Pathogenic GATA2 genetic variants utilize an obligate enhancer mechanism to distort a multilineage differentiation program

Significance

Germline genetic alterations of the gene encoding GATA2, a master regulator of hematopoiesis, create a predisposition to develop blood diseases including myelodysplasia and acute myeloid leukemia. Although genetic variation alters GATA2 function, the molecular and cellular consequences of the variants are incompletely understood. We demonstrate that pathogenic GATA2 variants have both unique and GATA2-like attributes and retain the capacity to activate an enhancer-dependent mechanism, which generates a fragmented differentiation program. This mechanism elevates C/EBPε levels and generates a feedforward loop, representing a paradigm of GATA factor function and cellular differentiation.

Abstract

Mutations in genes encoding transcription factors inactivate or generate ectopic activities to instigate pathogenesis. By disrupting hematopoietic stem/progenitor cells, GATA2 germline variants create a bone marrow failure and leukemia predisposition, GATA2 deficiency syndrome, yet mechanisms underlying the complex phenotypic constellation are unresolved. We used a GATA2-deficient progenitor rescue system to analyze how genetic variation influences GATA2 functions. Pathogenic variants impaired, without abrogating, GATA2-dependent transcriptional regulation. Variants promoted eosinophil and repressed monocytic differentiation without regulating mast cell and erythroid differentiation. While GATA2 and T354M required the DNA-binding C-terminal zinc finger, T354M disproportionately required the N-terminal finger and N terminus. GATA2 and T354M activated a CCAAT/Enhancer Binding Protein-ε (C/EBPε) enhancer, creating a feedforward loop operating with the T-cell Acute Lymphocyte Leukemia-1 (TAL1) transcription factor. Elevating C/EBPε partially normalized hematopoietic defects of GATA2-deficient progenitors. Thus, pathogenic germline variation discriminatively spares or compromises transcription factor attributes, and retaining an obligate enhancer mechanism distorts a multilineage differentiation program.

Contrasting with models in which mutations elicit strictly gain-of-function or loss-of-function consequences, vast human genetic variation can elicit nonbinary functional consequences. This scenario exemplifies genes with pathogenic consequences for the hematopoietic system, e.g., DDX41 encoding an RNA helicase that regulates innate immunity and erythropoiesis (1–3), and GATA2 encoding a transcriptional regulator of hematopoietic stem and progenitor cell (HSPC) development and function (4). Though hundreds of germline and acquired GATA2 variants have been identified (5–8), many variants cannot be deemed unequivocally as pathogenic or benign. The variants disrupt critical sequences, create frameshifts that lower protein levels, and alter sequences of unknown function.

Inherited germline GATA2 coding sequence variants cause variably penetrant GATA2 deficiency syndrome in the adult or in pediatric contexts characterized by predisposition to myelodysplastic syndrome (MDS), acute myeloid leukemia (AML), and immunodeficiency (9–11). Noncoding germline variants disrupt a GATA2 intronic +9.5 enhancer, reducing GATA2, and establishing conditional pathogenicity that can be triggered to bone marrow failure and MDS/AML (12–15). Though somatic GATA2 variants can occur in adult and pediatric MDS/AML patients (7), whether these variants instigate or promote pathogenesis is unclear. Analyses with diverse systems indicate that GATA2 variation generates proteins with impaired DNA binding, chromatin occupancy, transcriptional regulation, and/or differentiation-regulatory activities (16–20). As +9.5 enhancer variants disrupt HSPC generation, activity, and/or regeneration (12–15, 21), and coding variants can reduce GATA2 activities (16–20), insufficient levels/activities of wild-type GATA2 underlie GATA2 deficiency syndrome. Variants discriminatively impair certain, but not all, GATA2 activities, and variants can retain activity or exert exaggerated activities in vitro (17–19). Considering that GATA2 activates and represses several thousand genes (22–25), partially disrupting GATA2-dependent genetic and protein networks by reducing GATA2 levels/activities, thereby reducing or elevating select network components, may derail mechanisms governing HSPC generation and function. GATA2 promotes the development of erythrocytes, megakaryocytes, and myeloid cells, including mast cells, dendritic cells, and eosinophils (4, 26–30). Fragmented networks are of considerable interest, given opportunities to supplant diminished and attenuate exaggerated mechanisms.

To elucidate mechanisms underlying how clinical genetic variation impacts GATA2 function, we developed rescue systems involving GATA2 replacement at near-physiological levels in GATA2-deficient primary and immortalized fetal progenitors (18–20). We demonstrated that the T354M variant, which modifies the DNA-binding C-finger, was defective in inducing erythroid, but not myeloid, colony formation (18). The R307W variant, which modifies the N-finger, generated more myeloid colonies than GATA2, and at several genes tested, either lost or retained activity. By contrast, a germline insertion of nine amino acids between the GATA2 zinc fingers (9aa-Ins) (20) was defective in chromatin occupancy, chromatin remodeling, and transcriptional regulation at many of its targets genome-wide (19). The insertion more severely impaired activation vs. repression, and 9aa-Ins retained activity at a minority of target loci (19).

Herein, we utilized a panel of clinical variants to dissect how GATA2 controls transcription and differentiation and how variants impact these processes. T354M lost a subset of its activities to control transcription and differentiation yet was competent to activate genes expressed in eosinophils. GATA2 and T354M functioned through a Cebpe enhancer that dictates the levels of the transcription factor C/EBPε, and C/EBPε was required for context-dependent GATA2-mediated transcriptional activation. Ablation of the enhancer down-regulated C/EBPε in progenitors, which abrogated select GATA2 and T354M activities. These studies unveiled an obligate enhancer and transcription factor feedforward mechanism underlying context-dependent GATA2 activity. Though GATA factors operate via context-dependent mechanisms (31–33), the molecular determinants that commission GATA factor activity at a restricted target gene cohort and preclude its activity at others are unknown. GATA2 occupies thousands of chromatin sites (19, 22, 34–36), but this is the only example of a GATA2-regulated enhancer, other than two Gata2 enhancers (12, 37–40), that is a vital mediator of GATA2-dependent functions in HSPCs. Surprisingly, this mechanism, which establishes the genetic framework for eosinophil development, is retained by disease variants that are considered to be loss of function. Rather than inactivating GATA2, commensurate with expectations for a canonical loss-of-function mutation, retaining the enhancer mechanism supports only one component of the differentiation program, thus generating a distorted program that is insufficient to fulfill requirements for multilineage hematopoiesis.

Results

Neomorphic GATA2 Disease Variants Selectively Retain and Acquire Activities.

Previously, we demonstrated that expressing GATA2, T354M, or R307W in GATA2-deficient myeloerythroid progenitors lacking the Gata2 −77 enhancer (−77^−/− Lin⁻ cells) increased myeloid colonies, while only GATA2 increased erythroid colonies (18). As the potential mechanistic similarities and differences between GATA2, T354M, and R307W activities are unresolved, we used this rescue system to test how GATA2 disease-linked variation impacts GATA2 function. To facilitate mechanistic studies, we adapted the rescue system to a 3-d liquid culture. GATA2, T354M, and R307W were retrovirally expressed in GATA2-deficient cells cultured with IL-3 and SCF (Fig. 1 A and B). The levels of expressed wild-type and mutant GATA2 were comparable to that of endogenous GATA2 (Fig. 1B). While GATA2 and R307W induced endogenous GATA2 as described (20), T354M had weaker activity to elevate GATA2 (Fig. 1A). While -77^+/+ cells produced erythroid and myeloid cells, −77^−/− cells generated predominantly myeloid cells (SI Appendix, Fig. S1A) as described (37). GATA2 or disease variant expression in −77^−/− cells did not obviously alter cell morphology (SI Appendix, Fig. S1A). Flow cytometric analysis revealed that 54% of −77^+/+ cells expressed erythroid markers. −77^−/− cells were devoid of erythroid markers and did not undergo erythroid differentiation (SI Appendix, Fig. S1B).

Fig. 1.

GATA2 disease variants are defective in regulating differentiation and transcription in a genetic rescue system. (A) Schematic representation of the genetic rescue system. (B) Left: Representative western blot analysis of −77^−/− fetal liver cells expressing HA-GATA2 or variants with anti-HA antibody. Right: Representative western blot analysis of −77^+/+ and −77^−/− fetal liver cells expressing HA-GATA2 or variants with anti-GATA2 antibody. (C) Flow cytometric analysis of GFP⁺ cells stained for monocytic cells (CD11b⁺CD115⁺), granulocytic cells (CD11b⁺CD115⁻), macrophage (F4/80⁺CD11b⁺), mast cells (FcεR1α⁺Kit⁺), and eosinophils (CD11b⁺SiglecF⁺). (D) Quantitation of flow cytometry data (n = 5). (E) Heatmap showing the result of RT-qPCR analysis of erythroid, mast cell, and eosinophil genes (n = 5). (F) Flow cytometric analysis of GFP⁺ cells stained for monocytic cells (CD11b⁺CD115⁺), granulocytic cells (CD11b⁺CD115⁻), macrophages (F4/80⁺CD11b⁺), mast cells (FcεR1α⁺Kit⁺), and eosinophils (CD11b⁺SiglecF⁺). (G) Quantitation of flow cytometry data (n = 4). * < 0.05, ** < 0.01, and *** < 0.001.

We tested whether GATA2 or disease variant expression in the rescue system alters the progenitor immunophenotype. GATA2, T354M or R307W expression reduced monocytic (CD115⁺CD11b⁺) and increased granulocytic cells (CD115⁻CD11b⁺) (Fig. 1 C and D). Regardless of GATA2 or T354M expression, CD115 and CD11b expression were unaltered in the uninfected GFP^- population (SI Appendix, Fig. S1C). Since GATA2 suppresses monocytic differentiation and activates Tpsb2 and Ear2, which are expressed in mast cells and eosinophils, respectively (41), we analyzed macrophage, mast cell, and eosinophil differentiation. The −77^−/− cells generated increased macrophage-like cells (F4/80⁺CD11b⁺) (Fig. 1 C and D), consistent with −77^−/− mouse embryos (37, 40, 42). GATA2 deficiency syndrome patients exhibit monocytopenia yet have normal macrophage levels in bone marrow (43). Expressing GATA2 or variants reduced F4/80⁺CD11b⁺ cells. After a 3-d culture, GATA2, but not disease variants, increased cells with a mast cell immunophenotype (FcεR1α⁺Kit⁺) (Fig. 1 C and D). As mast cell differentiation requires 2 to 3 wk of culture, these results suggest that GATA2 primed mast cell differentiation in precursors. Giemsa staining did not reveal mast cells at this early stage (SI Appendix, Fig. S1A). As GATA1 and GATA2 promote eosinophil development (44, 45), and GATA2 induces Gata1 expression in myeloid progenitors (40), we quantified CD11b and SiglecF expression as metrics of eosinophil differentiation (46). −77 enhancer deletion decreased eosinophil (CD11b⁺SiglecF⁺) numbers (Fig. 1 C and D). GATA2 and disease variants increased CD11b⁺SiglecF⁺ cells.

We evaluated whether transcriptional activities provide mechanistic insights into the control of cellular transitions. RT-qPCR analysis revealed that GATA2 activated the erythroid genes Gata1 and Zfpm1 to 67% and 69% of −77^+/+ Lin^- cells, respectively (Fig. 1E). T354M activated Gata1 and Zfpm1 to 20% and 39% of −77^+/+ cells. R307W activated Gata1 and Zfpm1 to 40% and 49% of −77^+/+ cells. These genes are expressed at low levels in progenitors. Zfpm1 encodes the GATA1 coregulator FOG1, and GATA1 increases Zfpm1 mRNA (47) and FOG1 protein (48) expression. GATA2 did not activate hallmark genes of differentiating erythroblasts (25), e.g., Hbb-b1, Ahsp, Alas2, Slc4a1, and Epb4.9 (Fig. 1E). Similarly, GATA2, but not T354M or R307W, regulated mast cell genes. By contrast, T354M and R307W activated eosinophil genes, and in certain cases, their activities exceeded that of GATA2 (Fig. 1E). T354M and R307W activities were not identical, as T354M, but not R307W, strongly activated Epx (Fig. 1E). R307W activated Ear2 to a greater extent than GATA2, whereas T354M did not increase Ear2 expression (Fig. 1E). Consistent with the differentiation-regulatory activity, GATA2 activated mast cell genes and eosinophil genes, while T354M and R307W activated only eosinophil genes.

Comparable analyses were conducted to determine whether additional patient germline variants in the zinc fingers [N-finger, A318T; Inter-zinc finger space, 9aa-Ins; C-finger, T355Δ, R362P, and R362Q (7, 20, 49–51)] behave similarly or distinctly from the GATA2 variants tested. A318T, R362P, and R362Q have been described as somatic or germline variants (7). Alphafold revealed the positioning of T354 adjacent to the C-finger Zn²⁺ center (C349, C352, C370, and C373) and is predicted to form hydrogen bonds with C349 and C352 (SI Appendix, Fig. S1D). With the methionine substitution for threonine, a hydroxyl group was exchanged for a methyl thioester side chain. This change is predicted to result in loss of one of the three hydrogen bonds with C352. The model did not predict alterations due to substitution of R362 with proline (R362P) or glutamine (R362Q). R307 in the N finger can form hydrogen bonds with L315 and N317, and these interactions are maintained in the predicted R307W structure. A318 in the alpha chain of the N finger forms a hydrogen bond with the surrounding Y322 residue and is adjacent to the Zn²⁺ center. Alanine to threonine substitution of this residue was not predicted to alter these interactions.

R362Q and A318T resembled GATA2 and T354M in suppressing monocytic differentiation. They also induced granulocytic and eosinophil differentiation (Fig. 1 F and G). Only R362Q elevated mast cell lineage markers and weakly activated mast cell genes (Fig. 1 F and G). However, FcεR1α expression was significantly lower than GATA2 (SI Appendix, Fig. S1E). Although 9aa-Ins and R362P had weaker differentiation-regulatory activity, they increased granulocytic and decreased monocytic cells. T355Δ was largely inactive. These data extend the repertoire of GATA2 disease variants that share a T354M activity profile and reveal the relationship between gene expression defects and the loss of certain, but not all, differentiation potentials. As the variants suppressed monocytic and promoted eosinophil differentiation, without regulating mast cell and erythroid differentiation, the GATA2- and disease variant-instigated gene expression programs differed significantly in their capacities to induce or support progenitor transitions.

GATA2 Disease Variants Create a Fragmented Hematopoietic Progenitor Transcriptome.

Since the GATA2 disease variants shared select GATA2 activities to promote differentiation, we utilized RNA-seq to gain insights into how T354M and R307W variant mechanisms differ from that of GATA2. GATA2-, T354M-, and R307W-regulated genes were not universally identical. Out of 794 GATA2-activated genes, 278 genes were T354M-activated, and 501 genes were R307W-activated (Fig. 2A). Out of 516 GATA2-repressed genes, T354M and R307W repressed 268 and 275 genes, respectively. The top 20 GATA2 or variant-activated genes revealed similarities and differences. While GATA2, T354M, and R307W regulated Prg2, Prg3, Il5ra, Chst13, Hdc, Rnase12, and Pdzph1, other genes were regulated mutually exclusively by GATA2 or disease variants (SI Appendix, Fig. S2A). Cftr, Gm29018, Eml5, Dach1, AC159193.1, and Hao1 were regulated solely by T354M. Mcpt8 was regulated solely by GATA2. Ces1d, Heph, Nepn, and 5330417C22Rik were regulated solely by R307W. Thus, the disease variants create aberrant progenitor transcriptomes.

Fig. 2.

Context-dependent T354M occupancy of chromatin sites and transcriptional regulation. (A) Venn diagram depicting overlap between GATA2- and disease variant-regulated genes. (B) Heatmap depicting RT-qPCR analysis of GATA2-regulated genes (n = 5). (C) RT-qPCR analysis of GATA2-regulated mRNAs and primary transcripts in GATA2- or T354M-expressing cells (n = 5). (D) Representative western blot analysis of hi-77^−/− cells expressing HA-GATA2 or variants with anti-HA antibody. (E) Representative image of Giemsa-stained cells. Cells were cultured for 3 d. (F) Heatmap depicting RT-qPCR analysis of GATA2-regulated mRNAs identified with GATA2- or disease variant-expressing hi-77^−/− cells (n = 6). (G) Quantitative ChIP analysis at GATA2-occupied sites in target genes with anti-HA antibody (n = 5). (H) Quantitative ChIP analysis at GATA2-occupied sites in target genes with anti-GATA2 antibody (n = 4). (I) CUT&RUN profile of GATA2 at Ms4a3 locus (GSE171384). (J) PCR-genotyping assay for −77^−/−Ms4a3+27.6^−/− allele. (K) RT-qPCR analysis of Ms4a3 mRNA levels (n = 4). * < 0.05, ** < 0.01, and *** < 0.001.

GSEA analysis revealed that GATA2-activated genes in −77^−/− cells conform to a hallmark mast cell signature (Dataset Table S1, P = 0.0029). T354M also regulated mast cell genes (P = 0.041); out of 128 of these genes, GATA2 activated 41 and repressed 8. T354M activated 19 and repressed 17. Thus, GATA2- and T354M differed in their capacities to regulate mast cell genes (P = 0.0023) (SI Appendix, Fig. S2B, and Dataset Table S1). Although multiple signatures were unique to GATA2 vs. T354M and R307W, all three regulated genes conformed to IFN gamma response, complement, and angiogenesis signatures (Dataset Table S1).

RNA-Seq analysis revealed that GATA2, T354M, and R307W activated 260, 129, and 70 genes uniquely and repressed 159, 210, and 131 genes uniquely, respectively (Fig. 2A). RT-qPCR validation of the RNA-Seq data confirmed that GATA2, but not R307W and T354M, activated Tpsab1, Slc6a9, Fam198b, Itga2, Espn, and Ms4a2 (Fig. 2B). T354M and R307W activated Ncam1 and Nrg4 to a greater extent than GATA2 (Fig. 2B). However, T354M and R307W did not regulate an identical gene cohort. R307W, but not T354M, activated Ces1d and Heph (Fig. 2B). T354M, but not R307W, activated Ctsg, Elane, Ms4a3, Rab38, Cpa3, Anxa1, and Cldn15 (Fig. 2B).

Since GATA2 increased Gata1 expression and mast cell and eosinophil differentiation, we analyzed genes reported to be activated in GATA1⁺ Pre-GMs that can differentiate into mast cells, eosinophils, erythroid cells, and megakaryocytes (52) (SI Appendix, Fig. S2C). The top activated genes from GATA1⁺ Pre-GMs were compared to our RNA-Seq data from the rescue assay. −77 enhancer deletion decreased expression of 65% of these genes (SI Appendix, Fig. S2C). GATA2 or R307W expression activated many of these genes in −77^−/− Lin⁻ cells, and T354M activated 60% of the genes (SI Appendix, Fig. S2C). To test whether T354M-regulated mRNAs reflect transcriptional changes, we quantified mRNA and primary transcripts. GATA2 and T354M increased mRNAs and primary transcripts (Fig. 2C), indicating that T354M retains context-dependent transcriptional regulatory activity.

Molecular Determinants of GATA2 Disease Variant Transcriptional-Regulatory Activity.

T354M is considered to be a loss-of-function variant with compromised DNA binding and transcriptional-regulatory activities (16, 17, 53, 54). In principle, T354M might activate genes via DNA binding to WGATAR motifs, tethering to DNA-bound factors, or indirectly by regulating another gene. We quantified activity in immortalized −77^−/− Lin⁻ cells (hi-77^−/− cells) (43) by expressing GATA2, T354M, or R307W (Fig. 2D). GATA2 or disease variant expression in hi-77^−/− cells did not alter cellular morphology (Fig. 2E). As detected with −77^−/− Lin⁻ cells, while the transcriptional regulatory activity of variants was attenuated at a GATA2 target gene cohort, including Tpsb2, the variants retained activity to regulate other GATA2 target genes, including Prg2 and ectopic target genes (Fig. 2F). Expression of other variants at levels resembling GATA2 and T354M revealed a similar regulatory behavior as T354M or R307W. R362Q regulated T354M-regulated genes, and A318T regulated R307W-regulated genes. Although 9aa-Ins, R362P, and T355Δ had weaker activities, they regulated select target genes e.g., Ctsg (SI Appendix, Fig. S2 D and E).

We tested whether T354M occupies GATA2-bound chromatin sites identified in our prior studies. While GATA2 and T354M occupied the Gata2 +9.5 enhancer and Bcl2l1 intron similarly, at other loci, little to no T354M occupancy was detected (Fig. 2G). We identified GATA2-occupied sites in genes analyzed in this study and compared GATA2 and T354M occupancy at these sites. HA-GATA2 occupied Il1rl1 −10.6 kb, Ikzf2 +171 kb, Tpsb2 -73 kb, Irf8 -242 kb, Ms4a3 +27.6 kb, Slc45a3 −20 kb, Anxa1 −19 kb, Epx +116 kb, Cebpe +6 kb, Hdc −17 kb, Elane +147 kb, Il18r1 int3, and Cpa3 −1.9 kb (Fig. 2G). Little to no T354M occupancy was detected at Tpsb2 −73 kb, Il1rl1 −10.6 kb, and Slc45a3 −20 kb. T354M and GATA2 occupied Ikzf2 +171 kb, Irf8 −242 kb, Hdc −17 kb, Anxa1 −19 kb, Ms4a3 +27.6 kb, Ctsg +9 kb, and Epx +116 kb (Fig. 2G). ChIP with anti-HA and anti-GATA2 antibodies yielded similar results (Fig. 2H). GATA2 occupancy was low in hi-77^−/− cells that express 70 to 80% less GATA2 vs. hi-77^+/+ cells. CRISPR/Cas9-mediated deletion of Ms4a3 +27.6 kb severely decreased T354M-mediated activation of Ms4a3 (Fig. 2 I–K and SI Appendix, Fig. S2F) without affecting expression of the neighboring GATA2-regulated gene, Ms4a2 (SI Appendix, Fig. S2G). T354M regulation of Ms4a3 was 88.6% lower in −77^−/−Ms4a3+ 27.6^−/−clone 1 (P = 0.001) and 77.4% lower in −77^−/−Ms4a3+ 27.6^−/−clone 2 (P = 0.005).

Since T354M DNA binding is compromised in vitro, but occupancy at select chromatin sites is retained, we tested whether sequences mediating DNA binding are required for T354M function. As the C-finger of GATA factors binds DNA with sequence-specificity (55–57), and the N-finger stabilizes binding at certain sites (58), we analyzed N-finger and C-finger cysteine mutants predicted to disrupt fingers (Fig. 3A). C370A abrogated GATA2- and T354M-mediated activation of Prg2, Cpa3, Epx, and Hdc (Fig. 3B). C295A abrogated GATA2- and T354M-mediated activation of Prg2 and Epx. Whereas C295A did not affect GATA2-mediated Cpa3 and Hdc activation, combining C295A with T354M abrogated T354M-mediated activation of Cpa3 and Hdc (Fig. 3C). We also tested T354M-specific target genes. C370A disrupted expression of most of these genes save Ctsg (Fig. 3D). C295A abrogated activation of all tested T354M-specific targets (Fig. 3D). Although these mutations inhibited GATA2 and T354M chromatin occupancy at certain sites (Fig. 3E and SI Appendix, Fig. S3A), C295A occupied Cpa3 -1.9 kb, Hdc -17 kb, and Bcl2l1int, and C295A activated Cpa3 and Hdc (Fig. 3E and SI Appendix, Fig. S3A).

Fig. 3.

Zinc finger requirements for T354M-mediated transcriptional regulation. (A) Schematic representation of zinc finger structures. (B) Left, Representative western blot analysis of HA-GATA2 or variants using anti-HA antibody. Right, RT-qPCR analysis of GATA2-regulated mRNAs in cells expressing GATA2 or variants (n = 5). (C) Left, Representative western blot analysis of HA-GATA2 or variants using anti-HA antibody. Right, RT-qPCR analysis of GATA2-regulated mRNAs in cells expressing GATA2 or variants (n = 5). (D) RT-qPCR analysis of T354M-regulated mRNAs in cells expressing GATA2 or variants (n = 5). (E) Quantitative ChIP analysis at GATA-occupied sites in target genes with anti-HA antibody (n = 3). (F) RT-qPCR analysis of Irf8 mRNA in cells expressing GATA2 or variants (n = 5). (G) Left, Representative western blot analysis of HA-GATA2 or variants using anti-HA antibody. Right, RT-qPCR analysis of GATA2-regulated mRNAs in GATA2- or variant-expressing cells (n = 4). * < 0.05, ** < 0.01, and *** < 0.001.

We tested whether zinc fingers are required for GATA2- and T354M-mediated repression. The innate immune genes repressed by GATA2 include Irf8 encoding an interferon-γ-activated transcription factor (19, 42, 43, 59). C370A, but not C295A, lost activity to repress Irf8 expression (Fig. 3F). A variant harboring C295A and T354M could not repress Irf8 (Fig. 3F). Thus, the C-finger is critical for Irf8 repression, and the N-finger is important only in the context of the T354M-altered C-finger.

As T354M shares select GATA2 attributes, we tested whether common or distinct mechanisms underlie their transcriptional regulatory activities. GATA factor N- and C-terminal sequences contribute to transcriptional activation. We generated deletion mutants to compare sequences mediating GATA2 vs. T354M transcriptional regulatory activities. While T354MΔ410-480 resembled T354M in regulating Prg2, Cpa3, and Epx, T354MΔ61-251 lost the capacity to regulate these genes (Fig. 3G). T354MΔ61-251 activated T354M-specific genes Cldn15, Ctsg, and Ms4a3 (SI Appendix, Fig. S3B). Deletion of 61-251aa attenuated GATA2-mediated activation of Cpa3, Epx, and Hdc, but not Prg2 (SI Appendix, Fig. S3C). Thus, the T354M N terminus confers context-dependent transcriptional regulation.

Considering that T354M resides in the C-finger, as an alternative strategy to assess a C-finger requirement for T354M function, we mutated RNRK, which resides external to the C-finger (396 to 399aa). GATA1 and GATA3 structural analyses implicated RNRK binding to the DNA minor groove (60). RNRK deletion abrogated DNA binding by a GATA1 C-finger peptide in vitro, and inclusion of this motif enables the N-finger to bind DNA resembling the C-finger (61). RNRK is conserved among GATA factors, and a R396W/Q variant in this motif occurs in GATA2 deficiency syndrome (7). R396A/R398A variants abolished GATA2- and T354M-mediated activation of Prg2, Cpa3, Epx, Hdc, and Ms4a3, and the variants were defective in chromatin occupancy at GATA2-binding sites (SI Appendix, Fig. S3 D and E). Without T354M, R396A/R398A retained activity to repress Irf8 (Fig. 3F). These results demonstrate that despite T354M naked DNA binding deficits, T354M utilizes sequences that promote DNA binding to occupy chromatin and regulate a GATA2 target gene cohort and genes that are not normally GATA2-regulated. The chromatin sites occupied by GATA2 and T354M vs. GATA2-only did not exhibit an obvious difference in single, palindromic, or double WGATAR motifs (Dataset Table S2).

We tested whether T354M transcriptional-regulatory activity is concordant with differentiation-regulatory activity. T354M and C295A increased granulocytic and suppressed monocytic differentiation of −77^−/− Lin⁻ cells (SI Appendix, Fig. S4A). Combining C295A with T354M, but not with GATA2, abrogated granulocytic differentiation (SI Appendix, Fig. S4A). C370A was not competent to suppress monocytic differentiation or increase granulocytic differentiation. Although C295A was active, combining C295A with T354M attenuated activity (SI Appendix, Fig. S4 B and C). C295A and C370A were not competent to induce mast cell differentiation, highlighting the N- and C-finger importance (SI Appendix, Fig. S4D). The R396A/R398A variant, combined with T354M, lost T354M activity to reduce monocytic and increase eosinophil differentiation. Without T354M, R396A/R398A promoted monocytic differentiation and reduced eosinophil differentiation (SI Appendix, Fig. S4 E and F). C295A and C370A did not activate the GATA2-specific targets Gata1, Il1rl1, Tpsb2, Samd14, and Ms4a2 that have important roles in mast cell development and/or function (SI Appendix, Fig. S5 A and B).

TAL1 and LDB1 facilitate GATA2 function at Gata2 +9.5 and Samd14 +2.5 enhancers (62) and can colocalize with GATA2 on chromatin (22, 34, 63, 64). ChIP-Seq (GSM2231902) analysis revealed TAL1 and the LDB1 interactor LMO2 reside at Ctsg +9 and Ms4a3 +27.6 sites (SI Appendix, Fig. S6 A and B) that harbor WGATAR, and TAL1-binding E-boxes. ChIP analysis with anti-TAL1 and anti-LDB1 antibodies revealed TAL1 and LDB1 occupancy at Gata2 +9.5, Samd14 int1, Dapp1 int3, Elane +147, Ikzf2 +171, Epx +116, Ctsg +9, Cpa3 -1.9, Ms4a3 +27.6, and Cebpe +6 (Fig. 4 A and B) that harbored E-boxes. GATA2 expression increased TAL1 and LDB1 occupancy at Samd14 int1, Dapp1 int3, Cpa3 -1.9, and Ms4a3 +27.6 (Fig. 4 A and B). T354M expression increased TAL1 and LDB1 occupancy at Ikzf2 +171, Cpa3 -1.9 Ms4a3 +27.6, and Cebpe +6 (Fig. 4 A and B).

Fig. 4.

Genetic ablation of TAL1 reveals its context-dependent requirement for GATA2 and T354M function. (A) Quantitative ChIP analysis with anti-TAL1 antibody at GATA2-occupied sites in target genes (n = 4). (B) Quantitative ChIP analysis with anti-LDB1 antibody at GATA2-occupied sites in target genes (n = 4). (C) PCR-based genotyping assay for −77^−/−Tal1^−/− allele. (D) Top, Representative western blot analysis of TAL1. Bottom, RT-qPCR analysis of Tal1 mRNA level in −77^−/− and −77^−/−Tal1^−/− cells. (E) RT-qPCR analysis of GATA2 target gene mRNAs in cells expressing GATA2 or variants (n = 5). (F) Quantitative ChIP analysis at GATA2-occupied sites in target genes with anti-HA antibody (n = 3). (G) Model of collaborative transcriptional regulation by GATA2/T354M and TAL1. * < 0.05, ** < 0.01, and *** < 0.001.

To test whether TAL1 is essential, modulatory, or not required for GATA2 and T354M activities, CRISPR/Cas9 was used to generate Tal1^−/− hi-77^−/− cells (Fig. 4 C and D). The cells tolerated the nullizygous mutation, which attenuated GATA2- and T354M-mediated activation of Samd14, Ms4a2, Ms4a3, Ctsg, and Elane without affecting Cebpe and Hdc (Fig. 4E). TAL1 co-occupies chromatin and cooperates with GATA2 to regulate transcription at select loci yet it is dispensable for GATA2-mediated activation of Cebpe, despite occupancy at Cebpe +6 kb. GATA2 and T354M occupied chromatin in TAL1-deficient cells (Fig. 4 F and G). Thus, GATA2 and T354M chromatin occupancy can be TAL1-independent, and, in certain contexts, TAL1 maximizes the transcriptional response.

An Obligate Enhancer-Dependent Feedforward Mechanism Distorts a Multilineage Differentiation Program and Exemplifies a GATA Factor Paradigm.

T354M activates certain GATA2 target genes, e.g., Prg2, even though GATA2 and T354M occupancy was undetectable. In principle, this regulation might be indirect, involving the regulated expression of another transcription factor. We mined our quantitative proteomic data from fetal liver CMP/GMP population progenitors and RNA-seq data from primary fetal liver Lin⁻ progenitors (43) and hi-cells (19), which revealed that lowering GATA2 decreases Cebpe mRNA and protein expression (Fig. 5A). Cebpe encodes the transcription factor C/EBPε that promotes granulopoiesis (65–67). C295A or C370A mutations, combined with GATA2 or T354M, disrupted GATA2- and T354M-mediated regulation of Cebpe (Fig. 5B). GATA2 and T354M occupied chromatin near a Cebpe+6 kb enhancer (Fig. 5C and SI Appendix, Fig. S7A) that was reported previously (66). CRISPR/Cas9-mediated deletion of Cebpe+6 kb (Fig. 5D and SI Appendix, Fig. S7B) abrogated GATA2- and T354M-induced Cebpe mRNA and protein (Fig. 5 E and F) and Prg2, Prg3, Epx, and Ms4a3 expression in two hi-77^−/−Cebpe+6^−/− clonal cell lines (Fig. 5F and SI Appendix, Fig. S7C).

Fig. 5.

Genetic ablation of a Cebpe enhancer reveals the vital importance of C/EBPε for context-dependent GATA2 and T354M function. (A) Left, Comparison of protein levels [normalized using MaxLFQ from the prior analysis (43)] in CMP/GMP. Center, mRNA levels (TPM) in Lin⁻ progenitors calculated using RSEM (68). Right, mRNA levels (TPM) in hi-77^+/+ or hi-77^−/− cell lines. (B) RT-qPCR analysis of Cebpe mRNA in cells expressing GATA2 or variants (n = 5). (C) ChIP-Seq profile of transcription factors at Cebpe locus in hi-77^+/+ cells mined from dataset GEO:GSE224904. (D) PCR-genotyping assay for −77^−/−Cebpe+6^−/− allele. (E) Left, Representative western blot analysis of C/EBPε. Right: Quantification of C/EBPε protein in hi-77^−/− cells expressing HA-GATA2 or variants (n = 4). (F) RT-qPCR analysis of GATA2 target gene mRNAs in cells expressing GATA2 or variants (n = 4). (G) ChIP-Seq profile of transcription factors at Prg2/3, Epx, Ms4a2/3, Csf2rb, and Anxa1 in mouse myelocytes mined from dataset GEO:GSE73844. (H) RT-qPCR analysis of GATA2 target gene mRNAs in hi-77^+/+ cells and hi-77^+/+Cebpe+6^−/− cells (n = 6). (I) Representative western blot analysis of HA-C/EBPε by using anti-HA antibody. Right, RT-qPCR analysis of C/EBPε target gene mRNAs (n = 5). (J) Top, Flow cytometric analysis of GFP⁺ cells stained with CD11b and CD115. Bottom, Quantitation of flow cytometry data (n = 5). * < 0.05, ** < 0.01, and *** < 0.001.

Analysis of granulocyte ChIP-seq data (69) revealed that C/EBPε occupied chromatin near Prg2, Prg3, Epx, Ms4a3, and Cebpe. C/EBPε occupied chromatin near Prg2, Prg3, Epx, Ms4a3, and Cebpe. GATA2 and C/EBPε occupancy overlapped at Epx and Cebpe (Fig. 5 C and G). GATA2 and C/EBPε overlap was detected at the Csf2rb promoter and Anxa1 -19. Cebpe+6 enhancer ablation in hi-77^+/+ cells decreased Prg2, Prg3, and Ms4a3 expression (Fig. 5H and SI Appendix, Fig. S7D). C/EBPε expression elevated Prg2, Prg3, Ms4a3, and Epx expression in hi-77^−/− cells (Fig. 5I). Expressing C/EBPε normalized the enhanced monocytic differentiation of −77^−/− Lin⁻ cells (Fig. 5J). Although ablating Cebpe+6 kb decreased expression of Slc7a8, flanking Cebpe, C/EBPε did not activate Slc7a8.

RNA-Seq analysis in −77^−/− Lin⁻ cells and proteomic analysis comparing −77^+/+ CMP/GMP and −77^−/− CMP/GMP (43) revealed that Slc7a8 is a GATA2 target gene (SI Appendix, Fig. S7 E and F). Thus, GATA2 regulates Slc7a8 through the Cebpe+6 kb enhancer. RNA-Seq analysis revealed 96 neighboring pairs (Dataset Table S3) of GATA2 and T354M targets like Cebpe and Slc7a8. As exemplified by Cebpe+6, T354M shares certain enhancers with GATA2 to regulate transcription (SI Appendix, Fig. S7G). GATA2 and T354M function through Cebpe+6 kb to control expression of C/EBPε and other target genes constitutes a type I coherent feedforward mechanism to establish the myeloerythroid progenitor transcriptome and regulate cell state transitions (Fig. 6).

Fig. 6.

Pathogenic GATA2 genetic variants utilize an obligate enhancer-dependent feedforward mechanism to distort a multilineage differentiation program. GATA2 and T354M utilize C/EBPε and TAL1 to regulate a target gene cohort and multilineage differentiation. The fragmented activity of T354M generates a lineage-distorted differentiation program involving increased eosinophil, but not erythroid or mast cell, differentiation. As other GATA2 pathogenic variants share this attribute, this may represent a hallmark of GATA2 pathogenic variants. Considering the involvement of C/EBPε as a suppressor of MLL-rearrangement AML (70), and GATA2 as a suppressor of MDS/AML, the GATA2-C/EBPε paradigm links two vital systems that ensure the normal development and function of the hematopoietic system.

Discussion

Unlike simplistic scenarios in which pathogenic genetic variation establishes gain-of-function or loss-of-function consequences, nonbinary outcomes with partial gain or loss of function and retention of select physiological activities can be confounding to interpret vis-à-vis pathogenic mechanisms. Although GATA2 pathogenic coding variant DNA binding activity can be compromised, we demonstrated that T354M and additional variants retain the capacity to occupy select GATA motif- and E-box-containing chromatin sites and to regulate target genes. In certain contexts, therefore, T354M and other disease variants share GATA2 activities. However, the context-dependent similarities are accompanied by overt differences. This mechanistic deviation disrupts vital GATA2 functions that ensure the generation and function of HSPCs and differentiated hematopoietic cell progeny. Partial retention of activity was described (18), but the mechanisms were not elucidated.

Like canonical transcriptional mechanisms, GATA2 colocalizes on chromatin with a host of transcription factors, including TAL1 and ETS factors (31). However, the factors and coregulators essential for GATA2 function are largely undefined. Tal1 ablation in hi-77^−/− cells attenuated GATA2 activity at a subset of target genes. GATA1 utilizes the obligate coregulator FOG1 at many, but not all, of its target genes in erythroid cells (47, 71). The factor and coregulator requirements for GATA1 function differ at distinct loci (25, 32, 71). TAL1 is expressed in erythroid cells, and while many correlations exist, it has not been demonstrated to be essential for GATA1 function. C/EBPε was required for GATA2-mediated activation of Prg2, Prg3, and Epx, and T354M utilized C/EBPε to increase Prg2, Prg3, Epx, and Ms4a3 expression.

C/EBPε is up-regulated upon granulocytic differentiation, and targeted ablation of Cebpe disrupts granulopoiesis and impairs neutrophil and eosinophil function (65, 67). The homozygous mutant mice survive in a sterile environment but bacterial infection yields a lethal MDS-like disease. Cebpe expression is induced by G-CSF signaling (72), the related factor C/EBPα (73), and the Cebpe+6 enhancer (66). Like GATA2 (9, 10), human CEBPE variants are associated with increased leukemia risk, although unlike the GATA2-linked pediatric and adult MDS/AML predisposition, CEBPE variants are associated with increased risk of childhood B-cell acute lymphocytic leukemia (74–76). CEBPE mutation also causes an immunodeficiency termed neutrophil-specific granule deficiency (SGD) (77). While it is unclear whether SGD is linked to MDS/AML, patients with mutations in SMARCD2, a collaborator of C/EBPε, developed MDS (78).

Both GATA2 and T354M occupied Cebpe+6, and the enhancer ablation in hi-77^−/− cells reduced Cebpe expression and abrogated GATA2- and T354M-mediated regulation of Cebpe, Prg2, Prg3, Epx, and Ms4a3. GATA2 and C/EBPε occupied chromatin at these genes. Besides compromising Cebpe transcription, Cebpe+6 ablation prevented induction of the neighboring gene encoding the solute transporter SLC7A8. In contrast to Cebpe, which is regulated by T354M to a greater extent than by GATA2, Slc7a8 was regulated by GATA2 to a greater extent. Thus, GATA2 and T354M share an enhancer to activate neighboring genes, yet the transcriptional outputs differ quantitatively. C/EBPε was also required for GATA2 function in wild-type (hi-77^+/+) cells. Although the determinants for CEBPE enhancer activation were unclear, MEF2D binds and represses the enhancer, reducing CEBPE expression (70). MEF2D and CEBPE expression are anticorrelative in AML and B-ALL, MEF2D is required for MLL-rearranged AML, and MEF2D activity to repress CEBPE expression may underlie this activity (70). Considering the MEF2D-CEBPε relationship, it is attractive to propose that MEF2D opposes GATA2-dependent CEBPE enhancer activation and GATA2 may suppress the MEF2D leukemogenic mechanism. The GATA2-C/EBPε axis conforms to a coherent feedforward loop, representing a paradigm for GATA factor function, and establishes an important component of the progenitor transcriptome.

Since T354M exerted transcriptional-regulatory activity in certain contexts, we compared the molecular determinants of GATA2 and T354M function. T354M, but not GATA2, often required the N-finger. While C295A activated Cpa3 and Hdc, C295A/T354M was inactive. Similarly, although C295A occupied multiple chromatin sites, C295A/T354M occupancy was only detected at Gata2 +9.5 and Bcl2l1 intron. The same consequences were evident with differentiation. While T354M and C295A suppressed monocytic and eosinophil differentiation, C295A/T354M severely compromised these activities. Thus, the N-finger is critical for T354M activity. Both N- and C-fingers can engage in protein–protein interactions. GATA factor N-fingers bind FOG proteins (47), and the C-finger can bind multiple proteins including PU.1 (79) and IKZF2 (80). Analogous to the GATA1 N-finger stabilizing C-finger-mediated DNA binding at palindromic GATA motifs (58), the N-finger might stabilize C-finger-dependent protein interactions under conditions in which T354M compromises C-finger structure. Alternatively, the N-finger might stabilize DNA binding of the T354M-compromised C-finger. The GATA1 N-finger does not contribute to GATA motif binding, but its deletion destabilizes interactions between GATA1 and DNA and decreases transactivation. The R216Q variant in the GATA1 N-finger from patients with X-linked thrombocytopenia with thalassemia and gray platelet syndrome does not stably bind palindromic GATA motifs (81).

Consistent with the aberrant T354M-instigated transcriptional program, T354M did not promote multilineage differentiation. Similarly, low GATA2 in −77^−/− Lin^- cells was insufficient to support mast cell, eosinophil, and erythroid differentiation. Although eosinophil loss with reduced GATA2 levels has not been reported, ectopically elevated GATA2 increases eosinophils (44). GATA2 and T354M rescued eosinophil gene expression, and reducing Cebpe expression via enhancer ablation disrupted GATA2- and T354M-mediated activation of Prg2, Prg3, and Epx expression. C/EBPε regulates genes (e.g., Prg2, Prg3, and Epx) encoding eosinophil secondary granule proteins (82). Expressing C/EBPε in −77^−/− Lin⁻ cells suppressed the excessive monocytic differentiation (Fig. 5J). Although peripheral blood and bone marrow analyses reveal monocytopenia as a human GATA2 deficiency syndrome attribute, macrophages are abundant in bone marrow from GATA2 deficiency syndrome patients with R398W or intron 5 (equivalent to murine +9.5 enhancer) variants (43). Mice lacking Epx and Prg2 produce fewer eosinophils (83). Patients with GATA2 pathogenic variants can exhibit eosinophilia (84), and pathogens detected in these patients can promote eosinophilia (85). Regarding the inability of GATA2 variants to promote mast cell differentiation, CD34⁺ cells harboring GATA2 heterozygous mutation have reduced activity to differentiate into mast cells, and mast cells from patients with GATA2 mutation exhibit a decreased response to IgE-FcεR1-mediated activation (86).

The fetal origin of the progenitors may constitute a limitation of the rescue system. As MDS/AML occurs in children and adults, the insights developed from fetal progenitor analyses may or may not be applicable to MDS/AML. However, Nras^G12D/+; p53^R172H/+ mice develop AML that is associated with reduced GATA2 levels and differential gene expression that overlaps significantly with GATA2-regulated genes in the rescue system (87). This result provides evidence that the rescue system yields insights relevant to mouse AML, and considering the 98% sequence identity between mouse and human GATA2, we propose that these findings can be extrapolated to human systems.

In summary, GATA2 utilizes C/EBPε in a coherent feedforward mechanism to establish an essential component of the myeloerythroid progenitor cell transcriptome. While pathogenic variants analyzed bind chromatin, regulate transcription at select loci, and utilize this mechanism, they institute a fragmented differentiation program. A mechanistically unique attribute of the T354M program is the disproportionate reliance on the N-finger, which may reflect a strategy to surmount functional limitations arising from the disease variant-compromised C-finger. There may be collateral consequences of this mechanism, e.g., perturbation of normal N-finger functions. Other GATA2 pathogenic variants shared T354M attributes, indicating a broader applicability of the mechanisms. As certain variants, such as T355Δ, had weaker activity, it will be instructive to evaluate potential genotype–phenotype relationships in GATA2 deficiency patients. Disease-linked zinc-finger variants have been described for other GATA factors, including GATA1 in congenital erythropoietic porphyria (88) and X-linked gray-platelet syndrome (89), GATA3 in breast cancer (90) and acute T cell leukemia (91), and GATA4 in cardiovascular disease (92). The insights described herein provide a foundation to decipher relationships between GATA factor structure and mechanisms that establish cell type-specific transcriptomes, which drive differentiation programs and suppress pathogenesis.

Methods

Cell Culture.

Murine fetal liver hematopoietic precursor cells were cultured in IMDM supplemented with 20% FBS, 1% penicillin/streptomycin, 4% IL3 conditioned medium, and 4% SCF conditioned medium. Cells were cultured in a humidified incubator at 37 °C and 5% carbon dioxide. Immortalized murine fetal liver-derived hematopoietic progenitor cells (hi-77^−/−) were cultured in OPTI-MEM media containing 10% FBS, 1% SCF-CM, 1% penicillin/streptomycin, 1% L-glutamine, and 30 mM β-mercaptoethanol. After infection with retroviral vector, hi-77^−/− cells were selected by puromycin to enrich for cells expressing GATA2 or disease variants. All the methods used in this study including qRT-PCR, RNA-Seq, flow cytometry, protein analysis, ChIP, and gene editing are described in SI Appendix, supporting methods. Materials used in this study are listed in the table of key resources.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (PDF)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. The datasets used in this study are available in GEO database under accession nos: GSE133606 (93), GSE77029 (94), GSE171384 (95), GSE84328 (96), GSE199464 (97), GSE224904 (98) and GSE73844 (99).