Leukemia (2007) 21, 311–319 & 2007 Nature Publishing Group All rights reserved 0887-6924/07 $30.00 www.nature.com/leu ORIGINAL ARTICLE

T-, B- and NK-lymphoid, but not myeloid cells arise from human CD34 þ CD38ÀCD7 þ common lymphoid progenitors expressing lymphoid-specific genes

I Hoebeke1,3, M De Smedt1, F Stolz1,4, K Pike-Overzet2, FJT Staal2, J Plum1 and G Leclercq1

1Department of Clinical Chemistry, Microbiology and Immunology, Ghent University Hospital, Ghent, Belgium and 2Department of Immunology, Erasmus University Medical Center, Rotterdam, The Netherlands

Hematopoietic stem cells in the bone marrow (BM) give rise to share a direct common progenitor either, as CLPs were not all blood cells. According to the classic model of hematopoi- found in the fetal liver.5 Instead, fetal B and T cells would esis, the differentiation paths leading to the myeloid and develop through B/myeloid and T/myeloid intermediates. lymphoid lineages segregate early. A candidate ‘common 6 lymphoid progenitor’ (CLP) has been isolated from The first report of a human CLP came from Galy et al. who À þ CD34 þ CD38À human cord blood cells based on CD7 expres- showed that a subpopulation of adult and fetal BM Lin CD34 sion. Here, we confirm the B- and NK-differentiation potential of cells expressing the early B- and T-cell marker CD10 is not þ À þ CD34 CD38 CD7 cells and show in addition that this capable of generating monocytic, granulocytic, erythroid or population has strong capacity to differentiate into T cells. As megakaryocytic cells, but can differentiate into dendritic cells, CD34 þ CD38ÀCD7 þ cells are virtually devoid of myeloid B, T and NK cells. These LinÀCD34 þ CD10 þ cells homo- differentiation potential, these cells represent true CLPs. To 7 unravel the molecular mechanisms underlying lymphoid com- genously expressed CD38. According to Ishii et al. expression þ mitment, we performed genome-wide gene expression profiling of the chemokine receptor CXCR4 on BM CD34 cells would on sorted CD34 þ CD38ÀCD7 þ and CD34 þ CD38ÀCD7À cells. be sufficient to restrict their differentiation potential to the Interestingly, lymphoid-affiliated genes were mainly upregu- lymphoid lineage. A human CMP was recently also identified in þ lated in the CD7 population, while myeloid-specific genes the LinÀCD34 þ CD38 þ fraction of BM and cord blood. These were downregulated. This supports the hypothesis that lineage CMPs are CD45RAÀ and express low levels of IL-3Ra.8 commitment is accompanied by the shutdown of inappropriate þ gene expression and the upregulation of lineage-specific In cord blood, expression of CD10 on CD34 cells does not genes. In addition, we identified several highly expressed discriminate progenitor cells with lymphoid-restricted potential genes that have not been described in hematopoiesis before. from multipotent cells.9 However, Hao et al.9 detected in the Leukemia (2007) 21, 311–319. doi:10.1038/sj.leu.2404488; most primitive CD34 þ CD38À cord blood fraction a subpopula- published online 14 December 2006 tion expressing CD7, an antigen that was previously identified Keywords: human; hematopoiesis; stem cells; cord blood; lymphoid progenitor on early human T-lymphoid progenitors, and they showed that single CD34 þ CD38ÀCD7 þ cord blood cells can generate B cells, NK cells and dendritic cells, but are devoid of myeloid or Introduction erythroid differentiation potential. T-cell potential was not addressed by these investigators. All blood cells ultimately derive from a rare population of In a recent study, Haddad et al.10 compared the differentiation hematopoietic stem cells in the bone marrow (BM) that are potential of cord blood CD34 þ CD45RAhiLinÀCD10 þ cells, multipotent and have the ability to self-renew. According to the which correspond to the BM CLP, with that of cord blood classic model of hematopoiesis, all lymphoid cells (T, B and CD34 þ CD45RAhiCD7 þ cells, which comprise the natural killer (NK) cells) develop through a common precursor CD34 þ CD38ÀCD7 þ CLP, as these uniformly express CD45RA. stage, the so-called ‘common lymphoid progenitor’ (CLP), and The authors showed that the differentiation potential of accordingly, cells from the myeloid lineages share a ‘common CD34 þ CD45RAhiCD7 þ cells is skewed toward the T/NK myeloid progenitor’ (CMP). This model was supported by the lineages, while CD34 þ CD45RAhiLinÀCD10 þ cells predomi- prospective isolation of cell populations with CLP and CMP nantly possess B-cell differentiation potential. Additionally, both 1,2 function from the murine BM. Recent evidence, however, populations retain some degree of myeloid differentiation indicated that BM CLPs are not physiological T-cell progenitors, capacity. Gene expression data from microarray analyses as early thymic progenitors (ETPs) do not have the CLP supported their conclusions. 3 4 phenotype and CLPs are not present in the peripheral blood. In the present study, we confirm that the CD34 þ CD38ÀCD7 þ Instead, the thymus is most likely seeded by a multipotent cord blood population is lymphoid-committed and we show that progenitor. During fetal hematopoiesis, B and T cells do not it also has strong T-lymphoid differentiation potential in fetal thymus organ culture (FTOC). To investigate the molecular Correspondence: Professor G Leclercq, Department of Clinical mechanisms driving lymphoid commitment, we studied the Chemistry, Microbiology and Immunology, Ghent University Hospital, differential gene expression between CD34 þ CD38ÀCD7À multi- 4 Blok A, De Pintelaan 185, B-9000 Ghent, Belgium. þ À þ E-mail: [email protected] potent cells and CD34 CD38 CD7 CLP cells using Affymetrix 3Current address: I Hoebeke, Diabetes Research Center, Brussels Free GeneChip technology. University (VUB), Laarbeeklaan 103, B-1090 Brussels, Belgium. 4 Current address: Laboratory of Molecular Cell Biology, Institute of Materials and methods Botany and Microbiology, Katholieke Universiteit Leuven and Department of Molecular Microbiology, Flanders Interuniversity Institute of Biotechnology (VIB), Leuven, Belgium Cell sorting Received 31 August 2006; accepted 19 October 2006; published Cord blood was obtained and used following the guidelines of online 14 December 2006 the Medical Ethical Commission of the Ghent University Molecular characterization of human cord blood CLP I Hoebeke et al 312 Hospital. Within 12 h after collection of human umbilical cord Flow cytometry blood samples, mononuclear cells were isolated over a Before labelling with antibodies, cells were pre-incubated Lymphoprep density-gradient (Axis-Shield PoC AS, Oslo, 15 min with anti-mouse FcRgII/III (clone 2.4.G2, a kind gift of Norway) and CD34 þ cells were isolated by positive selection with Dr J Unkeless, Mount Sinai School of Medicine, New York, NY, MACS magnetic beads (Miltenyi Biotec, Bergisch Gladbach, USA) and human IgG (Miltenyi Biotec) to block murine and Germany). Cells were labelled with anti-CD34-allophycocyanin human Fc receptors, respectively. Cells were incubated with (APC), anti-CD38-phycoerythrin (PE) and anti-CD7-fluorescein appropriate amounts of combinations of the following mouse isothiocyanate (FITC) monoclonal antibodies (BD Biosciences, anti-human monoclonal antibodies: CD19-PE, CD34-APC, San Jose, CA, USA) and CD34 þ CD38ÀCD7 þ and CD4-APC, CD33-FITC, CD14-FITC (all from BD Biosciences), CD34 þ CD38ÀCD7À cells were sorted with a FACSVantage CD56-APC and CD8b-PE (both from Immunotech, Beckman Cell sorter (Becton and Dickinson Immunocytometry Systems Coulter, Fullerton, CA, USA). Cell populations containing (BDIS), San Jose, CA, USA). The purity of the sorted cells was mouse leukocytes (from FTOC) were simultaneously stained checked on a FACSCalibur (BDIS) and was always 495%. with anti-mouse CD45-CyChrome (BD Pharmingen, San Diego, Sorted cells were either directly used in MS-5 co-culture or CA, USA). After 45 min, cells were washed with ice-cold FTOC, or either stored in 200 ml TRIZOL (Invitrogen, Carlsbad, PBS þ 1% BSA þ 0.1% NaN3, propidium iodide (4 mg/ml) was CA,USA)atÀ701C for later RNA isolation and use in added and cells were analyzed on a FACSCalibur. Propidium microarray experiments or real-time PCR. iodide positive and mouse CD45 þ cells, representing dead cells and mouse leukocytes, respectively, were excluded from analysis, which was performed with CellQuest software (BDIS).

Co-culture on MS-5 stromal cells The differentiation of stem cells to most lymphoid (except RNA isolation and amplification The TRIZOL lysates of different sorts, corresponding to a total of T cells) and myeloid cell types can be accomplished in vitro by culture in the presence of the appropriate human recombinant 100 000 sorted cells, were pooled and total RNA was extracted and purified on an RNeasy column (Qiagen, Venlo, The Nether- cytokines on a feeder layer of the murine stromal cell line MS-5.11 Four days before their use in co-culture experiments, MS-5 lands) according to the instructions of the manufacturers. The RNA was concentrated to 10 ml with Microcon YM-50 columns cells (kindly provided by L Coulombel, Institut Gustave Roussy, (Millipore, Billerica, MA, USA) and subjected to Degenerative Villejuif, France) were seeded in 96-well plates at a density of 3 Oligonucleotide Primer- (DOP) mediated amplification. The 5 Â 10 cells per well. Co-cultures were initiated by incubating detailed protocol, which was developed in our lab, is available human sorted cells in 200 ml Iscove’s Modified Dulbecco’s upon request. Briefly, mRNA was first reverse transcribed using a Medium (IMDM) (Invitrogen) supplemented with 5% human AB T7-promoter oligo(dT) primer. After RNase H treatment, second serum (Valley Biomedical, Winchester, VA, USA), 5% fetal calf strand synthesis was initiated using the 22-nt DOP-primer. In vitro serum (FCS), 100 U/ml penicillin, 100 mg/ml streptomycin, 2 mM transcription of the cDNA with T7 RNA-polymerase was carried L-glutamin (all from Invitrogen) and the following cytokines: SCF out to generate cRNA, which was used in a second round of (50 ng/ml), FL (50 ng/ml), TPO (10 ng/ml), IL-2 (5 ng/ml), IL-7 amplification using random hexamers for synthesis of the first (20 ng/ml) and IL-15 (10 ng/ml) (mix 6) (all cytokines from R&D strand, and the T7-promoter oligo(dT) primer for synthesis of the Systems, Abingdon, UK). After 20 days of culture at 371C and second strand. cDNA was transcribed and biotin-labelled using 7% CO , the progeny of the cells was counted with a Burker 2 ¨ the ENZO BioArray HighYield RNA Transcript Labeling Kit hemocytometer excluding dead cells with Trypan blue, and (ENZO, Farmingdale, NY, USA) according to the manufacturer’s their phenotype was determined by flow cytometry. instructions. Biotinylated cRNA was purified on an RNeasy column and its quality was determined on an Agilent 2100 BioAnalyzer (Agilent, Palo Alto, CA, USA). RNA isolation and amplification was performed twice for both CD34 þ CD38ÀCD7 þ Fetal thymus organ culture and CD34 þ CD38ÀCD7À cells (biological duplicates). Thymic lobes were isolated from fetal day 15 nonobese diabetic-severe combined immunodeficient (NOD-SCID) mice obtained from our own pathogen-free breeding facility. Animals Microarray analysis were treated according to the guidelines of the Laboratory Affymetrix microarray analysis was performed according to Animal Ethical Commision of the Ghent University Hospital. Minimum Information About a Microarray Experiment (MIAME) Each lobe was placed in a well of a Terasaki-plate and 25 ml guidelines (www.mged.org/Workgroups/MIAME/miame.html). complete IMDM medium containing 1000 human cells was Biotinylated cRNA was fragmented and hybridized to Affymetrix added. The plates were inverted and incubated at 371C with 7% GeneChip arrays according to the guidelines of the manufac- CO2 for 72 h. After this ‘hanging drop’ culture, during which the turer. In a first experiment, biotinlylated cRNA was hybridized precursor cells migrate into the thymic lobes, the lobes were to Affymetrix HG-U133A arrays, while in a second experiment it transferred to a Nuclepore polycarbonate membrane (Whatman, was hybridized to Affymetrix HG-U133 Plus 2.0 arrays. Brentford, UK) on a Gelfoam sponge (Pharmacia & Upjohn, Statistical analysis of the microarray data was performed as Kalamazoo, MI, USA) soaked in complete IMDM medium described before.12 Briefly, after background removal and supplemented with 10% human AB serum and cultured for 32 quantile normalization by Robust Multi-chip Average (RMA) 13 days at 371C with 7% CO2. After the first 14 days, half of the analysis, the raw perfect match (PM) probe intensity levels medium was replaced with fresh medium. Thymocytes were were used in a per probe set two-way analysis of variance harvested by mechanical disruption of the thymic structure and (ANOVA) (with factors ‘probe’ and ‘cell population’) to generate viable cells were counted by Trypan blue exclusion. Cells were an average expression level for the two biological repeats and a stained with appropriate antibodies and analyzed by flow P-value for the significance of the difference between the cytometry. CD34 þ CD38ÀCD7 þ and CD34 þ CD38ÀCD7À cell popula-

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 313 tions. The P-values were adjusted for multiple testing using Sidak step-down adjustment and differences with adjusted P o0.05 were considered significant.

Bioinformatics The differentially expressed genes were categorized according to Gene Ontology (GO) terms using the Affymetrix NetAffx center (http://www.affymetrix.com/analysis/index.affx) and the freely available programs Onto-Express14 and Ease (http:// david.niaid.nih.gov/david/ease.htm). Further information on the genes was gained by manually searching OMIM (http:// www.ncbi.nlm.nih.gov/entrez/query.fcgi?db ¼ OMIM) and Gene- Card databases (http://www.genecards.org/), which both have links to PubMed literature references. Expression pattern information was derived from GeneNote, a database of human genes and their expression profiles in healthy tissues based on microarray experiments performed on the Affymetrix HG-U95 set (http://genecards.weizmann.ac.il/genenote/), and SymAtlas (http://symatlas.gnf.org/SymAtlas/). Mapping of the genes on pathways and networks was done using the commercial package Ingenuity Pathways Analysis (Ingenuity, Redwood City, CA, USA). MatchMiner (http://discover.nci.nih.gov/matchminer/ index.jsp) was used to find common genes in the lists of differentially expressed genes of our study and those of Dik et al.12 and van Zelm et al.15

Real-time PCR analysis Total RNA was extracted from sorted cells using TRIZOL and Figure 1 Isolation of CD34 þ CD38ÀCD7 þ and CD34 þ CD38ÀCD7À was DNAse treated on an RNeasy column according to the cells from human umbilical cord blood. (a) Indication of sorting manufacturer’s guidelines. The RNA was concentrated with strategy. Cells in the upper left and upper right quadrants of the right dot plot showing CD7 expression on electronically gated Microcon YM-50 columns and oligo(dT)-primed reverse tran- þ À CD34 CD38 cells were sorted. (b) Reanalysis of sorted cells showed scription was performed with SuperScript II (Invitrogen). Real- purity of 495%. time PCR analysis with Sybr Green I (Eurogentec, Seraing, Belgium) was performed with an ABI PRISM 7000 (Applied Biosystems, Foster City, CA, USA) using the standard tempera- with CD7 þ cells and 30 wells with CD7À cells. After 20 days of ture protocol (40 cycles of 10 min 951C, 15 s 951C, 60 s 601C). culture, cells were counted and stained with antibodies for Reaction mixtures contained 300 nM of forward and reverse CD19, CD56 and CD33 to identify B cells, NK cells and primers and 0.04% BSA (Sigma, St Louis, MO, USA). Primers myeloid cells, respectively. On cells from the 20 wells with the were designed using Primer Express 2.0 software (Applied highest cell number, a second staining was performed with Biosystems) and sequences can be found in Supplementary antibodies against the hematopoietic stem cell marker CD34 Table S4. Expression levels were normalized to the expression of and the monocyte marker CD14. Five wells initiated with CD7 þ 16 the reference gene HPRT using the DDCT method. cells did not contain enough cells for reliable flow cytometry analysis and were excluded from analysis. Cytokine mix 6 is optimized for expansion of CD34 þ stem cells and differentia- Results tion towards monocytes, but also supports differentiation towards B and NK cells.17 CD7À cells proliferated well and CD34 þ CD38ÀCD7 þ cord blood cells generate high most cells differentiated into the myeloid lineage (CD33 þ ), numbers of B and NK cells but few myeloid cells in while also a low number of cells remained CD34 þ . On the co-cultures on MS-5 stromal cells contrary, the wells initiated with CD7 þ cells contained mainly To confirm that CD34 þ CD38ÀCD7 þ cord blood cells are B and NK cells, almost no CD34 þ cells and few myeloid cells. committed to the lymphoid lineage, we stringently sorted both Absolute total cell numbers were 13-fold reduced compared to CD34 þ CD38ÀCD7 þ (hereafter called CD7 þ ) and CD34 þ wells initiated with CD7À cells. Remarkably, the proportion of B CD38ÀCD7À (hereafter called CD7À) populations (Figure 1) and and NK cells in the different wells initiated with CD7 þ cells was determined their differentiation potential in MS-5 co-cultures highly variable, with most wells being enriched in either of the with a mix of six cytokines (SCF, FL, TPO, IL-2, IL-7 and IL-15) two cell types, and only five wells containing similar numbers of permitting both lymphoid and myeloid differentiation. In initial B and NK cells. This explains the large s.d. on the average experiments, we noticed that the CD7À cells proliferated much frequencies of B and NK cells. faster than the CD7 þ cells. Therefore, even the slightest contamination of the CD7 þ population with CD7À cells would obscure the real differentiation potential of the CD7 þ cells. For CD34 þ CD38ÀCD7 þ cord blood cells efficiently this reason, the sorted cells were plated at 10 cells per well, generate T cells in FTOC which ensures that most wells only contain CD7 þ cells. Figure 2 The co-culture system on MS-5 stromal cells does not support gives an overview of a typical experiment consisting of 50 wells T-cell differentiation. The capacity of the CD7 þ and CD7À cells

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 314

Figure 2 CD34 þ CD38ÀCD7 þ cord blood cells have lymphoid-restricted differentiation potential. (a) Flow cytometric analysis of MS-5 co-cultures with cytokine mix 6 (SCF, FL, TPO, IL-2, IL-7 and IL-15) after 20 days. Dot plots for three wells of 50 wells initiated with 10 CD7 þ cells and three wells of 30 wells initiated with 10 CD7À cells are shown. (b) Mean frequencies of B cells (CD19 þ ), NK cells (CD56 þ ), myeloid cells (CD33 þ ) and HSCs (CD34 þ ) obtained with CD7 þ cells and with CD7À cells. Error bars represent the s.d. (c) Average absolute cell number obtained from 10 cells after 20 days of MS-5 culture with cytokine mix 6.

to generate T cells was compared in FTOC. Three independent experiments were initiated with 1000 cells per thymic lobe. Thymocytes were harvested after 32 days of organ culture and analyzed for the expression of CD4 and CD8b (Figure 3). At that time, cultures initiated with CD7 þ cells clearly had generated a higher percentage of double positive (DP) thymocytes than those initiated with CD7À cells. Also from the absolute cell number it is clear that CD7 þ cells performed much better in FTOC than CD7À cells: CD7 þ cells generated about 10 times more cells than the CD7À cells (15 019775668 versus 14 549717 886). These data indicate that, as expected from a CLP, CD7 þ cells generate T cells with a much faster kinetics than the more immature CD7À cells.

RNA amplification and microarray analysis Figure 3 CD34 þ CD38ÀCD7 þ cord blood cells have strong T-cell þ À þ generation capacity. Flow cytometric analysis of FTOC after 32 days of As a result of the rare nature of the CD34 CD38 CD7 cord þ À blood population, it was necessary to perform RNA amplifica- culture starting from CD7 and CD7 sorted human cord blood cells. Percentages of CD4 þ CD8b þ double positive thymocytes obtained in tion to obtain a sufficient amount of RNA to hybridize on the three independent experiments are indicated in the upper right GeneChips. For each of two separate experiments, total RNA quadrant. was isolated from a total of 100 000 sorted CD34 þ CD38ÀCD7 þ and CD34 þ CD38ÀCD7À cells pooled from three to eight separate sorts, in each of which one to three cord blood units were pooled. This guaranteed a normalization correlation coefficient between the two biological repeats. As of interindividual sources of variation. shown in Figure 4, the correlation between the biological In the first microarray experiment, cRNA was hybridized to repeats was very high (correlation coefficients of 0.92 and 0.96 Affymetrix HG-U133A arrays, which are comprised of more for the CD7À and the CD7 þ arrays, respectively), which than 22 000 probesets representing 18 400 transcripts, including allowed us to use the average expression values of the two 14 500 well-characterized human genes. In the second experi- repeats for further analysis. Of the 22 215 probesets that were ment, cRNA was hybridized to Affymetrix HG-U133 Plus 2.0 present on both GeneChips, 201 were significantly differentially arrays, which are comprised of more than 54 000 probesets expressed (adjusted Po0.05) (see Supplementary Table S1). Of representing 47 000 transcripts, including 38 500 well-charac- these, 110 probesets representing 101 genes were upregulated terized human genes. The expression values of the 22 215 in the CD7 þ population, while 91 probesets representing 89 common probesets in both experiments were used to calculate a genes were downregulated in the CD7 þ population.

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 315 Validation of microarray data transporters and signal transduction molecules. The list of The mRNA expression of the markers used for cell sorting, downregulated genes is significantly enriched for structural namely CD34, CD38 and CD7, as measured by the microarrays, components of the ribosome, components of the cytoskeleton, correlated well with their cell surface expression on the sorted signal transduction molecules and molecules involved in populations. To further validate the microarray data, the protein biosynthesis and cell proliferation. The differentially expression of 11 randomly chosen transcripts was analyzed by expressed genes were also imported into Ingenuity Pathways real-time PCR on unamplified RNA isolated from newly sorted Analysis software to identify functional relationships between cells. Except for RGS2, fold changes obtained by real-time PCR genes based on known interactions in the literature (Table 1). analysis correlated well with those obtained by the microarray Interestingly, the biological function ‘hematological system analysis (Figure 5). This indicates that the differences in gene development and function’ is associated with three of the four expression detected by the microarray analysis are reliable and highest ranked biological networks. that the linear amplification of the RNA before the microarray Next, we did an extensive literature survey for the differen- analysis did not induce a bias in the relative presence of tially expressed genes. The gathered gene information is added individual RNA molecules. to Supplementary Tables S2 and S3. About one-third of both upregulated and downregulated genes has been reported to be expressed in hematopoietic tissue. These genes are underlined. Remarkably, more than 75% of the differentially expressed Analysis of the microarray data transcription factors have been involved in hematopoiesis. In Using annotation software, including Onto-Express, Ease and addition, many differentially expressed genes are involved in the NetAffx Analysis Center from Affymetrix, the significantly cell proliferation, gene expression regulation, cytoskeleton differentially expressed genes were grouped into functional regulation and protein degradation. In the following paragraphs, categories (see Supplementary Tables S2 and S3). Figure 6 we discuss these genes in more detail. shows the GO terms of the categories Molecular Function, Biological Process and Cellular Component that have at least three genes annotated to them. Gene categories that are Genes involved in hematopoiesis significantly over-represented in the list of upregulated genes þ Several genes that were significantly upregulated in the CD7 include transcription factors, RNA-binding molecules, compo- population play an essential role in lymphoid development. nents of the ubiquitin-protein ligase complex, splice factors, IGHM, BTK, PRKCB1 and BCL6 are essential for B-cell development; SMARCA4 and SATB1 are essential for T-cell development, while ADA and TCF4 (alias E2-2) are required for both B- and T-cell development. Other upregulated genes with a known function in lymphoid cells include the chemokine receptor CCR9 and the trancription factors MEF2A and RUNX3. In addition, many significantly upregulated genes are highly expressed in lymphoid tissues, suggesting a possible role in lymphoid development. These include MCM3AP, TRAF3IP3, CHD3, ARHGAP25, FLJ22635, c11orf21, TRAF4, TNFAIP3, EVL, FLJ13197, BASP1 and ADAM28. Several other significantly upregulated genes are expressed in the hematopoietic system, namely BAALC, ITGA4 (alias CD49D), TRIB2, TNFAIP2, PSCD4, P2RY14, HLX1, EMR2, LRRFIP1, TRIM33, TRAF5, NALP1 and OSBPL3 (alias ORP3). Figure 4 Correlation between the microarray experiments. Expres- sion levels for each of the 22 215 common probesets after hybridiza- Among the significantly downregulated genes a considerable tion on the HG-U133A chip (first experiment) and the HG-U133 Plus number of genes are affiliated to myeloid differentiation. For 2.0 chip (second experiment) are shown. The correlation coefficient R instance, the transcription factors TAL1 (alias SCL), CEBPB (alias is indicated. NF-IL6), EVI1 and IFI16 are well known for regulating myeloid

Figure 5 Validation of microarray data by real-time PCR. The fold change (expression in CD7 þ cells relative to expression in CD7À cells) of a selection of differentially expressed genes as determined by microarray analysis and real-time PCR analysis. Bars represent the s.d. on the mean of duplicate PCR reactions, except for RGS2 and PRKCB1, where they represent the s.d. on the mean of biological repeats. mRNA levels were normalized to HPRT mRNA expression.

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 316

Figure 6 Functional profile of significantly upregulated and downregulated genes constructed by Onto-Express. GO terms in the categories Molecular Function, Biological Process and Cellular Component which have at least three genes annotated to them are shown. The number of probesets and percentage of probesets annotated to each GO term are indicated. Note that a probeset can be annotated to more than one GO term. GO terms that are significantly overrepresented in the lists of up- and downregulated genes (Bonferroni-corrected Po0.05) are shown in black bars.

development. TNFSF10 (alias TRAIL), RGS2 and DLK1 (alias although the downregulation detected by the microarray Pref-1) are also assumed to be involved in myeloid differentia- analysis was not statistically significant. Several of the down- tion. IL1B, PRG1, UROD, PRDX2 and ICAM4 are specifically regulated genes might be associated with HSC function, and expressed in myeloid cell types. In addition, several down- their downregulation might, therefore, be correlated with the regulated genes are expressed in myeloid leukemias, suggesting onset of differentiation. For instance, long-term (LT) repopulat- a possible role in myeloid development. These include MLLT3, ing activity of HSCs is contained entirely in the HSC fraction BEX1, ARHGEF12, PTPN11 (alias SHP2), MN1 and S100A6 expressing HMGB3,18 and accordingly, LT-HSCs express higher (alias Calcyclin). Several other significantly downregulated levels of HMGB3.19 Also NDN and MLLT3 are upregulated in genes are expressed in the hematopoietic system, some of them LT-HSCs.19 Elimination of reactive oxygen species by catalase also in the lymphoid lineage. Possibly the expression of these (CAT) was recently shown to be required for HSC self-renewal20 genes must be downregulated to allow lymphoid differentiation, and accordingly, CAT is upregulated in LT-HSCs.19 The down- while they might be upregulated in later stages of lymphoid regulation of other transcripts implicated in oxidoreductase development. This category of genes includes HMGB3, the activity (PRDX2, ALDH6A1, AKR7A2 and FLJ22222) might also transcription factors KLF2, ETV5 and ELF-1, the T-cell differ- correlate with the loss of self-renewal capacity by the CD7 þ entiation protein MAL, and the T-cell specific transcription population. Other downregulated genes expressed in the factor GATA3. GATA3 is clearly downregulated in the CD7 þ hematopoietic system include ICAM2, BST2, TXNIP, CXCR4 population according to our real-time PCR data (see Figure 5), and LEPR.

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 317 Table 1 Biological networks generated by Ingenuity Pathways Analysis

Network Genes in network Score Top functions

1 ADA, BTK, C1QBP, CAT, CEBPB, CXCR4, EVI1, FHL2, GADD45A, H2-D1, HSPA1A, HSPB1, 52 Cellular growth and proliferation, cell IFI16, IFITM3, IL1B, IRF8, LR8, MAL, MIA, PHLPB, PMAIP1, PRKCB1, PTPN11, RGS2, death, hematological system RPS6, RPS6KA2, S100A6, SERPINE2, SMARCA4, TCF4, TNFAIP2, TNFAIP3, TNFSF10, development and function TRA1, TRAF4 2 CASP10, CDK2, CDKN2B, CKS2, CYCS, DDX21, FTH1, GADD45G, GSK3A, GTF2I, HSPB1, 20 Cellular function and maintenance, KLF2, LGALS3, LRRFIP1, MARCKS, MYC, NALP1, PDCD8, PNN, PPIA, PPP1R8, PRDX1, small molecule biochemistry, cell death PRDX2, PSEN2, PTEN, RPL35, RPL36A, RPS27, SF3B1, SF3B3, SHMT2,TF,TFR2, TFRC, TNFRSF8 3 ABCA1, APP, ARG1, ARHGEF12, BZW2, CD59, DOCK1, EVL, F13A1, FLOT1, FPRL1, G0S2, 20 Hematological system development and G6PD, IL13, ITGAE, KIF5A, KIF5B, KIF5C, KNS2, MN1, MUC5AC, NHLH1, PLS3, PTK2, RELA, function, tissue morphology, organismal RUNX3, SATB1, SLC2A5, SPON1, TAL1, TAX1BP3, TGFB1, TNFAIP2, TNS, USF2 survival 4 ADCYAP1, AKAP12, ANK2, BASP1, BCL6, BST2, CABP1, CCL6, CTDSPL, CYBB, FPRL1, 17 Cellular movement, hematological HMGN2, HOXA9, ICAM4, ITGA4, ITGB7, ITPR1, KLF6, MADCAM1, MCM3AP, NCOR2, ORM1, system development and function, RAB31, RALBP1, RB1, RBBP6, RPSA, SCN3A, SERPINB9, SKIIP, SLC1A3, SLC8A1, TNF, TPR, immune response UBR2 5 ARF1, ARF5, ATP1B1, ATR, BCL3, BLM, CHD3, CHEK1, CHEK2, EEF1E1, IFI16, IGHM, ING1, 17 Cell cycle, DNA replication, IQSEC1, KIAA0992, KNTC1, MSN, NBS1, NDN, PHYH, PIP5K1B, PLK1, PSCD4, RAB14, recombination, and repair, cancer RAD17, RDX, RPA1, RPA2, RXRB, SMARCB1, THRAP2, THRB, TP53, VIL2, WRN 6 CKS2, DLK1, E2F4, EGFR, ELF1, EXOSC5, GADD45B, HMGB3, KITLG, KLF6, LEP, LSM5, 17 Cancer, cellular growth and MACF1, MAPK8, NFIB, PLSCR1, PRG1, RIPK4, SERPINE2, SFPQ, SFTPC, SKI, SMAD4, proliferation, gastrointestinal disease SMURF2, SNRP70, TGFB1, TNFRSF17, TOP1, TPM1, TRA1, TRAF5, TRIM33, TXNIP, USF2, WAP 7 ACTN1, ACTR2, ACTR3, ADAM28, ANP32A, ARHGAP5, CDC42, CDH5, CDK5R2, COX7A2L, 15 Cell-to-cell signaling and interaction, CRYGD, DAF, EGF, EIF4A1, ETV5, F2, G6PD, HGFAC, HNRPL, ITSN2, KHDRBS1, KHDRBS3, cellular assembly and organization, cell MEF2A, MKNK1, NACA, NF1, NR4A3, PDCD4, RASGRP3, SET, SETBP1, SIM1, THBD, VEGF, morphology WAS 8 AGT, AMBP, C3, CCR9, CD209, CFH, CLDN1, EDNRA, ERG, GRK5, HLX1, HNRPA1, HSD11B1, 15 Molecular transport, cellular movement, ICAM2, IFIT1, IGL@, IL4, IL15, ITIH1, ITIH2, ITIH3, ITIH4, ITIH5, JUN, LSP1, MAFF, MST1, MYH7, cell-to-cell signaling and interaction NFE2, NPR3, OXT, RNPC2, TBX21, TJP2, TNS Genes in bold are differentially expressed in the microarray analysis. Only networks with a significance score of 3 or higher (Pp0.001) are shown.

Genes involved in proliferation and apoptosis also found amongst the significantly downregulated genes: The list of upregulated genes contains many genes known to act C20orf67, LSM5, KHDRBS3 (alias T-STAR), BZW2 and HSPA1A as negative regulators of cell proliferation, including LPIN1 and CTDSPL (alias SCP3). The differential expression of factors (Lipin1), MCM3AP, ATR, POLS, SMARCA4, MACF1, KNTC1, of the transcription/translation machinery might be an important RBBP6, TRIB2 and NF1. Many downregulated genes, on the mechanism for regulating lymphoid-specific gene/protein ex- other hand, are positive regulators of cell proliferation. These pression in the CD7 þ CLP cells. There is evidence that include LEPR, IFI16, S100A6 (alias Calcyclin), CKS2, LAPTM4B, alternative splicing of mRNA, which permits the generation of HSPA1A and several ribosomal proteins such as RPL35, RPS27, several proteins from one gene by alternative exon usage, can be RPS27L, RPS6 and RPL36A. A decrease in the expression of modulated in a cell type- or developmental stage-specific mRNAs that encode ribosomal proteins accompanies shut-off of way.22 Tissue-specific splicing can be the result of concentration cell division.21 However, some downregulated genes are differences of ubiquitously expressed splicing factors, but tissue negative regulators of cell proliferation: NDN, KHDRBS3 (alias type- or developmental stage-specific splicing factors have also T-STAR), TXNIP and GADD45A. In addition, several genes been described.23 involved in apoptosis were differentially expressed: NALP1, RERE, TRAF4 and TRAF5 were upregulated, while PMAIP1 was downregulated. Overall, this expression pattern of positive and negative regulators of proliferation tends to point to an intrinsic lower proliferative capacity of the CD7 þ population compared Genes involved in regulation of the cytoskeleton to the CD7À population. Many differentially expressed genes are involved in the regulation of the cytoskeleton. The list of upregulated genes contains MACF1, Palladin, TBCD and DOCK1. More cytoske- Genes involved in transcription/translation regulation leton-related genes are significantly downregulated: PLS3, In addition to transcription factors, also many other proteins KIAA1102, TNS1, ACTN1, PIP5K1B, ARHGEF12, MARCKS, involved in gene expression were differentially expressed. The ACTR2, HSPB1 and GABARAPL1. Although little is known list of significantly upregulated genes contains for instance many about the cytoskeleton of hematopoietic stem and progenitor genes involved in splicing, such as HNRPA1, CCNL1, SFPQ, cells,24 it might be involved in the regulation of differentiation in SF3B1, RNPC2, PNN and METTL3. EIF4A1 is a translation several ways. The cytoskeleton is a scaffold for various signal initiation factor, and NACA is required for the intracellular transduction pathways.25 In addition, alterations in its structure translocation of newly synthesized polypeptides. TPR is can lead to the relocation and clustering of certain cytoskeleton- implicated in the import of proteins into the nucleus, SLC38A1 linked surface molecules. Moreover, the cytoskeletal organiza- is an amino-acid transporter and the RNA helicase DDX21 is tion can regulate gene expression by controlling the nuclear involved in ribosomal RNA synthesis. These types of genes are import of certain transcription factors.26

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 318 Genes involved in protein degradation CD34 þ CD45RAhiCD7 þ cord blood population, but these cells A number of genes coding for proteins involved in the ubiquitin- were not selected for CD38 negativity. These CD34 þ proteasome system of proteolysis, were significantly upregu- CD45RAhiCD7 þ cells display strong NK- and T-cell potential, lated: TNFAIP3 (alias A20), UBR2, TRIM33 and USP34. Other but also substantial myelo/erythroid potential.10 This is not proteolytic enzymes that were upregulated are ADAM28 and surprising because most CD34 þ CD45RAhiCD7 þ cells are DPEP2. Proteins degraded by the ubiquitin-proteasome pathway CD38 þ and, as mentioned in the paper by Hao et al.9 and as include cyclins and other regulators of the cell cycle, and we noticed ourselves (data not shown), even low expression of transcription factors.27 Protein degradation is thus a mechanism CD38 on CD34 þ CD7 þ cells is sufficient to confer these cells to regulate the concentration of many regulatory proteins. myeloid differentiation potential. CD38 is a marker expressed on more mature progenitors of all lineages. Therefore, the CD34 þ CD45RAhiCD7 þ population is presumably heteroge- Discussion neous and contains both lymphoid- and myeloid-committed progenitors, which can also express low levels of the CD7 The CD7 þ subpopulation of CD34 þ CD38À human cord blood antigen. Accordingly, the CD34 þ CD45RAhiCD7 þ population cells was identified by Hao et al.9 as a primitive CLP population expresses increased levels of the T-cell receptor g chain (cDNAs with the ability to generate B, NK and dendritic cells, but with TRG, TRGC2 and TRGV9), terminal deoxynucleotidyl transfer- no potential for myeloid or erythroid differentiation. In the ase (DNTT), and strikingly many genes that are specifically present study, we confirmed the B- and NK-differentiation expressed in myeloid cells: calgranulin A (S100A8), the potential of this cell population in a co-culture assay using MS-5 macrophage colony stimulating factor receptor CD115 (CSF1R), stromal cells. In addition, we showed that this cell population the myeloid-specific transcription factor CEBP/D (CEBPD), and has strong T-cell differentiation potential in hybrid human- the neutrophil granule proteins lysozyme (LYZ), myeloperoxi- mouse FTOC. Therefore, CD34 þ CD38ÀCD7 þ cord blood cells dase (MPO), elastase 2 (ELA2), azurocidin (AZU1), cathepsin G have full lymphoid differentiation potential and are true CLPs. (CTSG) and proteinase 3 (PRTN3). None of these genes were Gene expression profiling of CD34 þ CD38ÀCD7 þ cells and expressed by our CD34 þ CD38ÀCD7 þ population. their CD7À counterparts using Affymetrix oligonucleotide Regarding T-cell differentiation, it is still an open question microarrays revealed the differential expression of many whether multipotent or lymphoid-committed progenitor cells transcription factors, cell cycle genes, signal transduction home to the thymus. In addition to their T-cell differentiation molecules and proteins involved in gene expression and potential, fetal CD34hiCD1aÀ thymocytes possess strong NK- cytoskeleton regulation. Many upregulated genes are lym- cell potential, as well as B- and dendritic cell potential, but lack phoid-affiliated, whereas many downregulated genes are related granulocyte/macrophage potential.32 Weerkamp et al. recently to the myeloid lineage. The majority (146 of 190) of the showed that the adult CD34 þ CD1aÀ thymocyte subset has significantly differentially expressed genes are also differentially differentiation potential not only for B, NK and dendritic cells, expressed between one or more consecutive differentiation but also for myeloid and erythroid lineages,33 leading to the stages of early human T- and B-cell development12,15 (Supple- hypothesis that the adult human thymus is seeded by a mentary Table S5). Of those, 80 are differentially expressed multipotent stem cell-like progenitor instead of a CLP. Still, during T- and B-cell development, 38 only during T-cell the multilineage differentiation capacity of both fetal and adult development and 28 only during B-cell development. For many early thymocytes remains to be shown at the single cell level, of those genes the differential expression (up- or downregula- thus the possibility exists that, analogously to the corresponding tion) between CD7À and CD7 þ cells is the same as between murine DN1 thymocyte stage (CD44 þ CD25À),34 the human cord blood CD34 þ LinÀ and the most immature T- or B-cell CD34 þ CD1aÀ thymic subset is heterogeneous and contains stage (CD34 þ CD38ÀCD1À and pro-B, respectively). The pre- multiple progenitors with different differentiation potential. ferential expression of lymphoid-affiliated genes in Recently, two possible models for seeding of the thymus have CD34 þ CD38ÀCD7 þ cells is in agreement with gene expression been discussed:35 either a multipotent HSC-like cell seeds the studies on murine hematopoietic stem and progenitor cells, thymus or multiple precursor cells with differential lineage which showed that CLPs express markers of B, T and NK cells differentiation capacities enter the thymus. Therefore, it might but no myeloid markers, and conversely, CMPs express be possible that T-committed progenitors in the human thymus granulocytic/monocytic and megakaryocytic/erythroid markers derive from CD34 þ CD38ÀCD7 þ thymic immigrants, while the but no lymphoid markers.28,29 The low-level ‘promiscuous’ other thymic cell progenitors derive from other cells in the expression of lineage-specific genes before commitment to a CD34hiCD45RAhiCD7 þ cord blood population. particular lineage, is referred to as ‘lineage priming’.30 Upon In conclusion, molecular characterization of human cord commitment to a particular lineage, cells upregulate the blood CD34 þ CD38ÀCD7 þ cells showed preferential expres- appropriate lineage-specific genes, and suppress the inappropri- sion of lymphoid-affiliated genes, which is compatible with the ate genes of the alternative lineages. Surprisingly, murine HSCs restricted potential of these cells to differentiate into T (this express many myeloid-affiliated genes but almost no lymphoid- study), B, NK and dendritic cells (Hao et al.9 and this study). In affiliated genes, while their immediate progeny, multipotent addition, gene expression analysis revealed several interesting progenitors (MPP), exhibit both myeloid and lymphoid promis- genes for further study. Overexpression and RNA interference of cuity.28,29 Moreover, murine HSCs express a lot of genes selected genes will learn whether expression of these genes is affiliated to non-hematopoietic tissues,29 which may explain critical for the developmental transition of multipotent stem their reported capacity to differentiate into non-hematopoietic cells to the CLP stage. cell types.31 Consistent with these data, our list of significantly downregulated genes also contained a number of non-hemato- poietic genes. Acknowledgements Our study is the first to describe a global gene expression profile of a human CLP. In their recent study,10 Haddad et al. We thank the staff of the Bloedtransfusiecentrum Oost-Vlaanderen determined the gene expression in the related for the supply of umbilical cord blood samples and Marie-Jose´ De

Leukemia Molecular characterization of human cord blood CLP I Hoebeke et al 319 Bosscher for lymphoprepping. We are grateful to Dick De Ridder 17 De Smedt M, Reynvoet K, Kerre T, Taghon T, Verhasselt B, for help with microarray analysis and to Inge Van de Walle for Vandekerckhove B et al. Active form of Notch imposes T cell fate excellent technical assistance. This work was supported by grants in human progenitor cells. J Immunol 2002; 169: 3021–3029. 18 Nemeth MJ, Curtis DJ, Kirby MR, Garrett-Beal LJ, Seidel NE, from the Fund for Scientific Research Flanders (Belgium) and from Cline AP et al. Hmgb3: an HMG-box family member expressed in the Ghent University concerted research program. primitive hematopoietic cells that inhibits myeloid and B-cell differentiation. Blood 2003; 102: 1298–1306. 19 Forsberg EC, Prohaska SS, Katzman S, Heffner GC, Stuart JM, References Weissman IL. Differential expression of novel potential regulators in hematopoietic stem cells. PLos Genet 2005; 1: e28. 1 Kondo M, Weissman IL, Akashi K. Identification of clonogenic 20 Ito K, Hirao A, Arai F, Matsuoka S, Takubo K, Hamaguchi I common lymphoid progenitors in mouse bone marrow. Cell 1997; et al. Regulation of oxidative stress by ATM is required for 91: 661–672. self-renewal of haematopoietic stem cells. Nature 2004; 431: 2 Akashi K, Traver D, Miyamoto T, Weissman IL. A clonogenic 997–1002. common myeloid progenitor that gives rise to all myeloid lineages. 21 Bortoluzzi S, d’Alessi F, Romualdi C, Danieli GA. Differential Nature 2000; 404: 193–197. expression of genes coding for ribosomal proteins in different 3 Allman D, Sambandam A, Kim S, Miller JP, Pagan A, Well D et al. human tissues. Bioinformatics 2001; 17: 1152–1157. Thymopoiesis independent of common lymphoid progenitors. Nat 22 Baklouti F, Huang SC, Tang TK, Delaunay J, Marchesi VT, Benz EJJ. Immunol 2003; 4: 168–174. Asynchronous regulation of splicing events within protein 4.1 4 Schwarz BA, Bhandoola A. Circulating hematopoietic progenitors pre-mRNA during erythroid differentiation. Blood 1996; 87: with T lineage potential. Nat Immunol 2004; 5: 953–960. 3934–3941. 5 Katsura Y. Redefinition of lymphoid progenitors. Nat Rev Immunol 23 Stoss O, Olbrich M, Hartmann AM, Konig H, Memmott J, 2002; 2: 127–132. Andreadis A et al. The STAR/GSG family protein rSLM-2 regulates 6 Galy A, Travis M, Cen D, Chen B. Human T, B, natural killer, and the selection of alternative splice sites. J Biol Chem 2001; 276: dendritic cells arise from a common bone marrow progenitor cell 8665–8673. subset. Immunity 1995; 3: 459–473. 24 Levesque J-P, Simmons PJ. Cytoskeleton and integrin-mediated 7 Ishii T, Nishihara M, Ma F, Ebihara Y, Tsuji K, Asano S et al. adhesion signaling in human CD34+ hemopoietic progenitor cells. Expression of stromal cell-derived factor-1/pre-B cell growth- Exp Hematol 1999; 27: 579–586. stimulating factor receptor, CXC chemokine receptor 4, on 25 Juliano RL. Signal transduction by cell adhesion receptors and the CD34+ human bone marrow cells is a phenotypic alteration cytoskeleton: functions of integrins, cadherins, selectins, and for committed lymphoid progenitors. J Immunol 1999; 163: immunoglobulin-superfamily members. Annu Rev Pharmacol 3612–3620. Toxicol 2002; 42: 283–323. 8 Manz MG, Miyamoto T, Akashi K, Weissman IL. Prospective 26 Janmey PA. The cytoskeleton and cell signaling: component isolation of human clonogenic common myeloid progenitors. Proc localization and mechanical coupling. Physiol Rev 1998; 78: Natl Acad Sci USA 2002; 99: 11872–11877. 763–781. 9 Hao Q-L, Zhu J, Price MA, Payne KJ, Barsky LW, Crooks GM. 27 Doherty FJ, Dawson S, Mayer RJ. The ubiquitin-proteasome Identification of a novel, human multilymphoid progenitor in cord pathway of intracellular proteolysis. Essays Biochem 2002; 38: blood. Blood 2001; 97: 3683–3690. 51–63. 10 Haddad R, Guardiola P, Izac B, Thibault C, Radich J, Delezoide A-L 28 Akashi K, He X, Chen J, Iwasaki H, Niu C, Steenhard B et al. et al. Molecular characterization of early human T/NK and Transcriptional accessibility for genes of multiple tissues and B-lymphoid progenitor cells in umbilical cord blood. Blood 2004; hematopoietic lineages is hierarchically controlled during early 104: 3918–3926. hematopoiesis. Blood 2003; 101: 383–389. 11 Robin C, Pflumio F, Vainchenker W, Coulombel L. Identification of 29 Miyamoto T, Iwasaki H, Reizis B, Ye M, Graf T, Weissman IL lymphomyeloid primitive progenitor cells in fresh human cord et al. Myeloid or lymphoid promiscuity as a critical step blood and in the marrow of nonobese diabetic-severe combined in hematopoietic lineage commitment. Dev Cell 2002; 3: immunodeficient (NOD-SCID) mice transplanted with human 137–147. CD34+ cord blood cells. J Exp Med 1999; 189: 1601–1610. 30 Hu M, Krause D, Greaves M, Sharkis S, Dexter M, Heyworth C 12 Dik WA, Pike-Overzet K, Weerkamp F, de Ridder D, de Haas EFE, et al. Multilineage gene expression precedes commitment in the Baert MRM et al. New insights on human T cell development by hemopoietic system. Genes Dev 1997; 11: 774–785. quantitative T cell receptor gene rearrangement studies and gene 31 Krause DS, Theise ND, Collector MI, Henegariu O, Hwang S, expression profiling. J Exp Med 2005; 201: 1715–1723. Gardner R et al. Multi-organ, multi-lineage engraftment by a single 13 Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, bone marrow-derived stem cell. Cell 2001; 105: 369–377. Scherf U et al. Exploration, normalization, and summaries of high 32 Haddad R, Guimiot F, Six E, Jourquin F, Setterblad N, Kahn E et al. density oligonucleotide array probe level data. Biostatistics 2003; Dynamics of thymus-colonizing cells during human development. 4: 249–264. Immunity 2006; 24: 217–230. 14 Khatri P, Draghici S, Ostermeier GC, Krawetz SA. Profiling gene 33 Weerkamp F, Baert MR, Brugman MH, Dik WA, de Haas EF, Visser expression using Onto-Express. Genomics 2002; 79: 266–270. TP et al. The human thymus contains multipotent progenitors with 15 van Zelm MC, van der Burg M, de Ridder D, Barendregt BH, de T/B-lymphoid, myeloid and erythroid lineage potential. Blood Haas EFE, Reinders MJT et al. Ig gene rearrangement steps are 2005; 107: 3131–3137. initiated in early human precursor B cell subsets and correlate with 34 Porritt HE, Rumfelt LL, Tabrizifard S, Schmitt TM, Zu´n˜iga-Pflu¨cker specific transcription factor expression. J Immunol 2005; 175: JC, Petrie HT. Heterogeneity among DN1 prothymocytes reveals 5912–5922. multiple progenitors with different capacities to generate T cell and 16 Livak KJ, Schmittgen TD. Analysis of relative gene expression data non-T cell lineages. Immunity 2004; 20: 735–745. using real-time quantitative PCR and the 2ÀDDCT method. Methods 35 Weerkamp F, Pike-Overzet K, Staal FJ. T-sing progenitors to 2001; 25: 402–408. commit. Trends Immunol 2006; 27: 125–131.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia