–215 (2002).

53. Fontenot, J.D. et al. Regulatory T cell lineage specification by the forkhead transcription factor foxp3. Immunity 22, 329–341 (2005).

54. Shen, A. & Higgins, D.E. The 5′ untranslated region-mediated enhancement of intracellular listeriolysin O production is required for Listeria monocytogenes pathogenicity. Mol. Microbiol. 57, 1460–1473 (2005).

55. Lauer, P., Chow, M.Y., Loessner, M.J., Portnoy, D.A. & Calendar, R. Construction, characterization, and use of two Listeria monocytogenes site-specific phage integration vectors. J. Bacteriol. 184, 4177–4186 (2002).

56. Brockstedt, D.G. et al. Listeria-based cancer vaccines that segregate immunogenicity from toxicity. Proc. Natl. Acad. Sci. USA 101, 13832–13837 (2004).

57. Chu, H.H., Moon, J.J., Kruse, A.C., Pepper, M. & Jenkins, M.K. Negative selection and peptide chemistry determine the size of naive foreign peptide-MHC class II-specific CD4+ T cell populations. J. Immunol. 185, 4705–4713 (2010).

58. Moon, J.J. et al. Naive CD4+ T cell frequency varies for different epitopes and predicts repertoire diversity and response magnitude. Immunity 27, 203–213 (2007).

59. Malhotra, D. et al. Transcriptional profiling of stroma from inflamed and resting lymph nodes defines immunological hallmarks. Nat. Immunol. 13, 499–510 (2012).

60. Berzins, S.P., Boyd, R.L. & Miller, J.F. The role of the thymus and recent thymic migrants in the maintenance of the adult peripheral lymphocyte pool. J. Exp. Med. 187, 1839–1848 (1998).

61. de Hoon, M.J., Imoto, S., Nolan, J. & Miyano, S. Open source clustering software. Bioinformatics 20, 1453–1454 (2004).

62. Eisen, M.B., Spellman, P.T., Brown, P.O. & Botstein, D. Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA 95, 14863–14868 (1998). ������

npg

© 2

016

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.