The miR-144/451 locus is required for erythroid homeostasis

Recent studies revealed an essential role for Ago2 and the miRNA pathway during mammalian erythroid differentiation (O’Carroll et al., 2007). To identify miRNA loci that contribute to terminal erythropoiesis we used the following set of criteria: expression of the candidate miRNAs should be (a) down-regulated substantially in the absence of Ago2, (b) abundant in erythroblasts, and (c) restricted to the erythroid lineage. Expression profiling revealed that miR-451 was abundantly expressed, but completely absent in Ago2^−/− erythroblasts (Fig. 1, A and B). This is in stark contrast to other miRNAs that were down-regulated but still expressed. The gene encoding miR-451 is located on chromosome 11, and encompasses a noncoding transcriptional unit that also encodes miR-144 that resides 5′ of miR-451 (Fig. S1 A). Mir-144 was also abundantly expressed in erythroblasts and markedly down-regulated in the absence of Ago2 (Fig. 1, A and B). Expression analysis of both miR-451 and miR-144 within lineage-restricted lymphocytes, granulocytes/monocytes, and erythroid cells revealed that the locus was selectively expressed within the erythroid lineage (Fig. 1 C; Lu et al., 2008). The expression of miR-451 was restricted to the fetal liver in embryonic day (E) 16.5 mouse embryos, the major site of hematopoiesis and erythropoiesis at this stage of development (Fig. 1 D). Furthermore, both miR-451 and miR-144 were expressed in fetal liver–derived erythroid cultures under media conditions that promote proliferation or terminal differentiation (Fig. 1 E). Indeed, consistent with our identification of miR-144/451 as putative miRNAs required for terminal erythropoiesis, the miR-144/451 locus was recently identified as a target gene of the essential erythroid transcription factor Gata1 (Dore et al., 2008).

MiR-451 and miR-144 met all our criteria for miRNAs with a key function during mammalian terminal erythropoiesis; therefore, we proceeded to generate a mouse miR-144/451 cluster-null allele (miR-144/451⁻) by deleter-Cre–mediated (Schwenk et al., 1995) excision of the loxP-flanked miR-144/451 allele (Fig. S1 B). MiR-144/451^−/− mice were born in Mendelian numbers, survived throughout adulthood without increased rates of mortality, and showed no developmental defects in the lymphocyte or granulocyte-monocyte lineages (unpublished data). However, analysis of adult hematopoiesis in miR-144/451^−/− mice revealed defects during terminal erythropoiesis. Peripheral blood showed a mild anemia with reduced red cell numbers and hematocrit and hemoglobin values, as well as increased red cell distribution width, indicating anisocytosis (Fig. 2 A and Fig. S2 A). The miR-144/451^−/− mice also displayed splenomegaly, indicating compensation of anemia by stress erythropoiesis (Fig. 2 B). The anemia and stress erythropoiesis was confirmed by elevated erythropoietin concentrations (Fig. 2 C) and increased numbers of splenic c-kit–positive erythroid precursors in miR-144/451^−/− mice (Fig. S2, B and C). Analysis of erythroid development revealed ineffective erythropoiesis in the bone marrow of miR-144/451^−/− mice characterized by the normal presence of early basophilic erythroblasts (Ter119^hi CD71^hi), but a marked decrease in the frequency and number of mature orthochromatophilic (Ter119^hi, CD71^neg) erythroblasts (Fig. 2, D and E). In the mouse, inefficient erythropoiesis in the bone marrow can be compensated by stress erythropoiesis in the spleen (Broudy et al., 1996; Bauer et al., 1999). Analysis of splenic erythroblast populations in miR-144/451^−/− mice revealed an ∼10-fold increase in the absolute numbers of early erythroblasts (Fig. 2 E), but normal or slightly elevated numbers of mature orthochromatophilic erythroblasts, indicating a major loss of cells during late erythroid maturation. Thus, miR144-451 deficiency impairs late erythroid differentiation, which is partly compensated by expansion of the early erythroid compartment in the spleen (Fig. 2, D and E, and Fig. S2 D). A function for miR-144/451 in both zebrafish and mouse erythropoiesis has been reported. However, data from zebrafish were conflicting as to the phenotypic consequences of miR-451 inhibition (Dore et al., 2008; Du et al., 2009; Pase et al., 2009). In the mouse, contrary to our observations, it was reported that inhibition of miR-144/451 impaired the proerythroblast to basophilic–erythroblast transition (Papapetrou et al., 2009). This discrepancy likely arises from the lentiviral decoy method of miRNA inhibition used.

The mild anemia in resting miR-144/451^−/− mice may underrepresent the importance of the miR-144/451 locus to the hematopoietic homeostasis under situations of stress. We therefore challenged wild-type and miR-144/451^−/− mice with phenylhydrazine, a chemical that induces hemolytic anemia, and assayed the potential of the mice to recover. Over half of the miR-144/451^−/− mice died within 6 d after the administration of phenylhydrazine, whereas all wild-type littermates made a full recovery (Fig. 2 F). These results indicate that the miR-144/451 locus is critical for a robust hematopoietic system.

Although unlikely, one could argue that the defects observed in the miR-144/451^−/− mice are not hematopoietic autonomous and that they do not represent a developmental defect, but are a consequence of the response to a hemolytic anemia. The hematopoietic-intrinsic nature of the developmental defects in miR-144/451^−/− mice was revealed through the analysis of lethally irradiated mice reconstituted with miR-144/451^−/− bone marrow. These reconstituted mice present the same defects as germline-deficient miR-144/451^−/− mice (Fig. S3). We next used an in vitro fetal liver–derived erythroid cell culture and differentiation system (Dolznig et al., 2001; von Lindern et al., 2001) to investigate the intrinsic nature of the developmental defect. Fetal livers from wild-type and miR-144/451^−/− E12.5 embryos were placed in culture with a cocktail of defined cytokines that promote the selective expansion of erythroid precursors. Under these conditions miR-144/451^−/− erythroid precursors proliferated at the same rate as the respective wild-type controls (Fig. 3 A). After 3 d of expansion, the cultures were switched to differentiation medium. MiR-144/451^−/− cultures normally underwent the final rounds of proliferation before entering terminal erythropoiesis (Fig. 3 B), as well as hemoglobinization (Fig. S2 E). In addition, no differences in the fraction of apoptotic cells were detected in the absence of miR-144/451 (Fig. 3 C). However, size reduction and enucleation of miR-144/451^−/− cultures was inefficient (Fig. 3, D and E, and Fig. S2 E). MiR-144/451 deficiency did not block any one process, per se, but rendered terminal erythroid differentiation inefficient. Thus, the in vitro system recapitulated the defective erythropoiesis observed in vivo in miR-144/451^−/− mice. These observations prove that the miR-144/451 cluster is not required for expansion of early erythroblasts, but is intrinsically required for terminal erythroid development.

We next wanted to explore the extent and nature of the deregulated genes in the absence of miR-144/451 during terminal erythropoiesis. To this end, we decided to use the in vitro culture model, as it presented advantages over ex vivo isolated erythroblasts. The miR-144/451^−/− erythroblasts in vivo are subjected to, among other things, altered cytokine levels and microenvironments. We detected a significant deregulation of gene expression in miR144/451^−/− erythroblasts (Fig. 4 A). Transcriptome analysis from cells with either loss or gain of function for a given miRNA has been used to reliably identify miRNA targets because of transcript destabilization after RISC binding (Lim et al., 2005; Giraldez et al., 2006; Baek et al., 2008; Selbach et al., 2008; Lewis et al., 2009). To identify miR-144/451 target genes, we therefore used the Sylamer algorithm (van Dongen et al., 2008) to analyze the gene expression profiles derived from wild-type and miR-144/451^−/− erythroid precursors. Sylamer searches for significant enrichment of 6-8 nt motifs across an ordered gene list. For miRNA analysis, the identification of miRNA seed match enrichment in the 3′ UTRs of deregulated genes from genome-wide experiments indicates direct miRNA involvement (van Dongen et al., 2008).

Sylamer analysis using all 7-nt motifs complementary to seed regions of mouse miRNAs revealed a highly significant (P < 8 × 10⁻¹³; Bonferroni corrected < 7 × 10⁻¹⁰) enrichment for the 7-nt seed match of miR-451 (AACGGTT) in the 3′ UTRs of genes in a cohort of 588 genes, most of which were up-regulated in miR144/451^−/− erythroid precursors (Fig. 4 B). Enrichment of the seed sequence of miR-144 (ATACTGT) was observed, but was less significant (P < 4 × 10⁻⁶, Bonferroni corrected < 4 × 10⁻³). Genes with 6-nt seed matches for miR-144 were far less affected than those containing miR-451 sites (Fig. 4 C). However, genes containing 6-nt seed matches for both miR-144 and miR-451 displayed a slight additive effect and were more deregulated than genes with miR-451 seed matches alone. The deregulation of a representative panel of the Sylamer-identified miR-144/451 seed-containing genes was validated by quantitative RT-PCR (qRT-PCR; Fig. 4 D). Of the 588 up-regulated genes identified by Sylamer analysis in the miR-144/451^−/− erythroblasts, 98 had one or more 6-nt seed matches for miR-451, 226 had seed matches for miR-144, and 57 of these contained seed matches for both (Table S1). Many of these genes have reported roles in terminal erythropoiesis (Table S1), but the majority are loci whose precise expression are only now linked to normal erythroid development (Table S1). This analysis is not suggestive of a single, or few critical miR-144 or miR-451 targets, but rather that the miR-144/451 locus regulates many genes whose collective deregulation derails normal terminal erythropoiesis. We did not find evidence of lineage-inappropriate gene expression, excluding the possibility that the miR-144/451 cluster functions to maintain lineage identity. Together, these data suggest that the miR-144/451 locus functions as a tuner of gene expression, analogous to the function of miR-223 in neutrophils (Baek et al., 2008).

Many of our observations suggest that within the miR-144/451 cluster, miR-451 is likely to play the pivotal role in terminal erythropoiesis. First, it displays a far higher expression level than miR-144. Second, our bioinformatics analyses showed a greater statistical enrichment for the miR-451 seed in the up-regulated cohort of genes in the miR-144/451 mutants. Third, miR-451 exerted a greater impact on the level of target gene expression than miR-144. To explore the possibility that miR-451 plays a key role in regulating terminal erythropoiesis, we generated and analyzed miR-451–null (mir-451^−/−) mice (Fig. S1 C). The phenotype of miR-451^−/− mice was largely indistinguishable from that of miR-144/451^−/− mice (Fig. 5, A–D). Together, these results indicate that deregulation of miR-451 targets is sufficient to impair erythroid development.

In this study, we show that miR-451 expression is strictly dependent on Ago2 function. Indeed, while this manuscript was under review it was shown that biogenesis of miR-451 is thus far unique in that it is Dicer independent but requires the endonuclease activity of Ago2 (Cheloufi et al., 2010; Cifuentes et al., 2010). The phenotype of miR-144/451^−/− and miR-451^−/− mice is milder than that of Ago2^−/− mice (O’Carroll et al., 2007), indicating that additional miRNA loci regulate terminal erythropoiesis. In addition, the viability of miR-451^−/− mice suggests that the perinatal lethality in homozygous endonuclease-dead Ago2^ADH mutant mice cannot be accounted for solely by the loss of miR-451 expression. Analogous to the granulocytic and the lymphoid lineages, we now show that a miRNA-encoding locus regulates erythroid development in vivo. We demonstrate that loss of miR-144/451 or miR-451 does not block any one process, but renders terminal erythroid differentiation inefficient. We show that the miR-144/451 cluster tunes the expression of many genes. Gata1 confers the erythroid lineage-specific expression of miR-144/451 (Dore et al., 2008). Therefore, Gata1 not only ignites the program of terminal erythroid differentiation, but also ensures that gene expression is tuned through the activation of the miR-144/451 locus.

RNA was isolated using Trizol (Invitrogen) according to manufacturer’s instructions. 500 ng of total RNA was labeled and hybridized to miRCURY LNA microRNA arrays V.11 (Exiqon) for miRNA profiling. Array Express accession no. E-TABM-996. qRT-PCR were performed using TaqMan miRNA assays (Applied Biosystems) for miR-144 and miR-451 and snRNA U6, respectively. In situ hybridizations were performed on cryosections with specific LNA probes (Exiqon).

The MiR-144/451 targeting strategy allows Cre-mediated deletion of transcription start site (TSS) for the miR-144/451 cluster, as well as the hairpin that encodes miR-144. This deletion simultaneously deletes miR-144 and transcription initiation of the cluster, leading to the loss of both miR-144 and miR-451 expression. The miR-144/451 gene-targeting vector (pDTA-miR-144/451) carries a neomycin resistance (neo^r) gene flanked with two frt sites and a loxP site (neo^r-frt2-loxP cassette) located 3′ of the TSS and a second loxP site between miR-144 and miR-451.

The MiR-451 targeting strategy allows Cre-mediated deletion of the hairpin that encodes miR-451. This deletion leaves the TSS and miR-144 transcription intact, resulting in the loss of miR-451 expression. The miR-451 gene-targeting vector (pDTA-miR-451) carries a loxP site between miR-144 and miR-451 and a neomycin resistance (neo^r) gene flanked with two frt sites and a loxP site (neo^r-frt2-loxP cassette) located 3′ of the hairpin that encodes miR-451.

The targeting constructs were transfected into IB10 ES cells. Southern blotting of the individual ES cell clones was used to identify homologous recombinants. Targeted ES cells were used to generate mice heterozygous for the miR-144/451– and miR-451–targeted alleles. These mice were then crossed to the FLP-expressing transgenic mice (FLPeR) to remove the frt flanked neo^r cassette, and then to Del-Cre mice. These genetic manipulations resulted in the generation of miR-144/451 and miR-451 loxP alleles, and miR-144/451 and miR-451 null alleles, respectively. All analysis was performed on mice that did not carry the FLPeR or Del-Cre alleles.

All of the mice were bred and maintained in EMBL Mouse Biology Unit, Monterotondo, in accordance with Italian legislation (Art. 9, 27. Jan 1992, n^o116) under license from the Italian Ministry of Health.

Isolation of spleen and bone marrow, and subsequent FACS analysis were performed as previously described (Socolovsky et al., 2001; Pronk et al., 2007). Blood parameters were analyzed using flow cytometry-based hematology (Hemavet HV950FS; Drew Scientific). The analysis of plasma EPO levels were performed using the Quantikine Mouse/Rat immunoassay (R&D Systems).

Femur and spleen isolated from adult wild-type and miR-144/451^−/− mice was fixed in paraformaldehyde and embedded in paraffin. Hematoxylin and eosin (H&E) staining was performed on 8 μm sections. Images were acquired on a microscope (DM6000B; Leica) with 10×, 0.25 numerical aperture dry lens, and 63×, 1.3 numerical aperture oil immersion lens fitted with a DFC290 camera and processed with Adobe Photoshop (Adobe Systems).

Adult mice were sacrificed and bone marrow cells isolated and transferred (2 × 10⁶ cells per recipient) into lethally irradiated (875R) C57BL/6 mice. Bone marrow–reconstituted mice were maintained on medicated water (Baytril; Bayer) for 4 wk after reconstitution and analyzed 3 mo after transfer.

Adult male mice were administered with phenylhydrazine at a dose of 48 mg/kg by intraperitoneal injection on two successive days, as previously described (Gutiérrez et al., 2008).

The in vitro expansion and differentiation of erythroid cells from E12.5 fetal livers was performed as previously described (Dolznig et al., 2001; von Lindern et al., 2001). Cell size reduction during differentiation was indirectly assayed by FSC-A using FACS analysis. To assess morphology, cells were spun onto glass slides, air dried, and stained with o-dianisidine (Neutral Benzidine; Sigma-Aldrich) in 1% methanol and Diff Quick staining set (Medion Diagnostics AG). Images were acquired with a microscope at 63×, 1.30 numerical aperture oil immersion lens fitted with a DFC290 camera and processed with Adobe Photoshop. The appearance of nascent reticulocytes was enumerated using a modified FACS protocol as described (Kingsley et al., 2006). Instead of Hoechst 33342, cells were stained for 30 min at 37°C with 20 µM Vybrant Violet (Invitrogen). CFSE labeling and dilution was used to monitor the number of cell divisions upon terminal differentiation as described (Gutiérrez et al., 2008). Annexin V FACS staining (Annexin V apoptosis detection kit; eBioscience) was used to enumerate the number of apoptotic cells in the cultures.

Live proliferating erythroblasts (FSC-A^hi) from eight independent erythroid cultures (four wild-type and four miR-144/451^−/−) was sorted by FACS to obtain highly homogeneous cell populations for gene expression profiling. Total RNA was isolated using Trizol (Invitrogen) according to the manufacturer’s instructions and profiled using Illumina mouse WG-6 v2.0 Bead arrays. Probe level data from all eight arrays were obtained from the Illumina GenomeStudio software and processed using R/BioConductor using a variance stabilizing transformation and robust spline normalization as implemented in the lumi package (Du et al., 2008). Differential expression between wild-type and miR-144/451^−/− samples was assessed by fitting linear models and using an empirical Bayes approach to shrink the estimated variance, by means of the limma package (Smyth, 2004). Probe annotations and mappings were obtained from Illumina (MouseWG-6_V2_0_R2_11278593_A). The gene expression data are deposited in Array Express under the accession no. E-MTAB-277.

Total RNA was reverse transcribed by Superscript II reverse transcription (Invitrogen) using random hexamer primers. Quantitative PCR was performed on a LightCycler 480 PCR instrument (Roche) using 2x SYBR green I master (Roche). The expression data were normalized to Hprt and Pgk1 reference genes and analyzed by REST software (Pfaffl et al., 2002) using pairwise fixed reallocation and randomization test to estimate relative normalized fold change between wildtype and miR-144/451^−/− erythroblast samples.

The gene-expression data for each probe were sorted according to t-statistic from most significantly up-regulated to most significantly down-regulated in the knockout samples. Each probe ID was mapped to a mouse transcript, and then to a 3′ UTR sequence. The 3′ UTR sequences were obtained by mapping probes to transcript sequences from Vega, Ensembl, or RefSeq, in that order of preference. An ordered list of 19,600 unique transcripts was produced in this fashion, each with a corresponding 3′ UTR sequence. Sylamer was run on this dataset with Markov correction set to filter trinucleotide biases and analyzing enrichment of 7-nt motifs.

Fig. S1 illustrates and validates the generation of the miR-144/451 and miR-451 null alleles, respectively. Fig. S2 depicts histological analysis of peripheral blood, BM, and spleen of miR-144/451–deficient mice, as well as FACS analysis enumerating erythroid precursors. Morphological analysis of in vitro erythroid cultures is also shown. Fig. S3 contains the analysis of lethally irradiated mice reconstituted with miR-144/451^−/− bone marrow. Table S1 lists the Sylamer-identified target genes of miR-144/451.

We gratefully acknowledge the support E. Perlas and V.Benes of EMBL’s Histology and Genecore core facilities, respectively. We also acknowledge the services of Jeanette Rientjes from Monash University’s Gene Recombineering Facility.

The authors have no conflicting financial interests.