Direct regulation of p53 by miR-142a-3p mediates the survival of hematopoietic stem and progenitor cells in zebrafish

Hematopoietic stem and progenitor cells have the capacity to self-renew and differentiate into all blood cell lineages, and thus sustain life-long homeostasis of the hematopoietic system. Although intensive studies have focused on the orchestrated genetic network of hematopoietic stem and progenitor cell specification and expansion, relatively little is known on the regulation of hematopoietic stem and progenitor cell survival during embryogenesis. Here, we generated two types of miR-142a-3p genetic mutants in zebrafish and showed that the loss-of-function mutants displayed severe reduction of hematopoietic stem and progenitor cells. Further analysis showed that the diminished proliferation and excessive apoptosis in miR-142a-3p mutants were attributed to the increased p53 signaling. Mechanistically, we demonstrated that miR-142a-3p directly targets p53 during hematopoietic stem and progenitor cell development, and the hematopoietic stem and progenitor cell survival defect in miR-142a-3p mutants could be rescued by loss of p53. Therefore, our work reveals the significance of the miR-142a-3p-p53 pathway in controlling hematopoietic stem and progenitor cell survival, and thus advances our understanding of the role of p53 in vertebrate hematopoiesis.

MicroRNAs are reported to be highly expressed in hematopoietic tissues including bone marrow [17] and emerging evidence has shown that they are implicated in hematopoiesis in different organisms [18][19][20]. In mouse, miR-125 is highly expressed in hematopoietic stem cells (HSCs) to expand HSC numbers in vivo [21]. In vitro, mouse fibroblasts were reprogrammed into hemogenic endothelial cells that differentiated into CD45 + cKit + cells (HSPCs) between day-20 and day-35 culture; in these HSPCs, the highest enrichment of microRNAs including miR-142 was observed [22]. In Xenopus, miR-142-3p has an essential role in establishing the mesodermal lineage that contributes to both HSC emergence and vasculogenesis [23]. In zebrafish, miR-142-3p modulates neutrophil development by controlling its maturation [24]. MiR-142 is also an essential regulator of lymphocyte ontogenesis and megakaryopoiesis [20,25]. In our previous study, suppression of miR-142a-3p by antisense morpholinos shows that HSPC development is impaired when miR-142a-3p is knocked down. Further microarray data and functional assays demonstrate that irf7 acts as a direct target of miR-142a-3p and that the miR-142a-3p-irf7 signal pathway regulates the formation and differentiation of definitive HSPCs possibly through the inflammatory signaling [26]. Another group has reported that ectopic expression of miR-142-3p leads to the reduction of primitive erythrocyte progenitor cells and HSCs in zebrafish [27]. Thus, the role of zebrafish miR-142a-3p in definitive hematopoiesis remains elusive, largely because of different methods used by different groups.
P53, a tumor suppressor involved in many disorders including hematological malignancy [28], can activate DNA repair proteins when DNA is under sustained damage, leading to apoptosis and G1/S cell cycle arrest [29][30][31]. p53 is a critical transcription factor for cell cycle regulation in hematopoietic cells and is involved in the quiescence, self-renewal, senescence and apoptosis of HSCs [32,33]. HSCs from hypermorphic p53 mice, which constitutively activates P53, fail to expand [34] and HSC self-renewal in p53 − / − mice is enhanced [35]. In the zebrafish Diamond-Blackfan model, the downregulation of erythroid marker gata1 is modulated by p53 [36]. Intriguingly, loss of p53 facilitates reprogramming murine fibroblasts to hematopoietic progenitor cells and p53 − / − reprogrammed cells efficiently generated erythroid, megakaryocytic, myeloid, and lymphoid lineages [37]. In addition, suppression of p53 can exert hematopoietic rescue effect on genetic alterations including mysm1 deficiency [38], and p53 − / − HSCs are more resistant to irradiation [32]. Taken together, p53 may be closely linked to the cell cycle and apoptosis and thus critical for the survival and stemness maintenance of HSPCs. Gfi-1, Necdin and P21 were the downstream targets of P53 in mediating HSC quiescence and Puma was responsible for HSC apoptosis [28,29]. However, the upstream regulator(s) of p53 governing HSC development is largely unclear. Figure 1 Generation of miR-142a mutants by TALENs. (a) Schematic diagram showed the TALEN target sites at the miR-142a locus. (b and c) Indel mutations induced by TALENs showed a 13-base-pair deletion in the miR-142a locus. Green arrow shows the initiation of the mutation. (d) Predicted RNA secondary structure of wild-type and mutated pri-miR-142a by TALENs using the RNA Structure Software (version 4.6, Rochester, NY, USA). (e) Expression of miR-142a-3p was absent in 142 T − / − embryos at 4 days post-fertilization (dpf) by WISH. Red dashed circles mark the expression of miR-142a-3p in the thymus and red arrowheads mark the expression of miR-142a-3p in the CHT. (f) qPCR demonstrated that the expression of miR-142a-3p and miR-142a-5p was nearly undetectable in 142 T − / − embryos at 4 dpf. The expression of miR-142a-3p and miR-142a-5p was normalized to U6 (mean ± s.d., n = 3, **Po 0.01).

miR-142-3p in hematopoiesis
In this work, we demonstrate that p53 is a direct target of miR-142a-3p during zebrafish definitive hematopoiesis, and the hematopoietic defects in miR-142a-3p mutants can be rescued by loss of p53, which deepens our understanding of the role of p53 in mediating miR-142a-3p function in HSPC development in zebrafish.

Results
Generation of miR-142a-3p mutants by TALENs and CRISPR/Cas9 In order to substantiate our previous findings that loss of miR-142a-3p leads to decreased HSPCs [26], both TALENs and CRISPR/Cas9 were used to generate genetic mutants of miR-142a. First, we designed a pair of TALEN arms that target two sites inside of the miR-142a gene (Figure 1a). The TALEN mRNAs were injected into WT embryos (F0) and after genotyping of F2 embryos, a mutant, named 142 T − / − , was identified. 142 T − / − embryos were born at the expected Mendel ratio, and they could develop into fertile adults. According to the sequencing result, 13 base pairs were deleted in the locus of the miR-142a gene (Figure 1b and c) and this disrupted the secondary structure of miR-142a as well as the formation of mature miR-142a-3p (Figure 1d). Further analysis revealed that the expression of miR-142a-3p in both the thymus and caudal hematopoietic tissue region was undetectable ( Figure 1e). Consistently, real-time quantitative PCR (qPCR) result demonstrated that miR-142a-3p and another member of miR-142a, miR-142a-5p, were severely decreased (Figure 1f), indicating the loss-offunction of miR-142a-3p and miR-142a-5p.
To generate a large deletion of the miR-142a gene by CRISPR/Cas9, a pair of gRNAs flanking the miR-142a locus was designed and co-injected into zebrafish WT embryos with Cas9 mRNA (Supplementary Figure S1a). After F2 screening, a miR-142a mutant was identified, named 142 C − / − , and the sequencing result displayed a 986-base-pair deletion in the locus of miR-142a, which was able to be employed to distinguish the WT siblings and mutants by PCR (Supplementary Figure S1b). Impaired secondary RNA structure of miR-142a and failure to form the mature miR-142a-3p also occurred in 142 C − / − embryos (Supplementary Figure S1c  Taken together, two zebrafish genetic mutants of miR-142a were generated in which miR-142a-3p was disrupted and thus facilitated functional investigation of miR-142a in zebrafish hematopoiesis.

Loss of miR-142a-3p leads to decreased HSPCs in zebrafish
Previous studies implicate that loss of miR-142a-3p leads to decreased erythrocytes as well as vascular defects [39,40]. Thus, we examined whether miR-142a-3p mutants display such developmental defects. With a somitic marker myod to define specific developmental stages, WISH results showed that primitive hematopoiesis including erythroid progenitors (labeled by scl and gata1) and myeloid cells (labeled by lyz and pu.1) was unaffected in 142 T − / − embryos at 24 and 36 hours postfertilization (hpf), respectively (Supplementary Figure  S2a  To confirm the HSPC defects in miR-142a-3p morphants as reported in our previous study [26], the HSPC marker runx1 was examined in 142 T − / − embryos and WISH result displayed its reduction in the aorta-gonad-mesonephros (AGM) at 26 and 36 hpf ( Concerning the absence of miR-142a-3p and miR-142a-5p in 142 T − / − , it is critical to determine whether the decrease in HSPCs is also attributed to the absence of miR-142a-5p. In fact, knockdown of miR-142a-5p by antisense morpholino showed normal expression of HSPC marker runx1 (Supplementary Figure S3c), excluding the involvement of miR-142a-5p in HSPC regulation. Therefore, we concluded that the defects of HSPCs in 142 T − / − embryos were only attributed to the absence of miR-142a-3p.
To demonstrate whether the defect of HSPCs in 142 T − / − can recover during adulthood, the kidney of adult mutant was sectioned and stained with hematoxylin and eosin to test the number of hematopoietic cells. It was observed that both the cellularity and the volume of the pronephros decreased dramatically in 12-week-old 142 T − / − adult fish (Figure 2g), confirming the continuous defect of definitive hematopoiesis in adult zebrafish upon miR-142a deletion. To test whether there exists hypermaturation of neutrophils in 142 T − / − as reported previous [24], we also sorted myeloid cells using flow cytometry (FACS) and conducted Wright-Giemsa staining of eosinophils, macrophages and neutrophils. However, our analysis showed that the morphology of myeloid cells in between mutants and WT siblings seems no difference (Supplementary Figure S4), suggesting normal maturation of neutrophils in 142 T − / − embryos.
Previous work showed that irf7 is a direct target of miR-142a-3p, and miR-142a-3p morphants display miR-142-3p in hematopoiesis increased expression of irf7 [26]. Using WISH and qPCR, we were able to detect that there was an ectopic expression of irf7 in 142 T − / − embryos (Figure 3a and b). Moreover, knockdown of irf7 in 142 T − / − embryos displayed a partial rescue on HSPCs at both 26 and 36 hpf (Figure 3c), confirming that irf7 functions downstream of miR-142a-3p to mediate HSPC development.
Taken together, miR-142a-3p is indispensable for the HSPC emergence and differentiation, and its direct target irf7 is upregulated in 142 T − / − embryos, confirming the findings in our previous study [26].   Figure S5a). Therefore, cells in this region including HSPCs underwent excessive apoptosis in both types of miR-142a mutants.
On the basis of the decrease in proliferation and the increase in apoptosis of HSPCs in 142 T − / − embryos, we referred to our previous microarray data that contained upregulated genes in miR-142a-3p morphants [26] and found that apoptotic genes including p53 were greatly upregulated (Supplementary Figure S5b). Because of the dual roles of p53 in both cell cycle arrest and apoptosis and the hint from the microarray, we next focused on the relationship between miR-142a-3p and p53.

T − / − embryos display ectopic expression of p53
To elucidate whether the expression of p53 in miR-142a mutants is consistent with the microarray data in morphants, we first examined the expression of p53 and its downstream target p21. qPCR showed that the expression levels of p53 and p21 were increased at 26 and 36 hpf (Figure 5a). To find out whether p53 is increased in HSPCs, frozen section of 142 T − / − embryos stained with p53 probe was performed and the result displayed that p53 was ectopically expressed especially in the dorsal aortal region (Figure 5b). To further support this notion, we sorted cmyb-positive cells from the trunk region of 142 T − / − embryos at 36hpf for qPCR analysis. Result showed that expression of p53 was increased in cmyb-positive cells in 142 T − / − (Figure 5c). Taken together, we concluded that p53 was specifically upregulated in HSPCs of 142 T − / − embryos.

MiR-142a-3p inhibits p53 by direct binding
On the basis of several studies revealing that p53 could act as a direct target of microRNAs [41] and the increase of p53 in 142 T − / − embryos, we sought for the possibility that p53 might act as a direct target of miR-142a-3p. Systematic bioinformatic analysis using RNA hybrid website (BiBiServ2) revealed that there is a miR-142a-3p-binding site in the coding sequence (CDS) but not the 3'-UTR of zebrafish p53, suggesting that miR-142a-3p may be a direct regulator of p53.

miR-142-3p in hematopoiesis
To confirm the direct targeting of p53 by miR-142a-3p, we cloned the WT p53-coding region as well as mutated sequences (mismatched with miR-142a-3p seed sequence) containing the predicted miR-142a-3p-binding site into luciferase and GFP reporter vector (Figure 6a, Supplementary Figure S6a and b). Luciferase activity and GFP expression were examined, respectively. Overexpression of the miR-142a-3p duplex in HEK293T cells and zebrafish embryos suppressed the luciferase activity of WT but not the mutated p53-coding region in a dosedependent manner (Figure 6b, Supplementary Figure S6a), indicating that p53 is a direct target of miR-142a-3p. Moreover, injection of GFP reporter fused with the WT p53 but not the mutated p53-coding region together with the miR-142a-3p duplex displayed a marked inhibition of GFP activity (Supplementary Figure S6b). Taken together, these data strongly suggest that p53 acts as a direct target of miR-142a-3p.
To further demonstrate the inhibition of p53 by miR-142a-3p directly, we ectopically expressed the full-length p53 cDNA fused with Myc-tag in the HEK293T cells. The WT or mutated p53 CDS was cotransfected with the miR-142a-3p duplex. Consistent with the reporter assays, overexpression of miR-142a-3p could attenuate Myc-P53 expression in WT p53 but not the mutated p53-transfected cells (Figure 6c and d), further supporting the specific binding of miR-142a-3p to p53.

P53 mediates the decrease in HSPCs upon loss of miR-142a-3p
To determine whether p53 upregulation mediated the HSPC defects in 142 T − / − embryos, we first injected p53 morpholino (MO) into 142 T − / − embryos. WISH and qPCR results showed that p53 MO could partially rescue the decreased expression of runx1 in 142 T − / − embryos at 26 and 36 hpf (Figure 7a and b). Furthermore, we generated a double knockout of miR-142a-3p and p53 by outcrossing adult 142 T − / − with p53 M214K − / − . WISH showed that the expression of runx1 at 36 hpf was markedly restored in 142 T − / − p53 M214K − / − embryos, compared with 142 T − / − p53 M214K+/+ embryos (Figure 7c), consistent with the results by p53 MO knockdown. To determine whether there is a synergy effect between irf7 and p53 in miR-142a-3p-mediated HSPC development, we knocked down both irf7 and p53 in 142 T − / − embryos. Double knockdown of irf7 and p53 fully rescued the HSPC defect (Figure 8a and b), demonstrating that irf7 and p53 act synergistically downstream of miR-142a-3p in HSPC development. In addition, the expression of p53 and irf7 was examined in miR-142a mutants injected with irf7 MO or p53 MO, respectively. qPCR showed that knockdown of irf7 or p53 had no effect on the expression of each other (Figure 8c and d), revealing that irf7 and p53 are parallel targets of miR-142a-3p.
In order to test that p53 inhibition specifically restored apoptosis in the AGM region, we injected p53 MO into 142 T − / − embryos. As expected, p53 MO injection led to alleviated apoptosis in both the trunk and AGM region of 142 T − / − embryos (Supplementary Figure S7a), strengthening the cause-consequence relationship between p53-induced apoptosis and HSPC defects in miR-142a-3p mutant.
In addition, we also examined the expression of differentiated hematopoietic markers in p53 MOinjected 142 T − / − mutants. As shown in Figure S7b, the defects of differentiated populations including erythrocytes, neutrophils and T cells in 142 T − / − embryos were also partially restored by p53 knockdown, supporting that p53 functions downstream of miR-142a-3p in HSPC differentiation.

Discussion
In this work, by generating two types of miR-142a-3p mutants using both TALENs and CRISPR/Cas9 methods, we demonstrate that miR-142a-3p has an essential role in definitive hematopoiesis. Decreased HSPCs and their derivatives, as well as ectopic expression of p53 and irf7, were observed in miR-142a-3p mutants. Loss of p53 could rescue the defects of HSPCs in miR-142a-3p mutants and direct binding assay together support that p53 is a direct target of miR-142a-3p, and the newly identified miR-142a-3p-p53 axis has a critical role in HSPC survival and thus protects emerging HSPCs and their derivatives from apoptosis. In addition, p53 acts synergistically with the previously reported irf7 downstream of miR-142a-3p in controlling HSPC development [26] (Supplementary Figure S8).
Consistent with our previous finding that miR-142a-3p is highly expressed in the runx1 + /cmyb + cells in the trunk region that labels HSPC populations in zebrafish [26], two independent groups also showed evidence that the counterpart of miR-142a-3p is enriched in HSPC population in Xenopus and mice [22,23]. Studies in Xenopus demonstrated that miR-142a-3p targets TGFβ and functions as a master regulator in HSPC lineage specification and thus sits at the apex of the hierarchy programming definitive hemangioblast [23]. This is consistent with our work and illustrates the crucial function of miR-142a-3p in definitive hematopoiesis across vertebrates.
However, there are also different views on the function of miR-142a-3p in HSPCs based on morpholino or duplex injection [27,40]. It was reported that ectopic expression of the miR-142-3p duplex (concentration of 20 μmol l −1 ) suppresses the primitive erythrocyte progenitors (gata1 + ) and HSC formation (cmyb + ) [27]. We also performed miR-142a-3p duplex injection and found that the dose of duplex they used caused nonspecific side effects in embryos, such as heart edema, whereas lower dose (concentration of 5-0 μmol l −1 ) showed increased expression of HSPC marker runx1 at 26 hpf as well as cmyb and rag1 at 5 dpf. To exclude the delusive dose effect of morpholino or duplex, in this work, we chose to generate miR-142a genetic mutants for functional investigation.
Recently, by targeting two miR-142 genes including miR-142a and miR-142b, the mutants of miR-142 by miR-142-3p in hematopoiesis zinc-finger nucleases (ZFN) method are generated, and these miR-142-3p mutants display neutrophil defects but with relatively normal HSPCs [24]. Consistently, our 142 T − / − embryos also displayed reduced number of neutrophils (marked by lyz) by WISH at 4 dpf ( Figure 2f). Furthermore, the kidney, which contains more neutrophils compared with other organs [42], was shrunken in 142 T − / − adult fish (Figure 2g). Notably, hematoxylin and eosin staining result revealed that the cellular density of neutrophils was decreased during the loss of miR-142. According to the Giemsa staining, the maturation of neutrophil was normal, as different subtypes of neutrophils with various morphologies were all observed (Supplementary Figure S4). This result is discrepant with the previously reported neutrophil phenotypes in miR-142a/b mutants [24]. On the basis of the deficiency of miR-142a and miR-142b genes, there is a possibility that the reported neutrophil hypermaturation in miR-142a/b mutants might be due to the deletion of miR-142b. Moreover, neutrophils at different maturation stages can be observed in WT and 142 T − / − embryos; therefore, the ratio of naive versus mature neutrophils varies with development.
As a tumor suppressor, p53 is also a critical transcription factor in hematopoietic cells and it is involved in the quiescence, self-renewal, senescence and apoptosis of HSPCs [32,33]. It also modulates the cell cycle arrest of HSPCs through p21 [32]. Notably, mounting evidence suggests that HSPCs are highly sensitive to cellular stress, cell cycling and the microenvironment, which are closely related to the function of p53 [43,44]. In zebrafish, mutation in cleavage and polyadenylation specificity factor 1 (cpsf1) displays definitive HSC defects due to a p53-dependent apoptotic cell death [45]. In mouse, the depletion of HSPCs in mutants of Mysm1 and Pot1b could be restored by loss of the P53 [38,46]. These findings support that HSPC survival is tightly controlled by various factors and that p53mediated apoptosis is the key to maintain the survival of HSPCs. Our work here demonstrates that in zebrafish embryogenesis, the activation of P53 in the AGM region leads to excessive apoptosis, which impairs the survival of the earliest specified HSPCs. Consistent with our findings, super-p53 mice, carrying one extra gene dose of p53, results in significant acquisition of repopulating capacity but decreased survival of HSCs Xinyan Lu et al. 9 in recipients [47]. Direct and indirect evidence sheds light upon the hypothesis that the dosage of p53 should be strictly controlled to maintain the normal HSPC survival. Although several studies have identified the downstream targets of p53 in regulating HSPCs, the direct upstream regulators of p53 in modulating definitive hematopoiesis are not well defined. Here, we show that miR-142a-3p directly binds to the coding region of p53 and blocks its function, and then reveals the crucial role of the miR-142a-3p-p53 axis in HSPC survival.
In our previous study, irf7 knockdown only partially rescued the HSPC defects in miR-142a-3p morphants [26], suggesting that other targets of miR-142a-3p may be involved in this process. In fact, double knockdown of irf7 and p53 could fully rescued the HSPC defects and the rescue effect was better compared with irf7 knockdown, demonstrating that there is a synergistic effect of both irf7 and p53 downstream of miR-142a-3p in HSPC development. Tight control of HSPC survival is crucial for HSPC development at early stages in vertebrates. It has been reported that scl-a-deficient zebrafish embryos display more apoptosis and failure to maintain the survival of nascent HSPCs in the AGM [48]. In mice, knockout of gata2 in hematopoietic cells using Vav-Cre shows increased apoptosis and attenuated survival of HSPCs [49]. Interestingly, a recent report also showed that miR-142 − / − CD4 + DCs underwent excessive apoptosis, compared with their littermate counterparts [50]. Although developmental defects in HSPCs were not examined in miR-142 − / − mice, we speculate that a similar HSPC phenotype would be expected at the onset of definitive hematopoiesis. Recently, inhibition of p53 and its downstream targets p21 or miR-34a has been demonstrated to enhance the reprogramming efficiency of induced pluripotent stem (iPS) cells generation [51,52]. Moreover, repression of p53 can efficiently induce CD34 + miR-142-3p in hematopoiesis cord blood cells to iPSCs [53]. These lines of evidence in stem cells give clues that excessive level of p53 might dampen the emergence of stem cells and miR-142a-3p could protect the nascent emerging HSPCs from unwanted apoptosis, and this protection is critical for HSPC development during embryogenesis.
In summary, our studies demonstrate that miR-142a-3p directly targets p53 to regulate HSPC survival in zebrafish embryos. In addition to our previous report in which irf7 is a direct target of miR-142a-3p in controlling HSPC formation and differentiation, here we discover that p53 acts downstream of miR-142a-3p to control the survival of HSPCs.

Materials and Methods
Generation of miR-142a mutants and RNA secondary structure prediction Zebrafish miR-142a mutants were generated by TALEN and CRISPR/Cas9, as previously described [54,55]. The injection dose of miR-142a TALEN mRNA was 250 pg per embryo. The injection doses of gRNA and cas9 mRNA were 100 and 500 pg per embryo, respectively. The TALEN-generated miR-142a mutants (142 T − / − ) had a 13-base-pair deletion and could be verified by sequencing (Supplementary Table S1). The CRISPR/Cas9-generated miR-142a mutants (142 C − / − ) caused a 986-base-pair deletion, and the structure of mature miR-142a-3p in both mutants was disrupted. The 142 C − / − could be genotyped by a combination of different lengths of PCR products (Supplementary Table S2). The microRNA secondary structure was predicted by the RNA Structure software (version 4.6).

Morpholinos and microinjection
Antisense morpholinos were ordered from Gene Tools including p53 MO and irf7 MO. The p53 MO: 5′-GCG CCATTGCTTTGCAAGAATTG-3′ was injected at the dose of 4 ng per embryo and the irf7 MO was conducted as previously described [26].

Quantitative RT-PCR and western blot analyses
Trunk regions of embryos were dissected for qPCR and western blot analyses. Quantitative RT-PCR and western blot analyses were performed as described previously [57]. Antibody used for western blot analysis included anti-Runx1(1:200, AnaSpec, Fermont, CA, USA) and anti-Myc (1:500, Cell Signaling, Danvers, MA, USA). Anti-Myc antibody was used as a tag to test the binding affinity of miR-142a-3p to p53.

Reporter assay
The WT and mutated p53-coding regions were cloned and then inserted into the EcoRI and SpeI sites of the pGL3luciferase vector, which was kindly provided by Dr Peifeng Li (Institute of Zoology, CAS, Beijing, China). The WT p53coding region was mutated from 'CACUGC' to 'GUGAUG'. Wild-type or mutated pGL3-p53 CDS co-transfected with negative control or the miR-142-3p duplex (Invitrogen, Shanghai, China) were transfected into HEK293T cells using Lipofectamine2000 (Invitrogen). The luciferase activity was measured using the luciferase activity assay kit (Promega). Luciferase and GFP reporter assay in zebrafish was conducted as described previously [24,58].

Immunostaining, BrdU incorporation and TUNEL assay and section
Live zebrafish embryos injected with BrdU were fixed in 4% paraformaldehyde (PFA) overnight and then replaced with methanol at − 20°C for 2 hpf. Embryos were permeabilized with proteinase K, refixed in 4% PFA for 20 min and then rinsed in PBST, followed by incubation in 2 N HCl for 1 h. Then, embryos were blocked for 1 h in blocking solution and incubated with primary anti-BrdU antibody (Roche, Mannheim, Germany) and secondary Alexa Fluor 488-conjugated anti-mouse antibody (Invitrogen). The BrdU-labeled embryos were treated with 30% sucrose and then washed three times using PBST and embedded in O.C.T. medium (SAKURA, Torrance, CA, USA). The embryos were sectioned using LEICACM1900 cryostats. Detection of both cmyb RNA and mitosis marker pH3 and BrdU simultaneously was conducted as described previously [59].
Embryos at 24 hpf in methanol were washed with PBST and then permeabilized with proteinase K. After being washed with PBST, embryos were subjected to TUNEL assay according to the instruction of the TUNEL staining kit (Roche) at 4°C overnight.

Flow cytometry of transgenic fish and whole-kidney marrow
Flow cytometry analysis and cell sorting were based on the forward scatter and side scatter as described previously [42]. All experiments were conducted using the FACSCalibur flow cytometer (BD, San Jose, CA, USA). Data was analyzed using the FlowJo software Version 7.6.1.

Hematoxylin and eosin and Wright-Giemsa
Hematoxylin and eosin staining was performed on the serial sections of larval zebrafish. Wright-Giemsa staining was performed as described previously [42].

Microscopy
Embryos were observed using a Nikon microscope for digital image capture. Confocal images of live cmyb:EGFP embryos, TUNEL and BrdU assays were taken on confocal microscope (Nikon A1 laser microscope, Nikon A1, Melville, NY, USA). All images were assembled using Photoshop (Adobe Systems, Mountain View, CA, USA).

Statistical analysis
Values are presented as mean ± s.d. Student's t-test was used to compare the means of different groups and P-values o 0.05 were considered to be significant (*Po 0.05, **Po 0.01).