Themis controls thymocyte selection through regulationof T cell antigen receptor–mediated signaling

Guo Fu1,5, Sebastien Vallee1,5, Vasily Rybakin1, Marielena V McGuire1,4, Jeanette Ampudia1,Claudia Brockmeyer2, Mogjiborahman Salek2, Paul R Fallen1,4, John A H Hoerter1, Anil Munshi1,Yina H Huang1,4, Jianfang Hu1, Howard S Fox3,4, Karsten Sauer1, Oreste Acuto2 & Nicholas R J Gascoigne1

Themis (thymocyte-expressed molecule involved in selection), a member of a family of proteins with unknown functions, is

highly conserved among vertebrates. Here we found that Themis had high expression in thymocytes between the pre–T cell

antigen receptor (pre-TCR) and positive-selection checkpoints and low expression in mature T cells. Themis-deficient thymocytes

showed defective positive selection, which resulted in fewer mature thymocytes. Negative selection was also impaired in

Themis-deficient mice. A greater percentage of Themis-deficient T cells had CD4+CD25+Foxp3+ regulatory and CD62LloCD44hi

memory phenotypes than did wild-type T cells. In support of the idea that Themis is involved in TCR signaling, this protein was

phosphorylated quickly after TCR stimulation and was needed for optimal TCR-driven calcium mobilization and activation of

the kinase Erk.

T cell development in the thymus follows an ordered progression,from the CD4�CD8� double-negative (DN) stage (subcategorized asstages DN1–DN4) through the CD4+CD8+ double-positive (DP) stageto the CD4+CD8� or CD8+CD4� single-positive (SP) stage, thatdepends on a complex interaction of signaling pathways1,2. In DN3cells, variable-diversity-joining (VDJ) rearrangement and expressionof the T cell antigen receptor (TCR) b-chain and its association witha pre-TCR a-chain results in signaling that drives cells through theb-selection checkpoint into the TCRlo DP stage. Here, Va-Ja re-arrangement occurs and a mature ab TCR is expressed on the cellsurface. Sequential rearrangements that delete previous rearrange-ments are often required before a TCR is produced with sufficientability to interact with complexes of self peptide and major histo-compatibility complex (MHC) to induce the positive-selection differ-entiation program that leads to the differentiation of SP thymocytes3.Cells that do not receive this positive selection signal eventually diethrough lack of stimulation, whereas those cells whose TCR binds toostrongly to self peptide–MHC complexes undergo activation-inducedapoptosis referred to as ‘negative selection’1,2.

Signaling through the TCR is tightly regulated during positiveselection. Preselection DP cells express about 10% of the amount ofcell surface TCR found on post-selection TCRhi DP and SP cells,yet preselection DP cells are more sensitive to TCR stimulation byweak ligands than are TCRhi cells4,5. The low TCR expression on

preselection DP cells is actively maintained by signaling by proteintyrosine kinases of the Src family and ubiquitination-mediated degra-dation6,7, and these controls enforce TCR a-chain allelic exclusion8,9.Development through the positive-selection checkpoint can be dis-rupted by mutation of genes encoding elements of the TCR signalingcascade, such as the kinases Zap70, Itk, Lck and Fyn or the Vavguanine nucleotide–exchange factor10,11. The regulation of signalingthrough the mitogen-activated protein kinase Erk is extremely impor-tant in distinguishing between positive and negative selection12–15.Similarly, regulated mobilization of calcium in response to TCRstimulation is related to the discrimination between positive ornegative selection4,16,17.

Here we identify a previously unknown gene and proteininvolved in thymocyte positive and negative selection. Two othergroups have also identified this gene and protein, and in consensuswe refer to the protein as ‘thymocyte-expressed molecule involvedin selection’ (Themis)18,19. We found that it was expressed in atightly regulated way during T cell development. It was expressedmainly in late DN and DP thymocytes and was downregulated afterpositive selection. Themis was tyrosine-phosphorylated withinseconds of TCR stimulation. It is a member of a small family ofunknown function and is highly conserved among vertebrates. Ithas no known conserved domains other than a polyproline region.We show that Themis is important in regulating thymocyte

Received 7 April; accepted 8 June; published online 13 July 2009; doi:10.1038/ni.1766

1Department of Immunology and Microbial Science, The Scripps Research Institute, La Jolla, California, USA. 2T cell Signalling Laboratory, Sir William Dunn School ofPathology, University of Oxford, Oxford, UK. 3Department of Molecular and Integrative Neurosciences, The Scripps Research Institute, La Jolla, California, USA. 4Presentaddresses: United States Army Medical Research and Materiel Command, Fort Detrick, Maryland, USA (M.V.M.); Oxford Gene Technology, Begbroke Science Park, Oxford,UK (P.R.F.); Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri, USA (Y.H.H.); and Department of Pharmacology,University of Nebraska Medical Center, Omaha, Nebraska, USA (H.S.F.). 5These authors contributed equally to this work. Correspondence should be addressed to N.R.J.G.(gascoigne@scripps.edu).

848 VOLUME 10 NUMBER 8 AUGUST 2009 NATURE IMMUNOLOGY

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

development through TCR signaling and in particular through theregulation of calcium influx and phosphorylation of Erk.

RESULTS

Themis encodes a previously unknown protein

We constructed a cDNA subtraction library with cDNA from TCRa-deficient (Tcra�/�) thymus from which genes also expressed inrecombination-activating gene 1–deficient (Rag1�/�) thymus wereremoved20. We further analyzed a clone with differences in mRNAexpression in Tcra�/� versus Rag1�/� thymocytes (Fig. 1). We isolatedfull-length cDNA clones from a mouse thymocyte library and froma human Jurkat T cell library. The main open reading frame of themouse Themis gene encoded a protein of 72.8 kilodaltons. The humanThemis protein was 81% identical (86.4% similarity) to the mouseprotein (Fig. 1a). The gene was present in public databases (GenBank:mouse gene, NM_178666; RIKEN cDNA, E430004N04Rik; human

gene, NM_001010923), and orthologs were present in mammals, aswell as in birds (Gallus gallus) and bony fish (Danio rerio; Supple-mentary Fig. 1).

Although Southern blot analysis showed only a single Themis-hybridizing band in mouse genomic DNA (data not shown), whichsuggested a single-copy gene with no close relatives, BLAST analysis ofthe mouse genome identified two paralogs: BC013712 (which encodesIcb-1)21 and 9130404H23Rik. These encode proteins whose functionsremain uncharacterized. Themis was more closely related to Icb-1(32% identity) than to 9130404H23Rik (25% identity; Supplemen-tary Fig. 2). Mouse Themis is on chromosome 10, location A4; humanTHEMIS is at chromosome 6q22.33 (C6orf190). The sequences ofThemis, Icb-1 and 9130404H23Rik showed no identifiable conserveddomains by PFAM or SMART searches, but there was a highlyconserved proline-rich sequence forming three overlapping PXXP(where ‘X’ is any amino acid) putative Src homology 3 (SH3)

a

d

e

Tap1–/–

PBSTap1–/–

OVATap1–/–

VSV

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

WT

Mm 1 MALSLEEFV SLDLRTLPRVLEIQ GIY EGS YEMFGNECC STGEVIKITGLK KK AEICE I GCES PFELPMNFPGLFK ADKTPYL E Y S F V L I MM GA G QK VM SI Hs 1 MALSLEEFV SLDLRTLPRVLEIQ GIY EGS YEMFGNECC STGEVIKITGLK KK AEICE I GCES PFELPMNFPGLFK ADKTPYL E H A L I F V II Q E LQ IV TM .

Mm EITRT IGPSRLGHPCFYH KDIKLENLIIKQGE I NSVEEI GE V C V RNHQ HSF LPLSQEGEFYECEDE IYTLKEIVEWKIPKNRTRT VN L P RF N TL N G V S T HHs EITRT IGPSRLGHPCFYH KDIKLENLIIKQGE I NSVEEI GE V C V RNHQ HSF LPLSQEGEFYECEDE IYTLKEIVEWKIPKNRTRT IH Q Q ML D IM S A A T N R

Mm V LTDFSNKWDSTNPFP DFYGTLILKPVYEIQGV KF KDI RILPSLDVEVKDITDSYDANWFLQLLST DLFEMTSKEFP V EV E GNHLPQS K E L Q V D V A V ISQHs V LTDFSNKWDSTNPFP DFYGTLILKPVYEIQGV KF KDI RILPSLDVEVKDITDSYDANWFLQLLST DLFEMTSKEFP V EV E GNHLPQS N K M R I E I T I APE

Mm ILQ KTIVIHKKYQASRILASEIRSNFPKRHFLIP SYKGKFKRRPREFPTAYDL IAKS KE LHVVATKAFH H LS VSVGDQFLVH SETTEV RE I Q R T TL KE P HHs ILQ KTIVIHKKYQASRILASEIRSNFPKRHFLIP SYKGKFKRRPREFPTAYDL IAKS KE LHVVATKAFH H LS VSVGDQFLVH SETTEV PG T E E P SP DK S Q

Mm EG KV NVL CEK L K E A LPLYMEGGFVEVIHDKKQY ISELC QF PFNVKV VRDLSI D LAATPGLQLEEDITDSYLLISDFANP VF TR T V N TR D Q Q T CW A KD I E .Hs EG KV NVL CEK L K E A LPLYMEGGFVEVIHDKKQY ISELC QF PFNVKV VRDLSI D LAATPGLQLEEDITDSYLLISDFANP LC IK V A I K SY A L P K RL S EE V T

Mm ECWEIP RLNMTV LV S A L VR VEEITEEQYYMMRRYESS SHPPPRPPKHPS EE KLTLL LAEERT LPKS K HHVD KKL MS R NGS LP DAG LQ SF L A M S IN L S RP PHs ECWEIP RLNMTV LV S A L VR VEEITEEQYYMMRRYESS SHPPPRPPKHPS EE KLTLL LAEERT LPKS K HHVD KKL VG Q SNF RD EPF TL A V T T VD P R IT H ..

Mm G DS G QND D E S GSDES Q RAPV F VA V RQK KH PLQPQAPL .......Hs G DS G QND D E S GPNQA L KVLI S LV E KER NR ATAIAETFKNEKHQK

101100

201200

301300

401400

500500

600598

Exp

ress

ion

(rel

ativ

e)

b

5.7

(kb)

4.3

GAPDH

Liver

Brain

Kidney

Heart

Thym

us

Spleen

Rag1–/

– thym

us

Tcra–/

– thym

us c

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

Exp

ress

ion

(rel

ativ

e)

DN1DN2

DN3DN4

DPlo

DPint

DPhi

CD4+

CD8m

Figure 1 Structure, expression and localization of Themis in wild-type mice. (a) Alignment of predicted amino

acid sequences of mouse (Mm) and human (Hs) Themis proteins; the polyproline region is underlined, and

amino acid identity is presented as white letters on a black background. (b) RNA blot analysis of Themis RNA

in various tissues. GAPDH (encoding glyceraldehyde phosphate dehydrogenase) serves as a loading control.

Results are representative of three independent experiments. (c) Real-time RT-PCR analysis of Themis mRNAexpression by subsets of B6 thymocytes sorted by flow cytometry; results are presented relative to b-actin

expression. DPlo, DPint and DPhi indicate DP thymocytes subcategorized by low, intermediate or high cell surface

TCR expression, respectively; CD8m, mature CD8+ SP. Data are representative of three experiments (error bars,

s.d.). (d) In situ hybridization of mouse thymus with Themis probe. Top, brightfield (left) and darkfield (right)

image showing Themis expression as black dots (left) or white dots (right); bottom, brightfield image of the

corticomedullary junction (cortex, right; medulla, left). Original magnification, �10 (top) or �40 (bottom).

Results are representative of two independent experiments with mice each. (e) Real-time RT-PCR analysis

of Themis expression by thymocytes from wild-type mice (WT) and from OT-I TCR–transgenic Tap1�/� mice

(Tap1�/�) injected intravenously with PBS or with OVA or vesicular stomatitis virus (VSV) peptide; results are presented relative to b-actin expression.

OT-I-transgenic Tap1�/� mice do not have any mature T cells or thymocytes specific for the injected peptide, so they would not be expected to undergo

apoptosis due to cytokine release. Flow cytometry did not show a greater percentage of dead cells in thymocytes from treated mice or any change in

the percentage of DP or SP subsets (data not shown). Data are representative of four experiments (error bars, s.d.).

NATURE IMMUNOLOGY VOLUME 10 NUMBER 8 AUGUST 2009 849

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

domain–recognition motifs near the carboxyl terminus (residues 555–563 (all numbering refers to mouse protein); Fig. 1a). The proline-rich sequence was conserved in Icb-1 and 9130404H23Rik (Supple-mentary Fig. 2). There was a putative bipartite nuclear-localizationsignal between residue 330 and residue 346.

Themis expression in T cell development

We analyzed Themis expression by RNA blot of various tissues(Fig. 1b). Themis was expressed in the thymus and to a lesser extentin the spleen but was not detectable in nonlymphoid tissues. Ofthe two main transcripts, the 5.7-kilobase transcript was the mostabundant. It was readily detectable in several transformed cell linesof thymic origin as well but was undetectable in transformed B celllines (Supplementary Table 1). Reference to the BioGPS geneexpression atlas database22 confirmed its very restricted expressionpattern, in that it was detected in large amounts in thymus, smalleramounts in mature T cells and very small amounts elsewhere.Notably, Icb-1 expression is restricted to B cells, macrophagesand dendritic cells, and 9130404H23Rik is expressed specifically inthe intestine (Supplementary Fig. 3).

Real-time RT-PCR analysis of thymocyte subsets showed that Themisexpression was low in DN1 and DN2 cells and was upregulated in DN3cells; its expression remained high in TCRlo immature preselection

DP thymocytes and was downregulated in post-positive-selectionTCRint and TCRhi cells (Fig. 1c). Themis expression remained low inmature SP subsets. Positive selection is also marked by the migrationof immature thymocytes from the thymus cortex toward the medulla,where only mature SP thymocytes that have survived the selectionprocess are found. In situ hybridization showed that Themis expressionwas detectable mainly in cortical thymocytes, with little expression inmedullary thymocytes (Fig. 1d). Thus, Themis expression is high inimmature cortical thymocytes but low in mature medullary thymocytes.

The data reported above suggested that Themis expression may bedownregulated by stimulation through the ab TCR. To explore thispossibility, we used the OT-I TCR–transgenic mouse, which expressesa TCR that recognizes a peptide derived from ovalbumin (OVA). A setof peptides is available with different affinities for the OT-I TCR, andpositive and negative selection of OT-I TCR–transgenic thymocytesis blocked in mice that lack MHC class I expression, such as micedeficient in the transporter Tap-1 (Tap1�/� mice)15,23–25. We thereforeinjected OT-I Tap1�/� mice (whose thymocytes are blocked at thepreselection DP stage25) with peptides to stimulate the TCR in vivo26.Injection of antigenic OVA peptide resulted in downregulation of Themisexpression, but injection of a nonstimulatory vesicular stomatitis viruscontrol peptide did not (Fig. 1e). Basal Themis expression was lower inTap1-expressing thymocytes than in Tap1-deficient cells, presumably

Themis+/+ Themis–/–

3.8

2.8

90

14.6

3.4

79

c

f

b

49

1.39.5

39

2.1

1.18.4

87

AND Themis+/–

AND Themis–/–

8.1

9.04.6

76

2.0

3.44.0

89

OT-I Themis+/– OT-I Themis–/–

d

e

CD

4

CD8

a

CD8

10.3

4.01.9

83

Themis+/+ Themis–/–

2.1

2.01.0

94

CD

4Themis+/+

Themis–/–

P < 0.05

DN ISP DP CD8+ CD4+

P < 0.01

P < 0.01

Themis+/+, Thy-1.1, Ly5b

Themis–/–, Thy-1.2, Ly5b

CD8

CD8

Themis+/–

CD3

CD

69

1 23

4

5

1 23

4

5

CD8

CD

4

(69.8%) (6.18%) (8.52%) (3.73%)(7.67%)1 2 3 4 5

(0.68%)(79.4%) (4.86%) (2.99%)(7.42%)

CD

4

CD

4

1

10

100

Cel

l (%

)Themis–/–

Figure 2 Defective positive selection in Themis�/� mice. (a) Flow cytometry of thymocytes from Themis+/+ and Themis�/� mice (intercross from a line backcrossed

ten times to the B6 strain); right, frequency of thymocyte subpopulations. Numbers adjacent to outlined areas (left) indicate percent cells in the DN (bottom left),DP (top right), CD8+ SP (bottom right) and CD4+ SP (top left) populations. ISP (right), immature CD8+ cells identified by lack of CD3 expression. P values,

Student’s t-test. Data are representative of one experiment with seven mice (right) or more than three experiments (left). (b) Differences in surface expression

of CD69 and CD3 used to identify thymocyte populations of different maturity in Themis+/� mice (top row) and Themis�/� mice (bottom row). Numbers

adjacent to outlined areas at far left indicate subpopulations gated at right. Parentheses above plots indicate percent thymocytes in each subpopulation. Data are

representative of two independent experiments with two mice of each genotype. (c) Flow cytometry of thymocyte subpopulations from irradiated Thy-1.2 Ly5a mice

reconstituted with a 1:1 mixture of bone marrow from Thy-1.2 Ly5b Themis�/� and Thy-1.1 Ly5b Themis+/+ mice; donor populations were identified by expression

of Thy-1.1 or Thy-1.2 by Ly5b+ cells. Data are representative of three separate experiments with similar results with two chimeric mice per group. (d,e) Flow

cytometry of thymocytes from Themis+/+ and Themis�/� mice expressing the AND TCR27 (d) or the OT-I TCR24 (e). P o 0.0005, CD4+ SP cells (d) and CD8+

SP cells (e) in Themis+/+ versus Themis�/� thymi (t-test). Data are pooled from over three separate experiments each with four (d) or eight (e) Themis+/+ mice

or six (d) or twelve (e) Themis�/� mice. (f) Thymus sections from Themis+/+ and Themis�/� mice, stained with hematoxylin; the medulla shows lighter staining.

Original magnification, �10. Results are representative of two experiments.

850 VOLUME 10 NUMBER 8 AUGUST 2009 NATURE IMMUNOLOGY

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

because of the Themis downregulation induced by positive selection.These data indicate that stimulation through the TCR by peptide-MHCcomplexes induces downregulation of Themis.

Defective positive selection of Themis�/� thymocytes

To identify the function of Themis in T cell development, we generatedThemis-deficient mice. We prepared a construct targeting the firstexon of Themis, replacing it with a neomycin-resistance cassette, andtransfected this construct into 129Sv embryonic stem cells (Supple-mentary Fig. 4). We screened and selected these embryonic stem cellsand injected them into blastocysts by standard methods. One progenywas over 90% chimeric and we bred this mouse. We backcrossedThemis+/� offspring to the C57BL/6 (B6) strain for eight generationsor more and then intercrossed the resultant mice to generate homo-zygous Themis�/� mutants. Disruption of Themis resulted in inter-mediate or complete loss of expression of Themis protein in Themis+/�

or Themis�/� mice, respectively (Supplementary Fig. 4c).Themis-deficient mice were viable, were born at the expected

mendelian frequency and presented no gross abnormalities. Totalthymocyte numbers were similar to those of wild-type mice (Supple-mentary Fig. 5). The proportion and number of DP thymocytes wasslightly but significantly greater, whereas the proportion and numberof mature SP cells, particularly CD4+ SP cells, was lower in Themis�/�

mice than in wild-type mice (Fig. 2a and Supplementary Fig. 5). Toidentify the developmental block in the Themis�/� mice, we sub-categorized thymocytes into five populations on the basis of differ-ences in the expression of CD3 and CD69 and then analyzed theexpression of CD4 and CD8 (Fig. 2b). Population 1, the most immatureTCRloCD69lo cells, was composed mostly of DN and DP cells. Thispopulation was present in similar proportions in the Themis�/� andThemis+/+ mice. Population 2, which had a TCRintCD69lo phenotypeand corresponded to preselection DP cells, was present in a slightlyhigher proportion in Themis�/� mice, as expected from the largerpercentage of DP cells in these mice (Fig. 2a). Population 3 partiallyupregulated CD69 and increased TCR expression relative to popula-tion 2, and was present in a slightly lower proportion in Themis�/�

mice. Themis-deficient thymi showed considerable depletion ofpopulation 4 (TCRhiCD69hi; post-positive-selection thymocytes). Inwild-type thymi, most of this population was CD4+ SP or CD8+ SP,whereas in Themis-deficient mice, the cells were mostly CD4+CD8int

or CD8+ SP. Finally, Themis�/� thymi contained very few TCRhiCD69lo

cells of population 5, which corresponded to mature SP cells readyfor export to the periphery. From these data, it is clear that theThemis-deficient cells were mostly blocked at the earliest stage ofpositive selection during which CD8 downregulation is initiated andthat very few Themis-deficient thymocytes developed into a fullymature phenotype.

To establish if Themis acts in a T cell–intrinsic way, we did a bonemarrow–reconstitution experiment (Fig. 2c). We injected bone mar-row cells from Themis+/+ Thy-1.1 Ly5b mice and Themis�/� Thy-1.2Ly5b mice into irradiated Thy-1.2 Ly5a recipient mice and allowedthem to reconstitute the thymus for 6 weeks. We gated Ly5b-expressingdonor cells and then separated them into Thy-1.1+ (Themis+/+) andThy-1.2+ (Themis�/�) populations. This flow cytometry demonstratedthat the defect in development from DP to SP was intrinsic toThemis�/� thymocytes, as Themis�/� thymocytes had lower propor-tions of CD4+ SP cells and greater proportions of DP cells than didThemis+/+ cells developing in the same environment.

We next analyzed the effect of Themis deficiency on thymocytedevelopment in mice expressing the MHC class II–restricted ANDTCR transgene27 or the MHC class I–restricted OT-I TCR transgene24

(Fig. 2d,e). The defect in thymocyte development was greater in TCR-transgenic Themis-deficient mice than in nontransgenic polyclonalThemis-deficient mice. In AND TCR–transgenic thymi, developmentwas blocked at the DP stage, with very few cells proceeding tothe CD4+ SP stage (Fig. 2d). The maturation arrest in the ANDThemis�/� thymus first manifested at the start of the transition fromCD4hiCD8hi to CD4hiCD8int with little or no progression to themature SP stage (Supplementary Fig. 6a). In OT-I thymi, developingMHC class I–restricted cells pass through the CD4hiCD8int stagebefore upregulating CD8 and downregulating CD4 (ref. 28). Theconsiderable decrease at this stage in OT-I Themis�/� thymi (Fig. 2eand Supplementary Fig. 6b) suggests that like AND TCR–transgenicand nontransgenic mice, OT-I Themis�/� mice have a considerabledefect at the earliest stage of positive selection. The architecture of thethymus was altered in Themis�/� mice, with the medulla being smallerand more fragmented than wild-type medulla (Fig. 2f); this defect isindicative of impaired T cell maturation29.

Involvement of Themis in negative selection

TCR stimulation of DP cells and SP cells also induces negativeselection. Thus, we investigated the function of Themis in nega-tive selection in a ‘superantigen’-mediated deletion model. B6- andB10.D2-derived mouse strains express the superantigens Mtv-8 andMtv-9, which cause deletion of Vb5+, Vb11+ and Vb12+ TCR clono-typic thymocytes when the MHC molecule I-E is also present30,31. Tointroduce I-E, which is absent from B6-backcrossed Themis�/� mice,we bred those mice to B10.D2 mice, which express I-E. Comparisonof T cell populations in H-2b/b Themis+/� thymi (which lack I-E) andH-2b/d Themis+/� thymi (which express I-E) showed superantigen-mediated deletion of T cells expressing Vb5, Vb11, and Vb12 (Fig. 3).Vb6+ thymocytes (and Vb7+ thymocytes; data not shown) do notrecognize Mtv-8 and Mtv-9 and were not deleted. In fact, theirproportion was slightly higher in mice in which deletion of theother clonotypes had occurred. Superantigen-mediated deletion wassignificantly less efficient in Themis�/� mice than in Themis+/� H-2b/d

mice, although it was clearly partially active.We also analyzed thymocyte development in mice expressing the

MHC class I–restricted TCR transgene HY. Here, positive selection infemales leads to a large CD8+ SP population, but negative selectionpredominates in males, resulting in few DP cells or SP cells32,33.HY TCR–transgenic females lacking Themis resembled AND TCR–transgenic and OT-I TCR–transgenic Themis�/� mice in that few

CD

8+ T

cel

ls (

%)

CD

4+ T

cel

ls (

%)

H-2b/b W

T

Vβ6 Vβ5, Vβ11, Vβ12

H-2b/d W

T

H-2b/b KO

H-2b/d KO

P < 0.001P < 0.001

P < 0.001

P < 0.001

P < 0.001

P < 0.001

0

2

4

6

8

10

12

0

5

10

15

20

H-2b/b W

T

H-2b/d W

T

H-2b/b KO

H-2b/d KO

H-2b/b W

T

Vβ6 Vβ5, Vβ11, Vβ12

H-2b/d W

T

H-2b/b KO

H-2b/d KO

H-2b/b W

T

H-2b/d W

T

H-2b/b KO

H-2b/d KO

Figure 3 Negative selection defect in Themis�/� mice. Flow cytometry of the

expression of various Vb elements (below graphs) in CD4+ or CD8+ peripheral

blood T cells from B6 mice (H-2b) bearing wild-type (WT) or mutated (KO)

Themis alleles, backcrossed (H-2b/d) or not (H-2b/b) onto the B10.D2 (H-2d)

strain. Vb5, Vb11 and Vb12 are deleted by the Mtv-8 and Mtv-9

superantigens expressed in B6 and B10.D2 mice but only when I-E is also

expressed; thus, deletion occurs on H-2b/d but not H-2b/b backgrounds. Vb6

is not deleted by these Mtv superantigens. Each symbol represents a single

mouse; small horizontal lines indicate the mean. P values, Student’s t-test.

Data are representative of three independent experiments.

NATURE IMMUNOLOGY VOLUME 10 NUMBER 8 AUGUST 2009 851

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

thymocytes progressed to become mature CD8+ SP cells (Supplemen-tary Fig. 7a). Male HY TCR–transgenic Themis+/� mice had consider-ably fewer thymocytes than did female HY TCR–transgenic mice, anda large proportion of these cells were HY-TCR+ clonotypic cells withlow expression of CD4 and CD8 (ref. 34). These cells are believedto escape negative selection by loss of the coreceptor. In thymi frommale HY TCR–transgenic Themis�/� mice, negative selection wasapparently incomplete, in that expression of CD4 and CD8 wasmuch lower than normal but not as low as in the male Themis+/�

mice, and the clonotype-negative cells found in the male Themis+/�

mice were not present (Supplementary Fig. 7b). Thus, both negativeand positive selection were lower in the absence of Themis, whichsuggests that absence of Themis leads to a general defect in TCR-mediated signaling.

Peripheral T cells in Themis�/� mice

The deficiency in SP thymocytes led to fewer peripheral T cells inThemis�/� mice (Fig. 4a,b). The percentage of B cells was higherbecause of the loss of T cells (data not shown). The deficiency inperipheral T cells was most evident in the CD4+ population, but bothCD4+ and CD8+ cells were significantly fewer in number. Analysis ofTCR-transgenic mice confirmed those data, showing very few matureperipheral CD4+ cells in AND TCR–transgenic Themis�/� mice

(Fig. 4c) and few CD8+ cells in OT-I TCR–transgenic Themis�/�

mice (Fig. 4d). In the CD4+ population, cells with a regulatory T cellphenotype (CD25+Foxp3+) were greater in frequency in Themis�/�

mice (21%) than in Themis+/+ mice (10%; Fig. 4e), although theabsolute number of Themis�/� regulatory T cells was about 50% thenumber in Themis+/+ mice (data not shown).

The expression of CD62L and CD44 in peripheral T cells is used todistinguish naive T cells and memory T cells. Themis�/� mice had alower percentage of CD62LhiCD44lo naive-phenotype cells and ahigher proportion of CD62LloCD44hi memory-phenotype cells(Fig. 4f). Expression of CD122+ (interleukin 2 receptor-b) was alsohigher in Themis�/� mice (data not shown). Thus, the few peripheralcells in Themis-deficient mice have a phenotype reminiscent ofmemory cells, a frequent observation in partially lymphopenic micethat probably reflects homeostatic expansion of T cell populations inthe periphery35.

We then tested the ability of Themis�/� spleen T cells to respond tostimulation with antibody to CD3 (anti-CD3; Fig. 5). CD69 upregu-lation was much lower in Themis-deficient CD4+ T cells and CD8+

T cells than in wild-type T cells. The proliferation of both CD4+

Themis�/� T cells and CD8+ Themis�/� T cells in response toanti-CD3 was very poor compared with that of Themis+/+ T cells.Themis+/+ and Themis�/� CD8+ T cells had similar surface expressionof TCR, but TCR expression was slightly lower on the surface ofThemis�/� CD4+ cells (Fig. 5a).

Themis in early TCR signaling

We investigated the function of Themis in the TCR signaling cascadeby immunoprecipitation and immunoblot analysis (Fig. 6). Initially, aproteomic study of the global dynamics of TCR-directed tyrosinephosphorylation by stable isotopic labeling of amino acids in cellculture identified many previously unknown potential signaling com-ponents in Jurkat cells, including C6orf190 (C.B., O.A. and M.S.,unpublished data), called SPoT (for ‘signaling phosphoprotein specificfor T cells’; called ‘Themis’ here). Decreasing Themis expression throughthe use of small interfering RNA in Jurkat cells resulted in muchlower production of interleukin 2 induced by stimulation with anti-TCR and anti-CD28 (C.B., O.A. and M.S., unpublished data). Thedata obtained by stable isotopic labeling of amino acids in cell culture

b c d16.3

10.2

6.1

6.8

20.7

4.0

1.8

3.5

AND Themis+/–

AND Themis–/–

3.3

6.5

1.0

1.9

24 3.9

7.964.2

7.7 4.3

29.558.5

32.6 0.8

5.661

3.0 1.4

34.461.3

f

CD

62L

CD44

OT-I Themis+/–

OT-I Themis–/–

e

WT KO WT KO

0

10

20

30

40

50

0

10

20

30

40

50

CD62LloCD44hiCD62LhiCD44lo

WT KO WT KO

CD62LloCD44hiCD62LhiCD44lo

P < 0.001P < 0.001

P < 0.001

P < 0.001

Themis+/+

Themis–/–

CD4+ WT

CD4+ WT

CD8+ KO

CD8+ KO

CD

4+ c

ells

(%

)C

D8+

cel

ls (

%)

CD

4

CD8

CD

4

CD8

26.2

9.4Themis+/+

Themis–/–

CD

25

Foxp3

0

8

4

12

16

20

Cel

ls (

%)

CD4+ CD8+

Themis+/+

Themis–/–

P < 0.01

12

10

8

6

4

2

0777063564942352821147

CD4+ Themis–/–CD4+ Themis+/+CD8+ Themis–/–CD8+ Themis+/+

P < 0.01

a

Cel

ls (

×06 )

Time (d)

Figure 4 Phenotype of peripheral T cells in Themis�/� mice. (a,b) Quantification

by flow cytometry of CD4+ and CD8+ T cells in the spleens of B6 Themis+/+

and Themis�/� mice of various ages (a, left; horizontal axis) or 6-weeks-old

(a, right, and b). (a) Right, average frequency of each population. Left, each

dot represents an individual mouse; small horizontal lines indicate the mean.

Data are representative of over three experiments (a, left; b) or one experiment

with seven mice (a, right; error bars, s.d.). (c,d) Expression of CD4 and

CD8 on spleen cells from mice expressing the AND TCR transgene (c) orOT-I TCR transgene (d). P o 0.0009, AND Themis+/+ versus AND Themis�/�;

P o 0.003, OT-I Themis+/+ versus OT-I Themis�/� (Student’s t-test). Data

are representative of at least three separate experiments with four AND mice

of each genotype, seven OT-I Themis�/� mice or five OT-I Themis+/+ mice.

Numbers in quadrants (b–d) indicate percent CD4+CD8� cells (top left)

or CD4�CD8+ cells (bottom right). (e) Expression of CD25 and Foxp3

by Themis�/� and Themis+/+ splenocytes, gated on CD4+ cells.

Numbers adjacent to outlined areas indicate percent CD25+Foxp3+ cells.

P ¼ 0.000028. Data are representative of two independent experiments

with four mice of each genotype. (f) Expression of CD62L and CD44 on

lymph node T cells from Themis+/+ mice (WT) and Themis�/� mice (KO),

with gating on CD4+ or CD8+ cells (left), and frequency of CD4+ or CD8+

cells in the CD62LhiCD44lo and CD62LloCD44hi subpopulations (right).

Numbers in quadrants (left) indicate percent cells in each. Each symbol

(right) represents an individual mouse; small horizontal lines indicate

the mean. Data are representative of three independent experiments.

852 VOLUME 10 NUMBER 8 AUGUST 2009 NATURE IMMUNOLOGY

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

suggested that Themis was phosphorylated on tyrosine residues asearly as 30 s after TCR stimulation and that this phosphorylation waslower at 5 min after TCR stimulation; these phosphorylation kineticswere similar to those of other known TCR-proximal signaling proteins(such as Lat and SLP-76; data not shown).

To confirm the data reported above, we stimulated Jurkat cells withanti-CD3, immunoprecipitated cell lysates with antibody to phos-phorylated tyrosine and incubated immunoprecipitates with anti-serum to the carboxy-terminal sequence of human Themis. Thistreatment identified a phosphorylated protein of the expected sizefor Themis (B72 kilodaltons; Fig. 6a) that increased in abundanceafter 30 s of stimulation, then decreased. Immunoprecipitation withanti-Themis and immunoblot analysis with antibody to phosphoryl-ated tyrosine provided direct evidence that Themis was itselfphosphorylated on a tyrosine residue, which indicated that it isa TCR-dependent protein tyrosine kinase substrate. We obtained

similar data with human CD4+ peripheral blood T cells (Fig. 6b). Inmouse thymocytes, Themis was tyrosine-phosphorylated within30 s of TCR crosslinking, and phosphorylation was undetectable after3 min (Fig. 6c). Together, these data show that in mouse andhuman cells, Themis is an early target of TCR-controlled proteintyrosine kinases, and they suggest that it is a component of the TCRsignaling pathway whose function and/or recruitment may depend ontyrosine phosphorylation.

Defective TCR signaling in Themis�/� thymocytes

Because the proline-rich sequence in Themis and related proteinswas predicted to be a binding site for SH3 domains, we testedan SH3 domain array to screen for binding to the amino-terminalhalf and carboxy-terminal half (including the proline-rich sequence)of Themis. The proline-rich sequence–containing region showed sub-stantial binding to the SH3 domain of phospholipase C-g1 (PLC-g1;

CD8+

CD4+

2 3 4 50

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

CD69

CFSE

c

bα-CD3 Med α-CD3

Med α-CD3Med α-CD3

71.744.1

5.25.3

64.839.9

0

20

40

60

80

100 Med

6.110.4

0

20

40

60

80

100

Cel

ls (

% m

ax)

Cel

ls (

% m

ax)

0

20

40

60

80

100

a

CD3

CD3

CD8+CD4+

CD8+CD4+

Cel

ls (

% m

ax)

Cel

ls (

% m

ax)

b

US 2 10

IB: Themis

IB: p-Tyr

IB: Themis

α-CD3 (min)

dWCL IP: Themis

IB: PLC-γ1

IB: ThemisUS 2 10 30 US 2 10 30

α-CD3 (min) α-CD3 (min)

Themis+/+

a

US 0.5 10

IB: Themis

IB: p-Tyr

IB: Themis

IP: Themis

α-CD3 (min)

IP: p-Tyr

e Themis–/–Themis+/+IP: Itk

IB: PLC-γ1

IB: Themis

IB: Itk

α-CD3 (min) α-CD3 (min)US 2 10 30 US 2 10 30

f

IP: Itk

WCL

α-CD3 + α-CD4 (min)

IB: Itk

IB: Itk

IB: PLC-γ1

IB: Themis

US 0.5 1.5 3

c

α-CD3 +α-CD4 (min)

α-CD3 +α-CD4 (min)

Themis–/–Themis+/+

IB: Themis

WCL

IP: p-TyrIB: Themis

US0.

51.

5 3 10 US0.

5 1.5 3 10

IB: p-Zap70

IB: ItkIP: p-Tyr

IB: p-PLC-γ1

IB: p-p42/44

IB: p42/44

α-CD3 +α-CD4 (min)

α-CD3 +α-CD4 (min)

gThemis–/–Themis+/+

US 0.5

1.5 3 10 US 0.

51.

5 3 10

WCL

WCL

WCL

WCL

Figure 6 Themis is part of the TCR signal cascade. (a,b) Immunoblot (IB) analysis of Themis and phosphorylated tyrosine (p-Tyr) in lysates of Jurkat cells (a)

or fresh human peripheral blood T cells (b) left unstimulated (US) or stimulated with anti-CD3 for various times (below plots), then immunoprecipitated (IP)

with anti-Themis serum or monoclonal antibody to phosphorylated tyrosine. (c) Immunoblot analysis of Themis (arrows, right margin) in ‘rested’ Themis+/+

and Themis�/� thymocytes left unstimulated or stimulated with anti-CD3 plus anti-CD4, detected in whole-cell lysates (WCL) and after immunoprecipitation

with antibody to phosphorylated tyrosine. (d) Immunoblot analysis of PLC-g1 and Themis in freshly isolated B6 thymocytes left unstimulated or activated

with anti-CD3, lysed and immunoprecipitated with anti-Themis. (e) Immunoblot analysis of PLC-g1, Themis and Itk in freshly isolated Themis+/+ or Themis�/�

thymocytes left unstimulated or stimulated with anti-CD3, lysed and immunoprecipitated with anti-Itk. In some experiments, samples were depleted ofmature thymocytes by magnetic-activated cell sorting with anti-CD53 and cells were stimulated with anti-CD3 plus anti-CD4, with similar results (data not

shown). (f) Immunoblot analysis of PLC-g1, Themis and Itk in lysates of ‘rested’ B6 thymocytes left unstimulated or activated with anti-CD3 plus anti-CD4,

detected in whole-cell lysates and after immunoprecipitation with anti-Itk. (g) Immunoblot analysis of phosphorylated (p-) Zap70, Itk, phosphorylated PLC-g1and phosphorylated and total Erk1/2 (p-p42/44 and p42/44) in lysates of ‘rested’ Themis+/+ and Themis�/� thymocytes left unstimulated or stimulated with

anti-CD3 plus anti-CD4, detected in whole-cell lysates and after immunoprecipitation with antibody to phosphorylated tyrosine. Results are representative of

five (a), two (b,d,f) or three (c,e,g) experiments.

Figure 5 Defective activation of Themis-deficient

T cells. (a) CD3 expression on gated CD4+

(CD25�) and CD8+ Themis�/� (blue) and

Themis+/+ (red) spleen cells. (b) CD69

upregulation on Themis�/� (blue) and Themis+/+

(red) spleen cells left unstimulated (Med) or

activated by crosslinking for 5 h with anti-CD3

(a-CD3), analyzed on gated CD4+ or CD8+ cells.Numbers above and below bracketed lines indi-

cate percent CD69+ cells. (c) Flow cytometry of

dilution of the cytosolic dye CFSE by Themis�/�

(blue) and Themis+/+ (red) spleen cells labeled

with CFSE and left unstimulated or stimulated

in vitro for 3 d with anti-CD3, analyzed on gated

CD4+ or CD8+ cells. Data are representative of

at least three independent experiments.

NATURE IMMUNOLOGY VOLUME 10 NUMBER 8 AUGUST 2009 853

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

data not shown). To confirm a Themis–PLC-g1 interaction in thymo-cytes, we stimulated B6 thymocytes with anti-CD3, then immuno-precipitated cell lysates with Themis-specific antiserum. Althoughthere was some interaction between Themis and PLC-g1 beforestimulation, the amount of PLC-g1 that precipitated together withThemis was higher at 2 min after stimulation, but this ultimatelyreturned to a weaker interaction (Fig. 6d). As Itk is a tyrosine kinaseof the Tek family that is required for activation of PLC-g1 and forTCR signaling36, we tested the possibility of an interaction betweenItk and Themis. Anti-Itk precipitated together with both Themisand PLC-g1 from freshly isolated, anti-CD3-stimulated thymocytes(Fig. 6e). It seemed that that Itk-Themis and Itk–PLC-g1 interactionsincreased after CD3 crosslinking. To test that, we allowed thymocytesto ‘rest’ for 2 h in culture before activating them with anti-CD3plus anti-CD4. As reported before37, PLC-g1 interacted constitutivelywith Itk in thymocytes; after stimulation, the interaction increasedslightly before decreasing after 30 s (Fig. 6f). Interaction of Themiswith Itk was not constitutive but was induced within 30 s ofTCR stimulation.

To determine which signaling pathways are defective in Themis�/�

thymocytes, we analyzed immunoblots of Themis+/+ and Themis�/�

thymocytes activated with anti-CD3 plus anti-CD4 (Fig. 6g). We foundno difference in the phosphorylation of Zap70, PLC-g1 or Itk (or Lat;data not shown) in Themis+/+ and Themis�/� thymocytes in responseto TCR stimulation. However, Erk showed less phosphorylation inThemis�/� cells. Erk1 (p44) phosphorylation was delayed in Themis�/�

compared with its phosphorylation in Themis+/+ cells, and Erk2 (p42)phosphorylation was lower in Themis�/� cells at 30 and 90 s after

stimulation (Fig. 6g and Supplementary Fig. 8a) and was not alwaysdetectable at 3 min. We did this experiment three times with similarresults (Supplementary Fig. 8b). Phosphorylation of the kinases Raf,Mek and Jnk was not affected by Themis deficiency (data not shown).

To investigate other aspects of TCR signaling, we assessed preselec-tion DP thymocytes from OT-I TCR–transgenic Tap1�/� mice.Stimulation with anti-CD3 plus anti-CD4 or with OVA–H-2Kb

tetramers (Fig. 7a) induced a slightly slower and smaller initial peakof store-operated Ca2+ influx in Themis�/� cells relative to Themis+/+

cells. This difference in Ca2+ influx, although relatively subtle, wasvery reproducible. The release of endoplasmic reticulum Ca2+ storeswas very low in the thymocytes, and we found no difference betweenThemis�/� cells and Themis+/+ cells.

Organization of the cytoskeleton at the immunological synapse isregulated by TCR signaling38. We therefore imaged actin polymeriza-tion in OT-I Tap1�/� thymocytes expressing or lacking Themis as theyinteracted with OVA peptide presented by EL4 mouse lymphoma cells(Fig. 7b). The actin-polymerization pattern in Themis-sufficientthymocytes was similar to that in OT-I lymph node T cells, butThemis-deficient thymocytes showed a considerable defect in actinpolymerization at the immunological synapse. A useful correlate ofstimulation of thymocytes through the TCR by positive or negativeselecting ligands is CD69 upregulation15,25,39,40. We found that theOT-I Themis�/�Tap1�/� preselection DP thymocytes were less able toupregulate CD69 in response to OVA peptide than were their OT-IThemis+/+Tap1�/� counterparts (Fig. 7c). These data suggest thatThemis is involved in regulating the TCR signaling cascade, albeitwith relatively small effects on early events in the cascade.

DISCUSSION

The development of ab T cells is critically dependent on signalingthrough the TCR and can be disrupted by deficiencies in the receptoritself and in many signaling molecules and mediators. In this work, wehave identified a previously unknown protein, Themis, that was requiredfor passage through the positive-selection checkpoint. Themis deficiency

0

10

20

30

40

50

60

70

0 100 101 102 103 104 105 106 0 100 101 102 103 104 105 106

OVA (pM)

Themis+/+

Themis–/–Themis+/+

Themis–/–

a

0

300

600

900

1,200

1,500

OVA (pM)

b

c

CD

69+ (

%)

CD

69 (

geo

mea

n)

LN Tap1–/–Themis+/+

LN Tap1–/–

Themis+/+Tap1–/–

Themis–/–LN

Tap1–/

–

Them

is+/

+

Tap1–/

–

Them

is–/

–

Tap1–/

–

Them

is+/

+

Tap1–/

–

Them

is–/

–LN

0 100 200 300 400 500Time (s)

00.30.60.91.2

0 100 200 300 400 500Time (s)

0

0.5

1.0

1.5

0 100 200 300 400 500Time (s)

00.51.01.5

Indo

-1 fl

uore

scen

ce(v

iole

t/blu

e)

WTKO1KO2

WTKO1KO2

WTWTWT

SAv CaCl2 SAv CaCl2K

b-OVA

tetramer CaCl2

Tap1–/–Themis–/–

Sig

nal/b

ackg

roun

d

Syn

apse

/mem

bran

e

Membrane Synapse12 5

4

3

2

1

0

10

8

6

4

2

0

*

Figure 7 Signaling in Themis-deficient thymocytes. (a) Store-operated Ca2+

influx in preselection DP Themis+/+ or Themis�/� OT-I Tap1�/� thymocytes

stimulated by crosslinking with biotinylated anti-CD3 and anti-CD4 with

streptavidin (SAv; left and center) or with OVA–H-2Kb tetramer (Kb-OVA;

right) in Ca2+-free medium, followed by the addition of CaCl2; cells were

labeled with CFSE, indodicarbocyanine or no label, then cells were mixed,

incubated with the calcium indicator Indo-1 AM and assayed in one

tube to test Themis+/+ or Themis�/� cells at the same time in the sameenvironment. Ca2+ flux was measured by changes in the violet/blue

fluorescence ratio of Indo-1 AM. Left, three samples of Themis+/+ cells

tested against each other; center and right, Themis+/+ cells tested

against cells from two different Themis�/� mice (KO1 and KO2). Data

are representative of three independent experiments with a total of nine

Themis�/� mice and five Themis+/+. (b) Actin polymerization in lymph

node–derived naive OT-I T cells (LN; top left) or preselection DP thymocytes

(top, center and right) incubated with EL4 cells (large central cell in each)

loaded with OVA peptide and stained for F-actin with fluorescence-labeled

phalloidin. Original magnification, �100. Below, F-actin in the immunological

synapse and the rest of the T cell membrane, calculated from imaging

data above: left, actin signal at the membrane within or outside the synapse

(boxes, s.e.m.; crossbar, median; plus symbol, mean; bars, 5th and 95th

percentile range; dots, data points outside this range); right, ratio of F-actin

in the synapse to that in the nonsynapse membrane (each circle represents

a single data point; long horizontal lines lines, mean; short horizontal lines

at ends, s.e.m.). *P o 0.05, Tap1�/�Themis�/� versus Tap1�/�Themis+/+

(Student’s t-test). Data are representative of two experiments. (c) CD69

upregulation in preselection DP OT-I Tap1�/� Themis+/+ or Themis�/�

thymocytes stimulated in vitro with OVA presented by EL4 cells39,40.

Data are representative of three experiments.

854 VOLUME 10 NUMBER 8 AUGUST 2009 NATURE IMMUNOLOGY

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

resulted in a defect in development that was greater for CD4+ cells thanfor CD8+ cells. However, this defect was evident in both MHC class I–and MHC class II–restricted TCR-transgenic lines in which thymocytedevelopment was blocked at the earliest stages of positive selection.A more pronounced phenotype in TCR-transgenic thymocytes is acommon finding in mice deficient in signaling molecules, as the normalTCR repertoire is plastic enough to select for TCRs with higher or loweraffinity, which allows the signal strength to be modulated to permitselection10. The substantial developmental block in the TCR-transgenicThemis�/� mice, together with the evidence indicating that negative-selection signaling was also deficient in Themis�/� mice, suggestedinvolvement of Themis in the TCR signaling cascade rather than in adevelopmental program. Because Themis is expressed in immaturethymocytes and is downregulated by stimulation through the TCR, asoccurs during positive selection, the signaling function of Themis islikely to be most important in DP thymocytes before or during positiveselection. In agreement with that idea, Themis�/� preselection thymo-cytes responded poorly to TCR stimulation compared with Themis+/+

thymocytes, as measured by CD69 upregulation. Themis�/� thymocytesalso had less actin polymerization at the immunological synapse, as inItk-deficient and Vav-deficient cells. These results are consistent with adefect in TCR signal transduction39–42.

Themis was tyrosine-phosphorylated within 30 s of TCR stimula-tion, which suggests that it may be involved early in signal transductionthrough the TCR. Moreover, it interacted in an inducible way with Itkand PLC-g1, although phosphorylation of these proteins was notaltered in Themis�/� cells. Deletion of Itk has a profound effect oncalcium signaling but a relatively minor effect on thymocyte develop-ment36,41. In contrast, Themis deficiency had a substantial effect ondevelopment but a small effect on calcium signals (for example,slightly slower and weaker Ca2+ flux). The effect of loss of Themison Erk signaling was also subtle and manifested as a lower strengthand duration of phosphorylation. These findings suggest that Themismight negatively regulate phosphatases that turn off Erk signaling.Notably, similarly small differences in the induction of phosphorylatedErk distinguish signals that induce positive and negative selection ofthymocytes14,15, and sustained Erk signals are required for positiveselection2,12,14. Additionally, there is evidence that slight changes incalcium signaling lead to profound differences in the activation ofmitogen-activated protein kinases in non–T cells43 and in preselectionthymocytes and that these differences distinguish the quality of TCRstimulation17. The differences in calcium and Erk signaling inThemis�/� versus Themis+/+ preselection thymocytes suggest thatThemis may be involved in setting the threshold for TCR responsesto selecting signals. Such a function for Themis could account for itsexpression mainly in pre-positive-selection thymocytes and wouldindicate that its function in TCR-mediated signaling may not be soimportant in the activation of peripheral T cells. Study of the normalfunction of Themis in peripheral T cell signaling must await thegeneration of mice with conditional deletion of Themis.

When given the opportunity to produce a full TCR repertoire,some Themis-deficient thymocytes passed through the positive-selectioncheckpoint and generated peripheral T cells. These cells respondedpoorly to TCR-mediated stimulation. Notably, in both CD4+ andCD8+ subsets, there was a greater prevalence of cells with a memoryphenotype (CD44hiCD62LloCD122+) and lower proportions of popu-lations with a naive phenotype. This may have been the result ofhomeostatic population expansion in a lymphopenic environment35

and is similar to the phenotype of mice that lack Itk. Memory phenotypeT cells in Itk-deficient mice have been referred to as ‘innate-like’ T cells44.This phenotypic similarity to Itk-deficient mice supports the idea that

Themis is involved in TCR signal transduction through Itk. We alsonoted a much larger proportion (although a smaller absolute number)of regulatory T cells in Themis�/� mice. Notably, Themis was identifiedas one of the transcripts most strongly downregulated by the transcrip-tion factor Foxp3 in response to TCR stimulation45, and its suppressionmay therefore be part of the regulatory T cell development program.

Themis, Icb-1 and 9130404H23Rik are members of a small andessentially uncharacterized family. Each has very restricted, cell type–specific expression and is highly conserved in vertebrates. Other than aproline-rich sequence, they have no predicted conserved domainsshared with other proteins. These properties indicate that the familyhas important and perhaps novel functions and structure. Here wehave shown that one of these family members, Themis, is a T cell–specific protein that serves a crucial function in positive selection bytransducing TCR signals in thymocytes.

METHODS

Methods and any associated references are available in the onlineversion of the paper at http://www.nature.com/natureimmunology/.

Accession codes. Mouse Genome Informatics: mouse Themis,210757; GeneID: human THEMIS, 387357.

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

ACKNOWLEDGMENTSWe thank J. Bradley, A. Field, Y. Mondal, S. Rose, W. Sakati, V. Sharma,J. Shepherd and J. Vasquez for technical help. Supported by the US NationalInstitutes of Health (GM048002 and AI073870 to N.R.J.G. and T32 AI07244to M.V.M. and J.A.H.H.), the Concern Foundation for Cancer Research (S.V.),American Chemical Society (Irving S. Sigal Fellowship to J.A.H.H.), theWellcome Trust (GR076558MA to O.A., C.B. and M.S.), the Sir WilliamDunn School of Pathology (O.A., C.B. and M.S.) and Lincoln College, Oxford(Sloane Robinson Graduate Award to C.B.). This is manuscript 19524 fromThe Scripps Research Institute.

AUTHOR CONTRIBUTIONSN.R.J.G., M.V.M., S.V. and G.F. designed the project; M.V.M. cloned Themisand did initial gene-sequencing and expression analysis; G.F., S.V., V.R. andJ.A. did the main part of the study analyzing the Themis-deficient mice, withcontributions from J.A.H.H., A.M. and J.H.; P.R.F. constructed the Themistargeting vector and did expression studies; C.B., M.S. and O.A. identifiedand analyzed human Themis phosphorylation; Y.H.H. and K.S. helped withT cell signaling assays; H.S.F. did the in situ hybridization; N.R.J.G. and S.V.wrote the manuscript with contributions from G.F., V.R., K.S., O.A., A.M.and M.V.M.; and G.F., S.V., V.R., C.B., J.A.H.H. and N.R.J.G. prepared figures.

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

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

reprintsandpermissions/.

1. Starr, T.K., Jameson, S.C. & Hogquist, K.A. Positive and negative selection of T cells.Annu. Rev. Immunol. 21, 139–176 (2003).

2. Werlen, G., Hausmann, B., Naeher, D. & Palmer, E. Signaling life and death in thethymus: timing is everything. Science 299, 1859–1863 (2003).

3. Malissen, B. & Malissen, M. in T cell receptors (eds. Bell, J.I., Owen, M.J. &Simpson, E.) 352–368 (Oxford University Press, Oxford, 1995).

4. Davey, G.M. et al. Preselection thymocytes are more sensitive to T cell receptorstimulation than mature T cells. J. Exp. Med. 188, 1867–1874 (1998).

5. Lucas, B., Stefanova, I., Yasutomo, K., Dautigny, N. & Germain, R.N. Divergent changesin the sensitivity of maturing Tcells to structurally related ligands underlies formation ofa useful T cell repertoire. Immunity 10, 367–376 (1999).

6. D’Oro, U., Vacchio, M.S., Weissman, A.M. & Ashwell, J.D. Activation of the Lck tyrosinekinase targets cell surface T cell antigen receptors for lysosomal degradation. Immunity7, 619–628 (1997).

7. Naramura, M., Kole, H.K., Hu, R.-J. & Gu, H. Altered thymic positive selection andintracellular signals in Cbl-deficient mice. Proc. Natl. Acad. Sci. USA 95,15547–15552 (1998).

8. Alam, S.M. & Gascoigne, N.R.J. Post-translational regulation of TCR Va allelicexclusion during T cell differentiation. J. Immunol. 160, 3883–3890 (1998).

9. Niederberger, N. et al. Allelic exclusion of the TCR a-chain is an active processrequiring TCR-mediated signaling and c-Cbl. J. Immunol. 170, 4557–4563 (2003).

NATURE IMMUNOLOGY VOLUME 10 NUMBER 8 AUGUST 2009 855

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

10. Fischer, A. & Malissen, B. Natural and engineered disorders of lymphocyte develop-ment. Science 280, 237–243 (1998).

11. Zamoyska, R. et al. The influence of the src-family kinases, Lck and Fyn, on T celldifferentiation, survival and activation. Immunol. Rev. 191, 107–118 (2003).

12. Mariathasan, S. et al. Duration and strength of extracellular signal-regulated kinasesignals are altered during positive versus negative thymocyte selection. J. Immunol.167, 4966–4973 (2001).

13. Fischer, A.M., Katayama, C.D., Pages, G., Pouyssegur, J. & Hedrick, S.M. The role oferk1 and erk2 in multiple stages of Tcell development. Immunity 23, 431–443 (2005).

14. McNeil, L.K., Starr, T.K. & Hogquist, K.A. A requirement for sustained ERK signalingduring thymocyte positive selection in vivo. Proc. Natl. Acad. Sci. USA 102,13574–13579 (2005).

15. Daniels, M.A. et al. Thymic selection threshold defined by compartmentalization ofRas/MAPK signalling. Nature 444, 724–729 (2006).

16. Kane, L.P. & Hedrick, S.M. A role for calcium influx in setting the threshold forCD4+CD8+ thymocyte negative selection. J. Immunol. 156, 4594–4601 (1996).

17. Freedman, B.D., Liu, Q.H., Somersan, S., Kotlikoff, M.I. & Punt, J.A. Receptor avidityand costimulation specify the intracellular Ca2+ signaling pattern in CD4+CD8+

thymocytes. J. Exp. Med. 190, 943–952 (1999).18. Johnson, A.L. et al. Themis is a member of a new metazoan gene family and is required

for the completion of thymocyte positive selection. Nat. Immunol. advance onlinepublication, doi:10.1038/ni1769 (13 July 2009).

19. Lesourne, R. et al. Themis, a T cell–specific protein important for late thymo-cyte development. Nat. Immunol. advance online publication, doi:10.1038/ni1768(13 July 2009).

20. McGuire, M.V., Suthipinijtham, P. & Gascoigne, N.R.J. The mouse Supt16h/Fact140gene, encoding part of the FACT chromatin transcription complex, maps close to Tcraand is highly expressed in thymus. Mamm. Genome 12, 664–667 (2001).

21. Treeck, O., Strunck, E. & Vollmer, G. A novel basement membrane-induced geneidentified in the human endometrial adenocarcinoma cell line HEC1B. FEBS Lett.425, 426–430 (1998).

22. Su, A.I. et al. A gene atlas of the mouse and human protein-encoding transcriptomes.Proc. Natl. Acad. Sci. USA 101, 6062–6067 (2004).

23. Alam, S.M. et al. T cell receptor affinity and thymocyte positive selection. Nature 381,616–620 (1996).

24. Hogquist, K.A. et al. T cell receptor antagonist peptides induce positive selection. Cell76, 17–27 (1994).

25. Hogquist, K.A. et al. Identification of a naturally occurring ligand for positive selection.Immunity 6, 389–399 (1997).

26. Fiorini, E. et al. Peptide-induced negative selection of thymocytes activates transcrip-tion of an NF-kB inhibitor. Mol. Cell 9, 637–648 (2002).

27. Kaye, J. et al. Selective development of CD4+ T cells in transgenic mice expressing aclass II MHC-restricted antigen receptor. Nature 341, 746–749 (1989).

28. Gallegos, A.M. & Bevan, M.J. Central tolerance to tissue-specific antigens mediated bydirect and indirect antigen presentation. J. Exp. Med. 200, 1039–1049 (2004).

29. van Ewijk, W., Hollander, G., Terhorst, C. & Wang, B. Stepwise development of thymicmicroenvironments in vivo is regulated by thymocyte subsets. Development 127,1583–1591 (2000).

30. Irwin, M.J. & Gascoigne, N.R.J. Interplay between superantigens and the immunesystem. J. Leukoc. Biol. 54, 495–503 (1993).

31. Scherer, M.T., Ignatowicz, L., Winslow, G.M., Kappler, J.W. & Marrack, P. Super-antigens: bacterial and viral proteins that manipulate the immune system. Annu. Rev.Cell Biol. 9, 101–128 (1993).

32. Kisielow, P., Bluthmann, H., Staerz, U.D., Steinmetz, M. & von Boehmer, H. Tolerancein T-cell-receptor transgenic mice involves deletion of nonmature CD4+8+ thymocytes.Nature 333, 742–746 (1988).

33. Teh, H.S. et al. Thymic major histocompatibility complex antigens and the abT-cell receptor determine the CD4/CD8 phenotype of T cells. Nature 335, 229–233(1988).

34. Teh, H.S., Kishi, H. & von Boehmer, H. Deletion of autospecific Tcells in Tcell receptor(TCR) transgenic mice spares cells with normal TCR levels and low levels of CD8molecules. J. Exp. Med. 169, 795–800 (1989).

35. Ge, Q., Hu, H., Eisen, H.N. & Chen, J. Different contributions of thymopoiesis andhomeostasis-driven proliferation to the reconstitution of naive and memory T cellcompartments. Proc. Natl. Acad. Sci. USA 99, 2989–2994 (2002).

36. Berg, L.J., Finkelstein, L.D., Lucas, J.A. & Schwartzberg, P.L. Tec family kinasesin T lymphocyte development and function. Annu. Rev. Immunol. 23, 549–600(2005).

37. Huang, Y.H. et al. Positive regulation of Itk PH domain function by soluble IP4. Science316, 886–889 (2007).

38. Burkhardt, J.K., Carrizosa, E. & Shaffer, M.H. The actin cytoskeleton in T cellactivation. Annu. Rev. Immunol. 26, 233–259 (2008).

39. Yachi, P.P., Ampudia, J., Gascoigne, N.R.J. & Zal, T. Nonstimulatory peptides con-tribute to antigen-induced CD8-T cell receptor interaction at the immunologicalsynapse. Nat. Immunol. 6, 785–792 (2005).

40. Yachi, P.P., Lotz, C., Ampudia, J. & Gascoigne, N.R.J. T cell activation enhancement byendogenous pMHC acts for both weak and strong agonists but varies with differentia-tion state. J. Exp. Med. 204, 2747–2757 (2007).

41. Liu, K.Q., Bunnell, S.C., Gurniak, C.B. & Berg, L.J. T cell receptor-initiated calciumrelease is uncoupled from capacitative calcium entry in Itk-deficient T cells. J. Exp.Med. 187, 1721–1727 (1998).

42. Labno, C.M. et al. Itk functions to control actin polymerization at the immune synapsethrough localized activation of Cdc42 and WASP. Curr. Biol. 13, 1619–1624 (2003).

43. Kupzig, S., Walker, S.A. & Cullen, P.J. The frequencies of calcium oscillations areoptimized for efficient calcium-mediated activation of Ras and the ERK/MAPKcascade. Proc. Natl. Acad. Sci. USA 102, 7577–7582 (2005).

44. Berg, L.J. Signalling through TEC kinases regulates conventional versus innate CD8+

T-cell development. Nat. Rev. Immunol. 7, 479–485 (2007).45. Marson, A. et al. Foxp3 occupancy and regulation of key target genes during T-cell

stimulation. Nature 445, 931–935 (2007).

856 VOLUME 10 NUMBER 8 AUGUST 2009 NATURE IMMUNOLOGY

A R T I C L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

ONLINE METHODSMice. Tcra�/� mice (B6.129S2-Tcratm1Mom/J), Rag1�/� mice (B6.129S7-

Rag1tm1Mom/J) and Ly5a congenic mice (B6.SJL-Ptprca Pep3b/BoyJ) on the B6

background were from the Jackson Laboratory. Other strains were bred at The

Scripps Research Institute and were maintained in accordance with the Animal

Care and Use Committee of The Scripps Research Institute.

Antibodies. Anti-CD3 (145-2C11), anti-CD4 (RM4.4, RM4.5 and GK1.5),

anti-CD8a (53-6.7), anti-CD25 (PC61.5), anti-CD62L (MEL14), anti-CD44

(IM7), anti-Itk (2F12), anti-PLC-g1 (10/PLCg), anti-Foxp3 (FJK-165), anti-

Vb5 (MR9-6), anti-Vb6 (RR4-7), anti-Vb11 (CTVB11) and anti-Vb12

(CTVB12b) were from eBioscience or BD Biosciences. Antibody to phosphory-

lated tyrosine (4G10) was from Upstate Biotechnology. Anti-p42/44 (3A7),

antibody to phosphorylated p42/44 (197G2), antiserum to phosphorylated

Zap70 (2704) and antiserum to phosphorylated PLC-g1 (2821) were from Cell

Signaling Technology. Antibody to mouse Themis was raised in rabbits against

the amino-terminal 294 amino acids of mouse Themis, made as a fusion to

glutathione S-transferase, purified and with the glutathione S-transferase

removed; the antiserum was affinity purified. Antiserum to human Themis

was made by immunization of rabbits with a peptide corresponding to the

carboxy-terminal 15 amino acids of human Themis (specificity, Supplemen-

tary Fig. 9).

Cloning of Themis. For the subtraction library, cDNA from Tcra�/� thymus

mRNA was hybridized with cDNA from Rag1�/� thymus mRNA; the Tcra�/�

thymus cDNA that remained unhybridized was used to construct the library

as described20. Clones randomly selected from the library were sequenced and

used to search Genbank to identify potentially new clones. A full-length clone

of human THEMIS was isolated by screening of a Jurkat cDNA library.

Real-time RT-PCR analysis. SYBR Green PCR Master Mix and a Prism 7900

was used for real-time PCR (primer sequences, Supplementary Table 2) and

data were analyzed with SDSv2.1 software (Applied Biosystems). Results were

normalized to b-actin expression.

Generation of Themis-deficient mice. Exon 1 of Themis was replaced by

the 1.6-kilobase phosphoglycerate kinase–neomycin-resistance cassette. The

target vector (pKO Scrambler) contained the flanking sequence of exon 1:

a 5¢ 2.8-kilobase short arm and a 3¢ 3.9-kilobase long arm sequence flanking

the neomycin-resistance phosphotransferase gene. Embryonic stem cells

(129SvEv strain) were electroporated with 25 mg linearized DNA per 1 � 107

cells and were cultured for 7 d in medium containing the aminoglycoside

G418 (400 mg/ml). Drug-resistant clones were expanded and homologous

recombination was confirmed by PCR and Southern blot analysis. Positive

clones were injected into blastocysts derived from B6 embryo donor mice

mated with C6D2F1 stud males (B6 � DBA2/J hybrids); the blastocysts were

then transplanted into pseudopregnant CD-1 recipients. Chimeras were bred

to C57BL/6J females for demonstration of germline transmission, followed by

backcrossing to the B6 strain for ten generations.

Ca2+ flux. Thymocyte suspensions were prepared and put in separate tubes.

One was loaded for 10 min at 37 1C with CFSE (carboxyfluorescein diacetate

succinimidyl ester; 20 nM) or for 5 min at 21 1C with indodicarbocyanine

(Cy5; 1 mg/ml) or was ‘mock’ treated, followed by washing. Equal numbers

of CFSE-labeled and/or Cy5-labeled cell populations were mixed with the

unlabeled sample; for example, OT-I Tap1�/�Themis+/+ (mock) plus OT-I

Tap1�/�Themis�/� (CFSE or Cy5). In some experiments, the CFSE- and

Cy5-treated and untreated pairs were reversed to ensure that CFSE or Cy5 did

not alter the results. Cells were suspended at a density of 1 � 107 cells per ml

in cRPMI (RPMI medium supplemented with 10% (vol/vol) FCS, 100 U/ml

of penicillin, 10 mg/ml of streptomycin, 292 mg/ml of glutamine, 50 mM

2-mercaptoehthanol and 25 mM HEPES, pH 7.3) and were incubated for

30 min at 37 1C in 5% CO2 with the calcium indicator Indo-1-AM (2 mM;

Molecular Probes). Cells were washed twice with cRPMI. Cells were treated

for 20 min on ice with biotin-conjugated anti-CD3 and anti-CD4 (RM4.4), as

well as peridinin chlorophyll protein–cyanine 5.5–conjugated anti-CD8 and

phycoerythrin-indotricarbocyanine–conjugated anti-CD4 (GK1.5). Cells were

washed once with cRPMI and once with cHBSS (Ca2+-free and Mg2+-free

Hank’s balanced-salt solution supplemented with 1% (vol/vol) FCS, 1 mM

MgCl2, 1 mM EGTA and 10 mM HEPES, pH 7.3) and were resuspended in

cHBSS. Cells were prewarmed to 37 1C before analysis and were kept at 37 1C

during event collection on an LSR II (Becton-Dickinson). For cell stimula-

tion, streptavidin (10 mg/ml; Jackson) was added to crosslink biotinylated

antibodies; alternatively, cells were stimulated with tetramers made as described

from refolded H-2Kb and b2-microglobulin15,23 or with iTag MHC tetramer

(Beckman-Coulter). CaCl2 (5 mM) was added during analysis, and maximum

Ca2+ flux was obtained by the addition of ionomycin (500 ng/ml; Calbiochem).

Mean fluorescence ratio was calculated with FlowJo (TreeStar).

T cell activation and proliferation assay. Lymphocytes from Themis+/+ (Thy-

1.1, B6.PL) mice or Themis�/� (Thy-1.2, B6) mice were mixed at a ratio of 1:1.

Cells were cultured for 5 h in the presence of soluble anti-CD3, were stained

with anti-CD69 and anti-Thy-1 allotypes and were analyzed by flow cytometry.

For proliferation assays, cells were stained for 10 min at 37 1C with 0.2 mM

CFSE. Staining was stopped by the addition of an equal volume of FCS. After

being washed, cells were cultured for 48 h with or without stimulants (such as

anti-CD3) in cRPMI. T cell proliferation was measured by CFSE dilution

analyzed by flow cytometry.

Thymocyte CD69 upregulation assay. Thymocytes (2 � 105 to 3 � 105) from

OT-I-transgenic Tap1�/� mice (Themis+/� or Themis�/�) were incubated for

5 h at 37 1C with 2 � 105 peptide-pulsed EL4 cells in round-bottomed 96-well

plates. Cells were stained and analyzed by flow cytometry as described39,40.

Gates defining CD69� versus CD69+ were determined with thymocytes

incubated with non–peptide-pulsed EL4 cells.

Immunoprecipitation and immunoblot analysis. Cells were lysed in buffer

containing 20 mM Tris, pH 7.5, 150 mM NaCl, 0.5% (vol/vol) n-dodecyl-b-

D-maltoside or Nonidet P-40, 20 mM NaF, 1 mM Na3VO4 and protease

inhibitor ‘cocktail’ (Roche). Cell lysates (0.5–1 mg protein) were immunopre-

cipitated overnight at 4 1C, then protein A–Sepharose beads (Amersham) were

added for 1 h. After being washed, samples were resolved by 4–12% SDS-PAGE

and blotted. Membranes were blocked for 1 h with 1% (vol/vol) Tween and 5%

(wt/vol) dry milk in PBS, incubated for 3 h at 21 1C or overnight at 4 1C with

primary antibody, washed and then incubated for 1.5 h with secondary

horseradish peroxidase–conjugated antibody, followed by detection of signal

by enhanced chemiluminescence (Pierce).

In situ hybridization. Thymi from 3- to 4-week-old mice were fixed and

embedded in paraffin and sections 5 mm in thickness were probed with antisense35S-UTP-labeled Themis RNA. Sections were treated with RNAase A, washed

(0.2� SSC at 42 1C), dried and coated with Kodak NTB2 emulsion, followed by

exposure for 1–2 weeks, then samples were developed and counterstained with

methyl green. Controls included hybridization with sense and unrelated probes.

Peptide stimulation. OVA peptide (SIINFEKL) or vesicular stomatitis virus

peptide (RGYVYQGL) was injected intravenously into OT-I Tap1�/� mice as

described26. After 1 h, the thymus was removed and RNA was extracted and

analyzed by real time RT-PCR.

Sequence alignments and comparisons. Sequences were aligned and compared

with the MultAlin multiple-sequence alignment program46 and results were

rendered with the ESPript utility for generating Postscript files47.

Statistical testing. Statistical differences were calculated with the mean-

difference hypothesis of the Student’s two-tailed t-test with the assumption

of different variances and a confidence level of 95%. GraphPad Prism was used

for calculations.

46. Corpet, F. Multiple sequence alignment with hierarchical clustering. Nucleic Acids Res.16, 10881–10890 (1988).

47. Gouet, P., Courcelle, E., Stuart, D.I. & Metoz, F. ESPript: analysis of multiple sequencealignments in PostScript. Bioinformatics 15, 305–308 (1999).

doi:10.1038/ni.1766 NATURE IMMUNOLOGY

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.