SENP1-mediated GATA1 deSUMOylation is critical for definitive erythropoiesis

We generated SENP1^lox/lox mice based on homologous recombination. We first generated global deletion of SENP1 by mating SENP1^+/lox mice with β-actin–Cre mice to obtain SENP1^+/− mice, with deletion of both the targeted region (exons 5 and 6) and the Neo gene (Fig. 1, A–C). SENP1^+/− mice appeared normal and fertile. We attempted to generate SENP1^−/− (SENP1 KO) mice by intercrossing SENP1^+/−, and no live SENP1 KO mice were obtained after weaning. We then carefully examined mouse genotyping and phenotypes at E13.5 and 16.5 of embryonic development and on the day of birth (postnatal day [P] 0). Typical genotyping of SENP1^+/+ (WT), SENP1^+/− (heterozygote), and SENP1^−/− (KO) mice are shown in Fig. 1 D. Deletion of SENP1 in SENP1 KO mice was verified by RT-PCR using primers amplifying exons 5–6 and by Western blotting with an anti-SENP1 antibody recognizing the C-terminal domain of SENP1. Neither SENP1 transcripts encoding for exons 5–6 (not depicted) nor SENP1 protein was detected in SENP1 KO E13.5 fetal liver (Fig. 1 E). The ratios of live and dead SENP1 KO embryos were analyzed (Fig. 1 F). Only three SENP1 KO littermates among 384 born pulps survived to 2 wk of age, and each had severe anemia (unpublished data). We focused on E13.5–16.5 to further characterize the SENP1 KO phenotype.

Starting at E13.5, SENP1 KO embryos and fetal livers were paler and smaller than those of WT and heterozygote. Both the SENP1 KO embryo and yolk sac have poor blood vessel development with fewer red blood cells, indicating anemia in SENP1 KO (Fig. 2 A). Erythropoiesis is initiated in the yolk sac at around E8.0 during normal mouse development (primitive erythropoiesis; Orkin and Zon, 2008). Starting around E11 until 1 wk after birth, the fetal liver is the major erythropoietic organ. Fetal liver in SENP1 KO was smaller in size (Fig. 2 A) and lighter in weight (Fig. 2 B). Histological examination of fetal liver sections revealed a marked decrease in the number of morphologically identifiable small-sized erythroid cells (erythroblasts) with condensed chromatin nucleus in SENP1 KO fetal liver (Fig. 2, C and D [quantification]) as well as by FACS with erythrocyte marker Ter119 (Fig. 2 E). Although decreased erythropoietic islands were clearly discernable in SENP1 KO fetal livers, fetal liver touch preparations by May-Giemsa staining analysis showed a disproportionate abundance of immature erythroblasts, such as proerythroblasts, compared with WT mice (Fig. 2 F).

To further determine if SENP1 KO mice display defects in erythropoiesis leading to anemia, hemoglobin levels in E13.5 fetal livers from SENP1 WT and SENP1 KO mice were analyzed by quantitative RT-PCR. The SENP1 KO fetal liver showed a marked decrease in adult forms of hemoglobin (Hbb-a1 and Hbb-b1) with a dramatic increase in the corresponding embryonic forms (Hbb-x and Hbb-y; Fig. 2 F, ratios of Hbb-a1/Hbb-x and Hbb-b1/Hbb-y). To determine the stage of erythroid differentiation at which SENP1 KO develop defects, we compared the abilities of fetal liver cells derived from WT and SENP1 KO E13.5 embryos to form CFU-e and more immature BFU-e. SENP1-deletion had no effects on the number of the BFU-e (Fig. 2 G, BFU-e). However, the number of CFU-e showed a significant (40%) decrease in SENP1 KO fetal liver cells (Fig. 2 G, CFU-e). Similar reduction in megakaryocyte CFU (CFU-mk), the closest differentiation pathway with erythropoiesis, was observed in SENP1 KO (Fig. 2 G, CFU-mk). In contrast, granulocyte/monocyte CFU (CFU-G/M) was not significantly altered (Fig. 2 G, CFU-G/M). Interestingly, SENP1 KO showed an increase in CFU-G concomitant with an increase in CFU-M (Fig. 2 G, CFU-G and CFU-M). This in vitro observation is consistent with the in vivo results that monocytes in SENP1 KO transplanted recipient mice are increased compared with WT mice (Fig. 3 C).

To further determine if SENP1 KO erythroid progenitor cells exhibit an intrinsic defect in erythropoiesis, we measured erythropoietic activity of fetal liver cells from WT and SENP1 KO after transplantation into myeloablated adult WT mice. To this end, E13.5 fetal liver cells of WT or SENP KO mice (CD45.2⁺) were respectively injected into irradiated Pep3B recipients (CD45.1⁺). The engraftment ratio in BM of recipient mice after fetal liver transplantation was analyzed by flow cytometry analysis using CD45.2 and CD45.1 surface marker after 6 mo. The results suggest that engraftment could reach >95%, and no difference was detected between WT and SENP1 KO fetal liver cells in the reconstitution of recipient BM (Fig. 3 A, percentage of CD45.2⁺ cells). We confirmed the genotype of the donated fetal liver cells 6 mo after BM transplantation (BMT) by genotyping blood samples with WT or SENP1 KO–specific primers (unpublished data). We then measured precursor cells in BM of recipient mice by flow cytometry with specific antibodies. Erythroid cells (TER-119⁺/CD45.2⁺ ratio) were significantly decreased in mice receiving KO versus WT fetal liver cells, whereas there was no significant difference in granulocyte maturation for KO versus WT donors (Gr-1⁺CD45.2⁺ ratio; Fig. 3 B). We also performed complete blood counts (CBCs) of the peripheral blood from WT, WT BMT, and KO BMT groups. Consistent with the decreased number of erythroid cells (TER-119⁺) and megakaryocytes (CD41⁺) in their BM, SENP1 KO recipients showed decreases in hemoglobin, hematocrit, and number of red blood cell compared with the recipients of WT cells. However, lymphocytes and monocytes were increased (Fig. 3 C). These results suggest that SENP1 KO fetal liver transplantation confers an anemic phenotype. It is of note that the effects of SENP1 deletion on transplanted mice were less than those of fetal liver, suggesting a potential compensatory event in recipient mice. Indeed, we observed enlarged spleens and extramedullary hematopoiesis with increased erythroid precursor clumps, megakaryocytes and other developing myeloid populations in the SENP1 KO transplanted mice (Fig. S1).

To analyze the molecular determinants of the SENP1 KO erythropoietic phenotype, we examined protein expression of several critical factors regulating erythropoiesis. Consistent with previously published results (Cheng et al., 2007), protein expression of Epo and its upstream regulator HIF1α was reduced in SENP1 KO fetal liver (Fig. S2). Notably, erythropoietic transcriptional factor GATA1 and its downstream target erythropoietin receptor (EpoR) were also reduced in protein expression (Fig. 4 A). GATA2, a transcriptional factor involved in primitive erythropoiesis (Tsai et al., 1994), was not significantly altered by SENP1 deletion. Furthermore, GATA1 activity in fetal liver was reduced (more than threefold) as determined by electrophoretic mobility shift assay (EMSA) using equivalent GATA1 protein input from fetal liver nuclear extracts (Fig. 4 B).

To examine if reduction of GATA1-responsive gene expression is not a result of decreased total numbers of erythroid cells in SENP1 KO, we isolated stage-matched erythroid cells by FACS sorting with anti-CD71 and anti-Ter119. Freshly isolated E13.5 WT or SENP1 KO fetal liver cells were doubly stained with anti-CD71 and anti-Ter119. Based on these two surface markers, the proportion of erythroid progenitors at different stages of differentiation, termed R1–R5, can be assessed as described previously (Zhang et al., 2003). In SENP1 KO fetal liver, a twofold increase in number of less mature erythroblasts (R1 and R2) and concomitant twofold decrease of more mature erythroblasts (R3–R5) were detected (Fig. 4 C), which is consistent with increased proerythroblasts in SENP1 KO fetal liver visualized by May-Giemsa staining (Fig. 2 F). We first examined GATA1 expression in these stage-matched erythroid cells. WT and SENP1 KO showed no differences in GATA1 transcripts (stage R1–R3) and protein expression (stage R2–R3; Fig. S3). We then compared GATA1-responsive genes in the R1 and R2 populations isolated from WT and SENP1 KO mice by quantitative RT-PCR. Transcripts for GATA1-inducible genes EpoR and Hbb-y (an embryonic hemoglobin expressed in R1 and R2 populations) were decreased in SENP1 KO cells. In contrast, GATA1-repressive genes, such as c-Myb and c-Kit, were up-regulated in SENP1 KO cells (Fig. 4 D, data from R2 populations). Similar patterns of gene expression were obtained from R1 populations but not in more mature R3 populations (unpublished data). These data suggest that GATA1 activity and its dependent gene expression were reduced in SENP1 KO erythroid progenitors, causing a defect in erythropoietic maturation and subsequent decrease in total number of erythroid cells observed in fetal liver. Decrease of GATA1 expression observed in SENP1 KO fetal liver (Fig. 4 A) is a result of a smaller number of total erythroid cells (Fig. 2, C–E).

SENP1 is an endopeptidase that removes SUMO from conjugated protein targets. We reasoned that deletion of SENP1 leads to GATA1 SUMOylation, reducing GATA1 activity. To test our hypothesis, we first examined if a SUMOylated form of GATA1 could be detected in SENP1 KO fetal liver in the presence of deSUMOylase inhibitor. The SUMOylated form of GATA1 was dramatically increased in SENP1 KO fetal liver, which was confirmed by immunoprecipitation with anti-SUMO followed by Western blotting with anti-GATA1 (Fig. 5 A). To determine if SENP1 directly deSUMOylates GATA1, GATA1 WT, or a SUMO-defective mutant, GATA1-K137R was coexpressed with SENP1 WT or a catalytic inactive form with a C603A mutation (SENP1 CA). As expected, SUMOylated GATA1 could be detected for GATA1 WT but not for the GATA1-K137R mutant. Importantly, SUMOylated GATA1 was diminished by SENP1 WT. In contrast, SENP1 CA increased SUMOylated GATA1 (Fig. 5 B), suggesting that the catalytic inactive form of SENP1 may function as a dominant-negative mutant.

To directly test the role of SENP1-mediated GATA SUMOylation in erythropoiesis, we determined effects of SENP1 CA on erythropoiesis in a human erythroleukemic cell line (HEL), which undergoes erythropoiesis in the presence of hemin (Martin and Papayannopoulou, 1982). HEL cells were infected with adenovirus expressing LacZ or SENP-C603A followed by treatment with hemin for 96 h, and erythropoiesis was measured for hemoglobin-positive cells by benzidine staining. Infection efficiency was >90% as measured by β-galactosidase staining (unpublished data). SENP1 CA significantly blocked hemin-induced erythropoiesis as indicated by a reduced number of benzidine-positive cells (Fig. 6, A and B [quantification]). Similar results were obtained for expression of EpoR and β-hemoglobin by quantitative RT-PCR (unpublished data).

It has been previously reported that erythropoietic deficiency of SENP1 mutant mice is partly a result of inappropriate HIF1α activity (Cheng et al., 2007), and we also observed reduction of HIF1α expression in SENP1 KO fetal liver (Fig. S2). We next determined relative contributions of GATA1 and HIF1α, two erythropoietic transcription factors regulated by SUMOylation, to SENP1-mediated erythropoiesis. To this end, SENP1 CA was coexpressed with GATA1 or HIF1α into HEL cells (Fig. S4 A, protein expression of the transgenes), and erythropoiesis was measured for hemoglobin-positive cells by benzidine staining. Results showed that coexpression of GATA1 or HIF1α partially, and together additively, rescued SENP1 CA–induced hematopoietic defects (Fig. 6 c). These data suggest that both GATA1 and HIF1α contribute to SENP1-mediated erythropoiesis.

Recruitment of GATA1 to the GATA1 site in the EpoR promoter is critical for EpoR expression and subsequent erythropoiesis (Zon et al., 1991). To directly determine the effects of SENP1 CA on GATA1 activity and EpoR transcription in HEL cells, we measured GATA1 binding to the EpoR promoter in SENP1 CA–infected cells by chromatin immunoprecipitation (ChIP) assay. HEL cells were infected and treated with hemin for 96 h, and nuclei were subjected to ChIP with anti-GATA1 followed by a PCR amplification of the GATA1 binding site in the EpoR promoter. An isotype IgG and a region on exon 8 were used as a negative control for immunoprecipitation and PCR amplification, respectively (Fig. 6 D). Expression of SENP1 had no effect on the total level of GATA1 (Fig. 6 E). However, expression of SENP1 CA significantly blunted the binding of GATA1 to the EpoR promoter region (Fig. 6, F and G [quantification], representative ChIP gel). These data strongly support a critical role of SENP1 in controlling GATA1 recruitment to the EpoR promoter and subsequent erythropoiesis.

Reduction of GATA1 recruitment to EpoR promoter in SENP1 KO cells could be a result of alterations of SUMOylated GATA1 in protein stability, nuclear localization, and protein and/or DNA-binding activity as demonstrated for other transcriptional factors (Müller et al., 2001; Seeler and Dejean, 2003; Gill, 2004), and we next examined if SUMO conjugation affects these activities of GATA1. SUMOylation, like other posttranslational modification, is a dynamic process. Only a small proportion of a given SUMOylated protein can be detected (Fig. 5), and it has been difficult to determine a suppressive effect of SUMO conjugation on a specific protein. To get around this issue, we used a SUMO fusion strategy, which has been applied to other SUMOylated proteins such as p53 and GATA4 (Wang et al., 2004; Carter et al., 2007). SUMO fusion to the N terminus of GATA1 (Fig. 7 A) had no effects on its expression or its stability as measured for its half-life in the presence of protein synthesis inhibitor cycloheximide (Fig. S5 A). GATA1 WT K137R with or without SUMO fusion showed a similar pattern of nuclear localization (Fig. 7 B; and Fig. S6, A and B [quantification]). However, SUMO fusion completely diminished the DNA binding activity of GATA1 as measured by EMSA using a GATA1 probe (Fig. 7 C).

Finally, we determined the functional significance of GATA1 SUMOylation in erythropoiesis. To this end, HEL cells were infected with retrovirus encoding GATA1 WT, GATA1-K137R, SUMO-GATA1, or SUMO-GATA1K137R (Fig. S7 A, expression). Erythropoiesis was measured for hemoglobin-positive cells by benzidine staining as described in Fig. 6, except with no hemin treatment. Expression of GATA1 WT, induced erythropoiesis, and expression of KR strongly induced erythropoiesis, as indicated by a reduced number of benzidine-positive cells. In contrast, SUMO fusion proteins (GATA1 WT and KR) failed to induce erythropoiesis (Fig. S7, B and C [quantification]).

To eliminate any endogenous GATA1 effects, we reconstituted various GATA1 forms into G1E cells, a GATA1-null erythroid precursor cell line defective in erythropoiesis (Welch et al., 2004). G1E cells were infected with retrovirus encoding GATA1 WT, GATA1-K137R, SUMO-GATA1, or SUMO-GATA1K137R. Expression of GATA1 proteins was determined by Western blotting (Fig. 8 A). Transcripts of GATA1 responsiveness were determined by quantitative RT-PCR. Overexpression of GATA1 WT and GATA1-KR strongly enhanced transcription of the GATA1-inducible genes (EpoR and β-hemoglobin) but reduced the GATA1-repressive genes (Myb). In contrast, SUMO conjugation of GATA1 diminished GATA1 responsiveness of these genes (Fig. 8 B).

To directly determine if diminished GATA1 responsiveness of gene expression is the result of a blocked recruitment of SUMO-GATA1 to the GATA1-responsive genes, we measured the binding of GATA1 proteins to the EpoR and β-hemoglobin promoters in GATA1-infected G1E cells by ChIP assay. The GAPDH gene promoter was used as a negative control (Fig. 8 C). Similar to Fig. 6, nuclei were subjected to ChIP with anti-GATA1 followed by a PCR amplification of the GATA1 binding site in the EpoR and β-hemoglobin promoters. An isotype IgG was used as a negative control for immunoprecipitation. Consistent with the DNA binding activity and the gene induction of GATA1 proteins, both GATA1 WT and GATA1-K137R bound to the EpoR promoter region, and GATA1-K137R was stronger than GATA1 WT as expected. However, SUMO-GATA1 failed to be recruited to the EpoR and β-hemoglobin promoters (a representative ChIP gel from the EpoR gene is shown in Fig. 8 D with quantification in Fig. 8 E).

In this paper, we report that we have created SENP1 KO mice based on a homologous recombination system. The SENP1 KO mouse tissues express neither the SENP1 transcript encoding exons 4–5, as verified by RT-PCR, nor the SENP1 protein, as verified by Western blotting with anti-SENP1 antibody recognizing the C-terminal catalytic domain. There have been two studies on SENP1 mutant mice, and mice from both these studies were derived from ES cell lines with a retroviral vector that had been randomly inserted into the enhancer region or the coding region of the SENP1 gene, causing inactivation of the gene (Yamaguchi et al., 2005; Cheng et al., 2007). Consistent with our SENP1 KO mice, these SENP1 mutant mice show severe fetal anemia. Cheng et al. (2007) have also demonstrated that SENP1, via a direct deSUMOylation of HIF1α, regulates HIF1α stability and HIF1α-dependent genes, including Epo, causing defects of erythropoiesis. Consistent with this finding, we also observed that HIF-1α protein and Epo gene expression were reduced in our SENP1 KO mice and that overexpression of HIF1α could partially rescue a hematopoiesis defect induced by SENP1 deletion. Importantly, we have found that GATA1 is another important target of SENP1 during hematopoiesis (Fig. 8 F).

The evidence for the role of SENP1 in regulation of GATA1 activity is compelling. The quantitative RT-PCR analysis for mRNA, Western blotting for protein expression, and EMSA for the DNA binding activity clearly show that the GATA1 activity is down-regulated in fetal liver of the SENP1 KO mice. We found that a SUMOylated form of GATA1 is accumulated in SENP1 KO mice, correlating with down-regulation of GATA1-dependent genes including EpoR and β-hemoglobin. We further show that SENP1 can directly deSUMOylate GATA1, modulating GATA1-dependent EpoR expression and erythropoiesis by in vitro assays. Gene-targeting and loss-of-function studies have proved that GATA1 plays an essential role in erythropoiesis and megakaryopoiesis (Yamamoto et al., 1990; Weiss and Orkin, 1995). Indeed, SENP1 KO mice show defects in both processes in fetal liver and in the BM with transplanted fetal liver cells. We observed increased lymphocytes and monocytes in SENP1 KO fetal liver–transplanted recipient. This is consistent with the data that GATA1 deficiency could enhance multipotential differentiation ability of erythroid-committed cells (Kitajima et al., 2006). SENP1 KO mice show a milder anemia phenotype than GATA1-deficient mice, which die at E10.5 as a result of severe anemia with arrest of erythroid maturation. This is expected because GATA1 in SENP1 KO cells are not entirely SUMOylated. Therefore, GATA1 activity and GATA1-dependent gene expression is reduced, but not completely eliminated, in SENP1 KO. Therefore, SENP1 KO mice show a similar phenotype to GATA1 knockdown (GATA1.05 and GATA1-low mice, which express only ∼5–20% of the WT GATA1 levels, respectively; McDevitt et al., 1997; Takahashi et al., 1997). GATA1 SUMOylation is balanced by SUMO ligases and SENPs. Further investigations are needed to identify the SUMO ligases responsible for GATA1 SUMOylation and the efficiency of the ligase on GATA1. There are six members of the SENP family (SENP1–3 and SENP5–7), and each SENP has different cellular locations and, likely, has different substrate specificities (Gong et al., 2000; Kim et al., 2000; Nishida et al., 2000; Best et al., 2002; Hang and Dasso, 2002; Kadoya et al., 2002; Zhang et al., 2002; Bailey and O’Hare, 2004). Whether or not other SENPs members could deSUMOylate GATA1 during erythropoiesis still needs to be determined.

The function of GATA1 can be regulated by an array of posttranslational modifications, including phosphorylation and acetylation as well as SUMOylation (Boyes et al., 1998; Cantor and Orkin, 2002; Collavin et al., 2004; Yu et al., 2005; Hernandez-Hernandez et al., 2006; Lamonica et al., 2006; Zhao et al., 2006). Phosphorylation of GATA1 and its role in regulation of GATA1 function has been extensively studied. GATA1 can be strongly phosphorylated at residue Ser26 of the N-terminal acidic activation domain by MAPK and at Ser310 of the DNA-binding domain by Akt in response to growth factors such as Epo (Yu et al., 2005; Zhao et al., 2006). Phosphorylation of GATA1 increases GATA-1–mediated transcription without significantly changing the DNA-binding affinity of GATA1. However, GATA1 acetylation by cofactor p300 enhances its DNA-binding and recruitment to chromatin (Boyes et al., 1998; Hernandez-Hernandez et al., 2006; Lamonica et al., 2006). In contrast, the role of SUMOylation of GATA1 has not been well defined. GATA1 SUMOylation was initially identified based on its interaction with the SUMO E3 ligase PIASy (Collavin et al., 2004). Given the fact that only a small proportion (<5%) of GATA1 can be SUMOylated in cells, an inhibitory effect of SUMOylation on GATA1 cannot be readily detected. We have taken three different approaches to address this issue. First, we delete the SUMO endopeptidase SENP1 or express a dominant-negative form of SENP1 to promote GATA1 SUMOylation. This enhanced GATA1 SUMOylation dramatically represses the DNA binding and the transcriptional activities of GATA1 in erythropoiesis. Second, we use GATA1-null cells to show that a non-SUMOytable GATA1 mutant (K137R) has a higher activity than GATA1 in DNA-binding and induction of erythropoiesis. Finally, we demonstrate that SUMOylated GATA1, in a GATA1-SUMO fusion form, blunts the DNA-binding activity and subsequent recruitment to the EpoR and β-hemoglobin gene promoters and erythropoiesis. Collectively, we conclude that GATA1 SUMOylation inhibits its DNA-binding and erythropoietic activity.

Our study is the first demonstration of a functional significance of GATA1 SUMOylation. The exact mechanism by which SUMO modification inhibits GATA1 DNA binding and the subsequent recruitment to chromatin is currently unclear. It is conceivable that SUMOylation of GATA1, which is the opposite of acetylation, alters GATA1 conformation in a way that is unfavorable for DNA interaction. Whether or not SUMOylation of GATA1 in cells prevents GATA1 acetylation or its interactions with other coactivators are under investigation.

The SENP1 targeting arms were isolated from BAC clone identified by screening a Research Genetics (Invitrogen) BAC library using SENP1 cDNA as a probe. The targeting vector was constructed in a pEASY-flox backbone to contain a loxP site inserted upstream of SENP1 exon 5 and a neomycin cassette (Neo) flanked by two loxP sites down stream of exon 6 using standard molecular procedures. The linearized targeting construct was electroporated into 129/C57B/6 ES cells, and the targeted clones were selected with G418 and gancyclovir. Resistant clones were screened for homologous recombination by PCR and confirmed by Southern blot analysis. Two independent SENP1^lox ES cell clones were injected into C57BL/6 blastocysts. Chimeras were further bred with C57BL/6 female for germline transmission. SENP1^lox/lox mice were mated with β-actin–Cre mice to mediate a recombination in vivo resulting in a complete deletion of SENP1 exon 5 and exon 6 and a frameshift of the downstream catalytic domain.

Embryos or fetal tissues were fixed for immunohistochemical and hematoxylin/eosin staining in 4% paraformaldehyde. For histological analysis, tissues were embedded in paraffin, cut into 5-µM sections, and stained with routine hematoxylin/eosin. For immunohistochemical staining, tissue sections were stained with certain antibodies (e.g., anti-GATA1 and anti-EpoR) as described previously (He et al., 2006; Zhang et al., 2008). Bound primary antibodies were detected using avidin-biotin-peroxidase (NovaRed peroxidase substrate kit; Vector Laboratories). Pictures from four random areas of each section and five sections per mouse were taken using a digital camera (Kodak) mounted on a light microscope (Carl Zeiss, Inc.). Images were quantified using the Matlab software (The Math Works, Inc.) as described previously (He et al., 2006; Zhang et al., 2008).

Cells were prepared from murine E13.5 fetal livers. After lysis of mature red blood cells by PharmLyse (BD), cells were mixed with MethoCult M3334 to detect BFU-e or MethoCult M3434 to detect CFU-e, CFU-G, CFU-G, and CFU-G/M assays (STEMCELL Technologies Inc.). Cells were plated in 35-mm dishes and cultured at 37°C with 5% CO₂ for 3–8 d. For the CFU-e assay, benzidine-positive CFU-e colonies were scored on day 3. For BFU-e assay, benzidine-positive BFU-e colonies were scored on day 8. For MegaCult-C assays, fetal liver cells were used according to the manufacturer’s protocols (STEMCELL Technologies Inc.). 50 ng/ml human thrombopoietin, 50 ng/ml human IL-11, 10 ng/ml murine IL-3, and 20 ng/ml human IL-6 (PeproTech) were used in these assays. The cultures were incubated at 37°C with 5% CO₂ for 6–8 d. Colonies containing more than three megakaryocytes (AchE⁺ cells) were considered CFU-MKs (Cheng et al., 2009). Duplicate assays were performed for each mouse.

BMT was performed in lethally irradiated mice (irradiation 2× 5.5 Gy within 3 h) with WT or SENP1 KO fetal liver cells. Fetal liver cells were first isolated from E13.5 embryos of WT or SENP1 KO mice (CD45.2⁺), and red cells were removed by PharmLyse and subsequent washing with PBS. Cells were counted and 10⁵ cells were injected into the tail vein of the Pep3B recipient (CD45.1⁺). Successful BMT was controlled by FACS analysis of peripheral blood with anti-CD45.1 and anti-CD45.2 antibodies or by genotyping of SENP1 gene 6 wk after BMT. The mice were then subjected to FACS analysis for specific cell markers or CBC analysis. All animal studies were approved by the institutional animal care and use committees of Yale University.

To analyze erythrocytes, megakaryocytes, and other cell types, the fetal liver cells or BM cells were stained with standard protocols using PE-Ter119, PE-Gr1, PE-CD41a, PE-CD71, or APC-Ter119 (BD). FACS analysis was performed with an LSRII flow cytometer (BD) and FlowJo software (Tree Star, Inc.). Isotype controls were used in each experiment.

Mice were anesthetized with Methoxyflurane (Medical Developments International) followed by retroorbital bleeding (∼100 µl) using EDTA-treated glass capillary tubes. The blood was evacuated into tubes with 5 µl of 0.2M EDTA to prevent clot formation. The CBC of blood sample was performed using a Hemavet 950FS (Drew Scientific) according to the manufacturer’s protocol.

Mammalian expression plasmids for SENP1 WT were described previously (Li et al., 2008). Expression plasmid for SENP1-K137R was a gift from E. Yeh (The University of Texas M.D. Anderson Cancer Center, Houston, TX). Expression plasmids GATA1 and K137R mutant were gifts from C. Santoro (Università del Piemonte Orientale, Novara, Italy). Expression plasmid for EpoR (B. Forget, Yale University, New Haven, CT) SUMO-GATA1 fusion plasmid was constructed by cloning GATA1 into pSUMOstar vector (LifeSensors). Adenoviral vectors expressing flag-tagged SENP1-C603A (pAd-SENP1CA) and pAd-GATA1 was constructed as we described for pAd-AIP1 (Zhang et al., 2008). HIF1α adenovirus stock was a gift from F. Giordano (Yale University). All adenovirus was amplified in 293T cells and purified by ultracentrifugation. HEL cells were infected with adenovirus at MOI of 100 for 48–72 h before subsequent assays. Retroviral vector pMY.IRES.EGFP and ecotropic packaging cell line PLAT-E were gifts from T. Kitamura (Tokyo University, Tokyo, Japan). GATA-1 WT, GATA1-K137R, Sumo-GATA1 WT, and Sumo-GATA1-K137R cDNAs were inserted into the multicloning site of pMY.IRES.EGFP. The retrovirus supernatant was obtained by a transient transfection of the retroviral vectors to the PLAT-E packaging cell line using Lipofectamine 2000 (Invitrogen). 10⁶ G1E cells were resuspended in 2 ml retrovirus supernatant with 1 µg polybrene per ml and the infection was done by spinoculation (2 h, 25°C, 2,500 rpm).

GATA1-null G1E cell was a gift from M.J. Weiss, (University of Pennsylvania School of Medicine, Philadelphia, PA) and was cultured in IMDM supplemented with 15% fetal calf serum, 1:10,000 monothioglycerol (Sigma-Aldrich), human recombinant erythropoietin (R&D Systems) at 2 U/ml, and kit ligand (PeproTech) at 50 ng/ml and penicillin/streptomycin at 37°C in 5% CO₂ in a humidified incubator as previously described (Weiss and Orkin, 1995). HEL cells were cultured in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal calf serum, 2 mM L-glutamine, and penicillin/streptomycin at 37°C in 5% CO₂ in a humidified incubator. To induce differentiation, 2 × 10⁵ cells per ml were cultured in the presence of 40 µM hemin (Sigma-Aldrich) for 3 d or as specified.

Cell lysis, protein preparation, and immunoblotting were performed as described previously (Li et al., 2008). For detection of SUMOylated proteins, 1 mM N-ethylmaleimide (deSUMOylase inhibitor) was added to cell lysates. Antibodies against anti-SENP1, GATA1, GATA2, EpoR, USF-2, and tubulin were obtained from Santa Cruz Biotechnology, Inc., anti-HIF1α was obtained from Novus Biologicals, and anti–Flag-POD was obtained from Sigma-Aldrich.

The double-stranded oligonucleotide containing a GATA consensus site from the EpoR gene promoter was used for EMSA. Preparation of nuclear extracts and EMSA were performed as previously described (Zhang et al., 1995). The ChIP assay was performed exactly as previously described (Letting et al., 2004; Ji et al., 2008) and analyzed by real-time PCR. All the primers are listed in Table S1. Antibodies included anti-GATA1 (ab1963; Abcam), a positive control (anti-RNA polymerase II), and a negative control (normal IgG).

Total RNA of fetal liver was isolated by using phenol/chloroform and isolated using RNeasy kit with DNase I digestion (QIAGEN). Reverse transcription was done by standard procedure (Super Script First-Strand Synthesis System; Invitrogen) using 1 µg of total RNA. Quantitative real-time PCR was performed by using iQ SYBR Green Supermix on iCycler Real-Time Detection system (Bio-Rad Laboratories). Specific primers for mouse GATA1, GATA2, EpoR, Epo, c-Myb, c-Kit, and various isoforms of hemoglobin genes were used. 18 S ribosomal RNA or β actin was used as internal control. All the primers are listed in Table S1. The relative amount of mRNA normalized with control group as indicated was quantified.

Fixation, permeabilization, and staining of cultured BAEC were performed as described previously (Li et al., 2008). Alexa Fluor 488 (green)–conjugated secondary antibodies (Invitrogen) were used. Confocal immunofluorescence microscopy was performed using a confocal microscope (TCS SP5; Leica), and the acquired images were transferred to Photoshop 6.0 (Adobe) to generate the final figures.

All data are expressed as means ± SEM. Statistical significance of differences between different conditions was performed using Prism (GraphPad Software, Inc.) with a two-tailed unpaired Student’s t test. A value of P < 0.01 was considered as statistically significant.

Fig. S1 shows enlarged spleen with extramedullary hematopoiesis in recipient mice with SENP1 KO fetal liver cells. Fig. S2 shows a decreased expression of HIF-1α and Epo in E13.5 of SENP1 KO fetal liver. Fig. S3 shows similar GATA1 expressions in stage-matched hematopoietic progenitors. Fig. S4 shows a rescue of SENP-CA–induced hematopoiesis by overexpression of GATA1 or HIF-1α. Fig. S5 shows no effects of SUMO conjugation on GATA1 protein stability. Fig. S6 shows a similar punctate staining from GATA1 WT and GATA1 mutants. Fig. S7 shows an inability of SUMO-GATA1 to induce erythropoiesis. Table S1 lists primers used in this study.

This work was supported by National Institutes of Health (NIH) grants HL077357, HL070295, HL65978, and HL085789 to W. Min. Funding was also provided by the Yale Center of Excellence in Molecular Hematology (NIH DK072442).

The authors have no conflicting financial interests.