of June 15, 2015.This information is current as

Knockout Mouse ModelsofDifferentiation by Transcriptional Analysis

A General Survey of Thymocyte

NguyenGeneviève Victorero, Béatrice Loriod and Catherine Denis Puthier, Florence Joly, Magali Irla, Murielle Saade,

http://www.jimmunol.org/content/173/10/6109doi: 10.4049/jimmunol.173.10.6109

2004; 173:6109-6118; ;J Immunol 

MaterialSupplementary

htmlhttp://www.jimmunol.org/content/suppl/2004/11/02/173.10.6109.DC1.

Referenceshttp://www.jimmunol.org/content/173/10/6109.full#ref-list-1

, 41 of which you can access for free at: cites 64 articlesThis article

Subscriptionshttp://jimmunol.org/subscriptions

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/ji/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/cgi/alerts/etocReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists All rights reserved.Copyright © 2004 by The American Association of9650 Rockville Pike, Bethesda, MD 20814-3994.The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

by guest on June 15, 2015

http://ww

w.jim

munol.org/

Dow

nloaded from

A General Survey of Thymocyte Differentiation byTranscriptional Analysis of Knockout Mouse Models1

Denis Puthier,2 Florence Joly,2 Magali Irla, Murielle Saade, Genevieve Victorero,Beatrice Loriod, and Catherine Nguyen3

The thymus is the primary site of T cell lymphopoiesis. To undergo proper differentiation, developing T cells follow a well-orderedgenetic program that strictly depends on the heterogeneous and highly specialized thymic microenvironment. In this study, weused microarray technology to extensively describe transcriptional events regulating �� T cell fate. To get an integrated view ofthese processes, both whole thymi from genetically engineered mice together with purified thymocytes were analyzed. Using miceexhibiting various transcriptional perturbations and developmental blockades, we performed a transcriptional microdissection ofthe organ. Multiple signatures covering both cortical and medullary stroma as well as various thymocyte maturation intermediateswere clearly defined. Beyond the definition of histological and functional signatures (proliferation, rearrangement), we provide thefirst evidence that such an approach may also highlight the complex cross-talk events that occur between maturing T cells andstroma. Our data constitute a useful integrated resource describing the main gene networks set up during thymocyte developmentand a first step toward a more systematic transcriptional analysis of genetically modified mice. The Journal of Immunology, 2004,173: 6109–6118.

W ith the systematic sequencing of cDNA libraries andthe advent of microarray analysis, it has become pos-sible to describe, at the tissue level, the expression

pattern of thousands of genes already known or yet to be charac-terized. As an example, multitissue experiments using microarraysprovided a broad source of information and constituted a first stepin the description of gene expression profiles (1, 2). However, atthe level of each organ, higher resolution is necessary to specifythe histological expression pattern of individual genes. This task iscurrently in progress, but production of informative samples oftenrequires microdissection or purification procedures (3). The exten-sive development of genetically modified mice in the past fewyears has led to the generation of numerous models of transcrip-tional perturbations, most of them exhibiting very precise devel-opmental blockades and thus quantitative cellular disequilibrium(4). Hence, the use of perturbation models for transcriptional anal-yses may provide us with samples highly enriched in as yet poorlydescribed maturation intermediates.

The thymus is the major site for T lymphocyte maturation. An-atomically, it is divided into a subcapsular region, a cortex, wheremost of the thymocyte differentiation takes place and a medulla,where newly generated T cells undergo final processes of matura-tion. Each of these compartments forms a specialized stromal mi-

croenvironment that is crucial to control T cell fate. This stroma isessentially composed of epithelial cells, dendritic cells, macro-phages, and fibroblasts. It provides developing T cells with essen-tial extracellular matrix components, cell surface ligands, and sol-uble factors. Lymphoid progenitors derived from the bone marrowenter the thymus at the corticomedullary junction, and then migrateto the subcapsular region where they undergo multiple cycles ofproliferation and progress to the CD4�CD8� double-negativestage (DN)4 through discrete maturation steps: CD44highCD25�

(DN1), CD44highCD25� (DN2), CD44lowCD25� (DN3), andCD44lowCD25� (DN4). During DN to CD4�CD8� double-posi-tive (DP) transition, thymocytes relocate to the cortical region.Most DP cells (�97%) die by neglect, because they do not rec-ognize any of the available MHC molecules expressed by thymicstromal cells. Ultimately, DP thymocytes are subjected to negativeselection events occurring both in the cortex and in the medulla,leading to the deletion of autoreactive clones and giving rise to thegeneration of MHC-restricted CD4 and CD8 single-positive Tcells (CD4�SP and CD8�SP) (5, 6). Although the molecular path-ways involved in thymocyte ontogeny have been the subject ofextensive studies, some of them based on microarray technology(7–9), our understanding of these mechanisms is still fragmented,and results often remain difficult to situate in an integrated anddynamic view of T cell maturation and T cell-stroma cross talk.

By taking advantage of the relative enrichment in specific thy-mocyte and stromal populations displayed by some genetically en-gineered mice, microarray analysis should reveal their character-istic transcriptional signatures. Identification of such signaturesmay highlight the functional specialization of each cell populationand should allow us to propose a putative role for still poorly

Technologies Avancees pour le Genome et la Clinique/ERM 206, Parc Scientifique deLuminy, Marseille, France

Received for publication April 26, 2004. Accepted for publication August 17, 2004.

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.1 D.P. is supported by grants from the Agence Nationale de Recherche sur le Sida.This work was supported by Institut National de la Sante et de la Recherche Medicaleand Ligue Nationale Contre le Cancer. Microarrays were developed and produced inthe microarray platform of Marseille genopole.2 D.P. and F.J. contributed equally to this work.3 Address correspondence and reprint requests to Dr. Catherine Nguyen, Technolo-gies Avancees pour le Genome et la Clinique/ERM 206, Parc Scientifique de Luminy,case 928, 13288 Marseille cedex 09, France. E-mail address: nguyen@tagc.univ-mrs.fr

4 Abbreviations used in this paper: DN, double negative; DP, CD4�CD8� double-positive thymocyte; CD4�SP, CD4 single-positive thymocyte; CD8�SP, CD8 single-positive thymocyte; KO, knockout; WT, wild type; DAPI, 4�,6�-diamidino-2-phe-nylindole; Ac.CD4�P, activated CD4� peripheral T cell; Ac.CD8�P, activated CD8�

peripheral T cell.

The Journal of Immunology

Copyright © 2004 by The American Association of Immunologists, Inc. 0022-1767/04/$02.00

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

characterized genes. In addition, because T cell and stromal de-velopment are interdependent events, the combined analysis of ge-netically engineered mouse models affecting one or both compart-ments should underline such cross-talk events. In the presentstudy, using a 8700 cDNA mouse microarray, we draw up the firsttranscriptional picture of the thymus as a whole by analyzing bothwhole thymi from well-defined knockout (KO) mice together withpurified thymocytes.

Materials and MethodsMice

Mice were housed in a specific pathogen-free animal facility. C57BL/6(wild type (WT)), RAG1°, and TCR�° mice were obtained from the Centrede Distribution, Typage et Archivage (Orleans, France). CD3-��5 (CD3�°),LAT°, and RelB° mice were kindly provided by Drs. M. Malissen and P.Naquet (Centre d’Immunologie de Marseille Luming, Marseille, France).Thymi were isolated from adult mice that were sacrificed between 4 and 6wk of age.

Organs and cells

Cells were cultured in RPMI 1640 supplemented with 10% heat-inacti-vated FCS, 2 mmol/L glutamine, 100 �g/ml streptomycin, 100 U/ml pen-icillin, and 2 � 10�5 M 2-ME. The LMTK� fibroblastic cell line and theP388D1 macrophage cell line were both obtained from the American TypeCulture Collection (Manassas, VA). MTE-1D is a thymic epithelial cellline (10). DC2.4 are immortalized dendritic cloned cells derived from bonemarrow of C57BL/6 mice (11). 427.1, 1308.1, and 6.1.1 thymic stromalcell lines were kindly provided by Dr. B. Knowles (The Jackson Labora-tory, Bar Harbor, ME) (12). Thymocytes were separated from stroma ac-cording to standard procedures. CD25� thymocytes were isolated fromRAG1° mice. After labeling with an anti-CD25 Ab, they were purifiedusing magnetic beads (Dynal Biotech, Oslo, Norway) according to manu-facturer’s guidelines. These samples contain CD44highCD25� (DN2) andCD44lowCD25� (DN3) and are referred to as CD25�DN. For DP purifi-cation, thymocytes were sorted using a FACSVantage cell sorter (BD Bio-sciences, Mountain View, CA) on the basis of costained profiles with ananti-CD4 (RM4-4; BD Pharmingen, San Diego, CA) and anti-CD8 (53-6.7;BD Pharmingen). For SP cell isolation, single-cell suspensions were firstincubated with a mixture of Abs composed of an FcR-blocking Ab (2.4G2)and either an anti-CD8 (H-59-101-2) or an anti-CD4 (H129.196), for 30min at 4°C followed by incubation with sheep anti-rat IgG magnetic beads(Dynal Biotech). The populations recovered were then stained with anti-CD4 (RM4-4) and anti-CD8 (53-6.7) (BD Pharmingen) and FACS sorted.

Microarray design

For microarray preparation, the following cDNA libraries (some of whichwere incomplete) were used: the NIA Mouse 15K cDNA clone set, 2NbMT(thymus), NbMLN (lymph node), and 3NbMS (spleen). Detailed descrip-tions of these cDNA libraries are available at the UniGene database web-site(www.ncbi.nlm.nih.gov/UniGene/lbrowse2.cgi?TAXID�10090, 2NbMT:Lib.544, 3NbMS: Lib.553, NbMLN: Lib.567, NIA 15K: Lib.8622). All ofthe libraries were cloned into pT3T7D-Pac vector, except for the NIA 15KMouse cDNA clone set, which was cloned into pSPORT1 vector. The NIAMouse 15K cDNA clone set is a rearrayed and resequenced set of 15,000bacterial clones derived from 11 embryo cDNA libraries (13). It was ob-tained through the Medical Research Council Geneservice (Cambridge,U.K.). 2NbMT, NbMLN, and 3NbMS are sequenced I.M.A.G.E. librariesthat were obtained from the Deutsches Ressourcenzentrum fur Genomfor-schung Resource Centre (Berlin, Germany). Bacterial clones were anno-tated according to UniGene using the SOURCE database (14) before bioin-formatics selection and robotic reorganization. A total of 4300 clones wasselected from the NIA 15K library and had been previously sequence-confirmed. However, the reorganized clone set was not subjected to rese-quencing before being used. The final clone set includes 8750 bacterialclones representing �7770 nonredundant UniGene clusters. This clone setencompasses almost all described genes (“known genes”) present in theselibraries and has also been supplemented with clones belonging to Uni-Gene clusters that contain at least one RIKEN cDNA clone entry (i.e., witha corresponding putative full-length cDNA) (15). Although the gene con-tent of this clone set is quite exhaustive, some key regulators of T cellmaturation are still missing (e.g., RAG1, RAG2, ZAP-70). About 10% ofthe genes included in this clone set are represented by two or more different

cDNA clones, providing internal controls to assess the reproducibility ofgene expression measurements.

Microarray preparation

PCR amplifications were performed in 96-well microtiter plates usingthe following primers: 5�-CCAGTCACGACGTTGTAAAACGAC-3�and 5�-GTGTGGAATTGTGAGCGGATAACAA-3�, which were spe-cific of the polylinker sequence of both vectors used. The reactions wereperformed as described (16) by transferring few Escherichia coli froma growth culture with a plastic 96-pin gadget (Genetix, New Milton,U.K.) to 100 �l of PCR mix, containing 10 mM Tris-HCl (pH 9), 1.5mM MgCl2, 50 mM KCl, 0.1% Triton X-100, 1.5 M betaine, 250 �MdATP, dTTP, dGTP, and dCTP, and 5 U of Taq polymerase (Promega,Madison, WI). The plates were incubated for 6 min at 94°C, before 38cycles of 94°C for 30 s, 65°C for 45 s and 72°C for 210 s, followed bya final elongation phase at 72°C for 10 min. Amplification productswere not quantitated, but their quality was systematically checked on1% agarose gels. Eleven percent of the bacterial clones were estimatedto be nonamplified, and 4% showed multiple bands after PCR amplifi-cation. Unpurified PCR products were then transferred to 384-well mi-croplates, before being evaporated, taken up in 40 �l of distilled water,and spotted onto nylon membranes (Hybond-N�; Amersham Bio-sciences, Saclay, France). This step was conducted using a Micro-Grid-II arrayer (Apogent Discoveries, Cambridge, U.K.) equipped witha 64-pin biorobotics printhead. A vector probe hybridization (5�-TCACACAGGAAACAGCTATGAC-3�) was performed as previously de-scribed and showed uniform signal intensities across individual mem-branes (17).

RNA extraction and cDNA labeling

Total RNA was isolated using TRIzol reagent (Invitrogen Life Technolo-gies, Carlsbad, CA), and RNA integrity was checked on denaturing agarosegels. Reverse transcription was performed as previously described, using 2�g of total RNA in the presence of [�-33P]dCTP (18). A large excess ofoligo(dT) primers was added during cDNA synthesis, and the labeled probewas annealed with poly(dA)80 to ensure the complete saturation of poly(A)tails. Hybridizations were conducted for 48 h at 68°C in 500 �l of hybrid-ization buffer (18). After washing, arrays were exposed to phosphor im-aging plates, which were scanned using a BAS 5000 (Fuji, Tokyo, Japan)at a 25-�m resolution.

Data processing and analysis

After image acquisition, hybridization signals were quantified using thelocally developed Bzscan software (64). All images were carefully in-spected, and spots with overestimated intensities due to neighborhood ef-fects were manually excluded. For each of the 66 arrays, background in-tensity was calculated on the basis of the average of the 400 lowest valuesand subtracted. We defined distinct classes of samples, each of them con-taining all of the replicates of the same sample type (e.g., three replicatesof RAG1° thymi define one class). Of note, all of the cell lines, includingdendritic cells, were grouped in the same class. For each sample, the 6000highest values were flagged. A cDNA clone was kept for analysis if itsexpression values were flagged in all of the samples of at least one class.Data were filtered (80%), log2 transformed, and centered relative to themedian for each gene and each array using the Cluster software (19). Toanalyze genes with the most fluctuating expression levels across the sam-ples, only clones exhibiting a SD of at least 0.35 were kept. Hierarchicalclustering (average linkage clustering metric) was applied to the dataset,and results were visualized with the Treeview software (19). All data areMIAME compliant and have been submitted to ArrayExpress database(www.ebi.ac.uk/miamexpress; accession no. E-MEXP-96). This articlecontains supplemental data (Tables SI, SII, SIII, SIV, and SV).5

Immunofluorescence

Primary Abs used were as follows: FITC-anti-CD54/ICAM-1 (BD Phar-mingen, San Diego, CA), anti-RANTES (AF478; R&D Systems, Minne-apolis, MN), anti-K5 (AF138; Covance Research, Berkeley, CA), and anti-CD74 (IN-1). Secondary Abs used were as follows: Alexa-568-goatanti-rabbit IgG, Alexa-477-goat anti-rat IgG, and Alexa-477-goat anti-mouse IgG. Cryosections (8 �m) of 8-wk-old WT thymi were generatedfrom OCT-embedded organs and mounted on glass slides. Sections werefixed with phosphate buffer containing 4% paraformaldehyde for 1 h beforestaining in a humidified chamber. Abs (in 1% BSA, 0.01% Triton, and 0.1

5 The on-line version of this article contains supplemental material.

6110 TRANSCRIPTIONAL MICRODISSECTION OF THE MOUSE THYMUS

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

M Tris buffer (pH 7.4)) were applied to the sections and incubated for 1 hand 30 min at room temperature. Slides were then incubated in blockingbuffer (3% BSA in 0.01% Triton, 0.1 M Tris buffer, and 5% goat serum)for 10 min before being incubated with Alexa-conjugated-secondary Ab(1:1000 in 1% BSA, 0.01% Triton, and 0.1 M Tris buffer). Between eachstep, slides were washed for 5 min with Tris buffer. Tissues were coun-terstained with 1 �l/ml 4�,6�-diamidino-2-phenylindole (DAPI) andmounted with Mowiol fluorescent mounting medium (Calbiochem, Darm-stadt, Germany). Fluorescent images were acquired by Zeiss (Oberkochen,Germany) LSM 510 confocal microscopy. All images were processed afterthe same exposure time and under the same magnification.

ResultsTranscriptional microdissection of the murine thymus

To reveal the transcriptional signatures of the cellular compart-ments that constitute the adult mouse thymus, we used a combi-nation of whole thymi and of purified cell samples. We assumedthat comparing the transcriptional status of both WT and KO micewould help us to reveal specific signatures of over- or underrep-resented cellular compartments. The RAG1°, LAT°, and CD3-��5

(Cd3�°) KO mice were chosen, because they display an early blockin T cell development and consequently a relative enrichment inDN T cells and stromal cells (4). Thymi from TCR�° KO mice thatlack mature T cells and differentiated medullary stromal cells wereused to highlight the specific transcriptional signatures of thesecompartments. Mice invalidated for RelB, a member of the NF-�Bfamily, were also used, because they are characterized by a disor-ganized medulla and a lack of medullary dendritic cells (20). Inaddition, RNA from the following sorted thymocyte populationswere prepared: CD25� T cells from RAG1° mice (correspondingto DN2 and DN3 stages, CD25�DN), thymocytes from WT ani-mals (DN, DP, CD4�SP, CD8�SP), and activated peripheral T

cells (Ac.CD8�P and Ac.CD4�P). Finally, cell lines and dendriticcells were also used to delineate genes involved in cellular metab-olism. All samples were hybridized to a mouse microarray con-taining 8750 cDNA selected from embryo, thymus, lymph node,and spleen libraries. After normalization and data filtering, 4686genes were kept for analysis. Hierarchical clustering was used togroup genes on the basis of their variation of expression over sam-ples. The same method was applied to group the experimentalsamples according to their transcriptional similarity. Results arepresented in Fig. 1A. As depicted in Fig. 1B, the hierarchical clus-tering procedure gave rise to a highly relevant classification of thesamples, divided in four major groups. Cell lines and activatedperipheral T cells were grouped together due to their metabolicactivity. Indeed, their main transcriptional characteristic relies onthe high expression of genes encoding proteins involved in generalmetabolism (Fig. 1A, cluster G, and data not shown). The thymifrom RAG1°, LAT°, and Cd3�° clustered together, whereas a thirdgroup contained all of the purified thymocytes and also naive Tcells (not shown). The fourth class of samples was composed ofwhole thymi from WT, TCR�°, and RelB° mice. As shown in Fig.1A, numerous transcriptional signatures clearly stood out (signa-tures A–G). These signatures were extracted and samples werereordered according to their types (the signature G correspondingto general metabolism will not be discussed herein) (Fig. 1C). Agene expression signature was named by either the cell type inwhich its component genes were expressed (for example, “DN Tcells” signature) or the biological process in which its componentgenes are known to function (for example, the “proliferation” sig-nature). Signatures of T cell maturation stages (DN, DP, and SP)as well as functional signatures (proliferation, TCR rearrangement)

FIGURE 1. Hierarchical clustering based on gene expression data obtained from 66 RNA samples. A, Each row represents a separate cDNA clone onthe microarray, and each column represents a separate mRNA sample. Green, black, and red squares indicate transcript levels lower than, equal to, or greaterthan the median level, respectively. Gray squares correspond to technically inadequate or missing data. The six different signatures that will be developedthereafter are indicated (A–F). B, The dendrogram generated by the hierarchical clustering procedure (see Fig. 1A) organizes the samples according to theirtranscriptional similarity and reveals four major groups, presented in different colors. Of note, naive CD4 peripheral T cells cluster with the group of purifiedthymocytes. C, Supervised representation of the six distinct (A–F) transcriptional signatures. (The samples are not subjected to a classification accordingto their gene expression similarities but according to their origin.) Abbreviations used are as follows: CD25�DN, CD25� cells purified from RAG1° mice;DN, DP, CD4�SP, CD8�SP, T cell populations sorted from WT animals; CD4�P, naive peripheral T cells sorted from WT animals; Ac.CD4�P,Ac.CD8�P, peripheral CD4 and CD8 T cells activated with coated anti-CD3 and soluble anti-CD28 (1 �g/ml).

6111The Journal of Immunology

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

were defined (Fig. 1C, signatures A–E). For the SP T cell signature(E), three slightly different signatures from Fig. 1A (E1, E2, E3)originally disseminated along the clustering figure were groupedtogether in Fig. 1C. This signature encompasses genes character-istic of both late DP and SP stages. Indeed, these genes are not, oronly weakly, expressed in the TCR�° samples. In addition, it alsocontains genes more specifically expressed at the SP stage, whoseexpression is maintained in peripheral activated T cells. Of note,although we could detect some signatures specific for activatedperipheral T cells, they will not be considered in the present paper.

Finally, a clear histological signature (F), differentiating the thymicstroma from developing thymocytes, was also defined.

Proliferation status of developing thymocytes

Close inspection of the genes composing the signature A (Fig. 1B)suggests that it corresponds to cell cycle. Bibliographic scanningof cluster A indicates that it is composed of genes that controltransitions between G1, G2, and S phases, or that monitor entry andexit of mitosis: Ccna2, Cdc2, Cdca3/Tome1, and Ube2c (Table I).

Table I. Detailed gene contents of cell cycle-related signature (cluster A)

FunctionalClassificationa Ug.Symbolb Ug.Mmc Ug.Hsd LL.ID (Hs)e

Associated-MEGf Function

Cell cycle regulator Cdc451 Mm.1248 Hs.114311 8318 MEG1485 Initiation of DNA replication.Cdc2a Mm.4761 Hs.334562 983 Involved in G1/S and G2/M phase transitions.Ccna2 Mm.4189 Hs.85167 890 Involved in G1/S and G2/M phase transitions.Cdca1 Mm.151315 Hs.234545 83540 Regulatory role in chromosome segregation.Cdc20 Mm.29931 Hs.82906 991 MEG3165 APC activator during G1 and mitosis.cdca3/Grcc8 Mm.22228 Hs.30114 83461 Regulator of entry and exit of mitosis.

Chromatine structure Hist2 Mm.258496 Hs.348668 126961 Histone Nucleosome structure.H2afx Mm.245931 Hs.147097 3014 Histone DNA repair.H2afz Mm.916 Hs.119192 3015 MEG2200 Histone Transcriptional control.Nasp Mm.7516 Hs.446206 4678 H1 histone binding protein. Histone transport.Slbp Mm.4172 Hs.251574 7884 MEG3280 mRNA histone binding.

Mitotic checkpointcomponent

Birc5 Mm.8552 Hs.1578 332 MEG3232 Apoptosis inhibitor expressed during G2 phase.

Bub3 Mm.927 Hs.418533 9184 MEG4512 Mitotic checkpoint component.Ube2c Mm.89830 Hs.93002 11065 Destruction of mitotic cyclins. Cell cycle

progression.

Replication Prim1 Mm.2903 Hs.82741 5557 MEG1249 Synthesis of oligoribonucleotide primers onDNA.

Top2a Mm.4237 Hs.156346 7153 MEG518 Topoisomerase.Pcna Mm.7141 Hs.78996 5111 MEG2077 Cofactor of DNA polymerase �.Smc211/Fin16 Mm.2999 Hs.43910 10592 MEG2321 Mitotic chromosome condensation. DNA

repair.

Transcriptional Srebf2 Mm.38016 Hs.108689 6721 Transcription regulator.regulation Zhx1 Mm.37216 Hs.12940 11244 Transcription factor.

Hmgb2 Mm.1693 Hs.434953 3148 DNA bending factor.

RNA maturation Sfrs3 Mm.6787 Hs.405144 6428 MEG6285 Splicing factor.Rrm2 Mm.99 Hs.226390 6241 MEG1938 Ribonucleotide reductase M2.Hnrpab Mm.24 Hs.81361 3182 Pre-mRNA processing, mRNA metabolism,

and transport.No15a Mm.29363 Hs.376064 10528 MEG418 Pre-rRNA processing.Dutp Mm.276215 Hs.367676 1854 MEG674 Deoxyuridine triphosphatase. DNTP

metabolism.Ddx39 Mm.271162 Hs.311609 10212 MEG1184 Nuclear RNA helicase.Silg41 Mm.123859 Hs.30035 6434 MEG6925 Homolog to sfrs10. Splicing factor.

Miscellaneous Ptprv/Esp Mm.4450 375048 Embryonnic stem cell phosphatase.Hpgd Mm.18832 Hs.77348 3248 Hydroxyprostaglandin dehydrogenase 15-

(NAD).Topk-pending Mm.24337 Hs.104741 55872 Serine/threonine kinase.Lrpap1 Mm.4560 Hs.75140 4043 MEG2907 Lipoprotein receptor-associated protein.Dnclc1 Mm.56455 Hs.5120 8655 Dynein L chain. Enzymatic activity.

Unknown 2310061F22Rik Mm.37324 Hs.79077 9780 MEG42992410017I18Rik Mm.34567 Hs.46680 51001 MEG58102810417H13Rik Mm.269025

a Functional classification is based upon bibliographic searches.b Mouse UniGene symbols.c Mouse UniGene cluster IDs (release no. 129).d Corresponding human UniGene cluster IDs (release no. 163) based on the HomoloGene database (www.ncbi.nlm.nih.gov/HomoloGene).e Corresponding Human LocusLink IDs based on the HomoloGene database.f Corresponding associated metagenes (Associated-MEG) were retrieved from the Kimlab database (http://cmgm.stanford.edu/�kimlab/multiplespecies/Supplement/

stab1.xls).

6112 TRANSCRIPTIONAL MICRODISSECTION OF THE MOUSE THYMUS

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

Many genes (Cdc20, Bub3, and Cdca1) are involved in microtu-bule-dependent processes that are hallmarks of dividing cells. An-other feature of this cluster is the presence of numerous genesencoding components of chromatin structure. It includes three hi-stones genes, Hist2 and two histones variants, H2afx and H2afz, aswell as Nasp and Slbp. Interestingly several genes present in clus-ter A actively participate in DNA replication (Cdc45, Prim1,Top2a, Pcna, and Smc2l1/Fin16).

To visualize these genes in terms of function, we took advantageof the classification proposed by Stuart et al. (21). They definedmetagenes as sets of orthologous genes conserved between Homosapiens, Drosophila melanogaster, Caenorhabditis elegans, andSaccharomyces cerevisiae. By compiling microarray experimentsfrom these four species, they organized these metagenes into agene-coexpression network revealing functional clusters con-served across evolution. Because mouse microarray data have notbeen included in their study, we searched for each gene of clusterA, the human orthologue, and then the corresponding metagene.Among the 36 genes of cluster A, 19 were found to have an as-sociated metagene (Table I). In the conserved gene coexpressionnetwork, most of them colocalized in an area defined by Stuart etal. (21) as corresponding to cell cycle function (Fig. 2). Dutp andRrm2, two genes known to be involved in RNA metabolism, wereincluded in this metacluster, suggesting that they have functionallinks with the cell cycle. Likewise, MEG5810 corresponds to astill-uncharacterized mouse gene (2410017I18Rik) that most likelyparticipates in cell cycle control. Remaining metagenes weremainly found in two regions, highly enriched in genes devoted togeneral transcription and translation. Two metagenes (MEG4299/2310061F22Rik, MEG2907/Lrpap1) were found scattered in thecoexpression network. Although we cannot formally exclude theirparticipation in cell cycle-related processes, they may reflect mea-surement errors or clone contaminations. Altogether, this resultunderscores the capacity of our procedure to group-coexpressedgenes in functional networks without any underlying assumption.

As expected, the expression levels of all the genes in cluster Afluctuated drastically among the samples. These variations clearlyreflected the proliferative status of the biological samples. The

highest proliferation rates were observed in activated peripheral Tcells and also in WT, RelB°, and TCR�° thymi. This last obser-vation can be correlated with the presence of both highly dividingimmature single-positive thymocytes and DP cells in these thymi.Among the most immature thymocyte populations, total DN Tcells, highly enriched in DN4 stage, were found to have the highestlevel of proliferation, compared with CD25�DN isolated fromRAG1° thymi. An intermediate level of proliferation was observedin DP T cells, whereas the lowest rate of cycling characterized theSP population, except for CD8�SP. Although this could reflectproliferative differences between CD4�SP and CD8�SP, this morelikely reveals the presence of a few contaminating and highly di-viding CD8� immature single-positive thymocytes in the cellpreparation.

Notch1, Smarca4/Brg1, and their targets are highly expressed inCD25�DN T cells

Genes from cluster B were highly expressed in the thymi frommice characterized by an early blockade in T cell differentiation, inthe CD25�DN T cells from RAG1° mice, and in DN T cells ob-tained from WT animals (Fig. 1B). Expression profiles of thesegenes in purified T cells are presented in Fig. 3 (only representativegenes are depicted). The complete data set including all genes (n �41) is available in Supplemental Table SI. Cluster B was highlyenriched in genes previously shown to be strongly expressed dur-ing the first thymocyte developmental stages. It includes cell sur-face markers such as CD25, CD104/Itgb4, and the Gfra1 (22).

FIGURE 2. Analysis of the cell cycle signature. Visualization of thegenes of cluster A in the conserved coexpression network proposed byStuart et al. (21). The metagenes corresponding to genes from cluster A(Table I) were plotted on the conserved coexpression network and visual-ized using online tools at http://cmgm.stanford.edu/cgi-bin/cgiwrap/kim-lab/multiplespecies/visualize.pl. The mouse UniGene symbol correspond-ing to each metagene is included. Red circle corresponds to metagenes (i.e.,sets of orthologous genes) that are devoted to cell cycle control throughoutevolution. Green and blue circles correspond to genes with conserved func-tions in general transcription and translation, respectively.

FIGURE 3. Gene expression programs induced during thymocyte de-velopment. The clusters B, C, D, and E correspond to the major stages ofthymocyte development (B, double-negative stage; C, gene expression pro-gram induced during TCR rearrangement; D, double-positive stage; and E,single-positive stage). These clusters are directly derived from Fig. 1C, butonly purified thymocyte and T cell samples are shown. To clarify, the samplesare ordered according to the developmental scale of thymocyte differentiation.For each cluster, only representative genes are depicted. The complete data setis available as Supplemental Tables SI, SII, SIII, and SIV.

6113The Journal of Immunology

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

Importantly, several transcriptional regulators, including Notch1,Smarca4/Brg1, Dtx1/Deltex1, and Hes1/Hry were specifically ex-pressed in DN T cells. Coregulated expressions of these genespointed out some complex transcriptional regulatory networks.Dtx1 and Hes1 are two well-established targets of Notch1.Smarca4/Brg1 encoding a component of the SWI/SNF chromatin-remodeling complex showed a coordinated expression with Cd25.Interestingly, Smarca4/Brg1 is known to be recruited in the Cd25core promoter at the CD25�DN stages (23). The notion of func-tional links between genes of cluster B was reinforced by the si-multaneous expression of several genes involved in calcium-de-pendent signaling pathways, as illustrated with the following:Calpain5/Capn5, FK506-bp5, the calcium/calmodulin-dependenteukaryotic elongation factor-2 kinase/Eef2k, and Camk2a/CaMKII, which directly regulates the calcium channel, Cacna1h,also found in this cluster (24). These data underline the high en-richment in this signature of key regulators of CD25�DN stages ofmaturation. They suggest that additional unknown genes may belinked to the Notch1 or Smarca4/Brg1 pathway. These results alsopoint out the functional link that may exist between Notch1,Smarca4/Brg1, and the calcium-signaling pathway.

Cluster C corresponds to genes up-regulated during the TCRrearrangement process

Genes from cluster C show a relatively high expression in all thethymi tested, in CD25�DN, and in DP T cells (Fig. 1B). In con-trast, they were strongly down-regulated in CD4 and CD8 T cells.Five of 55 genes were excluded from cluster C, because they werenot significantly different between immature thymocytes(CD25�DN and DP T cells) and mature thymic and peripheral Tcells (CD4�SP, CD8�SP, Ac.CD4�P, and Ac.CD8�P) ( p � 0.05,Student’s t test). Close inspection of cluster C showed that it in-cludes several genes whose expression is known to diminish uponT cell maturation (Fig. 3 and supplemental data, Table SII). Sev-eral cell surface markers were found in cluster C (CD24/Hsa,CD3�, CD3�, Ly6e, Ly6d, CD26, Cdw127/Il17r, and Sort1/Ntr3).Moreover, numerous nuclear factors, known to regulate TCR ex-pression and rearrangements, were highly represented and clus-tered in the same expression network. Cutl1 has been shown tobind matrix attachment region sequences, and it has been proposedto be involved in the control of DNA accessibility during V(D)Jrecombination (25). Sp2 is a zinc finger-containing protein that isable to bind the GT box element in TCR V � promoters (26).Tcf12/Heb is abundantly detected in thymocytes and is thought toregulate E-box sites present in many T cell-specific gene enhanc-ers, including TCR-� and TCR-� (27). Runx1/AML1 has beenshown to modulate the local accessibility of DNA to the V(D)Jrecombinase (28). In the same way, G22p1/Ku70 and Dntt/Tdt arecomponents of the DNA repair/V(D)J recombination machineryand are crucial for rearrangement and generation of diversity at theV(D)J junction, respectively. Importantly, Terf2/Trf2/TLP acts asa positive regulator of the Dntt/Tdt promoter (29). The presence ofPpp1r1c/Tarpp is in agreement with its recently suggested role inTCR rearrangement (30). Of note, 16 other still-uncharacterizedgenes share the same transcription profile, and some of them maybe functionally related to rearrangement events.

Cluster D is mainly composed of genes highly expressed in WT,RelB°, and TCR�° mice, and corresponds to DP T cellsignature

As shown in Fig. 3, genes from cluster D are exclusively inducedat high levels at the DP stage. DP thymocytes were characterizedby the expression of genes encoding cell surface proteins such asCd4, Cd8a, CCR9, Cd97, and Blr1/Cxcr5 (Supplemental Table

SIII). Among the transcriptional regulatory genes included in thiscluster, Ets2 is already known to be up-regulated at the DP stageof development (31). The nuclear orphan receptor Rorc/ROR� isan important transcription factor that is tightly regulated during Tcell development: it displays a high expression during the DP stageand is turned off as thymocyte maturation progresses (32). As sug-gested, ROR� may reduce cell division (as does cyclin G2/Ccng2)in developing thymocytes, until they are selected. Strikingly, alarge number of genes involved in TCR-mediated signal transduc-tion were induced at the DP stage, as cells become able to receivesignals through their TCR. Three of them are adaptor moleculesthat mediate intermolecular interactions. Adaptors are now knownto be as crucial for lymphocyte activation as are receptors andeffectors. Sh2d1a/SAP plays a dual functional role both as an adap-tor for FynT and as an inhibitor of SH2 domain-mediated interac-tions (33). Lcp2/Slp76 is an adaptor protein whose deficiency leadsto a severe block at the DN3 stage (34). Interestingly Sla/SLAP,also present in this cluster, associates with Lcp2/Slp76 and hasbeen shown to be highly expressed in DP thymocytes. It appears toplay an important role in positive selection by down-regulating theTCR at the cell surface (35). Two genes related to apoptosis phe-nomena followed the same expression program characteristic ofthe DP stage: the product encoded by Arl6ip/Armer is known toprotect cells from apoptosis, whereas ASC/TMS-1 encodes a pro-apoptotic protein containing a pyrin domain (PD) and a caspase-recruitment domain (CARD). Altogether, this cluster clearly high-lights the setting up of some key components of the TCR signalingmachinery at the DP T cell stage.

Cluster E defines the transcriptional status of single-positivethymocytes

Genes from cluster E revealed the terminal maturation status of SPthymocytes subjected to selection events. This signature presents abias in genes encoding products already known to be involved inthe positive selection of thymocytes (Supplemental Table SIV).This includes notably cell surface markers, components of the sig-nal transduction machinery, transcriptional regulators, and alsoseveral members of the cytoskeletal network. Among highly ex-pressed cell surface molecules are CD5, CD6, CD52, CD53,CCR7, Itgb2, and Itm2a. CD5 and CD53 have been shown to bedevelopmentally regulated and induced by TCR engagement fromthe CD4�CD8�TCRlow cell population stage during selection(36). CD6 is closely related to CD5, and both are physically as-sociated at the immunological synapse. Itm2a is a cell surface-expressed glycoprotein that appears to be transcriptionally up-reg-ulated during positive selection (37). Itgb2 is the �-chain ofLFA-1, the ligand for ICAM1. As for the DP signature, numerousspecific signal transducers are found and may regulate the thresh-old of TCR-mediated signals. ITK is involved in positive selection,presumably by fine-tuning the TCR signals and affecting the effi-ciency with which thymocytes complete their maturation (38).Sh2d2a/Ribp, a T cell adaptor molecule, binds to ITK and regu-lates TCR-mediated signals (39). The tyrosine phosphatase HCPH/Shp-1 regulates the strength of TCR-mediated signals in vivo and,in turn, helps to set the threshold for thymocyte selection (40).Among transcriptional regulatory factors expressed by developingthymocytes, Tox is a member of DNA binding HMG box proteinthat is up-regulated during positive selection of DP thymocytes(41). Egr1 is another transcriptional regulator of interest, becauseit promotes positive selection of both CD4 and CD8 SP cells (42).The Kruppel-like zinc finger transcription factor (Klf2/Lklf) is in-duced during the maturation of SP thymocytes and rapidly down-regulated after lymphocyte activation. It has been proposed tomaintain the selected thymocytes in a quiescent state enhancing

6114 TRANSCRIPTIONAL MICRODISSECTION OF THE MOUSE THYMUS

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

their survival (43). The large representation of genes involved inthe cytoskeleton architecture is another striking feature of thiscluster. Actin cytoskeletal rearrangements play an essential role inshaping the cell response to extracellular signals providing struc-tural frameworks for the spatial reorganization of signaling effec-tors. Components of the cytoskeleton network are found (tubulin�4 (tuba4), myosin (Myh9)) in addition to actin assembly regula-tors (thymosin �10 (Tmsb10), profilin 1 (Pfn1), Vasp, and calpo-nin 2). Interestingly, ITK also regulates TCR/CD3-induced actin-dependent cytoskeletal events (44). Of the 69 genes defining thisSP signature, 15 encode still-uncharacterized proteins that may becrucial regulators of the final steps of thymocyte maturation.

Stromal genes specifically devoted to thymocyte education

Genes from cluster F belong to the stroma signature: they arehighly expressed in thymi from RAG1°, Cd3�°, and LAT° mice,whereas relatively weakly expressed in all of the purified T cellpopulations tested (Fig. 1B). In addition, these genes are not highlyexpressed in all of the thymic epithelial or fibroblastic cell linestested, suggesting that they are likely to correspond to more spe-cific genes that may be devoted to the specialized task of thymo-cyte education. This signature contains 96 genes, and several ofthem have already been described in the thymic stroma (Supple-mental Table SV). This signature includes notably 1) the cell sur-face markers CD83, Ly75/CD205, ALCAM/CD166, VCAM1, andICAM1/CD54, 2) chemokines, TECK/CCL25, SDF1/CXCL12,Cxcl16, CCL5/RANTES, and CXCL10/IP-10, 3) nuclear factors,Whn/Foxn1 and Pax1, 4) proteins known to be involved in anti-genic presentation Ii/CLIP/CD74, H2-Eb1, H2-Aa, H2-Oa, Ca-thepsin L/Ctsl, and CIITA, 5) intermediate filaments, Krt2–8/K8,Krt1–18/K18, Krt2–5/K5, Krt1–17/K17, and finally 6) gene prod-ucts whose function remains to be elucidated, SPATIAL, TSCOT,thymus LIM protein (Tlmp-pending), and PRSS16. Beyond thisquite exhaustive listing of T cell maturation regulators, uncharac-

terized or poorly described genes may be of particular interest forfurther studies. Strikingly, a hierarchical clustering analysis re-stricted to whole thymi samples indicated that this set of genes wasable to distinguish between WT and TCR�° and RelB° thymi (Fig.4A). As minor variations in the high expression values observed inRAG1°, Cd3�°, and LAT° mice strongly influenced gene cluster-ing, these samples were excluded from analysis. This allows us todefine a new signature containing genes that display a low expres-sion level in RelB° and TCR�° mice as compared with WT mice(Fig. 4B, cluster F.1). This signature revealed a high enrichment ofgenes already known to be highly or specifically expressed in themedullary stromal cells (Fig. 4C). Among the most obvious onesare the two cytokeratin genes, Krt2–5 and Krt1–17 (shown in H.sapiens), regulators or components of the Ag presentation machin-ery, C2ta/CIITA and some of its target genes (Invariant chain/Ii,H2-Aa, H2-Eb1, H2-Oa) (45), the cell surface marker ICAM1,CXCL10 (shown in H. sapiens), and the CXCL16 chemokine. Toclearly establish the capability of our procedure to group genesspecific to the medullary stroma compartment, immunofluorescentstainings for some of them were performed. As previously re-ported, Krt2–5/K5 and the invariant chain Ii/CLIP/CD74 are al-most exclusively expressed in the medulla with some scatteredpositive cells in the cortex (Fig. 5, A and B) (46, 47). ConcerningICAM1/CD54, its expression is particularly strong in the medulla,although faint staining is also observed throughout the cortex asLepique et al. (48) recently reported (Fig. 5C). Ccl5/RANTES isalready known to be expressed in the thymus, although its preciseexpression pattern is not yet documented. Of particular interest andas underlined by its transcriptional profile, high levels of Ccl5/RANTES chemokine immunostaining were also found in the med-ullary areas (Fig. 5D). Finally, the weak expression of the genesfrom cluster F.1 in RelB° and TCR�° thymi demonstrates thatcross-talk events with SP thymocytes are required for theirinduction.

FIGURE 4. Genes defining the thymic stroma signature. A, Cluster F is composed of 96 genes that discriminate RAG1°, Cd3�°, and LAT° from RelB°,TCR�°, and WT thymi. Corresponding gene names are given in Supplemental Table SV. B, Exclusion of RAG1°, Cd3�°, and LAT° thymi leads to theidentification of F1 signature. C, The F1 signature is composed of 19 genes expressed in medullary stromal cells.

6115The Journal of Immunology

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

DiscussionAs systematic characterization of the expression profiles of genesacross tissues are currently in progress, data are needed to improveour knowledge concerning their precise expression pattern andtheir putative function (3). In the thymus, notably, the role of nu-merous genes has yet to be elucidated. The use of whole thymialong with purified samples constitutes the major originality of ourapproach. Indeed, analyzing whole organs from genetically engi-neered mice allows us to visualize the transcriptional perturbationsin targeted populations and their effects on cross-talk events withneighboring cells. However, purified cell samples are required toresolve the complexity of some of the signatures identified withtotal organs. As an illustration, purified CD25�DN were necessaryto separate the thymocyte and stromal components of the RAG1°-,Cd3�°-, and LAT°-derived signatures. It also proved to be usefulto delineate specific signatures of poorly represented cell popula-tions. For example, the differences that exist, at the T cell level,between TCR�° and WT mice were more clearly underlined by thesorted SP samples.

Interestingly, this procedure enables us to visualize the coordi-nate expression of components of regulatory networks. As an ex-ample, the DN specific signature contains several transcriptionalregulators, as well as some of their identified targets. Indeed, onceactivated, Notch1 is able to convert the transcription factor CBF1/Su(H) from a repressor to an activator state. Hes1 and Dtx1/Deltex1are two well-known targets of CBF1. Notch1 has also been pro-posed as a Cd25 regulator, although the promoter region of Cd25

does not display canonical CBF1/Su(H) binding sites (49). Be-cause Smarca4/Brg1 is recruited in the promoter region of Cd25(23), this supports the notion that multiple pathways overlap in thiscluster. The transcriptional analysis of Notch1 and Smarca4/Brg1conditional targeting would be of great interest to resolve such acomplexity and to help us to temporally organize these transcrip-tional processes. Importantly, the recent findings of Raya et al.(50), establishing a calcium-dependant activation of the Notch1pathway, is clearly in agreement with the simultaneous expressionof numerous calcium-related genes together with Notch1. Further-more, the high expression level of Hes1 and Dtx1/Deltex1 inRAG1°, CD3�°, and LAT° T cells implies that, in these thymi, themicroenvironment is able to deliver a Notch1 signal. Although anactive Notch1 pathway has been clearly established in early DNthymocytes (51), this underlines the capacity of our approach todetect receptor-ligand activation signals and cellular cross talks.Consequently, in a more systematic analysis of genetically modi-fied mice, this procedure could point out stromal defects in trig-gering developmental signals.

As underlined by the proliferation signature, some of the clus-ters identified include numerous genes involved in specific func-tions. We revealed a cluster containing important genes involvedin TCR expression and V(D)J recombination processes. The co-regulation of genes whose role is still unclear such as the tran-scription factor Runx1/AML1 strongly suggests that they are partof this functional pathway. Runx1 is highly expressed in corticalthymocytes (52). It has been suggested to be crucial for cell fatedetermination during DN to DP transition and to be involved inproliferation events (53). Nevertheless, it has been previously re-ported that TCR-� and TCR-� genes harbor AML1 binding sitesin their enhancer regions (54). Our data support the notion that itcould be involved in TCR rearrangement processes or in the reg-ulation of TCR gene expression. More generally, such functionalclusters may give substantial clues to infer putative functions forpoorly described genes.

One other interesting aspect of this work resides in the visual-ization of the initiation of the thymocyte signaling machinery. Forexample, we could define genes, related to a calcium signalingpathway and specifically expressed at the DN stage (Capn5,FK506-bp5, Eef2k, Camk2a/CaMKII, and Cacna1h). Calciumplays important roles in thymocytes, controlling differentiation,proliferation, and apoptosis, depending on the intensity and theduration of signaling events. Our result is to be linked to the recentfindings suggesting that DN thymocytes exhibit a greater ability tosupport capacitative Ca2� entry than do DP thymocytes (55). Thiscalcium-related property may contribute to the ligand-independentsignaling by pre-TCR complexes. In DN thymocytes, the potentCa2� mobilization induced by the pre-TCR may also lead to thetranscriptional induction of survival molecules (56). DP and SPclusters, as well, are highly represented by molecular componentsof the signal transduction machinery. Because thymocytes are ableto receive signals through their TCR, they may induce a transcrip-tional program leading to the assembly of a specialized transduc-tion system composed of numerous molecular sensors at the cellsurface, of coactivators, and of downstream transducers such asSh2d1a/SAP, Slp76, Sla/SLAP, ITK, Sh2d2a/Ribp, and SHP-1. Inaddition, the SP stage is marked by the expression of proteinsinvolved in cytoskeletal dynamics that may be required to improvethe efficiency of TCR-mediated signals. In this regard, systematictranscriptional analysis of mouse models for each of these com-ponents may lead to the dissection of TCR-mediated signal trans-duction pathways and to a better understanding of �� T cell se-lection processes. Such an approach may help to unravelconflicting results in this area.

FIGURE 5. Immunofluorescence analysis confirms the high medullaryexpression of four genes from cluster F1 in WT C57BL/6 mice (�200).Cryostat sections of whole thymi were stained with anti-K5/Krt2–5 (A),anti-Ii/Cd74 (B), anti-Icam1/Cd54 (C), and anti-Ccl5/RANTES (D). Allsections were counterstained using DAPI to distinguish between medullaryand cortical zones. From left to right, images correspond to Ab staining(left), Ab vs DAPI staining (middle), and DAPI staining (right). Abbrevi-ations: c, cortex; m, medulla; cmj, corticomedullary junction.

6116 TRANSCRIPTIONAL MICRODISSECTION OF THE MOUSE THYMUS

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

In contrast to most of the microarray-based studies of thymocytedevelopment, we considered thymocytes and microenvironment astwo compartments displaying an indissociable and coordinated de-velopment. Of 96 genes enclosed in the stromal signature, severalhave already been described as playing key roles in thymus de-velopment (e.g., Whn/Foxn1, Pax1). In the same way, this clusteralso contained specific genes that may be finely regulated. Spatialexpression is thought to depend on the three-dimensional organi-zation of the thymus, and Tscot expression is lost in epithelial celllines and may require extracellular signals to be maintained (57).The cluster corresponding to medullary stromal cells containsgenes highly expressed in WT as compared with RelB° andTCR�° thymi. As expected, it includes several genes whose ex-pression is known to depend on the RelB/NF-�B pathway in dif-ferent cellular models (RANTES/Ccl5, Cxcl10/IP-10, Icam1/CD54, MHC class II). More surprisingly, like Cd54/Icam1 and theMHC components, the great majority of the genes belonging tothis signature have been shown to be regulated by IFN-� (Ifit-3,Cxcl10/IP-10, Krt1–17/K17, and Oasl2) (10, 58–62). Altogether,these genes exhibit features of tightly regulated genes that reflectthe activated phenotype of UEA1� medullary thymic stromalcells, known to be absent in RelB° mice. Their expression is de-pendent on SP-derived signals, because they are not expressed inTCR�° thymi. Therefore, this cluster pinpoints crucial cross-talkevents that exist in the medulla. Lymphotoxin (LT�1�2) andLIGHT are known to be produced by SP and to deliver activationsignals to medullary stromal cells through the LT� receptor. Theyare presumably required for the transcriptional induction of thesegenes, although a putative role of IFN-� in medullary stromal cellactivation cannot be excluded.

Our analysis both at the RNA and protein levels additionallyshows for the first time that RANTES/Ccl5 expression is restrictedto the medullary stroma. Despite the fact that RANTES has beenshown to be induced in T lymphocytes under some circumstances,its expression in the thymus is predominantly found in the med-ullary stromal cells. In the same way, our data indicate that, as forhuman (63), Cxcl10 is expressed in medullary stromal cells. Thelow level of expression of Cxcl16, Cxcl10/IP-10, and RANTES/Ccl5 in TCR�° mice underscores the mandatory presence of SPthymocytes for the induction of an efficient chemokine network inthe medulla.

The present work underlines the benefit of using models of per-turbed development to describe complex transcriptional processesin the mouse. The combined analysis of a still-limited number ofmodels of transcriptional perturbation is sufficient to summarizethe major phenomena related to thymus function in an integratedway. The identification of functionally relevant signatures willbring functional clues to uncharacterized transcripts coexpressedwith well-known genes. Most importantly, the systematic analysisof KO models should be a very promising method to precisely andextensively microdissect the complex events occurring during Tcell development.

AcknowledgmentsWe are grateful to P. Naquet, N. Auphan, and B. Malissen for criticalreading of the manuscript, and to R. Houlgatte for helpful discussions. Wethank A. Carrier for providing RNA from purified CD25� DN T cells,A. Bourgeois for technical help, and N. Brun and M. Barad for cell sorting.We also thank Dr. B. B. Knowles for kindly giving us thymic stromal celllines.

References1. Su, A. I., M. P. Cooke, K. A. Ching, Y. Hakak, J. R. Walker, T. Wiltshire,

A. P. Orth, R. G. Vega, L. M. Sapinoso, A. Moqrich, et al. 2002. Large-scale

analysis of the human and mouse transcriptomes. Proc. Natl. Acad. Sci. USA99:4465.

2. Bono, H., K. Yagi, T. Kasukawa, I. Nikaido, N. Tominaga, R. Miki, Y. Mizuno,Y. Tomaru, H. Goto, H. Nitanda, et al. 2003. Systematic expression profiling ofthe mouse transcriptome using RIKEN cDNA microarrays. Genome Res.13:1318.

3. Higgins, J. P., L. Wang, N. Kambham, K. Montgomery, V. Mason,S. U. Vogelmann, K. V. Lemley, P. O. Brown, J. D. Brooks, and M. Van De Rijn.2003. Gene expression in the normal adult human kidney assessed by comple-mentary DNA microarray. Mol. Biol. Cell 15:649.

4. Fischer, A., and B. Malissen. 1998. Natural and engineered disorders of lympho-cyte development. Science 280:237.

5. Kishimoto, H., and J. Sprent. 1999. Several different cell surface molecules con-trol negative selection of medullary thymocytes. J. Exp. Med. 190:65.

6. Ghendler, Y., R. E. Hussey, T. Witte, E. Mizoguchi, L. K. Clayton, A. K. Bhan,S. Koyasu, H. C. Chang, and E. L. Reinherz. 1997. Double-positive T cell re-ceptor(high) thymocytes are resistant to peptide/major histocompatibility com-plex ligand-induced negative selection. Eur. J. Immunol. 27:2279.

7. DeRyckere, D., D. L. Mann, and J. DeGregori. 2003. Characterization of tran-scriptional regulation during negative selection in vivo. J. Immunol. 171:802.

8. Hoffmann, R., L. Bruno, T. Seidl, A. Rolink, and F. Melchers. 2003. Rules forgene usage inferred from a comparison of large-scale gene expression profiles ofT and B lymphocyte development. J. Immunol. 170:1339.

9. Schmitz, I., L. K. Clayton, and E. L. Reinherz. 2003. Gene expression analysis ofthymocyte selection in vivo. Int. Immunol. 15:1237.

10. Naquet, P., H. Lepesant, A. Luxembourg, P. Brekelmans, C. Devaux, andM. Pierres. 1989. Establishment and characterization of mouse thymic epithelialcell lines. Thymus 13:217.

11. Shen, Z., G. Reznikoff, G. Dranoff, and K. L. Rock. 1997. Cloned dendritic cellscan present exogenous antigens on both MHC class I and class II molecules.J. Immunol. 158:2723.

12. Faas, S. J., J. L. Rothstein, B. L. Kreider, G. Rovera, and B. B. Knowles. 1993.Phenotypically diverse mouse thymic stromal cell lines which induce prolifera-tion and differentiation of hematopoietic cells. Eur. J. Immunol. 23:1201.

13. Tanaka, T. S., T. Kunath, W. L. Kimber, S. A. Jaradat, C. A. Stagg, M. Usuda,T. Yokota, H. Niwa, J. Rossant, and M. S. Ko. 2002. Gene expression profilingof embryo-derived stem cells reveals candidate genes associated with pluripo-tency and lineage specificity. Genome Res. 12:1921.

14. Diehn, M., G. Sherlock, G. Binkley, H. Jin, J. C. Matese, T. Hernandez-Boussard,C. A. Rees, J. M. Cherry, D. Botstein, P. O. Brown, and A. A. Alizadeh. 2003.SOURCE: a unified genomic resource of functional annotations, ontologies, andgene expression data. Nucleic Acids Res. 31:219.

15. Hayashizaki, Y. 2003. RIKEN mouse genome encyclopedia. Mech. Ageing Dev.124:93.

16. Diehl, F., B. Beckmann, N. Kellner, N. C. Hauser, S. Diehl, and J. D. Hoheisel.2002. Manufacturing DNA microarrays from unpurified PCR products. NucleicAcids Res. 30:e79.

17. Tanaka, T. S., S. A. Jaradat, M. K. Lim, G. J. Kargul, X. Wang, M. J. Grahovac,S. Pantano, Y. Sano, Y. Piao, R. Nagaraja, et al. 2000. Genome-wide expressionprofiling of mid-gestation placenta and embryo using a 15,000 mouse develop-mental cDNA microarray. Proc. Natl. Acad. Sci. USA 97:9127.

18. Verdeil, G., D. Puthier, C. Nguyen, A. M. Schmitt-Verhulst, andN. Auphan-Anezin. 2002. Gene profiling approach to establish the molecularbases for partial versus full activation of naive CD8 T lymphocytes. Ann. NYAcad. Sci. 975:68.

19. Eisen, M. B., P. T. Spellman, P. O. Brown, and D. Botstein. 1998. Cluster anal-ysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA95:14863.

20. Naquet, P., M. Naspetti, and R. Boyd. 1999. Development, organization andfunction of the thymic medulla in normal, immunodeficient or autoimmune mice.Semin. Immunol. 11:47.

21. Stuart, J. M., E. Segal, D. Koller, and S. K. Kim. 2003. A gene-coexpressionnetwork for global discovery of conserved genetic modules. Science 302:249.

22. Kondo, S., H. Kishi, Y. Tokimitsu, and A. Muraguchi. 2003. Possible involve-ment of glial cell line-derived neurotrophic factor and its receptor, GFR�1, insurvival and maturation of thymocytes. Eur. J. Immunol. 33:2233.

23. Yeh, J. H., S. Spicuglia, S. Kumar, A. Sanchez-Sevilla, P. Ferrier, and J. Imbert.2002. Control of IL-2R� gene expression: structural changes within the proximalenhancer/core promoter during T-cell development. Nucleic Acids Res. 30:1944.

24. Welsby, P. J., H. Wang, J. T. Wolfe, R. J. Colbran, M. L. Johnson, andP. Q. Barrett. 2003. A mechanism for the direct regulation of T-type calciumchannels by Ca2�/calmodulin-dependent kinase II. J. Neurosci. 23:10116.

25. Goebel, P., A. Montalbano, N. Ayers, E. Kompfner, L. Dickinson, C. F. Webb,and A. J. Feeney. 2002. High frequency of matrix attachment regions and cut-likeprotein x/CCAAT-displacement protein and B cell regulator of IgH transcriptionbinding sites flanking Ig V region genes. J. Immunol. 169:2477.

26. Kingsley, C., and A. Winoto. 1992. Cloning of GT box-binding proteins: a novelSp1 multigene family regulating T-cell receptor gene expression. Mol. Cell. Biol.12:4251.

27. Barndt, R., M. F. Dai, and Y. Zhuang. 1999. A novel role for HEB downstreamor parallel to the pre-TCR signaling pathway during �� thymopoiesis. J. Immu-nol. 163:3331.

28. Hernandez-Munain, C., and M. S. Krangel. 1995. c-Myb and core-binding factor/PEBP2 display functional synergy but bind independently to adjacent sites in theT-cell receptor � enhancer. Mol. Cell. Biol. 15:3090.

6117The Journal of Immunology

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from

29. Ohbayashi, T., M. Shimada, T. Nakadai, T. Wada, H. Handa, and T. Tamura.2003. Vertebrate TBP-like protein (TLP/TRF2/TLF) stimulates TATA-less ter-minal deoxynucleotidyl transferase promoters in a transient reporter assay, andTFIIA-binding capacity of TLP is required for this function. Nucleic Acids Res.31:2127.

30. Kisielow, J., A. C. Nairn, and K. Karjalainen. 2001. TARPP, a novel protein thataccompanies TCR gene rearrangement and thymocyte education. Eur. J. Immu-nol. 31:1141.

31. Anderson, M. K., G. Hernandez-Hoyos, R. A. Diamond, and E. V. Rothenberg.1999. Precise developmental regulation of Ets family transcription factors duringspecification and commitment to the T cell lineage. Development 126:3131.

32. He, Y. W., C. Beers, M. L. Deftos, E. W. Ojala, K. A. Forbush, and M. J. Bevan.2000. Down-regulation of the orphan nuclear receptor ROR�t is essential for Tlymphocyte maturation. J. Immunol. 164:5668.

33. Li, C., C. Iosef, C. Y. Jia, V. K. Han, and S. S. Li. 2003. Dual functional rolesfor the X-linked lymphoproliferative syndrome gene product SAP/SH2D1A insignaling through the signaling lymphocyte activation molecule (SLAM) familyof immune receptors. J. Biol. Chem. 278:3852.

34. Pivniouk, V., E. Tsitsikov, P. Swinton, G. Rathbun, F. W. Alt, and R. S. Geha.1998. Impaired viability and profound block in thymocyte development in micelacking the adaptor protein SLP-76. Cell 94:229.

35. Sosinowski, T., N. Killeen, and A. Weiss. 2001. The Src-like adaptor proteindownregulates the T cell receptor on CD4�CD8� thymocytes and regulates pos-itive selection. Immunity 15:457.

36. Mitnacht, R., M. Tacke, and T. Hunig. 1995. Expression of cell interaction mol-ecules by immature rat thymocytes during passage through the CD4�8� com-partment: developmental regulation and induction by T cell receptor engagementof CD2, CD5, CD28, CD11a, CD44 and CD53. Eur. J. Immunol. 25:328.

37. Kirchner, J., and M. J. Bevan. 1999. ITM2A is induced during thymocyte selec-tion and T cell activation and causes downregulation of CD8 when overexpressedin CD4�CD8� double positive thymocytes. J. Exp. Med. 190:217.

38. Lucas, J. A., L. O. Atherly, and L. J. Berg. 2002. The absence of Itk inhibitspositive selection without changing lineage commitment. J. Immunol. 168:6142.

39. Rajagopal, K., C. L. Sommers, D. C. Decker, E. O. Mitchell, U. Korthauer,A. I. Sperling, C. A. Kozak, P. E. Love, and J. A. Bluestone. 1999. RIBP, a novelRlk/Txk- and itk-binding adaptor protein that regulates T cell activation. J. Exp.Med. 190:1657.

40. Carter, J. D., B. G. Neel, and U. Lorenz. 1999. The tyrosine phosphatase SHP-1influences thymocyte selection by setting TCR signaling thresholds. Int. Immu-nol. 11:1999.

41. Wilkinson, B., J. Y. Chen, P. Han, K. M. Rufner, O. D. Goularte, and J. Kaye.2002. TOX: an HMG box protein implicated in the regulation of thymocyteselection. Nat. Immunol. 3:272.

42. Bettini, M., H. Xi, J. Milbrandt, and G. J. Kersh. 2002. Thymocyte developmentin early growth response gene 1-deficient mice. J. Immunol. 169:1713.

43. Kuo, C. T., M. L. Veselits, and J. M. Leiden. 1997. LKLF: a transcriptionalregulator of single-positive T cell quiescence and survival. Science 277:1986.

44. Grasis, J. A., C. D. Browne, and C. D. Tsoukas. 2003. Inducible T cell tyrosinekinase regulates actin-dependent cytoskeletal events induced by the T cell antigenreceptor. J. Immunol. 170:3971.

45. Chang, C. H., and R. A. Flavell. 1995. Class II transactivator regulates the ex-pression of multiple genes involved in antigen presentation. J. Exp. Med.181:765.

46. Klug, D. B., C. Carter, E. Crouch, D. Roop, C. J. Conti, and E. R. Richie. 1998.Interdependence of cortical thymic epithelial cell differentiation and T-lineagecommitment. Proc. Natl. Acad. Sci. USA 95:11822.

47. Farr, A., P. C. DeRoos, S. Eastman, and A. Y. Rudensky. 1996. Differentialexpression of CLIP:MHC class II and conventional endogenous peptide:MHCclass II complexes by thymic epithelial cells and peripheral antigen-presentingcells. Eur. J. Immunol. 26:3185.

48. Lepique, A. P., S. Palencia, H. Irjala, and H. T. Petrie. 2003. Characterization ofvascular adhesion molecules that may facilitate progenitor homing in the post-natal mouse thymus. Clin. Dev. Immunol. 10:27.

49. Adler, S. H., E. Chiffoleau, L. Xu, N. M. Dalton, J. M. Burg, A. D. Wells,M. S. Wolfe, L. A. Turka, and W. S. Pear. 2003. Notch signaling augments T cellresponsiveness by enhancing CD25 expression. J. Immunol. 171:2896.

50. Raya, A., Y. Kawakami, C. Rodriguez-Esteban, M. Ibanes, D. Rasskin-Gutman,J. Rodriguez-Leon, D. Buscher, J. A. Feijo, and J. C. Izpisua Belmonte. 2004.Notch activity acts as a sensor for extracellular calcium during vertebrate left-right determination. Nature 427:121.

51. Harman, B. C., E. J. Jenkinson, and G. Anderson. 2003. Entry into the thymicmicroenvironment triggers Notch activation in the earliest migrant T cell pro-genitors. J. Immunol. 170:1299.

52. Lorsbach, R. B., J. Moore, S. O. Ang, W. Sun, N. Lenny, and J. R. Downing.2003. Role of Runx1 in adult hematopoiesis: analysis of Runx1-IRES-GFPknock-in mice reveals differential lineage expression. Blood 103:2522.

53. Sato, T., R. Ito, S. Nunomura, S. Ohno, K. Hayashi, M. Satake, and S. Habu.2003. Requirement of transcription factor AML1 in proliferation of developingthymocytes. Immunol. Lett. 89:39.

54. Kim, W. Y., M. Sieweke, E. Ogawa, H. J. Wee, U. Englmeier, T. Graf, and Y. Ito.1999. Mutual activation of Ets-1 and AML1 DNA binding by direct interactionof their autoinhibitory domains. EMBO J. 18:1609.

55. Haks, M. C., S. M. Belkowski, M. Ciofani, M. Rhodes, J. M. Lefebvre, S. Trop,P. Hugo, J. C. Zuniga-Pflucker, and D. L. Wiest. 2003. Low activation thresholdas a mechanism for ligand-independent signaling in pre-T cells. J. Immunol.170:2853.

56. Aifantis, I., F. Gounari, L. Scorrano, C. Borowski, and H. von Boehmer. 2001.Constitutive pre-TCR signaling promotes differentiation through Ca2� mobiliza-tion and activation of NF-�B and NFAT. Nat. Immunol. 2:403.

57. Kim, M. G., F. A. Flomerfelt, K. N. Lee, C. Chen, and R. H. Schwartz. 2000. Aputative 12 transmembrane domain cotransporter expressed in thymic corticalepithelial cells. J. Immunol. 164:3185.

58. Kakuta, S., S. Shibata, and Y. Iwakura. 2002. Genomic structure of the mouse2�,5�-oligoadenylate synthetase gene family. J. Interferon Cytokine Res. 22:981.

59. Singer, K. H., S. M. Denning, L. P. Whichard, and B. F. Haynes. 1990. Thymo-cyte LFA-1 and thymic epithelial cell ICAM-1 molecules mediate binding ofactivated human thymocytes to thymic epithelial cells. J. Immunol. 144:2931.

60. Smith, J. B., and H. R. Herschman. 1996. The glucocorticoid attenuated responsegenes GARG-16, GARG-39, and GARG-49/IRG2 encode inducible proteins con-taining multiple tetratricopeptide repeat domains. Arch. Biochem. Biophys.330:290.

61. Vanguri, P., and J. M. Farber. 1990. Identification of CRG-2: an interferon-inducible mRNA predicted to encode a murine monokine. J. Biol. Chem.265:15049.

62. Vogel, U., B. Denecke, S. M. Troyanovsky, R. E. Leube, and E. C. Bottger. 1995.Transcriptional activation of psoriasis-associated cytokeratin K17 by interfer-on-�: analysis of �-interferon activation sites. Eur. J. Biochem. 227:143.

63. Gattass, C. R., L. B. King, A. D. Luster, and J. D. Ashwell. 1994. Constitutiveexpression of interferon-�-inducible protein 10 in lymphoid organs and inducibleexpression in T cells and thymocytes. J. Exp. Med. 179:1373.

64. Lopez, F., J. Rougemont, B. Loriod, A. Bourgeois, L. Loi, F. Bertucci, P.Hingamp, R. Houlgatte, and S. Granjeaud. 2004. Feature extraction and signalprocessing for nylon DNA microarrays. BMC Genomics 5:38.

6118 TRANSCRIPTIONAL MICRODISSECTION OF THE MOUSE THYMUS

by guest on June 15, 2015http://w

ww

.jimm

unol.org/D

ownloaded from