GATA factor-regulated solute carrier ensemble reveals a nucleoside transporter-dependent differentiation mechanism

Developmental-regulatory networks often include large gene families encoding mechanistically-related proteins like G-protein-coupled receptors, zinc finger transcription factors and solute carrier (SLC) transporters. In principle, a common mechanism may confer expression of multiple members integral to a developmental process, or diverse mechanisms may be deployed. Using genetic complementation and enhancer-mutant systems, we analyzed the 456 member SLC family that establishes the small molecule constitution of cells. This analysis identified SLC gene cohorts regulated by GATA1 and/or GATA2 during erythroid differentiation. As >50 SLC genes shared GATA factor regulation, a common mechanism established multiple members of this family. These genes included Slc29a1 encoding an equilibrative nucleoside transporter (Slc29a1/ENT1) that utilizes adenosine as a preferred substrate. Slc29a1 promoted erythroblast survival and differentiation ex vivo. Targeted ablation of murine Slc29a1 in erythroblasts attenuated erythropoiesis and erythrocyte regeneration in response to acute anemia. Our results reveal a GATA factor-regulated SLC ensemble, with a nucleoside transporter component that promotes erythropoiesis and prevents anemia, and establish a mechanistic link between GATA factor and adenosine mechanisms. We propose that integration of the GATA factor-adenosine circuit with other components of the GATA factor-regulated SLC ensemble establishes the small molecule repertoire required for progenitor cells to efficiently generate erythrocytes.


Introduction
As a process that broadly informs stem cell biology, hematopoietic stem cells produce diverse progenitor cells that differentiate into blood cells, ensuring physiological homeostasis and the capacity to respond to stress [1][2][3]. Lineage-committed progenitor cells undergo drastic molecular and cellular transitions to generate blood cell types with overtly different phenotypes and functions. For example, erythroid progenitor cells differentiate into precursor cells that progressively mature into enucleated reticulocytes and erythrocytes [4]. In pathological states, such as anemia resulting from acute blood loss, a "stress erythropoiesis" mechanism is deployed to accelerate erythrocyte regeneration and oxygen delivery, thereby protecting cells and tissues [5,6].
In addition to informing stem cell biology, hematopoiesis represents a powerful system for addressing fundamental problems in molecular biology and genetics, including how complex genetic, protein and small molecule networks control cellular differentiation. The GATA transcription factors GATA2 and GATA1 instigate genetic networks in hematopoietic stem and progenitor cells (HSPCs), erythroid precursor cells and erythroblast progeny [7]. GATA2 is expressed in erythroid precursor cells, and as GATA1 increases, it acquires the capacity to repress Gata2 transcription [8]. This GATA switch often decreases or increases GATA factor target gene transcription and impacts hundreds to thousands of proteins in the erythroblast proteome [9][10][11][12]. The target genes include members of large gene families, e.g. G-proteincoupled receptors, zinc finger transcription factors and solute carrier (SLC) transporters.
In a mechanism in which multiple members of a large gene family are integral to a developmental process, it is instructive to compare and contrast the regulation and function of the family members. Contrasting with the six-member mammalian GATA factor family [1,7,[13][14][15][16], a large gene family may contain hundreds of members. In principle, a common mechanism may establish expression of multiple family members, or an array of mechanisms may be deployed. To address this problem in the context of networks governing erythropoiesis, we analyzed the >450 SLC transporters that dictate the small molecule repertoire of cells [17]. Several SLCs are implicated in erythroid biology [18][19][20][21][22][23][24] including the GATA1-induced gene Slc4a1 [25] encoding an anion transporter in erythroblasts and erythrocytes [26]. GATA1 instigates a zinc transporter switch, involving the importer Slc39a8 and exporter Slc30a1, that controls intracellular zinc and erythroid differentiation [12]. We reasoned that an SLC transporter cohort establishes/maintains the erythroblast small molecule repertoire as a vital element in GATA factor-dependent networks. Identifying essential SLCs will enable building of integrative models to explain how extracellular stimuli utilize small molecules to orchestrate cellular functions.
Given the large number of SLC transporters not studied in hematopoiesis, we evaluated GATA1-and GATA2-regulated genes in genetic rescue and enhancer-mutant systems, respectively, to identify GATA factor-regulated SLCs. GATA1 and GATA2 co-regulated >50 SLCs including amino acid, metal and nucleoside/nucleotide transporters. Embedded within this cohort are eight SLCs, including the equilibrative nucleoside transporter 1 (Slc29a1/ENT1), that share GATA factor occupancy at predicted intronic enhancer regions containing predicted cis-regulatory elements, suggesting direct transcriptional regulation. Loss-of-function studies indicated that an Slc29a1-dependent mechanism promotes erythroblast survival and differentiation and attenuates anemia in an acute anemia mouse model. As Slc29a1 transports adenosine, and GATA factors induced adenosine kinase, which converts adenosine to AMP, GATA factor-dependent networks contain an adenosine circuit that is expected to control vital biological processes.

A GATA1-and GATA2-regulated solute carrier gene ensemble
Since SLCs regulate a wide spectrum of intracellular small molecules, and very little is known about how GATA factors impact small molecule-dependent mechanisms, we conducted an analysis to identify all GATA1-and GATA2-regulated SLC genes in erythroblasts. We used our RNA-seq dataset [27] from the GATA1-null G1E-ER-GATA1 cell genetic rescue system [28] to identify GATA1-regulated SLC genes ( Fig 1A). Treatment of cells with β-estradiol (48 hours) activates a conditional GATA1 allele (ER-GATA1) stably expressed in proerythroblastlike G1E cells, thus inducing or repressing GATA1 target genes and stimulating erythroid differentiation [25]. This analysis revealed 165 GATA1-regulated SLC genes (�1.5 fold) (Fig 1B).
Using a prioritization strategy involving the magnitude of GATA factor regulation and reported hematopoietic-relevant human or mouse phenotypes, we delimited an SLC cohort for further study. One of the top 10 GATA1-repressed SLC genes within this cohort was Slc29a1, which encodes the nucleoside transporter Slc29a1/ENT1 that mediates uptake of adenosine and certain chemotherapeutic drugs [36]. The -77 +/+ vs. -77 -/erythroid precursor RNA-seq data [31] revealed GATA2 activation of Slc29a1 expression 5.4 fold (Fig 3I). By qRT-PCR, GATA1 decreased and -77 increased Slc29a1 expression 26 and 1.9 fold, respectively ( Fig 3J). Both GATA1 (GEO GSE32491) and GATA2 (GEO GSE18829) occupied the human SLC29A1 locus at a site harboring a consensus GATA motif. ChIP-seq with primary murine Ter119 + erythroblasts (GEO GSE30142) or undifferentiated G1E proerythroblasts (GEO GSE29196) revealed endogenous GATA1 and GATA2 occupancy, respectively, within the Slc29a1 first intron harboring two consensus GATA motifs ( Fig 3K). In aggregate, these results demonstrate that GATA2 and GATA1 activate and repress Slc29a1, respectively ( Fig  3L), and regulate an SLC ensemble enriched in amino acid, metal and nucleoside/nucleotide transporters.
The GATA2 occupancy peaks of the eight SLC genes contained multiple WGATAR and E-Box motifs ( Fig 4B). We asked if these motifs are enriched in these intronic sequences in comparison to randomly selected intronic sequences. The WGATAR and E-box motifs were enriched in the SLC intronic sequences relative to random intronic regions with the same length and a similar chromosomal location (p = 0.0054 and p = 0.0169, respectively). We asked if the regulatory attributes of this cohort are conserved in human. Four out of eight genes contain DNaseI hypersensitive sites (HSs) in human myeloid/erythroid cells [42] that overlap with GATA2 occupancy in K562 cells (GEO GSE18829) (S4 Fig). These sites are located in the same intron as the predicted murine enhancers, suggesting that the GATA factor regulatory mechanism may be conserved.

PLOS GENETICS
downregulation, erythroid precursors were expanded for two days, GFP-positive cells expressing shRNAs were isolated by flow cytometry, and mRNA was quantified (Fig 5A). The shRNAs efficiently reduced Slc29a1 mRNA without affecting mRNA for Slc29a3 encoding the equilibrative nucleoside transporter ENT3 (Fig 5B). Under culture conditions that favor expansion

PLOS GENETICS
over differentiation, Slc29a1 downregulation reduced cell numbers in two of the three Slc29a1targeting shRNAs relative to luciferase control shRNA-infected cells (37 ± 3.5% decrease, p = 0.037). After expanding cells for two days, erythroid precursors were transferred to erythropoietin-containing differentiation media for three days. Under differentiation conditions, Slc29a1 downregulation decreased cell numbers relative to luciferase control shRNA-infected cells (56 ± 7.6% decrease, p = 0.0025) (Fig 5C). To investigate the mechanisms, erythroid precursors were expanded for two days and differentiated for three days. Using flow cytometry with apoptotic markers (Annexin V and DRAQ7), we quantified apoptosis in Slc29a1 shRNA and control-infected primary erythroblasts (Fig 5D). Slc29a1 downregulation increased apoptosis relative to cells expressing luciferase shRNA (61 ± 2.1% increase, p < 0.0001) (Fig 5E), indicating that Slc29a1 promotes erythroblast survival during erythroid differentiation.

Genetic ablation of Slc29a1 attenuates steady-state erythropoiesis and erythrocyte regeneration in response to acute anemia
Based on the Slc29a1 activity to promote survival and differentiation of primary fetal liver progenitor cells ex vivo, we evaluated whether these findings can be extrapolated in vivo. We bred floxed-Slc29a1 mice with EpoR-Cre mice to specifically ablate Slc29a1 in erythroid cells (e-Slc29a1 -/-) (Fig 7A). qRT-PCR analyses revealed that Slc29a1 expression was specifically ablated in CD71 + erythroblasts isolated from bone marrow of e-Slc29a1 -/mice (Fig 7B). The e-Slc29a1 -/mice were fertile and had no obvious abnormalities. We compared hematological parameters by analyzing complete blood counts (CBC) of adult e-Slc29a1 -/and Slc29a1 f/f mice. e-Slc29a1 -/mice exhibited normal total Hb and HCT, slightly lower red blood cell (RBC) number (p < 0.05), and mildly increased MCV and MCH (p < 0.001 and p < 0.01, respectively), yet within the normal range relative to Slc29a1 f/f mice (Table 1).
In acute anemia, erythrocytes decline rapidly, and stress erythropoiesis is deployed to regenerate erythrocytes to attenuate the anemia. To determine the importance of erythroid Slc29a1 in acute anemia, we utilized a phenylhydrazine (PHZ)-induced hemolytic anemia model with elevated stress erythropoiesis. 12-week old e-Slc29a1 -/mice and Slc29a1 f/f mice were injected consecutively with PHZ, and CBC analysis confirmed the PHZ-induced anemia in Slc29a1 f/f mice on day six. RBCs and total Hb decreased significantly (p < 0.001 and p < 0.01, respectively), while MCV and MCH increased (p < 0.001) ( Table 1). However, PHZtreated e-Slc29a1 -/mice developed more severe anemia relative to PHZ-treated Slc29a1 f/f mice, exhibiting decreased RBCs (p < 0.05), total Hb (p < 0.001) and HCT (p < 0.05), and increased MCV and MCH (p < 0.05) ( Table 1). This analysis provided in vivo genetic evidence that erythroid Slc29a1 protects against PHZ-induced acute anemia.
In mice, stress erythropoiesis occurs predominantly in the spleen [5,43]. To test whether Slc29a1 counteracts PHZ-induced acute anemia in the spleen, we conducted flow cytometry to assess erythroid differentiation in e-Slc29a1 -/and Slc29a1 f/f mouse spleen with or without PHZ treatment. Without PHZ, erythroid differentiation in e-Slc29a1 -/and Slc29a1 f/f mouse spleen did not differ significantly (Fig 8A-8B). Similar to BM, following PHZ treatment, erythroid differentiation was greater in Slc29a1 f/f mice relative to e-Slc29a1 -/mice, which exhibited an attenuated transition from Baso-E (p < 0.01) to Ortho-E (p < 0.01) (Fig 8A-8B). In aggregate, GATA factor-regulated Slc29a1 promoted cellular survival and erythroid differentiation ex vivo, and Slc29a1 promoted steady-state erythropoiesis and erythrocyte regeneration in bone marrow and spleen in an acute anemia model in vivo. These results position Slc29a1, along with a host of other GATA factor-regulated SLCs, as an important component of the GATA factor-dependent genetic network that governs erythrocyte development.
Slc29a1 is the principle adenosine transporter in mammals. Another critical mechanism to control adenosine homeostasis involves the gene Adk, encoding adenosine kinase. Adenosine kinase phosphorylates adenosine to generate AMP, which can function as a signaling molecule to activate AMP-activated protein kinase (AMPK) [44] or act as a precursor molecule for important cellular processes. While targeted ablation of Adk causes neonatal hepatic steatosis [45] and vascular inflammation [46], Adk has not been studied in hematopoiesis.

PLOS GENETICS
Since Adk catalyzes adenosine phosphorylation to yield AMP, and erythroid Slc29a1 loss decreases AMP levels and AMPK phosphorylation in mature erythrocytes [48], we asked if Slc29a1 downregulation decreases intracellular AMP in primary erythroblasts cultured ex vivo under conditions that promote erythroid cell expansion and differentiation. We isolated HSPCs from murine E14.5 fetal livers, infected them with two different shRNAs targeting Slc29a1 or luciferase, and quantified AMP and ATP levels after sorting GFP-positive erythroblasts. This analysis revealed a 46 ± 7.6% (p = 0.033) decrease in AMP from cells expressing For the box-and-whisker plots, the box depicts the 25th to 75th percentiles of data, the median line, and whiskers ranging from minimum to maximum values. Statistical analysis was carried out using an unpaired 2-tailed Student's t test. � p < 0.05, �� p < 0.01, ��� p < 0.001, ���� p < 0.0001.
In summary, GATA2 directly increased Slc29a1 expression, and as GATA1 replaces GATA2 during erythroid differentiation, GATA1 repressed Slc29a1. GATA1 upregulated adenosine kinase, which controls metabolic processes, cellular signaling and adenosine homeostasis. Although it seems counterintuitive that GATA1 reduces Slc29a1 and elevates Adk, it is likely that multiple positive and negative regulatory components collectively control adenosine homeostasis. However, our discoveries elucidating this adenosine circuit embedded within GATA factor-instigated regulatory networks is proposed to be a critical link between GATA factor and adenosine mechanisms (Fig 9G).

Discussion
Regulatory networks governing cell and developmental processes often include multiple members of large gene families. In a given biological context, numerous family members may share a common mechanism governing their expression and/or function, or distinct mechanisms may be committed to individual family members. Herein, we addressed this problem with the 456 member SLC transporter cohort. We demonstrated that GATA1 and GATA2 co-regulate >50 SLC genes, encoding amino acid, metal, nucleoside/nucleotide and additional SLC transporters (Fig 10). Our prioritization strategy identified a cohort of eight SLCs, including Slc29a1, that contain GATA1-and GATA2-occupied predicted intronic enhancers. Loss-offunction studies with Slc29a1, an equilibrative nucleoside transporter, indicated that Slc29a1 promotes erythroblast survival and differentiation in physiological and stress states by controlling intracellular adenosine homeostasis. By promoting erythroblast differentiation in spleen and BM in response to hemolytic anemia, erythroblast Slc29a1 mitigated anemia.
How does adenosine flux, sensing and consumption impact the generation of billions of erythrocytes daily? Adenosine is phosphorylated by adenosine kinase to yield AMP, which functions as a signaling molecule or precursor for ATP synthesis [46]. In an opposing pathway, adenosine deaminase-catalyzed conversion of adenosine to inosine is important for de novo nucleotide biosynthesis to generate nucleic acids [49,50]. AMP is a physiological activator of AMP-activated protein kinase [43], which inhibits mTOR signaling [51] and induces autophagy [52] that mediates intracellular remodeling required for erythroid differentiation [53,54]. In addition, mTORC1 signaling controls mitochondrial biogenesis as a critical step in erythropoiesis [55]. GATA2 and GATA1 regulated Slc29a1 and adenosine kinase expression, thus constituting a GATA factor-adenosine circuit predicted to elevate AMP levels and stimulate AMP-dependent processes that impact erythrocyte development and regeneration. Nucleoside/nucleotide transporter involvement in GATA factor-instigated regulatory networks has not been described. Intracellular adenosine accumulation, e.g. in adenosine kinase deficiency, results in elevated s-adenosylhomocysteine, which inhibits s-adenosylmethionine (SAM)dependent transmethylation reactions [32]. Since SAM is a requisite cofactor for DNA, RNA and protein methylation reactions that underlie epigenetic control of gene expression [56,57], mechanisms governing adenosine homeostasis impact genome function.
The Slc29/ENT family consists of four members, with three mediating nucleoside transport and controlling a wide swath of cellular processes [58]. Slc29a1 transports adenosine with high affinity [36]. Previously, we demonstrated that targeted ablation of Slc29a1 in erythroid cells elevates plasma adenosine to oppose acute hypoxia-induced tissue damage [59]. Furthermore, Slc29a1 downregulation at high altitude generates cytoprotective extracellular adenosine that mitigates hypoxia-mediated cell and tissue injury [59]. If sufficient erythrocyte regeneration does not occur, this injury can be lethal [5,6]. Slc29a1 is also an important determinant of drug transport (e.g., ribavirin) into erythrocytes [60], suppresses mineralization of spinal tissues [61] and generates the "Augustine" blood group antigen [62]. However, its role in hematopoiesis and GATA factor mechanisms was not established. Slc29a3 (ENT3) also transports adenosine, but differs from Slc29a1 in its localization to intracellular compartments. Slc29a3 -/mice exhibit HSC defects involving dysregulated autophagy and 5' AMP-Activated Protein Kinase activity [63]. Slc29a2/ENT2, which has high affinity for inosine [64], was GATA1-repressed.
As certain SLCs control nutrient availability and metabolic activity, mechanisms that unify or segregate actions of distinct SLC transporters and their respective small molecules are of considerable interest. These mechanisms may establish crucial links between extracellular stimuli, metabolic state and cellular transitions, including differentiation. Of the 31 known iron, zinc, and copper transporters, 22 were GATA factor-regulated, representing the majority of SLC metal transporters. These essential trace metals are important structural and functional co-factors in erythroid biology, with the best studied being iron-containing heme [65]. The initial and rate-limiting step of heme synthesis in the mitochondrial matrix requires succinyl-CoA, derived from the tricarboxylic acid cycle, and glycine, which can be acquired through SLC transport mechanisms [24,66] or biosynthetic pathways.
In summary, we demonstrated that hematopoietic-regulatory GATA factors control a >50 SLC gene ensemble, and one member, Slc29a1, promotes erythroblast survival and erythroid differentiation. Transporter co-regulation may be a primary determinant, or one of multiple parameters, of an integrated network that dynamically controls the metabolome and metallome. It will be instructive to consider the mechanistic interconnections or independence of the transporters and their small molecule passengers in cellular survival, proliferation and differentiation, both in the context of hematopoiesis and broad biological contexts.

shRNA construction and retroviral infection
Three distinct miR-30 shRNAs targeting Slc29a1 or luciferase as a control were cloned into MSCV-PIG vectors (IRES-GFP) using Xho I and Bgl II restriction sites [12,68]. An empty MSCV-PIG vector was used as an additional negative control. MSCV-PIG vectors and pCL-Eco packaging vector (15 μg of each) were transfected into 293T cells to produce retrovirus targeting Slc29a1 or luciferase. For shRNA-mediated loss-of-function experiments, 3 x 10 5 HSPCs were spinoculated with 100 μl retrovirus, 8 μg/ml polybrene and 10 mM HEPES buffer at 1600 x g for 90 min at 30˚C. After expanding cells for two days, live GFP + cells were sorted from the cell population using a BD FACSAria cell sorter (BD Biosciences), and gene expression was quantified by qRT-PCR.

Quantitative real-time PCR analysis
Total RNA was purified with TRIzol (Life Technologies). RNA (1 μg) was DNase treated for 15 min and then heated at 65˚C for 10 min with 25 mM EDTA. For cDNA synthesis, DNase Itreated RNA was incubated with 125 ng of a 5:1 mixture of oligo-dT primers and random hexamers at 68˚C for 10 min. The RNA and primer mixture was incubated with Moloney MLV reverse transcriptase (Life Technologies), 10 mM DTT, RNAsin (Promega), and 0.5 mM deoxynucleoside triphosphates at 42˚C for 1 hour followed by heat inactivation at 98˚C for 5 min. qRT-PCR was conducted with Power SYBER Green Master Mix (Applied Biosystems) and a ViiA 7 Real-Time PCR system (Applied Biosystems).

Flow cytometry
For detection of erythroid markers, 1 x 10 6 cells were isolated by centrifugation at 300 x g for 5 minutes at 4˚C and washed once in PBS. Cells were incubated in 100 μl PBS with 1:100 antimouse Ter119-APC (Biolegend) and 1:100 anti-mouse CD71-PE (Biolegend) at 4˚C for 30 min in the dark. To quantify apoptosis following CD71/Ter119 staining, cells were washed once in 1X Annexin V Buffer (10mM HEPES (Sigma), 140 mM NaCl, 2.5 mM CaCl 2 ) and stained with 1:40 Annexin V-Pacific Blue (Thermo Fisher) and 1:100 membrane-impermeable dye DRAQ7 (Abcam) at room temperature for 20 minutes in the dark. Cells were isolated by centrifugation for 5 minutes, 300 x g at 4˚C and resuspended in 300 μl PBS. Flow cytometry was conducted with a BD LSRII flow cytometer (BD Biosciences), and data was analyzed with FlowJo software (FlowJo).
Bone marrow and spleen cells were analyzed to quantify erythroid differentiation. Antibodies (Biolegend) were used at a concentration of 1:100 unless indicated. Bone marrow and spleen cells were stained with Pacific Blue-conjugated Ter119 and PE/CY7-conjugated CD71 for one hour or overnight on ice. Cells were centrifuged, resuspended and analyzed with a BD LSRII flow cytometer (BD Biosciences).

AMP and ATP quantification
Equal numbers of cells were isolated by FACS, followed by centrifugation at 1300 x g for 5 min at 4˚C. Intracellular ATP concentration was quantified with Cell-Titer Glo Luminescent Cell Viability Assay (Promega) following manufacturers' instructions. Intracellular AMP was quantified with AMP-Glo Assay (Promega) following manufacturers' instructions. ATP and AMP analysis was conducted with a PheraStar Microplate Reader (BMG Labtech).

Mice and phenylhydrazine (PHZ) studies
Animal protocols were reviewed and approved by the Institutional Animal Welfare Committee of UT Health Science Center at Houston. Slc29a1 flox/flox (Slc29a1 f/f ) were obtained from Dr. Holger Eltzschig (UT Health Science Center at Houston). EpoR-Cre + mice were obtained from Dr. Stuart Orkin (Harvard Medical School). Slc29a1 f/f EpoR-Cre + mice (e-Scl29a1 -/-) were generated by crossing Slc29a1 f/f with EpoR-Cre + mice and genotyped as Slc29a1 f/f +/+ EpoR-Cre + [57]. Twelve-week-old (sex-matched, male and female) Slc29a1 f/f and e-Slc29a1 -/mice were used for experiments. Slc29a1 f/f and e-Slc29a1 -/mice were treated with vehicle or PHZ (Sigma Aldrich, P26252) at 50 mg/kg on day 0 and day 1 by intraperitoneal injection. At day 7, mice were sacrificed, and blood was collected for complete blood count (CBC) analysis.

Statistical analysis
Statistical analysis for ex vivo studies was conducted with the 2-tailed Student's t test to compare experimental and control samples. Asterisks denote statistical significance relative to controls. For in vivo studies, all data are expressed as the mean ± SD. Differences between the means of multiple groups were compared by two-way ANOVA, followed by a Tukey's multiple comparisons test. P < 0.05 was considered significant. Statistical significance was analyzed with GraphPad Prism 8 software (GraphPad).
Motif enrichment analysis for WGATAR and/or E-Box motifs was conducted by generating eight comparable sequences to those observed from the intronic enhancer regions for the eight SLC genes regulated by both GATA2-and GATA1. 10,000 groups of eight sequences were randomly generated, and the proportion of times out of 10,000 in which the observed intronic enhancer sequences contained a WGATAR and/or E-Box motif was used to calculate statistical significance (p-value). This enrichment analysis was carried out with R software.  [31] and control vs. β-estradiol-treated (48 h) G1E-ER-GATA1 cells (right) [27]. The scatter plots represent means ± SEM. TPM; Transcripts per million (n = 3). (B) Quantitative proteomics data [12] illustrating GATA1 upregulation of adenosine kinase isoforms (± 48 hours of β-estradiol treatment of G1E-ER-GATA1). (TIF) S1