Sumoylation of EKLF Promotes Transcriptional Repression and Is Involved in Inhibition of Megakaryopoiesis^{▿ †}

Cell lines.

Cells of the murine erythroleukemia (MEL) line MEL 745A and human embryonic kidney cell-derived line 293T were cultured in Dulbecco's modified Eagle's medium, and K562 (human erythroleukemia) cells were cultured in RPMI medium, both supplemented with 10% fetal calf serum and antibiotics. Stable K562 and MEL 745A cells were established after transfection with pSVneoHMTIIA-Ter containing wild-type (WT) or K74R EKLF under a zinc-inducible promoter (43) using DMRIE (Invitrogen) or Tfx 50 (Promega) reagent, respectively. Selection was with G418 at final concentrations of 400 μg/ml for K562 cells and 800 μg/ml for MEL 745A cells.

Assessment of erythroid differentiation of a MEL stable cell line.

MEL stable cells were induced with 80 μM Zn for 18 h to express WT and K74R EKLF. For erythroid differentiation, cells were treated with 5 mM hexamethylene bisacetamide (Sigma catalog no. H6260) for 4 days. Each day, RNA samples were collected for quantitative PCR (qPCR) analysis.

Assessment of megakaryocytic or erythroid differentiation of K562 stable cell line.

K562 stable cells were induced with 160 μM Zn for 18 h to express WT and K74R EKLF. For megakaryocytic differentiation cells were diluted to 10E5 cells/ml and treated with 3 nM TPA (phorbol 12-myristate 13-acetate [Sigma; catalog no. P8139]) for 4 days. Each day, cells were monitored by flow cytometry for the presence of megakaryocytic marker CD41 (10E5 [antibody labeled with Alexa 647 was a kind gift from B. Coller]) and analyzed by FACScalibur (Becton Dickinson, San Jose, CA), and RNA samples were collected for qPCR analysis.

For erythroid differentiation, cells were diluted to 2 × 10E5 and treated with 30 μM hemin (Sigma H5533) for 4 days. Each day, cells were monitored by staining with benzidine and RNA samples were collected for qPCR analysis.

Immunofluorescence studies and analysis of subcellular localization.

K562 cells were used to examine the expression and subcellular localization of the EKLF-yellow fluorescent protein (YFP) fusion together with green fluorescent protein (GFP)-SUMO fusion. Cells were additionally cotransfected with Ubc9 and PIAS1. After 36 h, cells were cytospun onto slides and fixed for 10 min at room temperature in 4% formaldehyde (diluted in phosphate-buffered saline [PBS]). Three subsequent wash steps were performed in PBS. The slides were then mounted in Vectashield (Vector Laboratories, Inc.) supplemented with DAPI (4′,6′-diamidino-2-phenylindole) to identify cell nuclei. YFP, GFP, and DAPI were visualized on a Zeiss LSM 510 META confocal laser-scanning microscope.

RT-qPCR.

Total RNA was isolated from K562, MEL, or lineage-depleted bone marrow cells using TRIzol reagent (Sigma). One microgram was then reverse transcribed using a Promega reverse transcription (RT) system kit with random hexamers. Real-time PCR was performed using a SYBR green kit (Applied Biosystems 4304886) and primer sequences as described in the supplemental material. Results were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Analysis was performed on an ABI Prism 7900 HT on 384-well plates.

Generation of transgenic mice with EKLF under PF4 promoter control.

EKLF expression was placed under the control of the megakaryocyte-specific platelet factor 4 (PF4) promoter by subcloning its cDNA (WT or K74R mutant) into the Nhe1 site of pPF4 (kind gift of K. Ravid [61]). Transgenic mice were obtained after injection of the linearized (NdeI) fragment into the pronucleus of one-cell-stage fertilized eggs (Mount Sinai Genetics Shared Research Facility). Genotyping was performed by genomic PCR using transgene-specific primers pF4 (AGAAAGTCACAGGAGCCACTG) and EKLF R2.75 (TCTGAGGAGACGCAGGATTTG). Expression was monitored by RT-PCR analysis of transgenic bone marrow RNA using EKLF/transgene-specific primers that spanned an intron within the transgene: pF4-beta3 (CAGCCACCACTTTCTGATAG) and pF4-EKLF5 (CTGCCCACGTGCTTTTTCAC). We obtained three founder mice for pF4-WT-EKLF and five founder mice for pF4-K74R-EKLF. The results of the experiments consist of the average obtained from two founders of each type.

Colony assays.

Bone marrow cells from the tibia and fibula of 6- to 12-week-old mice were isolated by flushing the marrow cavity in Iscove's modified Dulbecco's medium. Colony assays in methylcellulose or Megacult (Stem Cell Technologies) were performed as described previously (16) using transgenic or nontransgenic bone marrow cells that had been enriched by lineage depletion with StemSep magnetic beads (Stem Cell Technologies) that included interleukin-7R antibodies (eBiosciences) (34). Acetylcholinesterase assays were performed as recommended by the manufacturer (Stem Cell Technologies).

Bone marrow cells were suspended in PBS, cytospun onto a glass slide, air dried, and stained with May-Grunwald Giemsa (MGG) stains (Sigma) as per the manufacturer's recommendations shortly after harvesting and lineage depletion. Slides were scored by examination of 500 cells in duplicate under a ×100 oil immersion objective. Images were taken using a Zeiss Axiophot microscope.

EKLF is covalently modified by SUMO-1 at lysine 74.

Close examination of the EKLF amino acid sequence revealed one lysine that matched perfectly within the consensus sumoylation motif (i.e., ΨKXE, where “Ψ” represents a hydrophobic amino acid and “X” represents any amino acid). Within EKLF it is LKSE, where lysine 74 is a potential acceptor site for SUMO modification. This motif is located in the N-terminal transactivation domain of EKLF and is highly conserved (Fig. 1A). To determine whether EKLF undergoes sumoylation, we transiently transfected 293T cells with constructs encoding Flag-tagged EKLF, hemagglutinin (HA)-tagged SUMO-1, E2 enzyme-Ubc9, and one of the possible E3 ligases (PIAS1).

In the presence of the components of SUMO machinery, a band migrating slower than EKLF is detected that is missing in the sample with EKLF alone, suggesting that EKLF can be sumoylated (Fig. 1B, lanes 1 and 2).

To confirm that the LKSE motif contains the SUMO acceptor site, we generated a mutant in which Lys 74 was changed to a similarly charged Arg residue (K74R EKLF) and tested for its ability to be modified. Western blot analysis revealed that even in the presence of components of the SUMO machinery, the K74R mutation abolished sumoylation (Fig. 1B, lane 3). This suggests that K74 is indeed the sole functional sumoylation site.

To further prove that EKLF is sumoylated and that the slower-migrating band is due to SUMO modification, we used GFP-tagged SUMO-1 protein instead of HA-tagged SUMO-1. Upon cotransfection, we noticed that the slower-migrating band was retarded even more due to the presence of the larger GFP tag (Fig. 1B, compare lanes 4 and 6). As expected, this band did not appear in the sample containing K74R EKLF (lane 5). These data provide evidence that the slower-migrating form of EKLF originates from covalently conjugated SUMO-1 to lysine 74.

EKLF is sumoylated in erythroid cell lines and primary fetal liver cells.

As EKLF is an erythroid-specific transcription factor, we investigated whether it can be posttranslationally modified in MEL cells by the endogenous sumoylation machinery. We generated two stably transfected MEL cell lines: one with Flag-WT EKLF and a second with Flag-K74R EKLF as a negative control. After immunoprecipitation with anti-Flag antibodies, the proteins were immunoblotted with anti-EKLF and anti-SUMO antibodies. As shown in Fig. 1C, only WT EKLF is sumoylated by the endogenous SUMO machinery. In addition we examined the status of endogenous EKLF in E13.5 fetal liver cells by preparing protein extracts followed by immunoprecipitation with monoclonal anti-EKLF antibody. The subsequent Western blot probed with polyclonal anti-EKLF reveals two slower-migrating EKLF bands that are also recognized by the anti-SUMO antibody (Fig. 1D). Although the reason for the appearance of two bands is not clear, this is only observed in primary erythroid and not in transfected cell extracts. These results indicate that endogenous EKLF is sumoylated by endogenous proteins in vivo.

PIAS1 serves as an E3 ligase for EKLF sumoylation.

Previous studies have shown that proteins from the PIAS family can act as SUMO E3 ligases, and as such, they significantly enhance the level of sumoylation of the substrate (33). To examine whether PIAS1, PIAS3, or PIASy could act as a specific ligase for EKLF, we transfected 293T cells with constructs encoding Flag-EKLF, HA-SUMO-1, Ubc9, and Flag-tagged proteins from the PIAS family. After immunoprecipitation with anti-Flag antibodies, the proteins were subjected to Western blot analysis, first with anti-EKLF antibodies to detect the level of EKLF with its modification and second with anti-Flag antibody to detect the level of expression of PIAS proteins. As shown in Fig. 2A, we find that the slower-migrating form of EKLF appears only in the presence of PIAS1. Figure 2B confirms that all PIAS proteins were properly expressed. Comparison of lanes 4, 5, and 6 in Fig. 2A reveals that the presence of increasing amounts of cotransfected PIAS1 correlates with an increase in the intensity of the slower-migrating EKLF bands corresponding to sumoylated EKLF. No such bands are observed with the K74R mutant. In addition, in the presence of PIAS3 or PIASy, we do not detect any SUMO-conjugated EKLF. Thus, we conclude that PIAS1 promotes sumoylation of EKLF in a dose-dependent manner and that PIAS1 is the specific E3 ligase for EKLF. This also supports the idea that E3 SUMO ligases display substrate specificity.

Sumoylation of EKLF depends on the intact RING finger domain of PIAS1 and its interactions with Ubc9.

To further characterize the function of PIAS1 in EKLF sumoylation, first we determined whether the PIAS1 RING finger domain is important for its E3 ligase activity. We introduced three point mutations at positions C346G, C351G, and C356A within the RING finger domain. Cysteines 351 and 356 are involved in Zn²⁺ coordination, and changes of these amino acids will disturb the RING finger structure (Fig. 2C). The effect of these mutations was tested in vivo in 293T cells, which were cotransfected with constructs encoding Flag-EKLF, HA-SUMO-1, Ubc9, and increasing amounts of mutated Flag-PIAS1. Whole-cell extracts were subjected to immunoprecipitation with M2-agarose, and proteins were analyzed with anti-EKLF antibodies. As shown in Fig. 2D, the RING finger mutant fails to enhance sumoylation of EKLF as compared to the WT-PIAS1 (Fig. 2D, lanes 1 to 6). These data suggest that a functional RING domain of PIAS1 is important for its E3 ligase function in EKLF sumoylation.

To examine if PIAS1 needs to interact with Ubc9 during the ligation step, two additional mutants of PIAS1 were constructed. Amino acids located at the surface of PIAS1 that interact with Ubc9 were changed (40). We generated mutants PIAS1-I363S and PIAS1-PV375/6AA and analyzed them the same way as mutants of the RING finger. Both of these mutants are defective in promoting sumoylation of EKLF (Fig. 2D, lanes 7 to 9) (data not shown). The levels of expression of the particular PIAS1 mutants and WT PIAS1 are similar, as shown in Fig. 2E. Thus, we conclude that the interaction between Ubc9 and PIAS1 is essential for SUMO modification of EKLF.

The subcellular localization of EKLF is not altered upon SUMO modification.

Sumoylation has been shown to regulate the intracellular localization of a number of proteins (3, 46). We analyzed the subcellular localization of EKLF in K562 cells, a human erythroleukemia cell line that does not contain endogenous EKLF (12). We transfected these cells with YFP-WT EKLF or mutant YFP-K74R EKLF together with GFP-SUMO, Ubc9, and PIAS1. YFP was fused to the C terminus of EKLF and did not impair its sumoylation, as shown in Fig. 3C. The distribution of both YFP and GFP fusion proteins was revealed by direct fluorescence and shows that EKLF is localized into nucleus, predominantly as distinct, nuclear dot-like structures, independent of its sumoylation status. However, only WT EKLF overlaps with SUMO, unlike K74R EKLF (Fig. 3A and B). Taken together, these results indicate that sumoylation of EKLF is not likely involved in altering its intracellular localization.

Sumoylation affects EKLF transcriptional activity and promotes repression.

Because the SUMO acceptor site for EKLF is located within its transactivation domain, we focused our attention on determining whether this covalent protein modification can alter EKLF transcriptional activity. To address this question, we carried out a series of transactivation assays using a luciferase reporter assay. First we compared the sumoylation-deficient mutant K74R EKLF with WT EKLF protein with respect to their ability to activate the β-globin gene promoter, the major natural target of EKLF (12). K562 cells were cotransfected with this reporter plasmid and with either empty vector or constructs encoding WT or K74R EKLF alone or in the presence of SUMO machinery (Ubc9 and HA-SUMO-1). Cell extracts were then analyzed for luciferase activity. As shown in Fig. 4A, there is little difference in activity between WT and K74R EKLF with or without the presence of SUMO and Ubc9. To enhance the efficiency of sumoylation, we also included PIAS1 (as the E3 ligase) but its inclusion causes a dramatic decrease of activation of both WT and K74R EKLF (Fig. 4A), indicating that this inhibition is SUMO independent. Then, in a similar transactivation assay, we tested other known EKLF native targets, the promoter of the AHSP (α-hemoglobin stabilizing protein) gene and the promoters of the BKLF (basic Kruppel-like factor) (18, 47). We did not detect any difference in activity of either WT or K74R EKLF (Fig. 4B) (data not shown). We also did not detect any effect on endogenous chromosomal β-globin promoter activity in hexamethylene bisacetamide-differentiated MEL cells that had been stably transfected with WT or K74R EKLF and analyzed by quantitative RT-PCR (data not shown).

Having established that sumoylation does not alter the transcriptional capacity of EKLF towards activation of the β-globin or AHSP promoters, we next tested the functional consequence of sumoylation on the repression activity of EKLF. To test for this property EKLF was fused to an exogenous DNA binding domain (GAL4) because its own DNA binding domain, overlapping the zinc fingers, is involved in interactions with corepressors (9). To monitor whether repression activity of EKLF is affected by sumoylation, we again carried out a luciferase reporter assay. K562 cells were cotransfected with a reporter plasmid (5xGAL-TK-Luc) and with expression constructs encoding GAL domain alone or fusion proteins GAL-WT EKLF or GAL-K74R EKLF with or without Ubc9 and HA-SUMO-1.

The basal level of luciferase activity, driven by the thymidine kinase (TK) promoter in the presence of GAL domain alone, is repressed 60% in the presence of GAL-WT EKLF (Fig. 5A), as seen previously (9). Unexpectedly, the sumoylation-deficient mutant (GAL-K74R EKLF) does not behave as a repressor at all, but instead yields a 15-fold-higher transcriptional activity than the GAL-WT EKLF (Fig. 5A). Cotransfection with the components of the SUMO machinery (Ubc9 and HA-SUMO-1) has little effect, likely due to the already high level of endogenous SUMO-1 in K562 cells that allows for sufficient endogenous sumoylation of EKLF (data not shown).

To confirm that the effect of the K74R mutation on repression was a result of a lack of SUMO-1 modification rather than the lack of another potential lysine modification, we repeated the same experiment in two ways. First, we included the dominant-negative (DN), catalytically inactive form of Ubc9 that contains a single amino acid change (C93S) (23). DN Ubc9 suppresses function of the SUMO modification pathway, and thus GAL-WT EKLF should not undergo sumoylation and should behave like the mutant GAL-K74R EKLF. Indeed, when normalized to the level of reporter expression in the presence of WT EKLF (and thus in a repressed state), inclusion of DN Ubc9 leads to derepression and a dramatically higher level of reporter activity (Fig. 5B). The importance of the sumoylation site for this effect is shown by the fact that the K74R EKLF mutant level of activity is not affected by the addition of DN Ubc9. Second, we carried out an analogous experiment in the presence of WT SSP (SUMO-specific isoprotease), which cleaves off SUMO from its substrate proteins and thus leaves GAL-WT EKLF unsumoylated. This also leads to a more active (27-fold) GAL-WT EKLF when normalized to the control experiment performed in the absence of SUMO isopeptidase (Fig. 5C). Again, the GAL-K74R EKLF mutant is barely affected by SSP expression (Fig. 5C).

As a final test, we prepared plasmids that express a chimeric protein by fusion of GAL-WT or GAL-K74R EKLF with SUMO at its N terminus to mimic constitutively sumoylated EKLF irrespective of the K74R mutation. We analyzed them with the same luciferase reporter assay and found that both SUMO-containing fusions efficiently repress the activity as compared to the GAL domain alone (Fig. 5D) or even to GAL-WT EKLF (Fig. 5D, compare lane 2 with lanes 3 and 5).

Taken together, these studies demonstrate two roles for EKLF sumoylation. One is that sumoylation can directly and reversibly regulate EKLF activity as a repressor and suggests that SUMO modification promotes repression activity of EKLF, a result seen with other transcription factors. But the second observation was unexpected, as disruption of the SUMO acceptor site in EKLF does not simply lead to an EKLF protein with a basal, nonrepressive level of activity but rather to its dramatic conversion to a superactivation phenotype, showing that there is a direct correlation between the increase of transcriptional activity and reduction in the degree of sumoylation.

Sumoylation of the transactivation domain of EKLF promotes efficient binding of Mi-2β, the subunit of NuRD complex.

From previous work, it is known that the zinc finger domain of EKLF (amino acids 287 to 376) interacts with Sin3A/HDAC1 (9). However, as the sumoylation site is located in its transactivation (proline-rich) domain, we considered whether another repressor complex could be involved in SUMO-dependent repression. Prior to testing this idea, we first confirmed that a construct containing only the EKLF transactivation region fused to the GAL DNA binding domain could be (at least) derepressed or even superactivated by mutation of its SUMO site (Fig. 6A). Indeed, the derepression caused by the SUMO-deficient proline domain alone (18-fold) is comparable to that seen with full-length EKLF (15-fold), demonstrating that the zinc finger domain is not required for the SUMO-dependent repression. Next, using trichostatin A (TSA), an inhibitor of HDACs, we examined the possible involvement of an HDAC in SUMO-dependent repression. Indeed, in the presence of TSA, the proline domain of GAL-WT EKLF is derepressed (Fig. 6B), indicating that sumoylation and deacetylase activities both play a role in EKLF's repressive properties.

In light of these observations, and based on the published properties and similar behavior of the LIN-1 sumoylated protein from Caenorhabditis elegans or p73 (35, 38), we tested whether the Mi-2β subunit of NuRD complex is involved in sumoylated EKLF repression by monitoring whether they interact in vivo. The GAL-WT or K74R proline domain of EKLF together with Flag-tagged Mi-2β were expressed in 293T cells and subjected to immunoprecipitation using M2-agarose. We find that only the sumoylated proline domain shows any affinity towards the Mi-2β subunit of NuRD complex, unlike the SUMO-deficient point mutant (Fig. 6C). These data suggest that the Mi-2β/HDAC complex is involved in the SUMO-dependent repression property of EKLF and that this is specifically dependent on protein-protein interactions between SUMO-modified EKLF and Mi-2β.

Sumoylation of EKLF is involved in inhibition of megakaryocytic differentiation.

To explore the biological relevance of sumoylation of EKLF, we considered the processes whereby EKLF could play role as a repressor. We have recently found that EKLF exerts a negative effect upon the early stage of megakaryocytic differentiation, likely due to inhibition of critical megakaryocytic factors such as Fli-1 (16). Erythroid and megakaryocytic lineages share a common precursor, and the subsequent onset of committed erythro- or megakaryocytic progenitors depends on the balance between the transcription factors involved in this lineage determination. EKLF together with Fli-1, a megakaryocytic transcription factor, may be involved in cross-antagonistic interactions for the final lineage determination (16).

We tested for the possible involvement of sumoylation of EKLF in this process by using K562 cells. Although this is an erythroleukemic cell line, it can be induced to differentiate towards a megakaryocytic phenotype in response to specific chemical inducers (56 and see the supplementary material). To this end, we prepared stable K562 cell lines with either WT or K74R EKLF under the control of a zinc-inducible promoter. These were treated with ZnCl₂ for 18 h prior to induction of megakaryocytic differentiation with TPA and monitored for 4 days. Using flow cytometry, we measured the level of cell surface antigen CD41 as a marker specific for megakaryocytic development. The flow cytometry results show that initially both cell lines yield comparable histogram profiles for CD41 (Fig. 7A). However, the subsequent dynamics of megakaryocytic differentiation are significantly different in the two lines, as both the kinetics and extent of CD41⁺ accumulation are faster and greater in the SUMO-deficient mutant of EKLF than the WT (Fig. 7A and B). This suggests that the K74R EKLF is unable to inhibit megakaryocytic differentiation as efficiently as WT EKLF.

In parallel, we carried out qPCR analyses of the mRNA levels for EKLF, GPIIb (glycoprotein IIb, another megakaryocyte-specific marker), and Fli-1, which further confirm difference between WT and K74R EKLF (Fig. 7C). Overexpression of WT EKLF decreases normal Fli-1 message levels during megakaryopoiesis of embryonic stem cells in culture (16). In the present experiment, although the mRNA levels of the K74R mutant are more pronounced than those of WT EKLF at simultaneous time points (Fig. 7C), the levels of expression of megakaryocytic markers GPIIb and Fli-1 are much higher for K74R mutant than in the presence of the lower level of WT EKLF (Fig. 7C). These data suggest a significant functional consequence of the mutation in the SUMO acceptor site of EKLF: an inability to exert its inhibitory effect on megakaryocytic differentiation.

K562 cells can be also differentiated towards an erythroid phenotype. However, erythroid differentiation after hemin treatment of the induced WT or K74R EKLF stable K562 cell lines was unaffected, as judged by benzidine staining and analysis by RT-qPCR of γ-globins (data not shown). This is not surprising as γ-globins are not the natural target of EKLF (12).

The data obtained from differentiation of K562 cells with TPA led us to postulate that the constitutive expression of WT or K74R EKLF will not have equivalent effects in vivo within the megakaryocytic lineage, where EKLF is no longer expressed (16). To examine this prediction we cloned EKLF downstream of the megakaryocyte-specific PF4 promoter and established transgenic lines from these DNA constructs. PF4 is a protein expressed in megakaryocyte granules whose promoter has been successfully used in a number of studies to drive restricted expression (61). Lines that expressed the transgenes in the bone marrow were examined in more detail by cellular and molecular analyses, and results from these lines are more fully described below.

Lineage-depleted bone marrow cells were cytospun onto slides and monitored for the presence of megakaryocytes by acetylcholinesterase staining. Visual inspection of the field made it readily apparent that, already at harvest, bone marrow cells from transgenic mice expressing WT EKLF contain significantly less megakaryocytes than those from a nontransgenic line (Fig. 8A). Quantitation of MGG-stained slides revealed a threefold decrease in megakaryocyte cellularity in WT EKLF transgenic bone marrow (Fig. 8A). This demonstrates that misexpression of EKLF in the megakaryocyte lineage has a repressive effect on megakaryocyte formation, consistent with the present studies in K562 cells and with our recently published analyses (16). However, of particular relevance to EKLF sumoylation, visual inspection and quantitative analysis of transgenic bone marrow from the K74R EKLF line revealed no effect of mutant EKLF transgene expression on megakaryocyte cellularity (Fig. 8A).

We next determined whether transgene expression also altered megakaryocyte colony formation. Megacult colony assays (Fig. 8B) of lineage-depleted bone marrow cells show a similar trend to that seen with the cell assay: WT EKLF transgenic bone marrow contains fivefold less megakaryocyte colony-forming potential than the nontransgenic control, while the K74R-EKLF transgenic bone marrow is not affected (Fig. 8B). CFU-erythroid (CFU-E) colony formation is not significantly affected in any of these bone marrow samples. Quantitative RT-PCR analysis shows that K74R EKLF is expressed at an even higher level than that of WT EKLF (Fig. 8C), thus ruling out the trivial possibility that its lack of effect on megakaryocyte cell number and colony formation is a result of absent transgene expression. As expected, total EKLF levels are unaffected in any of the samples. Together with our data from K562 cells, we conclude that availability of the SUMO modification site in EKLF is absolutely critical for it to exert its normal cellular function during hematopoiesis related to inhibition of megakaryopoiesis.