Sox6 Is Necessary for Efficient Erythropoiesis in Adult Mice under Physiological and Anemia-Induced Stress Conditions

Background Definitive erythropoiesis is a vital process throughout life. Both its basal activity under physiological conditions and its increased activity under anemia-induced stress conditions are highly stimulated by the hormone erythropoietin. The transcription factor Sox6 was previously shown to enhance fetal erythropoiesis together and beyond erythropoietin signaling, but its importance in adulthood and mechanisms of action remain unknown. We used here Sox6 conditional null mice and molecular assays to address these questions. Methodology/Principal Findings Sox6fl/flErGFPCre adult mice, which lacked Sox6 in erythroid cells, exhibited compensated anemia, erythroid cell developmental defects, and anisocytotic, short-lived red cells under physiological conditions, proving that Sox6 promotes basal erythropoiesis. Tamoxifen treatment of Sox6fl/flCaggCreER mice induced widespread inactivation of Sox6 in a timely controlled manner and resulted in erythroblast defects before reticulocytosis, demonstrating that impaired erythropoiesis is a primary cause rather than consequence of anemia in the absence of Sox6. Twenty five percent of Sox6fl/flErGFPCre mice died 4 or 5 days after induction of acute anemia with phenylhydrazine. The others recovered slowly. They promptly increased their erythropoietin level and amplified their erythroid progenitor pool, but then exhibited severe erythroblast and reticulocyte defects. Sox6 is thus essential in the maturation phase of stress erythropoiesis that follows the erythropoietin-dependent amplification phase. Sox6 inactivation resulted in upregulation of embryonic globin genes, but embryonic globin chains remained scarce and apparently inconsequential. Sox6 inactivation also resulted in downregulation of erythroid terminal markers, including the Bcl2l1 gene for the anti-apoptotic factor Bcl-xL, and in vitro assays indicated that Sox6 directly upregulates Bcl2l1 downstream of and beyond erythropoietin signaling. Conclusions/Significance This study demonstrates that Sox6 is necessary for efficient erythropoiesis in adult mice under both basal and stress conditions. It is primarily involved in enhancing the survival rate and maturation process of erythroid cells and acts at least in part by upregulating Bcl2l1.


Introduction
Erythropoiesis or red blood cell formation is a vital process throughout life. It first occurs in the yolk sac of the mammalian embryo. Then referred to as embryonic or primitive erythropoiesis, it produces red blood cells (RBCs) that are large, nucleated, and short-lived, and that contain hemoglobin made with the products of embryonic globin genes. Embryonic erythropoiesis is taken over by definitive erythropoiesis when the liver starts developing around day 10 of mouse gestation. This organ is the fetus erythropoietic center. It releases RBCs that are small, enucleated, long-lived, and whose hemoglobin is made with the products of the alpha and beta adult globin genes. Erythropoiesis transiently switches to the spleen around birth, before primarily and definitively homing to the bone marrow. It remains a very active process throughout postnatal development, constantly adjusting its RBC output to the growing body size. Under physiological conditions, a healthy adult mouse or human maintains a basal level of erythropoietic activity in the bone marrow, renewing 1-2% of its RBC population on a daily basis. A number of conditions that result in severe hypoxia force erythropoiesis to quickly and efficiently increase its output and eventually renew the entire pool of RBCs in just a few days. This accelerated process is called stress erythropoiesis and mainly occurs in the spleen [1]. In the most severe cases, it is triggered by acute anemia due to hematological disorders, severe blood loss, or chemotherapy. These conditions can be fatal if stress erythropoiesis is impaired. It is therefore important to fully understand the molecular mechanisms that underlie basal and stress erythropoiesis to be able to propose improved strategies to stimulate stress erythropoiesis in patients with such conditions. Basal and stress erythropoiesis occur in several developmental cell stages and are controlled by complex molecular networks. Specific transcription factors govern the recruitment of hematopoietic stem cells to the erythroid lineage and the sequential differentiation and maturation of the cells into erythroid progenitors, proerythroblasts, erythroblasts, reticulocytes, and erythrocytes [2][3][4]. Best known are the Gata1 and Fog1 zinc finger factors, the Eklf erythroid Krüppel-like factor, and the Foxo3a forkhead factor [5][6][7][8][9][10]. Erythroid cell survival and proliferation are largely controlled by the hormone erythropoietin (Epo) [11][12][13], but also by the stem cell factor, glucocorticoids, and bone morphogenetic protein-4 [14][15][16]. Epo is produced by the kidney at a low level under basal conditions, and at a very high level in response to hypoxia. It binds to a specific receptor, EpoR, expressed almost exclusively on erythroid cells, and signals principally through the Jak/Stat and Akt/PI3K pathways. Many factors necessary for basal erythropoiesis, such as Epo and Gata1, are also required for stress erythropoiesis [1,17,18]. In contrast, molecules like the Stat5 transcription factor and Bcl-xL antiapoptotic factor are dispensable for basal erythropoiesis, but essential to ensure the high efficiency of stress erythropoiesis downstream of Epo [19][20][21][22].
Sox6 is a transcription factor with an Sry-related high-mobilitygroup DNA-binding (Sox) domain and a highly conserved coiledcoil homodimerization domain, but with no known transactivation or transrepression domain [23]. While many Sox proteins are required for cell fate specification and differentiation in discrete lineages, Sox6 forms with Sox5 and Sox13 a Sox group (SoxD) that act as significant modulators rather than obligatory determinants of cell fate and differentiation. For instance, Sox6 is co-expressed with Sox5 in chondrocytes and the two proteins act redundantly to potentiate transactivation of cartilage marker genes by the chondrogenic master Sox9 and thereby to achieve efficient chondrogenesis [24,25]. Sox5 and Sox6 are also co-expressed and acting redundantly in oligodendrocytes [26]. Their roles in this lineage are to delay gliogenesis by blocking the ability of Sox9 and Sox10 to bind and transactivate oligodendrocyte markers. Sox6 and its relatives thus have the ability to enhance as well as repress gene transcription. The determinants of their positive and negative influence on transcription include specific gene sequences and protein partners [23]. Sox6 2/2 mice die at birth or in the third week of age with skeletal, glial, and cardiac developmental defects [24][25][26][27][28][29]. They also feature a large proportion of nucleated definitive red blood cells and are anemic, despite an increased Epo level and liver volume [30]. Sox6, but neither Sox5 nor Sox13, is expressed in definitive erythroid cells and has cellautonomous roles in these cells [30,31]. It potentiates the ability of Epo to stimulate early erythroid cell survival and proliferation, and it also facilitates erythroblast and reticulocyte maturation, but the molecular mechanisms underlying these functions remain unknown [30]. The embryonic globin genes are currently the only genes whose expression is known to be altered in Sox6 2/2 erythroid cells. While these genes are silent in normal definitive erythroid cells, they are highly expressed in Sox6 2/2 cells. Sox6 was shown to bind to the proximal promoter of the embryonic epsilon-globin gene and to interact with Bcl11a to repress epsilonglobin gene expression [31,32]. It is unknown, however, whether ectopic expression of embryonic globins is responsible for the survival, proliferation or maturation defects of Sox6 2/2 erythroid cells or whether Sox6 also modulates the expression of other genes in erythroid cells. Sox6 2/2 /wild-type hematopoietic chimeras were shown to maintain expression of the embryonic epsilonglobin gene in adulthood and to have misshapen RBCs [33], but these mice did not address the question of the need for Sox6 in basal or stress erythropoiesis in adulthood.
In light of the importance of basal and stress erythropoiesis in adulthood, we undertook the present study to further uncover the importance and molecular roles of Sox6 in this process. We show that Sox6 continues to enhance erythroid cell development throughout adulthood in the mouse. Its contribution is significant in basal erythropoiesis and critical to ensure mouse survival and prompt recovery in stress erythropoiesis. Embryonic globins are scarce in the developing erythroid cells and RBCs of Sox6 mutants and their presence appears inconsequential. Sox6 effectively upregulates the expression of multiple late erythroid cell markers, and one of its direct targets is Bcl2l1, which encodes Bcl-xL, an important survival factor acting downstream and beyond Epo signaling in erythroid cells.

Mouse generation and analysis
Mice were used according to federal guidelines and as approved by the Cleveland Clinic Institutional Animal Care and Use Committee (protocol ARC 8511, originally approved on 11.29.2007 -latest annual renewal approved 11.24.2009). They were carrying Sox6 null alleles [25], Sox6 conditional null alleles [34], an ErGFPCre transgene [35], or a CaggCreER transgene [36], as described. Tamoxifen (Sigma-Aldrich, St Louis, MO) was dissolved at 20 mg/ml in corn oil and injected intraperitoneally at 9 mg per 40 g mouse 3 times at 2-day intervals [36]. Phenylhydrazine (Sigma) was dissolved at 6 mg/ml in PBS and injected intraperitoneally at 60 mg/kg on two consecutive days [37]. All mice were on a mixed 1296 C57BL/6J genetic background. Complete blood cell tests, serum Epo level assays, fluorescenceactivated cell sorting (FACS) analyses, and globin chain and ghost analyses were performed as described [30]. CD44 antibody was purchased from BD Biosciences (San Jose, CA). The life span of RBCs was measured by injecting 3 mg EZ-Link-sulfo-NHS-Biotin (Pierce, Rockford, IL) in adult mice and measuring the decay of labeled RBCs by fluorescence-activated cell sorting (FACS) using streptavidin-phycoerythrin [38]. FACS was carried out using an LSRII instrument (BD Biosciences) and data were analyzed using FlowJo software (Tree Star, Ashland, OR). For the assay of reactive oxygen species (ROS), blood cells were suspended in PBS supplemented with 2% FCS and loaded with 5 mM 5-(and 6-)chloromethyl-29,79-dichlorodihydrofluorescein diacetate (CM-H2DCFDA; Invitrogen, Carlsbad, CA) in the dark for 20 minutes at 37uC, 5% CO 2 . The oxidative conversion of CM-H2DCFDA to its fluorescent product was measured immediately by flow cytometry. A positive control was created by incubating cells with 3 mM tert-butylhydroperoxide (Invitrogen). Western blot with a hemoglobin beta/gamma/delta/epsilon antibody (SC-22718, Santa-Cruz Biotechnology, Santa Cruz, CA) was performed using a standard protocol.

In vitro cultures and assays
Erythroid colony formation assays and primary erythroid cell cultures were performed as described using the whole cell populations of adult erythropoietic tissues and TER119 2 fetal liver cells, respectively [30,42]. Reporter assays performed as described [39]. Briefly, Cos-1 cells were transfected with FuGENE6 (Roche, Madison, WI), a mouse Bcl2l1 reporter, a Sox6 [24], constitutively active Stat5 [21] or empty expression plasmid, and a pSV2betaGal plasmid as control for transfection efficiency. The Bcl2l1 reporter was made by cloning a 2.8-kb Bcl2l1 gene fragment containing the upstream enhancer [43], promoter, first exon and intron, and second exon up to the first codon into the pGL3 basic reporter plasmid (Promega, Madison, WI). This fragment was amplified from mouse genomic DNA using AAGATCAAGGGCAGCCCAC (FP) and CCATGGTC-CAAAACACCTGCTCACTTACTG (RP). Chromatin immunoprecipitation was performed as described [39]. Briefly, 3610 6 cells were crosslinked with 0.5% formaldehyde for 10 min followed by addition of glycine at 125 mM. Chromatin was sheared by sonication to fragments averaging 400 bp in buffer containing 1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1 and protease inhibitor cocktail (Sigma-Aldrich). Chromatin was pre-cleared on protein G-coupled Dynal magnetic beads (Invitrogen) followed by immunoprecipitation with antibodies against Sox6 [24] or Stat5 (sc-836X, Santa-Cruz Biotechnology) coupled to protein Gcoupled Dynal magnetic beads. No antibody and non-immune IgG were used as controls. TGGAAGTCCCTTTAG-GGTTTCGGA (FP) and TGGTTCCTCCATCGACCAGAT-CAA (RP) amplified a 236-bp Bcl2l1 promoter region; TGT-TGGACACCGACATCGAAAGGA (FP) and AGAAAGG-GACTGGCATCGAGACAT (RP) amplified a 177-bp Bcl2l1 first-intron region; and TGAGTCCTATCCTGGGAACCATCA (FP) and ATTTATAGGAACCCGGATGGTGGG (RP) ampli-fied a 284-bp Gapdh promoter region. PCR was done with 1 cycle at 94uC for 3 min; 40 cycles at 94uC for 20 sec, 59uC for 30 sec, and 72uC for 30 sec; and 1 cycle at 72uC for 10 min.

Statistical analysis
Numerical values were expressed as means 6 SD. The significance of differences between control and mutant mice was determined using the statistical Student's t-test. A p value less than 0.05 was considered significant.

Results
Sox6 is needed to facilitate erythroid development and RBC survival in basal erythropoiesis Sox6 was previously shown to be expressed in definitive erythroid cells in mouse fetuses and one-week-old pups, but was not tested in adult mice [30]. Using total RNA from bone marrow and spleen of 6-, 10-, and 20-week-old mice, we found that Sox6 remained highly expressed in bone marrow throughout adulthood (Fig. 1A). It was also highly expressed in the spleen of young adults, but became downregulated as mice aged. Expression of the adult beta-globin gene similarly decreased with age, reflecting fading of erythropoietic activity in this organ under steady-state conditions for erythropoiesis.
To test whether Sox6 controls erythropoiesis in adulthood, we generated mice harboring Sox6 conditional null alleles (Sox6 fl/fl ) [34] and the ErGFPCre allele [35]. ErGFPCre is an EpoR knock-in of a gene encoding GFP fused to Cre recombinase. It is highly and specifically expressed in erythroid cells. Sox6 RNA was undetectable in the bone marrow and spleen of 10-week-old Sox6 fl/fl ErGFPCre mice, proving efficient inactivation of Sox6 in erythroid cells (Fig. 1B). These mice looked externally normal at all ages and complete blood cell counts (CBC) revealed that most of their blood parameters were normal ( Fig. 1C; BD and VL, unpublished data). Noticeably, however, their hemoglobin level, hematocrit, and RBC mean cell volume were 5 to 13% lower than normal; they had a 16% increase in red cell distribution width, reflecting RBC anisocytosis, and a 47% increase in reticulocyte count, revealing reticulocytosis; moreover, their serum Epo level (sEpo) was 183% as high as normal, and their spleen was 152% as large as normal. In standard colony-forming assays in vitro, mutant spleen cells generated 32% more erythroid colony-forming units (CFU-Es) than wild-type cells and, although they were slower than controls, they eventually generated 50% more erythroid burst-forming units (BFU-Es) than controls (Fig. 1D). These data thus indicated that Sox6 ablation in erythroid cells results in compensated anemia in adult mice.
To uncover the cause of this phenotype, we analyzed erythropoietic tissues by FACS. We first used antibodies against CD71 (transferrin receptor) to label late proerythroblasts and erythroblasts, and antibodies against TER119 (glycophorin Aassociated protein) to label erythroblasts (Fig. 1E). Mutant bone marrow and spleen tissues showed identical phenotypes, with 25-28% fewer non-/early erythroid cells (CD71 2 /TER119 2 ) than control tissues, and a 35-38% increase in the ratio of erythroblasts (CD71 + /TER119 + ) to late proerythroblasts (CD71 + /TER119 2 ). Using an antibody against the CD44 cell adhesion glycoprotein [44], we found that the size of two populations of developing erythroid cells was significantly altered in mutant tissues: the proerythroblast population (CD44 high /TER119 2 ) and the erythroblast population (CD44 med /TER119 + ) (Fig. 1F). The population of non-erythroid and erythroid precursor cells (CD44 2/med / TER119 2 ) appeared unchanged, and we therefore used it to calculate that the population of proerythroblasts was depleted by 33%, whereas the population of erythroblasts was more affected, with an enlargement of 60%. We further analyzed the erythroblast population using the thiazole orange (TO) nucleic acid dye to measure chromatin condensation and enucleation, and using the cell forward scatter to measure cell condensation. We thereby found that the increased proportion of erythroblasts in mutant tissues was due to a 3-fold increase in the ratio of uncondensed (FSC high /TO high ) versus condensed erythroblasts (FSC low / TO med ), i.e., early versus late erythroblasts (Fig. 1G). We reached the same conclusion when we analyzed the cells using their FSC and CD44 antibodies instead of TO (data not shown). These data thus demonstrated that Sox6 is needed in adult mice to facilitate the development of erythroid cells. It is importantly needed at the proerythroblast stage and critically needed at the erythroblast stage.
We then analyzed circulating RBCs. We observed that mature RBCs (TO 2 /CD71 2 ) accounted for more than 95% blood cells in control mice, as expected, but for only 70% blood cells in mutant mice (Fig. 1H). In addition, while 67% control reticulocytes (CD71 + and/or TO + ) were retaining RNA longer than CD71 (TO + /CD71 low/2 ), 81% mutant reticulocytes were maintaining CD71 longer than RNA (TO 2 /CD71 + ). Reticulocytes were thus maturing slowly and abnormally in mutant mice.
Finally, decay profiling of biotin-labeled cells revealed that RBCs had a half-life reduced to 19 days in mutant mice compared to 25 days in control mice (Fig. 1I). We concluded that adult mice

Impaired erythroid development is a primary cause of anemia upon Sox6 inactivation
We next tested whether the abnormal development of erythroid cells in Sox6 fl/fl ErGFPCre mice was a direct consequence of Sox6 inactivation or was due to anemia-induced stress erythropoiesis as a result of reduced survival or malfunction of RBCs. We generated Sox6 fl/fl adult mice harboring a widely expressed CaggCreER transgene [36]. This transgene encodes a fusion of Cre recombinase with a mutant estrogen ligand-binding domain, which exclusively binds tamoxifen (TM). The CreER protein is thus able to recombine genes only when mice are treated with TM. Sox6 fl/fl CaggCreER mice showed no visual sign of disease before and for at least one month after TM injection (BD and VL, unpublished data). The Sox6 RNA level was normal in bone marrow and spleen before treatment, and became undetectable 7 and 14 days after treatment ( Fig. 2A). CBC parameters remained normal for at least 2 weeks ( Fig. 2B; and, BD and VL, unpublished data). The ratio of uncondensed versus condensed erythroblasts doubled within the first week (Fig. 2C), but abnormal reticulocytes appeared in the circulation only in the second week (Fig. 2D). Developing erythroid cells thus became abnormal before the reduced lifespan or other defects of mutant RBCs became consequential, indicating that impaired erythroid cell maturation is a primary cause of anemia upon Sox6 inactivation.
Sox6 is critically needed for efficient stress erythropoiesis The relatively mild erythroid defects of Sox6 fl/fl ErGFPCre adult mice in basal conditions prompted us to determine whether this phenotype would be exacerbated in stress erythropoiesis. We induced acute anemia by injecting Sox6 fl/fl and Sox6 fl/+ ErGFPCre control mice and Sox6 fl/fl ErGFPCre mutant mice with phenylhy- drazine (PHZ), a powerful oxidative agent [37]. As expected, all pre-existing RBCs decayed in control mice within a week (Fig. 3A). A similar decay was observed in mutant mice, demonstrating similar sensitivity of mutant and control RBCs to PHZ. The hematocrit of all mice decreased by half within 2 days (Fig. 3B). Control mice started to recover by day 4 and all survived and regained a normal hematocrit by day 10 ( Fig. 3B and C). In contrast, mutant mice maintained a low hematocrit for at least two more days and 25% of them died at day 4 or 5. The survivors started to recover by day 7, but their hematocrit was still 10-15% lower than normal at day 10.
A quick, massive surge in sEpo level is a critical, early response to acute anemia. This surge occurred similarly in control and mutant mice and led to a maximal sEpo level from day 2 to day 4 (Fig. 3D). The sEpo level returned to its basal value by day 7 in control mice, but only by day 10 in mutants (Fig. 3D). Hence, the sEpo level was not responsible for the impaired recovery of mutants. Stress erythropoiesis also involves hyperplasia of the spleen, due to erythropoietic burst. Control spleens enlarged about 8 times by day 4 and returned to their original size after day 10 (Fig. 3E). Mutant spleens reached the same final size several days later despite a larger original size and normal surge in sEpo level, and returned to their original size by day 14 only. Sox6 thus critically participates in the erythropoietic burst generated in response to acute anemia. Supporting this conclusion, the Sox6 RNA level strongly increased in control bone marrow and spleen by day 4 and through day 7, and remained undetectable in mutant tissues (Fig. 3F). Interestingly, the Sox6 RNA level increased as the sEpo level surge was fading, but it coincided with the time of death of mutant mice and with the delay in enlarging the spleen and regaining a normal hematocrit. Sox6 thus participates in stress erythropoiesis mostly beyond the Epo-dominated early phase.

Sox6 ensures timely maturation of erythroid cells in stress erythropoiesis
To identify the contribution of Sox6 in stress erythropoiesis, we analyzed erythropoietic tissues and blood samples of PHZ-treated Sox6 fl/fl and Sox6 fl/fl ErGFPCre mice. From day 2 to day 7 post treatment, control and mutant mice increased the proportion of erythroblasts (CD71 + /TER119 + ) versus late proerythroblasts (CD71 + /TER119 2 ) 10-20 fold in bone marrow and spleen ( Fig. 4A and B). Interestingly, the proportion of uncondensed to condensed erythroblasts increased up to 2 fold in control mice and up to 5 fold in mutant mice under stress erythropoiesis (Fig. 4C). Sox6 thus ensures speedy maturation of erythroblasts in both basal and stress erythropoiesis.
Reticulocytes could not be detected in the week that followed PHZ injection, likely due to interference of PHZ with RNA dyes (BD and VL, unpublished data). We therefore assayed reticulocytes using CD71 antibody only. The quick recovery of control mice from acute anemia could be explained by an increase in reticulocyte percentage that became significant by as early as day 3 and that remained at 20-30% between day 5 and day 8 before returning to normal by day 10 (Fig. 4D). In contrast, the inability of mutant mice to quickly recover could be explained by failure to increase the reticulocyte percentage before day 5. Surprisingly, this parameter became almost twice as high as in control mice from day 5 and remained elevated until mice fully recovered. By day 7, when all RBCs were less than 1-week old (Fig. 3A), almost 50% of them were already mature in control mice, but less than 20% were mature in mutants (Fig. 4E). By day 10, about 90% RBCs were mature in control mice, but only about 50% were mature in mutants. Sox6 thus critically intervenes in stress erythropoiesis to ensure prompt maturation of erythroblasts and reticulocytes.

Sox6 upregulates expression of Bcl2l1 and other erythroid terminal markers
To better understand how Sox6 controls erythroid cell development, we screened gene expression microarrays with bone marrow and spleen RNA from Sox6 fl/fl and Sox6 fl/fl ErGFPCre littermates at 3 months of age under physiological conditions and 4 days after induction of acute anemia by PHZ. Volcano plots of p values against gene expression fold changes revealed that the embryonic epsilon globin gene (Hbb-y) was by far the most affected gene, with an upregulation fold of more than 200 in all mutant samples (Fig. 5A).
Hbb-y was previously shown to be a direct target of Sox6 in erythroid cells [31,32], but it was not shown whether its ectopic expression impaired the development of erythroid cells. We tested this possibility because imbalanced production of globin chains in thalassemia results in impaired maturation and survival of erythroid cells in a manner resembling the Sox6-null phenotype [45,46]. Northern blot analysis revealed that the embryonic globin genes were expressed at a much lower level than adult globin genes in Sox6-null erythroid cells (Fig. 5B). Accordingly, embryonic globin chains were hardly if at all detected in RBC lysates (Fig. 5C). A detrimental consequence of the imbalanced production of globin chains in thalassemia is precipitation of globin chains in erythroid membranes. Precipitated chains can be detected in ghosts, i.e., red blood cell membrane extracts. Sox6-null ghosts contained ectopic proteins with an Mr close to 15 and 30k (Fig. 5D). These proteins, however, did not react with an antibody against beta, gamma, delta and epsilon globins (Fig. 5E and F), and mass spectrometry analysis revealed that the 15k and 30k species largely corresponded to H2a/2b and H1 histones, respectively, and did not contain any globin chain (PD and VL, unpublished data). These histones thus reflected the fact that Sox6 mutant mice release a significant proportion of nucleated erythroid cells in the circulation [30]. Another consequence of the abnormal production of globin chains in thalassemia is the generation of reactive oxygen species (ROS) [47]. ROS are very detrimental to cells, causing protein damage and premature death. Their production was not increased in the blood of Sox6-null mice (Fig. 5G). Together, these data thus indicated that the abnormal erythroid phenotype of Sox6 mutant mice is unlikely due to ectopic expression of embryonic globin genes.
Besides Hbb-y upregulation, most other gene expression changes were less than 4-fold in Sox6 mutant samples (Fig. 5A). Few genes had their expression changed under both steady-state and erythropoietic stress conditions, but many genes had their expression affected in both bone marrow and spleen under stress erythropoiesis ( Fig. 6A and Table S1). We selected a subset of upregulated and downregulated genes according to their known or predicted functions and fold expression change, and validated their expression changes by quantitative RT-PCR using RNA from fetal liver and from samples similar to those used for the microarrays. We confirmed that the genes for CD59a and CD59b, which restrict the lysis of human erythrocytes and leukocytes by the homologous complement, tended to be upregulated in the bone marrow and spleen of Sox6 mutant mice (Fig. 6B). All selected downregulated genes were confirmed in this assay and in northern blot ( Fig. 6B and C). They included Bcl2l1, which encodes the anti-apoptotic factor Bcl-xL; Eif5, which encodes the eukaryotic translation initiation factor 5; Lmo2, which codes for the LIM-domain only factor 2 and is critically needed for hematopoietic development; Rnf11, which codes for the E3 ubiquitin ligase ring finger factor 11; St5, which codes for the suppressor of tumorigenicity 5 factor; Tmcc2, the gene for the transmembrane and coiled-coil domain family 2; and Xpo7, the gene for exportin 7. Most of these genes were upregulated at day 4 and 7 during stress erythropoiesis in wild-type mice, coincidentally with erythroid terminal maturation and Sox6 expression (Fig. 3F and 6D). They were also upregulated when wild-type erythroid cells were reaching the early and late erythroblast stages at 24 and 48 h in primary cultures, respectively (Fig. 6E). These data thus indicated that Sox6 is needed to upregulate expression of late erythroid markers.
Bcl2l1 was previously shown to mediate erythroid cell survival and to be a direct target of the transcription factor Stat5 downstream of Epo signaling [19][20][21]. Interestingly, we found that upregulation of Bcl2l1 expression in response to Epo signaling required Sox6 (Fig. 6F). Furthermore, Sox6 was able to upregulate a Bcl2l1 reporter containing the most highly conserved non-coding Probes were for adult alpha-globin (Hba-a1) and beta-globin (Hbb-b1), and embryonic X-globin (Hba-x), Y-globin (Hbb-y), and Z-globin (Hbb-z). All had similar specific activity and length. X-ray films were exposed for 30 min for adult globin RNAs, and 4 days for embryonic globin RNAs. The latter were thus weakly expressed compared to the former. (C) Electrophoretic analysis of globins in RBCs from E12.5 wild-type embryo and control and mutant adult mice 0 and 4 days after PHZ treatment. Globin chains are indicated. Several beta-globin isoforms are seen due to mixed genetic background. (D) RBC ghosts from 3-day-old (P3) Sox6 +/+ , Sox6 +/2 , and Sox6 2/2 littermates, and 4-month-old Sox6 fl/fl (C) and Sox6 fl/ fl ErGFPCre (M) mice 0 day (steady-state) and 4 days (stress erythropoiesis) after PHZ treatment. Protein marker molecular weights and typical RBC proteins are indicated. Blue arrows, ectopic histones in mutants. Red arrows, proteins reacting with hemoglobin antibody against beta-, gamma-, delta-, and epsilon-globins. (E) Coomassie blue staining of globins in RBC lysates from control mice at E13.0, E15.5, and postnatal day 7 (P7) and western blot with hemoglobin antibody. (F) Western blot of ghosts from P3 mice with hemoglobin antibody. Red arrows, all samples have similar residual globin chains. (G) Flow cytometry shows that minimal ROS production in blood of control and mutant adult mice, except in control blood treated with tert-butyl hydroperoxide (TBHP). doi:10.1371/journal.pone.0012088.g005 regions of the gene, including an upstream enhancer, the proximal promoter and first intron (Fig. 6G). These regions were previously shown to control Bcl2l1 expression in erythroid cells in response to Epo signaling [21,43]. Sox6 was as potent as constitutively active Stat5 in upregulating the activity of the Bcl2l1 reporter, and the two factors had additive effects. Additionally, a chromatin immunoprecipitation assay in wild-type fetal liver cells provided evidence that Sox6, like Stat5, directly binds to the promoter and first intron of the endogenous Bcl2l1 gene in erythroid cells in vivo (Fig. 6H). These data thus strongly suggest that Bcl2l1 is a direct target of Sox6 downstream of Epo signaling and during erythroid terminal maturation.

Discussion
This study has uncovered that Sox6 is expressed in erythroid cells throughout life and acts as a significant enhancer of both basal and stress erythropoiesis in the adult mouse. Sox6 has important roles in the development of proerythroblasts and also has critical roles in promoting maturation of erythroblasts and reticulocytes and in extending the life span of RBCs. Its ability to repress the embryonic globin genes appears inconsequential. In contrast, its ability to directly enhance Bcl2l1 expression likely contributes to its ability to work with and beyond Epo signaling in promoting erythroid cell survival.
We demonstrated that Sox6 significantly contributes to ensuring efficient basal erythropoiesis by showing that Sox6 fl/fl ErGFPCre adult mice had compensated anemia. Clear signs of compensation included an increase of up to 2 fold of the reticulocyte count, sEpo level, spleen size, and a 32-50% increase in the number of erythroid colonies developing in vitro from spleen cells. It is very likely that erythroid defects are the cause of this anemia since wildtype adult mice strongly express Sox6 in erythropoietic tissues and since Sox6 fl/fl ErGFPCre mice inactivate Sox6 only in erythroid cells. Further supporting this mechanism, we showed that Sox6 fl/fl ErGFPCre mice had erythroid cell defects and decreased RBC life span, and that Sox6 fl/fl CaggCreER adult mice, which inactivated Sox6 only upon tamoxifen injection, developed erythroblast defects before anemia. The shorter half-life of RBCs must have contributed to cause anemia, but its value at 19 instead of 25 days was probably insufficient to justify the doubling of the sEpo level and spleen size. It is likely that abnormal maturation of RBCs reduced the functionality of these cells and thus contributed to causing hypoxia. The enlargement of the spleen could have several causes. We demonstrated that it was partly due to increased in erythropoietic activity, since more erythroid colonies could be generated in vitro from mutant spleens than from control spleens. We showed that it was also due to a larger population of erythroblasts. As previously reported for fetal liver erythroid cells [30], erythroblasts from adult bone marrow and spleen were impaired in maturing and thus delayed in entering the circulation. Interestingly, we found that the erythropoietic tissues of adult mutants had a reduced proportion of proerythroblasts, revealing that Sox6 is needed to promote expansion of proerythroblasts. We previously reported that Sox6 ablation in fetal liver primary erythroid cells resulted in decreased survival and proliferation of proerythroblasts [30]. Similar defects in proerythroblasts are thus likely to occur in Sox6 mutant adult mice. They would explain that erythropoietic tissues contained fewer proerythroblasts and that BFU-Es formed more slowly from mutant than from wild-type adult erythropoietic tissues in vitro. Our new data thus demonstrate that Sox6 enhances definitive erythropoiesis in a similar way in the fetus, young pup, and adult mouse. While this is not entirely surprising, we could not have extrapolated the conclusion previously drawn in young animals to adults without experimentation since erythropoiesis occurs at a much lower rate in the adult under steady-state conditions than in the rapidly growing fetus and pup, and since major differences exist in erythroid gene expression between fetuses and adult mice [48,49].
We demonstrated that Sox6 is more critically needed in adult mice subjected to stress erythropoiesis than in mice under steadystate conditions of erythropoiesis. Similar conclusions were previously reached for other factors, such as Bcl-xL and Foxo3a, and is explained by the fact that stress erythropoiesis is a much more productive process than basal erythropoiesis and thus critically relies on factors capable of boosting it. Sox6 fl/fl ErGFPCre mice showed a mortality of 25% when subjected to acute anemia, whereas all control mice survived. RBCs decayed at the same rate in control and mutant mice and the sEpo level increased similarly in both types of mice, ruling out increased sensitivity of mutants to PHZ or inability to generate a sufficient surge in Epo level. The hematocrit of control and mutant mice decreased by half within 2 days of PHZ injection, but remained at this low value much longer in mutant than control mice. Such a low value is not life threatening in a normal mouse. While the hematocrit of the mutant mice that died could have further dropped, additional defects could also have contributed to lethality. Reticulocytes were maturation impaired, maintaining a high level of CD71 longer than control cells. We tried to measure hemoglobin levels in the first week after PHZ treatment, but the high degree of hemolysis at that time did not permit accurate measurement. We previously showed that the RBC hemoglobin content of Sox6 2/2 fetuses was about 66% the normal value and that Sox6 2/2 primary erythroid cells were delayed in hemoglobinizing [30]. It is thus possible that the first RBCs released in mutant mice under stress were poorly hemoglobinized. Moreover, their immature shape and possibly other defects may have further reduced their ability to properly deliver oxygen to tissues.
We showed that Sox6 has unique roles in stress erythropoiesis. Sox6 is dispensable to generate the surge in sEpo level and the Epomediated early response to acute anemia, which includes spleen enlargement and massive production of proerythroblasts and early erythroblasts. Further supporting the notion that Sox6 is not involved in this early response was our observation that Sox6 expression increases in bone marrow and spleen only from day 4 to 7 after PHZ injection. We previously showed that Sox6 enhances the ability of Epo signaling to promote proerythroblast survival and proliferation in vitro [30] and we showed here that Sox6 is needed probably in the same way to expand the population of proerythroblasts in adult mice under physiological conditions. This function of Sox6, however, is probably largely compensated for in mutant mice subjected to acute anemia by the very high increase in sEpo level. In contrast, the major role of Sox6 to promote erythroblast maturation in the second phase of the stress erythropoietic response, including cytoplasmic and nuclear condensation, is not fully compensated. Such defects were seen under basal conditions, but they were worsened under stress conditions. Sox6 thus most critically control erythroid cells at the terminal maturation stages at all mouse ages in both basal and stress conditions for erythropoiesis.
We provided new insights into the mechanisms whereby Sox6 controls erythroid development at the molecular level by screening gene expression microarrays and validating data with various RNA and protein assays. Previous studies in fetuses and hematopoietic chimeras revealed that the embryonic globin genes are ectopically expressed in Sox6-deficient definitive erythroid cells [30][31][32][33]. We have extended these findings to adult mice in both basal and stress erythropoiesis conditions. Importantly, we have also demonstrated that the globin RNA and protein levels are negligible compared to those of adult globins and that these proteins do not precipitate in the erythroid membranes and do not increase the production of damaging reactive oxygen species, in contrast that what is seen when abnormal ratios of globin proteins are produced in thalassemia patients. Repression of embryonic globin genes is thus unlikely to be the most consequential function of Sox6 in erythroid cells. We found that many other genes were upregulated or downregulated in the absence of Sox6. The expression changes of these genes were only a few folds, a finding consistent with the fact that Sox6 does not have a transactivation or transrepression domain, and has been shown to act in other cell types as a positive or negative modulator of gene expression [23]. One of the downregulated genes was Bcl2l1, which encodes the anti-apoptotic factor Bcl-xL. Previous studies have shown that Bcl2l1 is a direct target of Stat5 downstream of Epo signaling in early erythroid cells [21], and that it is upregulated in late erythroid cells in a Stat5-independent manner to promote erythroblast and reticulocyte survival [22,[50][51][52]. In agreement with these studies, we showed that Bcl2l1 expression closely parallels Sox6 expression in late erythroid cells and is dependent upon both Epo signaling and Sox6. Further, we demonstrated that both Sox6 and Stat5 are able to upregulate a Bcl2l1 reporter in cultured cells, and their effects are additive rather than synergistic. This result is consistent with the possibility that Stat5 and Sox6 both contribute to mediate Epo signaling in vivo but act independently of each other rather than cooperatively. While Stat5 is more critical in early erythroid cells, Sox6 is more critical at later stages of erythroid cell development. Further studies in vivo and in vitro will be needed to fully decipher how Stat5 and Sox6 contribute to control Bcl2l1 expression at the various stages of erythroid cell development and in response to Epo signaling. Another interesting gene significantly downregulated in Sox6 mutants is Rnf11. Although its role in vivo remains unknown, it is tempting to speculate that the E3 ubiquitin ligase activity of its ring finger protein 11 product [53] may contribute to the important roles that ubiquitination fulfills in erythroid cells. It has been shown, for instance, that ubiquitination causes rapid inactivation of EpoR upon ligand binding [54] and participates in reticulocyte remodeling, including CD71 sorting into exosomes [55]. Poor ubiquitination could thus account for several defects in Sox6 mutants, including retention of CD71 on reticulocytes. Together, these results thus strongly suggest that Sox6 facilitates erythroid cell development by modulating the expression of a large panel of marker genes.
Definitive RBCs are more efficient than primitive RBCs because they are small, lack a nucleus, contain adult rather than embryonic globins, and are long-lived. We have shown here that Sox6 acts throughout adult life to facilitate erythropoiesis. It is important in both basal and stress erythropoiesis. It ensures that proerythroblasts properly expand and that RBCs are long-lived, but its most critical roles are to ensure that the output of definitive erythropoiesis is not delayed or decreased by the additional steps of cell maturation required for erythroblast condensation and enucleation. In addition to Epo treatments widely used to generate an adequate burst of early erythroid cells in patients with various anemia-causing conditions, one should thus also consider targeting Sox6 to ensure timely maturation and release, as well as long life of newly generated red blood cells.