|Home | About | Journals | Submit | Contact Us | Français|
GATA factors establish transcriptional networks that control fundamental developmental processes. Whereas the regulator of hematopoiesis GATA-1 is subject to multiple posttranslational modifications, how these modifications influence GATA-1 function at endogenous loci is unknown. We demonstrate that sumoylation of GATA-1 K137 promotes transcriptional activation only at target genes requiring the coregulator Friend of GATA-1 (FOG-1). A mutation of GATA-1 V205G that disrupts FOG-1 binding and K137 mutations yielded similar phenotypes, although sumoylation was FOG-1-independent, and FOG-1 binding did not require sumoylation. Both mutations dysregulated GATA-1 chromatin occupancy at select sites, FOG-1-dependent gene expression, and were rescued by tethering SUMO-1. While FOG-1- and SUMO-1-dependent genes migrated away from the nuclear periphery upon erythroid maturation, FOG-1- and SUMO-1-independent genes persisted at the periphery. These results illustrate a mechanism that controls trans-acting factor function in a locus-specific manner, and differentially regulated members of the target gene ensemble reside in distinct subnuclear compartments.
Posttranslational modifications establish versatile molecular switches that control fundamental cellular processes. A classic example is cAMP-dependent phosphorylation of CREB, which enhances its affinity for the coregulator and histone acetyltransferase CBP/p300 (Goodman and Smolik, 2000; Mayr and Montminy, 2001). By contrast to phosphorylation and other simple chemical modifications, certain posttranslational modifications involve the conjugation of small proteins, including ubiquitin or the related SUMO proteins, to recipient proteins. Ubiquitination functions to control proteasome-mediated proteolysis and also via proteasome-independent mechanisms (Kerscher et al., 2006). By controlling protein-protein interactions, sumoylation regulates diverse processes (Geiss-Friedlander and Melchior, 2007). Whereas transcription factor sumoylation can increase transcriptional repression by facilitating corepressor recruitment (Rosendorff et al., 2006), considerably less is known about how sumoylation promotes transcriptional activation (Gill, 2005; Lyst and Stancheva, 2007).
The consequences of signal-dependent posttranslational modifications can differ in distinct experimental and biological contexts. Given that transcription factors can utilize distinct mechanisms to regulate different loci (Bresnick et al., 2006; Kim and Bresnick, 2007), a modification might only affect a subset of the cognate target genes. GATA-1 (Evans and Felsenfeld, 1989; Tsai et al., 1989), which is crucial for controlling red blood cell, platelet, mast cell, and eosinophil production (Hirasawa et al., 2002; Migliaccio et al., 2003; Pevny et al., 1991; Shivdasani et al., 1997; Simon et al., 1992), activates or represses target genes with or without the cell type-specific coregulator FOG-1 (Crispino et al., 1999; Tsang et al., 1997), depending upon the locus (Johnson et al., 2007; Kim and Bresnick, 2007). As GATA-1 target genes regulated by FOG-1-dependent and -independent mechanisms are known (Kim and Bresnick, 2007), and the underlying mechanisms can be analyzed via genetic complementation in GATA-1-null cells (G1E) (Grass et al., 2003; Weiss et al., 1997), it is instructive to study GATA-1 as a model to understand how modifications affect distinct modes of activator function at endogenous loci.
The GATA-1 N-terminus contains multiple phosphorylation sites (Crossley and Orkin, 1994) and a single sumoylation site (K137) (Collavin et al., 2004). Deletion of amino acids 1–63, which removes two phosphorylation sites (Crossley and Orkin, 1994), is linked to the development of megakaryoblastic leukemia (Wechsler et al., 2002). Knock-in mice bearing Ser to Ala mutations of two sites within the N-terminus (S72 and S142) and also S310 (Zhao et al., 2006) exhibit normal steady-state erythropoiesis and responsiveness to acute stress in vivo (Rooke and Orkin, 2006). Deletion of amino acids 1–193 differentially affects target gene expression in G1E cells (Johnson et al., 2006). Since the 1–193 deletion eliminates K137, and protein-protein interactions requiring amino acids 1–193 are unknown, K137 might confer functionality to this important region.
The function of K137 sumoylation was initially assessed by expressing GATA-1 or GATA-1(K137R) in Xenopus embryos and scoring for globin and xGATA-1 mRNA as erythroid cell markers (Collavin et al., 2004). RT-PCR analysis revealed that GATA-1 and GATA-1(K137R) induced globin and xGATA-1 similarly. As GATA-1 and GATA-1(K137R) levels were not measured, and expression was assayed qualitatively at a single time, it is unclear whether GATA-1 and GATA-1(K137R) activities are similar or different. The SUMO ligase PIASy interacted with GATA-1 and increased K137 sumoylation upon overexpression of the factors in 293T cells (Collavin et al., 2004). In transiently transfected cells overexpressing GATA-1 and GATA-1(K137R) with PIASy, PIASy inhibited GATA-1- and GATA-1(K137R)-mediated transactivation of a luciferase reporter gene (Collavin et al., 2004).
Since GATA-1 is a critical regulator of hematopoiesis, it will be important to determine if K137 mediates endogenous target gene regulation, if similar levels of GATA-1 and K137 mutants function identically, and if K137 sumoylation accounts for the contribution of K137 to GATA-1 activity. Given the sophisticated knowledge of GATA-1 function (Kim and Bresnick, 2007), this is an attractive system for dissecting how sumoylation controls specific molecular steps instigated by a protein with dual activities to activate and repress genes. Using genetic complementation analysis, we demonstrate that K137 mediates GATA-1 function at certain, but not all, target genes, predominantly at FOG-1-dependent targets. Mutations of K137 and also V205 that mediates FOG-1 binding (Crispino et al., 1999) decreased chromatin occupancy at select sites, diminished FOG-1-dependent gene expression, and were rescued by tethering SUMO-1. While FOG-1- and SUMO-1-dependent genes migrated away from the nuclear periphery upon erythroid maturation, strikingly, FOG-1- and SUMO-1-independent genes persisted at the periphery. These results illustrate a mechanism in which sumoylation of a master regulator of hematopoiesis selectively controls its function at specific loci and constitutes the first example in which differentially regulated members of a target gene ensemble reside in distinct subnuclear compartments.
Conditional activation of an estrogen receptor ligand-binding domain fused to GATA-1 in GATA-1-null G1E cells recapitulates a normal window of erythropoiesis (Welch et al., 2004) and therefore represents a powerful system for dissecting GATA-1 function. We engineered G1E clonal lines stably expressing ER-GATA-1 or ER-GATA-1 with point mutations of K137 (Fig. 1A) at comparable levels (Fig. 1B). Whereas ER-GATA-1 migrated as major and minor ~75 and ~105 kDa bands, respectively, the K137A and K137R mutants migrated solely as ~75 kDa bands. Thus, K137 is required for recovery of the ~105 kDa species, consistent with K137 sumoylation in 293T and mouse erythroleukemia cells (Collavin et al., 2004). Unlike ER-GATA-1, ER-GATA-1(K137R) was incapable of inducing erythroid maturation (Fig. 1C).
To determine how K137 contributes to GATA-1 function at endogenous loci, we quantitated GATA-1 target gene expression after estradiol-mediated activation of ER-GATA-1 or the K137 mutants. Estradiol lacks activity in G1E cells devoid of ER-GATA-1, and estradiol-induced transcription in G1E-ER-GATA-1 cells reflects GATA-1 activity to control its physiological target genes (Johnson et al., 2007; Johnson et al., 2006; Welch et al., 2004). Even though the K137 mutants and ER-GATA-1 were expressed at comparable levels (Fig. 1B), the mutants were considerably less active than ER-GATA-1 in inducing βmajor, α-globin, Ahsp, and Slc4a1 expression (Fig. 1D). Mutation of K137 also reduced the induction of Alas2, Hebp1, Ptdss2, Abcb10, and Tac2, (Fig. 1D). The mutants were at least as effective as ER-GATA-1 in inducing Csf2rβ2, Fog1, Epb4.9, and Eklf, (Fig. 1D). Thus, K137 differentially mediates GATA-1 activity at endogenous loci. We also analyzed the activity of GATA-1 and GATA-1(K137R) lacking ER after transient electroporation into G1E cells (data not shown). However, the protein levels were considerably lower than ER-GATA-1, and therefore the induction of βmajor and Slc4a1 was lower (~4–5 fold). The K137R mutant lacking ER exhibited a significantly lower activity than GATA-1 (p = 0.024) in this assay.
Of the four known FOG-1-independent, GATA-1-activated genes (Crispino et al., 1999; Kim et al., 2007), Fog1, Epb4.9, and Eklf do not require K137 (Fig. 1D). As ER-GATA-1 activates Tac2 in a FOG-1-independent manner and subsequently represses it in a FOG-1-dependent manner (Johnson et al., 2006), Tac2 differs from other FOG-1-independent genes. Of the FOG-1-dependent genes (βmajor, α-globin, Ahsp, Slc4a1, Alas2, Hebp1, Ptdss2, Abcb10, and Csf2rβ2), all except Csf2rβ2 require K137 (Fig. 1D). Although GATA-1 activates Csf2rβ2 in a FOG-1-dependent manner (Kim et al., 2007), nothing is known about its transcriptional mechanism that distinguishs it from other FOG-1- and K137-dependent genes.
GATA-1 also represses genes through FOG-1-dependent and -independent modes (Kim and Bresnick, 2007). We tested whether K137 is required to repress the FOG-1-dependent targets Gata2 (Grass et al., 2003; Pal et al., 2004) and c-kit, and the FOG-1-independent target Lyl1 (Johnson et al., 2007). K137A and K137R mutants were considerably less effective than ER-GATA-1 in repressing Gata2, c-kit, and Lyl1 (Fig. 1D), indicating that K137 is not solely required for activation.
Although GATA-1 is sumoylated at K137 (Collavin et al., 2004), lysines can also be acetylated and methylated (Yang and Seto, 2008). The K137 requirement for context-dependent GATA-1 activity might therefore reflect such modifications. To determine if ER-GATA-1 is sumoylated in G1E-ER-GATA-1 cells, we analyzed lysates under conditions that restrict SUMO protease activity. In estradiol-induced G1E-ER-GATA-1 cells, ER-GATA-1 migrated as ~75 and ~105 kDa species, with the ~75 kDa species being much more abundant (Fig. 2A). Immunoprecipitation with anti-SUMO-1 antibody enriched the ~105 kDa species, with its abundance exceeding that of the ~75 kDa species. Taken together with the report of K137 sumoylation (Collavin et al., 2004), these results indicate that SUMO-1 conjugation to ER-GATA-1 forms the ~105 kDa species.
FOG-1-dependent and -independent activated targets were largely sensitive and insensitive, respectively, to the K137 mutation. The preferential K137 requirement for GATA-1-mediated, FOG-1-dependent activation highlights an unexpected link between FOG-1 and K137. The GATA-1 N-terminal zinc finger contacts multiple FOG-1 zinc fingers (Cantor et al., 2002; Crispino et al., 1999; Liew et al., 2005), and no evidence exists that K137 and/or nearby sequences modulate FOG-1 binding. This link might reflect a FOG-1 requirement for K137 sumoylation or for sumoylated K137 function. As ER-GATA-1 is sumoylated in three clonal lines of FOG-1-null hematopoietic precursor cells (Cantor et al., 2002; Johnson et al., 2007) (Fig. 2B), FOG-1 is not essential for K137 sumoylation, and K137 is not required for ER-GATA-1 to bind endogenous FOG-1 in a co-immunoprecipitation assay (Fig. S1).
To determine if sumoylation underlies the differential K137 requirement for GATA-1 activity at distinct loci, we tested whether tethering SUMO-1 to ER-GATA-1(K137R) rescues its activity. We stably expressed a SUMO-1 fusion to ER-GATA-1(K137R) in G1E cells (Fig. 2C). Multiple clonal lines were generated to permit comparisons of clones expressing ER-GATA-1(SUMO-1/K137R) and ER-GATA-1(K137R) at comparable levels (Fig. 2C). Quantitative analysis of target gene expression indicated that ER-GATA-1(SUMO-1/K137R) activity far exceeded that of ER-GATA-1(K137R) at βmajor, α-globin, Ahsp, and Slc4a1 (Fig. 2D). By contrast, ER-GATA-1(SUMO-1/K137R) and ER-GATA-1(K137R) similarly induced Csf2rβ2, a K137-insensitive gene (Fig. 1D).
We developed an assay to measure the capacities of transiently expressed ER-GATA-1 to regulate endogenous targets in G1E cells. Using the Nucleofector system, ER-GATA-1, ER-GATA-1(K137R), and ER-GATA-1(SUMO-1/K137R) were readily expressed (Fig. 2E). Although ER-GATA-1(K137R) was slightly higher than that of ER-GATA-1 (Fig. 2E), it activated βmajor, α-globin, and Slc4a1 considerably less than ER-GATA-1 (Fig. 2F). ER-GATA-1(SUMO-1/K137R) activity was comparable to or greater than that of ER-GATA-1 (Fig. 2F), despite its expression being slightly lower than ER-GATA-1 (Fig. 2E). These results recapitulate conclusions from the stable assay (Figs. 1D and and2D).2D). ER-GATA-1 sumoylation at K137, and SUMO-1-mediated rescue of ER-GATA-1(K137R) activity indicate that sumoylation controls GATA-1 activity at select loci.
As an alternative approach to test whether sumoylation underlies the K137 requirement, we mutated E139 neighboring K137, which is a component of the sumoylation consensus. The E139A mutant was not sumoylated (Fig. 2G, top), and E139 resembled K137 in being required for ER-GATA-1-mediated activation (Fig. 2G, bottom). In aggregate, analyses with K137 and E139 mutants and the SUMO-1 fusion protein indicate that sumoylation of K137 explains the critical function of this residue. Even though the single major isoform of ER-GATA-1 is abrogated by mutating K137, anti-SUMO-1 antibody immunoprecipitated the isoform, tethering SUMO-1 rescued the K137 mutation, and mutating the sumoylation consensus dysregulated gene expression analogous to the K137 mutation, one cannot unequivocally rule out the possibility that K137 is subject to additional modifications in distinct contexts.
As certain GATA-1 target genes are insensitive to K137 mutations, presumably the K137 mutants localize to the nucleus. However, since GATA-1 target genes differ in the GATA-1 level required for regulation (Johnson et al., 2006), a change in nuclear or subnuclear localization might dysregulate select targets. Biochemical (Fig. 2H) and cell biological (Fig. 2I) analyses indicated that both ER-GATA-1 and ER-GATA-1(K137R) localized in the nucleus. ER-GATA-1 localized in speckles throughout G1E nuclei (92% of cells), analogous to endogenous GATA-1 (Elefanty et al., 1996). ER-GATA-1(K137R) localized in a ring-like pattern in 41% of the cells, with the least immunoreactivity centrally in the nucleus (Fig. 2I). Immunostaining with anti-nucleolin indicated that in certain cases nucleoli resided centrally in the ring (data not shown). The ring-like distribution was observed in only 8% of ER-GATA-1-expressing cells. Transiently expressed GATA-1 and GATA-1(K137R) lacking ER also localized differentially in G1E cells (Fig. S2).
The aberrant subnuclear localization of ER-GATA-1(K137R) might limit its availability for target gene occupancy. Quantitative ChIP analysis was conducted to compare chromatin occupancy by ER-GATA-1 and the K137 mutants at upstream regulatory regions of the β-globin (HS3 and HS2) (Johnson et al., 2002) and α-globin loci (HS-26) (Anguita et al., 2002; Pal et al., 2004), as well as the βmajor (Johnson et al., 2002), α-globin (Anguita et al., 2004), and Ahsp promoters (Pilon et al., 2006). The Ey promoter of the β-globin locus, which is hypoacetylated in definitive erythroid cells (Bulger et al., 2003; Forsberg et al., 2000) and not occupied by GATA-1 (Johnson et al., 2002), was used as a negative control. ER-GATA-1 and the mutants occupied HS3 indistinguishably (p = 0.178). ER-GATA-1(K137R) and ER-GATA-1(K137A) occupied HS2 29% and 39% less (p = 0.031) and the βmajor promoter 56% and 75% less (p = 0.011), respectively, than ER-GATA-1 (Fig. 3A). The mutants occupied HS-26 slightly less than ER-GATA-1 (p = 0.128), whereas this defect was greater at α-globin and Ahsp promoters (p = 0.011 and 0.010, respectively) (Fig. 3B). While K137 was not essential for ER-GATA-1 chromatin occupancy, it enhanced occupancy at select sites, resembling FOG-1 facilitation of GATA-1 occupancy (greatest at the βmajor promoter, less at HS2, and irrelevant at HS3) (Letting et al., 2004; Pal et al., 2004).
Submaximal levels of K137 mutants at certain chromatin sites might disrupt all or perhaps select GATA-1-instigated molecular steps at these sites. K137 might also function post-chromatin occupancy. As GATA-1 induces Pol II recruitment to promoters and distal regulatory sequences (Johnson et al., 2001; Johnson et al., 2003; Johnson et al., 2002), we tested whether K137 mediates Pol II recruitment to the βmajor, α-globin, and Ahsp promoters and also to the constitutively active RPII215 promoter (Im et al., 2005). K137 mutations strongly reduced ER-GATA-1-mediated Pol II recruitment at βmajor, α-globin, and Ahsp (p = 0.006, 0.007, and 0.001, respectively) but not at RPII215 (Fig. 3C).
GATA-1 binds and recruits multiple coregulators (Blobel et al., 1998; Kim et al., 2007; Stumpf et al., 2006; Tsang et al., 1997). We tested whether K137 mediates FOG-1, CBP, and TRAP220 recruitment to the β-globin locus, as well as the α-globin and Ahsp promoters. At HS3, which the K137 mutants occupied normally (Fig. 3B), ER-GATA-1 and the mutants similarly recruited FOG-1 (Fig. 4A). At HS2, in which occupancy by the K137 mutants was reduced modestly (Fig. 3A), ER-GATA-1 and the mutants also similarly recruited FOG-1 (Fig. 4A). The mutants recruited little to no FOG-1 to the βmajor promoter (Fig. 4A), the site at which their occupancy was most reduced (Fig. 3A). At the α-globin and Ahsp promoters, the K137 mutation had little to no impact on FOG-1 recruitment (Fig. 4B). Thus, modest decreases (27–54%) in K137 mutant occupancy did not compromise FOG-1 recruitment, whereas a larger decrease (66%) correlated with loss of FOG-1 recruitment.
GATA-1 induces CBP occupancy at HS3 and the βmajor promoter, but CBP resides at HS2 independent of GATA-1 (Im et al., 2005; Letting et al., 2003). GATA-1-mediated CBP recruitment to HS3 was K137-independent, whereas K137 enhanced CBP recruitment to the βmajor (Fig. 4A), α-globin, and Ahsp promoters (Fig. 4B). At the β-globin locus, TRAP220 recruitment was K137-independent at HS3, modestly enhanced by K137 at HS2, and K137-dependent at the βmajor promoter (Fig. 4A). TRAP220 recruitment to α-globin and Ahsp promoters also required K137 (Fig. 4B). These results indicate that K137 mutants have the capacity to recruit all the coregulators, but not at all sites. If K137 is absolutely required for GATA-1 binding to a coregulator in all contexts, one would expect defective recruitment at all sites.
Since K137 sumoylation is preferentially required for FOG-1-dependent gene activation, the sumoylation contribution to GATA-1 activity might utilize V205, which contacts FOG-1 and is required for maximal FOG-1 recruitment (model 1). Alternatively, sumoylation might function independent of the V205 requirement (model 2) (Fig. 5A). If model 1 is valid, tethering SUMO-1 to a V205G mutant would have no functional consequence. If sumoylation bypasses the V205 requirement, tethering SUMO-1 to the V205G mutant should yield a mutant capable of activating FOG-1-dependent genes, despite the absence of the key residue mediating FOG-1 binding.
We tested whether SUMO-1 rescues V205G activity to regulate FOG-1-dependent target genes in the transient complementation assay. Although ER-GATA-1 and ER-GATA-1(V205G) (Fig. 5B) were expressed similarly (Fig. 5C), only ER-GATA-1 strongly activated βmajor and α-globin (Fig. 5D). Tethering SUMO-1 to V205G (Fig. 5B) rescued activity to 116% and 70% of the ER-GATA-1 value, respectively (Fig. 5D). Tethering SUMO-1 to a C261P mutant (Fig. 5E), which disrupts the integrity of the C-finger, abrogating DNA binding, did not rescue activity (Fig. 5F).
Since K137 regulates subnuclear localization and mediates FOG-1-dependent gene activation, we tested whether V205 also regulates subnuclear localization. ER-GATA-1(V205G) localized in a ring-like pattern (71% of the cells) (Fig. 6A), similar to ER-GATA-1(K137R) (Fig. 2I), whereas ER-GATA-1(SUMO-1/V205G) resembled ER-GATA-1. A V205G/K137R mutant exhibited the ring-like pattern in 79% of the cells (data not shown).
We tested whether tethering SUMO-1 to ER-GATA-1(V205G) rescues activity via enhancing FOG-1 binding. Multiple co-IP analyses indicated that while the V205G mutation reduced FOG-1 binding to ER-GATA-1, tethering SUMO-1 did not increase binding (Fig. 6B). Thus, SUMO-1 endows the V205G mutant with the activity to regulate FOG-1-dependent genes at a level comparable to that of ER-GATA-1, despite reduced FOG-1 binding. We reasoned that SUMO-1 might restore the capacity of the V205G mutant to recruit FOG-1 to chromatin. To analyze molecular components at endogenous loci in cells expressing wild-type or mutant proteins, G1E clonal lines stably expressing V205G and SUMO-1/V205G mutants were generated. Similar to the transient assay (Fig. 5D), tethering SUMO-1 to ER-GATA-1(V205G) rescued ER-GATA-1-mediated activation of endogenous targets (Fig. 6D). Quantitative ChIP analysis revealed that ER-GATA-1(V205G) (clone 48), expressed ~2 fold higher than ER-GATA-1 (clone 11) (Fig. 6C), occupied HS2 similar to ER-GATA-1 but significantly less than ER-GATA-1 at the βmajor promoter (Fig. 6E). V205G recruited significantly less FOG-1 and Pol II to both sites. Tethering SUMO-1 to V205G rescued all defects, resulting in even greater occupancy than ER-GATA-1 at HS2. Thus, disruption of K137 sumoylation or FOG-1 binding reduces ER-GATA-1 chromatin occupancy at select sites (Fig. 6F), and tethering SUMO-1 overcomes the V205 requirement for GATA-1-mediated recruitment of FOG-1 to chromatin.
To establish the topographic relationship between GATA-1 and target genes regulated via distinct mechanisms, we conducted 3D immuno-FISH under conditions in which nuclear morphology is preserved. We measured the localization of FOG-1- and sumoylation-dependent (β-globin, α-globin, and Slc4a1) and –independent (Epb4.9, Fog1, and Eklf) targets relative to nuclear periphery-associated lamin and quantitated the percent of loci localizing at the periphery (outer 20% of the nuclear radius), as well as the inner 4 shells, each constituting 20% of the radius.
In uninduced G1E-ER-GATA-1 cells, ~50% of the β-globin (Fig. 7A,B) and α-globin (data not shown) loci localized at the periphery. In β-estradiol-treated G1E-ER-GATA-1 cells, β-globin, α-globin, and Slc4a1 were distributed throughout the 5 shells, whereas ~50% of these loci localized at the periphery in cells expressing ER-GATA-1(K137R) (Fig. 7A,B). As ER-GATA-1(K137R) is defective in inducing maturation (Fig. 1C), and locus positioning in cells expressing this mutant resembles that of immature cells, this difference might be related to the inability of ER-GATA-1(K137R) to induce maturation. Epb4.9, Fog1, and Eklf localized similarly in cells expressing ER-GATA-1 or ER-GATA-1(K137R), with ~50% of these loci residing at the periphery. Thus, ER-GATA-1 instigates the migration of certain loci away from the periphery upon maturation, consistent with the report that the β-globin locus in murine fetal liver erythroblasts resides at the periphery in immature cells and migrates inward upon maturation (Ragoczy et al., 2006). However, our results demonstrate that other GATA-1 targets reside at the periphery independent of maturation, and strikingly these loci do not require FOG-1 and GATA-1 sumoylation; only the loci that migrate inward are FOG-1- and sumoylation-dependent.
Activators often interact with multiple coregulators and are subject to diverse posttranslational modifications. Thus, it can be challenging to ascertain the contribution of individual regulatory events at endogenous loci. Herein, we used genetic complementation analysis to dissect the role of a posttranslational modification and a coregulator in the transcriptional control of various members of a target gene ensemble.
GATA-1 activates or represses target genes depending upon chromosomal context, but mechanisms underlying context-dependent activity are not established (Kim and Bresnick, 2007). Activation and repression occur via FOG-1-dependent and -independent mechanisms, but why only certain genes require FOG-1 is unknown. FOG-1 occupies FOG-1-independent loci, and therefore either FOG-1 functions redundantly at such loci or does not exert essential functions at all loci. Activities attributed to FOG-1 include facilitating GATA-1 chromatin occupancy (Letting et al., 2004; Pal et al., 2004), as well as binding the NuRD complex (Hong et al., 2005; Rodriquez et al., 2005), CtBP (Fox et al., 1999; Katz et al., 2002), and TACC3 (Garriga-Canut and Orkin, 2004; Simpson et al., 2004). The NuRD complex mediates GATA-1-dependent repression in certain contexts (Hong et al., 2005; Rodriquez et al., 2005). The significance of the CtBP interaction is unclear, since a mutant Fog1 allele encoding FOG-1 defective in CtBP binding lacks a phenotype (Katz et al., 2002). TACC3 can sequester cytoplasmic FOG-1, prevent GATA-1 binding to FOG-1, and suppress erythroid differentiation (Garriga-Canut and Orkin, 2004).
Although GATA factors recruit multiple coregulators to chromatin and are subject to diverse posttranslational modifications (Kim and Bresnick, 2007), no evidence exists that a single coregulator or modification dictates the precise transcriptional output. We demonstrate that GATA-1-mediated regulation of FOG-1-dependent target genes requires K137 sumoylation. Of the known FOG-1-independent activated target genes, K137 was not required for control of three of these four genes, demonstrating that it controls GATA-1 activity at select loci. As tethering SUMO-1 to the K137 mutant rescued activity, and disruption of the sumoylation consensus inhibited activity, similar to mutating the sumoylation site, K137 sumoylation is a crucial molecular switch to control GATA-1 function at an important subset of its target genes. Tethering SUMO-1 also rescued the V205 mutant, which was associated with restoration of FOG-1 recruitment to chromatin, although FOG-1 binding did not require K137, and K137 sumoylation did not require FOG-1. One cannot rule out the possibility, however, that FOG-1 modulates sumoylation in specific compartments in vivo. While trans-acting factor sumoylation is often implicated in transcriptional repression, and in a limited number of cases, in activation (Geiss-Friedlander and Melchior, 2007; Hay, 2005; Lyst and Stancheva, 2007), we are unaware of reports of sumoylation dictating distinct transcriptional outputs at different endogenous loci.
The control of subnuclear localization by sumoylation is established (Gill, 2004), and sumoylation can mediate nuclear pore docking (Mahajan et al., 1997; Matunis et al., 1996). Sumoylation of SATB2 (Dobreva et al., 2003) and Sp3 (Ross et al., 2002) confers a nuclear periphery localization, which is not the case with GATA-1 sumoylation. The ring-like pattern of K137 and V205 mutants resembles the localization of wild-type Sp3 (Ross et al., 2002). While SATB2 sumoylation reduces chromatin occupancy and promotes repression (Dobreva et al., 2003), and Sp3 sumoylation promotes repression (Ross et al., 2002), GATA-1 sumoylation facilitates occupancy, thereby conferring transcriptional control.
Since the molecular basis for FOG-1-dependency or independency is unknown, our demonstration that GATA-1 instigates the migration of FOG-1/sumoylation-dependent, but not -independent genes away from the nuclear periphery is of particular interest. The immobile, periphery-associated genes lack the coregulator and posttranslational modification requirements that are hallmarks of the genes that relocalize internally during maturation. Given the massive nuclear condensation during late erythropoiesis, genes residing in distinct subnuclear compartments might require different regulatory factors, e.g. genes at specific sites might require factors to counteract formidable repressive forces associated with the increasingly condensed chromatin. The organization of other genes in more hospitable nuclear subdomains would in effect buffer them against inhibitory influences. Given the differential subnuclear localization of two classes of GATA-1 target genes, it will be particularly instructive to define cis-elements at their loci and factors that confer specific subnuclear localizations and sensitivity or insensitivity of localization to maturation. Defining such sequences and factors that underlie nuclear compartment-specific transcriptional requirements has potential to be broadly important for understanding how complex gene networks are established and regulated during development.
G1E cell clonal lines, in which ER-GATA-1 and mutant derivatives were not overexpressed relative to endogenous GATA-1, and FOG-1-null cells were maintained as described in Supplemental Experimental Procedures.
G1E cells (3 × 106) were resuspended in 100 μL Nucleofector solution R (Amaxa Biosystems), and transfected with 24 μg endotoxin-free Maxi-prep DNA (Qiagen). Cells were transfected twice with the Nucleofector II system (Amaxa Biosystems) using the G-016 program, allowing 24 h between transfections. Cells were treated with 1 μM β-estradiol at 6 h after both transfections when pGD-ER-puro constructs were transfected. Cells were harvested at 24 h after the second transfection. Protein and RNA were prepared from 1 × 106 and 2 × 106 cells, respectively.
Wild-type and mutant GATA-1 were expressed as fusions with the ER ligand-binding domain using the pGD-G1ER-puro retroviral vector (Tsang et al., 1997) unless stated otherwise. Vector construction details are provided in Supplemental Experimental Procedures.
Clonal lines expressing mutant proteins at levels comparable to ER-GATA-1 were selected. To confirm that passages of cells retained comparable expression, lines were re-analyzed periodically as described in Supplemental Experimental Procedures.
GATA-1 sumoylation and endogenous FOG-1 binding were analyzed as described in Supplemental Experimental Procedures.
The immunofluorescence and 3D Immuno-FISH conditions are described in Supplemental Experimental Procedures.
Total RNA was purified with TRIzol (Invitrogen) from identical cultures used for Western blotting and ChIP, and gene expression was analyzed as described in Supplemental Experimental Procedures.
ChIP analyses were conducted as described (Im et al., 2004) and in Supplemental Experimental Procedures.
This work was funded by NIH grants DK68634 and DK50107 (EHB) and American Heart Association Predoctoral Fellowships (S-IK, H-YL). We thank Dr. Ken Young for assisting with Wright-Giemsa staining.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.