©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology  aDVaNCE ONLINE PUBLICaTION �

A rt i c l e s

T cell development occurs in the thymus, which provides a unique microenvironment and presents ligands consisting of self peptide and major histocompatibility complex (MHC) molecules to T cell anti-gen receptors (TCRs). In the cortex of the thymus, low-affinity TCR interactions initiate positive selection signals in CD4+CD8+ double- positive (DP) immature thymocytes, which supports their survival and differentiation into CD4+ or CD8+ single-positive (SP) thymo-cytes. Positively selected cells move to the medullary region and, after several days, emigrate to the periphery. The term ‘positive selection’ is sometimes used to describe the entire process. However, in think-ing about molecular mechanisms, it is helpful to break this down into kinetically distinct processes, such as survival, allelic exclusion, lineage commitment and functional maturation.

Cortical DP thymocytes require interaction of the surface TCR with selecting peptide-MHC complexes to induce survival. Expression of the activation marker CD69, the TCR and the anti-apoptotic protein Bcl-2 is rapidly upregulated in cells of this population, which also undergo changes in the expression of many other genes1,2. Genetic deficiency in TCRs, MHC molecules, the co-receptors CD4 or CD8 or molecules in the TCR signaling pathway blocks this process2. Recombination-activating genes are rapidly repressed at this stage, which facilitates allelic exclusion. Expression of the chemokine receptor CCR7 is upregulated somewhat later and facilitates the migration of progenitor cells from the cortex to the medulla3. Lineage commitment occurs concurrently and involves downregulation of the gene encoding the inappropriate co-receptor and the initiation of genetic remodeling that will ultimately determine if the cell has helper or killer potential. Genetic deficiency in key transcription factors can block commitment to the CD4+ or CD8+ lineage4,5. Although line-age commitment is mechanistically independent of migration to the medulla, these processes are roughly concurrent. Thus, SP thymocytes

reside predominantly in the medulla; however, not all SP thymocytes are equivalent.

CD24hiQa2lo SP thymocytes have been defined as ‘semi-mature’ and have been shown to be susceptible to apoptosis when triggered through the TCR6. In contrast, mature SP thymocyte and thymic emigrants proliferate when triggered through the TCR6,7. Over the years, studies have shown that the expression of a small number of other cell-surface proteins, including CD69 and various cytokine and chemokine receptors, changes during maturation8,9. However, the molecular mechanisms that control the maturation of SP thymocytes have remained unclear.

Here we defined steps in the maturation of SP thymocytes through which SP thymocytes became equipped with mature functions, such as proliferation competency, emigration competency and cytokine ‘licensing’. For this, we used comprehensive microarray analysis, quantitative real-time PCR (qPCR) and flow cytometry, along with combinations of several gene-deficient and transgenic mouse models. We found that SP thymocytes received signals via both the cytokine TNF and type I interferons in the thymus and that only TNF-resistant mature thymocytes survived and became emigration competent and licensed to produce cytokines.

RESULTSThree SP stages defined by functionIn this study, we sought to determine the ideal markers for flow cytometry to define SP thymocyte stages by function. The ordered development of SP thymocytes has been characterized by the expres-sion of cell-surface makers (CD24, CD69, CD62L and Qa2)6,10; the carbohydrate epitope 6C10 on the glycoprotein Thy-1, recognized by the monoclonal antibody SM6C10 (ref. 9); or chemokine receptors CCR7 and CCR9 (ref. 11). However, these markers have not been well

The Center for Immunology, Department of Laboratory Medicine & Pathology, University of Minnesota, Minneapolis, Minnesota, USA. Correspondence should be addressed to K.A.H. (hogqu001@umn.edu).

Received 22 October 2015; accepted 19 February 2016; published online 4 April 2016; doi:10.1038/ni.3419

Late stages of T cell maturation in the thymus involve NF-B and tonic type I interferon signalingYan Xing, Xiaodan Wang, Stephen C Jameson & Kristin A Hogquist

Positive selection occurs in the thymic cortex, but critical maturation events occur later in the medulla. Here we defined the precise stage at which T cells acquired competence to proliferate and emigrate. Transcriptome analysis of late gene changes suggested roles for the transcription factor NF-B and interferon signaling. Mice lacking the inhibitor of NF-B (IB) kinase (IKK) kinase TAK1 underwent normal positive selection but exhibited a specific block in functional maturation. NF-B signaling provided protection from death mediated by the cytokine TNF and was required for proliferation and emigration. The interferon signature was independent of NF-B; however, thymocytes deficient in the interferon- (IFN-) receptor IFN-R showed reduced expression of the transcription factor STAT1 and phenotypic abnormality but were able to proliferate. Thus, both NF-B and tonic interferon signals are involved in the final maturation of thymocytes into naive T cells.

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

�  aDVaNCE ONLINE PUBLICaTION nature immunology

A rt i c l e s

correlated with functional maturation. For our comprehensive flow cytometry, we used mice with expression of green fluorescent protein (GFP) directed by the promoter of recombination-activating gene 2 (Rag2) via a bacterial artificial chromosome transgene (Rag2GFP), in which GFP expression (as Rag2-GFP) acts as a ‘molecular timer’ for differentiation events after positive selection10,12 and allows the exclusion of re-circulating mature T cells. To focus our analysis on conventional αβ T cells, we used a ‘dump strategy’ to exclude γδ T cells, invariant natural killer T cells (iNKT cells) and regulatory T cells (Treg cells) (Supplementary Fig. 1a). We found that the combination of staining for CD69 and MHC class I (MHCI) precisely defined SP thymocyte stages by function and could be used on cells of both the CD4+ lineage and CD8+ lineage; positively selected medullary thy-mocytes (TCRβ+CCR7+) clearly had three populations, on the basis of the expression of CD69 and MHC class I, including CD69+MHCI− cells, CD69+MHCI+ cells and CD69−MHCI+ cells13 (Fig. 1a). GFP expression gradually decreased in this order (Fig. 1a, top right), which indicated that CD69+MHCI− thymocytes preceded CD69+MHCI+ cells and that CD69−MHCI+ cells were the most mature population. After the initiation of positive selection, CD4+ SP (CD4SP) cells quickly repress CD8 expression, whereas CD8+ SP (CD8SP) cells extin-guish CD4 expression more slowly14. Thus, MHC class I–restricted medullary thymocytes were in the DP gate at the CD69+MHCI− stage and did not complete downregulation of CD4 expression until the CD69−MHCI+ stage (Fig. 1a). This gating strategy was confirmed in MHC class II–deficient mice, in which only CD8+ T cells could be positively selected in the thymus (Supplementary Fig. 1b).

To assess proliferation competence, we sorted the three popula-tions, labeled them with the proliferation-detection fluorescent dye CellTrace Violet and stimulated them in vitro with antibody to the invariant signaling protein CD3 (anti-CD3) plus antibody to the co-receptor CD28 (anti-CD28). CD69+MHCI− cells did not proliferate, while CD69+MHCI+ and CD69−MHCI+ thymocytes did (Fig. 1b), which suggested that upregulation of the expression of MHC

class I most precisely defined the boundary between proliferation- incompetent cells and proliferation-competent cells. Thus, we designated CD69+MHCI− population as semi-mature (SM) and des-ignated CD69+MHCI+ and CD69−MHC1+ populations as mature 1 (M1) and mature 2 (M2), respectively. To assess emigration and trafficking competence, we assessed expression of the sphingosine 1-phosphate (S1P) receptor S1PR1 (ref. 15), L-selectin (CD62L)16 and the transcription factor KLF2, which is required for the expression of S1PR1 and CD62L in thymocytes15–17. There was high expression of S1PR1 and CD62L on M2 cells but not on SM cells or M1 cells (Fig. 1c). Likewise, there was high expression of KLF2 only on M2 cells (Supplementary Fig. 1c). Thus, among proliferation-competent M1 and M2 thymocytes, only the most mature M2 cells were competent to emigrate. Finally, we assessed at which stage SP thymocytes became licensed to produce the cytokine TNF18. Only the M2 subset had a substantial population of TNF-producing cells following stimulation via anti-CD3 and anti-CD28 (Fig. 1d), and this fraction continued to increase among recent thymic emigrants (data not shown), consistent with published reports18.

Published staining combinations6,9 used to define two popula-tions (CD24hiQa2lo and CD24loQa2hi) or four populations (SP1–SP4) with the combination of CD69, Qa2 and monoclonal antibody SM6C10 directed against Thy-1 did not precisely distinguish between proliferation-incompetent (SM) cells and proliferation-competent (M1) cells, although this nicely distinguished emigration-competent cells from other cells (Supplementary Fig. 1d). The combination of staining for CCR7 and CCR9 together with CD69 on CD4SP thymo-cytes11 somewhat distinguished proliferation-incompetent (CCR9hi) CD4SP cells from proliferation-competent (CCR9int–lo) CD4SP cells (Supplementary Fig. 1d, left), but this staining panel did not separate CD8SP thymocytes very well (Supplementary Fig. 1d, right), as mature CD8SP cells had abundant expression of CCR9. Thus, expression of CD69 and MHC class I was the most effective measure of the late stages of T cell maturation.

Rag2-GFP+

thymocytes CCR7+TCRβ+

CC

R7

CD4SP

CD8SP

CD4SP

CD8SP

CD62L

TNF

CD4SP

CD8SP

CTV

Eve

nts

CD

4

S1P

R1

SS

C

CD8

TCRβ MHCI Rag2-GFP

M2

M2

M2

M2

M1

M1

M1M1

CD

69

9.3105

105

104

104

103

103

0

0105

105

104

104

103

103

0

0

105

105

104

104

103

103

0

0

105

105

104

104

103

103

0

0 105104103

–103

0

1051041030

103

104

105

0

SM

SM

M2

CD3+CD28Medium

M1SM

SM

M2M1SM

72.720.7

0.0

0.0 2.1 13.5

12.91.80.1

1.4 54.0

58.04.31.7

69.7

27.7

66.2

30.5

10080604020

0

60

40

20

250K200K150K100K50K

0

0

SM

Eve

nts

a b c

d

Figure 1 Three major SP stages defined by function. (a) Flow cytometry of thymocytes from Rag2GFP mice (n = 4), showing the expression of CCR7 and TCRβ by Rag2-GFP+ thymocytes (dump-gated as in Supplementary Fig. 1a) (top left), gates of three subsets of medullary thymocytes identified by expression of CD69 and MHC class I (top middle), and GFP expression by CD69+MHCI− (SM) cells, CD69+MHCI+ (M1) cells and CD69−MHCI+ (M2) cells (top right), as well as CD4- and CD8-expression profiles of the populations above (bottom). Numbers adjacent to outlined areas indicates percent cells in each. (b) Proliferation of the SM, M1 and M2 subsets of CD4SP thymocytes (sorted from C57BL/6 mice; top row) and CD8SP thymocytes (sorted from MHC class II–deficient mice; bottom row) labeled with CellTrace Violet (CTV), left unstimulated (Medium) or stimulated for 3 d with anti-CD3 and anti-CD28 (CD3+CD28), and analyzed by flow cytometry. (c) Expression of S1PR1 and CD62L on the subsets of CD4SP and CD8SP thymocytes identified in a. Numbers adjacent to outlined areas indicate percent S1PR1+CD62L+ cells. (d) TNF production by CD4SP or CD8SP thymocytes 4 h after stimulation with anti-CD3 and anti-CD28. Numbers adjacent to outlined areas indicate percent TNF+ cells. SSC, side scatter. Data are representative of four (a) or three (b–d) independent experiments.

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology  aDVaNCE ONLINE PUBLICaTION �

A rt i c l e s

Transcriptome analysis indicates a role for NF-B and IFNsDP thymocytes have a gene-expression profile substantially different from that of T cells19, consistent with the susceptibility of the former to apoptosis and the proliferation competence of the latter. However, our data and published results would suggest that the apoptosis- to-proliferation change occurs at the semi-mature–to–mature stage in the medulla, which we estimate to be at least 24 h after the initiation of positive selection signaling6,20. Thus, we hypothesized that the unique gene changes that occur late in the process might be most relevant to the understanding of this competence transition. To characterize these gene changes, we ‘mined’ microarray data from thymocytes isolated using the previously defined markers Qa2 and CD69 to des-ignate four stages. CD69−TCRβ− pre-selection DP (‘pre-DP’) cells and CD69+TCRβ+ post-selection DP (‘post-DP’) cells (Fig. 2) were sorted from the thymus of β2-microglobulin-deficient (B2m−/−) mice (data not shown), in which only MHC II–restricted CD4+ T cells undergo positive selection. Qa2−CD69+ (SM-M1) CD4SP cells and Qa2+CD69− (M2) CD4SP cells (Fig. 2) were sorted from thymocytes of Rag2GFP mice with a ‘dump’ channel that included the activating NK cell receptor NK1.1, the T cell–activation marker and cytokine receptor CD25 and the antibody GL3, directed against TCRδ (data not shown). The ‘dump’ and Rag2-GFP gates allowed the exclusion of iNKT cells, Treg and γδ T cells, as well as recirculating mature T cells, which can include up to 15% of CD4SP cells (Supplementary Fig. 2a). Through analyzing the microarray data, we identified genes with significant changes in expression and categorized them by their expression-kinetic patterns (Fig. 2). By far the largest number of gene-expression changes (2,060 genes up- or downregulated) were initiated during the cortical positive selection step from pre-DP to post-DP (Supplementary Fig. 2b). Some of these changes occurred only early (group A), some continued to change at later stages (group B) and a few were transient (group C) (Supplementary Fig. 2b). Notably, a substantial number of gene-expression changes occurred between the post-DP stage and SM stage (1,005 genes; groups D and E), whereas only a small number of genes altered their expression at the very end of medullary maturation (269 genes; group F), including a large number of genes whose expression changes were initiated earlier but then were reversed (Supplementary Fig. 2b). Using these expression-pattern groups, we defined sets of genes whose expression changed early or late during selection (Fig. 2).

To understand which upstream pathways and molecular factors might be acting at early and late stages, we used gene- set–enrichment analysis (GSEA) tools. Distinct stages showed enrich-ment for the expression of certain sets of gene that share a conserved cis-regulatory motif (called the ‘C3 motif ’ in GSEA) in their promot-ers and 3′ untranslated regions (Table 1). The promoters of genes induced or repressed during positive selection (pre-DP to post-DP) showed enrichment for various transcription factor–binding sites, including those motifs recognized by the transcription factors ATF, EGR, NF-κB, ELK1 or SRF, many of which have already been shown to be involved in positive selection21,22. Genes regulated by the transcription factor E2F showed a negative-enrichment score, indica-tive of downregulation. Among genes whose expression changed at

later time points during the post-DP–to–SM-M1 transition, there was enrichment only for the expression of gene sets regulated by E2F, NF-κB and transcription factors of the IRF family. E2F- regulated genes seemed to be upregulated at this stage, reflective of the transient nature of their repression during positive selection. Among the NF-κB-regulated genes, some changes were initiated earlier, at the pre-DP–to–post-DP transition, and others were initiated later, at the post-DP stage. Substantial enrichment for the expression of IRF-regulated genes was seen only at the post-DP–to–SM-M1 tran-sition. The number of gene changes observed at the latest stage of maturation was small (413), and there was no significant positive enrichment for any transcription factor–binding sites at this stage (normalized enrichment score of ≥1.6 or ≤–1.6). Because this analysis suggested that NF-κB- and IRF-regulated gene changes occurred late after positive selection, we focused on their effects in this study.

Requirement for kinase TAK1 in the transition from SM to M1Mice deficient in the expression of various genes encoding compo-nents of the NF-κB pathway, including those encoding the kinase TAK1 (encoded by Map3k7; called ‘Tak1’ here), the inhibitor of NF-κB (IκB) kinases (IKKs) IKK2 or IKKγ, the NF-κB subunit RelB or the ubiquitin-conjugating enzyme Ubc13, display nearly normal num-bers of total thymocytes but few peripheral T cells23. Among these, we focused on TAK1 as a central kinase in this pathway (Supplementary Fig. 3a). Consistent with published studies24–26, we found that mice with T cell–specific deficiency in TAK1 (which undergo deletion of loxP-flanked alleles encoding TAK1 by Cre recombinase expressed from the T cell–specific Cd4 promoter (Tak1fl/flCd4Cre)) had a nor-mal number of thymocytes relative to that of their Tak1fl/fl (control) littermates, but the proportion and absolute number of both CD4SP

Pre-DP

A (1,319) B (395)

Early Late

C (346) D (861) E (144) F (269)

Post-DP SM-M1CD4SP

M2CD4SP

Figure 2 Transcriptome analysis of late stages of thymocytes. Microarray analysis (bottom) of sorted pre-DP thymocytes, post-DP thymocytes, Qa2−CD69+ (SM-M1) thymocytes and Qa2+CD69− (M2) CD4SP thymocytes (top) categorizing gene-expression changes as six patterns (A–F) and two categories (Early and Late); numbers in parentheses indicate total genes in each pattern group. Data are pooled from three independent experiments.

Table 1 GSEA of microarray data

Pre-DP to post-DP Post-DP to SM-M1 SM-M1 to M2

Gene set # NES Gene set # NES Gene set # NES

ATF 3 1.97 E2F 14 1.84 None ≥1.60EGR 6 1.86 NF-B 1 1.62 None ≤−1.60CREB1 3 1.83 IRF 1 1.62E4F1 1 1.82 ATF 1 −1.60NF-B 4 1.78 EGR 2 −1.73ETS 3 1.73 PAX8 1 −1.62ELK1 1 1.71 NFAT 1 −1.60SRF 2 1.70PU.1 1 1.65STAT1 1 1.62NERF 1 1.61E2F 8 −1.62

GSEA of microarray data, showing gene sets whose expression underwent enrichment and that share a conserved transcription factor–binding motif in the promoters and 3′ untranslated regions, among genes whose expression changed (upregulated (bold) or downregulated (not bold)) at the pre-DP–to–post-DP, post-DP–to–SM-M1 and SM-M1–to–M2 transitions (column headings); only gene sets with a normalized enrichment score (NES) of ≥1.6 or ≤−1.6 are included here. #, number of gene sets related to that transcription factor that were labeled.

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

�  aDVaNCE ONLINE PUBLICaTION nature immunology

A rt i c l e s

thymocytes and CD8SP thymocytes were lower in Tak1fl/flCd4Cre mice than in their control littermates (Fig. 3a,b). We further investigated the stage at which TAK1 deficiency affected maturation and found a normal number of SM cells but a profound reduction in the number of M1 cells and M2 cells (of 10-fold and 100-fold, respectively) in Tak1fl/flCd4Cre mice, relative to the number of these cells their control littermates (Fig. 3a,b). This suggested that TAK1-dependent signals were not required for positive selection itself but were required for the ultimate survival and/or maturation of T cells. Mature iNKT cells, Treg cells and intraepithelial lymphocyte precursor thymocytes were almost completely absent from Tak1fl/flCd4Cre mice as well (Supplementary Fig. 3b).

To determine if TAK1 could be a key driver of late gene-expression changes during selection, we performed microarray analysis of puri-fied SM CD4SP thymocytes from Tak1fl/flCd4Cre and control mice. 381 genes were up- or downregulated by TAK1 deficiency (data not shown). Using a heat map to visualize the overall pattern of gene changes, we noted that TAK1-dependent genes were changed mainly at the late stage of maturation, rather than an early stage, in wild-type mice (Fig. 3c,d). Next we used Ingenuity pathway analysis to define potential upstream regulators on the basis of the P value of the overlap between our gene lists and those defined by the literature. The upstream regulators defined for early gene-expression changes included TCR, TGF-β and members of the ‘Id’ family of transcrip-tion factors, among others (Supplementary Fig. 4). In contrast, late gene-expression changes overlapped many interferon and IRF path-ways (Supplementary Fig. 4), consistent with the enrichment for IRF-binding sites in the promoters at transcripts that underwent late changes in expression (Table 1). Notably, results obtained for the upstream regulators defined for TAK1-dependent genes were very similar to those of transcripts that underwent late changes (yellow

or green in Supplementary Fig. 4), which suggested that TAK1- dependent processes were the dominant signals that drove the gene-expression changes late in positive selection.

Restoration of survival but not proliferation by TNF blockadeBecause activation of NF-κB protects cells from TNF-induced cell death27, we investigated whether the role of TAK1 signaling in positive selection might be mainly to allow survival at the mature stage or if it was required for differentiation and the acquisition of proliferation competence. To study this, we crossed Tak1fl/flCd4Cre mice with TNF-deficient (Tnf−/−) mice. In Tak1fl/flCd4CreTnf−/− mice, the number of M1 cells was completely restored to equivalence to that of Tak1fl/fl mice (Fig. 4a,b), which suggested that one role for TAK1 signals might be to protect cells from TNF-induced death. However, M2 cells were ‘rescued’ only partially, and Tak1fl/fl Cd4CreTnf−/− mice were profoundly lymphopenic in the periph-ery (data not shown). The expression of multiple other members TNF receptor family was upregulated during thymic maturation (Supplementary Fig. 5a,b) and, similar to the receptor TNFR1, one of these, DR3, contains a death domain, and cells expressing this recep-tor might also require NF-κB signals for protection from induced cell death. However, crossing Tak1fl/flCd4Cre mice with mice with transgenic expression of Bcl2 (data not shown) or mice genetically defi-cient in Bim, a proapoptotic member of the Bcl-2 family (Bcl2l11−/−), yielded a phenotype similar to that of Tak1fl/flCd4CreTnf−/− mice, with low numbers of M2 cells and peripheral lymphopenia (Supplementary Fig. 5c,d). Although TNF deficiency restored the number of mature cells, those cells were unable to proliferate in response to anti-CD3 plus anti-CD28 (Fig. 4c). These data suggested that TAK1 signals might have been required for differentiation, in addition to being required for cell survival.

WT vs TAK1

Early

Late

Tak1fl/fl

Tak1fl/flCd4Cre

1010

109

108

107

106Tot

al th

ymoc

ytes NS

108

107

106

105

104SM

CD

4SP

cel

ls NS 108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

M1

CD

4SP

cel

ls

M2

CD

4SP

cel

ls

SM

CD

8SP

cel

ls

M1

CD

8SP

cel

ls

* *

* *NS

M2

CD

8SP

cel

ls

10

11 10

Expression (fold)qPCR

Exp

ress

ion

(fol

d)m

icro

arra

y

Totalthymocytes

105

104

103

–1030C

D4

105

CD8

6.6 87.8

2.6

–103

1030

104

Tak1fl/fl Tak1fl/flCd4Cre

2.2 94.0

2.4 0.62.2

CD4SP

CD

69

37.6

30.6

28.2

88.1

8.0

0.7

CD8SP

105

104

103

0

MHCI

30.3

32.6

33.1

86.9

4.6

1.3

105

–103

1030

104

a

c

b

d

Figure 3 TAK1 is required for the SM-to-M1 transition. (a) Flow cytometry of total thymocytes from Tak1fl/fl and Tak1fl/flCd4Cre mice (top) and thymocytes in the CD4SP or CD8SP lineage (with the exclusion of CD1d– α-GalCer+ NKT cells, CD25+ Treg cells, CD44hi recirculating memory cells and GL3+ γδ T cells) (below). Numbers adjacent to outlined areas indicate percent CD4+CD8− cells (top left), CD4+CD8+ cells (top right), CD4−CD8+ cells (bottom right) or CD4−CD8− cells (bottom left) (top row), or CD69+MHCI− (SM) cells (top left), CD69+MHCI+ (M1) cells (top right) or CD69−MHCI+ (M2) cells (bottom right) (middle and bottom rows). (b) Quantification of total thymocytes (top) and SM, M1 and M2 CD4SP or CD8SP thymocytes (below) in Tak1fl/fl mice (n = 10) and Tak1fl/flCd4Cre mice (n = 10). Each symbol represents an individual mouse; small horizontal lines indicate the group mean (± s.d.). NS, not significant (P > 0.05); *P ≤ 0.001 (two-tailed unpaired Student’s t-test). (c) Microarray analysis of 381 genes expressed differentially (twofold or more (red, upregulated; blue, downregulated); P ≤ 0.05) by sorted SM-M1 CD4SP thymocytes from Tak1+/+ mice versus those from Tak1fl/flCd4Cre mice (WT vs TAK1; top row) or from Tak1+/+ mice and Tak1fl/flCd4Cre mice (below; ‘Early’ and ‘Late’ as in Fig. 2). (d) qPCR analysis of the expression of select TAK1-dependent genes separately sorted SM-M1 CD4SP thymocytes from Tak1fl/fl and Tak1fl/flCd4Cre mice, to confirm the microarray analysis results; PCR results are normalized to those of the control gene Hprt, and microarray results are those in Tak1fl/fl cells relative to those of Tak1fl/flCd4Cre cells. Each symbol represents an individual gene. Pearson correlation coefficient r = 0.46; P < 0.05 (two tailed). Data are representative of (a) or pooled from (b) nine independent experiments or are pooled from three independent experiments with biological triplicates (c,d; mean).

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology  aDVaNCE ONLINE PUBLICaTION �

A rt i c l e s

Restoration of proliferation but not licensing by IKK activityTAK1 signals result in the activation of both NF-κB and mitogen-activated protein kinases28. To determine which TAK1-dependent functions involve NF-κB, we used a transgene encoding constitutively active IKK2 (called ‘IKKCA’ here) to restore the activation of NF-κB in TAK1-deficient mice. We crossed mice expressing a Cre-inducible IKKCA-encoding transgene to Tak1fl/flCd4Cre mice (the use of Cd4Cre avoids the documented effects of IKKCA on the DN thymocytes29). Both the proportion and number of mature SP thymocytes in Tak1fl/fl Cd4CreIKKCA mice were greater than that of Tak1fl/flCd4Cre mice and showed no significant difference relative to that of the Cd4CreIKKCA control mice (Fig. 5a,b). Notably, the proportion and number of cells in the M1 and M2 subsets were ‘rescued’ to levels similar to those of Cd4CreIKKCA control mice (Fig. 5a,b). Furthermore, M1 and M2 cells from the thymus of Tak1fl/flCd4CreIKKCA mice proliferated in response to anti-CD3 and anti-CD28, as did those from control Tak1fl/fl mice, albeit to a slightly lesser degree (Fig. 5c). These results indicated that the activation of NF-κB driven by IKKCA restored phenotypic maturation and proliferation competence of TAK1- deficient SP thymocytes.

Finally, we assessed TNF production in the TAK1-deficient SP thymocytes restored via IKKCA. In contrast to the 15–20% of TNF-producing cells among the M2 population from Tak1fl/fl mice, only 1–2% of the M2 cells from Tak1fl/flCd4CreIKKCA mice produced TNF after stimulation with anti-CD3 and CD28 (Fig. 5d). Furthermore, Tak1fl/flCd4CreIKKCA mice had very few CD4+ or CD8+ T cells in the spleen, similar to Tak1fl/flCd4Cre mice (Fig. 5e,f). M2 cells from Tak1fl/flCd4CreIKKCA mice expressed normal amounts of S1PR1 (Fig. 5g), which challenged the possibility of a problem with the emigration competency of M2 cells and raised the possibility that those cells could not survive in the periphery. Thus, although

IKKCA restored several aspects of maturation in TAK1-deficient SP thymocytes, it did not restore cytokine-production competence or the establishment of T cell populations in the periphery.

Facilitation of IFN signaling by TAK1 independently of NF-BTo understand which TAK1-dependent gene-expression changes were dependent on NF-κB activity and which were independent of this, we measured the expression of 20 TAK1-dependent genes that were iden-tified by microarray and whose expression changes were confirmed by qPCR (Fig. 3c,d). Thus, we sorted SM and M2 cells from Tak1fl/fl, Tak1fl/flCd4Cre and Tak1fl/flCd4CreIKKCA mice and measured gene expression by qPCR. The expression of about half (9) of these genes was restored by the IKKCA-encoding transgene and the expression of about another half (11) was not (Fig. 6a). Notably, many of the latter genes were interferon-regulated genes, which suggested that TAK1 regulated interferon signaling in a pathway that could not be replaced by NF-κB activity. We noticed that expression of Stat1 mRNA itself was not restored by IKKCA (Fig. 6a); STAT1 is an essential media-tor of interferon signaling. Thus, we measured intracellular STAT1 protein. STAT1 expression increased from the SM stage to the M1 stage, and there was less total STAT1 protein in M1 and M2 SP thy-mocytes from Tak1fl/flCd4CreIKKCA mice than in those of Tak1fl/fl control mice (Fig. 6b), which suggested that these cells might have had reduced responsiveness to interferons.

Published studies have shown that interferon-β (IFN-β) is con-stitutively expressed in thymic medullary epithelial cells from naive mice30,31. Given that an interferon-regulated gene signature was apparent in medullary thymocytes in our microarray analysis, we sought to test the hypothesis that maturing thymocytes might respond to interferon constitutively produced in the thymus. To address this, we studied mice deficient in the receptor for type I interferons

Totalthymocytes

6.9

2.4

87.1 3.2 5.3

3.33.1 0.6

92.3 89.5

34.1 78.7 56.1

35.1 16.7 30.9

32.9 78.9 53.7

36.1 13.6 32.4

2.4 8.127.3

7.91.427.6

CD4SP

CD8SP

CD

69

1.02.7

Tak1fl/fl Tak1fl/flCd4CreTak1fl/flCd4Cre

Tnf –/–

105

105

104

104

103

103

102

–1020

0

1051041030

CD

4

105

104

103

0

CD8

MHCI

a

Tota

l thy

moc

ytes

SM

CD

4SP

cel

ls

NS NS

*

Tak1fl/fl

Tak1fl/flCd4Cre

Tak1fl/flCd4CreTnf –/–

1010

109

108

107

106

108

107

106

105

104 M1

CD

4SP

cel

ls

108

107

106

105

104

M2

CD

4SP

cel

ls 108

107

106

105

104 SM

CD

8SP

cel

ls 108

107

106

105

104

M1

CD

8SP

cel

ls 108

107

106

105

104 M2

CD

8SP

cel

ls 108

107

106

105

104

NS

NS NS *

b

SM M1 M2Medium CD3+CD28

CTV

8001.0K

400200

0

600

Tak1fl/fl

Tak1fl/fl

Cd4Cre

Tnf –/–

Eve

nts

–103 103 104 1050

c

Figure 4 Blockade of TNF activity restores maturation but not proliferation. (a) Flow cytometry of thymocytes from Tak1fl/fl, Tak1fl/flCd4Cre and Tak1fl/flCd4CreTnf−/− mice, showing CD4 and CD8 profiles (top) and the proportion of SM, M1 and M2 CD4SP or CD8SP cells (with exclusion of cells as in Fig. 3a) (below). Numbers adjacent to outlined areas as in Figure 3a. (b) Quantification of total thymocytes (top) or SM, M1 and M2 CD4SP or CD8SP thymocytes (below) from Tak1fl/fl mice (n = 10), Tak1fl/flCd4Cre mice (n = 10) and Tak1fl/flCd4CreTnf−/− mice (n = 4). Each symbol represents an individual mouse; small horizontal lines indicate the group mean (± s.d.). *P ≤ 0.001 (two-tailed unpaired Student’s t-test). (c) Proliferation of SM, M1 and M2 CD4SP thymocytes sorted from Tak1fl/fl and Tak1fl/flCd4CreTnf−/− mice and left unstimulated or stimulated for 3 d with anti-CD3 plus anti-CD28 (assessed as in Fig. 1b). Data are representative of (a) or pooled from (b) four independent experiments or are representative of two independent experiments (c).

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

�  aDVaNCE ONLINE PUBLICaTION nature immunology

A rt i c l e s

(Ifnar1−/− mice). Indeed, sorted mature thymocytes from Ifnar1−/− mice showed reduced expression of various genes relative to their expression in Ifnar1+/+ mice (Fig. 6c). The genes most affected were all NF-κB-independent interferon-regulated genes noted above (Fig. 6a), which demonstrated that medullary thymocytes responded to constitutively produced interferon. Among the genes with reduced expression were Stat1 and Irf7, which have been shown to be tar-

gets of constitutive type I interferon signaling, and STAT1 and IRF7 are essential for ‘priming’ cells for cytokine responsiveness32,33. Thus, next we closely assessed the phenotype and function of SP thymocytes from Ifnar1−/− mice. The number of SM, M1 and M2 CD4SP thymocytes was similar in Ifnar1−/− mice and Ifnar1+/+ mice (Fig. 6d). Many other maturation markers were also expressed normally on CD4SP and CD8SP thymocytes from Ifnar1−/− mice

Tak1fl/fl

Tak1fl/flCd4Cre

Tak1fl/flCd4CreIKKCA

Cd4CreIKKCA

1010

109

108

107

106

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

108

107

106

105

104

Tota

l thy

moc

ytes

SM

CD

4SP

cel

ls

M1

CD

4SP

cel

ls

M2

CD

4SP

cel

ls

M1

CD

8SP

cel

ls

M2

CD

8SP

cel

ls

SM

CD

8SP

cel

ls

NSNS

NS NS

NSNS

**

*

***NS***

NS******

Tak1fl/fl Tak1fl/flCd4Cre Tak1fl/flCd4CreIKKCA Cd4CreIKKCA

109

108

108

107

107

106

106

105

Tota

l spl

enoc

ytes

CD

4+ s

plen

ocyt

es

108

107

106

105

104

CD

8+ s

plen

ocyt

es

NS NS NSNS ***

Totalthymocytes

2.5

7.7 87.0 3.2 92.0 2.5 92.7

37.8

30.0 75.2 39.0 26.0

35.5

24.0

35.5 13.2 33.6 30.5

39.723.85.130.8

15.2

1.1

81.3 38.2 31.5

39.1

26.1

38.2

20.2

3.4 3.23.5 0.6

92.1

0.8 1.42.1

Tak1fl/fl Tak1fl/flCd4CreTak1fl/flCd4Cre

IKKCACd4Cre

IKKCA

105

105

104

104

103

103–103

0

105

104

103

0

0

CD

4C

D69

CD4SP

CD8SP

CD8

105104103–1030

MHCI

23.0

11.5 0.9 0.6 4.4

2.2 2.2 9.6

Tak1fl/fl Tak1fl/flCd4CreTak1fl/flCd4Cre

IKKCACd4Cre

IKKCA

Totalsplenocytes

105

105

104

104

103

103–103–103

0

0

CD

4

CD8

9.8 0.5 9.6 8.4

Tak1fl/fl Tak1fl/flCd4CreTak1fl/flCd4Cre

IKKCACd4Cre

IKKCA

CD4SPthymocytes

105

105

104

104

103

103

–102

102

0

0

S1P

R1

CD62L

a b

Eve

nts

SS

C

SM M1 M2

SM M1 M2

Medium CD3+CD28

250K

200K

150K

100K

50K

0

0.6 2.8 11.3

2.40.8 1.1

100

80

60

40

20

0

Tak1fl/fl

Tak1fl/fl

Tak1fl/fl

Cd4Cre

IKKCA

Tak1fl/fl

Cd4Cre

IKKCA

105104103–103 0

CTV

1051041030

TNF

c

d

e

f

g

2.7

Figure 5 IKK activity restores proliferation and maturation but not licensing or survival. (a) Flow cytometry of thymocytes from Tak1fl/fl, Tak1fl/flCd4Cre, Tak1fl/flCd4CreIKKCA and Cd4CreIKKCA mice, showing CD4 and CD8 profiles (top), and proportion of SM, M1 and M2 CD4SP or CD8SP cells (with exclusion of cells as in Fig. 3a) (below). Numbers adjacent to outlined areas as in Figure 3a. (b) Quantification of total thymocytes (top) or SM, M1 and M2 CD4SP or CD8SP thymocytes (below) from Tak1fl/fl mice (n = 10), Tak1fl/flCd4Cre mice (n = 10), Tak1fl/flCd4CreIKKCA mice (n = 6) or Cd4CreIKKCA mice (n = 4). (c) Proliferation of SM, M1 and M2 CD4SP thymocytes sorted from Tak1fl/fl and Tak1fl/flCd4CreIKKCA mice and left unstimulated or stimulated for 3 d with anti-CD3 plus anti-CD28 (assessed as in Fig. 1b). (d) TNF production by SM, M1 and M2 CD4SP cells obtained from Tak1fl/fl and Tak1fl/flCd4CreIKKCA mice and stimulated for 4 h anti-CD3 plus anti-CD28. Numbers adjacent to outlined areas indicate percent TNF+ cells. (e) Flow cytometry of total splenocytes from mice as in a. Numbers adjacent to outlined areas indicate percent CD4+ (CD4SP) T cells (top left) or CD8+ (CD8SP) T cells (bottom right). (f) Quantification of total splenocytes (top) and CD4+ or CD8+ T cells (below) in the spleen of Tak1fl/fl mice (n = 9), Tak1fl/flCd4Cre mice (n = 8), Tak1fl/flCd4CreIKKCA mice (n = 6) or Cd4CreIKKCA mice (n = 4). (g) Flow cytometry of CD4SP thymocytes from mice as in a. Numbers adjacent to outlined areas indicate percent S1PR1+CD62L+ T cells. Each symbol (b,f) represents an individual mouse; small horizontal lines indicate the group mean (± s.d.). *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001 (two-tailed unpaired Student’s t-test). Data are representative of (a,e,g) or pooled from (b) four independent experiments or are representative of (c,d) or pooled from (f) three independent experiments.

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology  aDVaNCE ONLINE PUBLICaTION �

A rt i c l e s

(data not shown). However, CCR7 expression was slightly higher on Ifnar1−/− CD8SP thymocytes than on Ifnar1+/+ CD8SP thymocytes, and Qa2 expression was much lower on M2 CD4SP and CD8SP cells from Ifnar1−/− mice than on those from Ifnar1+/+ mice (Fig. 6e). The response to interferon deficiency was cell intrinsic, as Ifnar1−/− progenitor cells injected intrathymically into wild-type recipient mice still showed substantial Qa2 loss (data not shown). We did not detect functional deficiency in mature thymocytes from Ifnar1−/− mice, in terms of proliferation or licensing (Supplementary Fig. 6). However, Ifnar1−/− cells have deficient responses to IL-6, IFN-γ and the cytokine M-CSF, consistent with their reduced expression of STAT1 and IRF7 (ref. 34). Thus, the thymocytes responded to interferon constitutively during development, and this might have primed their responsiveness to other cytokines.

DISCUSSIONOur data have provided a precise analysis of gene-expression changes that occur after positive selection and how they relate to function. While several gating strategies have been proposed for the demar-cation of functionally relevant stages6,9,11, our data suggested that upregulation of MHC class I expression most precisely marked the stage at which cells acquired the competence to proliferate. Other strategies failed in this context, although they were accurate in marking the stage at which cells became emigration competent. The only other cell-surface protein whose expression change correlated with the acquisition of cell division competence was the immunomod-ulatory receptor GITR. Qa2 is commonly used to mark the most mature thymocytes, but we found that the expression of Qa2 was strongly

dependent on type I interferon signaling in SP cells and was not associated with maturation itself. This is consistent with the observa-tion that Qa2 expression is dependent on the transcriptional regulator AIRE9,35, yet T cells in AIRE-deficient mice are not thought to have major maturational defects.

The observations noted above about Qa2 expression would suggest that IFN-β is produced by medullary thymic epithelial cells in the steady state. Some independent evidence supports this hypothesis30,31, yet further investigation is needed, given that constitutive production of IFN-β can also be driven by the microbiota36–38 or DNA dam-age39,40. Regardless of the source of IFN-β, an important question is how exposure to type I interferon during development changes the functional properties of T cells. There is a growing appreciation that constitutive or tonic interferon signaling maintains homeostasis and primes cytokine responsiveness in other hematopoietic cell types34. Published analysis of Ifnar1−/− mixed–bone marrow chimeras suggests an important role for constitutive interferon signals in T cell develop-ment and Treg cell homeostasis41, so further studies of how interferon alters the gene expression and function of T cells are warranted.

The main finding of our study was that NF-κB signaling was critical for late maturation processes, both for survival at the SP stage and for functional maturation. In terms of survival, it is well established that TNFR signals can trigger both activation of NF-κB and death, through signaling pathways that have been referred to as ‘complex I’ and ‘complex II’42. Signaling via complex I results in the activation of NF-κB and protection against complex II–mediated death. Our results established that TNF was present in the thymic environment and was able to mediate cell death, since TNF deficiency was able to ‘rescue’

100

10

1

0.1Gen

e ex

pres

sion

(fo

ld)

S1pr1

Ms4

a4b

CD62L

Smad

7

Ms4

a4clfi2

03Bcl3

Pla2g1

6Relb

Slfn1

lfi27I

2a lfit1

Xaf1Sta

t1ll2

rbc-

JunOas

3Oas

l2lsg

15

Restored by lKKCA Not restored by lKKCA

Tak1fl/flCd4Cre SM

Tak1fl/flCd4CrelKKCA SM

Tak1fl/fl (WT) M

Tak1fl/flCd4CrelKKCA M

100

10

1

0.1

0.01

Gen

e ex

pres

sion

(fo

ld)

S1pr1

Ms4

a4b

CD62L

Smad

7

Ms4

a4c

lfi203

Bcl3

Pla2g1

6Relb

Slfn1

lfi27l2

alfit

1Xaf

1Sta

t1ll2

rbc-

Jun

Oas3

Oasl2

lsg15

lfnar1+/+ M

lfnar1–/– M

100806040200

Eve

nts

SM M1 M2

STAT1

–103 0 103 104 –105

Tak1fl/fl Tak1fl/flCd4CrelKKCA

a c

b lfnar1+/+

lfnar1–/–

M2 CD4SP M2 CD8SP

500400300200100

0

0

CCR7

Eve

nts

103 104 105

200

150

100

50

0

Eve

nts

Qa2

–103 0 103 104 105

e

Totalthymocytes

105

104

103

0

CD

4

7.1 86.3

3.2 2.6

0 103 104 105

CD8

7.3 86.8

3.1 2.0

lfnar1+/+ lfnar1–/–

CD4SP

32.0

35.8

30.2

42.1

25.9

29.3

CD8SPC

D69

105

104

103

0

0MHCI

103 104 105

31.3

34.3

31.2

43.3

25.0

27.1

d

Figure 6 Interferon-regulated genes and STAT1 levels are not ‘rescued’ by IKKCA, and the interferon receptor IFN-αR is required for Qa2 expression but not for maturation. (a) Expression of mRNA from select TAK1-dependent genes (horizontal axis) in sorted CD24hiCD62Llo (SM) CD4SP thymocytes or CD24intCD62Lhi (mature (M)) CD4SP thymocytes from Tak1fl/fl mice (n = 2), Tak1fl/flCd4Cre mice (n = 2) and Tak1fl/flCd4CreIKKCA mice (n = 2), presented relative to that of Tak1fl/fl SM cells, and grouped by restoration to control expression by the IKKCA transgene (left) or not (right). (b) Expression of STAT1 by SM, M1 and M2 CD4SP thymocytes fromTak1fl/fl and Tak1fl/flCd4CreIKKCA mice. (c) Expression of mRNA from the TAK1-dependent genes in a (horizontal axis) in sorted CD24intCD62Lhi (M) CD4SP thymocytes from Ifnar1++ mice (n = 3) and Ifnar1−/− mice (n = 3), presented relative to that of Ifnar1++ CD24hiCD62Llo (SM) CD4SP thymocytes. (d) Flow cytometry of total thymocytes (top) from Ifnar1++ and Ifnar1−/− mice, showing CD4 and CD8 profiles (top), and the proportion of SM, M1 and M2 CD4SP or CD8SP cells (with exclusion of cells as in Fig. 3a) (below). Numbers adjacent to outlined areas as in Figure 3a. (e) Expression of CCR7 and Qa2 on M2 CD4SP thymocytes from Ifnar1++ mice (n = 3) and Ifnar1−/− mice (n = 3). Data are pooled from (a; mean) or representative of (b) two independent experiments or are pooled from (c; mean) or representative of (d,e) three independent experiments.

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

�  aDVaNCE ONLINE PUBLICaTION nature immunology

A rt i c l e s

the number of mature cells in TAK1-deficient mice. Those findings are consistent with the phenotype of mice deficient in the pro-survival factor c-FLIP43, which interferes with apoptotic signaling down-stream of death receptors. Although TNF was shown to be a source of the death signal, ‘rescue’ was not complete with TNF deficiency. Thus, it is possible that DR3, another death domain–containing member of the TNFR family expressed in medullary thymocytes, contributes as well.

Our data provided definitive evidence that NF-κB signals were required for late maturation. This is consistent with various reports in the literature showing normal positive selection but peripheral T cell lymphopenia in mice deficient in factors such as IKKγ, IKK1 and IKK2, and Ubc13. Subtle differences in the phenotypes might reflect differences in the rate of protein loss after Cre-mediated deletion in the different models (for example, with IKKγ) and/or redundancies with related components of the pathway (for exam-ple, with IKK1 and IKK2). Various receptors can activate the TAK1– NF-κB pathway. The TCR is an obvious candidate, since the groups of genes regulated by NF-κB showed enrichment for genes that changed both at positive selection and into the SM stage. However, the TCR activates TAK1–NF-κB through a Carma1–Bcl-10–Malt1 signalo-some, and mice deficient in those components do not exhibit this phenotype, although they do lack mature Treg cells44–46. We also feel the TCR is unlikely to be the sole source of TAK1 activation in con-ventional T cells because interaction of the TCR with MHC class II is required for survival from the post-DP stage to the SM stage, but it is not required for survival or maturation from the SM stage to mature stages8 (data not shown), although this is controversial47. Thus, there is a distinction between conventional T cells and Treg cells whereby both require activation of NF-κB through TAK1, but conventional T cells seem to receive sufficient NF-κB stimulation without the TCR. Signaling via the TGF-β receptor TGFβR might provide activation of TAK1 in medullary thymocytes. However, TGFβR was not identi-fied as a potential upstream regulator in our analysis of either genes that underwent late changes in expression or TAK1-dependent genes. Furthermore, mice with TGFβR deficiency in T cells do not show a maturation problem per se, although again they have impaired devel-opment of Treg cells and iNKT cells48,49.

We favor the hypothesis that multiple members of the TNFR family can provide TAK1 signals to promote the survival and maturation of developing thymocytes. These include TNFR1 and TNFR2, which bind to TNF. TNFR1 was expressed constitutively in thymocytes, while TNFR2 expression increased notably from the DP stage to the SP stage. Cells from mice with TNF deficiency alone did not have a maturation defect; thus, we propose that OX40, GITR, CD27 and DR3 provide redundant signals, as these receptors are either consti-tutively expressed (CD27) or induced by positive selection (OX40, GITR and DR3). The ligands for some of these (OX40L, GITRL and CD70) are known to be expressed in the thymus, particularly the thymic medulla50. This raises the possibility that thymocytes might need to access the medullary environment to receive signals for matu-ration. Given the data available so far, this seems not to be the case. Thymocytes that lack CCR7 fail to localize to the thymic medulla after positive selection and have defects in central tolerance, but mature and emigrate normally25. Likewise, mice that lack an organized med-ullary environment also have defects in central tolerance, but their thymocytes seem to mature normally11.

In summary, our data have revealed two critical features of T cell maturation that occurred after positive selection but before cells emigrated from the thymus. Activation of NF-κB occurred at the SM stage. NF-κB activity was critical for protecting cells from

complex II–mediated death downstream of TNFR1, as it does in many cell types. However, in thymocytes, NF-κB activity was also critical for maturation processes that allowed the cells to mount a proliferative response when stimulated through the antigen receptor. A second critical feature was that thymocytes responded to constitutively produced type I interferon in the medullary environment. This response was dependent on TAK1 but independent of NF-κB. Constitutive type I interferon signaling resulted in upregulation of the expression of STAT1 and IRF7 and primed T cells to respond to inflammatory cytokines.

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

Accession codes. GEO: microarray data, GSE74078.

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

ACKnoWledgmentSWe thank H. Chi (St. Jude Children’s Hospital) for Rag2GFP mice; M. Mescher (University of Minnesota) for Ifnar1−/− mice; A. Strasser (Walter and Eliza Hall Institute) for Bcl2l11−/− mice; K. Hayakawa (Fox Chase Cancer Center) for monoclonal antibody SM6C10; and M.A. Farrar for comments and review of the manuscript. Supported by the NIH (R01 AI088209 and P01 AI35296 to K.A.H.).

AUtHoR ContRIBUtIonSY.X. designed and performed experiments, analyzed data, and wrote the manuscript; X.W. performed experiments and analyzed data; S.C.J. provided reagents, animals and input for the preparation of the manuscript; and K.A.H. directed the research, 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. Huang, Y.H., Li, D., Winoto, A. & Robey, E.A. Distinct transcriptional programs in thymocytes responding to T cell receptor, Notch, and positive selection signals. Proc. Natl. Acad. Sci. USA 101, 4936–4941 (2004).

2. Mick, V.E., Starr, T.K., McCaughtry, T.M., McNeil, L.K. & Hogquist, K.A. The regulated expression of a diverse set of genes during thymocyte positive selection in vivo. J. Immunol. 173, 5434–5444 (2004).

3. Ueno, T. et al. CCR7 signals are essential for cortex-medulla migration of developing thymocytes. J. Exp. Med. 200, 493–505 (2004).

4. Setoguchi, R. et al. Repression of the transcription factor Th-POK by Runx complexes in cytotoxic T cell development. Science 319, 822–825 (2008).

5. He, X. et al. CD4-CD8 lineage commitment is regulated by a silencer element at the ThPOK transcription-factor locus. Immunity 28, 346–358 (2008).

6. Kishimoto, H. & Sprent, J. Negative selection in the thymus includes semimature T cells. J. Exp. Med. 185, 263–271 (1997).

7. Villunger, A. et al. Negative selection of semimature CD4+8−HSA+ thymocytes requires the BH3-only protein Bim but is independent of death receptor signaling. Proc. Natl. Acad. Sci. USA 101, 7052–7057 (2004).

8. Cowan, J.E. et al. The thymic medulla is required for Foxp3+ regulatory but not conventional CD4+ thymocyte development. J. Exp. Med. 210, 675–681 (2013).

9. Li, J. et al. Developmental pathway of CD4+CD8− medullary thymocytes during mouse ontogeny and its defect in Aire-/- mice. Proc. Natl. Acad. Sci. USA 104, 18175–18180 (2007).

10. McCaughtry, T.M., Wilken, M.S. & Hogquist, K.A. Thymic emigration revisited. J. Exp. Med. 204, 2513–2520 (2007).

11. Cowan, J.E. et al. Differential requirement for CCR4 and CCR7 during the development of innate and adaptive αβT cells in the adult thymus. J. Immunol. 193, 1204–1212 (2014).

12. Boursalian, T.E., Golob, J., Soper, D.M., Cooper, C.J. & Fink, P.J. Continued maturation of thymic emigrants in the periphery. Nat. Immunol. 5, 418–425 (2004).

13. Mingueneau, M. et al. Immunological Genome Consortium. The transcriptional landscape of αβ T cell differentiation. Nat. Immunol. 14, 619–632 (2013).

14. Singer, A., Adoro, S. & Park, J.H. Lineage fate and intense debate: myths, models and mechanisms of CD4- versus CD8-lineage choice. Nat. Rev. Immunol. 8, 788–801 (2008).

15. Matloubian, M. et al. Lymphocyte egress from thymus and peripheral lymphoid organs is dependent on S1P receptor 1. Nature 427, 355–360 (2004).

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology  aDVaNCE ONLINE PUBLICaTION �

A rt i c l e s

16. Arbonés, M.L. et al. Lymphocyte homing and leukocyte rolling and migration are impaired in L-selectin-deficient mice. Immunity 1, 247–260 (1994).

17. Carlson, C.M. et al. Kruppel-like factor 2 regulates thymocyte and T-cell migration. Nature 442, 299–302 (2006).

18. Priyadharshini, B., Welsh, R.M., Greiner, D.L., Gerstein, R.M. & Brehm, M.A. Maturation-dependent licensing of naive T cells for rapid TNF production. PLoS One 5, e15038 (2010).

19. Hoffmann, R. & Melchers, F. A genomic view of lymphocyte development. Curr. Opin. Immunol. 15, 239–245 (2003).

20. Ramsdell, F., Jenkins, M., Dinh, Q. & Fowlkes, B.J. The majority of CD4+8− thymocytes are functionally immature. J. Immunol. 147, 1779–1785 (1991).

21. Aliahmad, P. & Kaye, J. Commitment issues: linking positive selection signals and lineage diversification in the thymus. Immunol. Rev. 209, 253–273 (2006).

22. Carpenter, A.C. & Bosselut, R. Decision checkpoints in the thymus. Nat. Immunol. 11, 666–673 (2010).

23. Hogquist, K.A., Xing, Y., Hsu, F.C. & Shapiro, V.S. T Cell Adolescence: Maturation Events Beyond Positive Selection. J. Immunol. 195, 1351–1357 (2015).

24. Sato, S. et al. TAK1 is indispensable for development of T cells and prevention of colitis by the generation of regulatory T cells. Int. Immunol. 18, 1405–1411 (2006).

25. Kurobe, H. et al. CCR7-dependent cortex-to-medulla migration of positively selected thymocytes is essential for establishing central tolerance. Immunity 24, 165–177 (2006).

26. Wan, Y.Y., Chi, H., Xie, M., Schneider, M.D. & Flavell, R.A. The kinase TAK1 integrates antigen and cytokine receptor signaling for T cell development, survival and function. Nat. Immunol. 7, 851–858 (2006).

27. Liu, Z.G., Hsu, H., Goeddel, D.V. & Karin, M. Dissection of TNF receptor 1 effector functions: JNK activation is not linked to apoptosis while NF-κB activation prevents cell death. Cell 87, 565–576 (1996).

28. Adhikari, A., Xu, M. & Chen, Z.J. Ubiquitin-mediated activation of TAK1 and IKK. Oncogene 26, 3214–3226 (2007).

29. Jimi, E., Strickland, I., Voll, R.E., Long, M. & Ghosh, S. Differential role of the transcription factor NF-κB in selection and survival of CD4+ and CD8+ thymocytes. Immunity 29, 523–537 (2008).

30. Otero, D.C., Baker, D.P. & David, M. IRF7-dependent IFN-β production in response to RANKL promotes medullary thymic epithelial cell development. J. Immunol. 190, 3289–3298 (2013).

31. Lienenklaus, S. et al. Novel reporter mouse reveals constitutive and inflammatory expression of IFN-β in vivo. J. Immunol. 183, 3229–3236 (2009).

32. Gough, D.J. et al. Functional crosstalk between type I and II interferon through the regulated expression of STAT1. PLoS Biol. 8, e1000361 (2010).

33. Hata, N. et al. Constitutive IFN-α/β signal for efficient IFN-α/β gene induction by virus. Biochem. Biophys. Res. Commun. 285, 518–525 (2001).

34. Gough, D.J., Messina, N.L., Clarke, C.J., Johnstone, R.W. & Levy, D.E. Constitutive type I interferon modulates homeostatic balance through tonic signaling. Immunity 36, 166–174 (2012).

35. Dong, J. et al. Homeostatic properties and phenotypic maturation of murine CD4+ pre-thymic emigrants in the thymus. PLoS One 8, e56378 (2013).

36. Abt, M.C. et al. Commensal bacteria calibrate the activation threshold of innate antiviral immunity. Immunity 37, 158–170 (2012).

37. Kawashima, T. et al. Double-stranded RNA of intestinal commensal but not pathogenic bacteria triggers production of protective interferon-β. Immunity 38, 1187–1197 (2013).

38. Ganal, S.C. et al. Priming of natural killer cells by nonmucosal mononuclear phagocytes requires instructive signals from commensal microbiota. Immunity 37, 171–186 (2012).

39. Yu, Q. et al. DNA-damage-induced type I interferon promotes senescence and inhibits stem cell function. Cell Rep. 11, 785–797 (2015).

40. Härtlova, A. et al. DNA damage primes the type I interferon system via the cytosolic DNA sensor STING to promote anti-microbial innate immunity. Immunity 42, 332–343 (2015).

41. Metidji, A. et al. IFN-α/β receptor signaling promotes regulatory T cell development and function under stress conditions. J. Immunol. 194, 4265–4276 (2015).

42. Silke, J. The regulation of TNF signalling: what a tangled web we weave. Curr. Opin. Immunol. 23, 620–626 (2011).

43. Zhang, N. & He, Y.W. An essential role for c-FLIP in the efficient development of mature T lymphocytes. J. Exp. Med. 202, 395–404 (2005).

44. Jost, P.J. et al. Bcl10/Malt1 signaling is essential for TCR-induced NF-κB activation in thymocytes but dispensable for positive or negative selection. J. Immunol. 178, 953–960 (2007).

45. Molinero, L.L. et al. CARMA1 controls an early checkpoint in the thymic development of FoxP3+ regulatory T cells. J. Immunol. 182, 6736–6743 (2009).

46. Barnes, M.J. et al. Commitment to the regulatory T cell lineage requires CARMA1 in the thymus but not in the periphery. PLoS Biol. 7, e51 (2009).

47. Sinclair, C. & Seddon, B. Overlapping and asymmetric functions of TCR signaling during thymic selection of CD4 and CD8 lineages. J. Immunol. 192, 5151–5159 (2014).

48. Chen, W. & Konkel, J.E. Development of thymic Foxp3+ regulatory T cells: TGF-β matters. Eur. J. Immunol. 45, 958–965 (2015).

49. Rubtsov, Y.P. & Rudensky, A.Y. TGFβ signalling in control of T-cell-mediated self-reactivity. Nat. Rev. Immunol. 7, 443–453 (2007).

50. Mahmud, S.A. et al. Costimulation via the tumor-necrosis factor receptor superfamily couples TCR signal strength to the thymic differentiation of regulatory T cells. Nat. Immunol. 15, 473–481 (2014).

©20

16N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

nature immunology doi:10.1038/ni.3419

ONLINE METHODSMice. C57BL/6 mice were purchased from the National Cancer Institute. H2-Ab1−/−, B2m−/−, Tak1fl/fl, Cd4Cre, R26-StopFLIkk2ca (IKKCA) and Tnf−/− mice were obtained from Jackson Laboratories. Bcl2l11−/− and Ifnar1−/− mice were kindly provided by A. Strasser (Walter and Eliza Hall Institute, Melbourne, Australia) and M. Mescher (University of Minnesota, Minneapolis, MN), respectively. Tak1fl/flCd4Cre, Tak1fl/flCd4CreIKKCA, Cd4CreIKKCA, Tak1fl/flCd4CreTnf−/− and Tak1fl/flCd4CreBcl2l11−/− mice were generated by crossbreeding at the University of Minnesota. Rag2GFP and Klf2GFP mice were described10,51. Animals were maintained under specific-pathogen-free conditions at the University of Minnesota. All experimental procedures were approved by the institutional animal care and use committee at the University of Minnesota.

Flow cytometry, purification by magnetic-activated cell separation and cell sorting. Single-cell suspensions were stained for 20 min on ice with the indicated antibodies. Antibodies to the following were purchased from BD Biosciences: CD3 (145-2C11), CD4 (GK1.5), CD8a (53-6.7), CD24/HSA (M1/69), CD25 (PC61), CD27 (LG3A10), CD28 (37.51), CD30 (2SH12-5F-2D), CD40 (3/23), CD44 (IM7), CD62L (MEL-14), CD69 (H1.2F3), CD137/41BB (1AH2), MHCI/H2-Kb (AF6-88.5) and TCRbeta (H57-597). Antibodies to the following were purchased from Biolegend: CD120a/TNFRI (55R-286), CD120b/TNFRII, (TR75-89), CD199/CCR9 (CW-1.2), CD267/TACI (8F10), DR3 (4C12), and Qa2 (695H1-9-9). Antibody to S1PR1 (713412) was purchased from Bio-Techne. Antibody to CD134/OX40 (OX86) was purchased from eBioscience. CD197/CCR7 (4B12, Biolegend) staining was performed for 15 min at 37 °C at 37 °C. Fluorescent CD1d–α-GalCer tetramers preparation was described51. Data were collected using an LSR Fortessa instrument (BD Biosciences) and ana-lyzed using FlowJo (Tree Star). For intracellular Foxp3 (NRRF-30, eBio-science) staining, cells were stained using the eBioscience Foxp3 staining kit and manufacturer’s protocol. To isolate CD4SP subsets, we depleted CD8SP and DP thymocytes via negative selection with biotinylated anti-CD8b.2 (53-5.8, BD Biosciences) and StreptAvidin MACS beads (Miltenyi Biotech) by using MACS separation columns (Miltenyi Biotech) before cell sorting. For CD8SP subsets, we enriched CD8SP thymocytes via positive selection with biotinylated anti-TCRbeta (H57-597) before cell sorting. Cells were sorted on a FACSAria (Becton Dickinson) and results were reliably >90% of the target population.

Microarray analysis. RNA from sorted cells was extracted using RNeasy Mini kit (Qiagen) with on column DNase step (Qiagen) per the manufac-turer’s instructions. RNA was then quantified using a Nanodrop 2000/2000c spectrophotometer (Thermo Scientific). RNA (150–300 ng) was used for generating biotinylated cRNA through single-round amplification using the MessageAmpIII RNA Amplification kit following the manufacturer’s recom-mendations (Ambion). A total of 20 µg of biotinylated cRNA was fragmented and hybridized to Affymetrix murine 430 2.0 gene chips (Affymetrix), and scanned at the Biomedical Genomic Center (University of Minnesota) following standard procedures. Three independent RNA samples were analyzed. The microarray data were analyzed using GeneSpring GX 11 software (Agilent). The normalization was carried out using MAS5 algorithm. Probe sets were filtered by flag values (present/marginal as acceptable flags), and probe sets with P ≤ 0.05 (unpaired t-test) were consid-ered statistical significance, then probe sets showing a difference of twofold or more were considered differentially expressed. Gene-expression patterns

in Figure 2 were generated using Venn Diagram and Profile Plot operations of GeneSpring software.

GSEA and Ingenuity Pathway Analysis. GSEA was performed on microarray data using GSEA software (Broad Institute) per Broad Institute instructions52,53. C3 transcription factor target gene sets contain genes that share a transcription factor binding site defined in the TRANSFAC (ver-sion 7.4, http://www.gene-regulation.com/) database. Each of these gene sets is annotated by a TRANSFAC record. Pathway analysis was performed using Ingenuity Pathway Analysis software (Ingenuity System) according to its instructions.

Quantitative RT-PCR (qPCR). An RNeasy mini kit (Qiagen) and SuperScript III First Strand Synthesis SuperMix for qRT-PCR (Invitrogen) were used for the isolation of RNA and production of cDNA. FastStart Universal SYBR Green Master (Roche) and an ABI PRISM 7900HT sequence detection system (Applied Bioscience) were used for amplification and detection. Hprt (hypox-anthine guanine phosphoribosyl transferase) was used for normalization of samples. Primers were described in Supplementary Table 1.

CellTrace Violet cell proliferation analysis. Sorted cells were labeled with CellTrace Violet (CTV) (Molecular Probes) and cultured in complete medium (RPMI-1640 medium containing 5 mM HEPES pH7.5, 2 mM L-glutamine, 50 µM 2-mercaptoethanol, 50 U/ml of penicillin, 50 µg/ml ptreptomycin, 50 µg/ml gentamicin sulfate and 10% FBS) in a 96-well round bottom plate coated with anti-CD3 (145-2C11, BD Bioscience; 10 µg/ml) and anti-CD28 (37.51, BD Bioscience; 20 µg/ml) antibodies. After 3 days, cells were analyzed by flow cytometry after staining with Fixable Viability Dye eFluor 780 (eBio-science), which was used to exclude dead cells. The intensity of CTV dye is diluted by half for every cellular division.

TNF production assay. Total thymocytes were stimulated with plate-bound anti-CD3 plus anti-CD28 (identified above) in the presence of GolgiPlug (BD Bioscience). 4 h later, cells were treated with Cytofix/Cytoperm (BD Bioscience) for 30 min on ice after surface staining, and then stained for intra-cellular TNF (MP6-XT22, BD Bioscience).

Intracellular staining of STAT1. We stained freshly isolated thymocytes for surface markers (antibodies identified above), then treated them using Cytofix/Cytoperm (BD Biosciences) for 30 min on ice. Intracellular STAT1 was stained with anti–STAT1 rabbit polyclonal antibody (ab92506, Abcam) and subsequently with Alexa Fluor 647–conjugated goat anti-rabbit IgG (Invitrogen). Data were acquired on an LSRFortessa flow cytometer with FACSDiva software and analyzed using FlowJo (Tree Star).

Statistical analysis. s.d. and P values were determined using Prism soft-ware (GraphPad Software, Inc.). P values were calculated using a two-tailed unpaired Student’s t-test with 95% confidence interval.

51. Weinreich, M.A. et al. KLF2 transcription-factor deficiency in T cells results in unrestrained cytokine production and upregulation of bystander chemokine receptors. Immunity 31, 122–130 (2009).

52. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 102, 15545–15550 (2005).

53. Mootha, V.K. et al. PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267–273 (2003).