Molecular Dissection of Prethymic Progenitor Entry into the T Lymphocyte Developmental Pathway

This information is current as C. Chace Tydell, Elizabeth-Sharon David-Fung, Jonathan E. of October 1, 2021. Moore, Lee Rowen, Tom Taghon and Ellen V. Rothenberg J Immunol 2007; 179:421-438; ; doi: 10.4049/jimmunol.179.1.421 http://www.jimmunol.org/content/179/1/421 Downloaded from

Supplementary http://www.jimmunol.org/content/suppl/2008/03/13/179.1.421.DC1 Material

References This article cites 76 articles, 38 of which you can access for free at: http://www.jimmunol.org/ http://www.jimmunol.org/content/179/1/421.full#ref-list-1

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists by guest on October 1, 2021 • Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2007 by The American Association of Immunologists All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Molecular Dissection of Prethymic Progenitor Entry into the T Lymphocyte Developmental Pathway1

C. Chace Tydell,2 Elizabeth-Sharon David-Fung,2,3 Jonathan E. Moore, Lee Rowen,4 Tom Taghon,5 and Ellen V. Rothenberg6

Notch signaling activates T lineage differentiation from hemopoietic progenitors, but relatively few regulators that initiate this program have been identified, e.g., GATA3 and T cell factor-1 (TCF-1) (gene name Tcf7). To identify additional regulators of T cell specification, a cDNA library from mouse Pro-T cells was screened for genes that are specifically up-regulated in intrathymic T cell precursors as compared with myeloid progenitors. Over 90 genes of interest were iden- tified, and 35 of 44 tested were confirmed to be more highly expressed in T lineage precursors relative to precursors of B and/or myeloid lineage. To a remarkable extent, however, expression of these T lineage-enriched genes, including zinc finger transcription factor, helicase, and signaling adaptor genes, was also shared by stem cells (Lin؊Sca-1؉Kit؉CD27؊) and Downloaded from multipotent progenitors (Lin؊Sca-1؉Kit؉CD27؉), although down-regulated in other lineages. Thus, a major fraction of these early T lineage genes are a regulatory legacy from stem cells. The few genes sharply up-regulated between multipotent progenitors and Pro-T cell stages included those encoding transcription factors Bcl11b, TCF-1 (Tcf7), and HEBalt, Notch target Deltex1, Deltex3L, Fkbp5, Eva1, and Tmem131. Like GATA3 and Deltex1, Bcl11b, Fkbp5, and Eva1 were dependent on Notch/Delta signaling for induction in fetal liver precursors, but only Bcl11b and HEBalt were up-regulated between the first two stages of intrathymic T cell development (double negative 1 and double negative 2) corresponding to T lineage http://www.jimmunol.org/ specification. Bcl11b was uniquely T lineage restricted and induced by Notch/Delta signaling specifically upon entry into the T lineage differentiation pathway. The Journal of Immunology, 2007, 179: 421–438.

he circulating population of mature T lymphocytes is con- cell type, loss of the transcription factor causes a selective block of stantly regenerated as hemopoietic progenitors leave the the developmental pathway while gain of function of the transcrip- T bone marrow (BM)7 and home to the thymus where de- tion factor can accelerate differentiation into the lineage. Equiva- velopment and maturation occur (1–3). Cell-intrinsic regulatory lent “master regulators” of T cell development have not yet been factors that are up-regulated in a lineage-specific way play domi- found. More than eight known transcription factors are essential nant roles in lineage choice of hemopoietic precursors. In RBC for T cell development (reviewed in Refs. 4 and 5), but none of by guest on October 1, 2021 development, GATA1 acts as a central mediator of erythroid gene these exhibits the ability to instruct or accelerate entry into the T expression, and it is known that B cells are instructed specifically cell program. by early B cell transcription factor and Pax5. These transcription Two transitions in early T cell development are of particular factors can be considered “master regulators” because, for each interest for lineage choice mechanisms: the onset of T lineage gene expression (“specification”), and the final exclusion of any fate except a T cell fate (“commitment”). Both transitions occur among Division of Biology, California Institute of Technology, Pasadena, CA 91125 intrathymic, early T lineage cell populations (Pro-T cells), which Received for publication December 18, 2006. Accepted for publication April are still negative for the mature T cell markers CD4 and CD8 and 16, 2007. do not yet express TCRs. These double-negative (DN) cells within The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance the mouse thymus are divided into four stages on the basis of their with 18 U.S.C. Section 1734 solely to indicate this fact. expression of the surface markers Kit, CD25, and CD44 reviewed 1 This work was supported by grants (to E.V.R.) from the National Science Founda- in Refs. 6–8. The first thymocyte population, DN1, maintains tion (MCB-9983129) and National Institutes of Health U.S. Public Health Service much lineage plasticity and, under special conditions, is capable of (R01 CA90233 and R01 CA98925), by National Institutes of Health U.S. Public Health Service awards K08 AI054699 (to C.C.T.) and F32 AI068366 (to J.E.M.); and producing macrophages, NK cells, or dendritic cells, with a minute from the DNA Sequencer Royalty Fund at the California Institute of Technology. subset apparently capable of generating B cells as well (9–12). It 2 C.C.T. and E.-S.D.-F. contributed equally to this study. is not yet clear whether DN1 cells are distinct from prethymic 3 Current address: Theoretical Division, Los Alamos National Laboratory, Los precursors in gene expression pattern. However, as cells enter the Alamos, NM 87544. next stage, DN2, they express sharply increased levels of Pro-T 4 Current address: Institute for Systems Biology, Seattle, WA 98103. cell genes such as those encoding pT␣, CD3␧, CD25, Rag1, and 5 Current address: Department of Clinical Chemistry, Microbiology and Immunology, IL-7R␣ (CD127) (10, 13–16), and some rearrangement begins at University Hospital Ghent, Ghent University, Ghent, Belgium. the DJ␤ and VJ␥ TCR loci (17, 18). In hemopoietic precursors 6 Address correspondence and reprint requests to Dr. Ellen V. Rothenberg, Division of Biology 156-29, California Institute of Technology, Pasadena, CA 91125. E-mail (derived from fetal liver) that are differentiating in vitro in re- address: [email protected] sponse to Notch/Delta signaling, the first-appearing DN2 pheno- 7 Abbreviations used in this paper: BM, bone marrow; DN, double negative; Dtx3L, type cells display the same dramatic increase in expression of these Deltex3-like; pT␣, pre-TCR␣; qRT-PCR, quantitative real-time PCR; TCF-1, T cell genes (19). DN2 cells have undergone “specification” but are not factor-1. yet committed to the T lymphocyte pathway; a high proportion of Copyright © 2007 by The American Association of Immunologists, Inc. 0022-1767/07/$2.00 DN2 cells are still able to differentiate into NK cells, macrophages, www.jimmunol.org 422 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

or dendritic cells (10, 20–23). At the DN3 stage, thymocytes stop Thymus and BM samples were taken from animals 5–7 wk old. The dividing, further increase expression of the Pro-T differentiation animals used were bred and maintained under sterile conditions at Caltech. genes as well as Notch target genes (24), and undergo extensive cDNA library TCR rearrangements. Only at this stage do they become committed to a T cell fate in vivo. Cells only progress beyond the DN3 stage The C.B-17-scid thymocyte, random-primed, cDNA library was con- through successful TCR gene rearrangement and TCR-dependent structed in the pSPORT1 vector (Invitrogen Life Technologies), and was arrayed and spotted at high density onto Hybond-Nϩ nylon filters selection, at which time they graduate from Pro-T cell status, to (Amersham Biosciences) using the Q-BOT robot (Genetix) as described give rise to up to five types of T cells: ␣␤ CD4, ␣␤ CD8, ␥␦, NKT, previously (31). or regulatory T cells. Of all these developmental transitions, surprisingly little is Cell populations for library generation and library screening known about the stages encompassing “T lineage specification,” The two types of cells used as sources of RNA for the subtraction protocol that is, the DN1 to DN2 transition. The regulatory participants in were a bulk population of Pro-T cells and a population of progenitor/ these early stages have not been sufficiently characterized to ex- premyeloid cells. To obtain large numbers of Pro-T cells in the DN1-DN3 plain the outcome, although Notch/Delta signaling plays a role (19, stages, we took advantage of the Rag2 knockout mouse, in which thymo- cyte development arrests at DN3. In the wild-type mouse thymus, DN3 25). Traditional methods to identify all the transcription factors cells account for only 1% of thymocytes, but even without sorting, Rag2 that play key roles in early stages of T lineage specification have knockout thymocytes consist of 90% DN3 Pro-T cells, with the remaining found limited success, in part because transcription factors are typ- cells being DN1, DN2, NK, or thymic stromal cells. Because these cells were not pure Pro-T cells, we refer to them in the text as “Pro-T plus.” A ically present in low copy numbers. T cell precursors at the earliest Ϫ ϩ stages are represented in vivo at any one time by tiny numbers of progenitor/premyeloid population was obtained from Lin Kit BM cells

matured in culture toward a myeloid fate. Specifically, BM cells that were Downloaded from cells, providing very limiting material for standard microarray KitϩGr1ϪCD11bϪTer119ϪCD19Ϫ from Rag2ko mice were cultured for

analysis. In addition, small fold changes in transcription factor 42 h at 37°C in 5% CO2 in IMDM with 10% heat-inactivated FCS sup- abundance or changes in transcription factor ratios may generate plemented with IL-3 (100 ␮l of WEHI-3B cell supernatant/ml medium), stem ␮ dramatic shifts in cell state. Identification of truly novel regulatory cell factor (Kit ligand) (100 l of BHK-MKL cell supernatant/ml medium), and 10 ng/ml rIL-6 (PeproTech). At the time of harvest, the morphological factors and novel isoforms of known factors has also been limited appearance of the cells ranged from undifferentiated, blast-like cells to mature by the comprehensiveness of the microarrays used (26–29) and by granulocytes. the microarray’s nucleotide probe design. http://www.jimmunol.org/ To circumvent these problems, we used a subtractive hybrid- Cell populations for quantitative RNA expression analysis ization technique (30) to probe a mouse Pro-T cell cDNA mac- In addition to the subtraction populations, sorted cell populations were roarray library. The subtractive technique allowed us to enrich a obtained for analysis by qRT-PCR. Two hemopoietic progenitor popula- Pro-T cell probe for message not shared by progenitor/premyeloid tions, a Pro-B population, and a population of myeloid cells were all sorted ko Ϫ ϩ ϩ Ϫ cells and to identify genes (known and previously unknown) that from the BM of Rag2 mice. LSK CD27 cells were Kit Sca-1 CD27 Gr1ϪCD11bϪTer119ϪCD19Ϫ. LSK CD27ϩ cells were KitϩSca-1ϩCD27ϩ might be specifically up-regulated during the initiation of the T Gr1ϪCD11bϪTer119ϪCD19Ϫ. Pro-B cells were CD19ϩKitϩ/ϪGr1ϪCD11bϪ lineage program. Clones selected by the subtraction were se- Ter119Ϫ. Sorted BM myeloid cells were Gr1ϩCD11bϩKitϪTer119Ϫ. quenced and their patterns of expression characterized in detail Initial expression screening made use of the unsorted Rag2ko thymocytes by guest on October 1, 2021 using sensitive, quantitative real-time PCR (qRT-PCR) on a range called Pro-T plus. For more detailed analysis, five populations of DN Pro-T cells were purified from C57BL/6 mouse thymi by cell sorting, essentially as of highly purified cell populations. The Pro-T cell library was gen- Ϫ Ϫ Ϫ described previously (13, 24). Each of these DN subsets is CD4 CD8 CD3␧ erated by random priming of mRNA from SCID thymocytes, Ter119ϪF4/80Ϫ and Gr1Ϫ. DN1 cells are KitϩCD44ϩCD25Ϫ. DN2 cells are which consist of DN1–3 stage Pro-T cells and NK precursors (31). KitϩCD44ϩCD25ϩ. DN3a cells are KitlowCD44ϪCD25ϩCD27Ϫ. DN3b cells Ϫ ϩ ϩ Ϫ Ϫ The resulting macroarray library represents the actual spectrum of are CD44 CD25 CD27 , and DN4 cells are CD44 CD25 . All Abs used in transcripts in the DN2 and DN3 cell populations (those immedi- this study were obtained from eBioscience or BD Pharmingen. Thymocytes and BM cells were obtained from animals immediately ately preceding lineage commitment) and also contains multiple after euthanasia. Cells were incubated in CBSS (5.4 mM KCl, 0.3 mM

clones of genes with abundant transcripts, providing opportunities Na2HPO4, 0.4 mM KH2PO4, 4.2 mM NaHCO3, 137 mM NaCl, and 5.6 to sample alternate splice variants. mM D-glucose (pH 7.4))/1% BSA with clone 2.4G2 anti-CD32/CD16 Our subtraction protocol has identified genes that are specifi- (Fc␥RIII/II) supernatant on ice for 10 min, followed by washing and ad- cally enriched in T-lineage as opposed to early B or myeloid lin- dition of Abs for staining. Cells stained with biotin-conjugated Abs were washed through a layer of FCS before staining with streptavidin-PECy5 eage precursors. Enriched genes include novel transcription factor (eBioscience). Stained cells were sorted using FACS Aria cell sorter (BD candidates, chromatin remodeling factors, RNA binding molecules Immunocytometry Systems). All sorted fractions were reanalyzed imme- and helicases, a select group of signaling molecules and adaptors, diately for purity and all fractions used here were at least 96% pure. and novel or functionally uncharacterized genes. In this study, we Preparation of fetal liver cells as input for OP9-DL1 and OP9-control cultures is described below. present the resulting expression profiles of Ͼ35 of these genes expressed during T lineage specification. A remarkable feature of Generation of subtractive probe the whole ensemble of these Pro-T cell genes is the high frequency A Pro-T plus unselected cDNA probe and a T lineage-enriched subtracted of “legacy” genes that are expressed strongly in both stem cells probe were generated from the pool of enriched Pro-T cells (ϳ1 ϫ 107 and Pro-T cells, although down-regulated elsewhere. In this con- cells; Pro-T plus) and the population of myeloid and multipotent progen- text, the very select group of regulatory genes that are specifically itors (ϳ9 ϫ 106 cells) described above. RNA was isolated from each popu- induced coincidentally with T lineage specification takes on an lation of cells by the Qiagen RNeasy Lipid Minikit (Qiagen), and the RNA was unexpected significance. analyzed for purity and integrity by Agilent Bioanalyzer and RNA 6000 Nano- chips (Agilent). Messenger RNA was isolated from total RNA by the Ambion Poly(A) Purist Kit (Ambion). The mRNA was also evaluated for purity and Materials and Methods concentration by the Bioanalyzer. Mice The subtraction method was adapted from Rast et al. (30). cDNA was synthesized using enzymes and buffers from a Clontech Marathon cDNA C57BL/6J, B6.CB17-Prkdcscid/SzJ (B6-scid) (The Jackson Laboratory) synthesis kit (BD Biosciences), but with the LT7 random-BT primer: 5Ј- and B6-Rag2null mice (originally from E. Palmer, Basel Institute, Basel, (biotin)-CGGAGGTAATACGACTCACTATAGGGAGNNNNNN-3Ј (34 Switzerland) were bred and maintained in specific pathogen-free facilities nt). Qiagen PCR purification columns were used to purify samples between at Caltech. first-strand and second-strand synthesis stages. The Journal of Immunology 423

Different linkers are used for Selectate (the Pro-T plus-derived cDNA) number of spot pairs on a blot, or 768; other reasonable values of J do not or Driver (the progenitor/myeloid-derived cDNA) to avoid nonspecific appreciably change the estimates. An integer, I, was chosen so as to min- subtraction. Linkers contain a 3Ј dideoxy residue to prevent filling in of imize the difference between the Ith and (I ϩ J)th logarithms, and the mode overhang, and a 5Ј phosphate for blunt ended ligation to the cDNA. The was estimated by averaging these. Selectate linker sequences were 5Ј-GGGTGCTGTATTGTGTACTTGAA Clones selected by this method were sequenced from both the 5Ј and the CGGGCGGCCGCA-3Ј and 3Ј-dideoxy-CGCCCGCCGGCGT-P-5Ј. The 3Ј ends at the Institute for Systems Biology using standard procedures Driver linker sequences were 5Ј-GCCAACGTATGTAAGGTTGAGTTC designed for the Applied Biosystems 3730XL sequencer. Each sequence CGGGCAGGT-3Ј and 3Ј-dideoxy-CCCGTCCA-P-5Ј. Linkers were an- was filtered by RepeatMasker and analyzed by Blastn, Blastx, and BLAT nealed by placing the linker pairs in a 1:1 molar ratio at a concentration searches of National Center for Biotechnology Information (NCBI), Uni- totaling 1 ␮g/␮l in 10 mM Tris (pH 7.9) and 100 mM NaCl, heating in a versity of California, Santa Cruz (UCSC), and Ensembl databases to iden- heating block to 95°C for 5 min, then turning off the block and allowing the tify matches to known genes or genomic sequences (latest searches: build linkers to anneal as the block cools to room temperature (ϳ30 min). Li- m36 of mouse genome). Of the 1164 clones submitted, 1046 good se- gation efficiency was evaluated by comparing their electrophoresis migra- quences were obtained. tion on a 2% agarose gel relative to the untreated mixture. These linkers were ligated to either the Selectate or Driver cDNA for 16 h at 16°C with GOToolBox analysis DNA ligase. The GO-Stats function of the GOToolBox website (35) was used to per- cDNA with linkers attached was PCR amplified with LT7 primer (5Ј- form a hypergeometric analysis of statistically relevant over- or under- CGGAGGTAATACGACTCACTATAGG-3Ј) and a primer specific for ei- represented terms within our data set as compared with the Mouse Genome ther the Selectate linker (5Ј-GGGTGCTGTATTGTGTACTTGAACG-3Ј) Informatics database of genes. The Benjamini & Hochberg correction for or for the Driver linker (5Ј-GCCAACGTATGTAAGGTTGAGTTCC-3Ј) multiple testing was applied. Selected results of searches in the categories to produce 600 ng of product. The resulting Selectate was size selected for of Biological Processes and Cellular Components are reported herein. 300- to 500-bp product by electrophoresis of the PCR in an agarose gel, excising the appropriate region of the gel, and electroeluting cDNA from

Bioinformatic databases Downloaded from the gel. The electroeluate was precipitated and resuspended in 50 ␮lof water or T low E (10 mM Tris and 0.1 mM EDTA, (pH 7.8). Driver The following databases were used: www.ncbi.nlm.nih.gov, http:// product was precipitated and resuspended in 16 ␮l of water. genome.ucsc.edu, www.ensembl.org/Mus_musculus, http://crfb.univ-mrs. Size-selected Selectate was amplified by PCR (primers listed above), fr/GOToolBox/index.php, and www.informatics.jax.org. and 1 ␮g of Selectate was set aside for production of the unsubtracted library probe. Selectate (3 ␮g) was subjected to single-strand purification Coculture of fetal liver cells with OP9 cells by Dynal Streptavidin beads, according to the manufacturer’s instructions Hemopoietic progenitors cocultured with BM stromal cells (OP9 cell line)

(Invitrogen Life Technologies). RNase-free technique was used from this http://www.jimmunol.org/ will develop into B lymphocytes in vitro. When OP9 stromal cells are point forward for both Selectate and Driver. The Ambion MEGAshortscript kit transfected to stably express the Notch ligand Delta-like 1 (OP9-DL1), (Ambion) was used according to the manufacturer’s instructions to trans- progenitor cells will develop into T lymphocytes in coculture (36). Mouse late the Driver cDNA to RNA. Single-stranded Selectate DNA (200 ng) fetal liver cells (containing hemopoietic progenitors) were cocultured with from Pro-T plus cells was mixed with Driver RNA (30 ␮g) in a 10-␮l total OP9-control or OP9-DL1 cells exactly as described previously (19). In volume, denatured at 95°C, iced, then hybridized at 65°C for 40 h. Double- ϩ Ϫ short, Kit Lin (Lin ϭ Gr1, Ter119, F4/80, CD19) cells from day 14 to stranded and single-stranded products were separated by hydroxylapatite ϩ Ϫ 14.5 mouse embryo livers were obtained by FACS sorting. Kit Lin fetal chromatography (30, 32), and the eluate containing the single-stranded liver cells were cocultured with OP9 control cells or with OP9-DL1 cells product was desalted and concentrated. The single-stranded product was used (36). Cocultures were harvested for RNA analysis by forceful pipetting at to manufacture the subtracted, radioactive probe by Ambion Maxiscript kit indicated time points. In some experiments, to test the effect of delayed using Amersham 800 Ci/mM 32P-UTP.

addition or withdrawal of Notch signals, progenitor cells were transferred by guest on October 1, 2021 Subtractive hybridization protocol to secondary cultures at day 4. OP9-control and OP9-DL1 cocultures were harvested, KitϩCD27ϩLinϪ cells were isolated from each culture by sort- Macroarray filters were sequentially hybridized with cDNA from Pro-T ing, and these were each split and used to seed fresh monolayers of OP9 plus Rag2ko thymocytes, stripped of probe, then hybridized with subtracted control and OP9-DL1 cells, to be harvested at the indicated later time probe, i.e., probe enriched for mRNA that was not shared by the progen- points. itor/premyeloid population. Hybridization intensity for each probe was To compare the time courses of T lineage differentiation from distinct pre- measured by a PhosphorImager (Molecular Dynamics, GE Healthcare) using cursor subsets, KitϩLinϪ fetal liver cells were fractionated into KitϩSca-1ϩ BioArray software (Genetix). Representative data are shown in Fig. 1. (“LSK”), KitϩSca-1lowCD27ϩFlk2/Flt3 (CD135)ϩIL-7R␣(CD127)Ϫ (“Flkϩ”), On the macroarray blot, each clone is applied in duplicate as a “spot and KitϩSca-1lowCD27ϩCD135ϩCD127ϩ (“CLP-like”) subsets, as de- pair” with unique position and arrangement (see Fig. 1, A and B). Pixel scribed elsewhere (T. Taghon, M. A. Yui, and E. V. Rothenberg, submitted intensity for each spot pair was averaged, and systematic differences in for publication). These subsets were then cocultured with OP9-control or hybridization intensity across the blots were minimized by the application OP9-DL1 cells for 1–7 days before harvesting for RNA. of a whitening filter, much like the linear optimal filter or Wiener filter (33, 34). Its formula is: Quantitative real-time PCR

⌽͑ f ͒ ϭ S͑ f ͒2 / ͑S͑ f ͒2 ϩ N͑ f ͒2) qRT-PCR was performed on diluted samples of cDNA using SYBR Green PCR Master Mix in an ABI PRISM 7700 Sequence Detector (Applied where ⌽( f) is the whitening filter, S( f) is the signal in Fourier space, and Biosystems). In all figures comparing expression levels of multiple genes, N( f) is the noise in Fourier space. A Wiener filter is often used to filter the measurements for all genes shown were conducted on the same cDNA ⌬ random, usually small-scale, noise from data leaving mostly the large-scale samples. The Ct method was used for all expression measurements, with correlations; this is its normal use in image analysis. However, because our a fixed threshold to enable direct comparison between test genes and the clones were randomly placed on the blots, the only correlations expected in GAPDH standard. Primers were designed, using Primer3 software (37), to our data are the relatively small-scale ones between spot pairs. Hence, all have optimal melting temperatures and to cross introns. The primers were large-scale correlations are likely due to systematic noise such as inhomo- BLAST tested for gene specificity before being synthesized (Operon geneous probing or washing of the blots. Therefore, we used the whitening Biotechnologies). Each primer pair was evaluated for acceptable dose- filter to remove such correlations (J. E. Moore, unpublished results). response titration slopes and amplification. Primer sequences are as The ratio of the average spot-pair hybridization intensity before and follows: Ablim, forward, CTGGCAGCTCAGAGGAGTTC, and re- after subtraction was termed enrichment: (intensity for subtracted probe) Ϭ verse, CGCAGCTGGGATGATAATG; Aff3, forward, CAACAGAGA (intensity for the unsubtracted probe) ϭ enrichment. The logarithms of the GCAGCGCAACA, and reverse, CCCGTCTCCATATTGCACACTT; enrichments for 73,728 spot pairs were calculated, and a clone was deemed AI449175, forward, GCTCCTTCCCAGAAGACTCTC, and reverse, T to be significantly enriched when the logarithm of its enrichment was more CAGGCTCTTCAAAATGGTCTT; Akap8, forward, AAATTGAGAA than three SDs above the mode of its blot (Fig. 1C). Clones more than four ACGGCGTCAG, and reverse, AATGTGCGGCTTCAATCTTT; ␤-ac- SDs above the mode were selected for special attention (see Table I). The tin, forward, ACACCCGCCACCAG, and reverse, TACAGCCCGGGG modes were calculated by a process called “estimating the rate of an in- AG; Bat2, forward, ATACTGCCACAAGCCGAAAG, and reverse, T homogeneous Poisson process by Jth waiting times” (33), briefly outlined CAGGTCCACTCCACTGTCA; Bcl11a, forward, GTCTGCACACGG here. For each blot, the logarithms of the enrichments were ordered. A AGCTCTAA, and reverse, CACTGGTGAATGGCTGTTTG; Bcl11b, window size, J, was chosen, which for these calculations was (1)/24 the forward, GGGCGATGCCAGAATAGAT, and reverse, GGTAGCCTC 424 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

Table I. Other zinc finger proteins

Total Clones Enrichment qRT-PCR Comments

Exon matched Genes known to be up-regulated in T development IL-2ra 124 clones ‡ IL-2R, ␣ chain. regulation of T cell proliferation CD3g 61 clones ‡ CD3 Ag, ␥ chain. Component of T cell complex Tcrg 53 clones ‡ TCR ␥ chain CD3e 34 clones ‡ CD3 ⑀ chain Tcrb 22 clones ‡ TCR ␤ chain Tcf7 (TCF-1) 18 clones ‡ A,b B Transcription factor 7, aka TCF-1, T cell specific Thy1 17 clones ‡ Thymus cell Ag 1, ␪ Lat 5 clones ‡ Linker for activation of T cells, receptor-signaling protein Dntt 5 clones ‡ Deoxynucleotidyltransferase, terminal. Also known as Tdt Ly6a 5 clones ‡ Lymphocyte Ag 6 complex, locus A, aka Sca-1. Lck 3 clones ‡ Lymphocyte protein tyrosine kinase is involved in regulation of the TCR-signaling pathway Dpp4 2 clones ‡ Dipeptidylpeptidase 4, aka CD26 Cd24a ‡ HSA or Ly52 RNA binding Ddx17 † B DEAD box helicase protein 17, p72, reported to regulate transcription Downloaded from Ddx19b ‡ B, C DEAD box helicase protein 19b FUS 2 clones † A, B Pigpen or Fusion derived from t(12;16) malignant liposarcoma, transcriptional activator Helz † B, C Helicase with zinc finger domain Other zinc finger proteins Rad18 ‡ Ring-type zinc finger protein, DNA repair activity

Trim44 ‡ A, B Tripartite motif-containing 44, B-box domain http://www.jimmunol.org/ Lim-related proteins Ablim1 ‡ B Actin-binding LIM protein 1, deletions are linked to cancer Ctdsp1 ‡ A, B A small phosphatase with a NLI, nuclear LIM interacting, domain Ldb1 2 clones ‡ A, B LIM domain-binding 1, may interact with Scl in thrombopoiesis Protein metabolism, signaling, kinases, and ubiquitination Akap8 ‡ B A kinase (PRKA) anchor protein 8 Apoe 2 clones ‡ Apolipoprotein E, lipid transport and Ca2ϩ homeostasis Atxn2l † A, B Ataxin 2 spinocerebellar ataxia type II, may be part of by guest on October 1, 2021 cytokine-signaling system Ccnk † Cyclin K, may play a dual role in regulating CDK and RNA polymerase II activities Ddef1 † Development and differentiation enhancing 1, predicted to have transcriptional activity Dpp8 ‡ Dipeptidylpeptidase 8 Dtx3l 2 clones ‡ C Deltex3L, RING finger protein, similar to rhysin2 Fgfrl1 † B Fibroblast growth factor receptor-like 1 is involved in the negative regulation of cell proliferation. Fkbp5 † A, B, C FK506 binding protein 5, a novel T cell-specific immunophilin capable of calcineurin inhibition. Low in mature T cells. Gpr56 3 clones ‡ B An atypical G protein-coupled receptor, binds to tissue transglutaminase and acts in brain cortical development Gps1 ‡ G protein pathway suppressor 1, subunit 1 of the COP9 signalosome Grap † A, B GRB2-related adaptor protein, GRB2 may interact with LAT by the SH2 domain. It is a negative regulator of the Erk pathway. Huwe1 ‡ A, B HECT, UBA and WWE domain containing 1 (aka Ureb1) is an E3 ubiquitin protein ligase that inhibits the activity of tumor suppressor p53 protein Prkag3 † Protein kinase, AMP-activated, ␥3 noncatalytic subunit, regulates skeletal muscle glycogen content Prpf4b † Pre-mRNA processing factor 4 homolog B, a putative kinase Prss16 # 2 clones ‡ B# Thymus-specific serine protease precursor Prss16, a protein of the thymic stroma Ptpn7 ‡ B Protein tyrosine phosphatase, non-receptor type 7, a MAPK-specific protein tyrosine phosphatase. Same as HEPTP Rab2 ‡ A, B A G protein of the RAS oncogene family (Table continues) The Journal of Immunology 425

Table I. (Continued)

Total Clones Enrichment qRT-PCR Comments

Rabgap1 † A, B RAB GTPase-activating protein 1 Senp2 ‡ B SUMO/sentrin-specific protease, enhancer of Wnt Tgfbr2 ‡ TGF, ␤ receptor II Trappc2l † B Trafficking protein particle complex 2-like, formerly 1810017G16Rik Ube2l3 2 clones ‡ Ubiquitin-conjugating enzyme E2L3, an E2 ubiquitin protein ligase Uble1a ‡ Ubiquitin-like 1-activating enzyme E1A (SUMO-1- activating enzyme) Was (Wasp) † A Wiskott-Aldrich syndrome protein, may function as a signal transduction adaptor downstream of Cdc42 Wbp11 † WW domain binding protein 11, has protein phosphatase type 1 regulator activity, and is involved in RNA splicing Zcchc11 ‡ B Zinc finger, CCHC domain containing 11, reported interact with TIFA to modulate TLR signaling Regulators of transcription AI449175 ‡ A, B Kruppel-like zinc finger protein that has a KRAB domain Aff3 ‡ A, B Laf4, implicated in T cell leukemia Downloaded from Crsp7 ‡ A, B Cofactor required for Sp1 transcriptional activation Lef1 ‡ Lymphoid enhancer binding factor 1, positive regulator of transcription Mll1 ‡ A, B Myeloid/lymphoid or mixed-lineage leukemia 1, trithorax family protein Mll2 ‡ A, B Myeloid/lymphoid or mixed-lineage leukemia 2, trithorax

family protein http://www.jimmunol.org/ Mxd4 ‡ B Max dimerization protein 4, bHLH protein and predicted transcription factor Notch1 8 clones ‡ Notch gene homolog 1 is a positive regulator of transcription Notch3 3 clones ‡ Notch gene homolog 3 is involved in the negative regulation of cell differentiation Runx1 ‡ Runt-related transcription factor 1, AML1, has DNA- dependent transcriptional activity Tardbp 2 clones † TAR DNA-binding protein, predicted to regulate transcription Tcf12 (HEB) 4 clones ‡ A, B, C HEB, bHLH protein, transcriptional activator by guest on October 1, 2021 Zfp109 † A, B, C KRAB box zinc finger protein, predicted to regulate transcription Zfp27 † A, B KRAB box zinc finger protein, predicted to regulate transcription Zfp30 † A, B, C KRAB box zinc finger protein, predicted to regulate transcription 1700021K02Rik # 2 clones ‡ A, B# A stromal protein also known as Spatial, stromal protein associated with thymi and lymph nodes, a possible transcription factor Nucleotide binding Exosc10 ‡ Exosome component 10, a nucleolar protein with exonuclease activity Gcc1 ‡ Golgi-coiled coil 1 contains a GRIP domain Electron transport Txn2 ‡ Thioredoxin 2, a mitochondrial protein involved in electron transport Molecular function unknown Bat2 4 clones ‡ B HLA-B associated transcript 2, a novel MHC class III- encoded protein Eva1 18 clones ‡ A, B Epithelial V-like Ag 1, possible homotypic adhesion molecule, also expressed on thymic stroma. Mpv17 ‡ May be integral to the mitochondrial membrane Tmem131 3 clones ‡ A, B RW1 or Neg, predicted to contain 2 transmembrane domains. On chromosome 1, immediately adjacent to Zap70. 6030458C11Rik ‡ A RIKEN cDNA of unknown function

Intron with or without exon Regulators of transcription and other zinc fingers Baz2a ‡ Bromodomain adjacent to zinc finger domain 2A or TIP5, transcription termination factor I-interacting protein 5 Bcl11b (CTIP2) † B, C CTIP2, transcription factor required for T cell development (Table continues) 426 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

Table I. (Continued)

Total Clones Enrichment qRT-PCR Comments

Foxp1 † Forkhead box protein P1, a possible transcriptional repressor Lass5 ‡ A Longevity assurance homolog 5, homeobox domain Myb ‡ A, B, C Myeloblastosis oncogene, Myb is a SANT domain transcription factor Ncor1 † Nuclear receptor corepressor 1, transcriptional repressor reported to bind Runx1 Runx1 ‡ B Runt-related transcription factor 1, AML1 Rbm4 ‡ Reported to be involved in nuclear mRNA splicing, via spliceosome. Trim39 ‡ Tripartite motif protein 39, has a four-helical cytokine domain Calcium channel proteins Cacnb2 † Calcium channel, voltage-dependent, ␤2 subunit has high voltage-gated calcium channel activity Kcnn4 ‡ Potassium intermediate/small conductance calcium- activated channel, subfamily N, member 4, positive regulator of protein secretion. Protein metabolism, signaling, kinases, and ubiquitination Downloaded from Arpp21 † Protein phophatase type I regulator activity Csnk1e ‡ Casein kinase 1, ⑀, involved in Wnt signaling by DVL1 Fbxw4 ‡ F-box and WD-40 domain protein 4, domain structure suggests it may be involved in ubiquitin cycle and/or Wnt signaling Gna11 ‡ Guanine nucleotide binding protein, ␣11. Possibly

involved in G protein-coupled receptor protein complex http://www.jimmunol.org/ Saps1 † SAPS domain family, member 1, formerly KIAA1115 Taok2 ‡ TAO kinase 2, contains a serine/threonine protein kinase domain Molecular function unknown 5730419I09RIK ‡ May have synaptotagmin and calcium-dependent lipid- binding domains Tm6sf1 ‡ Transmembrane 6 superfamily member 1, integral to the membrane Malat1 ‡ A, B, C Metastasis associated lung adenocarcinoma transcript 1 (noncoding RNA) by guest on October 1, 2021 a Genes identified by the subtraction protocol as enriched in the Pro-T plus population relative to a myeloid/progenitor population are listed in Table I. Genes identified by clones with canonical exon matches are listed separately from genes identified by clones with matches to noncoding or intron and exon regions. Two genes selected by the subtraction that were found in thymic stroma but not in developing T cells are indicated by #. One clone was identified for each gene unless otherwise noted. Most genes listed were enriched by 4 SD above the geometric mean, ‡, and some were enriched by 3 SD, †. Selected genes were analyzed by qRT-PCR in various cell population sets, coded A, B, and C. Population set A is comprised of nonsorted Pro-T plus cells, sorted Pro-B cells, and nonsorted progenitor/premyeloid cells. Set B is sorted DN3 cells, stem-like progenitor cells (LSK CD27Ϫ), later multipotent progenitor (LSK CD27ϩ), sorted Pro-B cells, and sorted BM myeloid cells (see also Fig. 2). Set C includes LSK CD27Ϫ and LSK CD27ϩ progenitors as in set B, as well as sorted DN thymocyte subsets from earliest DN1 stage through ␤-selection to the DP stage (see Fig. 5). b Key: A, unsorted Pro-T plus, cultured progenitor/premyeloid and sorted Pro-B cells (data not shown); B, sorted DN3, LSK CD27ϩ, LSK CD27Ϫ, Pro-B, and BM myeloid populations (Fig. 2); and C, B populations with sorted DN1, DN2, DN3a, DN3b, and DN4 Pro-T subpopulations (Fig. 5). #, Present in thymic stroma (see Fig. 4); ‡, genes up-regulated by four SDs; and †, 3 SD up-regulation.

CACATGGTCAG; Crsp7, forward, ATGGTGGCAGTGTTGGAAGT, TGCTGCCACAGCTATTT, and reverse, CACCAACCTTTGCTGGAG and reverse, GGTTTTCTTGCGGACATCAT; Ctdsp1, forward, CCAG AT; Ldb1, forward, TGAAGTTGGCTCCACCTTAGT, and reverse, G TGAACAATGCGGACTT, and reverse, CCCATTCGCTGTAGGAAC CTCCTTCGGCGAGTACAG; MLL1, forward, TGCCCATAGCCCA TC; Ddx17, forward, AGACAAAGAGGCGCTGTGAT, and reverse, T, and reverse, TCTGTGAATGAGGC; MLL2, forward, GTGCAGCA CCTTTCCAGATCGGAACTCA; Ddx19b, forward, GCCAAGTAGAG GAAGATGGTGAA, and reverse, AGAGCAGCCAGCAGGTCTAA; CCTGCAAAC, and reverse, ACTTGCCCATCTGCTCAATC; Deltex1, Myb, forward, AGCGGGAATCGGATGAATCT, and reverse, GAGC forward, GAGGATGTGGTTCGGAGGTA, and reverse, CCCTCATA AGAAGAAGTTTCCCGATTT; Mxd4, forward, CCGAACAACAGGT GCCAGATGCTGT; Deltex3L, forward, CGGACACCTACGAGGTG CTTCACA, and reverse, CGCTTCAGAAGGCTCAGAGT; Prss16, AAG, and reverse, TTTCCAGGACAATGGTCACA; Eva1, forward, forward, CCCAAACAAGGGTGGTTAGA, and reverse, CTTGGCCA TCACAGCCCTTTGTCCTACA, and reverse, AGTTAGCGCATCTCC GTTCTGTGTTGA; Ptpn7, forward, CTTACACGCTGGACGCTACA, CACAG; FgfrL1, forward, TGCAAATACCATGGGCTACA, and re- and reverse, TCCAGGTCTTCAGGGTTGAC; PU.1, forward, GCGCT verse, GCTTGTGGATGACGATGAAG; Fkbp5, forward, AACGAAG GGCACCTTTTTGTAT, and reverse, CAATAATTTTACTTGTCTTT GAGCAACGGTAAA, and reverse, AATCGGAATGTCGTGGTCTT; AGTGGTTA; Rab2, forward, TGCCAAGACTGCGTCTAATG, and FUS, forward, CAGCAACGAGCTGGAGACTG, and reverse, TCTGG reverse, GCTGAGGGCCAATTTTAATG; Rabgap1, forward, CCTCC CTTAGGTGCCTTACACTG; GAPDH, forward, ACTCCACTCACGG CAGTGGTTCCTTACA, and reverse, GGGCGACATTAAAGATGAC CAAATTCA, and reverse, GCCTCACCCCATTTGATGTT; GATA3, AC; Scl, forward, CAACAACAACCGGGTGAAGA, and reverse, AT forward, GAGGTGGTGTCGCATTCCAA, and reverse, TTTCACAGC TCTGCTGCCTCCATCGTT; Senp2, forward, TAAGGTTCTCGGCA ACTAGAGACCCTGTTA; Gpr56, forward, TTGCAGCAGCTTAGCA CCATTC, and reverse, GGCTGGGATCTCATCAGTGT; Spatial, GGTA, and reverse, GTCTCCCAGGAAGCTCACAG; Grap, forward, forward, GACACAAGAGGCAGCCTACAG, and reverse, GGATGCA GTGTGACGAGCAACCACTGA, and reverse, TCCACAACTTCCA CCAGGAGGACTT; Tcf7 (aka T cell factor-1 (TCF-1)), forward, CA CGATGTC; Heb-alternate, forward, GTGCTTATCCTGTCCCTGGAA AGGCAGAGAAGGAGGCTAAG, and reverse, GGCAGCGCTCTCC TG, and reverse, TGGCTTGGGAGATGGGTAAC; Heb-canonical, for- TTGAG; Tmem13, forward, GCCCTCCCTAGACCCAACTG, and ward, GAGAAGAAGACCGCTCCATGAT, and reverse, TGGCTTGG reverse, GCTTCCAAGTAGGCTGTTCCA; Trim44, forward, TCTGT GAGATGGGTAAC; Helz, forward, TGATGGGCTATTTGGGTGTT, GTCCTGTGTCCAGTCATT, and reverse, CAGTCCACCGGAATCTT and reverse, CTGGAGGGCCATGTCATAGT; Huwe1, forward, GGT TGC; Zcchc11, forward, TGACAGTGCTTCAGGGATTG, and reverse, The Journal of Immunology 427

TAGCCTCTGCTCAGGTGTCA; Zfp27, forward, TTTTTGCCAGCA GCAGATAG, and reverse, CTGCACCACATCCCGATAG; Zfp30, for- ward, TGCCTACGAGAGGGATCTGT, and reverse, CCTTGTTCCAA CAGGGTGA; and Zfp109, forward, GCTGCTCAGAGGAAGCTGTA, and reverse, CCCCAGTGAAAGGCATCTTA. Data display as heat maps The heat maps were generated in the Excel program by arranging expres- sion data in a table with the genes forming the rows and the conditions forming the columns. For each gene, its expression data is normalized by dividing by the geometric mean of that gene’s maximum and minimum expressions. All of the normalized values between 1/͌3 and ͌3 are as- signed to the middle bin, yellow. Each step from dark blue (lowest ex- pression) to red (highest expression) in the rainbow represents a 3-fold decrease or increase, respectively, in the boundaries of the bin, except for the final bins; dark blue ranges from 0 to 1/27͌3, and red ranges from 27͌3 to infinity. Results Subtractive screening of arrayed library To identify previously uncharacterized genes that might act during the earliest stages of the T cell developmental program, we per- Downloaded from formed a hybridization screen for T-lineage-enriched transcripts in a macroarrayed cDNA library of ϳ70,000 clones from mouse FIGURE 1. Subtractive hybridization to macroarrays to identify Pro-T Pro-T (DN1-DN3) and pre-NK cells. This library, generated in our cell enriched cDNAs. A and B, Phosphorimages of a single region of a lab (31), had yielded novel and informative Pro-T cell transcripts macroarray blot are shown, demonstrating enrichment of two cDNAs in before (38) and provided an opportunity to recover unannotated subtracted probe. A was hybridized by unsubtracted Pro-T plus probe. B was hybridized with subtracted probe. cDNA clones are each represented genes as well as alternative transcripts that might not be repre- http://www.jimmunol.org/ by spot pairs. Note that two spot pairs, indicated by arrows, are much sented in microarrays. To establish a baseline, the library was ini- darker in the subtracted probe relative to the mostly unchanged background ko tially probed with Rag2 thymocyte cDNA (Pro-T plus; because clones. The enriched clones represented by these spot pairs were sequenced this mutation prevents ␤-selection, this population is primarily and identified as linker for activation of T cells (LAT) and IL-2 receptor ␣ DN3 cells). It was then probed with “subtracted probe,” consisting (IL2Ra). C, The results of the subtractive enrichment of one of the four of Rag2ko thymocyte cDNA from which message shared by a my- blots that comprise the whole arrayed library are shown here on a log scale. eloid-biased progenitor population was subtracted. T lineage spe- Each dot on the graph represents the change in hybridization intensity cific cDNAs were those that hybridized with specifically increased between unsubtracted and subtracted probes for a single clone in the mac- intensities to the subtraction-enriched probe (Fig. 1). Clones thus roarray library. Note the skew of the data reflects the ability of this method identified as enriched (see Materials and Methods) were se- to measure enrichment but not depletion between two probe samples. Three by guest on October 1, 2021 geometric SDs from the mode are indicated by the black lines and four quenced and mapped to their coordinates in the mouse genomic geometric SDs by the gray lines. The number of clones in each region of sequence. More than 1000 sequences were analyzed to retrieve the graph is denoted on the right. genes specific to early T lineage cells. An early indication of the robustness of the subtraction was evidenced by the fact that 348 clones, one third of the enriched exons. In addition, select matches that include intronic or imme- clones, were found to represent genes already known to be up- diately flanking sequences are listed, as long as they did not in- regulated in or unique to Pro-T cells (Table I). One of these genes, clude SINE or LINE homologies, because novel alternative splic- Tcf7 (TCF-1), encodes a transcription factor with known essential ing, polyadenylation, and promoter use isoforms would also be of roles in T cell development (39–41) while the others encode pre- interest. TCR and TCR components, signaling molecules (Lck and LAT), Ninety of the 92 genes listed in Table I were submitted to the mutagenic DNA polymerase DNTT (terminal deoxynucleoti- GOToolBox (35) for classification by Gene Ontology. Hypergeo- dyl transferase), and distinctive cell surface markers of Pro-T cells. metric statistical analysis was performed using the GOToolBox We excluded from consideration 217 clones that represented ribo- GO-Stats function. Only Spatial and Prss16 were omitted from this somal RNA, 51 clones of mitochondrial origin, and 120 clones analysis, for reasons described below. Select GO-Toolbox results with significant alignments to short or long interspersed nuclear are listed in Table II. Among the subtraction-enriched transcripts, elements (SINEs or LINEs). Also, 154 of the clones (24%) aligned those encoding transcriptional regulators were markedly over-rep- to unidentified RIKEN sequences in the databases or were not resented relative to the Mouse Genome Informatics (MGI) data- significantly similar to any known sequence in the bacterial or base ( p Ͻ 4 ϫ 10Ϫ6). Also significantly overrepresented in the animal NCBI database. (These sequences have been reported and enriched data set were transcripts predicted to encode components 8 are presented in Supplementary Table I.) The 92 genes repre- of the ubiquitin cycle ( p ϭ 0.0012), Wnt receptor signaling com- sented by the remaining clones are the focus of this report. ponents ( p Ͻ 3 ϫ 10Ϫ6), and proteins with a nuclear localization Ͻ ϫ Ϫ7 Candidate genes for early T cell function ( p 4 10 ). Wnt signaling and Notch signaling components as well as transcriptional regulators generally were of interest be- Table I lists the transcripts that were identified as enriched in Pro-T cause of the critical roles of these signaling pathways in early T plus cells relative to premyeloid cells by 3 (†) or 4 (‡) SDs above cell development (42–45). These results suggested that our sub- the mode. These genes were identified by high-quality sequence- traction-enriched clones could be a rich source of potential regu- Ͼ Ͼ matches (typically 500 bp, all were 100 bp) to documented latory genes for the early stages of T cell development. To verify the enrichment predicted by the subtractive screen, a 8 The online version of this article contains supplemental material. selection of genes identified as up-regulated by the screen was 428 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

Table II. Statistical overrepresentation of genes in the subtractiona analyzed for expression by qRT-PCR analysis of a sorted Pro-B cell population from Rag2ko mice and the two populations of cells ko Regulation of transcription ( p Ͻ 4 ϫ 10Ϫ6) used in the subtractive hybridization, i.e., Rag2 thymocytes Aff3 Foxp1 Ncor1 Tardbp (Pro-T plus), and cultured Rag2ko progenitor/premyeloid cells. Ex- Baz2a Lass5 Notch1 Tcf12 pression analyses also included Nulp and Nfe2 (data not shown), Bcl11b Lef1 Notch3 Tcf7 Ccnk Mxd4 Rab2 Zfp109 identified early in the study by a less stringent criterion. Of 23 Ddef1 Myb Runx1 Zfp30 genes tested from the set shown in Table I (qRT-PCR “A”), for all Ubiquitin cycle ( p ϭ 0.0012) except Ctdsp1, Was (Wasp), Rab2, and Senp2, the qRT-PCR re- Fbxw4 Rad18 Trim39 Uble1a sults showed higher expression in the Pro-T plus population as Huwe1 Senp2 Ube2l3 compared with progenitor/premyeloid cells (data not shown). Wnt receptor signaling ( p Ͻ 3 ϫ 10Ϫ6) These results justified a higher resolution analysis of the expres- Csnk1e Ldb1 Senp2 Tcf7 sion patterns of the enriched genes. Fbxw4 Lef1 Nuclear localization ( p Ͻ 4 ϫ 10Ϫ7) Comparison of gene expression patterns between Pro-T and Aff3 Exosc10 Mxd4 Tardbp Pro-B cells and multilineage progenitors Akap8 Fkbp5 Myb Tcf12 Bat2 Foxp1 Ncor1 Tcf7 The significance of new candidate regulatory genes for T cell Baz2a Gps1 Prpf4b Uble1a lineage determination could be quite different depending on Bcl11b Huwe1 Rad18 Wbp11 their expression pattern in the above cell populations. We iden- Ctdsp1 Lass5 Rbm4 Zcchc11 tified three general categories of expression: 1) inherited from a Dde1 Ldb1 Runx1 Zfp27 Downloaded from Ddx19b Lef1 Senp2 Zfp30 stem-cell precursor, a category we termed “legacy”; 2) ex- pressed in a general “pan-lymphoid” pattern; and 3) actually a The 90 genes selected by the subtraction protocol were analyzed by the GO-Stats function of the GOToolBox web site for statistically significant overrepresentation of induced in developing precursors through a T lineage-specific Biological Processes or Cellular Components relative to the Mouse Genome Infor- process. We analyzed the patterns of expression of 43 genes matics database. from the screen in sorted populations of wild-type mouse he- mopoietic cells. By using gene-specific qRT-PCR, we were able to compare expression quantitatively in highly purified, sorted http://www.jimmunol.org/ cells from very small populations. Pro-T plus cell populations by guest on October 1, 2021

FIGURE 2. Quantitative real-time PCR comparison of selected gene expression levels in prethymic progenitors, Pro-T cells, and other hemopoietic lineages. The patterns of expression of 38 genes selected by the subtractive screen were analyzed in five hemopoietic cell populations as indicated by the key at the bottom of the figure. Gene expression relative to GAPDH, averaged from three biological replicates, is graphed on a log scale. Error bars indicate plus and minus one geometric SD. The patterns of expression of subtraction-identified genes (B and C) are compared with those of six landmark genes (A). (See Materials and Methods for primers and details of acquisition of cell populations.) The same samples were used for all measurements shown. The Journal of Immunology 429

tential Wnt signaling modulator Senp2, the protein tyrosine phos- phatase Ptpn7 (He-PTP), and Deltex3-like (Dtx3L) (B lymphoma and BAL-associated protein, Rhysin-2), a RING finger ubiquitin li- gase related to the Notch-induced protein Deltex1 (Dtx1). Several other genes of unknown function, such as Bat2, Tmem131 (RW1 and Neg), and Eva1, were also included in the analysis based on their high representation among the subtraction-enriched cDNA clones. The populations used for this analysis were validated by anal- yses of regulatory landmarks for the stem cell to T cell transition (Fig. 2A), namely, the genes encoding the stem cell transcription factor SCL/Tal1, the T cell transcription factor GATA3, the my- eloid transcription factor PU.1, the Bcl11b relative that is required for B cell development, Bcl11a, and the direct Notch target gene Deltex1, which encodes an E3 ubiquitin ligase (47) (Fig. 2A). These showed the expected patterns of expression for the cell pop- ulations. SCL was expressed highly in the progenitor subsets but not the others; PU.1 was expressed highly in the progenitors and Pro-B cells and was further enriched in myeloid cells, but down- regulated in the Pro-T cells; GATA3 was up-regulated specifically in the Pro-T cells; and Bcl11a was highest in the Pro-B cells but Downloaded from specifically down-regulated in the Pro-T and myeloid cells. As shown in Fig. 2, the majority of the subtraction-selected genes was verified to show at least 2-fold more expression in Pro-T cells than in the sorted BM myeloid cells (yellow or white vs pink bars, Fig. 2), and in most cases the difference was at least 10-fold (please note log scale in Fig. 2). The exceptions were Senp2, http://www.jimmunol.org/ Ctdsp1, and Rab2, which showed weak enrichment if any. Trappc2l had Ͻ3-fold enrichment. The genes that were T-enriched showed various patterns of expression relative to Pro-B cells (blue bars) and to the stem and multipotent progenitor cells (green bars). However, as will be described below, remarkably few of these genes were truly T lineage restricted.

Dominance of “legacy genes” and pan-lymphoid genes in Pro-T by guest on October 1, 2021 vs premyeloid-enriched gene set Many of the genes that were differentially expressed between FIGURE 3. Heat map of gene expression levels in hemopoietic lin- Pro-T and premyeloid cells were expressed at similar levels in eages. The data depicted in graphs in Fig. 2 were used to construct a Pro-T cells and in Pro-B cells (Յ2ϫ difference and/or within er- clustered heat map. The log average of the maximum and minimum ex- ror), implying functions shared in early T and B lineage develop- pression level for each gene was set to mid-range (yellow). Each color step ment. These pan-lymphoid genes include Aff3 (LAF4), Crsp7, indicates a 3-fold change in expression level from blue (lowest expression Mll1, Mll2, Mxd4, Zfp27, Ddx17, Trim44, Zcchc11, Gpr56, Grap, level) to red (highest expression level). Population averages with SDs are shown in Fig. 2. and Akap8. Although boundaries between classes are not sharply defined, Myb, FUS, Ablim, Huwe1, and Bat2 could be considered pan-lymphoid as well. All of these genes except Ablim and Grap were compared not only with sorted Gr-1ϩMac-1ϩ myeloid were also expressed at similar levels in the multilineage LSK cells and CD19ϩ Pro-B cells from Rag-knockout BM, but also CD27ϩ precursors, implying that their lymphoid function may be with sorted populations of enriched hemopoietic stem/progenitor inherited from a pluripotent precursor. These similarities are evi- cells (LinϪKitϩSca-1ϩCD27Ϫ) and multipotent lymphomyeloid dent in the heat map shown in Fig. 3. progenitors (LinϪKitϩSca-1ϩCD27ϩ) (19, 46). These results are Genes specifically up-regulated as part of the T lineage devel- presented in Fig. 2 as qRT-PCR graphs and in Fig. 3 as a clustered opmental choice would be expected to be expressed more highly in heat map. Pro-T cells than in either Pro-B or BM myeloid cells (Ͼ2ϫ), and The genes that we selected from Table I to test for expression a number of genes were found to have this pattern. However, even pattern included those encoding known transcription factors and within this set, the majority had expression levels in one or both of chromatin modifying proteins Bcl11b, HEB (aka Tcf12, two dis- the progenitor populations (LSK CD27Ϫ and LSK CD27ϩ) similar tinct promoter isoforms tested), MLL1, MLL2, Mxd4, Myb, and to (within 2ϫ) that found in the Pro-T cell population. These genes four likely zinc finger transcriptional repressors (Zfp109, Zfp27, include Zfp109, Zfp30, Ldb1, Rabgap1, Ptpn7, and Ddx19b. Genes Zfp30, and AI449175), as well as known or suspected transcrip- such as Myb, AI449175, and Helz were also expressed most highly tional modulatory factors Aff3, Crsp7, Ablim, Ctdsp1, and Ldb1. in the Pro-T cell samples, but the magnitudes of their up-regulation We also tested genes encoding RNA-binding proteins and heli- relative to stem and progenitor cells were only 2–3ϫ. None of cases Ddx17, Ddx19, FUS, and Helz; zinc finger factors with other these transcripts were as T lineage enriched as GATA3 or Tcf7 less-characterized roles such as Trim44 and Zcchc11; signaling (Fig. 2A) or even as the canonical form of HEB (HEBcan). These molecules and adaptors such as Gpr56, Grap, Rabgap1, Rab2, newfound genes therefore are not specifically induced during T Akap8, and Fkbp5; the E3 ubiquitin ligase Huwe1 (Ureb1); po- lineage specification, but instead represent multipotent precursor 430 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO Downloaded from http://www.jimmunol.org/

FIGURE 4. Stromal-specific genes and a gene shared by stromal cells and thymic lymphocytes. A, Expression of three genes previously considered to be exclusive to the thymic stroma was measured in sort-purified DN3 thymocytes (denoted DN3) and in unsorted DN3-enriched cells with stromal contamination, Pro-T plus. Although Pro-T plus samples indicate high levels of expression of Eva1, Spatial, and Prss16, only Eva1 expression is expressed equally in sort-purified DN3 cells. B, The pattern of Eva1 expression was measured in LSK prethymic cells, Pro-B and myeloid cells, and in DN thymocyte subsets from the earliest stages through ␤-selection (see text and Fig. 4 for details). Up-regulation of Eva1 occurs at DN2, peaks at DN3a, and declines after ␤-selection. Eva1 mRNA is present at low levels in Pro-B cells and is found at background levels in myeloid cells. C, Structures of novel Eva1 by guest on October 1, 2021 transcripts identified in this study and aligned with the mouse genome on chromosome 9 by BLASTn (red blocks identified as “Non-canonical sequences” in Ensembl genome browser). Nine of them are depicted in this Ensembl alignment. The GenBank accession numbers for these expressed sequence tags are EL773010, EL773011, EL773012, EL773013, EL773014, EL773015, EL773016, EL773017, and EL773018. These novel isoforms all affect promoter use and/or splicing patterns at the 5Ј end of the Eva1 gene. For sequences of these novel Eva1 cDNAs, see Supplemental Table I.8

legacy genes that T lineage cells continue to express, even while Subtraction-enriched thymic stromal genes and a gene with other lineages down-regulate them. shared lymphoid and stromal expression Because mRNAs for the library construction and the subtraction protocol were obtained from nonsorted Rag2ko and SCID thymo- Genes specifically up-regulated in T lineage precursors cytes with some contamination by stromal epithelial cells, our Against this background, the T lineage specificity of a select group screen would be expected to enrich for stromal-specific cDNAs as of genes from our screen stood out (Figs. 2 and 3). These included well as for thymic lymphocyte-specific ones. Two genes selected transcripts of three genes encoding transcription factors, Tcf7 by the subtraction, encoding the serine protease Prss16 and Spatial (TCF-1), Bcl11b and HEBalt (the alternative promoter use form of (Titest, 1700021K02Rik), were also specifically expressed in the HEB); the RING finger protein Dtx3L; the signaling adaptor pro- Pro-T plus population (Fig. 2C and data not shown). However, tein Fkbp5; and two products of unknown function, Tmem131 their transcripts were not found in sorted hemopoietic populations (RW1, Neg) and Eva1 (epithelial V Ag). Tmem131 and Fkbp5 (Fig. 4A), in accord with their annotation as stromal specific genes. were up-regulated by slightly less than an order of magnitude (9- Unlike Spatial or Prss16, a third “stromal” annotated gene, Eva1 (ep- fold) from precursors (Fig. 3, green to gold). HEBalt levels in ithelial V-like Ag 1), was verified to be expressed in sort-purified Pro-T cells were much higher than those in LSK cells, but were DN3 cells (Fig. 4A). Eva1, thought to be a homotypic adhesion mol- also up-regulated substantially in Pro-B cells, in agreement with ecule and previously found only in thymic stroma, liver, and other previous report (48). Bcl11b was unusual for the magnitude and epithelial tissues (49), was expressed within the T lineage in a stage- specificity of its up-regulation (Fig. 3, dark blue to red), even specific and transient way, beginning at the DN2 stage and peaking at greater than the up-regulation of known T lineage factor Tcf7 the preselection DN3a (24) stage (Fig. 4B). It is possible that Eva1 (TCF-1, Fig. 3, blue to orange) and comparable to that of Deltex1 mediates homotypic adhesion interactions between thymocytes and (Fig. 3, dark blue to red). These genes are investigated in more thymic stroma. Expression of Eva1 by DN3 cells would have easily detail below. been overlooked in immunohistochemical assays (49) because the The Journal of Immunology 431 Downloaded from http://www.jimmunol.org/

FIGURE 5. Distinct stage-specific regulation of different Pro-T-specific genes: fine-scale developmental regulation of Pro-T cell-enriched genes en- coding Bcl11b (J), HEBalt (K), Dtx3L (F), and Fkbp5 (E) and novel zinc finger factor genes Zfp109 (D) and Zfp30 (C) is compared with that of reference genes encoding GATA3 (T lineage) (A), Myb (legacy) (B), PU.1 (G), SCL (H), and Bcl11a (progenitor- and non-T lineage) (I), and Dtx1 (Notch target) (L). Gene expression was analyzed by qRT-PCR in nine hemopoietic cell populations as described in Fig. 4B and graphed on a log scale relative to GAPDH expression. Cell populations and their relationships to T cell development are shown in the schematic (bottom). They include LSK CD27Ϫ stem cell-like hemopoietic progenitors (dark green bars), LSK CD27ϩ multipotent progenitors (light green), sorted DN T cells from earliest, DN1 stage through ␤-selection to DN4 stage (orange, gold, yellow, pale yellow, and white), and Pro-B cells (blue) and BM myeloid cells (pink). Populations of each cell type were purified on three separate dates, and the expression levels for the genes analyzed are reported here as the average Ϯ geometric SD among the three

independent biological replicates. by guest on October 1, 2021

percentage of DN3 cells among thymocytes in the wild-type thymus is low (ϳ1%). We found nine noncanonical transcripts for Eva1, a benefit of the macroarray library, that appear to encode at least four novel transcripts with previously unreported exons or promoter re- gions (Fig. 4C and Supplemental Table I).8 Regulation of Eva1 is potentially interesting because the Eva1 gene is located on chromo- some 9 in the only significant physical cluster of Pro-T cell genes identified in our study. It is immediately adjacent to cd3e (within 32 kb) and within 161 kb of Mll1, which flanks the cd3g/cd3d/cd3e cluster on the other side (data not shown). Stage-specific onsets of regulatory gene expression in T lineage precursors The potential roles of the few T lineage-specific transcription fac- tors and signaling molecules identified in Fig. 2 should depend on the developmental stages at which they are induced. Tcf7 (TCF-1) has been extensively studied (29–31), and its expression shown to increase gradually through the DN1–DN4 progression (8), but the FIGURE 6. Heat maps of LSK and/or DN subset expression of select other genes are less well characterized. To determine the timing of genes. The gene expression heat maps were generated as was the map up-regulation of these T lineage-biased genes, we analyzed their shown in Fig. 3. The expression data shown as graphs in Fig. 5 were expression in 5 subpopulations of T cell precursors sorted from combined with data from additional genes and are presented here as a heat wild-type mouse thymus (DN1, DN2, DN3a, DN3b, and DN4), in map. B, Expression of 10 additional genes in DN subsets (relative to ␤-ac- tin) is shown as a heat map with the classic Notch target gene HES1 as a direct comparison with the two subpopulations of hemopoietic progenitors (LinϪKitϩSca-1ϩCD27Ϫ and LinϪKitϩSca-1ϩ reference (geometric mean levels from two independent biological series). ϩ The sorted DN populations used for this panel were from different prepa- CD27 ), Pro-B cells, and the sorted BM myeloid population used rations than those used for A and Fig. 5 (E.-S. David-Fung, data not above. This comparison spans the range of early T lineage mile- shown). stones: entry into the thymus during the transition to the DN1 432 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO stage; “specification” at the DN1 to DN2 transition; “commit- a modulator of the glucocorticoid receptor (57). These signaling ment” and proliferation arrest at the DN2 to DN3a transition; molecules and cell surface receptors thus appear to be most and ␤-selection or ␥␦-selection, via DN3b and DN4 intermedi- strongly expressed at a stage coinciding with T lineage commit- ates (13, 24, 50). ment, cell cycle arrest, and TCR gene rearrangement, but after The results are shown in Fig. 5 as qRT-PCR graphs, and Fig. 6 initial T lineage specification has begun. shows DN subsets Ϯ LSK population expression results for a more The genes with similar levels in prethymic progenitor cells and extensive set of genes in heat map form. For developmental ref- Pro-T plus cell fractions (Fig. 2) confirmed their “legacy” patterns erence standards, we measured GATA3 expression as a model T of expression by the continuity and constancy of their expression lineage-specific positive regulator (Fig. 5A); Myb expression as a patterns throughout the early DN stages. Myb expression showed key regulator used by both multipotent progenitor cells and Pro-T little change from prethymic stages throughout the Pro-T stages, cells (Fig. 5B); and PU.1 and SCL as progenitor-cell regulators with only a gentle increase from DN1 to DN3 and a steeper drop that are shut off precipitously between the DN1 and DN3 stages after ␤-selection (Fig. 5B). Two relatively novel KRAB-domain (Fig. 5, G and H). The Notch signaling target gene Deltex1 (Dtx1) zinc finger transcription factors with “legacy” patterns of expres- was also analyzed (Fig. 5L). Deltex1 was not transcribed in the two sion, Zfp30 and Zfp109, were up-regulated between the LSK hemopoietic progenitor populations (see Fig. 5M, detection thresh- CD27Ϫ and LSK CD27ϩ stages of prethymic differentiation and olds), but its expression was up-regulated ϳ100-fold above back- continued their expression through the DN1 to DN3 stages with a ground levels at the DN1 and DN2 stages, in agreement with the decrease after ␤-selection (Fig. 5, C and D). critical role of Notch signaling in T cell specification. Interest- Gene expression analyses in these DN and BM subsets were ingly, it showed a further up-regulation at the DN3a stage to also conducted for Helz, Ddx19b, Tmem131 (Fig. 6A), and further Ͼ2000-fold over the background (Fig. 5L), suggesting a second analysis of DN thymocyte subsets was performed on Aff3, Grap, Downloaded from discrete phase of Notch activity (51). Ldb1, Mll1, Mll2, Trim44, Atxn2l, Tcf7, and FUS, in comparison Only two of the regulatory factors in our study were primarily with the Notch target gene HES1 (Fig. 6B). None of these matched induced during the DN1 to DN2 transition. One was the promoter- the T lineage specification-associated induction of Bcl11b. Aff3 use variant of HEB known as HEBalt (Fig. 5K). Its dramatic up- actually declined steadily from the DN2 to the DN4 stage after an regulation was followed by an ϳ10-fold decline after the DN3a early plateau. Tcf7, FUS, Grap, Helz, Ldb1, Mll1, Mll2, Tmem131, stage, in agreement with previous report (48), and consistent with and Trim44 remained steady or increased gently to the DN3 stage, http://www.jimmunol.org/ the early hit-and-run positive function this basic helix-loop-helix with a decline thereafter; but the range of expression was narrow. factor variant appears to play in T cell development (48). Ddx19b followed the same pattern, after an initial drop between The gene with the most singular pattern of expression in our the LSK CD27ϩ prethymic stage and DN1 stage. In a companion analysis encodes Bcl11b, a zinc finger factor that usually acts as a study of Ͼ80 Pro-T cell-expressed transcription factors (E.-S. transcriptional repressor (52–54). Bcl11b appeared strictly T lin- David, G. Buzi, L. Rowen, R. Butler, R. A. Diamond, M. K. eage-specific relative to stem and progenitor populations and, in Anderson, and E. V. Rothenberg, manuscript in preparation), only contrast to HEBalt, is expressed at only trace levels in Pro-B and Bcl11b demonstrated T lineage specificity and Ͼ100-fold up-reg- myeloid cells (Fig. 5J). In addition, unlike all the other T lineage ulation at the DN1 to DN2 transition. by guest on October 1, 2021 genes, Bcl1b transcripts showed little expression in DN1 cells, but increased 500-fold between the DN1 and DN2 stages, with only a Bcl11b induction by Notch/Delta signaling fewfold further up-regulation to the DN3a stage (Fig. 5J). The Notch/Delta signaling induces expression of the known T lineage magnitude of this increase dwarfed the increase seen in GATA3 regulatory genes GATA3 and Tcf7 with a characteristic time expression over the same interval (Fig. 5A). Unlike HEBalt, course in fetal liver-derived hemopoietic precursors (19, 58, 59). Bcl11b expression then remained fairly level through the DN4 OP9 stromal cells normally support B cell differentiation of he- stage, and its expression continued in peripheral T cells (52) (data mopoietic precursors, but OP9 cells engineered to express the not shown). Bcl11b up-regulation was accompanied by the recip- Notch ligand Delta-like1 (OP9-DL1 cells) support T cell develop- rocal down-regulation of its relative, Bcl11a, which was strongly ment. Time course analysis of hemopoietic precursor cells in co- expressed in progenitor cells and non-T cells but down-regulated culture with OP9 control or OP9-DL1 cells provides a second way by two orders of magnitude between the DN1 and DN3 stages to look at the earliest events involved in T lineage specification, (Fig. 5I). This analysis implies that the ratio of Bcl11b to Bcl11a separable from any technical issues about the correct identification in thymocytes shifts dramatically during progression from DN1 to and purification of precursor subsets. We therefore cultured fetal DN4, by over four orders of magnitude, a finding that is particu- liver-derived hemopoietic precursors on OP9-DL1 or OP9-control larly relevant in light of reports that chromosomal translocations stroma and compared the expression kinetics of Bcl11b with those affecting Bcl11b expression are found in ϳ20% of pediatric, T cell of Tcf7, Deltex1, the T lineage gene CD3␧, Eva1, and legacy or acute lymphoblastic leukemias (55, 56). pan-lymphoid transcription factor genes as shown in Fig. 7. Sam- Other signaling genes and transcription factor genes selected by ples were obtained as described previously (19), representing the screen showed less dramatic increases with earlier or later 2-day intervals in a time course of 10 days of culture overall. peaks of expression. Dtx3l, encoding a putative interaction partner Fig. 7 shows that Bcl11b was strongly and specifically up-reg- of Deltex1, roughly paralleled Deltex1 (Fig. 5, F and L) and Eva1 ulated in fetal liver cells in response to culture with OP9-DL1 (Fig. 4B) in expression in T lineage populations. Showing little (navy blue line) but not when cultured with OP9-control cells that expression in the stem/progenitor cells, Dtx3L was already detect- do not express Notch ligand (turquoise line). The kinetics of ably up-regulated at the DN1 stage and increased to a peak in the Bcl11b induction were strictly dependent on the timing of expo- DN3a stage. Like several Notch pathway targets (24), Dtx3L was sure to DL1. When cells were initiated in culture on OP9-control sharply down-regulated after ␤-selection in the DN3b and DN4 and then shifted to OP9-DL1 after a delay of 4 days (see Materials stages. FK506-binding protein5, Fkbp5, was already expressed at and Methods), Bcl11b induction was also delayed (Fig. 7, yellow significant levels in the multilineage precursor populations, but its line). In contrast, induction of Bcl11b through these stages de- T lineage-specific up-regulation also reached a peak at the DN3 pended on continued signaling from Notch/DL1 interaction, be- stage, albeit with changes of lower amplitude (Fig. 5E). Fkbp5 is cause when the cells were removed from OP9-DL1 to OP9-control The Journal of Immunology 433 Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 7. Temporal regulation of Pro-T and legacy genes during induction of T cell specification by Notch/Delta signaling. Time courses of gene expression were analyzed by qRT-PCR in fetal liver cells cocultured with either OP9 control cells or OP9 cells expressing the Notch ligand DL1. The results of four culture conditions are shown. Top panels, Schematic of the experiment. Lower panels, Quantitation of gene expression in time course samples by qRT-PCR. Fetal liver-derived precursors (day 0) were cultured for up to 10 days total with OP9-control stromal cells (nonpermissive for T lineage) or OP9-DL1 stroma (T lineage inducing). To distinguish gene expression responses that respond to continuous Notch/DL1 signaling from those that do not, cells retaining progenitor phenotype (KitϩCD27ϩ) after 4 days of culture were repurified from both kinds of cultures, then each sample was split and used to seed both OP9-control and OP9-DL1 secondary cultures. Samples from days 6–10 were generated either from cells that were replated on the original type of stroma or from duplicate samples that were switched to the opposite stromal type, as indicated by the color code: DL1 to DL1 (navy blue ϭ continuous T lineage inducing), DL1 to control (magenta ϭ abortive exposure to T lineage inducing), control to control (aqua ϭ T lineage nonpermissive, B lineage inducing), and control to DL1 (yellow ϭ delayed exposure to T lineage inducing). Expression levels of the indicated genes are graphed relative to GAPDH levels on a log scale. after 4 days, Bcl11b expression was down-regulated (Fig. 7, ma- temporal pattern as Tcf7 (Fig. 7) (19), but with even greater T genta line). The dependence of Bcl11b induction on Notch/Delta lineage restriction of expression. Thus, Bcl11b is an integral part signaling was comparable to that of Deltex1 (Fig. 7), but its re- of a T lineage-specific regulatory program that is induced in the sponses were temporally blunted. Deltex1 reached a plateau within first stages of response to Notch/Delta signaling in hemopoietic 4 days of stimulation (Fig. 7, navy line) and was immediately progenitors. turned on or off in response to the addition or removal of DL1 (Fig. For most of the other genes identified as T cell specific in this 7, yellow and magenta lines), whereas Bcl11b required more than screen, the OP9-DL1 response kinetics of fetal liver progenitors (Fig. 6 days of stimulation to reach a plateau and showed slower induc- 7) gave results consistent with their steady-state expression patterns in tion and de-induction in response to changes in DL1 stimulation adult prethymic cells and DN thymocytes (Figs. 5 and 6). Fkbp5 (Fig. (Fig. 7). This pattern would be consistent with more complex reg- 7) and Tmem131 (data not shown) were up-regulated in response to ulatory requirements than Notch signaling alone, or a longer Notch/Delta signaling, in agreement with their identification as Notch mRNA half-life than Deltex1, or both. Bcl11b showed the same target genes (28), but with a very shallow change in magnitude. Eva1 434 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

FIGURE 8. Impact of developmental preprogramming on kinetics of Bcl11b induction by Notch/Delta signaling. Expression of Bcl11b in cells differ- entiating from three populations of hemopoietic progenitors was compared by qRT-PCR at time points for up to 7 days on either OP9 control stroma (f) or OP9-DL1 stroma (Ⅺ). Bcl11b expression relative to GAPDH is graphed on a log scale (left axis). For each progenitor population, the percentage of cells cocultured on OP9 DL1 (Ⅺ) that had progressed passed the DN1 stage of T cell development at the indicated time points is also graphed on a linear scale ϭ ϭ ϩ Ϫ ϩ Ϫ ϩ ϩ

(line graphs, right axis). CLP-like common lymphoid progenitor-like cells. LSK stem-like cells. Flk : Lin Kit Sca-1 CD27 Flk2/Flt3 multipotent Downloaded from progenitors. Bcl11b expression under conditions that promote B cell development (f, OP9 control) remained at very low/background levels, whereas Bcl11b expression in cells cultured under T cell-promoting conditions (Ⅺ, OP9-DL1) up-regulated dramatically by day 2 of coculture. Results shown are from one of two independent experiments that gave similar results. The rate at which each progenitor population matures down the T lineage is most rapid for CLP-like cells, slowest for LSKs, and intermediate for Flkϩ cells. The rate at which each population proceeds down the T cell pathway in response to Notch signaling, as indicated by the appearance of DN2 and later cells, is tightly correlated to the level of Bcl11b expression. http://www.jimmunol.org/ was also induced in a Notch/Delta dependent way. The legacy genes sponses. Fig. 8 shows that none of these populations express Myb and Mll1 showed virtually unchanging expression from fetal detectable Bcl11b initially (“0” time points). When cocultured with liver progenitor throughout the differentiation time course, in the pres- OP9-DL1, Bcl11b is rapidly up-regulated in the CLP-like cells, to ence or absence of DL1, in keeping with their shared use in T, B, and levels approaching maximal within 2 days, while the LSK cells stem cells. take 7 days to reach the same level and the Flkϩ cells require at The OP9 kinetic assays did discriminate between Bcl11b and least 3 days (Fig. 8, Ⅺ). These kinetics are in excellent agreement HEBalt usage in T vs B lineage differentiation. HEBalt is ex- with the time it takes for each population to generate DN2 and later pressed in B lineage as well as T lineage precursors (48), but in the stage cells in vitro (Fig. 8, line graphs) (T. Taghon, M. A. Yui, and by guest on October 1, 2021 adult, in vivo-derived populations its expression appeared to be E. V. Rothenberg, submitted for publication). Thus, the duration of strongly biased toward the T lineage (Fig. 2B). In fetal liver pre- Notch-Delta signaling required to turn on Bcl11b depends on the cursors differentiating in vitro, however (Fig. 7), HEBalt was in- initial developmental state of the responding cells. duced in the absence of DL1 (turquoise line, magenta line; B cell conditions) only a fewfold less strongly than in the presence of DL1 (navy line, yellow line; T cell conditions). It also appeared to Discussion be expressed at substantial levels in the fetal liver-derived starting In this study, we have used a gene discovery approach to search for populations (Fig. 7), as in the BM LSK populations (Fig. 2B). In new regulatory factors that participate with GATA3, Tcf7 (TCF- contrast, Bcl11b expression was completely dependent on the T 1), and Notch signaling in initiating T lineage specification. The lineage differentiation conditions. Thus, our screen identifies method used was focused on broad-scale de novo identification of Bcl11b as a singularly specific early component of the T cell pro- gene transcripts that are specifically up-regulated in early T lineage gram in vivo and in vitro. cells relative to other hemopoietic progenitors. This screen yielded a number of candidate transcription factors and signaling mole- Kinetics of Bcl11b induction depend on developmental state of cules that, previously, have been uncharacterized or only circum- prethymic precursors stantially linked with T cell development. Using quantitative RT- The exponential increase of Bcl11b RNA expression over a 4- to PCR with highly purified cell populations, we have tracked 6-day period, as shown in Fig. 7, suggested that in addition to the expression of many of the candidate regulatory genes in detail Notch-dependent process inducing Bcl11b transcription, the fre- across the transition from stem cell to committed T lineage cell. quency of cells competent to express the gene may also be in- This analysis has generated two main results. First, against expec- creasing or that Bcl11b may exert a positive feedback effect on its tation, it has established the preponderance of legacy genes in early own expression. We therefore tested whether the kinetics of T cell development. Most of the regulatory genes that showed Bcl11b induction under these conditions were dependent on the preferential expression in T lineage precursors, as compared with developmental status of the input cells. We took advantage of the B or myeloid lineage cells, actually represent a direct, continuous, fact that distinct subsets of prethymic precursors in the fetal liver quantitatively stable inheritance from pluripotent hemopoietic pro- progress to the DN2 stage with faster or slower kinetics in the genitors. Second, just two new regulatory factors have emerged OP9-DL1 system: CLP-like (LinϪKitϩCD27ϩCD135ϩCD127ϩ) from this screen with early T lineage up-regulation comparable to cells differentiate faster, while stem-like LSK cells (LinϪKitϩ that of Tcf7: HEBalt, a functionally distinct promoter variant of Sca-1ϩ) show a lag (T. Taghon, M. A. Yui, and E. V. Rothenberg, HEB, which is also used in early B cells, and Bcl11b, the zinc submitted for publication), and “Flkϩ” cells (LinϪKitϩ finger transcription factor and tumor suppressor, which is fully T Sca-1lowCD27ϩCD135ϩCD127Ϫ) cells give intermediate re- lineage specific. The Journal of Immunology 435

Table III. Summary of microarray results from recent reportsa

Tabrizifard et al. Hoffmann et al. Dik et al. Lee et al. Petrie lab (24) Melchers lab (25) Staal lab (26) McCune lab (27)

Affymetrix Mouse Affymetrix Mouse Affymetrix Human Affymetrix Human MG-U74A chip Mu11k A&B U95Av2 U133 A&B

Ablim —b — — n/r Aff3 — n/r n/r n/r AI449175 — — — — Akap8 n/r n/r n/r n/r Bat2 n/r n/r Flat n/r Bcl11b — — Peaks at an equivalent Enriched vs stroma but no developmental stage up-regulation between ITTP and DP Bcl11a — — Peaks at ISP CD4ϩ Flat Crsp7 — — — n/r Ctdsp1 — — — n/r Ddx17 n/r n/r Flat n/r Ddx19b — n/r n/r n/r Deltex1 n/r n/r ? n/r Dtx3L — — — n/r Eva1 n/r — n/r n/r Downloaded from Fgfrl1 — — — n/r Fkbp5 n/r n/r Higher in ISP CD4ϩ Slow rise from ITTP to CB4 FUS n/r n/r n/r n/r GATA3 Peaks at DN3 Rises through CD4ϩ High at SP CD4ϩ Rises from ITTP to peak at DP ϩ

Gpr56 n/r — Drops at SP CD4 Drops from ITTP to DP http://www.jimmunol.org/ Grap — — n/r n/r HEBalt — — — — HEBcan Rises through smDP n/r Highest at DP CD3Ϫ n/r Helz — n/r n/r n/r Huwe1 n/r n/r n/r n/r Ldb1 Peaks at smDP n/r n/r n/r MLL1 Peaks at DN2 n/r n/r n/r MLL2 — — n/r n/r Mxd4 n/r n/r Only high at UCB n/r Myb Peak at DN2 and Peak at DN3 Peaks at CD34ϩ38ϩ1aϩ 6-fold drop between DP

again at smDP and SP CD4ϩ by guest on October 1, 2021 Ptpn7 — — High at DP CD3- and again n/r at SP CD8ϩ Rab2 Peaks at smDP n/r High at DP CD3ϩ n/r Rabgap1 — — n/r n/r SenP2 — — — n/r Tcf7 Rises through smDP n/r Peaks at DP CD3ϩ and DP Enriched vs stroma but flat CD3Ϫ between ITTP and DP Tmem131 n/r — n/r n/r Trappc2l n/r — — n/r Trim44 — — Flat n/r Zcchc11 — — n/r n/r Zfp109 — — — — Zfp27 — n/r — — Zfp30 n/r n/r — n/r

a Expression of the panel of genes identified in Fig. 2 as reported by four recent microarray-based bioinformatics articles. The first two columns of the table summarize investigations of mouse hematopoiesis. Small double-positive thymocytes (smDP) represent the development stage subsequent to DN4. CD4ϩ indicates a population of naive CD4ϩ thymocytes. The third and fourth columns summarize analyses of human T lymphopoiesis. Human ISP CD4ϩ cells are roughly equivalent to the mouse DN3 stage (78, 79). Single-positive CD4ϩ (SP CD4ϩ) are naive CD4ϩ thymocytes. DP CD3Ϫ and DP CD3ϩ human thymocytes are similar to mouse double-positive thymocytes. UCB in the Dik et al. study refers to “stem cell-like” CD34ϩ cells from umbilical cord blood, presumed to correspond to a mouse LSK CD27Ϫ population. The ITTP population referred to in the Lee et al. article are similar to the human ISP CD4ϩ poulation (and to mouse DN3 cells), and CB4 indicates a population of naive CD4ϩ T cells obtained from umbilical cord blood. b —, The Affymetrix microarray used by the authors does not have a probe for the gene; n/r, the authors did not report expression for the gene either because it was not the focus of the study or because expression fell below the limits of detection; Flat, gene expression was reported but is consistent throughout the populations evaluated by the authors.

This study is distinguished from related studies of early T cell studied in T cell development, including those encoding transcrip- development in part by its focus on the transition from prethymic tional regulators Aff3, MLL2, Zfp27, Zfp109, and Zfp30, RNA- progenitors (specifically LSK CD27Ϫ and LSK CD27ϩ from BM binding proteins Helz, FUS and Ddx19b, signaling component and LinϪKitϩCD27ϩ fetal liver precursors) into the first intrathy- Grap, the immunophilin Fkbp5, the Lim-binding protein Ldb1, as mic stages. Although several other studies have compared popu- well as uncharacterized products like Eva1, Tmem131, and lations from DN2 through ␤-selection, the earlier transitions have Trim44 (Table III). A comparison of our findings and the results of remained more obscure. Furthermore, the de novo, gene cloning microarray studies by Hoffmann et al. (27), Tabrizifard et al. (26), approach of our work identified a number of genes not previously Dik et al. (28), and Lee et al. (29) shows general agreement where 436 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO results for comparable samples were given, but also highlights the impact of microarray chip comprehensiveness and probe design. T development Bcl11b Deltex1 Microarray data from these sources regarding the genes in Fig. 2 Dtx3L Ddx19b are presented in Table III. Hoffmann et al. (27) and Tabrizifard et AI449175 Eva1 Ablim HUWE1 al. (26) studied T cell development in the mouse, but the Hoffmann Fkbp5 Legacy Trappc2l Heb-alt work did not examine prethymic or DN1 populations and Tabriz- MLL2 Tmem131 ifard et al. (26) started only with DN1 cells and reported results Aff3 Akap8 Ptpn7 Tcf7 only for transcription factors. These studies were constrained by Bat2 Crsp7 Ddx17 FUS (Fgfrl1) Grap Gpr56 Ldb1 the limits of their microarrays as shown by the dashes in Table III MLL1 Mxd4 Rabgap1 (not included on the chip used) and possibly also by detection Trim44 Zfp109 Zfp27 threshold issues (Table III – “n/r”, gene expression not reported). Zfp30 Zcchc11 Dik et al. (28) evaluated gene expression in human hemopoietic Progenitor ϩ PU.1 cells roughly equivalent to the LSK CD27 through DN transi- Scl tions and extended to the late CD4 and CD8 single-positive T cell Myeloid lineage stages. These authors also noted selective up-regulation of Bcl11b at a point in human T cell development similar to our murine data. However, they found Bcl11a expression reinduced in CD4ϩ ISP cells, representing a later T lineage population, and this was not supported by our findings. The recent work by Lee et al. (29) used FIGURE 9. Legacy genes as well as T lineage-specific genes con- a highly comprehensive human microarray chip but their analysis tribute to the early T cell regulatory state. This Venn diagram represents Downloaded from focuses on the later transitions, from ␤-selection to naive CD4ϩ T legacy genes as those genes that are maintained with similar expression cell. Without early thymocyte and hemopoietic populations for in hemopoietic progenitors and in early T cells, but which are down- comparison, the Lee paper answers different questions about gene regulated in early myeloid development. A selection of legacy genes is listed in the overlapping area between progenitor and T development. expression in T cell development. The genes found in this study that are most specifically up-regulated at Identification of additional T lineage-specific regulators is sig- a part of the T lineage program (i.e., expression in Pro-T plus is at least http://www.jimmunol.org/ nificant because the known T lineage-specific transcription factors, 2-fold that of LSK CD27ϩ and the SDs do not overlap) are listed in the GATA3, Tcf7, and HEB (canonical form), do not appear to in- yellow region of the Venn. crease in mRNA expression sharply enough to account for the dramatic onset of T lineage differentiation gene expression, during the DN1 to DN2 “specification” transition (10, 13, 15). In vitro, a marized in Fig. 9). These legacy genes include many molecules 100-fold up-regulation of differentiation genes encoding CD3␧ and that could play roles in signaling cascades that trigger differenti- pT␣ at the DN1 to DN2 transition is accompanied by only a 2- to ation as well as in functions shared by multipotent progenitors and 4-fold greater expression of GATA3 and Tcf7, with no detectable early T cells, and they will be of substantial interest for further increase in known Notch target gene expression to suggest en- study. by guest on October 1, 2021 hanced Notch signaling (24). Therefore, any additional factors that The only two genes that emerged from this screen showing might provide a gain in T lineage-specific regulatory function dur- sharp up-regulation during T lineage specification both turned out ing the DN1 to DN2 transition would be of great interest as com- to encode transcription factors. The alternate promoter use form of binatorial participants in the lineage specification process. A priori, HEB, HEBalt, was previously identified in this arrayed cDNA li- it was assumed that many transcription factors would be up- brary on the basis of its enrichment in Pro-T cells relative to pe- regulated in this interval as the cells began T lineage differen- ripheral lymphocytes (4, 31, 48). In the T lineage, its expression is tiation, and that the challenge would be to detect a subset with confined to stages before ␤-selection, it is positively regulated by functional importance. The results instead revealed remarkable both Notch signaling and canonical HEB, and its expression is continuities between the multilineage stem and progenitor cells down-regulated sharply if lymphoid precursors are diverted to a and the Pro-T cells. myeloid fate (48, 72, 73). Most recently, HEBalt has been shown The strategy we used to select the genes of interest was delib- to play a hit-and-run accelerating role in early T cell development, erately designed to enable us to recover legacy genes as well as distinct from that of canonical HEB (48). There is circumstantial strictly T lineage-specific ones. This decision was based on the evidence from conditional expression studies with the basic helix- evidence for low-level “multilineage priming” of stem cells for loop-helix transcription factor E2A that increases in effective E- expression of genes used in other hemopoietic lineages (60, 61). protein activity could provide some of the rate-limiting inductive The surprise was that the Pro-T cell-enriched genes identified by function for T lineage differentiation genes (69, 71). E2A itself this approach were so dominated by legacy genes, many of them appears to be regulated mostly through a dramatic increase at the virtually unchanging in their levels of expression from the multi- protein level between the DN1 and DN2 stages (74), even though potent progenitor to the Pro-T cell state. It is tempting to speculate its RNA levels remain relatively constant. The basis for this effect that this inherited assemblage of regulatory factors may contribute is not yet known. Thus, it is intriguing that the sharp increase in to the remarkable maintenance of developmental plasticity in in- HEBalt mRNA expression could provide a functionally distinctive trathymic Pro-T cells until just before ␤-selection (10, 12, 17, 22, interaction partner (48) for E2A protein, or a decoy partner for its 62–65) (T. Taghon, M. A. Yui, and E. V. Rothenberg, submitted Id-family antagonist, in the same developmental interval. An E2A for publication). Where genetic evidence is available, it confirms stabilization role would be consistent with the accelerating func- that legacy genes like Myb and Mll1 are indeed required for T tion that HEBalt has recently been shown to provide for entry into lineage differentiation (66–71), but they are not induced de novo in the T lineage pathway (48). this process. In addition to transcription factors, transcripts encod- The highly lineage-specific, Ͼ100-fold up-regulation of Bcl11b ing signaling molecules, adaptors, phosphatases, predicted RNA- in the DN2 stage (Fig. 5 and Ref. 55) stands alone as the most binding molecules, helicases, and others identified in our screen remarkable regulatory discontinuity we found to mark the entry were all found to be shared with prethymic BM LSK cells (sum- into the T lineage program. In a study of human lymphopoiesis, The Journal of Immunology 437

Dik et al. (28) also report dramatic up-regulation of human Bcl11b 2. Pelayo, R., R. Welner, S. S. Perry, J. Huang, Y. Baba, T. Yokota, and P. W. Kincade. at a stage similar to the murine LSK CD27ϩ to DN1 transition. 2005. Lymphoid progenitors and primary routes to becoming cells of the immune system. Curr. Opin. Immunol. 17: 100–107. Although already known to be expressed in a T lineage-specific 3. Bhandoola, A., A. Sambandam, D. Allman, A. Meraz, and B. Schwarz. 2003. way, Bcl11b was not expected to represent such a rarity. The tight Early T lineage progenitors: new insights, but old questions remain. J. Immunol. 171: 5653–5658. coupling between the timing of Bcl11b up-regulation and the ca- 4. Anderson, M. K. 2006. At the crossroads: diverse roles of early thymocyte tran- pability of a precursor cell to begin T cell differentiation indicates scriptional regulators. Immunol. Rev. 209: 191–211. that this gene responds to developmental stage-specific regulatory 5. Rothenberg, E. V., and T. Taghon. 2005. Molecular genetics of T cell develop- ment. Annu. Rev. Immunol. 23: 601–649. inputs in addition to Notch signaling. Continued Bcl11b expression 6. Borowski, C., C. Martin, F. Gounari, L. Haughn, I. Aifantis, F. Grassi, and is also closely correlated with maintenance of T lineage identity: H. von Boehmer. 2002. On the brink of becoming a T cell. Curr. Opin. Immunol. when DN2 and DN3 stage thymocytes are diverted to the myeloid 14: 200–206. 7. Hayday, A. C., and D. J. Pennington. 2007. Key factors in the organized chaos pathway by forced expression of PU.1, Bcl11b is one of the genes of early T cell development. Nat. Immunol. 8: 137–144. that is sharply down-regulated (72). This makes the regulatory 8. Ceredig, R., and T. Rolink. 2002. A positive look at double-negative thymocytes. system controlling Bcl11b a uniquely powerful indicator for the Nat. Rev. Immunol. 2: 888–897. 9. Benz, C., and C. C. Bleul. 2005. A multipotent precursor in the thymus maps to mechanism of T lineage specification. The key question raised by the branching point of the T versus B lineage decision. J. Exp. Med. 202: 21–31. our findings is what role Bcl11b may play during T lineage spec- 10. Balciunaite, G., R. Ceredig, and A. G. Rolink. 2005. The earliest subpopulation of mouse thymocytes contains potent T, significant macrophage, and natural ification as this gene undergoes its own rapid induction. Steady- killer cell but no B lymphocyte potential. Blood 105: 1930–1936. state analysis of germline Bcl11b knockout mice has shown only 11. Sambandam, A., I. Maillard, V. P. Zediak, L. Xu, R. M. Gerstein, J. C. Aster, that it is essential for normal passage from the DN3 to the DP stage W. S. Pear, and A. Bhandoola. 2005. Notch signaling controls the generation and differentiation of early T lineage progenitors. Nat. Immunol. 6: 663–670. (75) and has not identified any function yet in the DN1-DN2 tran-

12. Porritt, H. E., L. L. Rumfelt, S. Tabrizifard, T. M. Schmitt, J. C. Zuniga-Pflucker, Downloaded from sition. It is notable, however, that in the absence of Bcl11b, DN3 and H. T. Petrie. 2004. Heterogeneity among DN1 prothymocytes reveals mul- cells are quite abnormal and incapable of differentiation to the DP tiple progenitors with different capacities to generate T cell and non-T cell lin- eages. Immunity 20: 735–745. stage even when a TCR transgene is supplied or p53 checkpoint 13. David-Fung, E. S., M. A. Yui, M. Morales, H. Wang, T. Taghon, R. A. Diamond, control is removed (76, 77). Indeed, it is now known that abnor- and E. V. Rothenberg. 2006. Progression of regulatory gene expression states in malities of Bcl11b expression are found in ϳ20% of childhood T fetal and adult pro-T cell development. Immunol. Rev. 209: 212–236. 14. Tremblay, M., S. Herblot, E. Lecuyer, and T. Hoang. 2003. Regulation of pT␣ cell acute lymphoblastic leukemias (55, 56). Thus, Bcl11b activity gene expression by a dosage of E2A, HEB, and SCL. J. Biol. Chem. 278: during the DN2 stage, when it is first up-regulated, may guide the 12680–12687. http://www.jimmunol.org/ 15. Gounari, F., I. Aifantis, C. Martin, H. J. Fehling, S. Hoeflinger, P. Leder, generation of normal DN3 cells. Bcl11b could contribute to normal H. von Boehmer, and B. Reizis. 2002. Tracing lymphopoiesis with the aid of a T cell development either by acting directly on T lineage genes pT␣-controlled reporter gene. Nat. Immunol. 3: 489–496. through its newly reported positive regulatory activity (52), or by 16. Wang, H., R. A. Diamond, J. A. Yang-Snyder, and E. V. Rothenberg. 1998. Precocious expression of T cell functional response genes in vivo in primitive participating in the repression of residual antagonists of the T cell thymocytes before T lineage commitment. Int. Immunol. 10: 1623–1635. program, such as PU.1 or SCL/Tal-1, to enforce T lineage 17. King, A. G., M. Kondo, D. C. Scherer, and I. L. Weissman. 2002. Lineage commitment. infidelity in myeloid cells with TCR gene rearrangement: a latent developmental potential of pro-T cells revealed by ectopic cytokine receptor signaling. Proc. In summary, this study has defined about ninety genes enriched Natl. Acad. Sci. USA 99: 4508–4513. in early T cell precursors as compared with precursors embarking 18. Capone, M., R. D. Hockett, Jr., and A. Zlotnik. 1998. Kinetics of T cell receptor ␤, ␥, and ␦ rearrangements during adult thymic development: T cell receptor by guest on October 1, 2021 on myeloid differentiation, but of ϳ40 tested, the majority are ϩ ϩ rearrangements are present in CD44 CD25 pro-T thymocytes. Proc. Natl. already highly expressed in prethymic multilineage precursors and Acad. Sci. USA 95: 12522–12527. merely sustained in a T lineage-specific way. Using detailed quan- 19. Taghon, T. N., E. S. David, J. C. Zu´n˜iga-Pflu¨cker, and E. V. Rothenberg. 2005. Delayed, asynchronous, and reversible T lineage specification induced by Notch/ titative expression profiling in phenotypically defined subsets and Delta signaling. Genes Dev. 19: 965–978. in kinetic analyses of differentiation triggered by Notch/Delta sig- 20. Wu, L., C. L. Li, and K. Shortman. 1996. Thymic dendritic cell precursors: naling, one gene emerges as most intimately linked with the T relationship to the T lymphocyte lineage and phenotype of the dendritic cell progeny. J. Exp. Med. 184: 903–911. lineage specification process. This gene, Bcl11b, has had its func- 21. Shen, H. Q., M. Lu, T. Ikawa, K. Masuda, K. Ohmura, N. Minato, Y. Katsura, tions only hinted at in previous work and should provide a rich and H. Kawamoto. 2003. T/NK bipotent progenitors in the thymus retain the field for future study. potential to generate dendritic cells. J. Immunol. 171: 3401–3406. 22. Schmitt, T. M., M. Ciofani, H. T. Petrie, and J. C. Zu´n˜iga-Pflu¨cker. 2004. Main- tenance of T cell specification and differentiation requires recurrent notch recep- tor-ligand interactions. J. Exp. Med. 200: 469–479. Acknowledgments 23. Ikawa, T., S. Fujimoto, H. Kawamoto, Y. Katsura, and Y. Yokota. 2001. Com- We thank Drs. Mark Leid (Oregon State University) and Dorina Avram mitment to natural killer cells requires the helix-loop-helix inhibitor Id2. Proc. (Albany Medical College) for stimulating discussions about Bcl11b and for Natl. Acad. Sci. USA 98: 5164–5169. 24. Taghon, T., M. A. Yui, R. Pant, R. A. Diamond, and E. V. Rothenberg. 2006. generously sharing data before publication; John Cortese (Georgia Tech Developmental and molecular characterization of emerging ␤- and ␥␦-selected Research Institute) for helpful advice on terminology; Mary Yui (Califor- pre-T cells in the adult mouse thymus. Immunity 24: 53–64. nia Institute of Technology) for providing independent, confirmatory RNA 25. Maillard, I., T. Fang, and W. S. Pear. 2005. Regulation of lymphoid development, samples; Rochelle Diamond and Stephanie Adams (California Institute of differentiation, and function by the Notch pathway. Annu. Rev. Immunol. 23: 945–974. Technology) for expert advice and help with flow cytometry; Scott Bloom 26. Tabrizifard, S., A. Olaru, J. Plotkin, M. Fallahi-Sichani, F. Livak, and H. T. (Institute for Systems Biology) for excellent sequencing; Rob Butler, Petrie. 2004. Analysis of transcription factor expression during discrete stages of Robin Condie, Natasha Bouey, and Ruben Bayon (California Institute of postnatal thymocyte differentiation. J. Immunol. 173: 1094–1102. Technology) for excellent mouse care; and Shirley Pease and the Geneti- 27. Hoffmann, R., L. Bruno, T. Seidl, A. Rolink, and F. Melchers. 2003. Rules for gene usage inferred from a comparison of large-scale gene expression profiles of cally Engineered Mouse Service staff (California Institute of Technology) T and B lymphocyte development. J. Immunol. 170: 1339–1353. for timed pregnant mice. 28. Dik, W. A., K. Pike-Overzet, F. Weerkamp, D. de Ridder, E. F. de Haas, M. R. Baert, P. van der Spek, E. E. Koster, M. J. Reinders, J. J. van Dongen, et al. 2005. New insights on human T cell development by quantitative T cell receptor Disclosures gene rearrangement studies and gene expression profiling. J. Exp. Med. 201: The authors have no financial conflicts of interest. 1715–1723. 29. Lee, M. S., K. Hanspers, C. S. Barker, A. P. Korn, and J. M. McCune. 2004. Gene expression profiles during human CD4ϩ T cell differentiation. Int. Immunol. 16: 1109–1124. References 30. Rast, J. P., G. Amore, C. Calestani, C. B. Livi, A. Ransick, and E. H. Davidson. 1. Petrie, H. T., and P. W. Kincade. 2005. Many roads, one destination for T cell 2000. Recovery of developmentally defined gene sets from high-density cDNA progenitors. J. Exp. Med. 202: 11–13. macroarrays. Dev. Biol. 228: 270–286. 438 GENE DISCOVERY FOR T CELL SPECIFICATION IN VIVO AND IN VITRO

31. Anderson, M. K., G. Hernandez-Hoyos, R. A. Diamond, and E. V. Rothenberg. Disruption of the BCL11B gene through inv(14)(q11.2q32.31) results in the ex- 1999. Precise developmental regulation of Ets family transcription factors during pression of BCL11B-TRDC fusion transcripts and is associated with the absence specification and commitment to the T cell lineage. Development 126: of wild-type BCL11B transcripts in T-ALL. Leukemia 19: 201–208. 3131–3148. 57. Wochnik, G. M., J. Ruegg, G. A. Abel, U. Schmidt, F. Holsboer, and T. Rein. 32. Britten, R. J., D. E. Graham, and B. R. Neufeld. 1974. Analysis of repeating DNA 2005. FK506-binding proteins 51 and 52 differentially regulate dynein interaction sequences by reassociation. Methods Enzymol. 29: 363–418. and nuclear translocation of the glucocorticoid receptor in mammalian cells. 33. Press, W. H. 1986. Numerical Recipes: The Art of Scientific Computing. Cam- J. Biol. Chem. 280: 4609–4616. bridge University Press, Cambridge. 58. Schmitt, T. M., R. F. de Pooter, M. A. Gronski, S. K. Cho, P. S. Ohashi, and 34. Wiener, N. 1948. Cybernetics or Control and Communication in the Animal and J. C. Zuniga-Pflucker. 2004. Induction of T cell development and establishment the Machine. J. Wiley, New York. of T cell competence from embryonic stem cells differentiated in vitro. Nat. 35. Martin, D., C. Brun, E. Remy, P. Mouren, D. Thieffry, and B. Jacq. 2004. Immunol. 5: 410–417. GOToolBox: functional analysis of gene datasets based on gene ontology. Ge- 59. Ho¨flinger, S., K. Kesavan, M. Fuxa, C. Hutter, B. Heavey, F. Radtke, and nome Biol. 5: R101. M. Busslinger. 2004. Analysis of Notch1 function by in vitro T cell differentiation 36. Schmitt, T. M., and J. C. Zu´n˜iga-Pflu¨cker. 2002. Induction of T cell development of Pax5 mutant lymphoid progenitors. J. Immunol. 173: 3935–3944. from hematopoietic progenitor cells by Delta-like-1 in vitro. Immunity 17: 60. Hu, M., D. Krause, M. Greaves, S. Sharkis, M. Dexter, C. Heyworth, and 749–756. T. Enver. 1997. Multilineage gene expression precedes commitment in the he- 37. Rozen, S., and H. Skaletsky. 2000. Primer3 on the WWW for general users and mopoietic system. Genes Dev. 11: 774–785. Methods Mol. Biol. for biologist programmers. 132: 365–386. 61. Miyamoto, T., H. Iwasaki, B. Reizis, M. Ye, T. Graf, I. L. Weissman, and 38. Chen, F., L. Rowen, L. Hood, and E. V. Rothenberg. 2001. Differential tran- ␤ K. Akashi. 2002. Myeloid or lymphoid promiscuity as a critical step in hemato- scriptional regulation of individual TCR V segments before gene rearrange- poietic lineage commitment. Dev. Cell 3: 137–147. ment. J. Immunol. 166: 1771–1780. 62. Carlyle, J. R., A. M. Michie, C. Furlonger, T. Nakano, M. J. Lenardo, C. J. Paige, 39. Verbeek, S., D. Izon, F. Hofhuis, E. Robanus-Maandag, H. te Riele, and J. C. Zu´n˜iga-Pflu¨cker. 1997. Identification of a novel developmental stage M. van de Wetering, M. Oosterwegel, A. Wilson, H. R. MacDonald, and marking lineage commitment of progenitor thymocytes. J. Exp. Med. 186: H. Clevers. 1995. An HMG-box-containing T cell factor required for thymocyte 173–182. differentiation. Nature 374: 70–74. 40. Ioannidis, V., F. Beermann, H. Clevers, and W. Held. 2001. The ␤-catenin-TCF-1 63. Shortman, K., D. Vremec, L. M. Corcoran, K. Georgopoulos, K. Lucas, and pathway ensures CD4ϩCD8ϩ thymocyte survival. Nat. Immunol. 2: 691–697. L. Wu. 1998. The linkage between T cell and dendritic cell development in the Downloaded from 41. Goux, D., J. D. Coudert, D. Maurice, L. Scarpellino, G. Jeannet, S. Piccolo, mouse thymus. Immunol. Rev. 165: 39–46. K. Weston, J. Huelsken, and W. Held. 2005. Cooperating pre-T cell receptor and 64. Lee, C.-K., J. K. Kim, Y. Kim, M.-K. Lee, K. Kim, J.-K. Kang, R. Hofmeister, TCF-1-dependent signals ensure thymocyte survival. Blood 106: 1726–1733. S. K. Durum, and S. S. Han. 2001. Generation of macrophages from early T 42. Zediak, V. P., I. Maillard, and A. Bhandoola. 2005. Closer to the source: Notch progenitors in vitro. J. Immunol. 166: 5964–5969. and the nature of thymus-settling cells. Immunity 23: 245–248. 65. Weerkamp, F., M. R. Baert, M. H. Brugman, W. A. Dik, E. F. de Haas, 43. Wu, L. 2006. T lineage progenitors: the earliest steps en route to T lymphocytes. T. P. Visser, C. J. de Groot, G. Wagemaker, J. J. van Dongen, and F. J. Staal. Curr. Opin. Immunol. 18: 121–126. 2006. Human thymus contains multipotent progenitors with T/B lymphoid, my-

44. Weerkamp, F., M. R. Baert, B. A. Naber, E. E. Koster, E. F. de Haas, eloid, and erythroid lineage potential. Blood 107: 3131–3137. http://www.jimmunol.org/ K. R. Atkuri, J. J. van Dongen, L. A. Herzenberg, and F. J. Staal. 2006. Wnt 66. Ernst, P., J. Wang, and S. J. Korsmeyer. 2002. The role of MLL in hematopoiesis signaling in the thymus is regulated by differential expression of intracellular and leukemia. Curr. Opin. Hematol. 9: 282–287. signaling molecules. Proc. Natl. Acad. Sci. USA 103: 3322–3326. 67. Ernst, P., J. K. Fisher, W. Avery, S. Wade, D. Foy, and S. J. Korsmeyer. 2004. 45. Staal, F. J., and H. C. Clevers. 2005. WNT signalling and haematopoiesis: a Definitive hematopoiesis requires the mixed-lineage leukemia gene. Dev. Cell 6: WNT-WNT situation. Nat. Rev. Immunol. 5: 21–30. 437–443. 46. Igarashi, H., S. C. Gregory, T. Yokota, N. Sakaguchi, and P. W. Kincade. 2002. 68. Fidanza, V., P. Melotti, T. Yano, T. Nakamura, A. Bradley, E. Canaani, Transcription from the RAG1 locus marks the earliest lymphocyte progenitors in B. Calabretta, and C. M. Croce. 1996. Double knockout of the ALL-1 gene blocks bone marrow. Immunity 17: 117–130. hematopoietic differentiation in vitro. Cancer Res. 56: 1179–1183. 47. Liu, W. H., and M. Z. Lai. 2005. Deltex regulates T cell activation by targeted 69. Emambokus, N., A. Vegiopoulos, B. Harman, E. Jenkinson, G. Anderson, and degradation of active MEKK1. Mol. Cell. Biol. 25: 1367–1378. J. Frampton. 2003. Progression through key stages of haemopoiesis is dependent 48. Wang, D., C. L. Claus, G. Vaccarelli, M. Braunstein, T. M. Schmitt, J. C. on distinct threshold levels of c-Myb. EMBO J. 22: 4478–4488. Zuniga-Pflucker, E. V. Rothenberg, and M. K. Anderson. 2006. The basic helix- 70. Lieu, Y. K., A. Kumar, A. G. Pajerowski, T. J. Rogers, and E. P. Reddy. 2004. by guest on October 1, 2021 loop-helix transcription factor HEBAlt is expressed in pro-T cells and enhances Requirement of c-myb in T cell development and in mature T cell function. Proc. the generation of T cell precursors. J. Immunol. 177: 109–119. Natl. Acad. Sci. USA 101: 14853–14858. 49. Guttinger, M., F. Sutti, M. Panigada, S. Porcellini, B. Merati, M. Mariani, 71. Allen, R. D., III, T. P. Bender, and G. Siu. 1999. c-Myb is essential for early T T. Teesalu, G. G. Consalez, and F. Grassi. 1998. Epithelial V-like antigen (EVA), cell development. Genes Dev. 13: 1073–1078. a novel member of the immunoglobulin superfamily, expressed in embryonic 72. Franco, C. B., D. D. Scripture-Adams, I. Proekt, T. Taghon, A. H. Weiss, epithelia with a potential role as homotypic adhesion molecule in thymus histo- M. A. Yui, S. L. Adams, R. A. Diamond, and E. V. Rothenberg. 2006. Notch/ genesis. J. Cell Biol. 141: 1061–1071. Delta signaling constrains reengineering of pro-T cells by PU.1. Proc. Natl. Acad. 50. Rothenberg, E. V., and C. J. Dionne. 2002. Lineage plasticity and commitment in Sci. USA 103: 11993–11998. T cell development. Immunol. Rev. 187: 96–115. 73. Dionne, C. J., K. Y. Tse, A. H. Weiss, C. B. Franco, D. L. Wiest, M. K. Anderson, 51. Ciofani, M., and J. C. Zuniga-Pflucker. 2005. Notch promotes survival of pre-T and E. V. Rothenberg. 2005. Subversion of T lineage commitment by PU.1 in a cells at the ␤-selection checkpoint by regulating cellular metabolism. Nat. Im- clonal cell line system. Dev. Biol. 280: 448–466. munol. 6: 881–888. 74. Engel, I., C. Johns, G. Bain, R. R. Rivera, and C. Murre. 2001. Early thymocyte 52. Cismasiu, V. B., S. Ghanta, J. Duque, D. I. Albu, H. M. Chen, R. Kasturi, and development is regulated by modulation of E2A protein activity. J. Exp. Med. D. Avram. 2006. BCL11B participates in the activation of IL2 gene expression in ϩ 194: 733–745. CD4 T lymphocytes. Blood 108: 2695–2702. 53. Avram, D., A. Fields, K. Pretty On Top, D. J. Nevrivy, J. E. Ishmael, and 75. Wakabayashi, Y., H. Watanabe, J. Inoue, N. Takeda, J. Sakata, Y. Mishima, M. Leid. 2000. Isolation of a novel family of C H zinc finger proteins implicated J. Hitomi, T. Yamamoto, M. Utsuyama, O. Niwa, et al. 2003. Bcl11b is required 2 2 ␣␤ in transcriptional repression mediated by chicken ovalbumin upstream promoter for differentiation and survival of T lymphocytes. Nat. Immunol. 4: 533–539. transcription factor (COUP-TF) orphan nuclear receptors. J. Biol. Chem. 275: 76. Inoue, J., T. Kanefuji, K. Okazuka, H. Watanabe, Y. Mishima, and R. Kominami. ␣␤ 10315–10322. 2006. Expression of TCR partly rescues developmental arrest and apoptosis ␣␤ Ϫ/Ϫ 54. Avram, D., A. Fields, T. Senawong, A. Topark-Ngarm, and M. Leid. 2002. of T cells in Bcl11b mice. J. Immunol. 176: 5871–5879. COUP-TF (chicken ovalbumin upstream promoter transcription factor)-interact- 77. Okazuka, K., Y. Wakabayashi, M. Kashihara, J. Inoue, T. Sato, M. Yokoyama, ing protein 1 (CTIP1) is a sequence-specific DNA binding protein. Biochem J. S. Aizawa, Y. Aizawa, Y. Mishima, and R. Kominami. 2005. p53 prevents mat- Ϫ ϩ Ϫ Ϫ 368: 555–563. uration of T cell development to the immature CD4 CD8 stage in Bcl11b / 55. Su, X. Y., V. Della-Valle, I. Andre-Schmutz, C. Lemercier, I. Radford-Weiss, mice. Biochem. Biophys. Res. Commun. 328: 545–549. P. Ballerini, M. Lessard, M. Lafage-Pochitaloff, F. Mugneret, R. Berger, et al. 78. Spits, H. 2002. Development of ␣␤ T cells in the human thymus. Nat. Rev. 2006. HOX11L2/TLX3 is transcriptionally activated through T-cell regulatory Immunol. 2: 760–772. elements downstream of BCL11B as a result of the t(5;14)(q35;q32). Blood 108: 79. Carrasco, Y. R., C. Trigueros, A. R. Ramiro, V. G. de Yebenes, and 4198–4201. M. L. Toribio. 1999. ␤-Selection is associated with the onset of CD8␤ chain 56. Przybylski, G. K., W. A. Dik, J. Wanzeck, P. Grabarczyk, S. Majunke, expression on CD4ϩCD8␣␣ pre-T cells during human intrathymic development. J. I. Martin-Subero, R. Siebert, G. Dolken, W. D. Ludwig, B. Verhaaf, et al. 2005. Blood 94: 3491–3498.