Differential use of SCL/TAL-1 DNA-binding domain in developmental hematopoiesis

An important biologic process is the specification of tissue stem cells during development and their subsequent controlled differentiation to form the large number of specialized cell types. The molecular decisions that regulate these processes depend, in large part, on the establishment of specific transcriptional programs, mediated by the combined activities of ubiquitously expressed and tissue-specific transcription factors. Hematopoiesis provides a well-studied and accessible system to investigate transcriptional mechanisms involved in lineage specification and terminal differentiation. One of the critical hematopoietic-specific transcriptional regulators is the basic helix-loop-helix (bHLH) protein SCL.

SCL was discovered through its involvement in translocation events and is involved in up to 60% of cases in T-cell acute lymphoblastic leukemia.¹  In normal hematopoiesis, it plays key roles in both specification of progenitors and terminal erythroid differentiation. Loss-of-function studies have shown that SCL is required for specification of the first primitive (embryonic) hematopoietic cells from a yolk sac mesodermal precursor at the origin of both blood and endothelium, the hemangioblast.^2-4  As development proceeds, SCL is then required in the cellular processes leading to specification of definitive (adult) hematopoietic stem cells (HSCs) in the embryo proper.^5-8  Functions for SCL in the maintenance and commitment of mouse short-term and human nonobese diabetic–severe combined immunodeficient mouse repopulating HSCs have been proposed,^9-11  whereas SCL is not thought to be involved in mouse long-term HSC biology.^12,13  Separately, SCL is necessary for proper development and maturation of erythroid and megakaryocytic progenitors, normal differentiation of mast cells, as well as in neural development.^12,14-20 

SCL belongs to the family of class II tissue-specific bHLH proteins. Similarly to the prototype of this class, MyoD, it binds specific DNA sequences, E-box (CANNTG) motifs, in cis-regulatory elements as a heterodimer with the E-proteins.²¹  Moreover, it interacts with multiple protein partners to regulate transcription. In erythroid cells, where its function has been most studied, it can positively as well as negatively modulate transcription of target genes by recruiting cofactors (activators or repressors) and chromatin remodelling proteins (Schuh et al²²  and references therein).²³ 

DNA binding has long been considered as a prerequisite for the transcriptional activity of bHLH proteins, as exemplified by muscle-specific proteins of this family. The DNA-binding domain of MyoD is crucial in conferring myogenic properties to a nonmyogenic cell line.²⁴  Similarly, myogenin is unable to induce the myogenic program without DNA binding.²⁵  We suggested that this might not be strictly true for SCL because some of its functions during hematopoietic differentiation can be mediated by DNA-binding–independent mechanisms.¹⁵  However, there are important caveats with these findings. The assays used were based on rescue as opposed to loss-of-function experiments. Importantly, SCL expression was forced at unregulated levels in in vitro embryonic stem (ES) cell differentiation assays and in the zebrafish mutant embryo cloche. Moreover, the consensus suggests that, although ES cell–derived hematopoiesis faithfully recapitulates primitive, yolk sac hematopoiesis, in vitro–differentiated ES cells do not give rise to definitive, adult hematopoietic stem cells and their progeny.²⁶  Thus, to rigorously and systematically assess the DNA binding–dependent and -independent activities of SCL in developmental (primitive and definitive) hematopoiesis, an in vivo physiologic model was needed. We anticipated that this analysis would have broader implications for our understanding of how regulatory proteins achieve their functions.

To address this question, we have created a knockin mouse model expressing a mutant form of SCL that has no DNA-binding activity. In contrast to early lethality and complete absence of hematopoiesis in scl-null embryos, hematopoietic cells are specified in the DNA-binding homozygous mutant mice. Terminal erythroid cell maturation, however, is severely compromised and leads to lethality from embryonic day E14.5. This conclusively shows that SCL DNA-binding–independent activities are necessary and sufficient for hematopoietic stem/progenitor cell specification, whereas SCL direct DNA-binding activity is critically required in terminal erythroid maturation. We propose that the differential use of distinct functional domains is likely to be a general property of developmental regulators. In this context, dissection of the molecular mechanisms that regulatory proteins engage at different stages of development or in distinct cellular contexts should uncover novel transcriptional regulatory networks.

This study was conducted in compliance with the Home Office regulations at the University of Oxford. The targeting construct containing the mutations aimed at replacing the RER residues by alanines in SCL basic domain (details in Document S1, available on the Blood website; see the Supplemental Materials link at the top of the online article) was electroporated into mouse E14Tg2a ES cells maintained in BHK-21-Glasgow MEM (Invitrogen, Carlsbad, CA). On selection with G418 (0.3 mg/mL) and Gancyclovir (0.2 mM), resistant colonies were isolated and characterized by Southern blotting. To excise the floxed neomycin cassette, positive clones (scl^Wt/RERneo) were transiently transfected with a vector expressing a Cre recombinase/GFP fusion protein (gift from T. Enver, University of Oxford). Forty-eight hours after transfection, sorted GFP-positive cells were selected for G418 sensitivity. Excision of the neomycin cassette was further confirmed by Southern blotting. scl^Wt/RERΔneo ES cells with a normal karyotype were then injected into C57Bl/6 blastocysts and introduced into pseudo-pregnant C57Bl/6 females. Germ line transmission was obtained and heterozygotes were produced. The polymerase chain reaction (PCR)–based genotyping strategy relied on the presence of a PvuII restriction site created by the mutation introduced into the scl locus.

Heterozygous scl^WT/RERneo ES cells were grown in 1.8 mg/mL G418.²⁷  Clones were picked and homozygosity was characterized by Southern blot analysis. The neomycin cassette was excised as described for scl^WT/RERneo to generate scl^RER/RER ES cells. Clones were analyzed by Southern blotting and allele-specific PCR as mentioned above. For in vitro hematopoietic differentiation, wild-type E14Tg2a, scl^RER/RER, and scl^−/−5 ES cells were differentiated into embryoid bodies (EBs) in IMDM as described.²⁸  At day 3 of development, EBs were disaggregated by trypsin treatment, and cells were replated in 1% methylcellulose in VEGF-containing medium. Colonies (transitional, blast, and secondary EBs) were scored after 4 days of culture and individually expanded in liquid culture on Matrigel-coated wells in IMDM supplemented with cytokines that support the growth of both hematopoietic and endothelial lineages, as described.²⁸  Hematopoietic colony-forming potential of the nonadherent cells was assayed on replating in 1% methylcellulose.²⁸ 

Cells from yolk sac (day E8.5), fetal liver (days E12.5 and 14.5), and adult bone marrow (total or common myeloid progenitor [CMP] and megakaryocyte-erythroid progenitor [MEP] populations) were subject to progenitor assays. At least 5 independent experiments were performed for each assay. Replating conditions are described in Document S1.

CMPs and MEPs were isolated as described.^29,30 

Fetal liver cells from day E12.5 embryos were expanded for 3 days, and Ter119⁻ erythroid progenitors were purified.^22,31  This fraction, referred to as day 0 population, was further expanded or induced to differentiate toward the erythroid lineage.³¹ 

RNA was extracted from day 0 and day 2 fetal liver populations using the RNAeasy Micro RNA isolation kit (Qiagen, Valencia, CA) and cDNA was synthesized using the Sensiscript kit (Qiagen). Ready-made primer and probe mixes from Applied Biosystems (ABI; Foster City, CA) were used for c-kit, gpa, eklf, pb4.2, and alpha globin. Primer sequences for beta globin and gata1 are in the Document S1. Samples were analyzed in duplicates using an ABI Prism 7000 sequence detection system.

For chromatin immunoprecipitation (ChIP) experiments, primers and 5′-6-carboxyfluorescein-3′-6-carboxy tetramethylrhodamine–labeled probes were selected from unique sequences in the pb4.2, eklf, and gpa loci and appropriate external controls using Primer Express (see Document S1). Input and immunoprecipitated material were analyzed in duplicates relative to a sequence in the gapdh locus as previously described.³² 

ChIP assays were performed as described.²² 

Hematopoietic colonies were imaged on an inverted IX51 Olympus microscope (Tokyo, Japan) with a Jenoptik C14 camera (Jena, Germany). The objectives used were 10×/0.3 NA (numeric aperture), 20×/0.4 NA, and 4×/0.13 NA; see figure legends for details). MGG/benzidine-stained cytospins were imaged on a BX60 Olympus microscope with a QImaging camera (Surrey, ON); all images were taken through a 40×/0.5 NA objective. Embryos were imaged on an SMZ1500 Nikon microscope (Tokyo, Japan) with a Nikon digital camera Dxm1200F. The Openlab (version 3; Improvision, Coventry, United Kingdom) software was used for image acquisition and images were exported into Adobe Photoshop (version CS2; Adobe Systems, San Jose, CA) for processing.

To abolish SCL DNA-binding activity, point mutations were introduced into exon VI of the scl genomic locus by homologous recombination in ES cells (targeting construct Figure 1Bi). These mutations substituted 3 highly conserved DNA-binding residues³³  in SCL basic domain with alanine residues (amino acids RER at positions 195-197; Figure 1A). Southern blot analysis identified heterozygous scl^Wt/RERneo ES cell clones positive for homologous recombination (Figure 1Bii,C). On transient expression of a Cre recombinase, 2 ES subclones harboring the deletion of the floxed neomycin resistance gene (scl^Wt/RERΔneo; Figure 1Biii,C) were generated. Both clones had a normal karyotype (not shown), and the presence of the mutation in exon VI was confirmed by allele-specific digestion of PCR products (Figure 1D). One of the clones was then used for blastocyst injection, and mouse lines heterozygous for the SCL DNA-binding mutation (thereafter referred to as scl^Wt/RER) were generated.

Initial breeding produced scl^RER/RER embryos at several developmental stages as well as scl^RER/RER newborn mice (as shown in Tables 1 and 2, see next subsection). Before embarking on extensive breeding programs, we fully characterized the mutant SCL^RER protein to ensure that substitution of the 3 residues (RER) in SCL basic domain had specifically inhibited DNA-binding activity. Inability to bind DNA, unperturbed levels of expression, and nuclear localization were confirmed from tissues isolated from homozygous embryos or adult mice (Figure S1A-C).

As shown in Table 1, mating of heterozygous scl^Wt/RER mice only produced 5% of homozygous scl^RER/RER mice (as opposed to 25% for a mendelian ratio), suggesting embryonic lethality of approximately 80% of homozygotes. The surviving scl^RER/RER mice developed to adulthood. To document time of death, timed matings were set up between homozygous males and heterozygous females. scl^Wt/RER and scl^RER/RER embryos were observed at a mendelian frequency up to gestation day E12.5 (Table 2). Lethality was observed at day E14.5 and increased until day E17.5. Equivalent frequencies were observed in the progeny of heterozygous mice (data not shown).

In conclusion, we successfully generated a mouse model expressing a mutant form of SCL that was specifically impaired in its DNA-binding capacity. Contrary to the conventional SCL knockout embryos,^2,4  the mutant homozygous scl^RER/RER embryos survive at the yolk sac stage (day E9.5). The lethality observed at the fetal liver stage indicates a specific nonredundant requirement for SCL direct DNA-binding activity in definitive hematopoiesis.

In vitro experiments have suggested that the scl^−/− embryonic phenotype is due to failure of hematopoietic commitment from the hemangioblast.³  The fact that no lethality was observed at the yolk sac stage in the scl^RER/RER embryos therefore suggested that the hemangioblast was specified and that its hematopoietic potential was expressed. To confirm this hypothesis, we used the ES cell in vitro differentiation system that gives easy access to this early embryonic hematopoietic compartment. The in vitro equivalent of the hemangioblast, called blast colony-forming cell (or BL-CFC), develops in ES-derived EBs from day 2.5 to day 4 of differentiation.²⁸  We derived scl^RER/RER ES cells from heterozygous clones (Figure S1D) and submitted them to in vitro differentiation, in parallel with wild-type and scl-null ES cells.⁵  At day 3 of EB development, the cells were replated in conditions favoring hemangioblast development (see Figure 2A for the culture steps involved in the blast colony assay). Remarkably, colonies that developed from scl^RER/RER cultures were very similar in terms of number and morphology to those observed from wild-type cultures (Figure 2B right and left, C) and previously described as blast colonies (progeny of the BL-CFCs²⁸ ). In contrast, scl-null EB-derived cells mainly gave rise to large numbers of secondary EBs (Figure 2B middle, C). To characterize their hematopoietic and vascular potential, blast colonies were transferred onto matrigel-coated plates. Both wild-type and scl^RER/RER blast colonies developed into a typical layer of adherent (endothelial) cells surmounted by nonadherent (hematopoietic) cells (Figure 2D right and left). In contrast, the scl^−/− EB structures generated an adherent layer and a compact inner core of cells (Figure 2D middle). The hematopoietic nature of the nonadherent wild-type and scl^RER/RER cells was confirmed on replating in medium supplemented with hematopoietic growth factors (Figure 2E). In summary, the DNA-binding mutation did not affect SCL activity in blast colony-forming cells, and hematopoietic cells were specified from scl^RER/RER hemangioblasts.

These data are therefore the first in vivo indication in a mammalian system that the direct DNA-binding activity of SCL is dispensable for its earliest role in hematopoietic development.

The first hematopoietic cells to emerge in the embryo are primitive erythroid progenitors in the yolk sac blood islands at day E7.5, that mature progressively after entering the circulation.³⁴  To examine the effects of absence of SCL DNA-binding activity on primitive erythropoiesis, day E9.5 yolk sacs were dissected. Homozygous scl^RER/RER yolk sacs were pale compared with wild-type controls and heterozygotes (Figure 3Ai-iii). Blood examination showed mature basophilic erythroblasts in control embryos (Figure 3Aiv,v), whereas the presence of immature proerythroblasts in the scl^RER/RER blood suggested a block or a delay in maturation in mutant embryos (Figure 3Avi arrows). We then examined the differentiation potential of the erythroid progenitors in colony assays. Erythroid colonies derived from day E8.5 control yolk sacs showed the presence of mature, benzidine-positive, and enucleated erythrocytes (Figure 3Avii,viii). In contrast, mutant scl^RER/RER colonies contained immature proerythroblasts and fewer fully differentiated hemoglobinized red cells (Figure 3Aix arrows). Together, these data indicate a requirement for SCL DNA-binding activity in the maturation of primitive erythroid cells.

Concomitant with the decline of this transient, first wave of erythropoiesis, definitive hematopoietic cells develop autonomously in the embryo proper and, by day E11.5, have colonized the fetal liver.³⁵  At day E12.5, mutant scl^RER/RER embryos were pale compared with controls (Figure 3Bi-iii). Circulating blood mainly consisted of primitive erythrocytes, as expected at that developmental stage³⁴  (Figure 3Biv-vi). In agreement with yolk sac analyses, blood derived from homozygous scl^RER/RER embryos contained cells of an immature phenotype (Figure 3Bvi arrows). At day E14.5, when lethality is first observed (Table 2), the surviving mutant scl^RER/RER embryos were smaller and paler than their control littermates (Figure 3Bxiii-xv). At that stage, definitive erythrocytes have entered the bloodstream and outnumbered primitive red cells. MGG/benzidine staining of mutant blood films detected benzidine-negative cells and only few enucleated, benzidine-positive erythrocytes (compare Figure 3Bxviii with Figure 3Bxvi,xvii).

We then performed further analyses on day E12.5 embryos, an age when lethality is not observed (Table 2) but when the phenotype is expressed and compensation mechanisms (as shown by the incomplete penetrance of the phenotype; Table 1) are less likely to mask the effects of loss of SCL DNA-binding activity.

Immunophenotypic characterization of the fetal liver hematopoietic compartments showed significant differences between mutant homozygote and wild-type samples (Figure 3C). Decrease in the homozygous mutant CD71⁺/Ter119⁺ compartment showed late defects in erythroid maturation. The increase in the mutant early erythroid compartments (corresponding to BFU-E, CFU-E, and proerythroblast stages) denotes either an accumulation of immature progenitors and a corresponding decrease in more mature CD71⁺/Ter119⁺ cells because of defective erythroid maturation or a response to the fast turnover of later precursors. Regarding the other cellular populations, an expansion of the scl^RER/RER myeloid compartment was documented (see Document S1 for a quantitative analysis of the data).

The in vitro growth potential of E12.5 hematopoietic progenitors was then analyzed in colony assays. No difference in numbers of CFU-E, BFU-E, megakaryocyte, myeloid, and mixed colonies was observed between mutant and control samples (Figure S2A). On morphologic examination, mutant homozygous CFU-E colonies appeared poorly hemoglobinized (Figure 3Bvii-ix), contained an increased proportion of immature erythroblasts, and were devoid of terminally differentiated, benzidine-positive erythrocytes (Figure 3B compare ix and xii with vii,viii and x,xi).

Because SCL can have survival functions in red cells,³⁶  and to further characterize the affected erythroid populations, E12.5 fetal liver cells were subject to cell-cycle and apoptosis analyses. No difference in cell-cycle progression (G₁, S, G₂/M phases) was observed between wild-type and mutant populations (Figure 3D). Presence of very few events in the sub-G₁ population (Figure 3D) suggested low apoptotic rates in both mutant and wild-type samples. This was confirmed on analysis of fetal liver sections by transferase-mediated dUTP nick end labeling (TUNEL) assay. Very few apoptotic cells were detected in both scl^RER/RER and wild-type control samples, that represented 0.13% of total fetal liver cells (Figure 3E; data not shown).

In summary, SCL DNA-binding mutant homozygous embryos are anemic and show late defects in erythroid maturation that do not seem to be due to impaired proliferative or survival capacities. These defects result in most homozygous fetuses dying from anemia and show a critical in vivo function for SCL DNA-binding activity in embryonic/fetal erythropoiesis. Analysis of heterozygous embryos showed phenotypes very similar to those described for wild-type embryos, indicating that haploinsufficiency does not obviously affect SCL activity.

Surviving adult scl^RER/RER mice did not present with obvious morphologic or behavioral abnormalities and showed no increase in death rate. Examination of blood variables, however, showed mild microcytic hypochromic anemia (Table 3). The reticulocyte count was increased. Red cell anisocytosis was shown by decreased mean corpuscular volume (MCV) and an increase in red cell distribution width (RDW). The homozygous scl^RER/RER mice also showed decreased hemoglobin (HGB), hematocrit (HCT), and mean corpuscular hemoglobin (MCH). The rest of the peripheral blood count was normal (not shown). Examination of blood films confirmed the anemia in the scl^RER/RER mice. May-Grünwald-Giemsa (MGG) staining showed the presence of reticulocytes and poorly hemoglobinized target cells compared with control wild-type cells (Figure 4Ai,ii) and Brilliant Cresyl Blue (BCB) staining confirmed a 4-fold increase in the number of reticulocytes in blood derived from the mutant mice (Figure 4Aiii,iv and data not shown). In summary, these blood analyses indicate defective erythropoiesis in the adult scl^RER/RER mice.

To survey the adult hematopoietic compartments, we conducted an immunophenotypic analysis of bone marrow committed progenitors and hematopoietic cells. The proportion of common myeloid CMP and megakaryocyte-erythroid MEP progenitors was not affected in mutant scl^RER/RER mice (Figure S2B). In addition, and contrary to what we observed in fetal liver hematopoiesis, the proportions of scl^RER/RER hematopoietic populations were not affected (Figure 4B).

We then studied the potential of bone marrow hematopoietic progenitors in colony assays. The homozygous mutant compartments (BFU-E, CFU-E, megakaryocytic, and myeloid) were not reduced compared with controls (Figure S2C). Qualitatively, however, as observed in fetal liver replating assays, mutant CFU-E colonies were poorly hemoglobinized and, on cytospins, deficient in mature benzidine-positive erythrocytes compared with wild-type controls (Figure 4Av-viii).

Compensatory mechanisms triggered by anemia often include an increase in erythropoiesis in spleen. Splenomegaly was observed in scl^RER/RER mice (Figure 4C), indicative of an expanded extramedullary erythropoiesis. To confirm this, we analyzed the splenic erythroid compartment by flow cytometry based on expression of the cell surface markers CD71 and Ter119.³⁷  Four populations (I to IV) corresponding to stages of progressive erythroid maturation were defined (Figure 4D). Analysis of cells isolated from scl^RER/RER mice consistently showed an increased number of splenic red cells in regions II and III, reflecting perturbed maturation from early to late normoblasts. Morphologically, populations I (proerythroblasts) and II (early normoblasts) derived from scl^RER/RER mice were similar to those derived from control wild-type mice (Figure 4E). The scl^RER/RER mutant population III was composed of smaller, poorly hemoglobinized cells with abnormal membranes and devoid of enucleated erythrocytes (Figure 4E). Finally, the mutant population IV presented with microcytic red blood cells, in agreement with what has been observed in adult blood.

To further investigate the erythroid progenitor compartment in a stress situation, we induced anemia in the animals by phenylhydrazine injections or bleeding. Homozygote, heterozygote, and wild-type mice recovered similarly as assessed by the reestablishment of pretreatment hematocrit and reticulocyte count over a period of 17 days (Figure S2D; data not shown), thereby confirming the unaffected potentiality of the early splenic erythroid progenitor compartment in vivo.

We concluded that the increase in the early scl^RER/RER splenic erythroid compartments was probably to compensate for impaired terminal differentiation.

To characterize the molecular defects associated with SCL DNA-binding mutation in erythroid cells, we analyzed the expression levels of selected regulators of erythroid differentiation (known or potential SCL target genes) in material isolated from day E12.5 fetal livers. We used an in vitro differentiation system to expand c-kit⁺/CD71⁺/Ter119⁻ erythroid precursors (day 0) that subsequently underwent differentiation to obtain CD71⁺Ter119⁺ erythrocytes after 48 hours (day 2; Figure S3).

Analysis in immature erythroid populations (day 0; Figure 5A; data not shown) showed genes whose expression was either not affected in mutant scl^RER/RER cells (c-kit, gata1, gpa, alpha and beta globin, scl) or down-regulated (eklf and pb4.2). In more mature scl^RER/RER erythroid cells (day 2; Figure 5A; data not shown), expression of the genes was either unchanged (c-kit, alpha and beta globin) or significantly decreased (gata1, gpa, eklf, pb4.2, and scl) compared with wild-type controls.

As shown above, SCL DNA-binding activity is necessary to achieve maximal expression of scl in late (day 2) erythroid cells, thereby showing either an autoregulatory mechanism or secondary effects of SCL DNA-binding mutation. To distinguish between these 2 possibilities, we performed ChIP from material isolated from both early (Ter119⁻) and late (Ter119⁺) erythroid populations and analyzed recruitment of SCL to its own + 40 enhancer. This enhancer is active in fetal liver red cells, contains E-box motifs, binds SCL in an erythroid cell line,³⁸  and could therefore mediate transcriptional autoregulation. In scl^RER/RER Ter119⁻ cells, binding of SCL to + 40 enhancer was not affected, but it was impaired in more mature scl^RER/RER Ter119⁺ cells (Figure 5B). This decrease in SCL binding correlated with the changes in scl expression levels. We concluded that maximal expression of scl in the late stages of erythroid maturation is likely to rely on an autoregulatory mechanism requiring direct DNA binding.

To directly link perturbed gene expression with impaired SCL DNA-binding activity, we analyzed the in vivo interaction of SCL with regulatory regions of genes whose expression was dysregulated in scl^RER/RER cells. We performed ChIP from material isolated from Ter119⁻ (day 0) and Ter119⁺ erythroid populations and focused our analyses on pb4.2, eklf, and gpa genes. The proximal promoter regions of these genes contain E-boxes and Gata motifs. These sequences are bound in hematopoietic cell lines by SCL and protein partners (E2A, Lim-only 2 [LMO2] and LIM domain binding protein 1 [LDB-1]), as well as GATA1 or GATA2, that drive gene expression in functional assays.^39-41 

When assayed for SCL recruitment in Ter119⁻ cells, these promoters clearly showed different requirements for SCL DNA-binding activity in material isolated from wild-type versus mutant mice (Figure 6). In scl^RER/RER cells, we observed a complete loss of SCL binding on the pb4.2 promoter. In contrast, there were only statistically insignificant or partial decreases in SCL recruitment on eklf and gpa promoters, respectively. These differences in SCL binding correlated with the extent in PolII recruitment, levels of acetylation of histone H3 (a mark of increased chromatin accessibility), the extent of reduction in gene expression (fold changes in brackets; Figure 6), and the binding of 2 members of the pentameric complex, E2A and LDB1. Of note, we observed unaffected weak binding of GATA1 on all 3 loci (data not shown).

We then repeated the ChIP experiments from material isolated from Ter119⁺ cells. In these cells however, a decrease in SCL recruitment in mutant scl^RER/RER cells could not solely be attributed to the lack of DNA-binding activity but also to decreased SCL levels (Document S1; Figure S4A). In contrast with the observations in Ter119⁻ cells, significant decrease in recruitment of SCL, E2A, LDB-1, and PolII, as well as in histone acetylation, was observed on the promoter region of each locus in scl^RER/RER cells (Document S1; Figure S4B).

In summary, in early Ter119⁻ erythroid progenitors, we report a strict dependence on SCL direct DNA-binding activity for the transcriptional regulation of pb4.2 promoter and a requirement for maximal expression of eklf. Importantly, in mature Ter119⁺ cells, scl expression is subject to an autoregulatory mechanism leading to decreased levels of the SCL protein in scl^RER/RER cells.

In contrast to conventional scl knockout embryos, the striking observation that the DNA-binding mutant scl^RER/RER embryos survive at days E9.5 and 10.5 but die at later stages is the first in vivo demonstration of the differential use of SCL's mechanisms of action during development.

Remarkably, direct DNA-binding activity of SCL is not required for development of the first yolk sac hematopoietic cells because primitive erythroid cells, absent in scl^−/− embryos,²  were observed from scl^RER/RER yolk sacs. Substantiating these findings, hematopoietic cells develop from scl^RER/RER, but not scl-null, ES cell–derived hemangioblasts.

Previous loss-of-function studies have shown that SCL is required to establish definitive hematopoiesis (see “Introduction”). Moreover, analysis of conditional adult scl-null mice shows that SCL is absolutely necessary for the development of erythroid (BFU-E) and megakaryocyte progenitors (see “Introduction”). Our study shows that direct DNA binding is not required for the function of SCL in the establishment of definitive hematopoietic stem/progenitor cells in scl^RER/RER mice, because similar numbers of wild-type and mutant hematopoietic colonies were observed in progenitor assays from day E12.5 fetal livers.

Anemia was documented in scl^RER/RER embryos throughout development as well as in the surviving adult mutant mice. This was due to defective terminal erythroid maturation rather than to defective erythroid progenitor specification. These defects very probably resulted in the embryonic death observed from day E14.5.

As observed in days E8.5 and E9.5 embryos, SCL DNA-binding–dependent activity is required for terminal primitive erythroid cell maturation (this study and Juarez et al⁴² ).

At day E12.5, the maturation block in definitive erythropoiesis resulted in decreased numbers of mature CD71⁺/Ter119⁺ red cells and paucity of terminally differentiated erythrocytes. This therefore establishes that SCL's critical functions in fetal red cell maturation are mediated through direct DNA binding. No evidence of increased apoptosis or proliferation defects was observed that could have explained the decrease in erythroid precursors and the defect in maturation. Altogether, our data are in favor of a role for SCL in terminal maturation and hemoglobinization rather than control of survival or proliferation. Interestingly, expansion of the myeloid populations (CD61⁺, Mac-1⁺, Gr-1⁺; Figure 3C) suggests a possible role for SCL in repressing the macrophage/granulocyte program as myeloid precursors differentiate.

In our survey of gene expression, the gene most affected by the SCL mutation in fetal liver erythroid cells was that coding for protein band 4.2 (PB4.2). PB4.2 is a main component of the red blood cell membrane. The pb4.2 knockout mice have mild anemia.⁴³  The decrease in pb4.2 mRNA levels observed in SCL mutant erythroid cells is therefore likely to contribute to the anemia. Decreased levels of the erythroid-specific transcription factor eklf and, possibly, of its target genes might have also contributed to the phenotype of the scl^RER/RER mutant embryos.⁴⁴  Importantly, the requirement of SCL direct DNA-binding for activation of its own expression in maturing erythroid cells confirms the autoregulatory mechanism previously hypothesized,³⁸  and it suggests that one of the key functions of SCL DNA-binding activity might be to maintain its own expression in the terminal stages of red cell maturation. Finally, the erythroid phenotype may also result from the perturbation of expression of genes involved in other pathways important in erythropoiesis that remain to be characterized.

The long-term consequences of SCL deletion on adult erythropoiesis were recently analyzed in a conditional knockout model.¹⁶  Anemia, splenomegaly, and maturation defects in erythroid cell populations were reported. The phenotypic similarities between this model and ours suggest that most of SCL's functions in adult erythroid differentiation are mediated by its direct DNA-binding activity. Interestingly however, an expansion of the MEP compartment was documented in the scl-null model but was not observed in the scl^RER/RER mice suggestive of an SCL DNA-binding independent function in this compartment.

On the basis of the discussion above, we propose that SCL's critical functions in specification decisions are direct DNA binding–independent, whereas, for terminal maturation, its activity is critically mediated by direct DNA binding (Figure 7).

Why do different mechanisms of action of SCL predominate in different cellular contexts? The answer is likely to lie in the biologic function of the cells and the nature of the target genes/protein complexes.

We postulate that high DNA affinity mediated through direct binding ensures robust protein/DNA interactions in differentiating cells, probably to be required for tight regulation of expression of key target genes throughout maturation. This is exemplified by our data on the regulation of expression of pb4.2, important for terminally differentiating red cells. We show a strict dependence on SCL DNA-binding activity for its full expression, supported by the fact that 2 E-box motifs in pb4.2 regulatory sequences are functionally important.³⁹  At eklf and gpa promoters, appreciable SCL binding is detected in scl^RER/RER erythroid cells, suggesting that direct and indirect recruitment is also likely to occur in a dynamic process. Alternatively, it is possible that expression of only very few key genes requires SCL direct DNA-binding activity, such as pb4.2 in Ter119⁻ cells and scl itself in Ter119⁺ cells. The positive autoregulation of scl expression, by engaging the cells in a forward momentum,⁴⁵  could therefore be one of the key events driving terminal erythroid differentiation.

In cell fate decisions, at least 2 nonexclusive main mechanisms can account for SCL DNA-binding independent activities: (1) indirect recruitment to DNA through protein-protein interactions or (2) sequestration “off DNA” of other developmental regulators that mitigate against hematopoietic cell specification. In this context, note that functionally important residues are present at critical positions in loop domains of both SCL⁴⁶  and the sequestering non–DNA-binding HLH-only protein, Id1.⁴⁷  Regulation of expression of c-kit, a crucial hematopoietic cytokine, supports our first hypothesis. c-kit was previously shown to be regulated by SCL indirect DNA-binding activity,⁴⁸  and, in agreement with this, its expression was not affected in scl^RER/RER Ter119⁻ erythroid progenitors. Of course, one cannot exclude the possibility that direct DNA binding might also occur in specification decisions but that this activity is not critical for hematopoietic development.

In early hematopoietic development, the recent demonstration that SCL-LMO2 interaction is required^46,49  suggests that SCL is likely to nucleate the same multiprotein core complex in specification processes as it does in terminally differentiating erythroid cells.⁵⁰  Whether this complex is recruited to DNA through other DNA-binding transcription factor(s) to regulate gene expression or fulfils its function off DNA by sequestering other nuclear regulators or both remains to be determined.

Other transcriptional regulators exert DNA binding–independent mechanisms of action. However, in contrast to what we describe here for SCL, these mechanisms are either associated to direct DNA-binding functions (as for GATA1 and PU.1 mutually antagonistic activities in lineage decisions^51,52 ) or remain to be shown in a physiologic, loss-of-function model (as for the bHLH protein dHAND⁵³ ). To our knowledge, the glucocorticoid receptor (GR) is the only other example in which specific abolishment of DNA-binding function in an in vivo setting has unveiled distinct mechanisms of action. However, the GR DNA binding–independent activities are not required for lineage specification, but participate in the control of physiologic pathways and in tissue organization.⁵⁴ 

Thus, to date, SCL is the first example of a transcription factor exhibiting differential mechanisms of action in distinct developmental/cellular contexts, relying on DNA binding–independent functions in lineage specification. We propose that functional assessment of DNA binding, or other essential activities, of key proteins during development might show that differential use of functional domains is a property common to many transcriptional regulators.

We thank Robert Sumner (Biomedical Services, University of Oxford) for mouse blastocyst injection and chimera production and Ann Atzberger (MRC Molecular Hematology Unit) for FACS sorting. We also thank Boris Guyot, Eduardo Anguita, and Douglas Vernimmen for advice on real-time PCR and ChIP assays; Jackie Sharpe and Jackie Sloane-Stanley for invaluable help; and Lauren Aronson for technical assistance. We thank Doug Higgs and Bill Wood for critical reading of the manuscript, and Roger Patient and Tariq Enver for helpful discussions.

This work was supported by the MRC and a grant from the Leukaemia Research Fund (M.T.K.). M.T.K. was the recipient of a Karim Rida Said foundation scholarship in the early stages of this work.

Contribution: M.T.K. designed and performed research, analyzed the data, and wrote the paper; H.C. performed experiments; P.V. designed research and wrote the paper; and C.P. designed research, analyzed the data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Catherine Porcher, MRC Molecular Haematology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS, United Kingdom; e-mail: catherine.porcher@imm.ox.ac.uk.