Cooperative signaling between cytokine receptors and the glucocorticoid receptor in the expansion of erythroid progenitors: molecular analysis by expression profiling

Hematopoiesis requires tight control of survival, proliferation, and differentiation of progenitors to generate and maintain correct numbers of mature cells in the periphery. This process is dependent on cooperation between lineage-specific and more general growth factors and hormones (for review, see Lapidot and Petit¹ ). Steady-state erythropoiesis involves a limited number of “differentiation divisions”^2,3  and is controlled by erythropoietin (Epo), while stress erythropoiesis, induced by hypoxia, is coregulated by Epo, stem cell factor (SCF), and glucocorticoid receptor (GR) ligands, for example, hydrocortisone or dexamethasone (Dex).^4,5  Epo signaling is required for survival and terminal differentiation of erythroid progenitors, since mice lacking Epo or the Epo receptor (EpoR) die at midgestation with a lack of definitive erythrocytes.^6,7  In the absence of SCF or its receptor c-Kit, erythrocytes are generated, but the mice die at birth with severe anemia, and their bone marrow contains reduced numbers of erythroid progenitors (erythroid burst-forming units [BFU-Es] and erythroid colony-forming units [CFU-Es]).^8-11  Mice carrying a transactivation-deficient mutation of the GR (GR^dim/dim)¹²  are viable, show normal erythropoiesis under normal conditions, but fail to respond to hypoxia and show no stress-induced erythropoiesis in spleen.⁴ 

In cell culture, the cooperative action of Epo, SCF, and Dex induces prolonged expansion of primary, proerythroblast-like, c-Kit–expressing erythroid progenitors, which appears to mimic stress erythropoiesis.^4,13  Additionally, immortal cultures can reproducibly be established from p53^–/– erythroid progenitors that remain fully factor dependent and differentiation competent and retain a stable diploid genome, which allowed unlimited expansion (eg, line I/11).^3,13,14  In these models, Epo is strictly required both for terminal differentiation and for expansion. In the absence of Dex, SCF promotes additional cell divisions and delays red cell differentiation, but is unable to cause a complete differentiation arrest.^3,13,15,16  On the other hand, activation of the GR upon induction of terminal differentiation by Epo hardly affects differentiation except for reduced hemoglobin accumulation.¹⁴  Thus, the cooperation of SCF, Epo, and Dex is required to allow expansion of erythroid progenitors and arrest differentiation. The receptors for Epo and SCF are located in a joint signaling complex at the plasma membrane and activate common signaling pathways, such as the phosphatidylinositol-3-kinase (PI3K) and the mitogen-activated protein kinase/extracellular signal–regulated kinase (MAPK/ERK) cascade,^3,17,18  albeit to a different extent. In addition, the EpoR is solely responsible for signal transducer and activator of transcription 5 (Stat5) activation, a pathway known to cooperate with the GR through synergistic and/or antagonistic effects on transcription.^19-21 

To analyze how cooperative signaling by Epo, SCF, and Dex may control gene expression relevant for progenitor renewal, we used expression profiling using cDNA microarrays. Since the existing libraries of expressed sequence tag (EST) sequences derived from mouse tissues may lack cDNA sequences derived from erythroid-specific genes, we also generated a microarray specific for transcripts from hematopoietic cells to complement the genome-wide arrays. We show that combining this novel microarray with microarrays containing known genes/ESTs was suitable to characterize erythroid-specific gene expression patterns. Thus, a set of genes could be identified (ie, BTG1, VDUP1, CXCR4, and RIKEN29300106B05) whose regulation by Epo or Epo/SCF was counteracted by Dex while a similar antagonistic effect of Epo versus SCF was never observed.

Primary erythroid progenitors were expanded from fetal livers of wild-type or GR^dim/dim (C57Bl/6 × 129 mixed background) E12.5 embryos as described.^12,13  Briefly, dissected fetal livers were suspended in serum-free medium (StemPro-34; Life Technologies, Gibco BRL, Carlsbad, CA) containing Nutrient supplement according to the instructions of the manufacturer, and connective tissue was removed by filtering through a cell strainer (70 μM) (Becton Dickinson, Franklin Lakes, NJ). These primary cell suspensions, as well as the immortal I/11 erythroblast line³  were cultivated in StemPro-34 medium supplemented with 2 U/mL human recombinant Epo (Erypo; Cilag, Baar, Switherland), 100 ng/mL murine recombinant SCF (R&D Systems, Minneapolis, MN), 10^–6  M Dex (Sigma, St Louis, MO), and 40 ng/mL insulin-like growth factor 1 (IGF-1) (Promega, Madison, WI). Cultures were maintained at densities between 2 and 4 × 10⁶ cells per milliliter and analyzed daily for cell number and cell size distribution by means of an electronic cell counter (CASY-1; Schärfe-System, Reutlingen, Germany). Primary cell populations were used for experiments until day 10 after seeding.

For acute stimulation with cytokines and hormones, proliferating I/11 erythroid progenitors or cells from primary cultures were washed twice in phosphate-buffered saline (PBS) and seeded at 4 × 10⁶ cells per milliliter in plain Iscove modified Dulbecco medium (IMDM) or StemPro-34, containing Nutrient supplement, without factors. After 4 hours of starvation in this medium, factors were added as indicated (2 U/mL Epo; 100 ng/mL SCF; 10^–6 M Dex; or 3 × 10^–6 M GR antagonist ZK112.993 [ZK], a kind gift from Schering, Berlin, Germany). Cells were harvested after the indicated times.

Total RNA was prepared from 2 to 5 × 10⁷ cells with the use of the Trizol LS reagent according to the manufacturer's instructions (Life Technologies, Gibco BRL). For each sample, 15 μg RNA was separated in 1.4% agarose gels containing formaldehyde, transferred to a nylon membrane (Gene Screen; NEN DuPont, Wilmington, DE), fixed by ultraviolet (UV) irradiation, and hybridized sequentially with various ³²P-labeled cDNA probes as described.¹⁴ 

The custom-made hematopoietic microarray contains approximately 9000 cDNA spots enriched for erythroid- and T-cell–specific cDNAs. RNA isolated from 3T3 fibroblasts and EpH4 epithelial cells was subtracted from expanding I/11 cells and quiescent CD4⁺ T-cell RNA by suppression-subtractive hybridization (SSH) (by means of PCR-Select cDNA Subtraction Kit; Clontech, Palo Alto, CA). Polymerase chain reaction (PCR) products of the enriched 9000 cDNAs were spotted onto poly-lysine–coated glass slides by a chip-spotting robot developed by A. Massimi for the microarray facility at the Albert Einstein College of Medicine of Yeshiva University (New York, NY) and built by Promedia Association (Antwerp, Belgium). The protocols used for the complete process can be found at http://sequence.aecom.yu.edu/bioinf/microarray/protocol.html (accessed July 23, 2003) and in Cheung et al.²²  Prior to hybridization with total RNA, the quality of this RNA was determined by means of a bioanalyzer (2100 Bioanalyzer; Agilent), according to the manufacturer's instructions. For a single hybridization, 30 μg total RNA was reverse transcribed into cDNA by means of oligo(deoxythymidine)_12-18 (oligo(dT)_12-18) primer (Life Technologies) and 4 μL cyanin 5–uridine 5′-triphosphate (Cy5-UTP) (CyDye; Amersham Biosciences, Freiburg, Germany), while control total RNA was labeled with Cy3-UTP. The microarrays were hybridized overnight with the labeled probes. The scanning was performed with a GenePix 4000A scanner (Axon Instruments, Union City, CA), and the analysis used the GenePix program (Axon Instruments). For the cluster analysis, the data were prefiltered by excluding all genes that did not show any signal after all hybridizations. All spots have been included that showed signals in at least 2 experiments out of 5 or 6 and analyzed by means of GeneSpring (Silicon Genetics, Redwood City, CA).

RNA was treated with DNase I (Stratagene, Amsterdam, Netherlands), 1 U/μg RNA, incubated 30 minutes at 37°C, and heat-inactivated 10 minutes at 70°C. RNA concentration was determined on a Pharmacia LKB Ultrospec III UV-visible spectrophotometer (Pharmacia, Uppsala, Sweden). Poly(A)⁺ mRNA was purified from isolated total RNA (10 to 20 μg) with the oligotex mRNA minikit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The cDNA was synthesized from 1 μg total or poly(A)⁺ RNA in a 40 μL reverse-transcription reaction mixture containing 2 μg random hexamers (Amersham Pharmacia Biotech, Piscataway, NJ) or 0.25 μg oligo(dT)_12-18 (Life Technologies) and 40 mM deoxynucleoside triphosphates (dNTPs), heated 5 minutes at 65°C, and incubated 10 minutes at room temperature. Then, 20 μL 5 × first-stranded buffer (250 mM Tris-HCl [tris(hydroxymethyl)aminomethane-HCl], pH 8.3; 375 mM KCl; 15 mM MgCl₂), 200 mM dithiothreitol, 40 U RNasin and 200 U Superscript II RNase H^– reverse transcriptase (Life Technologies) were added. The reaction was incubated 2 hours at 42°C, heated 3 minutes at 95°C, and quick-chilled on ice. The cDNA was diluted 1:10 to 1:400 prior to PCR amplification.

Gene-specific primers corresponding to BTG1 (L16846), CXCR4 (D87747), glucocorticoid-induced leucine zipper (GILZ,NM_010286), MMP-2 (XM_125113), ribonuclease inhibitor (IMAGE:1366946), RIKEN9230106B05 (AK020317), and VDUP1 (BC011212) were obtained from Invitrogen Life Technologies or Sigma-Genosys (The Woodlands, TX). The sequence of the primers used for the amplification were as follows: BTG1, forward 5′TGCAGGAGCTGCTGGCAG3′, reverse 5′TGCTACCTCCTGCTGGTGA3′; CXCR4, forward 5′TGACTTCGAGAGCATTGTGC3′, reverse 5′TGACTCCGTGGAGACGGAAG3′; GILZ, forward 5′TGTATCAGACCCCCATGGAG3′, reverse 5′TCCATGGCCTGCTCAATCTTG3′; MMP-2, forward 5′TGACCTTGACCAGAACACCA3′, reverse 5′TGCCAGGAGTCCATCCTTG3′; ribonuclease inhibitor, forward 5′TCCAGTGTGAGCAGCTGAG3′, reverse 5′TGCAGGCACTGAAGCACCA3′; RIKEN9230106B05, forward 5′TAGGCTAAAGTGCCACGGAC3′, reverse 5′TACATGCCATTCCTGTGGCA3′; and VDUP1, forward 5′TCCTTGCTGATCTACGTCAG3′, reverse 5′TGTCTTGAGAGTCGTCCACA3′.

The real-time reverse-transcription PCR (RT-PCR) assay involves TaqMan technology (PE Applied Biosystems model 7900 sequence detector; Applied Biosystems, Weiterstadt, Germany), which combines rapid thermo-cycling with online fluorescence detection of the PCR products. The reactions were performed in a volume of 25 μL of a mixture containing 4 μL cDNA dilution, primers at 5 μM, and 12.5 μL 2 × SYBR Green PCR Master mix (PE Biosystems, Warmington, United Kingdom) containing Amplitaq Gold DNA polymerase, reaction buffer, dNTP mix with dUTP, passive reference, and the double-stranded DNA (dsDNA)–specific fluorescence dye SYBR Green I. Samples were placed in a 96-well plate, capped, and placed in the TaqMan sequence detector. The amplification program consisted of 50°C with 2-minute hold (AmpErase urasil-DNA glycosylase [UNG] incubation), followed by 95°C with 10-minute hold (AmpliTaq Gold Activation), and then by 40 cycles of denaturation at 95°C for 15 seconds, annealing at 62°C for 30 seconds, and extension at 62°C for 30 seconds. All the different primer pairs have a similar optimal PCR annealing temperature. Acquisition of the fluorescence signal from the samples was carried out at the end of the elongation step. To confirm amplification specificity, the PCR products from each primer pair were subjected to melting curve analysis.

While Epo is required for both expansion and differentiation of erythroid progenitors, SCF and Dex cooperate with Epo to promote expansion and arrest differentiation (Figure 1A). To examine the synergistic interactions and the molecular events downstream of the respective cooperating signaling pathways, we first optimized the conditions to detect joint gene regulation by these factors in I/11 erythroid progenitors. Cells were deprived of growth factors for 4 hours and restimulated by Epo (2 U/mL) and SCF (100 ng/mL) plus an agonist of the GR (10^–6 M Dex) or an antagonist of the GR (3 × 10^–6 M ZK). The expression levels of potential target genes for cooperative Epo/SCF/Dex signaling were then analyzed by Northern blot (Figure 1B). While expression of the Epo-target genes PIM-1 and CIS was enhanced and/or prolonged by exposure to ZK as compared with Dex treatment, activation of suppressor of cytokine signaling 1 (SOCS1) was almost completely repressed in the absence of Dex. However, Dex alone did not induce SOCS1 (data not shown). As observed in other cells,²³  Dex reduced expression of its own receptor in cooperation with Epo and SCF. The expression kinetics of GATA1 and IκBα were initially similar in the presence or absence of Dex, but persisted in the presence of ZK at late time points. Given that c-MYB expression is rapidly down-regulated at the onset of differentiation, while GATA-1 expression is even transiently increased (Figure 1A; also Dolznig et al¹³ ), the decreasing expression of c-MYB compared with GATA1 in the presence of Epo, SCF, and ZK (18 and 24 hours) indicates that the differentiation program has been activated (Figure 1B). In contrast, the ratio between c-MYB and GATA1 remains constant in the presence of Epo, SCF, and Dex. Steady-state activation of EpoR and c-Kit in the presence of either Dex or ZK showed much weaker expression changes in these genes (data not shown). This suggested that acute stimulation of EpoR/c-Kit signaling in the presence or absence of Dex is the method of choice to analyze cooperative effects on gene expression by the 3 receptors required for renewal.

Although a large choice of EST libraries is available for gene expression studies, these libraries are still likely to be devoid of genes specifically expressed in red blood cells or their progenitors. Therefore, we generated a novel microarray, enriched for specific hematopoietic transcripts. Messenger RNA prepared from proliferating erythroid progenitors (I/11 cells)³  and resting CD4⁺ T-cells (freshly isolated by fluorescence-activated cell sorter [FACS]) was used to generate an erythroblast- and a T-cell–specific library by SSH, by means of mRNA from a mixture of fibroblasts (NIH 3T3) and epithelial cells (EpH4) for subtraction. Individual cDNA clones from the 2 subtracted libraries were spotted onto glass slides, which were hybridized with Cy5-labeled cDNA derived from I/11 erythroid cells (red) and Cy3-labeled cDNA derived from CD4⁺ T-cells (green) (Figure 2A). As expected, most of the spotted clones were specific either for erythroid progenitors (red) or mature T-cells (green) (Figure 2A), while only a minority of the genes represented generally expressed ones.

To evaluate the quality of the libraries produced and to identify the genes represented by the erythroid- or T-cell–specific clones, 192 randomly selected clones from the erythroid-specific library and 96 from the T-cell specific library were sequenced (Table 1). Despite the SSH protocol, the erythroid library was redundant; for instance, beta-globin was found 59 times in the 192 erythroid-specific sequences analyzed. Other erythroid-specific genes like the carbonic anhydrase I or II (CAI, CAII), or Rhag (Rhesus-associated antigen) were also found more than once (Table 1). However, less abundantly expressed genes like p53, eEF1α1, p-selectin, PKCδ, or RNA helicases were also present on the microarray. In addition, we found several sequences without significant homology to sequences in the public databases, most likely representing unknown genes.

First, we used the hematopoietic microarray to study gene regulation by Dex in I/11 erythroid progenitors during acute Epo/SCF stimulation. After factor withdrawal for 4 hours, cells were incubated in the presence of Epo, SCF, and either Dex or the GR antagonist ZK for 2, 6, 18, and 24 hours and processed for mRNA isolation. Cy3-labeled cDNA was generated from Epo/SCF/ZK–treated cell samples, while Cy5-labeled cDNA was prepared from the Epo/SCF/Dex–treated cells. From each time point, the mixed Cy5/Cy3 probes were hybridized to the hematopoietic microarrays. To test for reproducibility, the microarrays were also hybridized with cDNAs labeled with the opposite dye combination, yielding similar results (Figure 2B).

After quantification, the results from the 2 hybridizations were combined and analyzed by means of the clustering program GeneSpring, applying a hierarchic clustering algorithm and a pseudo–color visualization matrix. Data for Dex–up-regulated and Dex–down-regulated genes are shown in Tables 2 and 3. As mentioned above, the SSH-derived cDNA arrays contained multiple copies of hematopoietic-specific, abundantly expressed genes. Rhag and BTG1 (B-cell translocation gene 1) were represented at least 48 and 18 times, respectively, always up-regulated by Dex (Tables 2 and 3). A large cluster of down-regulated sequences represented mainly beta-globin cDNAs (617 times on the microarray). The redundancy combined with the microarray hybridization data performed in duplicate allowed a direct estimation of the reproducibility of the data.

Four genes up-regulated by Dex in the presence of Epo/SCF (RIKEN9230106B05, VDUP1, CXCR4, and BTG1; represented on the microarray 6, 4, 1, and 18 times, respectively) were analyzed in more detail. Dex-dependent up-regulation of these genes is shown in Figure 3A, plotted as ²log values of the fold changes between (1) Epo/SCF plus Dex and (2) Epo/SCF plus ZK. The reproducibility of the data is evidenced by the small standard deviation obtained in 2 independent experiments plus concomitant color switch of the hybridization probes (Figure 3A). To validate Dex-controlled expression, the amount of transcript before and after induction by Epo/SCF/Dex or Epo/SCF/ZK was determined by real-time RT-PCR in 2 independent experiments, again in duplicate determinations (Figure 3B).

Except for minor differences, real-time RT-PCR confirmed the data obtained by the cDNA microarray analysis. In fact, the fold-change values were mostly larger when measured by real-time RT-PCR. Following factor deprivation, all 4 genes were down-regulated by Epo/SCF in absence of Dex (ie, in ZK) while coactivation of the GR by Dex diminished this down-modulation or, in case of RIKEN9230106B05, even resulted in up-regulation relative to the factor-deprived control.

To validate our data by an independent approach, we analyzed the expression of putative GR-target genes in erythroid progenitors derived from fetal livers of GR^dim/dim mice. These mice carry a glucocorticoid receptor mutated in the dimerization domain, rendering it unable to contribute to gene transactivation.¹²  Erythroid progenitors from fetal livers of GR^dim/dim embryos or wt littermates were expanded for 7 days. RNA was directly isolated from proliferating cells without factor depletion or stimulation. Microarray analysis showed that all genes up-regulated in the presence of Dex were expressed at lower levels in the GR^dim/dim erythroid progenitors compared with their wt counterpart (Tables 2 and 3). On the other hand, beta-globin, which is strongly down-regulated by the addition of Dex, was up-regulated in the GR^dim/dim erythroid progenitors as compared with wt cells (Tables 2 and 3). Furthermore, real-time RT-PCR analysis confirmed that the selected genes RIKEN9230106B05, VDUP1, BTG1, and CXCR4, found to be up-regulated in the presence of Dex plus Epo/SCF, were down-regulated in GR^dim/dim when compared with wt erythroid progenitors (Figure 3C).

To unravel how the individual factors Epo, SCF, and Dex regulate gene expression and how they cooperate in various combinations, we obviously had to analyze a larger number of potential target genes than were present on our hematopoietic microarray. For that purpose, we used cDNA microarrays containing approximately 17 000 different EST sequences (represented once or maximally twice [17K]; details available at: http://www.imp.univie.ac.at/2003/Beug/erythroblast_renewal/. Accessed on July 23, 2003). The I/11 erythroid progenitors were factor deprived for 4 hours and treated for 2 hours with Epo, Dex, or SCF alone, as well as with combinations of Epo/SCF (ES), Epo/SCF and Dex (ESD) and Epo/SCF and ZK (ESZK). The cDNA from these cells was labeled with Cy5 and hybridized to both the hematopoietic and the 17K cDNA microarray. In parallel, Cy3-labeled cDNA from erythroid progenitors factor deprived for 6 hours was included as a control. First, we defined prefiltering conditions by determining to what extent the different factors affected overall gene expression. Upon stimulation by Epo or SCF, 1.29% and 1.24% of the genes on the 17K cDNA microarray were more than 1.75-fold up- or down-regulated (Table 4). Yet a similar number of Dex-regulated genes was obtained only when genes up- or down-regulated by a ratio exceeding 1.5 were included (Table 4). Therefore, we set the prefiltering threshold to 1.5 (fold induction or repression) for Dex targets and to 1.75 for Epo and SCF targets, being aware of the risk of incorporating a larger number of nonrelevant and false-positive genes in the resulting gene lists.

Applying these thresholds to the combined results of the 9K hematopoietic microarray and the 17K microarray, we performed cluster analysis by means of the GeneSpring program. Epo and SCF induced similar patterns of gene expression changes: genes up-regulated by Epo were never down-regulated by SCF or vice versa (Figure 4A-B). In addition, expression changes induced by Epo but not by SCF were always present in the Epo/SCF (ES) combination, suggesting that effects of the differentiation-inducer Epo are not neutralized by the renewal factor SCF. While a small number of genes were differentially regulated by Dex plus Epo/SCF (Figure 4A), Dex alone caused a completely different pattern of gene regulation (Figure 4B).

According to their regulation in response to either single agents (ie, Epo, SCF, Dex), Epo/SCF, or Epo/SCF/Dex in comparison with Epo/SCF/ZK, genes were clustered into various profile groups (Table 5). Profile group 1 shows that 54 genes were up-regulated by Epo (Table 5) (group 1A), whereas 36 genes were down-regulated by Epo (group 1B). Similarly, with SCF 63 genes were up-regulated (group 2A) and 58 genes down-regulated (2B). In profile group 3, we distinguished between genes regulated by Dex alone (3A has 21 genes up; 3B, 33 genes down), or by Dex alone and by Dex in combination with Epo and SCF (3A′, 18 up; 3B′, 2 down), or only by Dex in combination with Epo and SCF (3A″, 106 up; 3B″, 72 down). Although genes coregulated by Dex and Epo/SCF were a minority, we considered them to be the most interesting, since genes subject to this coregulation seemed to be the best candidates to give insight into the mechanism of renewal. Table 6, therefore, lists the genes showing up- or down-regulation by Dex, Epo/Dex (profile group 5), SCF/Dex (profile group 6), or Epo/SCF/Dex (profile group 7).