Hematopoietic Lineage Commitment: Role of Transcription Factors

Transcription

1 Hematopoietic Lineage Commitment: Role of Transcription Factors John H. Kehrl B Cell Molecular Immunology Section, Laboratory of Immunoregulation, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland, USA Key Words. Hematopoiesis Myelopoiesis Lymphopoiesis Transcription Lineage Differentiation Abstract. This review focuses on the roles of transcription factors in hematopoietic lineage commitment. A brief introduction to lineage commitment and asymmetric cell division is followed by a discussion of several methods used to identify transcription factors important in specifying hematopoietic cell types. Next is presented a discussion of the use of embryonic stem cells in the analysis of hematopoietic gene expression and the use of targeted gene disruption to analyze the role of transcription factors in hematopoiesis. Finally, the status of our current knowledge concerning the roles of transcription factors in the commitment to erythroid, myeloid and lymphoid cell types is summarized. Introduction Hematopoietic stem cells have the ability both to self-renew and to differentiate into multipotent progenitors. These two features are readily evident in the ability of these cells to reconstitute the hematopoietic system and to sustain long-term hematopoiesis following bone marrow ablation [ 11. The mechanisms controlling the decision of a hematopoietic stem cell to self-renew versus differentiate are unknown. Theoretically, a balance between symmetrical cell divisions that results in self-renewal versus that which results in differentiation could maintain the stem cell pool as well as provide a constant source of multipotent progenitors. This Correspondence: Dr. John H. Kehrl, B Cell Molecular Immunology Section, Laboratory of Immunoregulation, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD 20892, USA. Received October 5, 1994; accepted for publication October 5, OAlphaMed Press /95/$5.00/0 balance could be regulated by a stochastic mechanism, environmental signals, or both. Alternatively, hematopoietic stem cells could divide asymmetrically, producing daughter cells with different capacities for self-renewal and differentiation. This would maintain the stem cell pool and generate multipotent progenitors. Asymmetrical cell divisions are a fundamental property of embryonic development where the generation of cell diversity requires daughter cells to have distinct fates, not dissimilar to the commitment to various lineages during hematopoiesis. In principle, different daughter cells can arise in two ways. They may differ from each other at the time of cell division (i.e., an intrinsic property of the daughter cells). For example, a parental cellular factor such as a transcription factor or a cell surface receptor may be distributed disproportionately to the daughter cells. Alternatively, the daughter cells may be similar at the time of cell division, but become different due to the subsequent exposure to distinct environmental signals such as a hormone or cytokine (i.e., a signal extrinsic to the daughter cells). Examples of both intrinsically and extrinsically determined asymmetric cell divisions have been found in developmental pathways [2]. While it is unknown whether totipotent stem cells undergo asymmetrical cell divisions, there is evidence that multipotent progenitors do. The analysis of the colony forming potentials of individual daughter cells has demonstrated that approximately 10% of the cell divisions of multipotent progenitors are asymmetric [3-51. This frequency is not altered by the presence of a variety of cytokine combinations [5]. The commitment of hematopoietic cells to various cell lineages occurs as multipotent progenitors progress through an irreversible descending hierarchy of differentiation steps that STEM CELLS 1995;13:

2 224 Hematopoietic Lineage Commitment eventually results in the production of mature hematopoietic cells. As hematopoietic lineage commitment proceeds, there is a concomitant loss in the capacity of committed progenitors to differentiate to other cell types. Two major models to explain lineage commitment and hematopoietic differentiation have been proposed. In one model, lineage commitment is regulated by external signals such as hematopoietic growth factors and/or progenitor-stromal cell interactions [6-81. This is similar to the extrinsic signal model discussed above in the context of asymmetrical cell divisions during embryonic development. Since their discovery, hematopoietic growth factors have been logical candidates for directing hematopoietic cell lineage commitment. However, while some studies have supported such a role, others suggest that growth factors contribute only a minor role in determining cell types. These latter studies propose a second model, whereby a stochastic mechanism accounts for lineage commitment [5, In this model, growth factors expand already committed cell types but do not bias the cells toward a specific lineage. The fate of the daughter cells is intrinsically determined. A third model that combines aspects of both these models has also been proposed [13]. Whichever model of hematopoietic lineage commitment is correct, transcription factors are likely to play pivotal roles in specifying lineage phenotypes. In the growth factor/stromal cell model, engagement of specific receptors on hematopoietic progenitors would generate downstream signals that eventually result in the enhancement or repression of transcriptional rates of a select group of genes necessary for lineage commitment and hematopoietic differentiation. In the stochastic model, random activation of a cascade of interacting transcription factors would regulate the transcriptional rates of the structural and regulatory genes necessary for determining a specific cell lineage. In both models, autoregulatory and negative regulatory elements in promoters and enhancers of lineagespecific genes are likely to assist in specifying cell lineages. In addition, passive regulatory mechanisms such as the reorganization of chromatin patterns and alterations in the methylation status of regulatory and structural genes would help account for the irreversible nature of lineage commitment. Approaches Used for the Identification of Transcription Factors Likely to Be Important in Hematopoietic Lineage Commitment One of the most direct approaches has been to identify cis-elements important in the regulation of lineage-specific genes. Such cis-elements present in the promoter(s), enhancer(s), or locus control region(s) of lineage-specific genes can be used to identify specific DNA binding proteins (likely transcription factors) with which they interact. An example of such an approach was the identification of GATA-1, a member of the zinc finger family of transcription factors that was first identified in the analysis of the tissue-specific expression of the globin genes [14, 151. Transfection studies localized a strong erythroid specific enhancer in the chicken p globin locus. The enhancer contained binding sites for a novel factor designated as Eryfl (now called GATA-1). GATA-1 has two highly conserved zinc fingers reminiscent of the DNA binding domain of the steroid receptor family. It is expressed at low levels in multipotent progenitors, and at high levels in erythroid cells and two related cell lineages, mast cells and megakaryocytes (see below under Erythroid Lineage Commitment). Transcription factors that transform hematopoietic cells are also likely candidates for regulating lineage commitment. For example, the v-myb gene transduced by the avian myeloblastosis virus (AMV) induces myeloid leukemias in chickens and transforms myeloid cells in vitro [16]. The DNA binding domain of v-myb is composed of two imperfectly conserved 52 amino acid repeats located near the amino terminus of the protein. The cellular homolog of v-myb, c-myb, is present at high levels in myeloid, lymphoid, and erythroid progenitors, but not in mature hematopoietic cells [17]. Ectopic expression of c-myb blocks hematopoietic differentiation, and expression of c-myb in conjunction with another myeloid specific transcription factor, NF-M, is sufficient to induce markers of myeloid differentiation in fibroblasts and erythroid cell lines (see below under Myeloid Lineage Commitment) [ 181. Another approach has been to examine the expression levels of known transcription factors in purified progenitors (and/or cell lines) and attempt to correlate those expression levels with commitment to specific lineages. An example of

3 Kehrl 225 such an approach is the ongoing attempt to correlate the expression levels of various homeobox genes with specific hematopoietic cell lineages [ 191. Homeobox genes were first discovered in Drosophila melanogaster, where they were found to be essential for pattern formation during early embryogenesis [20]. Approximately 40 human homeobox genes designated as HOX genes are localized in four gene complexes on chromosomes 2, 7, 12 and 17 [21]. In addition, a number of other human homeobox genes have been found which map outside of these four chromosome loci. Many of the HOX genes as well as other homeobox genes are expressed in hematopoietic progenitors. Their potential importance in the regulation of hematopoiesis has been recently reviewed [19], although some of the data are summarized below (see Lymphoid, Erythroid, and Myeloid Lineage Commitment). Finally, cdnas from progenitor populations or progenitor cells can be used to isolate genes that encode for members of a specific transcriptional factor family using conserved sequence information. For example, GATA-2 and GATA-3 encode proteins that may be important in multipotent hematopoietic progenitors and T lymphocyte lineage commitment, respectively. They were discovered based on their homology to GATA-1 [22]. Use of Embryonic Stem Cells and Target Gene Disruption in the Analysis of Hematopoietic Lineage Commitment Murine embryonic stem (ES) cells have proven to be extremely useful for the analysis of hematopoiesis and lineage commitment. ES cells are totipotent cells derived from the inner cell mass of blastocysts that can be maintained in a pluripotent state in culture [23, 241. Genetically modified ES cells can resume embryonic development after injection into a blastocyst and transfer to a pseudo-pregnant foster mother [25]. The mutant ES cells can contribute to all cell lineages of the chimeric mice, including the germline. If germline transmission occurs, the offspring can be crossed to establish mice homozygotic or heterozygotic for the mutation. A growing number of genes have been disrupted in this manner, including several genes important in hematopoiesis. For example, GATA-1 was disrupted in male ES cells by homologous recombination [26]. GATA-1 is localized on the X chromosome and therefore conveniently exists as a single copy in male ES cells. As such, a single gene-targeting event was sufficient to eliminate GATA-1 from the ES cells. In chimeric mice generated with the GATA-1- ES cells, the modified ES cells contributed to all tissues tested with the exception of cells of the erythroid lineage, thus revealing a requirement for GATA-1 in normal erythroid differentiation. The role of GATA- 1 is discussed further in the erythroid lineage commitment section, and other examples of specific gene targeting and the consequences for hematopoietic lineage commitment are discussed below. Targeting a specific gene in ES cells may lead to heterozygotic mice that are phenotypically normal and homozygotic mice that die early in development. This makes the analysis of the functional consequences of a mutation difficult. In examining null mutations in genes important in lymphoid cell development, one solution is to use RAG-2 deficient mice [27]. RAG-2 is necessary for the normal rearrangement of the immunoglobulin and T cell receptor genes. In the absence of RAG-2 there is a lack of mature B and T lymphocytes [28]. When RAG-2-deficient blastocytes are injected with genetically modified ES cells, all the lymphoid cells that develop in the chimeric mouse arise from the ES cells [27]. Thus, the development and functional competence of lymphoid cells in the chimeric mice can be directly evaluated. In addition to their use in the generation of mutant mice, ES cells can be used to study early events in hematopoiesis in vitro. When ES cells are cultured for eight to ten days in semisolid methylcellulose medium, they differentiate into discrete colonies containing mixed hematopoietic lineages [ Erythroid and myeloid cells are easily recognized in 70 to 90% of the colonies. Replating of these colonies in secondary methylcellulose cultures results in the sequential development of erythroid, multilineage, macrophage, and mast cell progenitors. In addition, genes sequentially activated during normal hematopoiesis are activated with similar kinetics in the differentiating ES cells [33]. ES cells can also be induced to differentiate into lymphoid progenitors. ES cells cocultured with a stromal cell line and a mixture of cytokines differentiate into T and B lymphoid progenitors [34]. The progeny of ES cells induced in vitro can reconstitute the lymphoid

4 226 Hematopoietic Lineage Commitment compartments of SCID mice which are deficient in mature B and T lymphocytes [35] or of RAG-2 deficient mice [36]. This ES cell differentiation system can be used to analyze the consequences of either disrupting or enhancing the expression of a gene that is postulated to be important in hematopoiesis. Both antisense constructs and the more direct targeted disruption of a select gene have been used successfully in this system. The former approach was used to analyze the importance of the vav proto-oncogene in hematopoiesis [37]. Vav is restricted to hematopoietic cells, and structurally it has features of both a transcription factor and a signal transducing molecule. ES cells expressing high levels of VUV antisense RNA were blocked in their differentiation into hematopoietic cells. In addition, antisense vav ES cells failed to express two hematopoietic lineage-specific transcription factors, GATA-1 (mentioned above) and PU.l. PU. 1 is a member of the ets family, and it is expressed in most hematopoietic cells with the exception of T lymphocytes (see Lymphocyte and Monocyte Lineage Commitment) [38]. Thus, vav appears to be very important in early commitment to the hernatopoietic lineage. Similar studies should identify other genes critical for early hematopoietic lineage commitment. Erythroid, Mast Cell and Megakaryocyte Lineage Commitment Erythroid progenitors are committed to one of two lineages early in development. The primitive lineage cells develop in the yolk sack and predominantly synthesize fetal hemoglobins, while the definitive lineage cells develop in the fetal liver and predominantly synthesize adult hemoglobins. Eventually erythropoiesis shifts from the fetal liver to the bone marrow. Early studies of the regulation of globin expression provided the basis for unraveling some of the mechanisms important in commitment to the two erythroid lineages and the differentiation of erythroid progenitors [39]. From these studies GATA-1 emerged as a key regulator of globin expression and erythroid-specific gene expression, as well as megakaryocytic and mast cell-specific gene expression. As mentioned above, GATA- 1 is the founding member of a family of zinc finger-containing proteins (GATA-1 through GATA-4) that recognizes the core nucleotide sequence GATA [22]. Potential target sites for GATA-1 are found in the regulatory regions of virtually all erythroidspecific genes and are often associated with binding sites for the ubiquitous transcription factors SP-1 and NF-1 [40]. GATA sites are also found in the promoters of several megakaryocytic-specific genes, including platelet factor 4 and glycophorin TIb [ The mast cell carboxypeptidase A promoter also contains a GATA site, and mutation of this site disrupts function of the promoter [44]. Analysis of the levels of GATA- 1 expression in various erythroid precursors has shown a positive correlation with the degree of commitment to the erythroid lineage [45]. In addition, when certain transformed cell lines that exhibit a erythrocytic/megakaryocytic phenotype (GATA- 1') are induced towards a myeloid phenotype, there is a rapid downregulation in GATA-1 [46]. Studies of erythropoietin and its receptor have revealed that GATA- 1 transactivates the erythropoietin receptor, and erythropoietin itself enhances GATA-I expression, thus forming the basis for a positive regulatory loop [ In contrast to more mature erythroid precursors, multipotent progenitors express GATA-1 at very low levels, or not at all. However, another GATA family member, GATA-2, is present at high levels in these cells [45,46]. In fact, GATA-2 may serve as an important factor for progenitor cell proliferation as hematopoiesis from GATA-2- ES cells is severely impaired [50]. As erythroid cells mature, GATA-2 levels decline and GATA- 1 levels rise. As mentioned above, the studies with GATA-1- ES cells demonstrated a requirement for GATA- 1 in the differentiation of erythroid cells. Proof that GATA-1 was solely responsible for the phenotype was provided by rescuing normal erythroid differentiation by reintroduction of GATA-1 into the GATA-I-ES cells [51]. Additional experiments using GATA-I- ES cells induced to differentiate in vitro demonstrated the importance of GATA- 1 for both the primitive and definite erythroid progenitors. GATA-1 was found to be necessary for the development of primitive erythroid precursors as assessed by replating cells derived from differentiated GATA-I- ES cells in the presence of erythropoietin. Definitive precursors (assessed by replating differentiated ES cells in the presence of both erythropoietin and kit ligand) developed in the absence of GATA- 1, but were arrested at the proerythroblast stage. GATA-2

5 Kehrl mrna transcripts were markedly elevated in the GATA-1- proerythroblasts, suggesting that rising levels of GATA-I contribute to the downregulation of GATA-2 during erythroid lineage commitment [50]. Since GATA-1 plays such a key role in erythroid differentiation, understanding its genetic regulation is likely to provide further insights into erythroid lineage commitment. Analysis of its regulatory sequences has shown that full expression of GATA-I is dependent upon GATA-1 binding sites, thus establishing another positive regulatory loop [52]. While this again underscores the importance of GATA-1, it does not answer how GATA-I transcription is initiated. In addition to the GATA binding sites, the GATA-1 promoter contains multiple binding sites for ets proteins [53]. The v-ets oncogene, the first member of the ets family to be discovered, was found in the chicken E26 retrovirus [ 161. There are now multiple members of the ets family including Ets-1, Ets-2, PU.1, Erg, GABP-a: Elf-1, and Fli-1. Many of them are expressed in hematopoietic cells [54]. The conserved DNA-binding domain, the ets domain, contains approximately 85 amino acids, has no structural similarity to other known DNA binding motifs, and binds to the core nucleotide sequence GGAA. Several ets family members bind the GATA-1 promoter and transactivate reporter constructs containing the GATA-I promoter implicating them as candidates for initiating GATA-1 transcription [55]. Besides the identification of the GATA family of transcription factors, the studies of the regulatory mechanism controlling globin expression also identified locus control regions (LCRs). They are another key element likely important in tissue-specific gene expression and perhaps in lineage commitment. These sites were first detected in the human P-globin locus as DNase hypersensitive sites located kb upstream from the gene cluster [56]. When LCR elements are linked to globin genes and introduced transgenically, the expression of the transgene is independent of the site of integration and linearly related to the number of transgene copies integrated [57]. This contrasts with transgenic constructs without the LCR, where expression is dependent upon the chromosomal location of the integration site. This property of conveying position independent expression distinguishes LCRs from classical enhancers. Verifying the importance of these regions are naturally occurring deletions of LCR 227 sequences that account for some forms of P- thalassemia [58]. Functionally, LCRs help to activate nearby promoters even when genes are embedded in an area of inactive chromatin. The globin LCRs contain most of the cis-acting elem\ts necessary for high-level globin expres- Besides the GATA motifs, a tandem AP-1 sio? like motif in the P-globin LCR binds an erythrdid-specific factor designated NF-E2. NF- E2 is a hpterodimeric basic leucine zipper transcription, factor composed of a 45 kda hematopoietic-specific component and an 18 kda ubiquitous component [59, 601. NF-E2 was first distinguished from AP-1 by its faster mobility on gel shift assays, its restriction to erythroid and megakaryocytic cells, and its requirement for a G residue at -2 relative to the AP-1 core [ The 45 kda component of NF-E2 was affinity purified, tryptic peptides subjected to microsequencing, and cdna clones isolated. It was found to be a member of the basic leucine zipper family. All erythroid, hematopoietic progenitor, megakaryocytic and mast cell lines were found to express p45 NF-E2, while macrophage, lymphoid and fibroblast cell lines did not, an expression pattern very similar to that of GATA-1 [59]. The coexpression of GATA-1 and p45 NF-E2 in mast cells, erythroid cells and megakaryocytes, all of which are descendants of a common committed progenitor, underscores the importance of these two proteins in commitment to these lineages [41, 641. A third gene denoted SCL or tal-1 may also be important in erythroid commitment. SCL was originally identified at a breakpoint in a subset of patients with acute T cell leukemia; however, SCL is not normally transcribed in T cells, but rather is predominantly expressed in megakaryocytes, erythroblasts and the developing brain [65]. SCL is a member of the basic helix-loop-helix (bhlh) family of transcription factors. Among the family members are c- myc, myod, and E12/47 (see below under Lymphoid Lineage Commitment). Antibodies were used to study SCL expression in developing mice embryos. SCL was detected at day 7.5 post coitus (pc) approximately 24 h before formation of recognizable hematopoietic elements. SCL expression localized to the blood islands of the yolk sac and, subsequently, to fetal liver and spleen paralleling the appearance of hematopoietic activity [66]. Similar to GATA- 1, SCL is downregulated during the later stages

6 228 Hematopoietic Lineage Commitment of erythroid and megakaryocyte differentiation [67]. The SCL promoter contains two consensus GATA-1 sites making it a likely target for GATA-1 and may in part account for its coexpression with GATA-1 [68]. C-myb was mentioned above and is apparently important for all hematopoietic lineages. Targeted disruption of c-myb resulted in mutant mice that at day 15 pc were severely anemic [69]. Analysis of early hematopoiesis indicated that primitive erythropoiesis in the yolk sac was intact; however, definitive erythropoiesis in the fetal liver was markedly impaired. In addition, all other hematopoietic lineages were similarly affected. These results plus other in vitro studies [70] have suggested that c-myb is essential for proliferation of hematopoietic progenitors. There is also evidence that c-myb is important in regulation of CD34. It encodes a protein expressed on hematopoietic stem cells and lineage-specific progenitors. C-myb transactivates CD34 promoter constructs by interacting with multiple c-myb binding sites in the promoter region, and c-myb expression in a CD34-negative glioblastoma cell line results in the induction of CD34 expression [71]. Another likely transcriptional factor that has been shown to be important for early erythropoiesis is encoded by rbtn2 [72]. It belongs to a class of genes that encode for proteins with a LIM domain. This domain is cysteine-rich and is known to bind zinc [73]. It has two fingerlike structures that have been postulated to bind DNA although no evidence for direct DNA binding has been reported. rbtn2 is expressed in a variety of tissues with the highest level of rbtn2 transcripts observed in fetal liver [72]. Rbtn2 is localized in the nucleus of both primitive and definitive erythroid progenitors, but not in mature cells of the lymphoid or myeloid lineages. Recently, rbtn2 has been found complexed with SCL in the nucleus of erythroid progenitors [74]. Targeted dikruption of rbtn2 led to a failure of primitive erythropoiesis and embryonic lethality at day 10.5 pc. Yolk sac tissue derived from the homozygous mutant mouse failed to differentiate to erythroid cells in vitro although myeloid cells were detected. As well, double targeted ES cells failed to differentiate to hemoglobinized erythroid cells in vitro [72]. The phenotype observed with the rbtn2 null mutation is similar to that obtained by disruption of GATA-1. Both mutations result in a block of primitive and definitive erythropoiesis. This contrasts with the c-myb null mutation that results in a block in definitive but not primitive erythropoiesis. Together these studies provide a working model of erythroid lineage commitment and differentiation. High levels of GATA-2, c-myb, and likely some members of the ets family are necessary for early hematopoietic progenitors. Either a stochastic or external signal such as erythropoietin result in enhanced transcription of GATA-I and NF-E2. Increasing levels of GATA-1 result in a downregulation in GATA-2, and induction of SCL and perhaps rbtn2. GATA-1, NF-E2 and SCL transactivate erythroid specific genes including the globin genes and the erythropoietin receptor. Alterations in chromatin structure surrounding erythroid-specific genes occur presumably in conjunction with induction of these regulatory genes. Erythropoietin further increases GATA- I, enhancing GATA-I transcription, thus biasing the differentiating progenitor further along the erythroid lineage. This may be accounted for by a dosage effect as high levels of GATA-1 may be necessary to transactivate certain genes. Eventually, GATA-1 and SCL levels decline as erythroid progenitors further differentiate. Megakaryocyte and mast cell progenitors appear to proceed along the same initial commitment and differentiation pathway as erythroid progenitors. Other unidentified transcription factors may be important in regulating the later stages of differentiation and perhaps in the final commitment to the megakaryocytic or mast cell Iineages. Tables I and I1 summarize many of the transcription factors likely to be important in multipotent and erythroid progenitors, respectively. Lymphoid Lineage Commitment B and T lymphocytes arise from a common lymphoid progenitor and undergo complex developmental pathways composed of multiple steps that are defined by changes in the expression patterns of lineage-specific genes and the status of rearrangement of their respective antigen receptor genes. Here the focus will be on the role of transcription factors in commitment to the B and T lymphoid lineage, and not to their roles in the regulation of the later stages of lymphocyte development.

7 Kehrl 229 Table I. Transcription factors likely important In early hematopoietic progenitors Protein DNA-binding Cognate Candidate domain sequence targets GATA-2 [22] Zinc finger AGATAG GATA- 1 T A c-myb [17] myb CAACGG? PU.l [38] ets TGGGGAAGT? A A HLX [ 1021 vav [37]? Homeodomain???? The first stage in the B lymphocyte developmental pathway is the pro-b cell stage. This stage is characterized by a germline configuration of the immunoglobulin genes, i.e., no rearrangement of the p, K, or h genes, and the expression of several B lymphocyte-specific gene products including B220 (an isoform of the CD45 protein that is only expressed in B cells); the surrogate light chain proteins h5 and Vpre-P (part of the antigen binding receptor used by pro-b and perhaps pre-b cells); CD19 (a B lineage-specific plasma membrane protein that likely functions during antigen stimulation); and mb- 1 (a plasma membrane protein that is part of the antigen receptor complex and important in intracellular signaling) [ The promoters of several of these genes have been studied. For example, the mb-1 promoter is a target for ets family members and an early B cell-specific protein termed early B cell factor (EBF) or B-Lyf [ EBF is expressed at the pro-b cell stage and persists until the plasma cell stage. It shares limited homology with other DNA binding proteins with the exception of a short sequence similar to the consensus helix 2 of the bhlh domain. However, this region in EBF is not involved in DNA binding, but rather in protein dimerization. EBF has a potent transactivation capacity augmenting the expression of heterologous promoters with EBF sites in nonlymphoid cells [81]. Analysis of the Vpre-p promoter identified another early B lymphocytespecific DNA binding protein termed EBB-1 [82]. The EBB- 1 protein is apparently identical to the B cell-specific activator protein (BSAP) [83]. BSAP was initially described as the mammalian homolog of a DNA binding protein that recognizes a cis-element in the sea urchin late histone 2A and 2B genes [83]. Among hematopoietic cell types, BSAP was restricted to the B cell lineage with the exception of cells at the terminal stages of B cell differentiation, i.e., plasma cells. Purification and subsequent cloning of cdnas demonstrated that the BSAP protein was encoded by PAX-5 [84]. PAX-5 is a member of a family of genes originally identified as important in Table 11. Transcription factors likely important in lineage committment of erythrocytes, megakaryocytes and mast cells Protein GATA- 1 DNA-binding domain zinc finger AGATAG globin, glycophorin, Epo R, all erythroid specific genes, (Eryf 2, GF-I) T A platelet factor-4, gplb, gpllb, carboxypeptidase A [14,42,44,49] NF-E2 [ basic leucine TGCTGAGTCAC a- & P-globin LCRs, porphobinogen deaminase, zipper C C T ferrochelatase SCL [ basic helix- CANNTG? (TAL-1) loop-helix rbtn2 [72]?LIM??

8 230 Hematopoietic Lineage Commitment Drosophila melanogaster embryonic development and subsequently shown to be important in mammalian development [85]. The CD19 promoter has also been identified as a target gene for BSAP [86]. Recent gene targeting of PAX-5 resulted in a marked impairment of early B cell development (M. Busslinger, personal communication). Although EBF and BSAP share similar patterns of expression, they seem to regulate distinct sets of genes, as a promoter that contains functional binding sites for both proteins has not been isolated [87]. In contrast, DNA binding sites for members of the ets family have been found in the majority of B lineage-specific genes. Of the multiple ets family members, none of them are exclusively expressed in B lymphocytes. Ets-1 is found in both B and T cells; PU.l and Spi-B are expressed in all hematopoietic cells with the exception of T cells; and FLi-1, Erp, and Erg-3 are expressed in B as well as other cell types [54]. Although all ets family members recognize the GGA core sequence, differential binding by individual family members is specified by the flanking nucleotides [88, 891. One feature of ets family members is the requirement to interact with other factors in order to stimulate transcription. Examples include interaction with the bhlh protein El2 (see below) [90] and with NF-EMS, an apparently B cellspecific protein important in regulation of the immunoglobulin light chain [91]. TdT is also expressed early in the B cell lineage, but as well in the T cell lineage [92]. Analysis of the TdT promoter defined a functionally important binding site for ets family members and a binding site for a lymphocyte-specific protein LyF- 1 [92, 931. LyF-1 has been shown to be related to Ikaros, a zinc finger-containing protein first identified to interact with a G-rich sequence in the CD36 enhancer 187,941. The role of Ikaros in lymphoid lineage commitment is further discussed in the T cell section. Several members of the homeobox gene family are expressed in early B lymphoid progenitors. Oct-2 is predominantly expressed in B lymphocytes and has a POU-homeodomain [95]. It binds to an eight-base pair nucleotide sequence, ATGCAAAT, termed the octamer motif. It has been considered to be an important regulator of immunoglobulin gene expression as well as the expression of several other B lymphoid-specific genes. However, recent gene targeting of the Oct-2 locus in mice and a B cell line has demonstrated that Oct-2 is not essential for normal B lymphocyte development and immunoglobulin gene expression, but rather is likely important in late B cell differentiation [96, 971. Early expression of immunoglobulin genes may be dependent upon another POUhomeodomain protein Oct- 1 in association with the protein OCA-B. OCA-B is an apparently B lineage-specific protein implicated as a cofactor with Oct-1 and Oct-2 [98]. The homeodomain protein LH2 has also been found to be expressed at the early stages of B cell development as well as in T lymphoid cell lines [99]. Besides a homeodomain the LH2 protein also possess a LIM domain similar to rbtn2 described above. Two other homeobox genes termed HLWHB24 and HEX, respectively, have been found expressed in early B cell precursors as well as other progenitors [loo HLX was found to be expressed in CD34' but not CD34- bone marrow cells, and HLX antisense oligonucleotides impaired the proliferative response of the CD34' cells to interleukin 3 (IL-3) and granulocytemacrophage colony-stimulating factor (GM- CSF) [103]. However, the importance of the proteins encoded by the HLX, HEX, and LH2 gene in early lineage commitment remains to be determined. Another family of DNA binding proteins implicated in the regulation of gene expression in B cell progenitors is the bhlh group [104, Binding sites for bhlh proteins consist of variations of the nucleotide sequence CANNTG which are termed E boxes and are found in the promoter regions of many B lymphoid-specific genes and in the immunoglobulin (Ig) enhancers [106]. Most bhlh proteins bind DNA as homoor heterodimers. Ubiquitously expressed bhlh proteins often heterodimerize with more selectively expressed bhlh proteins to form a tissue-specific transcriptional factor complex. E2A encodes two polypeptides, El2 and E47, which are the prototypes for the bhlh family and which were identified on the basis of their interaction with a site in the immunoglobulin heavy chain enhancer [104, Despite the presence of E2A transcripts in most cell types, the binding of the E2A-encoded polypeptides (termed BCF1) to the pe2 and pe5 boxes present in the immunoglobulin enhancer is restricted to B cells [ 107, A B lineage-specific bhlh protein that heterodimerizes with E12/E47 has been

9 Kehrl sought to account for this B lineage specific expression of BCFl, but none have been identified to date. E2A-encoded polypeptides have been recently detected in pro-b cells via immunoblotting [109]. In addition, high levels of Id, another HLH protein, have been found in pro-b cells [ 110, Id lacks a basic domain and as such it heterodimerizes with other bhlh proteins, but the complex is unable to bind DNA. Thereby, Id functions as an inhibitor of bhlh protein function. The interaction between Id and the E2A-encoded polypeptides likely assists in the regulation of gene transcription during early B cell development. The Re1 family of transcription factors is known to be important in the regulation of gene transcription in lymphoid as well as other cell types [112]. Several members of this family regulate B cell immunoglobulin gene transcription and are likely necessary for the transition from pre-b cells to immature B cells [ 113, While it is likely that some Re1 family members are important in lymphoid lineage commitment, their actual role in this process is unknown. The last family of DNA binding proteins to be discussed in the context of B lymphocyte lineage commitment is the high mobility group (HMG) of proteins. The HMG box is an unusual DNA binding domain that interacts with the nucleotides in the minor groove of double-stranded DNA and induces a strong bend in the DNA helix [ It has been suggested that this bending may allow access of other DNA proteins or juxtapose distantly bound transcription factors, perhaps allowing the formation of a complex. The HMG proteins fall into two classes. Members of one class are expressed at relatively high levels, bind to DNA in a relatively nonspecific manner, lack a transcriptional activation domain, and may be involved in general chromatin structure. The other class of HMG proteins binds DNA in a sequence-specific manner, may or may not have a transcriptional activation domain, and some members have been found to have important roles in specifying cell lineages. Two members of this class, Lef-1 and sox-4, are expressed in pro-b and pre-b cells, but not in later stages of B cell development [116, They both have transactivation domains and are also present in T cell progenitors (see below), thus they may have a role in specification of the lymphoid lineage The commitment to the T cell lineage begins prior to the expression of CD3, CD4, CD8 (triple negative thymocytes) and prior to the rearrangement of T cell receptor genes. Triple negative thymocytes have been fractionated on the basis of CD44, CD25 (&-2Ra), and c-kit receptor expression [118]. The earliest T cell progenitors are CD44+, CD25-, and c-kit+. Subsequent developmental stages are delineated as follows: CD44+, CD25+, c-kit+; CD44-, CD25+, c-kir; and finally CD44-, CD25-, c-kit-. T cell receptor rearrangements are first detected at the CD44-, CD25+, and c-kir stage. CD25 is found on activated T cells, and its presence on triple negative thymocytes is accounted for by the expression of NF- KB and AP-1 in these cells [ Early thymocytes also secrete the T lineage-specific cytokine, IL-2. A critical transcription factor important in IL-2 gene regulation is NF-AT [ NF-AT is transiently expressed in triple negative thymocytes and later in mature T cells following cell activation via the T cell receptor complex. The presence of an activated phenotype by these early T cell receptor negative thymocytes is attributed to an activation event likely mediated by a thymocyte-stromal cell interaction [119]. The early expression of NF-AT may be crucial for thymocyte development. NF-AT was originally identified in studies of the IL-2 enhancer and is comprised of a Re1 related protein NF-ATp in conjunction with AP-1 [ NF-ATp is present in the cytoplasm and migrates to the nucleus following T cell activation where, in conjunction with AP-I, it binds to the NF-AT cis-element. Adjacent to the NF- AT site in the IL-2 promoter is a weak AP-1 site. In the absence of AP- 1, NF-AT does not bind to the NF-AT cis-element. Furthermore, mutations of the AP-1 site eliminate NF-AT-mediated transcriptional activation of NF-AT site-containing reporter plasmids [ Recently another NF- AT referred to as NF-ATc has been isolated. NF- ATc is highly related to NF-ATp and is also a distant Re1 family member [125]. The expression of NF-ATc was found to correlate better with IL-2 gene expression than NF-ATp, as NF- ATp has a much wider tissue distribution. The relative importance of these two proteins in IL-2 promoter activation remains to be elucidated. Other members of the NF-AT family which may be expressed in B lymphocytes as binding to an NF-AT cis-element have been detected using B cell nuclear extracts [ 1261.

10 232 Hematopoietic Lineage Commitment Similar to the analysis of transcriptional regulation of B lineage-specific genes, the studies of the T cell receptor genes and other T lineagespecific genes have led to the identification of several DNA binding proteins potentially important in T lymphocyte lineage commitment. Analysis of the regulatory regions of the CD36 gene of the CD3 T cell receptor complex identified an enhancer with T cell-restricted activity. One of the elements in the enhancer termed the G box (GAATGGGGGTGG) was used to clone a zinc finger-containing protein designated as Ikaros [94]. Ikaros has five zinc finger modules organized in two clusters reminiscent of a Drosophila gap protein called Hunchback. Ikaros transcripts were detected in mouse fetal liver at day 9.5 pc to 10.5 and at day 12 pc in fetal thymus, although yolk sac expression was not detected. Thus, expression of Ikaros correlates with the onset of T lymphopoiesis making it a candidate for having a role in T cell lineage commitment. lkaros transcripts were also found in mature T cells and certain B cell lines. Recently, Ikaros has been found to be related to the Lyf- 1 protein identified in the analysis of the TdT promoter [87]. TdT is expressed in both early T and B cell progenitors (see above in B lymphocyte section). Recent gene targeting of the Ikaros locus in the mouse resulted in a failure of B and T cell development (K. Georgopoulos, personal communication). Both early B and T cell progenitors are lacking, implicating Ikaros as a key factor in early specification of the lymphoid lineage. It has not been reported when in hematopoiesis Ikaros is first expressed. If present in multipotent progenitors, its expression would need to be suppressed in hematopoietic cells destined for other lineages. Careful study of Ikaros in early hematopoietic progenitors and forced expression in hematopoietic lineages other that lymphocytes should provide some intriguing information about the early steps in the commitment to the lymphoid lineage. Analysis of the transcription factors that regulate the expression of the T cell receptor-a and -6 genes [ 1271 and of the CD8 promoter [ 1281 has identified functional GATA-3 binding sites. GATA-3 mrna transcripts are detected in T, but not in B cell lines; however, they are not T lineage-specific since they have been detected in a number of other tissues [22]. Analysis of GATA-3 transcripts during mouse fetal thymocyte ontogeny demonstrates that transcripts are present by day 13/14 pc in the fetal thymus [127]. This precedes the expression of the CD8 which is first detected at day 16/17, but follows the expression of Ikaros making GATA-3 a potential target gene of Ikaros. Whether GATA-3 is as indispensable for T cell development as GATA-1 is for erythroid differentiation awaits the results of gene targeting experiments. Studies of the CD3-& and T cell receptor-a enhancers led to the identification of three HMG members expressed in early T cells, Tcf- 1, Lef- 1, and sox-4 [129]. The gene encoding Tcf-1 was cloned based on its affinity for the AACAAAG motif in the CD3-& enhancer [130]. Lef-1 was identified on the basis of its binding to the?tcaaag motif in the TCR-a enhancer and, in addition, a cdna was cloned by subtraction from a pre-b cell cdna library (see above) [116, Tcf-1 is expressed in all T lineage cells beyond the prothymocyte stage, although it is also present in a variety of other tissues as well. Sox-4 was cloned by homology to Tcf-1 and Lef-1 and was found expressed in T cell and pre-b cell lines and in the thymus [ sox-4 has a potent transactivation domain in addition to its HMG domain and is capable of transactivation of a multimerized sox-4 binding site fused to a minimal promoter [ The role of these genes in lymphoid lineage commitment remains to be delineated. Ets binding sites have been found in the enhancers associated with the T cell receptor genes and the 1L-2 gene, in the CD4 promoter, and in the 1L-2 receptor-p chain promoter [ 127, T cells express at least five members of the ets family including Ets- 1, Ets-2, GABPa, Elf-1, and Fli-1 [ Several ets binding sites including two in the T cell receptor p enhancer are located adjacent to a T/G rich motif termed the core site [ A DNA binding protein termed the core binding factor (CBF) binds to this region cooperatively with ets-1. CBF (also known as SL3-3, PEBP2, and SL3) binding sites have also been found in other TCR receptor enhancers and the CD3& gene [ CBF is a complex composed of at least two proteins. The CBFP subunit is a kda non-dna binding protein, and the CBFa-subunit is a kda DNA binding protein [141, CBFa is highly related or identical to the product of AMLl [ 143, AMLl is frequently involved in translocations in acute myeloid leukemia. It encodes a 250 amino acid protein related to the Drosophila segmentation protein, runt. Runt regulates the expression of pair-rule genes during Drosophila development

11 Kehrl 233 [ AMLl mrna transcripts were found in B and T lymphocyte, erythroleukemia, and chronic myelogenous leukemia cell lines [144]. The role of AMLl/CBF in early lymphoid development and in the commitment to other hematopoietic lineages remains to be clarified. Several homeodomain-containing proteins have been found to be expressed in the T cell lineage. LH-2 transcripts are present in both early T and B cell progenitors (see above) [99]. TCFpl encodes for a POU-homeodomain protein that was identified on the basis of its binding to a site in the T cell receptor p enhancer [146]. However, TCFPl mrna transcripts are also found in B lymphocytes. TCFpl binds not only to the E4 region in the TCRp enhancer, but also to the octamer motif [ Expression of TCFPI in early B and T cell progenitors has not been reported. HOXB7 mrna transcripts have been found expressed in a variety of different T cell lines and, interestingly, were found to be preferentially expressed in CD4 versus CD8 positive peripheral blood T cells [ The expression of HOXB7 during early T cell development has not been reported. Transcripts from HOXC4 were found in RNA prepared from T and B cell lines, but not from erythroid and myeloid cell lines. Again, levels in B and T cell progenitors were not reported [ While several of these genes may be important in B and/or T lymphoid lineage commitment, clearly further studies are needed before any conclusions can be drawn. In summary, the several transcription factors critical for lymphoid lineage commitment and differentiation along the B and T cell pathways have been identified (summarized in Table 111). Ikaros has emerged as a key factor in specifying the lymphocyte lineage. In the B lymphocyte developmental pathway, three transcription factors, BSAF', E2A, and EBF, are likely to be essential for early B cell development. There are several obvious candidates for essential transcription factors in specifying the T cell lineage although there are no reported gene targeting experiments in which the resulting phenotype is an early block in T cell development without an interference in the development of other cell lineages. Myeloid Lineage Commitment Less is known about the transcription factors necessary for myeloid than for either erythroid or lymphoid lineage commitment. Myeloid cells arise from a common precursor that differentiates into monocytes or neutrophils. Monocytes can undergo a further differentiation step in tissues to become macrophages. Again, the analysis of tissue-specific promoters led to the identification of several transcription factors likely important in myeloid lineage commitment [ The macrophage CSF (M-CSF) receptor is largely restricted to monocyteslmacrophages [ Analysis of the proximal portion of the M-CSF promoter identified a functional binding site for the ets family member PU.l [151]. This site was critical for high levels of expression of the M-CSF promoter in transfection studies. PU. 1 was also found to be important in the myeloidspecific expression of the CDIIb promoter [152]. PU.l RNA transcripts have been detected via in situ hybridization primarily in immature myeloid cells and in macrophages [ Early granulocyte precursors expressed PU. 1 mrna transcripts, while mature erythroid and mature granulocytes did not. In support of a role for PU.l in the lineage divergence between B lymphocytes and myeloid cells was a recent finding of differences in nuclear PU. I expression and PU. 1 function between macrophage cell lines and 3 cell lines [154]. PU.l was found to repress the activity of the IgH enhancer, but to transactivate the M-CSF promoter in B cells. In addition, the levels of PU. 1 in nuclear extracts were noted to be much higher in macrophage than in B cell lines [154]. Further research is needed to reconcile this study with other studies that have suggested a positive role for PU.l in the regulation of B lymphocyte lineage-specific genes [91, In support of a role for PU.1 in both myeloid and lymphoid progenitors is recent gene targeting of the PU.1 locus. In PU.1-deficient mice there was a defect in the generation of lymphoid and myeloid progenitors [ Studies of the myeloperoxidase promoter have also identified several functioniil cis-elements, one of which binds a myeloid restricted nuclear factor termed MyNFl [157]. A cdna clone for this factor has not yet been reported. A gene that encodes a novel zinc finger protein termed MZFl was found in a screen of a cdna library prepared from a patient with chronic myelogenous leukemia [ MZFl has two independent DNA binding domains that recognize two distinct but similar nucleotide sequences [159]. Each target sequence has a G-rich core. The expression pattern of MZFl transcripts is consistent with a role in myeloid