Establishing a hematopoietic genetic network through locus-specific integration of chromatin regulators

Andrew W. DeVilbiss, Meghan E. Boyer, and Emery H. Bresnick1

Department of Cell and Regenerative Biology, Wisconsin Institutes for Medical Research, Carbone Cancer Center, University of Wisconsin-Madison Blood Research Program, University of Wisconsin School of Medicine and Public Health, Madison, WI 53705

Edited by Sherman M. Weissman, Yale University School of Medicine, New Haven, CT, and approved July 23, 2013 (received for review February 11, 2013) The establishment and maintenance of cell type-specific transcrip- repression (9–11). GATA switches occur at numerous loci and tional programs require an ensemble of broadly expressed are frequently associated with altered transcriptional output. The chromatin remodeling and modifying enzymes. Many questions GATA-1–interacting coregulator Friend of GATA-1 (FOG-1) remain unanswered regarding the contributions of these enzymes mediates GATA-1–dependent activation and repression in a to specialized genetic networks that control critical processes, such context-dependent manner (12, 13). FOG-1 facilitates GATA-1 as lineage commitment and cellular differentiation. We have been chromatin occupancy (14, 15) and interacts with the nucleosome addressing this problem in the context of erythrocyte develop- remodeling and deacetylase (NuRD) chromatin remodeling com- ment driven by the transcription factor GATA-1 and its coregulator plex (16) containing the ATPase CHD4 (Mi2β), which is required Friend of GATA-1 (FOG-1). As certain GATA-1 target genes have for development of erythroid and other hematopoietic lineages little to no FOG-1 requirement for expression, presumably addi- (17, 18). GATA-1 also recruits the chromatin remodeler BRG1 tional coregulators can mediate GATA-1 function. Using a genetic (19, 20), the histone acetyltransferase CBP/p300 (21), and the complementation assay and RNA interference in GATA-1–null cells, mediator complex component Med1 (22, 23). BRG1 promotes we demonstrate a vital link between GATA-1 and the histone H4 expression of adult α-andβ-like globin genes in erythroid cells lysine 20 methyltransferase PR-Set7/SetD8 (SetD8). GATA-1 selec- (19, 24), CBP/p300 mediates GATA-1 function in at least certain tively induced H4 monomethylated lysine 20 at repressed, but contexts (21), and Med1 amplifies GATA-1 activity at select target not activated, loci, and endogenous SetD8 mediated GATA-1– genes (22). Given the crucial developmental functions of GATA dependent repression of a cohort of its target genes. GATA-1 used factors, it is reasonable to assume that the requisite coregulator different combinations of SetD8, FOG-1, and the FOG-1–interact- machinery is complex and involves considerable functional re- ing nucleosome remodeling and deacetylase complex component dundancy to ensure developmental fidelity. Because GATA-1 Mi2β to repress distinct target genes. Implicating SetD8 as a con- target genes differ in their requirements for FOG-1 (11, 12) text-dependent GATA-1 corepressor expands the repertoire of co- and GATA-1 K137 sumoylation, a modification that enhances regulators mediating establishment/maintenance of the erythroid GATA-1 activity at loci requiring FOG-1 (25), the ensemble of cell genetic network, and provides a biological framework for dis- coregulators mediating GATA factor function appears to be secting the cell type-specific functions of this important coregula- locus-specific. tor. We propose a coregulator matrix model in which distinct Unraveling mechanisms underlying locus-specific GATA factor combinations of chromatin regulators are required at different actions will provide key insights into how GATA factors function GATA-1 target genes, and the unique attributes of the target loci uniquely in distinct cell types and developmental stages. To ad- mandate these combinations. dress this problem, we conducted in silico data mining of the BioGPS database (http://biogps.gnf.org) (26) to identify chromatin GATA | epigenetics | genomics | erythroid Significance he precise regulation of complex transcriptional networks fi Tensures the delity of critical developmental processes. A Broadly expressed enzymes commonly change chromatin fundamental component of this regulation involves epigenetic structure and function. How ubiquitous chromatin regulators mechanisms that impose stringent constraints to restrict cis-ele- establish specialized patterns of gene activity is not un- ment occupancy by trans-acting factors and postchromatin oc- derstood. We identified an important link between a histone cupancy mechanisms involving the recruitment of broadly expressed methyltransferase and a transcription factor (GATA-1) that chromatin modifying and remodeling enzymes that chemically alter controls red blood cell development. We found that distinct or reposition nucleosomes. Histone modifications, such as acetyla- combinations of this enzyme and additional chromatin regu- tion and methylation, confer transcriptional repression or activation lators are required for GATA-1 to control transcription at dif- in a context-dependent manner. Although numerous enzymes ferent genetic loci. The resulting regulatory “matrix” provides modify and remodel chromatin, and knowledge on their bio- a conceptual framework for understanding how cell-restricted chemical mechanisms has advanced tremendously, many ques- factors use broadly expressed chromatin regulators to confer tions remain regarding how they establish and maintain genetic specialized gene-expression patterns that control important networks that control essential processes, including stem cell biological processes. self-renewal, lineage commitment, and cellular differentiation. Given the crucial red blood cell functions and common ther- Author contributions: A.W.D. and E.H.B. designed research; A.W.D. and M.E.B. performed research; A.W.D., M.E.B., and E.H.B. analyzed data; and A.W.D. and E.H.B. wrote apeutic scenarios demanding modulation of erythropoiesis (1), it the paper. is instructive to consider how epigenetic mechanisms control The authors declare no conflict of interest. hematopoietic stem cell (HSC) differentiation into multipotent This article is a PNAS Direct Submission. progenitors, lineage-committed progenitors, and ultimately Data deposition: The data reported in this paper have been deposited in the Gene Ex- erythrocytes. The transcription factor GATA-2 (2, 3) is required pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession nos. GSE49174 for the genesis and maintenance of HSCs (4), whereas GATA-1 and GSE48188). (5, 6) is crucial for erythrocyte, megakaryocyte, mast cell, and 1To whom correspondence should be addressed. E-mail: [email protected]. eosinophil development (7, 8). During erythropoiesis, GATA-1 This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. replaces GATA-2 at Gata2 chromatin sites, thus conferring 1073/pnas.1302771110/-/DCSupplemental.

E3398–E3407 | PNAS | Published online August 19, 2013 www.pnas.org/cgi/doi/10.1073/pnas.1302771110 Downloaded by guest on October 1, 2021 regulators enriched in erythroid cells, which may imply an impor- function was selective for Hbb-bh1 and Hbb-y, as the SetD8 PNAS PLUS tant erythroid function and identify novel GATA-1 coregulators. knockdown did not affect ER-GATA-1–mediated activation of This analysis revealed SetD8, the sole methyltransferase that adult Hbb-b1 and Hba-a1 expression (Fig. 1E). SetD8 knockdown catalyzes H4K20 monomethylation (H4K20me1) (27). Although did not alter expression of the established repressor of embryonic/ the SetD8 catalytic mechanism has been elucidated (28), many fetal β-like globin genes Bcl11a,orKlf1, which can induce BCL11A questions remain regarding its cell type-specific functions. SetD8 expression (43) (Fig. 1E). The SetD8 knockdown did not affect and H4K20me1 are dynamically regulated throughout the cell basal expression or GATA-1–mediated repression of the pro- cycle, and SetD8 degradation promotes cell cycle progression totypical GATA-1–repressed genes, Gata2, Lyl1,andc-Kit, (29). Targeted deletion of SetD8 blocks embryogenesis at the although Rgs18 basal activity was up-regulated and GATA-1– four- and eight-cell stages and impairs chromatin compaction mediated Rgs18 repression was reduced (Fig. 1E). siRNA-mediated (30). H4K20me1 has been correlated with transcriptional acti- knockdown of SetD8 mRNA by 80% in murine erythroleukemia vation and repression (27). H4K20me1 localizes predominantly cells did not alter Gata1 expression (Fig. S1). to nontranscribed chromatin regions on Drosophila polytene chro- Because the gene-expression analysis suggested that SetD8 mosomes (31) and to E2F-repressed genes in HeLa cells (32). represses a restricted cohort of GATA-1–regulated genes, we Functional studies in Drosophila provide evidence for SetD8- conducted transcriptional profiling to rigorously establish the dependent repression mechanisms (33). In contrast, analyses in SetD8-sensitive target gene ensemble. We compared the tran- HeLa cells and T lymphocytes revealed that H4K20me1 resides scriptional profiles of β-estradiol–induced G1E-ER-GATA-1 at actively transcribed chromatin (34, 35). We demonstrate that cells, each transfected with control or SetD8 siRNA. This anal- SetD8 is a context-dependent GATA-1 corepressor and provide ysis revealed only 97 significantly up-regulated genes upon SetD8 evidence for locus-specific mechanisms that integrate ensembles knockdown (Fig. 2A). Hbb-bh1 was among the top 10 highest up- of chromatin regulators, which we term a coregulator matrix model regulated genes (Fig. 2A). Only 11 genes were significantly down- of GATA factor function. regulated, with SetD8 expression declining to the greatest extent (Fig. 2B). Real-time RT-PCR analysis with up-regulated genes Results (Kank3, Ak1, Scamp5, Tnfrsf18, Rec8, and Dxcr) validated the SetD8-Dependent Target Gene Ensemble in Erythroid Cells. Given the microarray results (Fig. 2C). Gene Ontology (GO) analysis crucial SetD8 activity for early development, and the common revealed that these SetD8-regulated genes were associated with H4K20me1 mark in diverse systems, presumably SetD8 controls a wide spectrum of biological processes (Fig. 2D). The majority a broad spectrum of biological processes. As cell type-specific of SetD8-regulated genes differed from erythroid genes that en- SetD8 mechanisms and SetD8 function/regulation in the hemato- dow cells with the characteristic erythroid phenotype. In principle, poietic system are largely unexplored, we conducted siRNA-based the biologically and mechanistically disparate SetD8-regulated loss-of-function and genetic complementation analysis to evaluate genes might share a common chromosomal environment, such as its function in a physiologically relevant model of erythroid cell chromosomal location. However, evaluation of the chromosomal maturation, G1E-ER-GATA-1 cells (36) (Fig. 1A). G1E-ER- distribution of the top 20 SetD8-repressed genes residing on dif- GATA-1 cells were derived from GATA-1–nullizygous ES cells, re- ferent chromosomes revealed no enrichments at gross chromo- semble normal proerythroblasts, and stably express a conditionally somal features (e.g., telomeres or centromeres) (Fig. 2E). active allele of GATA-1 (ER-GATA-1) (37). β-Estradiol–mediated activation of ER-GATA-1 induces a GATA-1–dependent genetic SetD8 as a Context-Dependent GATA-1 Corepressor. Comparison of network and morphological changes, recapitulating a normal window GATA-1– and SetD8-regulated gene cohorts by real-time RT- of adult erythroid maturation (38). PCR revealed two modes by which SetD8 mediates GATA-1 siRNA-mediated knockdown of SetD8 mRNA (75% reduc- function. First, SetD8 was required for GATA-1 to repress tion) (Fig. 1B) and protein (Fig. 1C)inβ-estradiol–treated and Kank3 expression (Fig. 2C) and knocking-down SetD8 abrogated untreated cells did not induce gross changes in cell morphology and the repression. Second, SetD8 repressed expression of certain viability. To test whether SetD8 contributes to G1E-ER-GATA-1 GATA-1–activated genes. GATA-1 activated Scamp5 (Fig. 2C), GENETICS cell maturation, we quantitated expression of erythroid cell-sur- and knocking-down SetD8 yielded Scamp5 hyperactivation. A face markers (CD71 and Ter119) that delineate the maturation distinct gene cohort, exemplified by Ak1, was SetD8-, but not status (39). As expected, β-estradiol treatment for 48 h greatly GATA-1–, regulated (Fig. 2C). increased Ter119 staining. The SetD8 knockdown did not sig- To more rigorously establish interrelationships between GATA- nificantly influence Ter119 induction (Fig. 1D). Flow cytometric 1– and SetD8-regulated genes, we generated a new GATA-1– analysis of cell viability and apoptosis revealed no difference regulated gene dataset, because prior GATA-1 datasets were between control- and SetD8 siRNA-transfected cells (Fig. 1D). obtained with a different microarray platform and different Thus, G1E-ER-GATA-1 maturation, which recapitulates a com- cultures of G1E-ER-GATA-1 cells (38, 40). We conducted ponent of the process that yields erythrocytes, is insensitive to transcriptional profiling in nonspecific siRNA-transfected, unin- lowering SetD8 levels. duced, and β-estradiol-induced G1E-ER-GATA-1 cells to control GATA-1 activates and represses target gene transcription, for potential influences of nucleofection on ER-GATA-1 activity. thereby establishing a complex genetic network that orchestrates Because GATA-1–regulated and SetD8-activated genes did not erythropoiesis and erythroid cell function (11, 40–42). To de- overlap, we focused only on SetD8-repressed genes. A total of 47 termine if SetD8 contributes to the establishment and mainte- genes were coregulated by SetD8 and GATA-1 (2% of all GATA- nance of this genetic network, we asked whether the SetD8 1 targets). The SetD8-repressed genes segregated based on ER- knockdown altered the capacity of GATA-1 to regulate target GATA-1–responsiveness: GATA-1–activated, GATA-1–repressed, genes in the genetic complementation assay. The knockdown did and GATA-1–insensitive (Fig. 3 A and B). Although the majority not influence GATA-1–mediated activation of prototypical tar- (52%) of SetD8-regulated genes were GATA-1–insensitive, 32% get genes Alas2, Epb4.9, and Ahsp, although it modestly sup- were GATA-1–repressed (Fig. 3 A and B), and 16% were GATA- pressed Slc4a1 activation (Fig. 1E). We also tested whether the 1-activated (Fig. 3 A and B). GO analysis of the SetD8/GATA-1– knockdown influenced expression of Hbb-y and Hbb-bh1, which corepressed cohort revealed some of these genes function in encode embryonic/fetal β-like globin proteins and are expressed “B-cell proliferation,” including Cd81 and Tnfrsf13b. (Fig. 3 A at low levels in the adult or definitive G1E-ER-GATA-1 cells. and B). Two SetD8-repressed/GATA-1–activated genes were in- The SetD8 knockdown activated Hbb-bh1 and Hbb-y expression volved in “protein amino acid ADP-ribosylation (Fig. 3 A and 14- and 3.8-fold, respectively (Fig. 1E). This SetD8 repressive B). SetD8-regulated, GATA-1–insensitive genes were linked

DeVilbiss et al. PNAS | Published online August 19, 2013 | E3399 Downloaded by guest on October 1, 2021 Harvest Flow Rgs19,andLimd2 resembled Gata2 and c-Kit, direct GATA-1 A siRNA siRNA RNA/Protein Cytometry -estradiol target genes whose expression declines upon transition from  -estradiol (1M) -estradiol proerythroblasts to reticulocytes (Fig. 3C). SetD8 was highly Time (h) 024 48 72 expressed at all maturation stages. * * – B 1.0 C Although GATA-1 mediated repression can involve FOG-1

* * -3 0.8 38 SetD8 (12), GATA switches (10), and reduced occupancy of the he- 0.6

mRNA matopoietic transcription factor Scl/TAL1 (11), many questions 0.4 r

M x 10 50 Tubulin 0.2 remain unanswered regarding the underlying mechanisms. In

SetD8 -estradiol (Relative Units) 0 addition, we are unaware of reports in which SetD8 functions as -estradiol Control siRNA Control siRNA SetD8 siRNA a corepressor for any cell type-specific activator. To ask whether SetD8 siRNA GATA-1 directly controls the SetD8-regulated gene cohort, we 60 tested whether endogenous GATA-1 occupies the respective D 5 10 40 loci. Analysis of an endogenous GATA-1 ChIP-seq dataset from 4 10

High – 1.1 1.9 20 mouse erythroleukemia cells generated with our anti GATA-1

3 % Ter119 10

- Est 0 antibody revealed GATA-1 occupancy at, or in the vicinity of, 2 60 10 the SetD8/GATA-1–corepressed genes (Fig. 4A). GATA-1 oc- 0 Control SetD8 40 cupied Myo1g and Rgs19 intronic sites, Limd2, Vim, and Kank3 5 10 20

CD71 distal sites, and Vim, Kank3, and Clec10a promoter sites (Fig. 4 10 % Live Cells 0 4A). Analysis of histone modifications that demarcate enhancers 44 48.6 20 3 10 15 (H3acetylationatK27andH3monomethylation at K4) and pro- + Est 2 10 10 moters (H3 trimethylation at K4) from a mouse erythroleukemia

0 % Early 5 Control SetD8 Apoptosis cell ChIP-seq dataset revealed patterns largely predictable from 2 3 4 5 2 3 4 5 010 10 10 10 010 10 10 10 0 -estradiol the genomic location of GATA-1 occupancy (Fig. 4A). Of the 31 Ter119 Control siRNA SetD8 siRNA SetD8/GATA-1-corepressed genes, GATA-1 peaks were detected

800 at or in the vicinity of 24 genes (77%) (Fig. S2). Of these 24 genes, E Alas2 Slc4a1 Epb4.9 Ahsp 600 4000 200 GATA-1 occupied the promoter or gene body of 14 genes 600 * 3000 150 (promoter occupancy, 9 genes; gene body, 11 genes). 400 400 2000 100 To further assess whether the SetD8/GATA-1–corepressed 200 200 1000 50 genes occupied by endogenous GATA-1 are direct GATA-1 tar- mRNA Levels mRNA

(Relative Units) 0 0 0 0 45 4500 gets, we analyzed the kinetics of GATA-1–mediated repression of Hba-a1 80 Hbb-y 1200 Hbb-bh1 Hbb-b1 * * these genes. β-Estradiol treatment of G1E-ER-GATA-1 cells 30 60 3000 800 40 rapidly reduced primary transcript levels for the previously estab- 15 400 1500 lished direct GATA-1 targets Gata2 (9) and Lyl1 (45) by 70% after 20

mRNA Levels mRNA 12 h (Fig. 4B). Myo1g, Vim,andClec10a primary transcript levels

(Relative Units) 0 0 0 0 -estradiol were reduced rapidly by 1–2 h, and to a similar extent 12-h Control siRNA SetD8 siRNA postestradiol treatment, consistent with these genes being direct 1.6 2.5 Gata2 Lyl1 c-Kit Rgs18 1.5 Bcl11a 5 Klf1 2.0 4 GATA-1 targets (Fig. 4B). Kank3 and Rgs19 repression was slightly 1.2 * fi 1.5 1.0 3 slower (Fig. 4B). However, signi cant repression was apparent at 0.8 1.0 2 these loci by 1 and 2 h for Kank3 and Rgs19,respectively. 0.5 0.4 0.5 1 * Considering the SetD8 catalytic mechanism, presumably mRNA Levels mRNA (Relative Units) 0 0 0 SetD8 mediates repression of the GATA-1 targets by catalyzing -estradiol Control siRNA H4K20me1. We tested whether GATA-1 induces H4K20me1 at SetD8 siRNA SetD8/GATA-1-corepressed genes. To address this theory, we Fig. 1. Highly selective SetD8 functions in erythroid cells. (A) G1E-ER-GATA- conducted quantitative ChIP analysis with untreated or β-estradiol– 1 cells were transfected with nontargeting or SetD8-specific siRNA at t =0h. treated G1E-ER-GATA-1 cells using an anti-H4K20me1 anti- At t = 24 h, cells were transfected with the identical siRNA again, and were body. At the SetD8-sensitive, GATA-1–repressed genes Vim, β treated with -estradiol for 24 or 48 h. At t = 48 h, RNA and protein was Clec10a, Kank3,andRgs19, β-estradiol induced promoter- isolated, and flow cytometry was conducted at 72 h. Control cells were not A – treated with β-estradiol. (B) SetD8 mRNA levels were quantitated by RT-PCR associated H4K20me1 (Fig. 5 ). At GATA-1 activated promoters, ± < H4K20me1 levels were reduced or unaffected by ER-GATA-1 (n = 7, mean SE). **P 0.001. (C) SetD8 protein levels were measured by fi semiquantitative Western blotting. (D) G1E-ER-GATA-1 cell maturation was activation. ER-GATA-1 signi cantly reduced H4K20me1 at the quantitated by flow cytometry. (Left) Representative contour plot of G1E-ER- Alas2 promoter by 3.4-fold. H4K20me1 declined at the Slc4a1 GATA-1 cell maturation 48 h after β-estradiol treatment. (Right) Averages of promoter, but not significantly (P = 0.058). Low-level H4K20me1 the percent Ter119-high population, live cells, and early apoptotic cells (n =2, resided at the Hbb-b1 and Hba-a1 promoters, which was un- mean ± SD). (E) GATA-1-activated and -repressed target genes, globin genes, affected by β-estradiol treatment. ER-GATA-1 did not regulate and genes implicated in regulating hemoglobin switching (Bcl11a and Klf1) H4K20me1 at the constitutively repressed Necdin promoter, ± mRNA levels were quantitated by RT-PCR (n = 7, mean SE). which is insensitive to β-estradiol and SetD8 knockdown. To assess the mechanism by which ER-GATA-1 activation – to diverse cellular processes, including response to wounding, induces H4K20me1 at SetD8/GATA-1 regulated genes, we determined the relationship between GATA-1, SetD8, and stress, and apoptosis, (Fig. 3 A and B). Thus, expression pro- H4K20me1 occupancy. Quantitative ChIP analysis was con- filing demonstrated that SetD8 is a GATA-1 corepressor at ducted in β-estradiol–treated and untreated G1E-ER-GATA- a restricted cohort of GATA-1 target genes, and its quantitative 1 cells using anti–GATA-1,anti-H4K20me1,andanti-SetD8 contribution to GATA-1 function can be considerable. Because antibodies. At the Vim locus, β-estradiol-induced GATA-1 our analysis revealed genes that have not been studied in erythroid occupancy at sites −0.6 kb and +1.2 kb from the Vim tran- cells, we mined expression data obtained from murine primary scription start site (TSS) (Fig. 5B). SetD8 occupancy was also adult erythroblasts of differing maturation stage (44), focusing on detected at the Vim locus, and was maximal at the +1.2-kb SetD8/GATA-1–corepressed genes that will be the focus of sub- GATA-1 binding site (Fig. 5B). This finding indicates that SetD8 sequent mechanistic analyses. The expression of Vim, Clec10a, and GATA-1 can be cross-linked to overlapping chromatin regions.

E3400 | www.pnas.org/cgi/doi/10.1073/pnas.1302771110 DeVilbiss et al. Downloaded by guest on October 1, 2021 4 30 400 PNAS PLUS Kank3 Ak1 Scamp5 * A 9030617O03Rik Sdsl C * Pde6a Cpped1 3 * 300 Ak1 Zfp92 20 Kank3 Acaa2 * Phlda3 Fndc5 2 200 Scamp5 Dhrs7 Cpt1c Cd81 10 1 100

Dcxr Slc19a2 Levels mRNA

Hbb-bh1 Ncf2 (Relative Units) Tnfrsf18 Gna15 0 0 0 * Rec8 Shisa5 Clec10a chr5:122048289-122068947 -estradiol Aqp8 Flot1 Control siRNA A_55_P2093231 Rab44 Setd8 siRNA Amhr2 Exoc4 40 Serpinb2 Rps6ka1 Tnfrsf18 * Rec8 Dcxr Cpz Lage3 8 3 * Pros1 Tbc1d10c * * 30 Trp53inp1 Rnf169 6 2 Nme4 Vim 20 Mst1 chr13:38731361-38735655 4 Gsn Dcaf4 1 Lrdd ENSMUST00000086601 10 ENSMUST00000147634 Cyba 2

Tmem184a Arap3 Levels mRNA Glipr1 41156 (Relative Units) 0 0 0 Naip2 chr5:122048289-122068947 -estradiol Cdkn1a Dbp Control siRNA Sla Gm16515 Setd8 siRNA Cd80 Lgals9 Ephx2 NAP094642-001 Apoc1 Aen response to wounding p=1.9E-4 Mfge8 Fgd3 D regulation catalytic activity p=2.7E-4 chr11:120079199-120130099 Snx21 apoptosis p=2.8E-4 Bst2 Usp20 Art4 Alox5 regulation hydrolase activity p=2.9E-4 Tnfrsf13b Inf2 response to external stimulus p=6.9E-4 chr9:96664617-96683617 Lage3 cellular lipid metabolic process p=1.8E-3 Tle6 Bax Enc1 Tmem229b response to stress p=1.8E-3 Alox5ap Nr2f6 coagulation p=4.2E-3 Suox Itgb2 fatty acid metatabolism p=4.2E-3 Slc22a18 Limd2 Fam53b Tmem19 unsaturated fatty acid metabolism p=1.2E-2 Ccng1 Fam132a regulation caspase activity p=2.3E-2 Fig. 2. SetD8-dependent target gene Mgmt Arhgap9 regulation B cell activation p=2.7E-2 Man2b1 Gp5 ensemble. Microarray analysis was per- Gm3448 Vav1 response to radiation p=2.9E-2 P2rx1 regulation lymphocyte activation p=2.9E-2 formed using RNA from three indepen- Rgs19 10 regulation of signal transduction p=2.9E-2 Myo1g 6 dent SetD8 knockdown experiments. We Gm3448 Fold 4 acute inflammatory response p=3.6E-2 Khk Upregulated 3 positive regulation transport p=4.4E-2 compared RNA from β-estradiol–treated 1700016F12Rik 2.5 Rom1 2 0 5 10 15 20 (24 h) cells receiving nontargeting or Number of Enriched Genes SetD8 siRNA. (A) Genes significantly up- Setd8 Fold B Hk2 Downregulated regulated more than twofold upon SetD8 NAP096516-001 Trib3 4 knockdown. (B) Genes significantly down- Dnajc21 3 Sdad1 2.5 regulated more than twofold upon SetD8 2 Gm11974 knockdown. (C) Validation of array re- sults by RT-PCR (n = 3, mean ± SE). *P < E Chr. 1 Serpinb2, Phlda3 Chr. 12 9030617O03Rik 0.05. (D) GO analysis of genes up-regulated Gsn Chr. 14 Rec8 Chr. 2 by the SetD8 knockdown. (E)Murine Chr. 4 Trp53inp1, Tnfrsf18 Chr. 15 Amhr2 chromosomal distribution of the top 20 Chr. 5 Cpz Chr. 16 Pros1 up-regulated genes from the microarray fi Chr. 7 Cpt1c, Hbb-bh1, Apq8 Chr. 17 Nme4, Kank3 analysis (chromosome pro les adapted from the Integrated Genome Browser). Chr. 9 Scamp5, Mst1 Chr. 18 Pde6a Chromosomes are oriented with the Chr. 11 Dcxr, Clec10a centromere on the left.

Although β-estradiol treatment did not affect SetD8 occupancy, conventional G1E-ER-GATA-1 line (46), we used a clonal line H4K20me1 levels increased across a ∼7-kb region around the expressing lower levels of ER-GATA-1 to ensure that ER-GATA-1 TSS (Fig. 5B). GATA-1 occupied a site −0.4 kb from the Clec10a did not exceed expression levels of the mutants. Semiquantitative GENETICS TSS (Fig. 5C), and similar to Vim, SetD8 and GATA-1 occupancy Western blotting demonstrated that ER-GATA-1(V205G) protein overlapped (Fig. 5C). Also similar to Vim, β-estradiol treat- was expressed at least as high as ER-GATA-1 (Fig. 6B). Because of ment yielded a broad zone of H4K20me1 enrichment, whereas the low-level ER-GATA-1 expression, however, the magnitude of SetD8 occupancy was unaffected (Fig. 5C). These data indicate ER-GATA-1 responses at certain loci is less than with our typical that SetD8 precedes GATA-1 occupancy at these target genes, and G1E-ER-GATA-1 line. Whereas ER-GATA-1 repressed Gata2, are consistent with a model in which GATA-1 occupancy stim- Kank3,andRgs19, repression was lower in all of the ER-GATA-1 ulates SetD8 activity to induce H4K20me1 at the respective loci in (V205G) clones (Fig. 6B). In contrast, ER-GATA-1(V205G) re- a manner that is not restricted to the GATA-1 occupancy site. pressed Lyl1, Rgs18, Myo1g, Clec10a,andVim to an equal or greater extent than ER-GATA-1, indicating that disrupting the ER-GATA- Locus-Specific Integration of Chromatin Regulators: Evidence for 1-FOG-1 interaction did not affect repression of these genes (Fig. a Coregulator Matrix Model of GATA Factor Function. The activa- – tion or repression of GATA-1 target genes can be FOG-1–sensitive 6B). In aggregate, these results demonstrate that SetD8/GATA-1 – – corepressed genes are not dedicated to a single transcriptional mode or insensitive (11, 12). Thus, GATA-1 mediated repression of – – SetD8 target genes might be FOG-1–sensitive, FOG-1–insensitive, involving FOG-1, but rather are FOG-1 sensitive or insensitive. or both at distinct loci. We knocked-down FOG-1 in G1E-ER- The NuRD complex associates with FOG-1 (16) and is an – GATA1 cells to assess FOG-1–sensitivity of SetD8/GATA-1– important determinant of GATA-1 mediated regulation of corepressed genes (Fig. 6A). Whereas ER-GATA-1–mediated transcription and hematopoiesis (17, 18). We predicted that – – repression of Clec10a, Myo1g,andLyl1 was mildly sensitive to the FOG-1 regulated, SetD8/GATA-1 corepressed genes would knockdown, Kank3, Rgs19, Gata2,andc-Kit repression was abro- also require Mi2β, a key ATPase subunit of the NuRD com- gated (Fig. 6A). As an alternative approach, we evaluated FOG-1 plex (47, 48). siRNA-mediated knockdown of Mi2β mRNA sensitivity by stably expressing ER-GATA-1 or a mutant (ER- nearly ablated Mi2β protein (Fig. 7A, Left) and abolished GATA- GATA-1-V205G) defective in FOG-1 binding (12) in G1E cells. 1–mediated repression of the FOG-1–insensitive SetD8/GATA- Four clones of ER-GATA-1(V205G) –expressing cells were com- 1–corepressed gene Clec10a (Fig. 7A, Right). Repression of the paredwithanER-GATA-1–expressing clone. Because clonal lines FOG-1–sensitive genes Gata2, c-Kit,andRgs19 was significantly, typically express ER-GATA-1(V205G) at levels lower than our but modestly, reduced (Fig. 7A, Right). The FOG-1–sensitive

DeVilbiss et al. PNAS | Published online August 19, 2013 | E3401 Downloaded by guest on October 1, 2021 A 16 Genes protein amino acid ADP-ribosylation p=1.2E-2 01 23 Number of Enriched Genes 31 Genes regulation of B-cell proliferation p=4.2E-2 01 23 Number of Enriched Genes 50 Genes apoptosis p=9.3E-4 response to wounding p=1.6E-3 response to DNA damage stimulus p=5.5E-3 regulation caspase activity p=6.9E-3 coagulation p=1.3E-2 cellular response to stress p=1.8E-2 accute inflammatory response p=1.9E-2 GATA-1-Activated vesicle-mediated transport p=2.8E-2 NADP metabolic process p=3.4E-2 GATA-1-Repressed positive regulation of transport p=4.3E-2 response to radiation p=4.9E-2 GATA-1-Insensitive 0 510 Number of Enriched Genes

B GATA-1-Activated GATA-1-Repressed GATA-1-Insensitive Gene Fold Change Gene Fold Change Gene Fold Change Gene Fold Change Name GATA-1 SetD8 Name GATA-1 SetD8 Name GATA-1 SetD8 Name GATA-1 SetD8 Tmem184a 114 4.0 Vim -17.1 2.4 Ak1 - 11.0 Slc19a2 - 2.5 Trp53inp1 10.7 4.3 Limd2 -16.1 2.0 Phlda3 - 6.7 Ncf2 - 2.5 Gm16515 5.6 2.3 Clec10a -14.8 5.1 Cpt1c - 6.5 chr5:122048289-122068947 - 2.5 Aqp8 5.2 4.9 Myo1g -9.6 2.9 Dcxr - 6.4 Rab44 - 2.5 chr9:96664617-96683617 5.1 3.3 Kank3 -8.9 10.3 Tnfrsf18 - 6.1 Exoc4 - 2.5 Hbb-bh1 4.9 6.3 Arap3 -8.9 2.3 A_55_P2093231 - 4.8 Rps6ka1 - 2.4 Scamp5 4.9 6.6 Rgs19 -8.7 2.9 Cpz - 4.5 Lage3 - 2.4 Art4 4.8 3.3 Gp5 -6.7 1.9 Pros1 - 4.4 Rnf169 - 2.4 Amhr2 4.4 4.8 Apoc1 -6.5 3.4 Nme4 - 4.2 Dcaf4 - 2.3 Pde6a 4.0 12.0 Tnfrsf13b -5.5 3.3 Mst1 - 4.2 ENSMUST00000086601 - 2.3 Itgb2 2.9 2.0 Acaa2 -4.0 2.7 Gsn - 4.2 Sept4 - 2.3 Fndc5 2.8 2.7 Tle6 -4.0 3.3 Lrdd - 4.1 NAP094642-001 - 2.2 Fam132a 2.6 2.0 Arhgap9 -3.9 2.0 1700003M07Rik - 4.0 Aen - 2.2 Fgd3 2.3 2.2 Cyba -3.7 2.3 Naip2 - 3.8 Usp20 - 2.1 1700016F12Rik 2.0 2.8 Lgals9 -3.6 2.2 Cdkn1a - 3.7 Alox5 - 2.1 Gna15 1.9 2.5 Inf2 -3.1 2.1 Sla - 3.6 Bax - 2.1 Dbp -3.0 2.3 Cd80 - 3.6 Tmem229b - 2.1 Flot1 -2.8 2.5 Ephx2 - 3.5 Tmem19 - 2.0 Dhrs7 -2.8 2.7 Mfge8 - 3.4 Vav1 - 1.9 Glipr1 -2.8 3.8 chr11:120079199-120130099 - 3.4 Slc22a18 -2.6 3.1 Enc1 - 3.2 Rec8 -2.5 5.4 Suox - 3.2 Bst2 -2.5 3.4 Fam53b - 3.0 Alox5ap -2.4 3.2 Ccng1 - 3.0 Nr2f6 -2.3 2.1 Mgmt - 3.0 9030617O03Rik -2.3 13.5 Man21b - 2.4 Snx21 -2.1 2.2 Gm3448 - 2.9 Cpped1 -2.1 2.7 P2rx1 - 2.9 Tbc1d10c -2.1 2.4 Khk - 2.8 Shisa5 -1.8 2.5 Rom1 - 2.8 Cd81 -1.7 2.6 Sdsl - 2.8

15000 1500 6000 1500 Gata2 c-Kit Vim 4000 Limd2 Clec10a Rgs19 Setd8 C 150 60

10000 3000 1000 4000 1000 100 40 2000 5000 500 2000 500 50 20 1000 mRNA Levels 0 0 0 0 0 0 0 PBOR PBOR PBOR PBOR PBOR PBOR PBOR

P = Proerythroblast B = Basophilic Erythroblast O = Orthochromatic Erythroblast R = Reticulocyte

Fig. 3. SetD8 as a context-dependent GATA-1 corepressor. The set of genes up-regulated by SetD8 knockdown was subjected to sorting into one of three cat- egories based on their response to GATA-1 activation: GATA-1–activated, GATA-1–repressed, or GATA-1–insensitive. GATA-1–sensitivity values are based on microarray analysis comparing cells receiving control siRNA, and either no treatment or β-estradiol (n = 3 individual experiments). (A)Piechartdisplaystheper- centage of SetD8 regulated genes that fall into each of the three categories. GO analysis was conducted with all gene categories (David Bioinformatics, sorted by P value). (B)Genesfallingintoeachcategoryarelisted.(C) mRNA levels of GATA-1–repressed genes mined from the murine ErythronDB database (www.cbil.upenn. edu/ErythronDB/). The direct target genes Gata2 and c-Kit are shown as controls, and Vim, Clec10a,andRgs19 represent SetD8/GATA-1–corepressed genes.

gene Kank3 was Mi2β-insensitive (Fig. 7A, Right). These results were also repressed by Mi2β, including Hbb-bh1 and Clec10a (Fig. indicate that SetD8/GATA-1–corepressed genes are Mi2β-sensi- 7B). Only 12 genes were repressed by GATA-1, Mi2β, and SetD8. tive or -insensitive. No SetD8-activated genes were also activated by GATA-1, and To determine the extent to which GATA-1 requires combi- only one gene was activated by SetD8 and Mi2β (Fig. 7B). These β nations of SetD8 and Mi2 to mediate transcriptional repression, data suggest that although Mi2β can positively and negatively we conducted transcriptional profiling of β-estradiol–induced β coregulate genes with GATA-1, SetD8 is exclusively involved G1E-ER-GATA-1 cells treated with Mi2 siRNA or nontargeting – control siRNA. Knocking down Mi2β up-regulated 1,133 genes in GATA-1 mediated repression. The differential coregulator – and down-regulated 768 genes. Seventeen percent of GATA-1– requirements for GATA-1 mediated repression constitute repressed genes (240) were also repressed by Mi2β (Fig. 7B). amatrix(Fig.8A), supporting a model in which GATA factor Fifteen percent of GATA-1–activated genes (205) were also ac- function requires different coregulator combinations at distinct tivated by Mi2β. Additionally, 37% of SetD8-repressed genes (35) endogenous loci (Fig. 8B).

E3402 | www.pnas.org/cgi/doi/10.1073/pnas.1302771110 DeVilbiss et al. Downloaded by guest on October 1, 2021 PNAS PLUS A 54 GATA-1 Chr. 2

sdaeRecneuqeS 5 kb

10 H3K4me1

20 H3K27ac

40 H3K4me3

Vim 733 GATA-1 Chr. 11

10 kb

10 H3K4me1 50 H3K27ac Sequence Reads 40 H3K4me3

Limd2 193 GATA-1 Chr. 11

sdaeRec 1 kb

10 H3K4me1 neuqeS 20 H3K27ac

40 H3K4me3

Clec10a 62 GATA-1 Chr. 11

sdae 5 kb

R 10 H3K4me1 ecne

30 H3K27ac uqeS

55 H3K4me3

Myo1g 58 GATA-1 Chr. 17 sd 5 kb ae

RecneuqeS 10 H3K4me1

20 H3K27ac

40 H3K4me3

Kank3 GENETICS 150 GATA-1 Chr. 2

sdaeRecneuqeS 2 kb

10 H3K4me1

10 H3K27ac

30 H3K4me3

Rgs19 Fig. 4. Evidence that GATA-1 directly regulates SetD8 B 1.2 fi Vim Clec10a Myo1g Kank3 Rgs19 Gata2 Lyl1 target genes. (A) Endogenous GATA-1 ChIP-seq pro les 1.0 from mouse erythroleukemia cells at SetD8/GATA-1–core- pressed loci. GATA-1 ChIP-seq profiles are aligned with 0.8 H3K4me1, H3K27ac, and H3K4me3 profiles, all from 0.6 mouse erythroleukemia cells. (B) Kinetics of repression of 0.4 SetD8/GATA-1-coregulated genes. Primary transcript levels were measured at 0, 1, 2, 4, 8, 12, 24, and 48 h after

Primary Transcript 0.2 β-estradiol treatment (n =6± SE). The dashed vertical line (% Maximal Expression) 0 0 10203040500 10203040500 10203040500 10203040500 10203040500 10203040500 1020304050 in each panel in B represents the first time point at which Time (h) Time (h) Time (h) Time (h) Time (h) Time (h) Time (h) repression was statistically significant compared with t =0.

Discussion (49). SetD8 knockdown in HEK293 cells reduced expression of We describe evidence that the H4K20me1 methyltransferase several Wnt target genes, and SetD8 regulated Wnt target SetD8 is a context-dependent GATA-1 corepressor. A prior anal- genes in zebrafish (49). However, it was unknown whether SetD8 ysis had implicated SetD8 as a coactivator for the Wnt pathway mediates transcriptional control by a large or highly restricted co- factor LEF1/TCF4 (49). SetD8 overexpression and knockdown in hort of trans-acting factors, whether other developmental regulators 3T3 cells increased and decreased Axin2 expression, respectively use SetD8 to instigate cell type-specific transcriptional programs,

DeVilbiss et al. PNAS | Published online August 19, 2013 | E3403 Downloaded by guest on October 1, 2021 A GATA-1-Repressed GATA-1-Activated Control 0.20 Preimmune * 0.15 H4K20me1 * 0.10 * * * * * 0.05 * (Relative Units)

H4K20me1 Levels 0 -estradiol

Lyl1 c-Kit Alas2 Rgs18 Kank3 Rgs19 Slc4a1 Hba-a1 Hbb-b1 Necdin Gata2 1S Clec10a Vim (-1.1kb)

Clec10a B Chr. 2 Vim 1 kb C Chr. 11 1 kb

Fig. 5. Selective GATA-1-medi- 0.10 0.15 GATA-1 * GATA-1 ated H4K20me1 induction at 0.08 SetD8-repressed loci. (A) Quan- 0.1 0.06 titative ChIP analysis of promoter- * associated H4K20me1 levels at 0.04 * ± * 0.05 GATA-1 target genes (n =3, SE). *P < 0.05. (BandC) Quan- 0.02 * * * * titative ChIP analysis of ER-GATA- Relative Occupancy Relative Occupancy 0 0 0.06 1, SetD8, and H4K20me1 as a SetD8 SetD8 0.03 function of distance from the (B)

0.04 Vim TSS and (C) Clec10a TSS (n = 0.02 4, ± SE) An astersisk indicates a significant difference (P < 0.05) 0.02 0.01 between the untreated and β- estradiol–treated value for each Relative Occupancy 0 Relative Occupancy 0 amplified site. The TSS is depicted 0.80 H4K20me1 H4K20me1 * 0.3 as a vertical dashed line. Rabbit pre- 0.60 immune serum (for GATA-1 and ** H4K20me1) and purified mouse 0.2 * 0.40 * IgG (for SetD8) are graphed as * a horizontal dashed line, which re- * * * * 0.1 0.20 presents the average value from all sites at each respective locus. Vim Relative Occupancy 0 Relative Occupancy 0 -5 -4 -3 -2 -1 0 1 2 3 -15 -10 -5 0 5 10 and Clec10a loci are depicted at the Distance From TSS (kb) Distance From TSS (kb) top of B and C, respectively, with Untreated -estradiol Untreated -estradiol marks indicating the position of each amplicon.

and whether SetD8 commonly functions as a coactivator (34, 35, in diverse contexts. It is attractive to propose that the local 49) or a corepressor (31, 33). chromatin environment, higher-order chromatin structure, or Our results demonstrate that endogenous SetD8 functions pre- subnuclear neighborhood mandate the locus-specific mechanistic dominantly as a corepressor in G1E-ER-GATA-1 cells. Intrigu- requirements for transcriptional control. At a rudimentary level ingly, GATA-1 target genes differ in their requirements for SetD8 involving a single coregulator, FOG-1, FOG-1–sensitive GATA- and other coregulators. Reducing the level of endogenous SetD8 1 target genes are expelled from the nuclear periphery upon up-regulated mouse fetal and embryonic globin genes. Factors activation, whereas FOG-1–insensitive GATA-1 target genes that selectively repress the embryonic/fetal β-like globin genes are constitutively reside at the periphery (25, 50). Our coregulator of great interest, as increasing embryonic/fetal β-like globin gene matrix constitutes a unique foundation that can be extended to expression in human hemoglobinopathies involving mutated or yield a genome-wide perspective of the complex relationships reduced levels of adult β-globin is efficacious (43); existing clinical between GATA-1 and requisite coregulators at target genes with strategies are relatively nonspecific. Genome-wide expression unique attributes, including local chromatin environment, higher- analysis revealed that SetD8 repressed a restricted cohort of genes order chromatin structure, and subnuclear neighborhood. This in erythroid cells, some of which are GATA-1–regulated (e.g., matrix will be an exceptionally powerful tool to dissect parameters murine embryonic/fetal β-like globin genes). SetD8 also re- dictating context-dependent GATA factor functions and will permit pressed genes that were not GATA-1–regulated, including genes sophisticated modeling to evaluate how alterations in the regulatory implicated in B-cell biology. The majority of SetD8-repressed parameters influence GATA factor-dependent genetic networks genes in erythroid cells were nonerythroid genes. GATA-1 oc- and downstream physiological and pathophysiological outputs. cupied SetD8/GATA-1–corepressed genes, and kinetic studies Erythrocyte development requires cell-intrinsic and -extrinsic imply direct GATA-1 regulation. GATA-1 used SetD8 at mechanisms that control commitment of multipotent hematopoi- genes that are FOG-1–sensitive or –insensitive and those that etic precursors, massive gene-expression changes, and sequential are Mi2β-sensitive or -insensitive. maturation steps, including gross organelle remodeling, that pre- Our loss-of-function studies with SetD8 and other coregulators pare for enucleation (1, 44, 51). Analogous to GATA-1, FOG-1 provide evidence for a coregulator matrix model of GATA factor is a master regulator of erythropoiesis with broad roles to establish function. This model assumes that distinct combinations of the erythroid cell phenotype. We predict that select GATA-1 coregulators confer target-gene regulation in a locus-specific coregulators have more specialized functions to confer specific manner. Conceptually, this model differs from the paradigm in components of the GATA-1–dependent genetic network, such as which an activator or repressor function via a common mechanism components dedicated to controlling the induction of autophagy

E3404 | www.pnas.org/cgi/doi/10.1073/pnas.1302771110 DeVilbiss et al. Downloaded by guest on October 1, 2021 evitisnesnI1-GOFevitisneS1-GOF evitisneS8DteS PNAS PLUS 3 6 3 A Gc2ata -Kit 1lyL 81sgR Vim CMa01cel yo1gKank3 1sgR 9 Fog1 * 5 4 * 2 4 2 3 * * 3 * 2 1 2 1 * * * 1

mRNA Levels mRNA 1 * * (Relative Units) * 0 0 0 0 β-estradiol control siRNA Fog1 siRNA FOG-1 FOG-1 Sensitive Insensitive SetD8 Sensitive B 1.5 Gata2 Lyl1 Rgs18 Vim Clec10a Myo1g Kank3 Rgs19 102 -3 ER-GATA-1/ * * * * * * 76 * * * * * * * V205G 1.0 r

M x 10 52 α-Tubulin 0.5 mRNA Levels mRNA (Relative Units) ER-GATA-1(V205G) 0 WT VG WT VG WT VG WT VG WT VG WT VG WT VG WT VG ER-GATA-1 Untreated β-estradiol

Fig. 6. SetD8/GATA-1-corepressed genes differentially require FOG-1. (A) FOG-1 knockdown in G1E-ER-GATA-1 cells. Real-time RT-PCR quantitation of mRNA levels of FOG-1–sensitive and –insensitive genes, as well as SetD8/GATA-1–corepressed genes (n =5± SE). (B) Gene expression analysis of GATA-1 target genes in G1E-ER-GATA-1 or G1E-ER-GATA-1(V205G) mutant cells matched for ER-GATA-1 expression. White bars, untreated; black bars, β-estradiol–treated. One ER-GATA-1 and four ER-GATA-1(V205G) clonal lines. Each untreated sample was normalized to a value of 1. For G1E-ER-GATA-1 samples, error bars represent SD from two technical replicates. For G1E-ER-GATA-1(V205G) samples, error bars represent SE from four biological replicates. *P < 0.05.

(52), which is required for organelle remodeling (53, 54), or enu- Vim (encoding Vimentin) is down-regulated during erythropoiesis cleation (55, 56). The G1E-ER-GATA-1 system recapitulates a as a key step in maturation-associated cytoskeletal remodeling normal window of maturation (38), but ER-GATA-1 activation in (57). The loss of Vimentin-containing intermediate filaments has this system does not drive efficient enucleation. Furthermore, been proposed to be a prerequisite for enucleation (58). Extending regulatory events underlying the genesis of proerythroblasts cannot our studies to interrogate coregulator requirements for specific be studied in this system. Although SetD8 was not required for components of the GATA-1–dependent genetic network in more Ter119 induction in G1E-ER-GATA-1 cells grown under conven- complex systems will almost certainly reveal additional SetD8/ tional conditions in a 2-d maturation assay, it repressed embryonic/ GATA-1–coregulated biological processes, and the work described fetal β-like globin genes and additional genes expected to have im- herein provides foundational insights to guide such studies. portant roles in cellular physiology. Our results establish the mo- Beyond GATA-1 mechanisms, we expect that the coregulator lecular underpinnings of a pivotal biological mechanism in which matrix model can be extrapolated to the actions of trans-acting

A FOG-1 Sensitive FOG-1 Insensitive SetD8 Sensitive 260 Mi2 Clec10a Myo1g Mi2 2.0 Gata2 c-Kit Lyl1 Rgs18 Vim Kank3 Rgs19 1.5 -3 * GENETICS * 1.5 * 1.0 102 *

r 1.0

M x 10 * 0.5 Tubulin 52 0.5 * * mRNA Levels 0 -estradiol (Relative Units) 0 control siRNA -estradiol Setd8 siRNA control siRNA Mi2 siRNA Mi2 siRNA

B GATA-1-Repressed/Mi2-Repressed SetD8 Activated: (228) (11)

SetD8 Activated/ Mi2-Activated: (1) GATA-1-Activated: Mi2- GATA-1-Repressed Mi2-Repressed (1386) Activated: (1388) (1133) (768)

SetD8 Repressed/ SetD8 Repressed/ GATA-1-Repressed Mi2-Repressed (19) SetD8 Repressed (23) GATA-1-Activated/ GATA-1/SetD8/Mi2 (93) Mi2-Activated: Repressed (205) (12)

Fig. 7. SetD8/GATA-1-corepressed genes differentially require the NuRD component, Mi2β.(A)Mi2β knockdown in G1E-ER-GATA-1 cells. (Left) Western blot to detect Mi2β protein levels. (Right) Real-time RT-PCR quantitation of mRNA of GATA-1 target genes that are FOG-1–sensitive, FOG-1–insensitive, and SetD8- sensitive (n =5± SE) *P < 0.05. (B, Left) Venn diagram depicting the extent of overlap between GATA-1–repressed, Mi2β-repressed, and SetD8-repressed genes. (Right) Venn diagram depicting overlap of GATA-1-activated, Mi2β-activated, and SetD8-activated genes.

DeVilbiss et al. PNAS | Published online August 19, 2013 | E3405 Downloaded by guest on October 1, 2021 A Gata2 c-Kit Lyl1 Rgs18 Vim Clec10A Myo1g Kank3 Rgs19 Clec4d Cpa3 Rgs1 Treml2 FOG-1 SetD8 Mi2β

Insensitive Sensitive

B Gata2 Lyl1 Rgs18 Clec10a NuRD c-Kit Rgs1 Vim SetD8 Mi2β SetD8 Treml2 Myo1g NuRD FOG-1 Mi2β GATA-1 GATA-1 GATA-1 GATA-1 Fig. 8. (A) Matrix depicting GATA-1 coregulator requirement at distinct loci. (B) Models depicting GATA-1 utilization Kank3 Rgs19 Clec4d Cpa3 of different combinations of corepres- NuRD SetD8 NuRD sors at distinct loci. Although the model SetD8 β β Mi2 Mi2 depicts factors implicated to function FOG-1 FOG-1 FOG-1 nonredundantly at the specifictarget GATA-1 GATA-1 GATA-1 GATA-1 genes, it does not imply whether ad- ditional factors are present or absent at the respective regulatory elements.

factors functioning at endogenous loci in diverse biological Quantitative ChIP. ChIP analysis in G1E-ER-GATA-1 cells was conducted as contexts. However, further studies on endogenous coregulator described previously (59). Briefly, samples containing 5 × 106 cells were actions at endogenous loci are required. Comparative analyses of crosslinked in 1% formaldehyde for 10 min. H4K20me1 was immunopreci- other factor-specific or tissue-specific matrices will almost cer- pitated using rabbit polyclonal anti-H4K20me1 antibody (Millipore). SetD8 was immunoprecipitated using a mouse monoclonal antibody (Abcam tainly uncover broadly important principles. ab3798) and GATA-1 was immunoprecipitated using a rabbit polyclonal antibody developed by the Bresnick Lab. Rabbit preimmune serum (Cova- Materials and Methods nce) was used as a control. Samples were quantitated using RT-PCR (Applied Cell Culture. G1E-ER-GATA-1 and G1E-ER-GATA-1(V205G) cells were cultured Biosystems Viia 7). Quantity of DNA was determined by SYBR green fluo- ’ fi ’ in Iscove s modi ed Dulbecco s medium (IMDM; Gibco) containing 15% (vol/ rescence, and the amount of product was determined relative to a standard vol) FBS (Gemini), 1% penicillin-streptomycin (Gemini), 2 U/mL erythropoie- curve created from serial dilution of input chromatin. tin, 120 nM monothioglycerol (Sigma), 0.6% conditioned medium from a Kit μ ligand-producing CHO cell line, and 1 g/mL puromycin (Gemini). ER-GATA-1 Protein Analysis. Protein samples were isolated by centrifugation of 1 × 106 μ β – activity was induced by treating cells with 1 M -estradiol (Steraloids). FOG-1 cells from each condition, washing with cold PBS, and lysing in 1× SDS sample null hematopoietic precursor cells were maintained in IMDM (Gibco) con- buffer (25 mM Tris, pH 6.8, 2% β-mercaptoethanol, 3% SDS, 0.005% bro- taining 15% FBS (Gemini), 1% antibiotic/antimycotic (Gemini), and 10 ng/mL mophenol blue, 5% glycerol). Samples were boiled for 10 min and stored at IL-3 (R&D Systems). Mouse erythroleukemia cells were cultured in DMEM −80 °C. Samples were resolved by SDS/PAGE, and proteins were detected by (Gibco) supplemented with 5% (vol/vol) FBS (Gemini). semiquantitative Western blotting with ECL Plus (GE Healthcare). Antibodies used were anti-SetD8 (Millipore 07–316), anti–GATA-1 (Santa Cruz Biotech- RNA Interference. Dharmacon siGenome SmartPool siRNAs targeting mouse nology; sc-265), anti–α-tubulin (Millipore; clone DM1A, 05–829). Secondary 6 SetD8, Fog1, and Mi2β were electroporated into 3 × 10 G1E-ER-GATA-1 cells antibodies included goat anti-mouse-IgG-HRP, goat anti-rabbit-IgG-HRP, or using an Amaxa Nucleofector (Lonza) coupled with Nucleofection Kit R goat anti-rat-IgG-HRP (Santa Cruz Biotechnology; sc-2005, sc-2030, sc-2032). (Lonza), as described previously (22, 50). Nontargeting siRNA (Dharmacon) served as a control. siRNA transfections were conducted twice (at 0 h and at Flow Cytometry. For flow cytometry, 1 × 105 cells were isolated by centri- 24 h). For mRNA and protein analysis, transfected cells were treated with fugation (6 min, 168 × g), washed with ice-cold PBS, and resuspended in β-estradiol at 24 h and harvested at 48 h. For flow cytometric analysis, cells 100 μL Annexin V Binding Buffer (Invitrogen). Cells were incubated with 5 μL were treated with β-estradiol at 24 h and harvested at 72 h. Alexa Fluor 350-conjugated Annexin V (Invitrogen A23202) at room tem- perature for 15 min in the dark. Ice-cold Annexin V binding buffer was Real-Time RT-PCR. Total RNA was purified with TRIzol (Invitrogen). To prepare added (400 μL), followed by 30 μL of 150 μg/mL propidium iodide solution in cDNA, 1 μg RNA was annealed with 250 ng of a 5:1 mixture of random PBS. Samples were maintained on ice and were analyzed using a BD LSR II hexamer and oligo(dT) primers by heating to 68 °C for 10 min. The annealed Flow Cytometer. Annexin V Alexa Fluor 350 was detected with the UV laser RNA/primers were incubated with murine Moloney leukemia virus reverse- (detector at 355 nm, filter at 450/50 nm), and propidium iodide was detected transcriptase (Invitrogen), 10 mM DTT (Invitrogen), RNAsin (Promega), and with the green laser (detector at 561 nm, filter at 450/50 nm). Data were 0.5 mM deoxynucleoside triphosphates (dNTPs) at 42 °C for 1 h in a total analyzed using FlowJo 9.5.2 software. reaction volume of 20 μL. This mixture was heat-inactivated at 95 °C for 5 min, and then diluted to a final volume of 100 μL. RT-PCR reactions con- Statistical Analysis. Statistical significance was determined by Paired Student’s tained 1.5 μL cDNA, 10 μL Power SYBR Green Master Mix (Applied Bio- t test using web-based GraphPad software (www.graphpad.com). Statistical systems), appropriate primers, and water to a total volume of 20 μL. PCR analysis of genome-wide expression data were conducted using EDGE3 product accumulation was monitored by SYBR green fluorescence. Relative software (60). Statistical significance of GO terms was conducted with the expression was determined from a standard curve of serial dilutions of cDNA Web-based National Institutes of Health DAVID tool (http://david.abcc. sample. As an internal control, all RNA measurements were normalized to ncifcrf.gov/). 18S RNA levels. ChIP-Seq. ChIP-seq profiles for GATA-1, H3K4me1, H3K27ac, and H3K4me1 in Transcriptional Profiling. RNA samples from three independent SetD8- mouse erythroleukemia cells were generated using the University of California knockdown and Mi2β-knockdown experiments in G1E-ER-GATA-1 cells were at Santa Cruz Genome Browser (http://genome.ucsc.edu/). Endogenous GATA- used for microarray analysis. mRNA was isolated and used to synthesize 1 ChIP-seq data were generated by Sherman Weissman (Yale University, New Amino Allyl RNA (aRNA). aRNA was labeled and hybridized to 8 × 60k Mouse Haven, CT) with an anti–GATA-1 antibody developed by the Bresnick Whole Genome arrays (Agilent), and read using a G-2505C DNA Microarray laboratory (GEO accession GSM912907). Data for H3K4me1 (Abcam ab8895, Scanner with Surescan High Resolution (Agilent). Data were analyzed using GEO accession GSM1000073), H3K4me3 (Millipore 07–473 GEO accession EDGE3, a Web-based two-color mircroarray analysis software, coupled with GSM1000087), and H3K27ac (Abcam ab4729, GEO accession GSM1000142) Microsoft Excel. Heat maps were generated using Java TreeView software. were generated by Bing Ren (University of California, San Diego, La Jolla, CA).

E3406 | www.pnas.org/cgi/doi/10.1073/pnas.1302771110 DeVilbiss et al. Downloaded by guest on October 1, 2021 ACKNOWLEDGMENTS. We thank members of the E.H.B. group for critical com and National Institutes of Health Grant T32GM081061(toA.W.D.),andaUniversity PNAS PLUS ments. This study was funded in part by National Institutes of Health Grant DK50107, of Wisconsin Comprehensive Cancer Center Support Grant P30 CA014520.

1. Hattangadi SM, Wong P, Zhang L, Flygare J, Lodish HF (2011) From stem cell to red 30. Oda H, et al. (2009) Monomethylation of histone H4-lysine 20 is involved in cell: Regulation of erythropoiesis at multiple levels by multiple proteins, RNAs, and chromosome structure and stability and is essential for mouse development. Mol Cell chromatin modifications. Blood 118(24):6258–6268. Biol 29(8):2278–2295. 2. Zon LI, et al. (1991) Expression of GATA-binding proteins during embryonic 31. Nishioka K, et al. (2002) PR-Set7 is a nucleosome-specific methyltransferase that development in Xenopus laevis. Proc Natl Acad Sci USA 88(23):10642–10646. modifies lysine 20 of histone H4 and is associated with silent chromatin. Mol Cell 9(6): 3. Yamamoto M, et al. (1990) Activity and tissue-specific expression of the transcription 1201–1213. factor NF-E1 multigene family. Genes Dev 4(10):1650–1662. 32. Abbas T, et al. (2010) CRL4(Cdt2) regulates cell proliferation and histone gene 4. Tsai FY, et al. (1994) An early haematopoietic defect in mice lacking the transcription expression by targeting PR-Set7/Set8 for degradation. Mol Cell 40(1):9–21. factor GATA-2. Nature 371(6494):221–226. 33. Karachentsev D, Sarma K, Reinberg D, Steward R (2005) PR-Set7-dependent 5. Evans T, Felsenfeld G (1989) The erythroid-specific transcription factor Eryf1: A new methylation of histone H4 Lys 20 functions in repression of gene expression and is – finger protein. Cell 58(5):877–885. essential for mitosis. Genes Dev 19(4):431 435. fi 6. Tsai SF, et al. (1989) Cloning of cDNA for the major DNA-binding protein of the 34. Barski A, et al. (2007) High-resolution pro ling of histone methylations in the human – erythroid lineage through expression in mammalian cells. Nature 339(6224):446–451. genome. Cell 129(4):823 837. 7. Pevny L, et al. (1991) Erythroid differentiation in chimaeric mice blocked by a targeted 35. Cui K, et al. (2009) Chromatin signatures in multipotent human hematopoietic stem mutation in the gene for transcription factor GATA-1. Nature 349(6306):257–260. cells indicate the fate of bivalent genes during differentiation. Cell Stem Cell 4(1): – 8. Fujiwara Y, Browne CP, Cunniff K, Goff SC, Orkin SH (1996) Arrested development of 80 93. 36. Weiss MJ, Yu C, Orkin SH (1997) Erythroid-cell-specific properties of transcription embryonic red cell precursors in mouse embryos lacking transcription factor GATA-1. factor GATA-1 revealed by phenotypic rescue of a gene-targeted cell line. Mol Cell Proc Natl Acad Sci USA 93(22):12355–12358. Biol 17(3):1642–1651. 9. Grass JA, et al. (2003) GATA-1-dependent transcriptional repression of GATA-2 via 37. Gregory T, et al. (1999) GATA-1 and erythropoietin cooperate to promote erythroid disruption of positive autoregulation and domain-wide chromatin remodeling. Proc cell survival by regulating bcl-xL expression. Blood 94(1):87–96. Natl Acad Sci USA 100(15):8811–8816. 38. Welch JJ, et al. (2004) Global regulation of erythroid gene expression by transcription 10. Bresnick EH, Lee HY, Fujiwara T, Johnson KD, Keles S (2010) GATA switches as factor GATA-1. Blood 104(10):3136–3147. developmental drivers. J Biol Chem 285(41):31087–31093. 39. Zhang J, Socolovsky M, Gross AW, Lodish HF (2003) Role of Ras signaling in erythroid 11. Bresnick EH, Katsumura KR, Lee HY, Johnson KD, Perkins AS (2012) Master regulatory differentiation of mouse fetal liver cells: Functional analysis by a flow cytometry- GATA transcription factors: Mechanistic principles and emerging links to hematologic based novel culture system. Blood 102(12):3938–3946. – malignancies. Nucleic Acids Res 40(13):5819 5831. 40. Fujiwara T, et al. (2009) Discovering hematopoietic mechanisms through genome- fi 12. Crispino JD, Lodish MB, MacKay JP, Orkin SH (1999) Use of altered speci city mutants wide analysis of GATA factor chromatin occupancy. Mol Cell 36(4):667–681. fi to probe a speci c protein-protein interaction in differentiation: the GATA-1:FOG 41. Yu M, et al. (2009) Insights into GATA-1-mediated gene activation versus repression – complex. Mol Cell 3(2):219 228. via genome-wide chromatin occupancy analysis. Mol Cell 36(4):682–695. fi 13. Tsang AP, et al. (1997) FOG, a multitype zinc nger protein, acts as a cofactor for 42. Cheng Y, et al. (2009) Erythroid GATA1 function revealed by genome-wide analysis of transcription factor GATA-1 in erythroid and megakaryocytic differentiation. Cell transcription factor occupancy, histone modifications, and mRNA expression. Genome 90(1):109–119. Res 19(12):2172–2184. 14. Pal S, et al. (2004) Coregulator-dependent facilitation of chromatin occupancy by 43. Sankaran VG, Orkin SH (2013) The switch from fetal to adult hemoglobin. Cold Spring GATA-1. Proc Natl Acad Sci USA 101(4):980–985. Harb Perspect Medicine 3(1):a011643. 15. Letting DL, Chen YY, Rakowski C, Reedy S, Blobel GA (2004) Context-dependent 44. Kingsley PD, et al. (2013) Ontogeny of erythroid gene expression. Blood 121(6): regulation of GATA-1 by friend of GATA-1. Proc Natl Acad Sci USA 101(2):476–481. e5–e13. 16. Hong W, et al. (2005) FOG-1 recruits the NuRD repressor complex to mediate 45. Johnson KD, et al. (2007) Friend of GATA-1-independent transcriptional repression: A transcriptional repression by GATA-1. EMBO J 24(13):2367–2378. novel mode of GATA-1 function. Blood 109(12):5230–5233. 17. Gao Z, et al. (2010) FOG-1-mediated recruitment of NuRD is required for cell lineage 46. Johnson KD, Kim SI, Bresnick EH (2006) Differential sensitivities of transcription factor re-enforcement during haematopoiesis. EMBO J 29(2):457–468. target genes underlie cell type-specific gene expression profiles. Proc Natl Acad Sci 18. Miccio A, et al. (2010) NuRD mediates activating and repressive functions of GATA-1 USA 103(43):15939–15944. and FOG-1 during blood development. EMBO J 29(2):442–456. 47. Zhang Y, LeRoy G, Seelig HP, Lane WS, Reinberg D (1998) The dermatomyositis- 19. Kim SI, Bultman SJ, Kiefer CM, Dean A, Bresnick EH (2009) BRG1 requirement for long- specific autoantigen Mi2 is a component of a complex containing histone deacetylase range interaction of a locus control region with a downstream promoter. Proc Natl and nucleosome remodeling activities. Cell 95(2):279–289. Acad Sci USA 106(7):2259–2264. 48. Wade PA, et al. (1999) Mi-2 complex couples DNA methylation to chromatin 20. Kim SI, Bultman SJ, Jing H, Blobel GA, Bresnick EH (2007) Dissecting molecular steps in remodelling and histone deacetylation. Nat Genet 23(1):62–66. chromatin domain activation during hematopoietic differentiation. Mol Cell Biol 49. Li Z, Nie F, Wang S, Li L (2011) Histone H4 Lys 20 monomethylation by histone methylase SET8 mediates Wnt target gene activation. Proc Natl Acad Sci USA 108(8): 27(12):4551–4565. GENETICS – 21. Blobel GA, Nakajima T, Eckner R, Montminy M, Orkin SH (1998) CREB-binding protein 3116 3123. cooperates with transcription factor GATA-1 and is required for erythroid 50. Lee HY, Johnson KD, Boyer ME, Bresnick EH (2011) Relocalizing genetic loci into fi – differentiation. Proc Natl Acad Sci USA 95(5):2061–2066. speci c subnuclear neighborhoods. J Biol Chem 286(21):18834 18844. 22. Pope NJ, Bresnick EH (2010) Differential coregulator requirements for function of the 51. Merryweather-Clarke AT, et al. (2011) Global gene expression analysis of human erythroid progenitors. Blood 117(13):e96–e108. hematopoietic transcription factor GATA-1 at endogenous loci. Nucleic Acids Res 52. Kang YA, et al. (2012) Autophagy driven by a master regulator of hematopoiesis. Mol 38(7):2190–2200. Cell Biol 32(1):226–239. 23. Stumpf M, et al. (2006) The mediator complex functions as a coactivator for GATA-1 53. Kundu M, et al. (2008) Ulk1 plays a critical role in the autophagic clearance of in erythropoiesis via subunit Med1/TRAP220. Proc Natl Acad Sci USA 103(49): mitochondria and ribosomes during reticulocyte maturation. Blood 112(4): 18504–18509. 1493–1502. 24. Kim SI, Bresnick EH, Bultman SJ (2009) BRG1 directly regulates nucleosome structure 54. Zhang J, et al. (2009) Mitochondrial clearance is regulated by Atg7-dependent and and chromatin looping of the alpha globin locus to activate transcription. Nucleic -independent mechanisms during reticulocyte maturation. Blood 114(1):157–164. – Acids Res 37(18):6019 6027. 55. Ji P, Murata-Hori M, Lodish HF (2011) Formation of mammalian erythrocytes: 25. Lee HY, et al. (2009) Controlling hematopoiesis through sumoylation-dependent Chromatin condensation and enucleation. Trends Cell Biol 21(7):409–415. – regulation of a GATA factor. Mol Cell 36(6):984 995. 56. Keerthivasan G, Wickrema A, Crispino JD (2011) Erythroblast enucleation. Stem Cells 26. Wu C, et al. (2009) BioGPS: An extensible and customizable portal for querying and Int 2011:139851. organizing gene annotation resources. Genome Biol 10(11):R130. 57. Granger BL, Lazarides E (1983) Expression of the major neurofilament subunit in 27. Beck DB, Oda H, Shen SS, Reinberg D (2012) PR-Set7 and H4K20me1: At the crossroads chicken erythrocytes. Science 221(4610):553–556. of genome integrity, cell cycle, chromosome condensation, and transcription. Genes 58. Sangiorgi F, Woods CM, Lazarides E (1990) Vimentin downregulation is an inherent Dev 26(4):325–337. feature of murine erythropoiesis and occurs independently of lineage. Development 28. Couture JF, Dirk LM, Brunzelle JS, Houtz RL, Trievel RC (2008) Structural origins for the 110(1):85–96. product specificity of SET domain protein methyltransferases. Proc Natl Acad Sci USA 59. Im H, et al. (2004) Measurement of protein-DNA interactions in vivo by chromatin 105(52):20659–20664. immunoprecipitation. Methods Mol Biol 284:129–146. 29. Oda H, et al. (2010) Regulation of the histone H4 monomethylase PR-Set7 by 60. Vollrath AL, Smith AA, Craven M, Bradfield CA (2009) EDGE(3): A web-based CRL4(Cdt2)-mediated PCNA-dependent degradation during DNA damage. Mol solution for management and analysis of Agilent two color microarray experiments. Cell 40(3):364–376. BMC Bioinformatics 10:280.

DeVilbiss et al. PNAS | Published online August 19, 2013 | E3407 Downloaded by guest on October 1, 2021