The role and regulation of friend of GATA-1 (FOG-1) during blood development in the zebrafish

Transcription factor GATA-1 is the founding member of a small family of nuclear proteins that bind the DNA consensus sequences [(T/A)GATA(A/G)]¹  in a variety of tissues. GATA-1 was initially discovered as a nuclear protein that binds numerous GATA consensus motifs distributed throughout enhancers and promoters of erythroid-specific genes. Moreover, GATA-1 is a key regulator of erythropoiesis, as demonstrated by genetic studies in zebrafish and mice.^2-5  GATA-1 null mice showed complete ablation of primitive and definitive erythropoiesis resulting from arrested maturation^4,5  and apoptosis of erythroid cells.²  In addition, numerous gain- and loss-of-function experiments show that GATA-1 is important for the development of megakaryocytes,^6,7  eosinophils, and mast cells.^8-11  In addition to driving the maturation of several hematopoietic lineages, GATA-1 also inhibits the formation of alternate lineages by interacting with the myeloid transcription factor PU.1.¹¹  Specifically, studies in zebrafish and mice demonstrate that GATA-1 and PU.1 proteins antagonize each other during hematopoietic lineage specification.^12-15 

GATA-1 contains 2 zinc finger domains, which are highly conserved throughout vertebrate evolution. The C-terminal zinc finger is required for DNA binding, whereas the N-terminal finger stabilizes DNA binding and facilitates physical interaction with numerous proteins. Interactions between the GATA-1 N-finger and the multitype zinc finger protein, FOG-1 (Friend-of-GATA-1, zfpm1), appear to be particularly important.^16,17  GATA-1/FOG-1 complexes can function as activators for several erythroid and megakaryocytic genes and as repressors for others.¹⁸  In both mice and humans, GATA-1 missense mutations that disrupt FOG-1 interaction cause severe anemia and thrombocytopenia, partially recapitulating loss of GATA-1 phenotypes.¹⁷  In contrast to GATA-1–deficient mice, FOG-1 knockout animals show complete ablation of the megakaryocytic lineage,¹⁹  probably because GATA-2/FOG-1 complexes can partially compensate for loss of GATA-1.^7,18  Together these experiments demonstrate that FOG-1 acts as a critical cofactor for GATA-1 and GATA-2 in erythroid and megakaryocytic lineages.

Interestingly, FOG-1 antagonizes GATA-1 activities in other lineages. Thus, GATA-1–dependent eosinophil maturation is inhibited by FOG-1.⁸  A similar example has been reported for mast cell differentiation in which FOG-1 down-regulation is a prerequisite for mast cell development.^9,10  Ectopic expression of FOG-1 in committed mast cell progenitors reprograms them along erythroid/megakaryocytic lineages by repression of GATA-2, an essential GATA factor in mast and eosinophilic cells.^9,10  Together, these experiments suggest that erythroid and myeloid specification is achieved by activation and/or repression of FOG-1, which acts largely by modulating GATA factor activity through direct physical interaction.

FOG-1 is also expressed in nonhematopoietic tissues where it exerts GATA-1–independent functions, probably through interaction with other GATA family members. For example, in mice FOG-1 represses GATA-3 activity in naive T-helper cells during their development into T-helper 2 cells.²⁰  FOG-1 also plays a role in heart development, most probably through interactions with GATA factors 4, 5, or 6.^18,21-23  In zebrafish, the injection of an antisense morpholino directed against the homolog to murine FOG-1 resulted in embryos with a large pericardial effusion and a looping-deficient heart tube. This looping defect could be rescued by coinjection of cRNA encoding zebrafish or murine FOG-1.²⁴ 

Together, prior studies indicate that FOG-1 positively and negatively regulates the formation of hematopoietic and nonhematopoietic tissues through interactions with GATA proteins. However, the full extent of these interactions, the associated mechanisms, and their functional implications are unknown. Moreover, little is known about the regulation of FOG-1 expression, which must be exquisitely controlled to positively and negatively regulate GATA factors in different tissues. Using zebrafish genetics and molecular experiments, we show that zebrafish FOG-1 acts independently of GATA-1 and is required for T-cell development in the thymic organ. Our findings also reveal that loss of FOG-1 function significantly expands myeloid precursors at the expense of erythrocytes, similar to what occurs with loss of GATA-1 activity. Thus, FOG-1 is an essential component of GATA-1 activity during erythroid/myeloid cell fate determination. Finally, we used bioinformatics coupled with zebrafish transgenic methods to identify 2 conserved cis-regulatory motifs (CRM) in the FOG-1 locus, FE1 and FE2, which control its expression during hematopoiesis. The FE2 enhancer contains functional GATA-1 motifs, consistent with FOG-1 being a downstream target of GATA factors.

The nucleotide sequences reported in this paper have been submitted to GenBank with the accession numbers: AY515850 (FOG-1).

Wild-type (AB*, Tü) and germline transgenic zebrafish (Danio rerio) were kept and bred according to standard methods.²⁵  Maintenance of zebrafish transgenic lines is detailed in the supplemental data (available on the Blood website; see the Supplemental Materials link at the top of the online article). All studies described received full approval from the institutional animal care and use committee at Brigham & Women's Hospital.

A detailed description of the whole mount in situ hybridization (WISH) procedure and molecular probes is provided in the supplemental data.

To knock down the function of zebrafish FOG-1, we used morpholino (MO) oligomers.²⁶  Custom-synthesized MOs were obtained from Gene Tools, LLC. Approximately 0.05 to 0.08 pmol of MO was injected into embryos at the 1- to 2-cell stage. MOs targeting FOG-1 and control are listed in supplemental Table 1. The vlad tepes (vlt) genotyping was performed as previously described.³  Rescues assays were performed by coinjecting approximately 0.08 pmol of FOG-1MO with 100 pg of FOG-1 zebrafish cRNA.

MO against FOG-1 (∼ 0.08 pmol) was injected into fertilized eggs at the 1- to 2-cell stage from the transgenic zebrafish Tg(GATA-1:eGFP).²⁷  Approximately 100 to 200 embryos were collected from MO-injected and control MO-injected clutches. Disaggregated cells at 24 hpf were sequentially passed through 70-mm and 40-mm cell strainers, washed once in 0.9× phosphate-buffered saline buffer, and then pelleted by gentle centrifugation. The cells were resuspended in a final buffer containing 0.9× phosphate-buffered saline, 2% fetal bovine serum, and propidium iodide. Cells were sorted in a BD Biosciences FACSVantage SE machine, collected by cytospin centrifugation, and stained with Wright-Giemsa dye as described.¹⁴ 

We used the Gateway-compatible vectors (Invitrogen) to analyze gene function in transgenic zebrafish.²⁸  Descriptions of the molecular and transgenic techniques are detailed in the supplemental data.

The pCS2-TP plasmid, encoding the Tol2 transposase, was linearized, and 5′-capped Tol2 cRNA synthesized by in vitro transcription using SP6 RNA polymerase (Ambion). Embryos were coinjected at the 1-cell stage with 25 pg of Tol2 transposase cRNA and 50 pg of Tol2-flanked destination enhanced green fluorescence protein (eGFP) reporter plasmid.²⁸  The expression of the transgenic reporter eGFP was visualized at 24 hpf under a fluorescence microscope (Leica MZFLIII).

Mouse erythroleukemia (MEL) cells were initially grown in Dulbecco modified Eagle medium supplemented with 10% fetal calf serum, 2% penicillin/streptomycin, and 1% glutamine. MEL cells were induced to undergo erythroid maturation by the addition of dimethyl sulfoxide to final 1.7% (vol/vol) concentration for 24 hours before performing the chromatin immunoprecipitation (ChIP) assay. G1ER cells were grown as described²⁹  for ChIP analysis. ChIP assay was performed as previously described³⁰  with modifications. Protein-A DynaBeads (100.02D; Invitrogen) were preincubated with rabbit anti–rat IgG (H + L; Jackson ImmunoResearch) for 1 hour before incubation with the GATA-1 antibody (sc-265; Santa Cruz Biotechnology) for 3 hours. Chromatin from 4 × 10⁶ MEL cells was added and immunoprecipitated. The immunoprecipitated DNA was purified and analyzed by quantitative real-time polymerase chain reaction (PCR) using the QuantiTect SYBR Green PCR Kit (QIAGEN). A fragment 2 kb upstream of the GATA-1 HS1 site, which lacks the GATA-binding site, was used as an internal control, and fold of enrichment was calculated by the 2^ΔCt method. The primers used for PCR amplifications are listed in supplemental Table 2.

Pooled data were calculated as the mean plus or minus SD with the number of independent experiments indicated. Pairwise comparisons were performed by the Student t test and multiple comparisons by analysis of variance using the SAS 9.1.3 software (SAS Institute Inc; www.sas.com).

In zebrafish, the cells destined for hematopoiesis express GATA transcription factors in the lateral plate mesoderm (LPM). Subsequently, the LPM gives rise to the intermediate cell mass (ICM), the functional equivalent of the yolk sac blood islands in mammals.³¹  To assess the role of FOG-1 during zebrafish hematopoiesis, we used WISH to compare the expression patterns of FOG-1, GATA-1, and GATA-2 at early stages of embryogenesis. Zebrafish FOG-1 mRNA is strongly expressed as a maternal transcript during the 2-cell stage (Figure 1A) in a similar expression pattern to GATA-2 (Figure 1K). At the sphere stage, FOG-1 transcripts are observed along with zygotic GATA-2 mRNA (Figure 1B,L). At the 5-somite stage (5ss), the expression of GATA-1 is initiated in the LPM in a common region compared with FOG-1 and GATA-2 (Figure 1C,H,M). At the 15-somite stage (15ss), FOG-1 is coexpressed with those GATA-factors in the ICM (Figure 1D,I,N). At 24 hours postfertilization (hpf), FOG-1 and GATA-1 are strongly expressed in the ICM (Figure 1E,J), whereas GATA-2 levels are reduced in the same region (Figure 1O). We also detected expression of FOG-1 mRNA during organogenesis in the digestive tract. FOG-1 transcripts can be detected in the liver, hepatic duct, intestine, and pancreatic primordia (supplemental Figure 1A), similar to endodermal markers of the viscera, such as pdx1 (supplemental Figure 1B) and GATA-6 (supplemental Figure 1C). At 5 days postfertilization (dpf), FOG-1 is coexpressed with GATA-6 and pdx1, which delineate the liver, stomach, and intestine (supplemental Figure 1D-F). In the heart, FOG-1 transcripts are most intensely visible at the constriction of the atrial-ventricular boundary as well as nkx2.5 and GATA-6 (supplemental Figure 1G-I), consistent with previous observations.²⁴ 

Using zebrafish blood mutants defective at different stages of hematopoietic differentiation, we examined the genetic interactions and function of FOG-1. The zebrafish blood mutants involved in the early stages of hematopoiesis, such as cloche (clo, encoding a gene that specifies the formation of hematopoietic and vascular progenitors³² ) and vlad tepes (vlt, encoding a deficient form of GATA-1³ ), were evaluated for expression of FOG-1 in the ICM. In these mutants, the expression of FOG-1 is retained in neural structures of the head and in the heart but not in the ICM (Figure 1P-R). In contrast, the blood mutation frascati (frs, encoding a defect in the slc25a37 mitochondrial iron transporter³³ ), which affects erythroid maturation at a later stage than clo or vlt, displays normal levels of FOG-1 expression in the ICM (Figure 1S). Thus, the deficiency of erythropoietic progenitors in the ICM causes a drastic reduction in mRNA expression of FOG-1 during zebrafish hematopoiesis (Figure 1Q-R).

To test the role of FOG-1 in zebrafish early hematopoiesis, we depleted embryos of FOG-1 protein by injecting morpholinos²⁶  (hereafter referred to as FOG-1 MO). The FOG-1 MO was designed to target the splice-donor site of exon 4 to selectively interfere with FOG-1 mRNA processing. Using reverse transcription (RT)–PCR, we found that the injection of FOG-1 MO resulted in the production of an aberrantly spliced mRNA (supplemental Figure 2A; supplemental Table 1). Sequence analysis of the mis-spliced cDNA revealed that the variant results from exclusion of exon 5 (data no shown). As a result, a frame shift is generated that would lead to subsequent nonsense-mediated mRNA decay. Consistent with this prediction, we observed that the FOG-1 morphant had drastically reduced the steady-state FOG-1 mRNA compared with the control embryo (supplemental Figure 2B).

Embryos injected with FOG-1 MO show normal early hematopoiesis, as indicated by GATA-1 and scl expression in the erythroid progenitors of the LPM and ICM (supplemental Figure 3). In contrast, at 20 hpf, the expression of band3 was significantly reduced in FOG-1 morphants (supplemental Figure 3J). Interestingly, the overexpression of FOG-1 mRNA in wild-type embryos did not increase the number of band3⁺-erythroid cells (data not shown). This suggests that FOG-1 is not essential for specification of early erythroid progenitors but is required for maintenance or maturation of erythroid cells.

To quantify our data, the embryos were placed in 3 different categories (normal, reduced, and absent), depending on the level of hemoglobinization or expression for band3 and cd41 in erythrocytes and thrombocytes, respectively (Figure 2). The FOG-1 MO-injected embryos have a strong reduction of band3 expression in the ICM relative to wild-type embryos (Figure 2A-B,K). The loss of erythroid cells was verified by staining with o-dianisidine for hemoglobinized cells, which revealed either drastic reduction or complete absence of mature erythrocytes in FOG-1 morphants (Figure 2D-E,J). The specificity of the morphant phenotype was validated by injections with control MOs (supplemental Table 1), showing no anemia (data not shown). Furthermore, hemoglobinized cells (Figure 2J), band3 expression (Figure 2K), and cardiac morphant phenotype²⁴  were fully restored by coinjection of zebrafish FOG-1 cRNA in FOG-1 morphants (Figure 2C,F; supplemental Figure 4).

In zebrafish, thrombocytes are the hemostatic cellular equivalent of mammalian platelets.³⁴  To determine the effects of FOG-1 loss of function on thrombopoiesis at 4 dpf, we used a transgenic zebrafish line where the eGFP expression is under the control of the thrombocyte-specific cd41 promoter.³⁵  During zebrafish embryogenesis, the cd41 promoter directs the expression of eGFP reporter in hematopoietic stem cell in the dorsal aorta at 30 hpf. Previous studies demonstrate that these cd41⁺ hematopoietic stem cells exists only as transitory precursors and disappear by 2 dpf.³⁶  The Tg(cd41:eGFP) embryos injected with FOG-1 MO failed to generate eGFP⁺ mature thrombocytes by 4 dpf (Figure 2G-H,L). The effects were specific, as shown in the partial rescue of FOG-1 MO phenotype in embryos injected with FOG-1 cRNA (Figure 2H-I,L). The cd41 partial rescue could be explained by reduction of the exogenous cRNA activity after 4 dpf when the cd41 eGFP expression is first detected in zebrafish embryos.³⁵  Altogether, these results demonstrate that zebrafish FOG-1 is essential for erythroid and thrombocytic maturation.

In vertebrates, the thymic anlage forms from the pharyngeal endoderm.^37,38  The genes for lck and rag-1 are expressed in maturing T lymphocytes of the bilateral thymi in zebrafish.^37-40  The ikaros gene is expressed in all cells of the lymphoid lineage, including multipotential hematopoietic progenitors in the caudal hematopoietic tissue (CHT), a transient localization for definitive hematopoiesis in zebrafish.³¹  The effect of FOG-1 deficiency on lymphoid and thymic development was evaluated by WISH. The control and morphant embryos were placed in 3 different categories (normal, reduced, and absent), depending on the level of expression for lck, rag1, and ikaros in the thymi by WISH assay. Control embryos showed normal expression of lymphoid markers (Figure 3A,E,I arrow), whereas expression of those same markers in the FOG-1 morphant embryos was generally either reduced or absent (Figure 3B-D,F-H,J-L). In contrast, the morphant embryos showed a less severe effect on ikaros expression in the ventral CHT (Figure 3M-N insets). This result suggests that FOG-1 participates in T-lymphocyte maturation, but not during early lymphoid progenitor specification in the CHT. Previous studies in mouse have shown the endodermal requirement of pax9 expression during thymus development.⁴¹  We did not detect significant differences in the level of pax9 expression between control and FOG-1 morphant embryos (Figure 3O-Q). However, the patterning of the endodermal pouches was abnormal in FOG-1 morphants. This suggests that the initial phase of the endoderm specification does not require FOG-1 but that FOG-1 alone or in conjunction with a partner is required for proper endodermal patterning (Figure 3O-Q). We injected MOs into transgenic embryos that express eGFP regulated by the T-lymphocyte–specific lck promoter [Tg(lck:eGFP)].³⁸  The transgenic embryos injected with FOG-1 MO were sorted by the absence of eGFP⁺-T lymphocytes in the thymi, which indirectly reflects a defect in thymic development. To test whether endoderm-derived structures are normally developed in FOG-1 morphants, we used Alcian blue staining of cartilage. The morphant embryos (eGFP⁻) show gross hypoplasia of the pharyngeal arches (Figure 3S) compared with eGFP⁺ wild-type siblings (Figure 3R). This defect was complemented by coinjection with FOG-1 cRNA (supplemental Figure 4C). In summary, these data suggest that the deficiency of FOG-1 disturbs the formation of the later stages of endoderm-derived pharyngeal structures, such as pharyngeal arches and thymic anlage.

Given the close interrelationship between GATA-1 and FOG-1 during erythroid and megakaryocytic development, we evaluated whether FOG-1 acts in the context of GATA-1 interactions to specify T-cell progenitors during thymus formation. At 5 dpf, Tg(lck:eGFP) embryos injected with FOG-1 MO were sorted by phenotype (Figure 4A-C) and eGFP expression in the thymi (Figure 4G-L). Furthermore, o-dianisidine staining was used to assess erythropoiesis in the morphants (Figure 4D-F white arrows). Transgenic embryos injected with either GATA-1 MO or FOG-1 MO are profoundly anemic (Figure 4D-F). Knockdown of FOG-1 reduced eGFP expression in the thymi compared with the controls (Figure 4). To exclude a nonspecific MO effect, we coinjected the anti-p53 MO simultaneously with FOG-1 MO. The results demonstrate that the phenotypic alterations in zebrafish FOG-1 morphant embryos are specific and not caused by generalized apoptosis mediated by p53⁴²  (supplemental Figure 5). No differences in eGFP⁺-T lymphocytes were observed between control embryos and morphants injected with GATA-1 MO, indicating that GATA-1 is dispensable for FOG-1 activity during zebrafish thymic development (Figure 4H-K).

Zebrafish deficient in GATA-1 (vlt) shows expansion of myeloid cells. These data were interpreted to suggest that GATA-1 inhibits myelopoiesis by antagonizing PU.1 activity in hematopoietic precursors in the ICM.^14,15  Because FOG-1 is a critical cotranscription factor of GATA-1 in formation of the erythroid/megakaryocytic cell lineage, it may also play an essential role during myeloid cell fate determination.

To investigate the role of FOG-1 in zebrafish myelopoiesis, embryos injected with FOG-1 MO were processed by in situ hybridization to reveal the expression of PU.1 in myeloid progenitor cells and mpo in myelomonocytes. Quantification of the FOG-1 morphants indicated a significant increase in the number of myeloid cells expressing PU.1 (Figure 5A-C) and mpo (Figure 5D-F). Double in situ hybridization of FOG-1 morphants showed a reduction in the number of erythroid band3⁺ cells (Figure 5G-H) with concomitant increase of granulocyte-specific mpo⁺ cells (Figure 5I), suggesting that loss of FOG-1 redirects erythroid progenitor cells into myeloid precursors. We also tested whether the mpo expansion in GATA-1 null animals (vlt)^14,15  could be further increased by knockdown of FOG-1 function in vlt embryos. Injection of FOG-1 MO in vlt mutants produced even greater numbers of mpo-expressing cells in the common cardinal vein compared with vlt uninjected animals (Figure 5J-M). These results suggest that FOG-1 could act independently of GATA-1 in myeloid cell fate commitment.

To rigorously quantify the extent of the erythroid/myeloid switch in FOG-1 null embryos, we analyzed hematopoietic cells by flow cytometry⁴³  using the Tg(GATA-1:eGFP) transgenic line, which expresses the eGFP reporter in erythroid and myeloid progenitors of the rostral ICM.^14,15  The transgenic zebrafish injected with FOG-1 MO showed expansion of the anterior eGFP⁺-myeloid cells at the expense of the posterior eGFP⁺-erythroid cells (Figure 6A-D). Fluorescence-activated sorting of eGFP⁺ cells in uninjected control embryos showed that approximately 71% were erythrocytes located in a low forward scatter fraction (Figure 6E,G). Moreover, approximately 11% were myelomonocytic cells located in the high forward scatter population. Examination of eGFP⁺ cells in FOG-1 knockdown animals showed a significant reduction of erythrocytes together with an increase in myelomonocytes compared with controls (Figure 6F,H). The increase in the number of myelomonocytic cells provides further evidence that loss of FOG-1 preferentially shifts the hematopoietic progenitors to a myeloid fate in zebrafish. In summary, our findings demonstrate that loss of FOG-1 function in the presence of normal GATA-1 levels suppresses erythroid commitment and results in a switch to the myeloid cell fate commitment (Figures 5–6).

To gain further insight into the regulation of the FOG-1 genes during vertebrate development, we assessed their conservation and divergence through extensive sequence comparison over coding and noncoding domains for predicted transcriptional enhancer elements.⁴⁴  Using an in silico bioinformatics strategy, we identified FOG-1 CRM necessary for expression in the erythroid compartment. This analysis combines 2 parameters: (1) a positive regulatory potential (RP), which detects similarity to patterns in alignments distinctive for regulatory regions; and (2) conservation of binding site motifs for the essential erythroid transcription factor GATA-1.^45,46  Using this algorithm, 17 high-RP and GATA-1–binding sites (hiRP and GATA-1_BS) in the FOG-1 locus with shared homology in zebrafish, mouse, and human were identified (Figure 7A). Rather than testing each of the 17 hiRP and GATA-1_BS individually, extreme evolutionary sequence conservation was used as a filter to select for candidate regions with a high likelihood of enhancer activity.⁴⁴ 

Several evolutionarily conserved regions (ECRs) were found within the FOG-1 locus. These modules were plotted with the ECR browser server (http://ecrbrowser.dcode.org/) and compared with the relative position of mouse FOG-1 on chromosome 8 (Figure 7A). In intron 1 of the FOG-1 gene, we identified 4 candidate regulatory elements showing high nucleotide identity among vertebrate species. In Figure 7A, the exons are represented by the blue boxes with transcription directionally oriented left to right. The “hiRP” score is the log-likelihood measurement estimates of the probability that the sequence is involved in regulating expression. Matches to weighted matrices for potential GATA-1–binding sites were identified in the mouse sequences (track labeled “hiRP&GATA-1_BS,” indicated by black boxes). FE1 contains one of the largest ultra-conserved elements: approximately 250 base pairs with almost complete identity (∼ 90% nucleotide identity) among zebrafish, Xenopus, chicken, mouse, and human. For FE2, only chicken, mouse, and human showed higher conservation, whereas the zebrafish sequence is only partially conserved (∼ 50% nucleotide identity).

As shown in Figure 7A, the 4 putative GATA-1–binding sites FE1-FE4 have hiRP and GATA-1_BS, which led to the prediction of functional occupancy of these 2 sites by GATA-1 in regulating FOG-1 expression during erythropoiesis. ChIP assays using chromatin from Friend MEL cells were performed to test for in vivo occupancy by GATA-1 at these 4 putative sites (FE1-FE4). As shown in Figure 7B, FE2 is an in vivo binding site for GATA-1. We also tested whether FE1 or FE2 is a target of GATA-2. Considering that MEL cells express very low levels of GATA-2, we performed the ChIP assays using chromatin from uninduced and induced G1ER cells.²⁹  The results showed that none of the enhancers (FE1-FE2) was bound by GATA-2 (data not shown).

The FE1 to FE4 regions of the FOG-1 locus were individually cloned using the Gateway system into a Tol2 transposon-based vector. The final constructs contain the individual putative enhancers with a minimal TATA-box promoter driving an eGFP reporter²⁸  (supplemental Figure 7). Injection of either the mouse (Figure 7F-G) or zebrafish FE1 (Figure 7H-I) gives rise to transgenic zebrafish expressing eGFP in the ICM, mimicking the endogenous expression of FOG-1 mRNA (Figure 7C). In zebrafish, FOG-1 is also expressed in the developing heart from 24 through 48 hpf.²⁴  Analogous to the endogenous FOG-1 mRNA, FE1 directs expression of eGFP in the heart during early cardiac development and later during heart morphogenesis (supplemental Figure 6). None of the transgenic embryos carrying FE1 directed reporter activity to Rohon-Beard neurons (Figure 7F-I). In contrast, transgenic embryos carrying FE2 showed eGFP expression in both the ICM and Rohon-Beard (Figure 7J-K) in a manner comparable with FOG-1 mRNA expression (Figure 7C,E). Like FE1, FE2 directs expression of the eGFP reporter in the developing heart (supplemental Figure 6), suggesting that both enhancers play a role during early heart formation. No consistent or significant expression of the eGFP reporter was seen in any of the zebrafish transgenic embryos similarly injected with FE3 or FE4 (supplemental Table 3).

We demonstrate that FOG-1 is expressed in early erythroid progenitor cells in a pattern similar to that of GATA-1 and GATA-2 during zebrafish hematopoiesis. Our finding in FOG-1 knocked down embryos indicates that GATA-1 expression is independent of FOG-1 activity (supplemental Figure 3). Moreover, in the absence of GATA-1, FOG-1 expression is drastically reduced, suggesting that FOG-1 functions genetically downstream of GATA-1 (Figure 1R). Our data demonstrate that FOG-1 activity is necessary for erythroid cell maturation (Figure 2) but is not sufficient to expand the number of erythroid cells in the embryonic ICM (data not shown). These findings are generally consistent with the mouse model, in which the absence of either FOG-1 or GATA-1 perturbs erythroid/megakaryocytic cell formation.^6,7,19 

Relying on compelling evidence from the mouse model, FOG-1 is recognized as a cotranscriptional factor for GATA-1 during erythrocyte as well as megakaryocyte development.^6,7,19  However, the role of FOG-1 in differentiation of the other blood cell lineages remains unclear. In this study, we evaluated the effect of zebrafish FOG-1 knockdown during myeloid and lymphocyte development. Our in vivo analysis in zebrafish shows that FOG-1 activity is required for the expression of lymphoid markers in the thymi; however, the expression of ikaros in hematopoietic/lymphoid progenitors in the CHT region remains unaffected (Figure 3). This finding suggests that FOG-1 is specifically required for expression of T-lymphoid genes in the thymi but not during formation of hematopoietic/lymphoid ikaros⁺ progenitors in the CHT.³¹  Based on pax9a expression in the pharyngeal region of FOG-1 morphant embryos, we conclude that early endodermal specification is normal whereas patterning is abnormal (Figure 3O-Q). In zebrafish and mouse, there is evidence showing that FOG-1 requires GATA-4/-5/-6 during proper cardiac and endoderm development.^21-24  Although FOG/GATA interactions have been detected in zebrafish and mouse embryos, there has hitherto been no evidence for the role of FOG-1 during thymus development. Our results indicate that in zebrafish FOG-1 is required for proper patterning of endoderm, leading to disruption of normal arch architecture and defects in thymic and T-cell development. We also demonstrate that these phenotypic alterations are not caused by apoptosis mediated by p53 (supplemental Figure 5). Future transplantation experiments using FOG-1–deficient cells into a wild-type background would definitively help resolve the question of whether FOG-1 acts in a cell-autonomous or non–cell-autonomous manner for thymic anlage formation and residence of mature T lymphocytes.

From overexpression studies, it is clear that GATA-1 and PU.1 are able to specify erythroid and myeloid cell fate.^12-15  It is well documented that both GATA-1 and PU.1 cross-antagonize each other's activity.¹¹  However, the precise mechanism of how GATA-1 and PU.1 initiate the transcriptional network that determines the specification of erythroid/myeloid cells remains unknown. In the case of GATA-1/FOG-1 interactions, the situation is especially complicated by the establishment of different activating and repressive transcriptional complexes depending on the cellular context.^11,18  Given the cooperative interaction between FOG-1 and GATA-1,¹⁸  we found that the inhibition of FOG-1 function with MOs resulted in expansion of myeloblasts at the expense of erythroblasts (Figures 5–6). On the other hand, knockdown of FOG-1 in zebrafish GATA-1 mutants (vlt) increases the number of myeloid precursor cells, suggesting that FOG-1 itself could regulate myeloid specification (Figure 5J-M). Supportive of our observations, a recent ChiP analysis to functionally interrogate predicted GATA-binding motifs in the PU.1 locus demonstrated that FOG-1 occupies the PU.1 gene in the same regions bound by GATA factors.^47,48  Both GATA-1 and GATA-2 recruit FOG-1 to the PU.1 transcriptional regulatory elements. Interestingly, Chou et al also found that FOG-1 and its associated NuRD component⁴⁷  bound to an upstream regulatory enhancer called URE⁴⁹  at −14 kb, where they did not detect GATA factors. Thus, FOG-1 probably regulates PU.1 by GATA-dependent and -independent mechanisms.⁴⁷  Collectively, our data extend well-established observations indicating that the ability of GATA-1 to inhibit myelopoiesis requires FOG-1.

To determinate whether GATA-1 regulates FOG-1 in vivo, we used bioinformatics to predict whether functionally important GATA-binding motifs are present in the FOG-1 gene.^44-46  Using this model together with comparative genomic tools from the ECR browser (http://ecrbrowser.dcode.org/), we identified 4 predicted GATA-1-binding CRMs within the FOG-1 locus. The 4 CRMs (FE1-FE4) were conserved in evolution and contain canonical GATA-1 DNA-binding sites T/A(GATA)A/G.¹  Two of these regions, FE1 and FE2, individually drive the proper temporal and spatial expression of an eGFP reporter in hematopoietic organs in transgenic zebrafish (Figure 7). Using ChIP assays, we demonstrate that only FE2 is bound by GATA-1 (Figure 7B). We propose that GATA-1 binds to the FE2 enhancers to activate FOG-1 expression during hematopoiesis. As previously discussed, the mutual antagonism between GATA-1 and PU.1 during lineage specification^12-15  may be enhanced by the participation of FOG-1. In this context, the mutual activation of GATA-1 and FOG-1 expression would establish a sustained autoregulatory circuit and promote the lineage specification from myeloid to erythroid differentiation. We found that GATA-1 binds to FE2 but not FE1 (Figure 7B). This observation indicates that other transcription factors, besides GATA-1, could be involved in the activation of FE1 during hematopoiesis. Such considerations are pertinent in the context of GATA-2/FOG-1 interactions, which may be either positive or negative depending on the presence or use of additional partner protein.^11,18  Two recent studies proposed that GATA-2 acts to repress FOG-1 expression during mast cell development.^9,10  However, we did not find significant binding of GATA-2 on FE1-2 in G1E cells (data not shown). Thus, it is possible that GATA-2 regulates other enhancer elements within the FOG-1 promoter in mast cells. This study integrates results from 3 types of analyses and shows that the data from bioinformatics or ChIP binding do not reveal specific biologic functions. Our analysis of the in vivo transcriptional activity of the FE candidates using zebrafish transgenesis serves as a robust functional assay.

Although FOG-1 has been shown to participate in erythroid and megakaryocytic maturation in mice, its role in myeloid and T-lymphoid development in vivo remains completely unknown. Our study provides the first evidence that FOG-1 regulates cell fate determination between erythroid and myeloid development, as well as T-lymphoid maturation during thymic development in the zebrafish. We advance well-established observations indicating that GATA-1 requires FOG-1 to regulate myelopoiesis. Our strategy, combining bioinformatics algorithms with Tol2-Gateway cloning and transgenesis in zebrafish, greatly facilitates the rapid, functional analysis of conserved vertebrate regulatory elements that regulate tissue-specific expression. Here, we identified and characterized mouse and zebrafish FOG-1 transcriptionally active enhancers during hematopoiesis. We showed that the enhancers contain critical GATA motifs, consistent with FOG-1 being a direct target of GATA factors.

The authors thank members of the Paw laboratory and Mitchell Weiss for critical review of the manuscript; Xiao Ming Shang and Matthew Keefe for technical assistance; Christian Lawrence, Jason Best, and the crew of the zebrafish facility for the animal husbandry; Shuo Lin for providing the Tg(GATA-1:eGFP) transgenic zebrafish line; and Nathan Lawson for the pCS2-Tol2 cDNA.

This work was supported in part by grants from the National Institutes of Health (B.H.P., A.B.C., N.S.T., D.T., J.P.K.), William Randolph Hearst Foundation (B.H.P.), March of Dimes Foundation (B.H.P.), American Society of Hematology (A.B.C., G.C.S.), American Heart Association (J.D.C.), and Swiss National Science Foundation (G.E.A.).

National Institutes of Health

Contribution: J.D.A., G.E.A., J.J.C., J.D.C., D.M., N.B.L., E.S., G.C.S., W.H., N.S.T., A.J.D., B.A.B., Y.Z., S.A.W., D.T., M.Y., T.B.M., G.K., K.H., J.P.K., D.I.S., and B.H.P. designed experiments, performed research, and analyzed data; J.D.A, G.E.A., and B.H.P. wrote the manuscript; H.F.L. and R.I.H. provided reagents; and J.P.K., D.T., N.S.T., A.J.D., A.B.C., and B.H.P. designed experiments and edited the manuscript.

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

Current affiliations are as follows: Johnson & Johnson Pharmaceutical, Zürich (G.E.A); Baystate Medical Center, Tufts University (J.J.C); Columbia University, College of Physicians & Surgeons (N.B.L.); Pall Corporation (E.S.); University of Virginia (G.C.S); Huntsman Cancer Institute, University of Utah (W.H., N.S.T.); Massachusetts General Hospital (A.J.D.); Harvard Medical School (S.A.W.); University of California–San Diego (D.T.); University of Rhode Island (G.K.); and Merck Research Labs (K.H).

Correspondence: Barry H. Paw, Brigham & Women's Hospital, Hematology Division, 1 Blackfan Cir, Karp Bldg 06.213, Boston, MA 02115; e-mail: bpaw@rics.bwh.harvard.edu.