Leukemia (2008) 22, 1214–1225 & 2008 Nature Publishing Group All rights reserved 0887-6924/08 $30.00 www.nature.com/leu ORIGINAL ARTICLE

Dose-dependent repression of T-cell and natural killer cell by PU.1 enforces myeloid and B-cell identity

MB Kamath1, IB Houston1, AJ Janovski1, X Zhu1, S Gowrisankar2,3, AG Jegga2,4 and RP DeKoter1

1Department of Molecular Genetics, Biochemistry, and Microbiology, University of Cincinnati College of Medicine, Cincinnati, OH, USA; 2Division of Biomedical Informatics, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, USA; 3Department of Biomedical Engineering, University of Cincinnati, Cincinnati, OH, USA and 4Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH, USA

The Ets transcription factor PU.1, encoded by the Sfpi1, B-cell development.16 PU.1 was also shown to regulate functions in a concentration-dependent manner to promote macrophage/neutrophil cell fate decisions.17 Cells expressing myeloid and B-cell development and has been implicated in myeloid and lymphoid leukemias. To determine the conse- low levels of PU.1 had a mixed expression pattern, whereas quences of reducing PU.1 concentration during hematopoiesis, cells expressing high levels favored macrophage development 18 we analyzed mice with two distinct hypomorphic alleles of Sfpi1 by activating and repressing distinct subsets of genes. These that produce PU.1 at B20% (BN) or B2% (Blac) of wild-type studies demonstrate that PU.1 functions as a concentration- levels. Myeloid development was impaired in these mice, but dependent cell fate determinant. less severely than in Sfpi1 null mice. To identify the down- PU.1 is expressed at uniform levels in hematopoietic stem stream target genes that respond to changes in PU.1 concen- 19,20 tration, we analyzed ex vivo -3 dependent myeloid cells and common myeloid and lymphoid progenitors. PU.1 cell lines established from Sfpi1BN/BN, Sfpi1Blac/Blac and Sfpi1À/À increases during myeloid terminal differentiation and decreases 19,20 fetal liver cells. Unexpectedly, many T-cell and natural killer cell during B-cell terminal differentiation. Levels also decrease genes were expressed in Sfpi1À/À cells and repressed in a dose- after erythroid, T-cell or NK-cell commitment, and forced Blac/Blac BN/BN dependent manner in Sfpi1 and Sfpi1 cells. This expression of PU.1 results in a block to erythroid and T-cell pattern of dose-dependent T/NK-cell gene repression also development.19–22 PU.1 is critical in myeloid/erythroid fate occurred in ex vivo interleukin-7 dependent progenitor B cell 23–25 lines. These results suggest that PU.1 functions in a concen- decisions and interacts with the transcription factor GATA-1. tration-dependent manner to repress T-cell and natural killer PU.1 binds to GATA-1 on erythroid target genes, recruits cell fates while promoting myeloid and B-cell fates. repressive chromatin modification factors and promotes myelo- Leukemia (2008) 22, 1214–1225; doi:10.1038/leu.2008.67; poiesis at the expense of erythropoiesis.24–26 In contrast, it is published online 20 March 2008 unknown how T/NK-cell genes respond to changes in PU.1 Keywords: PU.1; repression; T-cell; transcription factor concentration or why PU.1 must be downregulated for terminal differentiation. Our laboratory recently generated a hypomorphic allele of Sfpi1 termed BN15 that results in failure of B-cell development, abnormal T-cell development, and hyperproliferation of Introduction immature myeloid cells. Analysis of Sfpi1BN/BN cultured cells revealed that they express B20% of wild-type PU.1 Transcription factors regulate hematopoiesis by activating or levels. This allele is unique because it results in a reduction repressing lineage-specific genes involved in cell fate decisions of PU.1 levels in all cell types.15 This is in contrast to a and lineage commitment. PU.1, encoded by the gene Sfpi1 in hypomorphic allele created by deletion of the upstream mice and Spi-1 in humans, belongs to the Ets transcription factor regulatory enhancer at À14 kb, which knocks down PU.1 in family.1 Over 100 target genes of PU.1 have been identified in most hematopoietic cell types but results in increased PU.1 myeloid and B cells. Knockout of Sfpi1 results in fetal or levels in T cells.8,12 perinatal lethality, absence of myeloid and B cells and abnormal In this study, we generated a second distinct hypomorphic T-cell and natural killer (NK)-cell development.2–5 Mutations allele of Sfpi1, termed Blac, to determine the concentration- that inactivate PU.1 are associated with acute myeloid leukemia dependent effects of PU.1 on target genes and lineage decisions. (AML) in humans and are sufficient to cause AML in mouse Analysis of Sfpi1Blac/Blac fetal liver cells cultured in interleukin-3 6–11 models. Altered PU.1 levels also cause T-cell leukemias in (IL-3) suggests that they express B2% of wild-type PU.1 protein 11,12 mice. levels. These cells fail to terminally differentiate as a conse- Early on, PU.1 was recognized to be expressed at significantly quence of low PU.1 expression and can be maintained as cell 13 higher concentrations in macrophages than B cells. High PU.1 lines. To determine gene regulation in response to varied PU.1 levels are required to generate macrophages both in vitro and in concentrations, we compared gene expression in Sfpi1BN/BN and 14–16 vivo. Low concentrations are required for terminal differ- Sfpi1Blac/Blac cells to Sfpi1-/- cells. With this unique allelic entiation of B cells, and forced expression results in a block to system, we can study the effects of three discrete concentrations of PU.1 at B20, B2 and 0% of wild-type levels. Our results Correspondence: Dr RP DeKoter, Department of Molecular Genetics, show that PU.1 both activates and represses distinct groups of Biochemistry, and Microbiology, University of Cincinnati College of genes. Genes activated in a gradient fashion included a cluster Medicine, 231 Albert Sabin Way, Cincinnati, OH 45267-0524, USA. E-mail: [email protected] of myeloid-specific direct PU.1 target genes, while genes Received 6 November 2007; revised 19 February 2008; accepted 21 repressed in a gradient fashion included clusters of erythroid- February 2008; published online 20 March 2008 specific and, unexpectedly, T-cell- and NK-cell-specific genes. PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1215 T/NK-cell genes were also repressed in a dose-dependent Ovation Biotin System protocol (NuGEN, San Carlos, CA, USA), manner in IL-7 dependent pro-B cells. In conclusion, our results hybridized to Affymetrix Mouse Genome Microarray 430 2.0 suggest that PU.1 functions in a concentration-dependent and analyzed as described previously.30 Data (Supplementary manner to promote myeloid or B-cell differentiation and Table S2) was analyzed with GeneSpring software (Agilent concurrently repress T-cell and NK-cell development. Technologies, Santa Clara, CA, USA) and normalized per chip to the 50th percentile and per gene to the median, with measurements less than 0.01 set to 0.01. Additionally, all Materials and methods samples were normalized relative to expression in Sfpi1À/À cells. We used the program Database for Annotation, Visualiza- Gene targeting of mice and cell culture tion and Integrated Discovery (DAVID) for gene functional BN/BN 15 Sfpi1 mice were generated as previously described. classification.31 Sfpi1Blac/Blac mice were generated by Cre-mediated excision of the neomycin gene by mating Sfpi1 þ /BN mice to EIIA-Cre mice (Jackson Laboratories). F1 mice lacking PGK-NEO were crossed Bioinformatic analysis to Black/Swiss mice (Jackson Laboratories) to remove the Cre We assembled a database of known direct PU.1 target genes transgene. Offspring lacking both PGK-NEO and the Cre (Supplementary Table S3). A total of 99 biochemically þ /Blac transgene were termed Sfpi1 . Excision was verified by characterized PU.1 binding site sequences from 93 validated PCR and by DNA-sequencing of PCR-amplified genomic DNA. PU.1 target genes were identified from the primary literature. Genotyping was performed with primers outlined in Houston These included the known core PU.1 binding site GGAA or 15 et al. Primers used to verify loss of Cre transgene are as AGAA and five flanking nucleotides on either side.32 Based on 0 0 follows: Cre1 (5 -CTAGGCCACAGAATTGAAAGATCT-3 ), Cre2 these sequences, a position weight matrix (PWM) was generated 0 0 0 (5 -GTAGGTGGAAATTCTAGCATCATCC-3 ), Cre3 (5 -GCGGT to identify consensus PU.1 sites and putative targets. A standard 0 0 CTGGCAGTAAAAACTATC-3 ), and Cre4 (5 -GTGAAACAGCA PU.1 matrix was also obtained from the TRANSFAC database.33 0 TTGCTGTCACTT-3 ). Timed matings were performed as de- A list of housekeeping genes was assembled as a negative 15 scribed. Colony-forming unit (CFU) assays were performed control for PU.1 regulation by finding genes with the ontology 27 using day 14.5 fetal liver cells as previously described. IL-3 ‘cell growth and/or maintenance’ that had similar expression dependent cell lines were generated by culturing day 14.5 fetal values (o3% difference) between all samples. To identify liver cells in complete Iscove’s modified Dulbecco medium putative PU.1 binding sites in promoters of activated, repressed À1 À1 (IMDM) with IL-3 (5 ng ml ), IL-6 (10 ng ml ) and SCF or housekeeping genes, we searched the conserved upstream À1 (100 ng ml ; Peprotech, Rocky Hill, NJ, USA) for 4 days, 1 kb regions identified based on the transcriptional start sites followed by passage every 4 days in complete IMDM containing annotated in the NCBI RefSeq database. The 17-species multiple À1 5ngml IL-3. Generation of pro-B cell lines has been alignment files for each of these upstream 1 kb sequences was 27 described. For switching experiments, cells were downloaded from the UCSC Golden Path database (http:// washed twice in complete IMDM and replated in triplicate in genome.ucsc.edu).34 Tffind, a computer program written in C, À1 À1 complete IMDM with IL-3 (5 ng ml ) or GM-CSF (0.5 ng ml ; was used to locate patterns of PU.1 sites within the multiple Peprotech). alignment files for each gene promoter. Tffind identifies matches to PWMs in conserved regions within any number of sequences in an alignment, searching sequentially through both the Flow cytometry and immunoblot forward and reverse complement strands.35 Putative PU.1 Flow cytometric analysis was performed on cells stained with binding sites were scored, based on their positions relative to the biotin-conjugated antibodies 15.3.2 (IgE), RB6-8C5 (Gr-1), the transcriptional start sites and their percent identity to the 2.4G2 (FcgRII/III), 1D3 (CD19), RA3-6B2 (B220) or phycoery- core PU.1 site from the PWM. An overall cut-off similarity of thrin-conjugated antibody 2B8 (c-Kit). Biotin-conjugated 70% was applied, with a 100% cut-off similarity required for the antibodies were visualized by secondary staining with strepta- core PU.1 site. For the current study, sites conserved across vidin–allophycocyanin conjugate (BD Pharmingen, San Diego, mouse and human were included. A nonparametric Mann– CA, USA). Immunoblotting was performed with rabbit anti-PU.1 Whitney test was used to analyze the statistical significance of polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA, differential representation of PU.1 sites between the groups of USA) and HRP-conjugated anti-rabbit secondary antibody genes. (Pierce, Rockford, IL, USA), as previously described.28

RT–PCR analysis Results Reverse transcription (RT)–PCR and real-time RT–PCR with a Cepheid SmartCycler were performed as previously de- Generation of mice with a second distinct hypomorphic scribed.28,29 Real-time RT–PCR quantitation was based on allele of Sfpi1 normalization to G6pdh. Primers are listed in Supplementary We previously reported the generation of mice with a Table S1. hypomorphic allele of the PU.1-encoding gene Sfpi1 termed BN, which expresses B20% of wild-type protein levels.15 To determine the consequences of removing the PGK promoter and Affymetrix GeneChips neomycin resistance gene, we performed Cre-mediated excision Two separate IL-3 dependent cell lines (n ¼ 2) were established of loxP elements (described in Materials and methods), thereby for each homozygous genotype. RNA was isolated from at least generating another allele of Sfpi1 termed Blac (Figure 1a). 7 6 Â 10 cells in each IL-3 dependent cell line (once for line n1, Genotypes were confirmed by PCR and Southern blotting Blac/Blac line n2 in duplicate, total triplicate analysis) with RNA-Bee (Figure 1b; data not shown). Sfpi1 mice were born at (Tel-Test Inc., Friendswood, TX, USA) and purified with RNeasy near Mendelian frequency (18%, Table 1). However, all died (Qiagen, Valencia, CA, USA). The RNA was labeled with the within 7 days of birth, as opposed to Sfpi1BN/BN mice, which

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1216

Figure 1 Generation of Sfpi1Blac/Blac mice. (a) Targeting strategy in which Cre-mediated excision was used to remove the neomycin resistance cassette in the BN allele to generate the Blac allele. LoxP sites are indicated with open triangles. (b) Genotyping PCR results with primers P1–P4, indicated as arrows in (a). (c) Survival curve for wild-type, Sfpi1BN/BN, Sfpi1Blac/Blac and Sfpi1À/À mice. (d) Colony-forming assays with fetal liver cells from wild-type, Sfpi1BN/BN, Sfpi1Blac/Blac and Sfpi1À/À mice in the indicated .

Table 1 Frequency of live births from matings of Sfpi1+/Blac mice from cultured cells and performed extensive RT–PCR analysis and sequencing of Sfpi transcripts. We found that the Sfpi1 Age Total Sfpi1+/+ Sfpi1+/Blac Sfpi1Blac/Blac transcript encoded by the Blac allele was spliced identically to the transcript encoded by the BN allele and also produced PU.1 E 14.5 40 8 (20%) 23 (58%) 9 (22%) protein lacking the first 31 amino acids (data not shown).15 To Day 1 76 16 (21%) 46 (61%) 14 (18%) determine the relative levels of Sfpi1 transcripts, we performed Day 8 72 22 (31%) 50 (69%) 0 Day 21 117 33 (28%) 84 (72%) 0 real-time RT–PCR using primers specific to exon 5, which is partially deleted in the Sfpi1À/À mouse.3 We validated these Genomic DNA was prepared from E14.5 fetuses, newborns at day 1, primers by analyzing RNA from fetal liver cells cultured in IL-3/ mice at day 8 after birth or at the time of weaning at day 21. Genotypes IL-6/SCF for 5 days and verified that PU.1 transcripts were were identified by PCR with primers as shown in Figure 1b. expressed in Sfpi1BN/BN cells at B20% of wild-type levels (Figure 2a left).15 In cells cultured in IL-3 for several weeks, we survived until weaning, and Sfpi1À/À mice, which died before found that PU.1 transcripts were expressed 7.8-fold lower in Blac/Blac BN/BN birth (Figure 1c). Colony-forming assays were performed to Sfpi1 cells than Sfpi1 cells, suggesting that the Blac B measure the frequency of fetal liver progenitor cells responsive allele expresses 2% of wild-type levels (Figure 2a right). PU.1 to granulocyte-colony stimulating factor (G-CSF), macrophage- protein levels from the BN allele were previously shown as 20% 15 Blac/Blac CSF (M-CSF), GM-CSF or a combination of interleukin-3 (IL-3), of wild type. Immunoblotting with Sfpi1 lysates could IL-6 and stem cell factor (SCF) (Figure 1d). Sfpi1Blac/Blac cells not reproducibly detect PU.1 protein at such low levels generated fewer colonies than Sfpi1BN/BN cells, which we have (Figure 2b); however, multiple lines of evidence suggest that previously shown to generate fewer colonies than wild type.15 these cells produce low levels of PU.1 activity (Figures 1c, d and Sfpi1À/À fetal liver cells generated colonies only in IL-3/IL-6/SCF 2e below). (data not shown).14 These results show that the Sfpi1Blac/Blac After several passages, wild-type cells grew more slowly than À/À BN/BN Blac/Blac phenotype is intermediate in severity between Sfpi1BN/BN and Sfpi1 , Sfpi1 or Sfpi1 cells (Figure 2c). Wild-type Sfpi1À/À, suggesting that PU.1 levels are between 0 and 20% of cell lines were found to be immature mast cells as measured wild type. by morphological analysis, high levels of c-Kit and the ability to bind IgE with high affinity (Figure 2d; data not shown). Previous studies have demonstrated that PU.1 is required for the PU.1 is expressed in a gradient fashion in ex vivo development of mast cells from fetal liver cells under these cultured cell lines culture conditions.36 Consistently, Sfpi1À/À cells morphologi- Wild-type, Sfpi1À/À, Sfpi1BN/BN and Sfpi1Blac/Blac fetal liver cells cally resembled immature blasts, expressed intermediate levels all proliferated in response to IL-3/IL-6/SCF (Figure 1d; data not of c-Kit, high levels of Gr-1 and FcgRII/III and did not bind IgE shown). Cells of all four genotypes expanded rapidly when (Figure 2d; data not shown). Unexpectedly, Sfpi1BN/BN and switched to liquid cultures containing IL-3. We prepared RNA Sfpi1Blac/Blac cells also resembled immature blasts and did not

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1217

Figure 2 IL-3 dependent cell lines. (a) Real-time RT–PCR with Sfpi1 exon 5 primers in fetal liver cells grown in IL-3 for 5 days to validate levels in Sfpi1BN/BN relative to wild type (left) and to establish transcript levels from IL-3 dependent cell lines of Sfpi1Blac/Blac relative to Sfpi1BN/BN (right). *** indicates Po0.01. (b) Western blot for PU.1 protein expression from cellular lysates of the indicated genotype. b-actin was used as a loading control. (c) Growth of wild-type, Sfpi1À/À, Sfpi1BN/BN and Sfpi1Blac/Blac fetal liver cells in IL-3. Cells were counted every 24 h from 0 to 72 h. (d) Flow cytometry of single cell suspensions from wild-type, Sfpi1À/À, Sfpi1BN/BN and Sfpi1Blac/Blac IL-3 dependent cell lines. Cells were gated for size and granularity and analyzed with antibodies to the indicated cell surface markers. Numbers indicate percentage of gated cells in each quadrant. (e) Ethidium bromide gel analysis of RT–PCR for mitf common, mitf-mc, FcgRIIb and FcgRIII in IL-3 dependent cell lines of the indicated genotype. b-actin was used as a loading control. bind IgE (Figure 2d; data not shown). Sfpi1BN/BN cells expressed detected by flow cytometric analysis was FcgRIII intermediate levels of c-Kit and Gr-1 and high levels of FcgRII/ (Figure 2d). III, while Sfpi1Blac/Blac cells had similar c-Kit, Gr-1 and FcgRII/III In conclusion, while wild-type cells differentiate into mast expression as Sfpi1À/À cells. This suggests that hypomorphic cells in IL-3, Sfpi1BN/BN (BN), Sfpi1Blac/Blac (Blac) and Sfpi1À/À Sfpi1BN/BN and Sfpi1Blac/Blac cells express levels of PU.1 that are (KO) ex vivo cell lines can be utilized as a unique Sfpi1 insufficient to promote mast cell differentiation.36 hypomorphic allelic series in which PU.1 is expressed at B20, Next, we performed RT–PCR analysis for several myeloid B2 or 0% of wild-type levels, respectively. This system can be genes (Figure 2e). Microphthalmia-associated transcription applied to study the effects of two discrete PU.1 levels on factor (Mitf) is critical for mast cell development and has several downstream target gene expression. identified isoforms, one of which is expressed exclusively in mast cells, Mitf-mc.37 Mitf common was readily detectable in cells of all four genotypes, while mitf-mc was expressed PU.1 activates and represses groups of target genes at exclusively in wild-type immature mast cells (Figure 2e). FcgRIIb distinct concentrations and FcgRIII are low-affinity receptors for IgG and are expressed To discover how genes respond to distinct levels of PU.1, we in both myeloid and lymphoid cell types (reviewed by performed whole genome microarray analysis to compare gene Nimmerjahn and Ravetch38). Our laboratory has previously expression in BN and Blac cell lines with gene expression in the validated FcgRIIb as a PU.1 direct target.15,28 Interestingly, KO cell line (Supplementary Table S2). The data was normal- FcgRIIb was expressed in a dose-dependent fashion, decreasing ized based on Affymetrix internal controls and averaged from from wild type to Sfpi1BN/BN to Sfpi1Blac/Blac and absent in three samples of each cell line (described in Materials and Sfpi1À/À cells, whereas FcgRIII was expressed in cells of all four methods). As expected, Sfpi1 was expressed in a gradient genotypes (Figure 2e). This suggests that the cell surface Fcg fashion: 2.25-fold in Blac cells and 7.69-fold in BN cells,

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1218 Table 2 PU.1 regulates genes in four modes of activation or repression

Affymetrix probe set ID Gene symbol Description Blac BN

1418747_at Sfpi1 SFFV proviral integration 1 2.35 7.69

High concentrations: activation 1422027_a_at Ets1 E26 avian leukemia oncogene 1 1.17 2.47 1448694_at Jun Jun oncogene 1.02 6.76 1418261_at Syk Spleen tyrosine kinase 0.98 2.25 1419848_x_at Tlr7 Toll-like receptor 7 1.02 2.46 1415989_at Vcam1 Vascular cell adhesion molecule 1 1.06 4.17

Low concentrations: activation 1420249_s_at Ccl6 Chemokine (C-C motif) ligand 6 2.41 2.30 1416529_at Emp1 Epithelial 1 10.72 4.67 1455251_at Itga1 Integrin-a 1 15.13 12.66 1440847_at Mtss1 Metastasis suppressor 1 5.38 4.86 1419132_at Tlr2 Toll-like receptor 2 2.80 2.84

Gradient: activation 1418982_at Cebpa CCAAT/enhancer binding protein, a 1.70 2.39 1420703_at Csf2ra Colony stimulating factor 2 receptor, a 1.91 3.45 1435477_s_at Fcgr2b Fc receptor, IgG, low affinity IIb 28.66 168.27 1421173_at Irf4 Interferon regulatory factor 4 7.47 21.98 1423547_at Lyzs Lysozyme 8.93 16.70 1424852_at Mef2C myocyte enhancer factor 2C 12.79 45.05 1415960_at Mpo Myeloperoxidase 118.74 251.91 1422928_at Ne neutrophil elastase 36.48 51.33 1420715_a_at Pparg Peroxisome proliferator activated receptor gamma 10.43 37.18 1419537_at Tcfec Transcription factor EC 140.74 422.78 1418162_at Tlr4 Toll-like receptor 4 36.77 121.26

Low requirement: activation 1418796_at Clec11a C-type lectin domain family 11, member a 2.52 0.95 1422279_at Fv1 Friend virus susceptibility 1 4.32 0.88 1448575_at Il7r receptor 2.53 1.21 1450495_a_at Klrk1 Killer cell lectin-like receptor subfamily K, member 1 2.51 0.75 1437155_a_at Wwtr1 WW domain containing transcription regulator 1 2.50 0.97

High concentrations: repression 1420802_at Il13 1.11 0.44 1418741_at Itgb7 Integrin-b 7 1.05 0.31 1452562_at Jarid1d jumonji, AT rich Interactive domain 1D 1.13 0.27 1457670_s_at Lmna Lamin A 1.05 0.43 1427628_a_at Tcrb-V8.2 T-cell receptor-b, variable 8.2 1.03 0.45

Low concentrations: repression 1421186_at Ccr2 Chemokine (C-C motif) receptor 2 0.15 0.24 1421647_at Cd1d2 CD1d2 antigen 0.25 0.68 1444295_at Neo1 neogenin 0.39 0.74 1452405_x_at Tcra T-cell receptor-a chain 0.41 0.56 1443115_at Tgfbr2 Transforming growth factor, b-receptor II 0.32 0.75 (Tgfbr2), transcript variant 2

Gradient: repression 1449991_at Cd244 CD244 natural killer cell receptor 2B4 0.75 0.18 1419178_at Cd3g CD3 antigen, gamma polypeptide 0.79 0.11 1423344_at Epor 0.90 0.59 1448716_at Hba-x Hemoglobin X, a-like embryonic chain in Hba complex 0.62 0.51 1420692_at Il2ra receptor, achain (CD25) 1.01 0.67 1427142_s_at Jarid1b Jumonji, AT rich interactive domain 1B 0.79 0.35 1421304_at Klra2 Killer cell lectin-like receptor, subfamily A, member 2 0.55 0.39 1445399_at Klrb1d Killer cell lectin-like receptor subfamily B member 1D 0.30 0.25 1421965_s_at Notch3 Notch gene homolog 3 0.86 0.30 1417986_at Nrarp Notch-regulated ankyrin repeat protein 0.88 0.40

Low requirement: repression 1449619_s_at Arhgap9 Rho GTPase activating protein 9 0.46 1.28 1456328_at Bank1 B-cell scaffold protein with ankyrin repeats 1 0.31 1.24 1418480_at Cxcl7 Chemokine (C-X-C motif) ligand 7 0.58 1.58 1420678_a_at Il17rb receptor B 0.48 1.50 1434914_at Rab6b RAB6B, member RAS oncogene family 0.43 1.02

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1219 Table 2 (Continued ) Affymetrix probe set ID Gene symbol Description Blac BN

No significant regulation: 1440341_at Csf1r Colony stimulating factor 1 receptor 1.20 1.10 1417065_at Egr1 Early growth response 1 0.71 0.55 1450665_at Gabpa GA repeat binding protein, a 0.99 1.05 1422046_at Itgam Integrin-a M (CD11b) 1.14 1.44 1418634_at Notch1 Notch gene homolog 1 1.04 1.18 Affymetrix Mouse Genome Microarray 430 2.0 expression data was analyzed with GeneSpring software (described in Materials and methods). Numbers indicate average expression (n ¼ 3) in Sfpi1Blac/Blac and Sfpi1BN/BN cells relative to Sfpi1À/À cells.

relative to KO (Table 2). Next, we wanted to determine Bioinformatic analysis of predicted PU.1 binding sites significant (X2-fold) genome-wide changes in gene expression We expected that we could determine if the gradient activated in response to the two discrete hypomorphic levels of PU.1. and gradient repressed genes were directly regulated by PU.1 by Out of a total 45 101 Affymetrix probe sets, 1135 searching for consensus-binding sites in their promoters. To were significantly activated in BN cells, 529 in Blac cells accomplish this, we compiled known and validated PU.1 target and 347 in both BN and Blac cells (Figure 3a left). PU.1 is genes (Supplementary Table S3) to create a computational PU.1- mainly thought of as an activator, but has recently been shown binding site matrix. With this position weight matrix, we probed to have repressor function as well.26,28,39 Therefore, we the promoters of 549 gradient activated and 400 gradient performed a similar analysis to identify genes that were repressed genes, as well as 580 housekeeping genes as a repressed by PU.1. 1186 probe sets were significantly repressed negative control, and identified conserved putative PU.1 in BN cells, 349 in Blac cells, and 171 in both BN and Blac cells binding sites (described in‘Materials and methods). Interestingly, (Figure 3a right). a majority of genes had PU.1 sites that were not conserved The patterns shown by the Venn diagrams suggest that genes between mouse and human and thus were excluded from this were regulated by PU.1 in four different modes of either analysis (data not shown). An example is transcription factor EC, activation or repression (Table 2). We decided to focus on genes which has five validated PU.1 binding sites in the mouse gene regulated in a gradient manner because dose-dependency promoter41 but none conserved in the human gene promoter. suggests that these genes may be direct targets of PU.1 We found that a similar number of genes in all three lists had (Figure 3b). Of the probe sets activated or repressed at both one predicted PU.1 binding site in their proximal promoters. BN and Blac expression levels of PU.1, 76.9% of the activated However, gene promoters from the gradient-activated list had a probe sets and 62.6% of the repressed probe sets were regulated significantly higher incidence of two or more predicted PU.1 in a gradient fashion (Figure 3a green). If the definition for binding sites than gene promoters from either the gradient- gradient regulation was relaxed such that activation in the Blac repressed list or the housekeeping-control list (Figure 4a). Next, cells did not have a X2-fold cutoff but was required to be we identified the positions of all of the predicted PU.1 binding anywhere between BN and KO values, the proportion of genes sites in the gradient-regulated lists and grouped them into activated or repressed in a gradient manner increased to 91.9% 100 bp regions (Figure 4b). The gradient-repressed list had of the probe sets activated or repressed at both the BN and Blac similar numbers of predicted PU.1 binding sites all along the levels of PU.1. This definition was applied to subsequent proximal promoter. The gradient-activated list had similar analyses of gradient regulated genes. numbers from À1000 to À300 relative to the transcription start To identify genes regulated by PU.1 in a gradient fashion, we site. However, the number of predicted PU.1 sites steadily utilized the web-accessible program Database for Annotation, increased in the most proximal 300 bp. Overall, the gradient Visualization and Integrated Discovery31 to divide our lists into activated genes had more predicted PU.1 binding sites in their functionally related groups (Figure 3c). The results show that the promoters than the gradient repressed or housekeeping genes, and many were clustered between À300 and the transcription two largest groups from both gradient-activated and gradient- start site. This suggests that activation versus repression of target repressed lists were receptors and transmembrane and genes by PU.1 might be mediated by different mechanisms, transcription factors, both of which are involved in development such as the context of the PU.1 binding sites. It will be important and function of the hematopoietic system. Other overlapping both to determine whether genes are repressed at the transcrip- groups included kinases, cytolytic proteases, and ion channels, tional level and if PU.1 directly interacts with repressed genes. which are involved in general cell homeostasis. Unique groups in the gradient-activated list were protein turnover, also involved in cell homeostasis, and protocadherins, which sense Gradient repressed genes include T-cell- and natural cell–cell interactions and propagate intracellular signaling killer cell-specific genes 40 (reviewed by Halbleib and Nelson ), an important function It has been established that PU.1 promotes myeloid develop- for immune cells. The Oxygen Transport group was unique to ment.14–16 Consistent with this, we found that the largest cluster the gradient-repressed list and included many globin genes that of genes in the gradient-activated list was myeloid specific and are normally expressed in erythroid cells. Globin genes have included known PU.1 targets (Table 2; Supplementary Table previously been shown as repressed by PU.1 via interaction with S3). We validated two representative target genes by real-time the erythroid transcription factor GATA-1.26 Overall, the RT–PCR and confirmed that the genes encoding neutrophil identities of genes activated or repressed by PU.1 in a gradient elastase (Ne) and transcription factor EC (Tcfec) were both manner were consistent with known functions of PU.1 in activated by PU.1 in a gradient fashion (Figure 5a). Additionally, regulating hematopoiesis. we were interested in the fact that analysis by DAVID showed

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1220

Figure 3 Genes regulated by PU.1 in a gradient fashion. (a) Venn diagrams representing the number of probe sets that were significantly (X2-fold) activated (left) or repressed (right) in BN (red) or Blac (yellow) cells versus KO cells. In the probe sets significantly regulated in both BN and Blac cells, the Gradient regulated subset is marked. (b) Heat maps created with GeneSpring software depicting levels of activation and repression of gradient-regulated lists. The color bar shows the brightest red as the highest level of activation and the brightest blue as the highest level of repression. Each line represents one probe set. (c) Pie charts classifying functions of gradient-activated (left) and gradient-repressed (right) lists. Identities of subsets with 10 or more probe sets are annotated.

that the largest category of genes regulated in a gradient manner hemoglobin X (Hba-x) were both repressed by PU.1 in a by PU.1 included receptors (Figure 3c). Therefore, we gradient manner. (Figure 5c). Unexpectedly, the second cluster performed a functional validation of the direct PU.1 target of gradient repressed genes included T-cell and NK-cell genes GM-CSF receptor-a, Csf2ra.42 BN cells, but not Blac or KO cells, (Table 2). Specifically, these genes were expressed in KO cells could be switched to GM-CSF dependence from IL-3 depen- and were repressed in a dose-dependent manner by increased dence, validating that PU.1 activated the Csf2ra gene signifi- PU.1 concentrations in Blac and BN cells. T-cell genes that were cantly in BN cells (Figure 5b; Table 2). repressed in a gradient fashion included CD25 (IL-2Ra, Il2ra) PU.1 represses erythroid development in favor of myeloid and CD3g (Cd3g). NK-cell genes that were repressed in a development.24–26 As expected, we found that one of the two gradient manner included CD244 NK-cell receptor 2B4 (Cd244) large clusters of genes in the gradient-repressed list was and the killer cell lectin-like receptors Klra2 and Klrb1d. erythroid specific (Table 2). We validated two representative We validated representative genes from the T-cell and NK-cell erythroid genes by real-time RT–PCR and confirmed that lineages by real-time RT–PCR, and both Cd244 and Il2ra were the genes encoding the erythropoietin receptor (Epor) and repressed in a gradient fashion (Figure 5d). Next, we performed

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1221 in BN/BN and Blac/Blac cells compared to WT cells (Figure 6b), confirming that PU.1 activates these genes.28 To determine if T/NK-cell genes were repressed in a gradient manner in response to two distinct hypomophic PU.1 concen- trations, we assessed transcription of Zap-70, CD244 and sterile transcripts of the constant regions of the TCRa and -b loci. Analysis by standard and real-time RT–PCR revealed that all four were repressed by PU.1 (Figure 6c). Next, we validated cell- surface protein expression levels of CD244 with flow cytometry and found that CD244 was expressed in Blac/Blac cells but not in WT or BN/BN cells (Figure 6b). In summary, several T/NK-cell specific genes were de-repressed in a gradient manner in PU. 1-hypomorphic ex vivo pro-B cell lines. This suggests that PU.1 represses T/NK-cell genes in both the myeloid and B-cell lineages.

Discussion

PU.1 regulates genes involved in hematopoietic cell fate decisions and enforces myeloid and B-cell gene expression programs after lineage commitment.14,16,18 PU.1 is down- regulated during early T-cell development,19,20 but the biologi- cal function of this is not known. In this study, we utilized PU.1 hypomorphic mice and established ex vivo cell lines to examine Figure 4 Bioinformatic analyses. Proximal promoters (1000 bp gene expression in response to three discrete concentrations of upstream of annotated transcription start site) of gradient activated B B and gradient repressed genes were scanned for predicted PU.1 binding PU.1: 20, 2 and 0% of wild-type levels. PU.1 is critical for sites. Numbers (a) and positions (average of each 100 bp) (b)of development of macrophages and neutrophils and has been identified consensus PU.1 binding sites are plotted. Housekeeping shown to repress erythroid-specific genes in hematopoietic genes were used as a negative control (a). progenitors and during myelopoiesis.23–26 Accordingly, in IL-3 dependent myeloid cell lines, many genes activated or repressed validations of cell-surface protein expression levels using flow in a dose-dependent manner were, respectively, myeloid or cytometry. CD244 was expressed only in KO cells but not in BN erythroid specific (Table 2; Figures 5a and c). Unexpectedly, or Blac cells, while CD25 was expressed in B10% of KO cells T-cell and NK-cell genes were repressed by PU.1 in a and repressed in Blac and BN cells in a dose-dependent manner dose-dependent manner in both IL-3 dependent myeloid and (Figure 5e). This data suggests that a novel function for PU.1 in IL-7 dependent pro-B cells (Table 2; Figures 5d, e and 6b, c). hematopoiesis may be to concurrently repress T-cell and NK- This suggests that PU.1 represses T/NK-cell genes to promote cell genes in myeloid cells. and enforce myeloid and B-cell identity, highlighting that selection of one lineage is concurrent with repression of alternate lineages. PU.1 represses T-cell- and natural killer-cell-specific Two groups have previously described in vitro systems to genes in cultured progenitor B cells examine the effects of PU.1 on gene expression by retrovirally PU.1 is expressed at all stages of B-cell development, but its infecting PU.1 cDNA into cytokine-dependent Sfpi1À/À myeloid function after B-cell commitment is not clear.28,43,44 Based on progenitor cells.36,45 The disadvantages of these approaches are the results described above, we hypothesized that PU.1 might that the levels of PU.1 expression cannot be controlled with also repress T-cell genes in B cells. Therefore, we re-examined precision and that induction of PU.1 results in terminally published microarray data from our lab, comparing gene differentiated myeloid cells. Our hypomorphic allelic system expression in wild type and Sfpi1À/À pro-B cells.28 Interestingly, has three major advantages over these approaches. First, many T-cell-specific genes were highly expressed in Sfpi1À/À changes in gene expression are likely independent of develop- pro-B cells but not in wild-type pro-B cells, including several mental stage, as myeloid differentiation is impaired in BN and T-cell receptor transcripts, some of which were also significantly Blac cell lines as a consequence of low PU.1 levels (Figures 2d regulated by PU.1 in the IL-3 dependent myeloid cells (Table 2). and e). Second, BN, Blac and KO cell lines grow at similar rates In order to analyze T-cell gene regulation with our system, we (Figure 2c), so gene expression is independent of changes in generated pro-B cell lines by placing wild-type, Sfpi1BN/BN and cell-cycle-related genes. Third, we can measure changes in gene Sfpi1Blac/Blac fetal liver cells into culture with stromal cells and expression in response to two distinct PU.1 concentrations. IL-7 (described in Materials and methods). Sfpi1À/À fetal liver Therefore, this cell culture system represents a powerful cells do not proliferate under these conditions because they do approach to study the effects of changes in PU.1 concentration not express the PU.1 direct target IL-7Ra, as previously on gene expression in hematopoietic cells. described.36 Wild-type cells generated rapidly proliferating The pattern of gene activation and repression observed in CD19 þ pro-B cell lines within 4 days, while both Sfpi1BN/BN response to three distinct PU.1 concentrations (Figure 3a) and Sfpi1Blac/Blac cells generated CD19 þ pro-B cell lines in suggests that PU.1 regulates downstream target genes in four culture with delayed kinetics (Figures 6a and b). Once distinct modes (Figure 7; Table 2). The ‘High Concentrations’ generated, Sfpi1 þ / þ (WT), Sfpi1BN/BN (BN/BN) and Sfpi1Blac/Blac mode is comprised of genes activated/repressed by B20% of (Blac/Blac) cell lines grew at similar rates. Surface expression of wildtype levels in BN cells, but not significantly regulated B220 and FcgRII/III were reduced in a dose-dependent manner by the B2% level in Blac cells (Figure 3a red). The ‘Low

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1222

Figure 5 Validations of gradient-regulated genes in IL-3-dependent cell lines. (a) Real-time RT–PCR for the myeloid genes Ne and Tcfec in BN and Blac cells relative to KO (set to 1). (b) Cytokine switch from media containing IL-3 into media containing either IL-3 or GM-CSF. Bars represent fold growth after 72 h relative to growth in media containing IL-3. (c) Real-time RT–PCR for the erythroid genes Epor and Hba-x in BN and Blac cells relative to KO (set to 1). (d) Real-time RT–PCR for the natural killer cell gene Cd244 and the T-cell gene Il2ra in BN and Blac cells relative to KO (set to 1). *Po0.10; **Po0.05; ***Po0.01. (e) Flow cytometry of single cell suspensions from BN, Blac and KO IL-3 dependent myeloid cell lines. Cells were gated for size and granularity and analyzed with antibodies to the indicated cell surface markers.

Concentrations’ mode contains genes activated/repressed in direct or indirect targets of PU.1. Intriguingly, two different Blac cells, but not further activated/repressed in BN cells laboratories have observed that mice with altered PU.1 (Figure 3a black). The ‘Gradient’ mode is comprised of genes expression frequently develop T-cell leukemia.11,12 Based on activated/repressed in a dose-dependent manner (Figure 3a our results, we speculate that inappropriate expression of T-cell green). The ‘Low Requirement’ mode contains genes activated/ genes in response to reduced PU.1 expression might contribute repressed in Blac cells, but then returned to near baseline in BN to T-cell leukemia. More work will be needed to identify the cells (Figure 3a yellow). Interestingly, several well-characterized molecular mechanism by which PU.1 represses T-cell gene PU.1 target genes, including M-CSFR (Csf1r) and Egr-1 (Egr1), expression in myeloid cells. were unaffected in Blac and BN cells (Table 2; Supplementary PU.1 is expressed throughout B-cell development until it is Table S3). We hypothesize that this is because the level of PU.1 downregulated upon differentiation into plasma cells.46 PU.1 is is below a threshold required to activate transcription of these required to generate B-cell progenitors, but its function after genes. B-cell commitment is not well understood, since B cells persist Our results suggest that an important function of PU.1 in and participate in immune responses after conditional inactiva- myeloid cells might be to enforce myeloid identity as well as tion of the Sfpi1 gene.43,44 Our results demonstrate that T/NK- prevent the expression of lineage-inappropriate genes such as cell genes are transcribed at high levels in cultured pro-B cells erythroid and T/NK-cell genes. It is known that PU.1 directly with reduced PU.1 expression (Figure 6c). Thus, we speculate interacts with GATA-1 at the promoters of erythroid target that an important function of PU.1 in B cells might be to repress genes,23,26 but we do not yet know whether T/NK-cell genes are T/NK-cell genes. Repression of lineage-inappropriate genes,

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1223

Figure 6 Activation of T/NK-cell genes in pro-B cells. (a) Growth of wild-type, Sfpi1BN/BN and Sfpi1Blac/Blac fetal liver cells in IL-7. Cells were counted every 48 h until cell count reached 10 million. (b) Flow cytometry of single cell suspensions from wild-type, BN/BN and Blac/Blac IL-7 dependent pro-B cell lines. Cells were gated for size and granularity and analyzed with antibodies to the indicated cell surface markers. Histograms have both unstained (open) and stained (colored) peaks. Numbers indicate mean fluorescence. (c) Real-time RT–PCR validation of the T-cell genes TCRb Constant region, TCRa Constant region and Zap-70, and the NK-cell gene Cd244 in BN/BN and Blac/Blac IL-7 dependent pro-B cell lines relative to WT (set to 1). *Po0.10; **Po0.05; ***Po0.01.

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1224 Conflict of interest/disclosure The authors declare no competing financial interests.

References

1 Klemsz MJ, McKercher SR, Celada A, Van Beveren C, Maki RA. The macrophage and B cell-specific transcription factor PU.1 is related to the ets oncogene. Cell 1990; 61: 113–124. 2 McKercher SR, Torbett BE, Anderson KL, Henkel GW, Vestal DJ, Baribault H et al. Targeted disruption of the PU.1 gene results in multiple hematopoietic abnormalities. EMBO J 1996; 15: 5647–5658. 3 Scott EW, Simon MC, Anastasi J, Singh H. Requirement of transcription factor PU.1 in the development of multiple hemato- Figure 7 Model for gene activation upon increasing concentrations poietic lineages. Science 1994; 265: 1573–1577. of PU.1. Target genes are regulated in four modes due to the different 4 Spain LM, Guerriero A, Kunjibettu S, Scott EW. T-cell development concentrations of PU.1. in PU.1-deficient mice. J Immunol 1999; 163: 2681–2687. 5 Colucci F, Samson SI, DeKoter RP, Lantz O, Singh H, Di Santo JP. Differential requirement for the transcription factor PU.1 in the including T-cell genes, is a well-characterized function of two generation of natural killer cells versus B and T-cells. Blood 2001; other transcription factors highly expressed in B cells, Pax-547 97: 2625–2632. and EBF.48 Therefore, PU.1 may cooperate with Pax-5 and EBF 6 Mueller BU, Pabst T, Osato M, Asou N, Johansen LM, Minden MD to repress T-cell genes and enforce B-cell lineage identity. It will et al. Heterozygous PU.1 mutations are associated with acute myeloid leukemia. Blood 2002; 100: 998–1007. be important to identify the degree to which PU.1 overlaps or 7 Vangala RK, Heiss-Neumann MS, Rangatia JS, Singh SM, Schoch differs from Pax-5 and EBF in this regard. C, Tenen DG et al. The myeloid master regulator transcription It is intriguing that an extremely low level of PU.1, such as factor PU.1 is inactivated by AML1-ETO in t(8;21) myeloid that observed in IL-3 dependent myeloid or IL-7 dependent pro- leukemia. Blood 2003; 101: 270–277. B cells generated from Sfpi1Blac/Blac fetal liver cells, can potently 8 Rosenbauer F, Wagner K, Kutok JL, Iwasaki H, Le Beau MM, repress target gene expression. Such low levels of PU.l would be Okuno Y et al. Acute myeloid leukemia induced by graded reduction of a lineage-specific transcription factor, PU.1. Nat predicted to exist in only one context in vivo, which is during Genet 2004; 36: 624–630. the silencing of PU.1 expression that occurs during erythroid or 9 Cook WD, McCaw BJ, Herring C, John DL, Foote SJ, Nutt SL et al. T-cell differentiation. PU.1 is expressed at the double-negative PU.1 is a suppressor of myeloid leukemia, inactivated in mice by (DN)1 and DN2 stages of early thymic T-cell development, but is gene deletion and mutation of its DNA binding domain. Blood dramatically downregulated upon differentiation into the DN3 2004; 104: 3437–3444. stage.49 Forced expression of PU.1 in pro-T cells efficiently blocks 10 Walter MJ, Park JS, Ries RE, Lau SK, McLellan M, Jaeger S et al. 22 Reduced PU.1 expression causes myeloid progenitor expansion differentiation at the DN3 stage. In addition, PU.1 in combina- and increased leukemia penetrance in mice expressing PML- tion with the myeloid transcription factor c/EBPa can reprogram RARalpha. Proc Natl Acad Sci USA 2005; 102: 12513–12518. 50 T-cell progenitors into myeloid lineage cells. It is clear from 11 Metcalf D, Dakic A, Mifsud S, Di Rago L, Wu L, Nutt S. these studies that PU.1 plays an important role in regulation of Inactivation of PU.1 in adult mice leads to the development gene expression during early T-cell development, during which, of myeloid leukemia. Proc Natl Acad Sci USA 2006; 103: as noted above, it is in the process of being silenced. 1486–1491. 12 Rosenbauer F, Owens BM, Yu L, Tumang JR, Steidl U, Kutok JL In summary, our cell culture system represents an excellent et al. Lymphoid cell growth and transformation are suppressed by a experimental system to investigate the mechanism of how target key regulatory element of the gene encoding PU.1. Nat Genet genes are regulated by PU.1 in a concentration-dependent 2006; 38: 27–37. manner. Taken together, our data suggests that a normal 13 Ross IL, Dunn TL, Yue X, Roy S, Barnett CJ, Hume DA. function of PU.1 might be to actively repress T-cell and/or Comparison of the expression and function of the transcription NK-cell gene expression in myeloid and B cells. factor PU.1 (Spi-1 proto-oncogene) between murine macrophages and B lymphocytes. Oncogene 1994; 9: 121–132. 14 DeKoter RP, Walsh JC, Singh H. PU.1 regulates both cytokine- dependent proliferation and differentiation of granulocyte/ Acknowledgements macrophage progenitors. EMBO J 1998; 17: 4456–4468. 15 Houston IB, Kamath MB, Schweitzer BL, Chlon TM, Dekoter RP. Reduction in PU.1 activity results in a block to B-cell We thank the Affymetrix GeneChip Microarray Core (Cincinnati development, abnormal myeloid proliferation, and neonatal Children’s Hospital Medical Center) for processing samples, Phil lethality. Exp Hematol 2007; 35: 1056–1068. Sanford of the Gene Targeted Mouse Service (University of 16 DeKoter RP, Singh H. Regulation of B lymphocyte and Cincinnati) for advice and H Leighton Grimes (Cincinnati macrophage development by graded expression of PU.1. Science Children’s Hospital Medical Center) and Brock Schweitzer 2000; 288: 1439–1441. 17 Dahl R, Walsh JC, Lancki D, Laslo P, Iyer SR, Singh H et al. (University of Cincinnati) for helpful discussions. MBK is a PhD Regulation of macrophage and neutrophil cell fates by the PU.1: candidate at University of Cincinnati, and this work is submitted C/EBPalpha ratio and granulocyte colony-stimulating factor. Nat in partial fulfillment of the degree requirements. This work was Immunol 2003; 4: 1029–1036. supported by National Institutes of Health grant AI052175 and 18 Laslo P, Spooner CJ, Warmflash A, Lancki DW, Lee HJ, Sciammas Ohio Cancer Research Associates grant 5407. R et al. Multilineage transcriptional priming and determination of Contribution: MBK, IBH and RPD designed and performed alternate hematopoietic cell fates. Cell 2006; 126: 755–766. 19 Back J, Allman D, Chan S, Kastner P. Visualizing PU.1 activity experiments and collected data. AJJ and XZ performed during hematopoiesis. Exp Hematol 2005; 33: 395–402. experiments. MBK, SG and AGJ analyzed and interpreted data 20 Nutt SL, Metcalf D, D’Amico A, Polli M, Wu L. Dynamic and performed statistical analyses. MBK and RPD drafted the regulation of PU.1 expression in multipotent hematopoietic manuscript. progenitors. J Exp Med 2005; 201: 221–231.

Leukemia PU.1 represses T/NK-cell genes in myeloid and B cells MB Kamath et al 1225 21 Schuetze S, Stenberg PE, Kabat D. The Ets-related transcription GATA-2 in the specification of myeloid cell fates. Immunity 2002; factor PU.1 immortalizes erythroblasts. Mol Cell Biol 1993; 13: 17: 665–676. 5670–5678. 37 Takemoto CM, Yoon YJ, Fisher DE. The identification and 22 Anderson MK, Weiss AH, Hernandez-Hoyos G, Dionne CJ, functional characterization of a novel mast cell isoform of the Rothenberg EV. Constitutive expression of PU.1 in fetal hemato- microphthalmia-associated transcription factor. J Biol Chem 2002; poietic progenitors blocks T-cell development at the pro-T-cell 277: 30244–30252. stage. Immunity 2002; 16: 285–296. 38 Nimmerjahn F, Ravetch JV. Fcgamma receptors: old friends and 23 Rekhtman N, Radparvar F, Evans T, Skoultchi AI. Direct interaction new family members. Immunity 2006; 24: 19–28. of hematopoietic transcription factors PU.1 and GATA-1: 39 Hu R, Sharma SM, Bronisz A, Srinivasan R, Sankar U, Ostrowski functional antagonism in erythroid cells. Genes Dev 1999; 13: MC. Eos, MITF, and PU.1 recruit corepressors to osteoclast-specific 1398–1411. genes in committed myeloid progenitors. Mol Cell Biol 2007; 27: 24 Rhodes J, Hagen A, Hsu K, Deng M, Liu TX, Look AT et al. 4018–4027. Interplay of pu.1 and gata1 determines myelo-erythroid progenitor 40 Halbleib JM, Nelson WJ. Cadherins in development: cell adhesion, cell fate in zebrafish. Dev Cell 2005; 8: 97–108. sorting, and tissue morphogenesis. Genes Dev 2006; 20: 25 Galloway JL, Wingert RA, Thisse C, Thisse B, Zon LI. Loss of gata1 3199–3214. but not gata2 converts erythropoiesis to myelopoiesis in zebrafish 41 Rehli M, Lichanska A, Cassady AI, Ostrowski MC, Hume DA. TFEC embryos. Dev Cell 2005; 8: 109–116. is a macrophage-restricted member of the microphthalmia-TFE 26 Stopka T, Amanatullah DF, Papetti M, Skoultchi AI. PU.1 inhibits subfamily of basic helix-loop-helix leucine zipper transcription the erythroid program by binding to GATA-1 on DNA and creating factors. J Immunol 1999; 162: 1559–1565. a repressive chromatin structure. EMBO J 2005; 24: 3712–3723. 42 Akagawa E, Muto A, Arai K, Watanabe S. Analysis of the 50 27 Houston IB, Huang KJ, Jennings SR, DeKoter RP. PU.1 immorta- promoters for human IL-3 and GM-CSF receptor alpha genes. lizes hematopoietic progenitors in a GM-CSF-dependent manner. Biochem Biophys Res Commun 2003; 300: 600–608. Exp Hematol 2007; 35: 374–384. 43 Polli M, Dakic A, Light A, Wu L, Tarlinton DM, Nutt SL. The 28 Schweitzer BL, DeKoter RP. Analysis of gene expression and Ig development of functional B lymphocytes in conditional PU.1 transcription in PU.1/Spi-B-deficient progenitor B cell lines. knock-out mice. Blood 2005; 106: 2083–2090. J Immunol 2004; 172: 144–154. 44 Ye M, Ermakova O, Graf T. PU.1 is not strictly required for B cell 29 Pfaffl MW. A new mathematical model for relative quantification development and its absence induces a B-2 to B-1 cell switch. in real-time RT-PCR. Nucleic Acids Res 2001; 29: e45. J Exp Med 2005; 202: 1411–1422. 30 Markey MP, Bergseid J, Bosco EE, Stengel K, Xu H, Mayhew CN 45 Henkel GW, McKercher SR, Maki RA. Identification of three genes et al. Loss of the retinoblastoma tumor suppressor: differential action on transcriptional programs related to cell cycle control and up-regulated in PU.1 rescued monocytic precursor cells. Int immune function. Oncogene 2007; 26: 6307–6318. Immunol 2002; 14: 723–732. 31 Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J 46 Cattoretti G, Shaknovich R, Smith PM, Jack H-M, Murty VV, et al. The DAVID gene functional classification tool: a novel Alobeid B. Stages of germinal center transit are defined by B cell biological module-centric algorithm to functionally analyze large transcription factor coexpression and relative abundance. gene lists. Genome Biol 2007; 8: R183. J Immunol 2006; 177: 6930–6939. 32 Kodandapani R, Pio F, Ni CZ, Piccialli G, Klemsz M, McKercher S 47 Nutt SL, Heavey B, Rolink AG, Busslinger M. Commitment to the et al. A new pattern for helix-turn-helix recognition revealed by the B-lymphoid lineage depends on the transcription factor Pax5. PU.1 ETS-domain-DNA complex. Nature 1996; 380: 456–460. Nature 1999; 401: 556–562. 33 Wingender E, Dietze P, Karas H, Knuppel R. TRANSFAC: a 48 Pongubala JM, Northrup DL, Lancki DW, Medina KL, Treiber T, database on transcription factors and their DNA binding sites. Bertolino E et al. Transcription factor EBF restricts alternative Nucleic Acids Res 1996; 24: 238–241. lineage options and promotes B cell fate commitment indepen- 34 Hinrichs AS, Karolchik D, Baertsch R, Barber GP, Bejerano G, dently of Pax5. Nat Immunol 2008; 9: 203–215. Clawson H et al. The UCSC Genome Browser Database: update 49 Tydell CC, David-Fung E-S, Moore JE, Rowen L, Taghon T, 2006. Nucleic Acids Res 2006; 34 (Database issue): D590–D598. Rothenberg EV. Molecular dissection of prethymic progenitor entry 35 Schwartz S, Elnitski L, Li M, Weirauch M, Riemer C, Smit A et al. into the T lymphocyte developmental pathway. J Immunol 2007; MultiPipMaker and supporting tools: alignments and analysis of 179: 421–438. multiple genomic DNA sequences. Nucleic Acids Res 2003; 31: 50 Laiosa CV, Stadfeld M, Xie H, de Andres-Aguayo L, Graf T. 3518–3524. Reprogramming of committed T-cell progenitors to macrophages 36 Walsh JC, DeKoter RP, Lee HJ, Smith ED, Lancki DW, Gurish MF and dendritic cells by C/EBPa and PU.1 transcription factors. et al. Cooperative and antagonistic interplay between PU.1 and Immunity 2006; 25: 731–744.

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

Leukemia