|Home | About | Journals | Submit | Contact Us | Français|
Growth factor independence-1 (GFI1) and GFI1B are closely related, yet differentially expressed transcriptional repressors with nearly identical DNA binding domains. GFI1 is upregulated in the earliest thymocyte precursors, while GFI1B expression is restricted to T lymphopoiesis stages coincident with activation. Transgenic expression of GFI1 potentiates T-cell activation, while forced GFI1B expression decreases activation. Both mice and humans with mutant Gfi1 display lymphoid abnormalities. Here we describe autoregulation of Gfi1 in primary mouse thymocytes and a human T-cell line. GFI1 binding to cis-element sequences conserved between rat, mouse and human Gfi1 mediates direct and potent transcriptional repression. In addition, dramatic regulation of Gfi1 can also be mediated by GFI1B. These data provide the first example of a gene directly targeted by GFI1 and GFI1B. Moreover, they support a role for auto- and trans-regulation of Gfi1 by GFI1 and GFI1B in maintaining the normal expression patterns of Gfi1, and suggest that GFI1B may indirectly affect T-cell activation through repression of Gfi1.
Growth factor independence-1 (Gfi1) and Gfi1B are two closely related oncogenes that play different but pivotal roles in hematopoiesis. Gfi1-deficient mice display both thymic and peripheral T lymphopenia, with severe abnormalities in pre-T-cell development (1–3). Furthermore, mutation of Gfi1 in humans induces defects in T lymphocyte function and development (4). Gfi1B is necessary for the development of megakaryocytes and for definitive erythropoiesis (5). Gfi1B deficiency is embryonic lethal by day E15 (5). Subsequently, no thymic or T-cell phenotype has been reported for Gfi1B deficiency; however, forced expression of GFI1B induces perturbed T-cell development and function (6). In accordance with the distinct hematopoietic phenotypes of Gfi1 and Gfi1B deficiency, one or the other factor is predominantly expressed in hematopoietic tissues of normal adult animals (7). Gfi1 is highly expressed in thymus, the site of T-cell development, while Gfi1B is the predominant factor expressed in spleen. Both factors are expressed in bone marrow (7).
The distinct physiological functions of Gfi1 and Gfi1B may seem somewhat surprising given the similarities of their defined biochemical functions (7,8). GFI1 and GFI1B are two members of a family of zinc-finger transcriptional repressors that are characterized by the presence of the SNAG (found in the Snail and GFI1 family of proteins) repression domain (8). Transcriptional repression by both GFI1 and GFI1B requires an intact SNAG domain (8). Specifically, transient transcription assays demonstrate that mutation of proline to alanine at amino acid 2 (P2A) ablates transcriptional repression of GFI1 and GFI1B (8). Furthermore, having nearly identical zinc-finger DNA binding domains, GFI1 and GFI1B bind the same consensus DNA sequence (7,9). Despite the extensive similarities of these proteins, each contains a region unique in amino acid sequence, the activities of which are yet to be defined. We therefore expect that these two factors have both redundant and unique biological roles in blood cell development.
GFI1 is expressed throughout T-cell development, and transgenic expression of GFI1 in T cells causes an increase in the response to activation mediated through the T-cell receptor (TCR) (6,10,11) or interleukin 2 (IL-2) (6). GFI1B is induced at stages of T lymphopoiesis that are coincident with TCR activation, and forced transgenic expression of GFI1B in T cells leads to decreased T-cell activation in both mature T cells in vitro and in thymocytes in vivo (6). While GFI1B transgenic mice display normal thymic cellularity, they share with Gfi1-deficient mice the characteristic of peripheral T lymphopenia. In addition, GFI1B transgenics display thymic abnormalities, including a defect in the formation of CD8 SP cells that can be partially corrected by simultaneous transgenic expression of GFI1 (6). Therefore, some of the observed GFI1B-mediated effects on T lymphopoiesis may result from the disruption of normal functions of GFI1 (6).
The antagonistic roles of GFI1 and GFI1B in T-cell biology might have been predicted from the frequency with which these two factors are activated in Moloney Murine Leukemia Virus (MoMLV)-induced T-cell tumors. Gfi1 is the second most frequently activated target of MoMLV insertion mutagenesis in T-cell malignancies, leading to overexpression of wild-type Gfi1; however, Gfi1B is rarely activated (12). Because overexpression of Gfi1 is tumorigenic, we decided to examine the normal regulation of the Gfi1locus. Here we show that homologous regions of the rat, mouse and human Gfi1 loci contain GFI1 recognition sequences that mediate autoregulation in both primary and transformed T lymphocytes. Moreover, such binding sites may also be targeted by GFI1B to repress Gfi1 expression directly.
The GFI1 and GFI1B transgenic mice were as described previously (6). All mice were housed in the Donald Baxter Barrier Facility at the University of Louisville and used in accordance with protocols approved by the university’s Institutional Animal Care and Use Committee.
Chloramphenicol acetyl transferase (CAT) reporter constructs and GFI1 and GFI1B expression plasmids used in transient transcription assays were as described previously (8). The luciferase reporters were generated by cloning the rat Gfi1 promoter and intron from the CAT reporter into pCBG99-basic (Promega, Madison, WI). Site-directed mutagenesis was performed with Quickchange II (Stratagene, Valencia, CA). Plasmids for in vitro transcription and translation were generated by cloning rat Gfi1 or mouse Gfi1B into the pcDNA3 vector (Invitrogen, Carlsbad, CA). Retroviral vector expression constructs were generated by cloning triple Flag-epitope-tagged GFI1 into the MIEV vector (13).
Jurkat T cells were maintained in RPMI 1640 with 10% fetal bovine serum (FBS), 1% l-Gln and 1% Pen/Strep (all from Invitrogen). For stable transfections, Jurkat cells were electroporated as described previously (8). The transgene constructs encoding GFI1 or GFI1B were cotransfected with empty pcDNA3.1 vector DNA at a ratio of 22:1. Transfected cells were selected in the presence of 1 mg/ml Geneticin (Invitrogen) and cloned by limiting dilution.
Transient transfections of Jurkat cells were performed with DMRIE-C (Invitrogen) according to the manufacturer’s protocol. Briefly, 4 × 105 cells per well were transfected in 24-well plates. Six wells were transfected for each condition, and two wells of transfected cells were combined for a single data point. Therefore, luciferase assays were performed in triplicate. Click Beetle Luciferase assays (Promega) were performed according to the manufacturer’s protocol.
293T cells were maintained in Dulbecco’s modified Eagle’s medium with 10% FBS, 1% l-Gln and 1% Pen/Strep (Invitrogen). Cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol with Opti-MEM low serum medium (Invitrogen) and 2.5 × 105 cells/well in 24-well plates that were pre-coated with poly-l-Lysine (0.1 mg/ml; Sigma, St Louis, MO). Transfections were performed in triplicate. CAT assays were performed as described previously (8).
For transduction experiments, Phoenix cells (14) were transiently transfected with retroviral constructs by means of CaPO4 (15), then co-cultured with Jurkat cells overnight. GFP+ cells were analyzed 4 days later, sorted on a FACSVantage (Beckton Dickinson) and then cloned by limiting dilution.
For northern blots, total RNA was extracted with Ultraspec RNA Isolation Solution (Biotecx Laboratories, Inc., Houston, TX) according to the manufacturer’s protocol, and poly-A+ RNA was obtained using the Oligotex Direct mRNA purification protocol (Qiagen Inc., Valencia, CA). Total RNA (20 µg) or Poly A+ RNA (5 µg) was electrophoresed in a 1% agarose-formaldehyde gel and transferred to MagnaGraph nylon membrane (Micron Separations, Inc., Westboro, MA). Membranes were probed in UltraHyb solution (Ambion, Austin, TX) according to the manufacturer’s protocol. Radioactive probes were generated by Prime-a-Gene random priming kit (Promega).
Western blot detection of GFI1 and GFI1B was performed as described previously (6). Nuclear extracts of U937 cells were made with NE-PER (Pierce, Rockford, IL) according to manufacturer’s instruction, and protein concentration was determined using BCA Protein Assay Reagent (Pierce). Primary antibody was a commercially available antiserum to the last 20 amino acids of GFI1 (sc-6357; Santa Cruz Biotechnology Inc., Santa Cruz, CA), anti-Flag-HRP (Sigma) or our mouse monoclonal specific for GFI1 (2.5D.17).
For resolution of the rat locus, primer extension was performed with an end-labeled oligonucleotide complementary to exon 1 sequence, hybridized to 50 µg total RNA from rat Nb-2 lymphoma cells or normal rat thymus, and the reaction products were analyzed by electrophoresis on a 7% polyacrylamide/7 M urea sequencing gel.
To characterize the exon–intron boundaries and genomic organization of Gfi1, an F344 rat genomic λDash clone (16) was digested with EcoR1 and the resulting fragments were subcloned in pBluescript SK (Stratagene). Next, oligonucleotide primers based on the Gfi1 cDNA sequence were used to probe Southern blots of the genomic subclones, for PCR amplification and for sequencing.
Cultures were harvested, and nuclei were isolated and subjected to incremental DNase I digestions as described previously (17).
Shifts were performed essentially as described previously (18). In vitro transcription and translation (IVT) was performed with the TNT T7-coupled reticulocyte lysate system (Promega) according to the manufacturer’s protocol. Jurkat nuclear extracts were made using a standard Dignam protocol (19). The sequences and identities of the four probes used are as follows: (i) probe B30 is a transcriptionally active, synthetic GFI1 recognition site in place of a site at –550, with rat promoter sequence surrounding. The sequence is 5′-GGGATCGCCACACCAAATCACTGCACCCCGGT-3′ and the mutant probe is 5′-GGGATCGCCACACCAGGTC ACTGCACCCCGGT-3′. (ii) Probe 2 is the GFI1 binding site at –701 of the rat promoter and surrounding promoter sequence. The sequence is 5′-GGAGCAAACCTCTGGG ATTGGGTGTCAAGGTA-3′ and the mutant probe is 5′-GGAGCAAACCTCAGGGACCGGGTGTCAAGGTA-3′. (iii) Probe 4 contains two GFI1 binding sites at –434 and –421 of the rat promoter. The sequence is 5′-GGTCCCGTCAA TCTGTGTCCTGAATCTGTGAC-3′, and the mutant probes are 5′-GGTCCCGTCGGTCTGTGTCCTGAATCTGTGAC-3′ (M1), 5′-GGTCCCGTCAATCTGTGTCCTGGGTCTG TGAC-3′ (M2) and 5′-GGTCCCGTCGGTCTGTGTC CTGGGTCTGTGAC-3′ (M3). (iv) Probe I contains two GFI1 binding sites in the rat intron1 that are conserved in mouse but not human. The sequence is 5′-GGA ACCCCCAATCAGTTCACCTAATCTCGGGT-3′, and the mutant probes are 5′-GGAACCCCCGGTCAGTTCAC CTAATCTCGGGT-3′ (M1), 5′-GGAACCCCCAATCAGT TCACCTGGTCTCGGGT-3′ (M2) and 5′-GGAACCCC CGGTCAGTTCACCTGGTCTCGGGT-3′ (M3). Bolded nucleotides are GFI1 core recognition sequences that were changed to generate mutant probes (AATC to GGTC). The underlined bases were added for labeling with Klenow (New England Biolabs, Beverly, MA) and [α-32P]dCTP (Amersham, Piscataway, NJ). Jurkat nuclear extracts (2.5 µg) or IVT protein (3–5 µl) were pre-incubated in binding buffer (18) at room temperature for 20 min, with cold oligonucleotide or antibodies [anti-GFI1 (sc-8558) or normal goat antiserum (sc-2028)] for competition and super-shift assays. Labeled probe (50 000 c.p.m.) was added, and the reaction continued at room temperature for 30 min. Samples were electrophoresed through a 6% non-denaturing polyacrylamide gel, which was dried and exposed to film.
Chromatin immunoprecipitation analysis (ChIP) assays were performed as described previously (20). Antibodies used for immunoprecipitation were rabbit anti-GFI1 (Pharmingen, catalogue no. 559680) and normal rabbit control, or goat anti-GFI1 and normal goat control (sc-2027, sc-8558 and sc-2028, respectively; Santa Cruz Biotechnology, Inc.). Each experiment was performed at least twice with similar results, and representative data are shown.
We cloned Gfi1 in an insertion mutagenesis screen for targets of MoMLV that could mediate progression of rat T-cell leukemias from IL-2-dependent to -independent growth (16). The identification of Gfi1 as an oncogene in rat T-cell leukemia lines provided a biological impetus for studying the transcriptional regulation of rat Gfi1. We therefore sought to examine the rat Gfi1 locus for regions of transcriptional control. Genomic clones from the rat locus were analyzed by restriction endonuclease digestion, followed by Southern blotting with Gfi1 cDNA probes and sequencing. We determined both the exon–intron boundaries of the gene (Fig. (Fig.1A)1A) and the organization of the rat Gfi1 locus (Fig. (Fig.1B).1B). Specifically, examination of the sequence of exon–intron boundaries revealed the presence of seven exons and six introns. The exon–intron boundary sequences are homologous to the previously published mouse locus (21), with minor exceptions. The genomic organization of rat Gfi1 is shown as a diagram in Figure Figure1B1B and spans 9.5 kb of sequence. These data were corroborated by the recently available rat genomic sequence in DDBJ/EMBL/GenBank (data not shown).
To determine potential regulatory regions, DNase hypersensitivity assays were performed (Fig. (Fig.1C)1C) with two different genomic probes (Fig. (Fig.1B,1B, I and II). Analysis of the Gfi1 locus in rat Nb2 and A2 leukemia cell lines with probe I revealed hypersensitive sites clustering around exon 1 (Fig. (Fig.1B,1B, arrows). Many of these sites occur within the first intron, but the major site of DNase hypersensitivity is just upstream of the first exon (Fig. (Fig.1B,1B, longer arrow, and C, arrow). In LE3Spl cells, two downstream hypersensitive sites were also identified with probe II (Fig. (Fig.1C)1C) and are indicated by arrows in Figure Figure1B.1B. With either probe, the pattern of DNase hypersensitivity was similar with or without prolactin addition to prolactin-dependent Nb2 cells, and with or without IL-2 addition to the IL-2-dependent LE3Spl cell line (data not shown). Thus, these regulatory regions are unlikely to be growth-factor dependent.
We next mapped the start site of transcription of the rat Gfi1 gene. The mouse Gfi1 locus has both a major transcription start site 5′ of exon 1a, and a minor one in the first intron (21). We therefore performed RNase protection analysis to locate the transcriptional start site(s) in the rat Gfi1 gene. In RNA from rat Nb2 lymphoma cells, a sequence proximal to exon 1 was identified as the single start site of transcription in the rat Gfi1 gene (Fig. (Fig.2D).2D). This sequence (Fig. (Fig.1D,1D, TCAGAGC) corresponds to the consensus sequence of an initiator element (Inr) (22). Inr elements are commonly utilized as transcription start sites in TATA-less lymphoid-specific genes such as terminal deoxynucleotidyl transferase (23). In agreement with this, we found no evidence of a canonical TATA box in the sequence of this region. Moreover, the relative position of the Inr corresponds to the major site of DNaseI hypersensitivity. In contrast to both mouse and human Gfi1, which generate multiple sized transcripts, the rat locus generates a single transcript (16). Unlike murine Gfi1, no start sites were detected in the first intron of the rat gene (data not shown). However, the rat Gfi1 start site is very similar to that of the major start site of the mouse gene.
The sequence around the transcription start site of the rat Gfi1 locus was compared with the sequences of the mouse and human Gfi1 loci. This analysis revealed a great deal of homology among rat, mouse and human sequences up to 808 bp upstream of the putative Inr of the rat locus (Fig. (Fig.2A,2A, gray areas). The extent of sequence homology is unusual given that the nearest protein coding sequences are at least 1.6 kb distal, emphasizing the probable importance of this region in the transcriptional control of Gfi1.
The Inr sequence is conserved in the mouse locus, and is located near the previously mapped start site of transcription for mouse Gfi1. In contrast, the Inr sequence is not conserved in the human sequences we identified. However, the reported major start site of the human Gfi1 promoter (24) is just upstream of the sequences shown and contains a consensus Inr element according to our analysis (data not shown). Thus, mouse, rat and human Gfi1 are likely to have similar cis-acting transcriptional control elements.
We examined the rat promoter sequence for transcription factor binding sites by computer-assisted matrix similarity analysis (MatInspector) (25). Transcription factor binding sites conserved among all three species are illustrated (Fig. (Fig.2).2). These include nuclear factor of activated T cells (NFAT), interferon response factor 2 (IRF2) and GATA, all of which respond to exogenous signals in T cells (26–28). Additional putative regulatory sites include recognition sequences for E4BP4/NFIL3, HoxA9, SP-1, and the Ets factors ETS1 and ELK1. Strikingly, several putative GFI1 binding sites that are conserved between mouse, rat and human loci were identified (Fig. (Fig.2).2). Alignment of sequences in the first intron of the rat locus with mouse and human Gfi1 revealed extensive homology in rodents and two potential GFI1 binding sites (data not shown).
The presence of GFI1 recognition sequences in the Gfi1 promoter suggested autoregulation. We previously reported low-level expressing GFI1-transgenic mice constructed with the Lck proximal promoter, and a rat Gfi1 cDNA that lacks the 3′ untranslated region (Fig. (Fig.3A).3A). We therefore examined the levels of endogenous Gfi1 transcript in these GFI1-transgenic thymocytes. Northern analysis with a murine Gfi1 3′-untranslated-region probe on Poly-A+-selected RNA from three GFI1-transgenic mouse thymi revealed an average 40% reduction in the level of endogenous Gfi1 transcript (Fig. (Fig.3B).3B). The modest effect on endogenous Gfi1 levels is in accord with the low level of transgene expression. As expected, western blot analysis of thymocyte whole-cell lysates revealed similar levels of overall GFI1 protein in GFI1-transgenic and littermate control thymocytes (Fig. (Fig.3C).3C). Analysis of CD2 promoter-driven GFI1 transgenic thymocytes revealed similar results (data not shown).
To determine if human Gfi1 could be autoregulated, we transfected a human T-cell line (Jurkat) with a selectable marker and the Lck-promoter GFI1-transgene construct (Fig. (Fig.3A),3A), and selected stable clones. Northern analysis of Poly-A+-selected RNA with a human Gfi1-specific probe revealed profoundly reduced levels of endogenous Gfi1 message in multiple independent GFI1-transfected clones, two of which are shown (Fig. (Fig.3D).3D). Western analysis confirmed high levels of GFI1 protein in the clones compared with control Jurkat cells (Fig. (Fig.3E).3E). To ensure that the observed repression is not the result of artifact induced by integration of the transgene or antisense generated by transgene concatemers, we repeated our analysis with independent clones of Jurkat T cells that were transduced with a GFI1-expressing retroviral vector (Fig. (Fig.3F).3F). Again, northern analysis revealed a profound decrease in endogenous Gfi1 expression (Fig. (Fig.3G).3G). Western analysis revealed increased expression of GFI1 in the transduced cell lines (Fig. (Fig.3H).3H). Therefore, both mouse and human Gfi1 respond to the level of GFI1 in a manner consistent with autoregulation.
To determine a molecular basis for the regulation of Gfi1 by forced GFI1 expression, we performed ChIP in Jurkat T cells. Cross-linked chromatin was immunoprecipitated with two independent antisera that are specific for GFI1. In templates generated from either antiserum immunoprecipitation, products of the amplification reaction were seen with primers directed to the rat/human homologous sequences in Figure Figure22 (Fig. (Fig.3I,3I, top and middle panels). In contrast, no product was seen in amplification reactions from control-antiserum templates (Fig. (Fig.3I,3I, top and middle panels), emphasizing the specificity of the immunoprecipitation. Furthermore, primers specific for a region of the Gfi1 locus that does not contain GFI1 binding sites by our analyses did not amplify a product (Fig. (Fig.3I,3I, bottom panel). GFI1 is expressed in U937 myeloid monocytic cells and ChIP analyses have shown that it is bound to several promoters in this cell line (20). Nonetheless, primers directed to the promoter of Gfi1 failed to amplify a product from U937 chromatin that had been immunoprecipitated with goat anti-GFI1 antibody (Fig. (Fig.3J),3J), and forced expression of GFI1 in U937 cells did not lead to silencing of Gfi1 transcription (data not shown). Therefore, in living Jurkat T cells, but not in U937 myeloid cells, GFI1 is specifically bound to the human Gfi1 sequences shown in Figure Figure22.
The GFI1/GFI1B consensus binding site is defined as TAAATCAC(A/T)GCA, with an absolute requirement for the AATC core (7,9). Mutation of AATC to GGTC ablates DNA binding (7,9). To determine which of the putative sites illustrated in Figure Figure22 can be bound by GFI1, we designed 30 bp oligonucleotide probes for use in electrophoretic mobility shift assay (EMSA) with nuclear extracts from human Jurkat or rat NB2 cell lines (data not shown). Both extracts shifted the same probes. A synthetic GFI1 binding site that has been shown to be active in transient transcription assays (B30) (8) was included for comparison. Incubation of Jurkat nuclear extract with the B30 probe resulted in the formation of two prominent protein–DNA complexes (Fig. (Fig.4A,4A, lane 2). As reported previously (9), excess unlabeled B30 oligonucleotide, but not excess ‘AATC to GGTC’ mutant B30, was able to compete for binding to the radiolabeled B30 oligonucleotide (Fig. (Fig.4A,4A, lanes 3 and 4).
Several different protein–DNA complexes were seen with individual probes from the mouse/rat/human homologous Gfi1 sequences in Figure Figure2,2, most of which were relatively minor in intensity and did not display GFI1-specific interaction. For simplicity, these minor GFI1-non-specific sites are not shown. However, T-cell nuclear extracts retarded the mobility of probes containing sites labeled 2 and 4 in Figure Figure2,2, and the sites in the first intron (I) in a GFI1-dependent manner (Fig. (Fig.4A,4A, lanes 6, 11 and 16). Binding of both radiolabeled probes 2 and I was competed by adding excess unlabeled self or excess unlabeled B30, but not by adding excess unlabeled ‘AATC to GGTC’ mutant B30 (Fig. (Fig.4A,4A, lanes 7–9 and 17–19). Surprisingly, binding of Jurkat nuclear extract to radiolabeled probe 4 was competed only by adding excess unlabeled probe 4, and not by adding either excess unlabeled B30 or mutant B30 (Fig. (Fig.4A,4A, lanes 12–14). We therefore conclude that the probes bind to three distinct protein complexes, two of which recognize sequences in the B30 probe, the other of which does not.
To determine whether these complexes contain GFI1, we performed supershift and competition analysis on each of the three probes (2, 4 and I). Figure Figure4B4B shows the results of these analyses performed with probe 2, which contains a single GFI1 binding site corresponding to –701 of the rat promoter (relative to the Inr). This probe shifts in a doublet pattern in the presence of Jurkat nuclear extract (lane 2), and the shift was ablated when the nuclear extract was pre-incubated with anti-GFI1 antibody (lane 3), whereas a control antibody had no effect (lane 4). Furthermore, while binding was competed away in the presence of excess cold probe 2 (lane 5), there was no competition in the presence of excess cold mutant 2 (lane 6). When the mutant probe 2 was radiolabeled and used alone, the doublet was not seen (lane 7). To confirm that GFI1 binds directly to probe 2, we performed EMSA with IVT FLAG-epitope-tagged GFI1 (Fig. (Fig.4B),4B), or DNA-binding-deficient GFI1 (data not shown). IVT GFI1 formed a complex with probe 2 that migrated at a significantly higher rate than that of complexes formed with Jurkat nuclear extract (compare lanes 2 and 9). Furthermore, pre-incubation of IVT GFI1 with anti-FLAG antibody resulted in the formation of a supershift, while a control antibody had no effect (lanes 10 and 11, respectively). DNA-binding-defective IVT GFI1 did not form a complex with probe 2 (data not shown). Taken together, these data suggest that GFI1 in Jurkat nuclear extract binds directly to probe 2 in the context of a larger complex, rather than as a monomer.
The same analysis was performed with probe 4, which contains two GFI1 binding sites at –434 and –421 of the rat promoter. Pre-incubation of Jurkat nuclear extract with anti-GFI1 antibody abated complex formation, whereas a control antibody did not (Fig. (Fig.4C,4C, lanes 3 and 4), and pre-incubation of IVT GFI1 with anti-FLAG antibody resulted in a supershift, whereas a control antibody did not (lanes 14 and 15). These results are similar to those observed with probe 2. Further analysis revealed that only one of the two GFI1 binding sites is necessary for GFI1 from Jurkat nuclear extract to bind to probe 4. Each site was mutated independently (4a mutant = M1 and 4b mutant = M2) or both sites were mutated together (M3), and these mutants were used in competition analyses. Both wild-type and M2 oligonucleotides effectively competed away the complex (lanes 5 and 7), whereas M1 and M3 had no effect (lanes 6 and 8). Furthermore, when the mutant oligonucleotides were labeled and used as probes, M1 and M3 did not form a complex, while M2 formed the same complex as the wild-type probe 4. These analyses clearly demonstrate that only site 4a binds to GFI1 in Jurkat nuclear extract, and that site 4b is dispensable for this binding.
The same analysis was performed with probe I, which contains two GFI1 binding sites from the first intron of rat Gfi1 (Fig. (Fig.4D).4D). These sequences are conserved between mouse and rat genes. The results of this analysis were similar to those obtained with probe 4. Pre-incubation with antibodies showed ablation of binding to Jurkat nuclear extract and supershift of IVT GFI1 (lanes 3 and 14, respectively), and competition analysis showed that only one of two putative GFI1 binding sites is necessary for the formation of complexes (lanes 5–11). We therefore conclude that GFI1 in protein complexes can bind to cognate binding sites that are present in mouse/rat/human conserved sequences shown in Figure Figure2,2, and in the first intron that is conserved in rodent genomes.
To determine if the GFI1-binding sites in Figure Figure44 mediate GFI1-responsive transcriptional repression, we performed transient transcription analyses. We constructed a CAT reporter driven by the rat Gfi1 promoter and first intron (–808 bp up to the beginning of exon 2), containing all of the DNase hypersensitive sites identified in the 5′ end of the gene (Fig. (Fig.1).1). Co-transfection of 293T cells with this reporter and a GFI1-expression construct resulted in repression of reporter activity (Fig. (Fig.5A).5A). However, co-transfection with a plasmid encoding a SNAG-repression-domain mutant of GFI1 (P2A) had no effect (Fig. (Fig.5A).5A). Furthermore, co-transfection of this reporter with increasing amounts of a plasmid encoding a fusion protein that consists of the herpes simplex virus VP16 transactivator and the zinc fingers of GFI1 results in dose-dependent activation of the Gfi1 reporter (Fig. (Fig.5B).5B). Therefore, the region of rat Gfi1 containing the proximal promoter and first intron is capable of binding and being repressed by GFI1 in 293T cells.
To permit similar experiments in Jurkat T cells, we changed to a luciferase reporter. We constructed a Click Beetle Luciferase reporter with the rat Gfi1 sequences (–808 bp up to the beginning of exon 2). Co-transfection of this reporter with an expression plasmid encoding GFI1 resulted in a decrease of reporter activity to an average of 55% of control. We next examined the requirement for sites 2, 4 and I, which bound GFI1-containing complexes during EMSA (Fig. (Fig.4).4). Surprisingly, mutation of either site 2 or site 4 individually had little or no effect on GFI1 repression. In contrast, mutation of both sites 2 and 4 decreased the GFI1-mediated repression of the reporter, such that the activity was an average of 75% of the control. Furthermore, mutation of GFI1 binding site I also decreased GFI1-mediated repression to a level comparable to that seen in the site 2 and 4 double mutant. Finally, mutation of GFI1 binding sites 2, 4 and I resulted in a complete lack of exogenous GFI1-mediated repression of this luciferase reporter. We conclude that promoter or intronic sites are sufficient for GFI1-mediated repression, but all sites are necessary for efficient GFI1-mediated repression.
Transgenic expression of GFI1B (Fig. (Fig.6A)6A) engenders defects in T lymphopoiesis, some of which can be rescued by simultaneous forced expression of GFI1 (6). Because GFI1 and GFI1B have nearly identical DNA binding domains (7), we previously reasoned that GFI1B may alter T lymphopoiesis by competing with endogenous GFI1 for DNA binding on specific promoters (6). We therefore sought to determine the extent to which transgenic expression of GFI1B disturbs the stoichiometry of GFI1/GFI1B in thymocytes. Dramatically, northern (Fig. (Fig.6B)6B) and western (Fig. (Fig.6C)6C) analysis of GFI1B-transgenic thymocytes revealed undetectable levels of GFI1. Thus, instead of competing with GFI1 for binding sites of target genes, transgenic GFI1B may occupy all of these binding sites in the absence of GFI1.
We next determined whether GFI1B directly represses Gfi1 in a manner similar to GFI1. First, we examined Jurkat T cells cotransfected with the GFI1B transgene construct and a selectable marker. Northern analysis of independent clones revealed that ectopic expression of GFI1B in Jurkat cells resulted in repression of endogenous Gfi1 (Fig. (Fig.6D).6D). Notably, higher levels of GFI1B protein correlated with lower levels of Gfi1 message (Fig. (Fig.6E).6E). Next, we examined the ability of GFI1B to bind sites 2, 4 and I. In fact, IVT GFI1B binds efficiently to all three probes, as shown by EMSA analysis with supershift (Fig. (Fig.6F).6F). Moreover, like GFI1, GFI1B expression constructs are capable of repressing the Gfi1-CAT reporter in 293T cells (Fig. (Fig.6G)6G) and in Jurkat T cells (Fig. (Fig.6H).6H). Finally, this repression is abrogated upon mutation of the GFI1 binding sites 2, 4 and I (Fig. (Fig.6H).6H). Therefore GFI1B may also regulate Gfi1 through the same sequences used by GFI1.
GFI1 is a transcriptional repressor protein that plays important biological roles in hematopoietic and neuronal cell development (1–3,29). Gfi1 was originally identified in a screen for genes that, upon deregulated expression by MoMLV insertion, engender progression of T-cell leukemias to IL-2-independent growth (16). While the frequency of MoMLV insertion in the Gfi1 locus indicates some selective advantage of forced GFI1 expression (12), validation of the oncogenic potential of GFI1 came from studies in transgenic mice. Transgenic overexpression of GFI1 is poorly oncogenic alone, but potently collaborates with either MoMLV infection or transgenic expression of MYC or PIM oncoproteins to cause leukemia (30). Therefore, cellular context may determine GFI1 oncogenic function. These data implicate GFI1 in T-cell biology. Indeed, mutation of Gfi1in mice and humans leads to lymphopenia (1,3,4). Since Gfi1 expression is critical for normal hematopoiesis, and overcoming regulatory control of Gfi1 accelerates oncogenesis, the transcriptional control of Gfi1 is of great interest. Here we have shown that GFI1, and the closely related GFI1B, both repress transcription of Gfi1. This is the first example of a gene targeted directly by GFI1, and the first example of a gene targeted by both GFI1 and GFI1B.
We present several lines of evidence supporting a direct mechanism of Gfi1 repression by GFI1. First, GFI1 binding sites in the rat Gfi1 promoter are conserved in the mouse and human Gfi1 loci. The conservation of these sequences is striking because the nearest coding sequence is at least 1.6 kb 3′ of the Inr. Secondly, in primary thymocytes, transgenic expression of GFI1 correlates with lower levels of endogenous Gfi1 steady-state mRNA levels. Thirdly, in the human Jurkat T-cell line, forced overexpression of GFI1, either by the Lck promoter or retroviral LTR, correlates with reduced endogenous Gfi1 steady-state mRNA levels. Fourthly, in EMSA analysis, in vitro-transcribed/-translated GFI1 and endogenous GFI1-containing nuclear complexes bind directly to sequences in Gfi1. Fifthly, mutation of these binding sites in reporter constructs ablates transcriptional response to GFI1 expression. Finally, ChIP analyses with two different antisera indicate that GFI1 binds specifically to mouse/rat/human conserved sequences in living Jurkat T cells. These in vitro and in vivo analyses provide strong support for the hypothesis that Gfi1 is autoregulated. In fact, Gfi1 autoregulation is evolutionarily conserved, as pag-3, the Caenorhabditis elegans ortholog of Gfi1, is autoregulated (31).
GFI1 binds to mouse/rat/human conserved sequences in living Jurkat T cells, but not in U937 myeloid cells. Interestingly, Gfi1–/– mice, which completely lack GFI1, have severe lymphopenia and profound neutropenia (1–3), while humans that carry a heterozygous dominant-negative mutation in Gfi1 display severe congenital neutropenia (SCN), but only mild lymphopenia (4). It is possible that expression of mutant GFI1 proteins in SCN patient T cells might lead to de-repression of the wild-type Gfi1 allele, providing compensatory expression of wild-type GFI1 and sufficient GFI1 activity in T cells. In contrast, the myeloid progenitor cell line U937 does not appear to exhibit Gfi1 autoregulation. If similar myeloid cells in SCN patients do not have compensation by the wild-type Gfi1 allele, then these cells would effectively lose most or all GFI1 function. This possibility warrants future investigation to determine if cell-type specific Gfi1 autoregulation provides an explanation for the differences in phenotype observed in Gfi1–/– mice when compared with SCN patients with a mutation in a single allele of Gfi1.
GFI1B can control Gfi1 expression. The abilities of GFI1 and GFI1B to repress endogenous Gfi1 seem comparable. Similar expression levels of the two factors in stably transfected Jurkat T cells resulted in similar repression. Both factors bound to the same DNA binding sites in EMSA analyses. Co-transfection of equal amounts of GFI1 and GFI1B expression plasmids with a Gfi1 promoter-driven reporter resulted in nearly equal levels of repression in 293T cells, and the repression was dependent on the same binding sites in Jurkat T cells. Thus, GFI1B is capable of regulating Gfi1.
The repression of Gfi1 by GFI1B potentially provides new insight into the transient expression of GFI1B in thymocytes that have been recently signaled for positive selection. Namely, GFI1B may be present to temporally repress Gfi1. Indeed, the expression of GFI1B during beta selection is followed by a decrease in GFI1 expression (32). In contrast, forced expression of GFI1 in immature double-positive thymocytes has been shown to be detrimental, resulting in a block to beta selection (32). The coincident expression of GFI1B at this stage of development may be necessary to maintain appropriate levels of Gfi1, thereby allowing the progression through this critical checkpoint.
GFI1B antagonism to GFI1-mediated T-cell activation may be understood as follows. First, in normal thymocytes the expression level of GFI1B is sufficiently low such that it does not completely repress Gfi1, and both GFI1 and GFI1B are expressed. This simultaneous expression results in competition for DNA-binding sites on target genes, which would allow proper signaling and selection of developing thymocytes. However, transgenic expression of GFI1B disrupts GFI1/GFI1B stoichiometry and results in the complete repression of Gfi1 transcription. Consequently, common target genes are occupied by GFI1B. Since GFI1 increases T-cell activation (6,10,11), GFI1B expression may indirectly inhibit T-cell activation through the repression of Gfi1. Additionally, GFI1B may have effects distinct from those of GFI1 on at least some target genes, as presumed occupancy by GFI1B does not necessarily result in the same effects as occupancy by GFI1.
The demonstration of GFI1 and GFI1B regulation of Gfi1, together with promoter analyses, suggests a very complex regulatory network controlling the expression of Gfi1. The extent of sequence homology between the rat Gfi1 promoter and murine and human Gfi1 loci emphasizes the probable importance of this region in the transcriptional control of Gfi1. Indeed, C.elegans and Caenorhabditis briggsae display extensive sequence homology in the promoter regions for the Gfi1 ortholog pag-3, even though they diverged 20–50 million years ago (33). Thus, the GFI family may be conserved signal transduction proteins that mediate pathways initiated by other transcription factors. Computer-assisted matrix similarity analysis (MatInspector) (25) of the rat/mouse/human homologous sequences in Figure Figure22 suggests that such factors may include NFAT, IRF2, GATA proteins, E4BP4/NFIL3, HoxA9, SP-1 and Ets proteins. Thus, in GFI1-expressing cells, a balance between positive regulators of Gfi1 expression, autoregulation and GFI1B repression should determine total Gfi1 levels. Future work should be directed towards identifying positive regulators of Gfi1 expression.
The authors would like to thank Stuart Orkin for providing mouse genomic Gfi1 clones for sequencing, and Morgan Jeffries for help with EMSA analyses. The authors are also grateful for excellent technical assistance from Rachel Rivoli and Natalie Claudio. This work was supported by Hope Street Kids, and in part by the Commonwealth of Kentucky Research Challenge Trust Fund and the Jewish Hospital Foundation. L.L.D. was supported by a National Science Foundation Graduate Research Fellowship. This work was supported in part by NIH PHS CA56110 (to P.N.T.), NIH PHS DK55820, NIH DK58161 and Borroughs-Welcome Fund SATR-1002189 (to M.H.)