NF90 Regulates Cell Cycle Exit and Terminal Myogenic Differentiation by Direct Binding to the 3′-Untranslated Region of MyoD and p21WAF1/CIP1 mRNAs*

NF90 and splice variant NF110/ILF3/NFAR are double-stranded RNA-binding proteins that regulate gene expression. Mice with targeted disruption of NF90 were engineered. NF90(–/–) mice were born small and weak and succumbed to perinatal death within 12 h because of neuromuscular respiratory failure. Lung inflation and morphology were normal in NF90(–/–) mice. The diaphragm and other skeletal muscles in NF90(–/–) mice demonstrated disorganized arrangement and paucity of myofibers, evidence of myocyte degeneration and increased apoptosis. The expression of myogenic regulators, MyoD, myogenin, and p21WAF1/CIP1, was severely decreased in NF90(–/–) mice. These myogenic transcription factors and cell cycle inhibitors are regulated in part through post-transcriptional mRNA stabilization. Northwestern blotting revealed that NF90 is the principal and specific p21WAF1/CIP1 and MyoD 3′-untranslated region RNA-binding protein in developing skeletal muscles. NF90 regulates transcription factors and a cell cycle inhibitor essential for skeletal muscle differentiation and for survival.

The Xenopus homolog of NF90 and NF110 is termed CCAATbox transcription factor (CBTF) and regulates the transcriptional activation of the hematopoietic regulatory factor, GATA-2 (19). In early development, CBTF is retained in the cytoplasm through interactions with maternal RNA. At the midblastula transition, degradation of maternal RNA releases CBTF for translocation into the nucleus, followed 4 h later by development of CCATT DNA binding activity and transcriptional activation of GATA-2 (20). The dsRBMs of CBTF are capable of binding both DNA and RNA (21), and intact dsRBMs are required for transcriptional control (15).
In addition to roles in transcriptional regulation, NF90 contributes to RNA splicing as a component of the spliceosome (11,22). NF90 specifically binds to AU-rich elements in the 3Ј-UTR of IL-2 mRNA, stabilizing IL-2 mRNA against rapid degradation, and contributing to the up-regulation of IL-2 gene expression in activated T-cells (23). NF90 mediates nuclear export of IL-2 mRNA to the cytoplasm through the interaction of the NF90 nuclear export signal with the nuclear export karyopherin, exportin-5 (24). Redistribution of NF90/MPP4 from the nucleus to the cytoplasm during mitosis is associated with increased phosphorylation (6,13).
In the cytoplasm, NF90 is associated both with translationally quiescent ribonucleoprotein (RNP) complexes and with ribosomes (25). Like many other mRNP components, NF90 is an autoantigen in human and experimental systemic lupus erythematosus (26). DsRBM proteins such as NF90/ILF3 are highly expressed in the testis, where they contribute to gene regulation by repressing the translation of nascent mRNAs (8). The mRNA encoding acid ␤-glucosidase is specifically translationally repressed by NF90/TCP80 (9). A search for other liver mRNAs specifically bound and translationally repressed by NF90/TCP80 identified aldolase B, complement protein 8 and fibronectin receptor ␤1 (9). The prototypic dsRBM protein, staufen, contributes to localized translation of specific mRNAs in mammalian neurons (27). Tau is a microtubule-associated protein that is preferentially localized in axons, and abnormal accumulations of tau form neurofibrillary tangles that contribute to the pathology of Alzheimer dementia (28). The 3Ј-UTR of tau mRNA contains AU-rich elements that contribute to axonal targeting; the axonal targeting element is specifically bound by NF90 and NF110/ILF3 (29). The interactions between cytoplasmic NF90/TCP80, translation elongation initiation factors ␣, ␤, and ␥ (14) and PKR (10,11,25,30) are likely to regulate translation of specifically bound mRNAs.
NF90 has been implicated in host antiviral responses (25, 30) (31, 32). Virally encoded structured RNAs interact specifically with NF90 dsRBMs, including the HIV TAR loop, the adenovirus VA2 RNA (30), hepatitis B RNA (33), and the 5Ј-and 3Ј-UTRs of the bovine virus diarrhea RNA virus (32). Specific binding of NF90 and NF45 to both the 5Ј-and 3Ј-UTRs of BVDV RNA virus was proposed to regulate the balance between viral replication and translation (32).
To further characterize the role of NF90 in regulation of gene expression and development, we have generated mice with targeted disruption of NF90. Our observation of perinatal lethality associated with muscle weakness and respiratory failure in NF90(Ϫ/Ϫ) mice revealed an unexpected contribution of NF90 to myogenic differentiation. Muscle development is regulated through the coordinated expression of myogenic regulatory factors that include MyoD, Myf5, myogenin, and Mrf4, and cyclin-dependent kinase inhibitors of the p21 WAF1/CIP1 family (34,35). MyoD and Myf5 direct primary myocyte differentiation, myogenin directs secondary differentiation, and p21 WAF1/CIP1 expression promotes cell cycle exit and terminal differentiation. In differentiating myoblasts and regenerating adult muscle, up-regulation of MyoD, myogenin, and p21 WAF1/CIP1 expression occurs concurrently through transcriptional mechanisms and post-transcriptional mRNA stabilization by specific RNA-binding proteins (35,36). Here, we demonstrate that MyoD, myogenin, and p21 WAF1/CIP1 are novel target genes essentially regulated by NF90. Our results led us to propose that the phenotype of muscle weakness and increased apoptosis in NF90(Ϫ/Ϫ) mice is explained by defects in skeletal myocyte differentiation arising from insufficient expression of myogenic regulatory factors.

EXPERIMENTAL PROCEDURES
Genomic Cloning and Targeting Vector Generation-The murine genomic NF90/ILF3 was isolated from a 129SvJ library using the human NF90 cDNA probe (performed at Incyte Genomics). The nucleotide sequence and intron-exon structure of the 29,085-bp gene was determined and deposited in GenBank TM (AF506968).
A targeting vector was designed for constitutive disruption of NF90 exons 2-4 (Fig. 1A). The 5Ј targeting arm was a 6.75-kb BamH1 restriction fragment digested from a genomic NF90 subclone, and the 3Ј targeting arm was 2.8 kb generated with PCR primers that introduced flanking EcoR1 restriction sites (sense primer b3: 5Ј-GGAGTGGGTCT-TCCTGGCCAGTTGCG-3Ј and antisense primer c3: 5Ј-GGCGAAT-TCTCTGCCAGCTATCAAGGA-3Ј). These targeting arms were cloned into pKO Scrambler NKTV-1903 targeting vector, which contains a positive selection neomycin resistance gene and a herpes thymidine kinase negative selection gene (Stratagene). Embryonic stem cells (R1) were transfected by electroporation and selected for resistance to G418 and ganciclovir.
Genotype Analysis by Southern Hybridization and PCR-Genomic DNA was isolated from ES cells or tail biopsy specimens, digested with HindIII (3Ј analysis) or EcoR1 (5Ј analysis), separated on a 0.7% agarose gel, followed by transfer to nitrocellulose membrane. Southern hybridization of the specific wild-type and targeted genomic fragments was performed using randomly primed 32 P-labeled probes outside the targeting arms (Fig. 1A), or an internal probe that hybridized to the neomycin gene, according to standard methods.
Expression Analysis by Northern and Western Hybridizations-Total RNA was extracted from newborn mice using TRIzol reagent (Invitrogen), and 20 g was fractionated in a 1% denaturing agarose gel, then transferred to nitrocellulose. Northern hybridization was performed using a randomly primed 32 P-labeled probe for NF90.
Extracted total proteins (20 g) from newborn mice were fractionated by SDS-PAGE (8% separating gel) and transferred to nitrocellulose membranes. Western immunoblotting was performed using rabbit polyclonal antisera (1:1000 dilution) generated against recombinant NF90 protein (5), followed by secondary goat anti-rabbit antibody (1:3000 dilution), and detection using enhanced chemiluminescence (Amersham Biosciences). Immunoblotting analysis of wild-type murine tissue extracts was performed using a monoclonal antibody that recognizes NF90 (anti-DRBP76, Transduction Laboratories 612154), followed by a goat anti-mouse secondary antibody (Santa Cruz Biotechnology SC-2031).
Histology, Immunohistochemistry, and in Situ TUNEL Analysis-Whole embryos or newborn mice with skin removed were fixed in 10% neutral buffered formalin followed by paraffin-embedding and tissuesectioning at 4-m thickness. Staining with hematoxylin and eosin was performed according to standard procedures.
For immunodetection of NF90, sections were deparaffinized and rehydrated, then boiled for 20 min in 10 mM citric acid in a microwave oven. Blocking of nonspecific mouse antigens was performed by incubation with 3% H 2 O 2 for 5 min followed by incubation with the Vector M.O.M. immunodetection kit (Vectorlabs) for 1 h. Anti-DRBP76 monoclonal antibody specific for NF90/ILF3/DRBP76 (Transduction Laboratories) was incubated at 1:100 dilution for 35 min at room temperature. The signal was amplified by using VECTASTAIN Elite ABC Reagent (Vector Laboratories) and visualized by a horseradish peroxidase reaction using stable diaminobenzidine (Invitrogen). In situ apoptosis was analyzed by TUNEL (TACS 2 TdT, Trevigen).

Generation of Mice with Targeted Disruption of NF90 Gene-
The targeting strategy for NF90 was to disrupt exons 2-4, eliminating the ATG translation initiation codon and introducing a translational frameshift that disrupts the open reading frame (Fig. 1A). The targeting vector was introduced into R1 embryonic stem cells by electroporation and transfected ES cells were subjected to positive and negative selection. 12 of 110 G418/gancyclovir-resistant ES cell clones showed appropriate gene targeting. The targeted allele shows a Southern hybridization signal at 9 kb, in addition to the wild-type allele with a hybridization signal of 7.4 kb (Fig. 1B,  lane 4). Targeted ES cell clones were further analyzed for correct integration using an internal probe for the neomycin resistance gene (data not shown). Three independently derived ES cell lines were injected into blastocyts (C57BL/6), and chimeric mice were derived. Two separate NF90-targeted ES cell lines transmitted through the germline. In mice ex- amined through 18 months of age, NF90(ϩ/Ϫ) heterozygote mice demonstrate no abnormal phenotypes.
Gross Phenotype of NF90(Ϫ/Ϫ) Mice-Following heterozygous intercrossing, homozygote NF90(Ϫ/Ϫ) mice were present in utero at the Mendelian frequency (Table I and Fig. 1C, lanes  4 and 8). At the time of delivery, all NF90(Ϫ/Ϫ) mice were smaller and weighed 35% less than their littermates (Table I and Fig. 1D). Approximately half of the NF90(Ϫ/Ϫ) mice died immediately after birth. The surviving NF90(Ϫ/Ϫ) mice appeared weak and showed almost no spontaneous movements. When lifted by their tails, they emitted a very weak cry compared with their littermates. Over the course of several hours, the NF90(Ϫ/Ϫ) mice became increasingly cyanotic compared with their littermates, and oxygen saturation was noted to be 65% in a cyanotic NF90(Ϫ/Ϫ) mouse, compared with 85-98% in littermates (Fig. 1D). Video analysis of breathing revealed that NF90(Ϫ/Ϫ) mice displayed more rapid and shallow tidal breathing, with no sighs, compared with their heterozygote and wild-type littermates. The NF90(Ϫ/Ϫ) mice never suckled for milk, and were ignored by their mothers. In over 400 live births, no homozygote NF90(Ϫ/Ϫ) mice animals survived beyond 12 h after delivery. Thus, the NF90 gene is essential for perinatal survival.
Gene Expression Analysis in NF90(Ϫ/Ϫ) Mice-We analyzed the expression of NF90 mRNA and protein in Day 1 NF90 homozygote, heterozygote, and wild-type littermates. The Northern blot analysis shows that wild-type mice express a single NF90 mRNA migrating at 3.4 kb ( Fig. 2A, lanes 2, 4, and  5), and heterozygote mice express two isoforms of NF90 mRNA, migrating at 3.4 and 3.1 kb ( Fig. 2A, lanes 1 and 3). These mRNA sizes are consistent with transcription of both wild-type and targeted NF90 alleles. The NF90(Ϫ/Ϫ) mice demonstrate expression of a single isoform of NF90 mRNA migrating at 3.1 kb ( Fig. 2A, lanes 6 and 7), and this size is consistent with transcription just of targeted alleles, which lack exons 2-4. The prominent intensity of the truncated 3.1-kb targeted NF90 mRNA in NF90(Ϫ/Ϫ) mice indicates that the endogenous NF90 promoter is functional following deletion of exons 2-4. Furthermore, the increase in intensity of the 3.1-kb truncated form of NF90 mRNA compared with the 3.4-kb wild-type form ( Fig. 2A,  lanes 6 and 7 versus 2, 4, and 5) suggests that the null phenotype (absent NF90 protein, Fig. 2B) is associated with increased transcription of the NF90 gene. Because the targeting strategy introduced a translational frameshift downstream of exon 4, there was no NF90 protein synthesized in NF90(Ϫ/Ϫ) mice (see below and Fig. 2B). Our prior studies demonstrated that transgenic expression of histidine-tagged NF90 protein potently suppressed the expression of endogenous NF90 protein (5). Here, the absence of NF90 protein in NF90(Ϫ/Ϫ) mice is associated with increased mRNA expression of the 3.1-kb targeted NF90 allele, through transcriptional or post-transcriptional up-regulation.
We analyzed protein expression of NF90 in extracts prepared from newborn wild-type, heterozygote, and homozygote NF90 mice by Western immunoblotting (Fig. 2B). In NF90 wild-type and heterozygote mice we observed a prominent immunoreactive band at 90 kDa, and a fainter band at 110 kDa, similar to the principal size observed in human Jurkat T-cell nuclear extracts (5) (Fig. 2B, lanes 1-5). The level of NF90 protein expression in heterozygote NF90 mice was slightly decreased from that in wild-type mice (Fig. 2B, lanes 1 and 3 versus 2, 4,  6). The NF90(Ϫ/Ϫ) mice showed complete absence of the 90and 110-kDa bands (Fig. 2B, lanes 6, 7 versus 1-5), demonstrating that the translational frameshift in the knock-out phenotype was associated with complete absence of NF90 and NF110 protein expression. Examination of the tissue-specific expression of NF90 and NF110 proteins in adult wild-type mice showed the greatest expression in testis, brain, and skeletal muscles (Fig. 2C, lanes 8, 2, 6), and moderate expression in heart, spleen, lung, liver, and kidney (Fig. 2C, lanes 1, 3, 4, 5,  and 7).
Histological Analysis of NF90(Ϫ/Ϫ) Mice-We performed careful histology on hematoxylin and eosin-stained sections of developing and newborn wild-type, heterozygote, and homozygote NF90 mice and did not discern any dramatic organ-specific developmental abnormalities. Consistent with the overall size reduction of the homozygote NF90-deficient mice, we observed a proportionate reduction in the size of all the internal organs, compared with wild-type and heterozygote littermates.
We characterized the tissue-specific expression and subcellular localization of NF90 protein by immunohistochemistry with a monoclonal antibody that specifically recognizes NF90 (Fig. 3). Wild-type mice show widespread expression of NF90 protein, predominantly in the nucleus of cells, including cardiomyocytes (Fig. 3A), lung epithelial and endothelial cells (Fig. 3B), certain hepatocytes (Fig. 3C), and most epithelial cells lining the stomach (Fig. 3D). We observed the strongest expression of NF90 in cerebral (Fig. 3E) and cerebellar (Fig.  3F) neurons, and in skeletal muscles of the back (Fig. 3G) and diaphragm (Fig. 3H) 3F) and epithelial cells (Fig. 3, B and D) of NF90(ϩ/ϩ) mice revealed moderate NF90 immunostaining in the cytoplasm. For comparison, NF90 immunostaining was completely absent in NF90(Ϫ/Ϫ) mice in cerebral (Fig. 3I) and cerebellar (Fig. 3J) neurons, and in skeletal muscles of the back (Fig. 3K) and diaphragm (Fig. 3L). These results support the specificity of the monoclonal antibody in recognizing NF90, and confirm that NF90(Ϫ/Ϫ) mice express no detectable NF90 protein.
As the cause of death of the homozygote NF90-deficient mice appeared to be respiratory failure, we examined the lung histology in detail. We found no evidence of lung developmental abnormalities, insufficient lung inflation, pulmonary edema, or hyaline membrane formation in the NF90(Ϫ/Ϫ) mice that might explain perinatal death because of respiratory failure. We detected no differences in the expression of lung surfactant genes SP-A, -B, and -C, and a 50% decrease in SP-D expression by RT-PCR in NF90(Ϫ/Ϫ) mice compared with wild-type littermates (data not shown). Surfactant protein D functions principally in lung host defense, and mice lacking SP-D are viable after birth (38).
Blood, Serum, and Lymphocytes Subset Analyses of NF90(Ϫ/Ϫ) Mice-Complete blood count analyses of the neonatal mice were difficult to perform because of the low amount of blood recoverable at sacrifice. However, there was a consistent suggestion of anemia in the NF90(Ϫ/Ϫ) mice (data not shown). White blood cell counts and differentials were similar between NF90(Ϫ/Ϫ) mice and wild-type littermates. There were no significant differences in the clinical chemistry measurements of hepatic or renal function between homozygote, heterozygote, and wild-type mice NF90 mice (data not shown).
Flow cytometry analyses of thymic and splenic lymphocyte subsets from neonatal NF90(Ϫ/Ϫ) and heterozygous littermates revealed no differences in the expression of CD4 and CD8 double positive and single positive thymocytes between NF90(ϩ/Ϫ) and NF90(Ϫ/Ϫ) mice (data not shown).

NF90(Ϫ/Ϫ) Mice Show Pathological Maturation of Skeletal
Muscles-Searching for a potential cause of perinatal respiratory failure that would be independent of the lung parenchyma, we carefully examined the skeletal muscle fibers of the diaphragm and discovered that NF90(Ϫ/Ϫ) mice exhibited a relative paucity of muscle fibers and generally thinner diaphragm muscles than heterozygote or wild-type littermates (Fig. 4, B versus A, and Fig. 3, L versus H). The abnormal features present in the diaphragm were also present in intercostal and back muscles of homozygote NF90(Ϫ/Ϫ) mice, including overall paucity of muscle fibers, wider variation of muscle fiber sizes, and some evidence of myofibers with strongly eosinophilic and homogeneous cytoplasm (Fig. 4, B versus A). These abnormal fibers were likely hypercontracted, possibly as a consequence of disrupted integrity of the sacrolemma inducing unstable ion fluxes. Nuclei within skeletal myofibers of NF90(Ϫ/Ϫ) mice appeared larger with more granular chromatin visible compared with nuclei from myofibers of heterozygote and wild-type littermates. We observed greater numbers of multinucleated giant fusing myocytes in the NF90(Ϫ/Ϫ) mice (Fig. 4, D versus C), suggesting that skeletal muscles in the NF90(Ϫ/Ϫ) mice are developmentally delayed, or that ongoing muscle degeneration may be triggering efforts at regeneration.
TUNEL Analysis of NF90-deficient Mice-The small size of the NF90(Ϫ/Ϫ) mice and their internal organs may be a consequence of a combination of generalized growth retardation, developmental delay, and/or increased apoptosis during development. To further understand the causes of small size and perinatal death in NF90-deficient mice, we performed in situ TUNEL staining. TUNEL-positive cells were more prominent in E19 and newborn NF90(Ϫ/Ϫ) mice compared with wild-type or heterozygote littermates, throughout multiple organs. The greatest numbers of TUNEL-positive cells were present in cortical neurons and skeletal muscles of the diaphragm in NF90(Ϫ/Ϫ) animals (Fig. 5, B and D versus 5, A and C). The increased presence of TUNEL-positive cells in NF90(Ϫ/Ϫ) animals correlated with the observed reduction in size and weight.
Reduced Expression of Myogenic Regulatory Factors, MyoD, Myogenin, and p21 WAF1/CIP in NF90(Ϫ/Ϫ) Mice-We analyzed neonatal skeletal muscle expression of MyoD, myogenin, and p21 WAF1/CIP by semi-quantitative RT-PCR (Fig. 6A). Expression of MyoD mRNA expression was greatly inhibited, and myogenin and p21 WAF1/CIP expression was reduced ϳ50% in NF90(Ϫ/Ϫ) compared with wild-type and heterozygote littermates (Fig. 6A, lanes 4 -6 versus 2, 3, and 1). We discerned slight decreases in the myogenin and p21 WAF1/CIP RT-PCR products in the heterozyogote compared with the wild-type littermates (Fig. 6A, lane 1 versus 2 and 3). We confirmed the reduction in myogenic regulatory factors in NF90(Ϫ/Ϫ) mice by Western analysis of skeletal muscle extracts (Fig. 6B). Protein expression of MyoD and p21 WAF1/CIP1 is markedly reduced in NF90(Ϫ/Ϫ) mice (Fig. 6B, lanes 4 -6 versus 2 and 3), which corresponds well with the results of mRNA expression (Fig. 6A). p21 WAF1/CIP1 and MyoD 3Ј-UTR RNAs Interact Specifically with NF90 -The 3Ј-UTRs of the murine and human p21 WAF1/CIP1 mRNA sequences show significant conservation in the proximal 350-nucleotide region that contains multiple AU-rich binding elements (39). This proximal 3Ј-UTR is predicted to adopt secondary structures that includes regions of double-stranded RNA and hairpin loops (Fig. 7A) (40). The MyoD 3Ј-UTR is also predicted to adopt extensive secondary structure (Fig. 7B). These RNA secondary structures are binding targets for specific regulatory RNA-binding proteins such as HuR (36) and NF90 (this work). We used Northwestern blotting to characterize specific interactions between a radioactively labeled murine p21 WAF1/CIP1 3Ј-UTR probe and testis, brain, and muscle extracts from adult wild-type mice, as well as brain and muscle extracts from neonatal NF90(Ϫ/Ϫ) mice (Fig.  7C). The p21 WAF1/CIP1 3Ј-UTR RNA probe specifically labels a 90 kDa protein that is present in wild-type tissues (Fig. 7C,  lanes 1-3), is completely absent in extracts of NF90(Ϫ/Ϫ) mice (Fig. 7C, lanes 4 and 5), and that comigrates with labeled, recombinant NF90 protein (Fig. 7C, lane 6). The MyoD 3Ј-UTR probe also specifically labels a 90 kDa protein in wild-type tissues (Fig. 7D, lanes 1 and 3) and recombinant NF90 protein (Fig. 7D, lane 6), and shows no labeling in extracts of NF90(Ϫ/Ϫ) mice (Fig. 7D, lanes 4 and 5). Both 3Ј-UTR probes showed no labeling of recombinant NF45 protein (Fig. 7, C and  D, lane 7). Taken together, these results identify NF90 as the principal specific p21 WAF1/CIP1 and MyoD 3Ј-UTR RNA-binding protein present in skeletal muscles. DISCUSSION Our study is the first to establish that NF90, a dsRNAbinding protein, is essential for perinatal survival in mice. The gross phenotype of NF90(Ϫ/Ϫ) mice is small size, weak cry, lack of spontaneous movements and death because of rapidly progressive respiratory failure. The relative contributions of diaphragmatic weakness, neurologic abnormalities or metabolic derangements to perinatal death are difficult to distinguish. Nevertheless, the immediate cause of death less than 12 h after delivery appears to be diaphragm muscle weakness and respiratory failure. We observed morphological correlations with respiratory failure of disorganized diaphragm architecture and muscle fibers with inclusions suggestive of apoptosis in the NF90(Ϫ/Ϫ) mice. A similar phenotype of perinatal death because of respiratory failure and decreased mass of diaphragm and intercostals muscles was observed in mice deficient in the activin-and transforming growth factor ␤-modulating protein, follistatin (41).
Targeted disruptions of other dsRBM-binding proteins in mice have been associated with varying phenotypes. RNA helicase is a nuclear transcriptional regulator that is the mammalian homolog of the Drosophila X-chromosome dosage-compensating factor, maleless. Targeted disruption of RNA helicase resulted in embryonic lethality before gastrulation, and was associated with enhanced TUNEL-positive staining (42). Zfr is a nuclear zinc finger RNA-binding protein with homology to NF90, expressed at highest levels in testis, ovary, and brain (43). Targeted disruption of Zfr resulted in early embryonic lethality at day 8 -9 as a conse-  Northwestern blots were performed using in vitro transcribed, radiolabeled 3Ј-UTR RNA probes for murine p21 WAF1/CIP1 (C) or MyoD (D). Murine tissue extracts (80 g each) (lanes 1-3, adult NF90(ϩ/ϩ); lanes 5 and 6, neonatal NF90(Ϫ/Ϫ)) and recombinant histidine-tagged NF90 and NF45 proteins (lanes 7 and 8, 10 g each) (5) were fractionated by SDS-PAGE, transferred to nitrocellulose, and hybridized with probed 32 P-lableled 3Ј-UTR probes. quence of developmental delay and enhanced apoptosis (44). Spnr is a cytoplasmic dsRNA-binding protein localized to microtubules in the testis, brain, and ovary that may contribute to translational activation (45). A hypomorphic allele of Spnr produced by a gene trap strategy resulted in small size, increased lethality and neurological and reproductive defects (46). Two independent knockouts of the cytoplasmic dsRNAactivated protein kinase, PKR, resulted in no obvious phenotype, and this may reflect the finding that each represents only a partial knockout (47). Near complete targeted disruption of the nuclear ADAR2 RNA-editing enzyme resulted in perinatal lethality by P12, because of increased seizure activity (48). Targeted disruption of nuclear ADAR1 resulted in embryonic lethality before E12, associated with widespread apoptosis (49). Together, these results suggest that nuclear dsRNA-binding proteins serve essential roles such as transcriptional regulation and RNA splicing. Cytoplasmic dsRNA-binding proteins contribute to mRNA stabilization, subcellular localization, and regulation of translation.
Essential and redundant contributions of myogenic regulatory factors (MRFs), MyoD, Myf5, and myogenin, and p21 WAF1/CIP1 family members to embryonic muscle development have been elucidated through gene targeting studies. Either MyoD or Myf5 can direct myogenic specification (50), whereas myogenin (51), and either p21 WAF1/CIP1 or p57 KIP2 are essential for normal myogenic differentiation (52). An essential role for MyoD in the development of the skeletal muscle of the diaphragm was revealed by the failure to obtain viable mice during backcrossing with mdx mice (lacking dystrophin) (53). Reductions in myogenin gene expression in mice resulted in embryonic or neonatal lethality characterized by severe reductions in skeletal muscles (51,54,55). Recent evidence suggests that Mrf4, an MRF previously considered to be involved in myogenic differentiation, may in fact be an early myogenic specification factor, serving a redundant function to that of MyoD and Myf5 (56). Up-regulation of p21 WAF1/CIP1 promotes cell cycle exit and terminal differentiation in skeletal muscles, characterized by the fusion of myoblasts into myotubes, and increased synthesis of muscle contractile proteins (57). p21 WAF1/CIP1 (Ϫ/Ϫ) mice develop normally however (58), and this was proposed to be because of redundant cell cycle regulation by p57 KIP2 (52). Mice that lacked both p21 WAF1/CIP1 and p57 KIP2 were not viable because of severely compromised muscle development as manifested by diminished diaphragm muscles. The absence of both cell cycle inhibitors resulted in overproliferation and increased apoptosis leading to a marked decrease in mature skeletal muscle fibers in vivo. The absence of p21 WAF1/CIP1 was associated with enhanced programmed cell death in differentiating myoblasts in vitro (59).
Mechanisms of post-transcriptional mRNA stabilization of MyoD, myogenin, and p21 WAF1/CIP1 have been studied in differentiating C2C12 myoblasts (36,60). Ultraviolet cross-linking of radiolabeled 3Ј-UTR RNA probes to cytosolic extracts of C2C12 cells labeled proteins of 37, 86, and 116 kDa (36). The 37 kDa protein was identified as the embryonic lethal abnormal vision (elav) protein, HuR, that binds AU-rich elements (AREs) and mediates RNA-stabilization and nuclear-to-cytoplasmic export. Knockdown of HuR protein expression by RNA interference in C2C12 cells inhibited myogenesis in vitro (60). However, in MDA-468 human breast cancer cells, although HuR bound p21 WAF1/CIP RNA, it was not a major modulator p21 WAF1/CIP expression or growth inhibition (39). HuR protein expression in young adult mice was prominent in intestine, thymus, and spleen, followed by the liver, and was almost undetectable in skeletal muscle, brain, kidney, lung, and heart (61). This low to absent expression of HuR in skeletal muscles raises a question about its contributions to post-transcriptional stabilization of MyoD, myogenin, and p21 WAF1/CIP1 during myogenesis in vivo.
Here, we demonstrate that NF90 is the principal and specific RNA-binding protein in skeletal muscles for MyoD and p21 WAF1/CIP1 . Gene-targeted NF90(Ϫ/Ϫ) mice, completely lacking NF90 protein, demonstrate severely reduced expression of MyoD, myogenin, and p21 WAF1/CIP1 . We propose that NF90(Ϫ/Ϫ) mice are deficient in post-transcriptional stabilization of MyoD, myogenin, and p21 WAF1/CIP1 mRNAs. The phenotype of impaired muscle formation associated with increased apoptotic myonuclei is likely the direct result of this coordinated down-regulation of multiple myogenic regulators, including those involved both in myogenic specification, cell cycle withdrawal, and differentiation. Because MyoD(Ϫ/Ϫ) (50) and p21 WAF1/CIP1 (Ϫ/Ϫ) (58) mice are each viable through compensatory expression of myogenic regulators, we reasoned that our phenotype of perinatal death because of diaphragmatic respiratory failure must arise from effects on more than one of these myogenic regulators. Mice deficient in MyoD (62), or in cell cycle inhibitors p21 WAF1/CIP1 and p57 KIP2 (52), exhibited overproliferation and increased programmed cell death. This is consistent with our observations of increased apoptosis in NF90(Ϫ/Ϫ) mice, especially prominent in skeletal muscles and brain. The overall growth stunting of NF90(Ϫ/Ϫ) mice (40% lower weight than littermates) is likely a consequence of abnormal myogenesis and organogenesis related to defects in exiting the cell cycle. Future studies will be directed at identifying other structured RNA targets involved in differentiation that are critically regulated through specific binding to NF90.