|Home | About | Journals | Submit | Contact Us | Français|
Fbxo45 is an F-box protein that is restricted to the nervous system. Unlike other F-box proteins, Fbxo45 was found not to form an SCF complex as a result of an amino acid substitution in the consensus sequence for Cul1 binding. Proteomics analysis revealed that Fbxo45 specifically associates with PAM (protein associated with Myc), a RING finger-type ubiquitin ligase. Mice deficient in Fbxo45 were generated and found to die soon after birth as a result of respiratory distress. Fbxo45−/− embryos show abnormal innervation of the diaphragm, impaired synapse formation at neuromuscular junctions, and aberrant development of axon fiber tracts in the brain. Similar defects are also observed in mice lacking Phr1 (mouse ortholog of PAM), suggesting that Fbxo45 and Phr1 function in the same pathway. In addition, neuronal migration was impaired in Fbxo45−/− mice. These results suggest that Fbxo45 forms a novel Fbxo45-PAM ubiquitin ligase complex that plays an important role in neural development.
Ubiquitin-dependent proteolysis is indispensable for various biological processes (3, 40). Protein ubiquitylation is mediated by several enzymes that act in concert, with a ubiquitin ligase (E3) playing a key role in substrate recognition (14). E3 enzymes contain specific structural motifs that mediate recruitment of a ubiquitin-conjugating enzyme (E2), with these motifs including HECT, RING finger, U-box, and PHD finger domains (30). The SCF complex consists of Skp1 (adaptor subunit), Cul1 (scaffold subunit), an F-box protein (substrate recognition subunit), and Rbx1 (also known as Roc1 or Hrt1; RING finger-containing subunit). Whereas Skp1, Cul1, and Rbx1 are common to all SCF complexes, the F-box protein is variable (with ~70 such proteins having been identified in humans) and confers substrate specificity.
Fbxo45 is an F-box protein that was originally isolated as an estrogen-induced protein (47). Human and mouse Fbxo45 genes comprise three exons and possess several consensus binding sequences for the estrogen receptor in the promoter region. Fbxo45 mRNA is rapidly induced on exposure of MCF-7 cells to 17β-estradiol (47). FSN-1, the Caenorhabditis elegans ortholog of Fbxo45, binds to RPM-1 (regulator of presynaptic morphology 1) together with CUL-1 and SKR-1, the C. elegans orthologs of mammalian Cul1 and Skp1, respectively (21, 46). RPM-1 belongs to an evolutionarily conserved family of proteins (the PHR family) that include Highwire (HIW) (Drosophila melanogaster), Esrom (Danio rerio), Phr1 (Mus musculus), and protein associated with Myc (PAM) (Homo sapiens), each of which contains a RING-finger domain that is required for its E3 activity (7, 20, 21, 27, 44). Complete loss of function of fsn-1 in C. elegans results in defects that are characterized by the simultaneous presence of overdeveloped and underdeveloped neuromuscular junctions (NMJs) and which are similar to, but not as pronounced as, those observed in rpm-1−/− mutants. These genetic findings support the notion that the functions of FSN-1 and RPM-1 are partially overlapping (21).
Although PHR family members interact with many potential targets (11, 24, 26, 31), genetic data have shown that one key substrate of RPM-1 and HIW is the mitogen-activated protein kinase kinase kinase known as DLK (dual leucine zipper kinase) in C. elegans and known as Wallenda in D. melanogaster, respectively. The abundance of this kinase is increased in rpm-1 or hiw mutants, and synaptic defects in the mutant worms and flies are suppressed by a loss of DLK or Wallenda. Furthermore, an increase in the level of DLK or Wallenda is sufficient to phenocopy the synaptic defects of the rpm-1 or hiw mutants (5, 27). PAM has also been shown to catalyze the ubiquitylation of tuberin (TSC2) and to regulate signaling by mTOR (mammalian target of rapamycin) in human cells (12).
To elucidate the physiological functions of Fbxo45 in mammals, we have now generated mice deficient in this protein. Analysis of the mutant mice revealed that Fbxo45 is required for normal neuromuscular synaptogenesis, axon pathfinding, and neuronal migration. Moreover, we found that Fbxo45 does not form an authentic SCF complex as a result of an amino acid substitution in the F-box domain, and we identified PAM as a binding partner of Fbxo45. The phenotype of Fbxo45−/− mice was found to be similar to that of Phr1−/− mice, especially with regard to the defects of neuromuscular synapse formation and of axon navigation. Our results indicate that three fundamental processes of neural development— axonal projection, synapse formation, and neuronal migration—may be linked by a common machinery consisting of the Fbxo45-Phr1 complex.
Genomic DNA corresponding to the Fbxo45 locus was isolated from E14 mouse embryonic stem (ES) cells by PCR using of LA-Taq polymerase (Takara). The targeting vector was constructed by replacing a 1.0-kb fragment of the genomic DNA containing exon 1 of Fbxo45 with internal ribosome entry site (IRES)-lacZ and PGK-neo-poly(A)-loxP cassettes. The vector thus contained 6.3- and 1.4-kb regions of homology located 5′ and 3′, respectively, relative to IRES-lacZ and the neomycin resistance gene (neo). A diphtheria toxin A cassette was ligated at the 3′ end of the targeting construct. The maintenance, transfection, and selection of ES cells were performed as described previously (28). The recombination event was confirmed by Southern blot analysis with a 0.5-kb fragment of genomic DNA that flanked the 3′ homology region (see Fig. Fig.3).3). The expected sizes of hybridizing fragments after digestion of genomic DNA with BglII and SpeI were 7.6 and 5.6 kb for the wild-type and mutant Fbxo45 alleles, respectively. Mutant ES cells were microinjected into C57BL/6 mouse blastocysts, and the resulting male chimeras were mated with C57BL/6 females. Germ line transmission of the mutant allele was confirmed by Southern blot analysis. Heterozygous offspring were backcrossed for 12 generations to C57BL/6 mice and were then intercrossed to produce homozygous mutant animals. For genotyping of embryos, DNA was extracted from the yolk sac or tail at embryonic day 13.5 (E13.5) to E18.5 and was analyzed by PCR with the primers PJL (5′-TGCTAAAGCGCATGCTCCAGACTG-3′), TS92 (5′-GGTTTCCCATCATTCATTTTCAGC-3′), and TS93 (5′-GCCTTTTGTTTGTTTGTTTGGG-3′). All mouse experiments were approved by the animal ethics committee of Kyushu University.
Polyclonal anti-Rbx1 was generated in rabbits by injection with full-length recombinant human Rbx1 produced in Escherichia coli. Polyclonal anti-Fbxo45, anti-DLK, and anti-ALK were kindly provided by H. Okano (Keio University), S. Ohno (Yokohama City University), and T. Yamamoto (Tokyo University), respectively. Polyclonal antisynaptophysin and anti-Cul1 were obtained from Zymed; monoclonal anti-Skp1, anti-Hsp70, and anti-Hsp90 were from BD Transduction Laboratories; polyclonal anti-phospho-p38 was from Promega; polyclonal anti-extracellular signal-regulated kinase (anti-ERK), anti-phospho-ERK, anti-c-Jun NH2-terminal kinase (anti-JNK), anti-phospho-JNK, anti-c-Jun, anti-phospho-c-Jun, and anti-TSC2 were from Cell Signaling; polyclonal antineurofilaments, monoclonal anti-FLAG (M2), and monoclonal anti-βIII-tubulin were from Sigma; monoclonal anti-hemagglutinin epitope (anti-HA) (HA11) was from Covance; polyclonal anticalretinin and anti-choline acetyltransferase were from Chemicon; polyclonal anti-LZK was from ABGENT; and monoclonal anti-Isl1 was from Developmental Studies Hybridoma Bank. Alexa 546-conjugated phalloidin and Alexa 594-conjugated α-bungarotoxin were obtained from Molecular Probes.
Frozen tissues were cut into small pieces and then homogenized for 5 min in an ice-cold solution containing 50 mM Tris-HCl (pH 7.5), 0.25 M sucrose, 1 mM EDTA, aprotinin (10 μg/ml), leupeptin (10 μg/ml), and 1 mM phenylmethylsulfonyl fluoride. The homogenate was diluted with an equal volume of 2× radioimmunoprecipitation assay buffer (0.3 M NaCl, 20 mM Tris-HCl [pH 7.5], 2% Nonidet P-40, 0.2% sodium deoxycholate, 0.2% sodium dodecyl sulfate [SDS], aprotinin [10 μg/ml], leupeptin [10 μg/ml], 1 mM phenylmethylsulfonyl fluoride), incubated on ice for 15 min, and centrifuged at 15,000 × g for 15 min at 4°C. The resulting supernatant (60 μg of protein) was then subjected to immunoblot analysis as described previously (36).
Tissue was fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS), embedded in paraffin, and sectioned at a thickness of 5 μm. Hematoxylin-eosin staining was performed as described previously (29). For immunohistofluorescence analysis, samples were fixed with 4% paraformaldehyde in PBS, embedded in OCT compound (Tissue Tek), and sectioned with a cryostat at a thickness of 5 μm. Immunohistofluorescence staining was then performed as described previously (29). Immune complexes were detected with secondary antibodies labeled with Alexa 546 or Alexa 488 (Molecular Probes), each at a dilution of 1:2,000. The specimens were examined with a fluorescence microscope, Eclipse E800 (Nikon) or Radiance2000 (Bio-Rad), and photographed. For whole-mount staining of nerves, embryos were killed, eviscerated, fixed for 1 h with 2% paraformaldehyde in PBS, and washed with 0.1 M glycine in PBS. Tissue was then dissected, permeabilized overnight at 4°C with 0.5% Triton X-100 in PBS containing bovine serum albumin (10 mg/ml), and incubated overnight at 4°C with rabbit antibodies to synaptophysin and to neurofilaments in PBS containing bovine serum albumin (10 mg/ml). It was then washed three times for 20 min with PBS, incubated for 2 h at room temperature with Alexa 488-conjugated goat antibodies to rabbit immunoglobulin G (Molecular Probes), washed again, and mounted between coverslips. For simultaneous detection of acetylcholine receptors (AChRs) and nerve terminals, tissue was stained with Alexa 594-conjugated α-bungarotoxin and antisynaptophysin, washed, and then incubated with Alexa 488-conjugated goat antibodies to rabbit immunoglobulin G.
Skeletal staining was performed as described previously (37).
Dissociated primary neuronal cultures were prepared from the cerebral cortex of Fbxo45−/− mouse embryos and their wild-type littermates at E18.5. The cells were maintained in Neurobasal medium (Gibco) supplemented with B27, penicillin, streptomycin, and l-glutamine.
Pregnant mice were injected intraperitoneally with 5-bromo-2′-deoxyuridine (BrdU) (50 mg per kilogram of body weight) at E14.5, and embryos were dissected and fixed at E18.5. Sections were prepared and stained to detect BrdU-positive nuclei with the use of a BrdU in-situ detection kit (BD Transduction Laboratories) and Alexa 546-conjugated streptavidin (Invitrogen).
The plasmids p3×FLAG-Fbxo45 (mouse) and pcDNA3-FLAG-PAM (human) were kindly provided by H. Okano (Keio University) and T. Patel (Loyola University), respectively. Complementary DNAs encoding Fbxw2 tagged with three copies of the FLAG epitope (3×FLAG-Fbxw2) (mouse), Fbxw2 tagged with HA (HA-Fbxw2) (mouse), 3×FLAG-Skp2 (human), HA-Skp2 (mouse), HA-Fbxo45 (mouse), mouse Fbxo45 deletion mutants tagged at their NH2 termini with HA, and human PAM deletion mutants (R2, R3, R4, R5, and R8) tagged at their NH2 termini with 3×FLAG were subcloned into pcDNA3 (Invitrogen). Human PAM deletion mutants R1, R6, and R7 were subcloned into p3×FLAG-CMV 7.1 (Sigma). HEK293T cells were transfected with expression plasmids by the calcium phosphate method and were cultured for 48 h before analysis. They were then collected, lysed in a solution containing 0.5% Triton X-100, and subjected to immunoprecipitation with anti-FLAG as described previously (13).
HEK293T cells were transfected with the use of the FuGene 6 reagent (Roche). After 24 h, the cells were incubated with cycloheximide (50 μg/ml) for 0 to 9 h and then subjected to immunoblot analysis.
HEK293T cells transiently expressing 3×FLAG-tagged Fbxo45 or mock-transfected cells were lysed in a solution containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Triton X-100, aprotinin (10 μg/ml), leupeptin (10 μg/ml), 10 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, 400 μM Na3VO4, 400 μM EDTA, 10 mM NaF, and 10 mM sodium pyrophosphate (buffer A). The lysate was centrifuged at 16,000 × g for 10 min at 4°C, and the resulting supernatant was applied to a column of Sepharose CL4B (GE Healthcare) conjugated with anti-FLAG (M2). The column was washed with buffer A, after which bound proteins were eluted with the 3×FLAG peptide (200 μg/ml), concentrated by precipitation with trichloroacetic acid, fractionated by SDS-polyacrylamide gel electrophoresis (PAGE), and stained with Coomassie brilliant blue G250. Protein bands were excised from the stained gel and washed, and stain was removed with acetonitrile. The proteins were reduced with 10 mM dithiothreitol, alkylated with 55 mM iodoacetamide, and digested for 16 h at 37°C with sequence-grade trypsin (Promega). The resulting peptides were sequentially extracted from the gel with 0.1% trifluoroacetic acid (TFA) in 2% acetonitrile, 0.1% TFA in 33% acetonitrile, and 0.1% TFA in 70% acetonitrile. The combined extracts were evaporated, and the residue was dissolved with 0.1% TFA in 2% acetonitrile.
Peptides were subjected to chromatography on a C18 column (0.2 mm by 5 cm; L-column) with a high-performance liquid chromatography system (Magic 2002; Michrom BioResources) that was directly coupled to an ion-trap mass spectrometer (LCQ-Deca; Finnigan) equipped with a nano-LC electrospray ionization source (AMR). Collision-induced dissociation (CID) spectra were acquired automatically in the data-dependent scan mode with the dynamic exclusion option. Uninterpretable CID spectra were compared with human entries in the International Protein Index (European Bioinformatics Institute) with the use of the MASCOT algorithm. Assigned high-scoring peptide sequences were manually confirmed by comparison with the corresponding CID spectra.
Given that most, but not all, known F-box proteins form an SCF complex with Skp1, Cul1, and Rbx1 (4, 17, 30), we examined whether Fbxo45 is able to form such a complex. Various F-box proteins tagged 3×FLAG were expressed in HEK293T cells, immunoprecipitated with antibodies to FLAG, and subjected to immunoblot analysis with anti-Skp1, anti-Cul1, and anti-Rbx1 (Fig. (Fig.1A).1A). Fbxw2 and Skp2 (also known as Fbxl1) bound to endogenous Skp1, Cul1, and Rbx1, whereas Fbxo45 interacted with Skp1 but not with Cul1 and Rbx1. These data suggested that Fbxo45 does not form an authentic SCF complex.
The crystal structure of human SCFSkp2 shows that the F-box domain of Skp2 interacts not only with Skp1 but also with Cul1 (49). The Cul1-Skp2 interaction appears to be mediated by two amino acid residues, Pro113 and Glu115, in the F-box domain of Skp2. In the loop region (amino acids 113 to 116) of Skp2, which forms the binding interface with Cul1, Pro113 participates in a hydrophobic interaction with Thr54 of Cul1 whereas Glu115 forms a hydrogen bond and a salt bridge with Tyr139 and His143, respectively, of Cul1. In particular, the electrostatic interaction between the negatively charged Glu115 and the positively charged His143 appears key to the interaction between Skp2 and Cul1 (18). Alignment of the amino acid sequences of various F-box proteins (Fig. (Fig.1B)1B) revealed that Pro113 and Glu115 of Skp2 are conserved among all such proteins examined with the exception of Fbxo45, in which the negatively charged Glu is replaced with positively charged Arg (Arg42). A repulsive force between positively charged residues, Arg42 of Fbxo45 and His143 of Cul1, thus likely inhibits the binding of Fbxo45 to Cul1. Indeed, replacement of Arg42 of Fbxo45 with Glu (R42E) conferred on Fbxo45 the ability to bind to Cul1 and Rbx1 (Fig. (Fig.1C).1C). Conversely, a Skp2 mutant (E115R) in which Glu115 is replaced with Arg did not bind to Cul1 and Rbx1 (Fig. (Fig.1C).1C). We therefore conclude that an amino acid substitution in the F-box domain of Fbxo45 prevents the interaction of Fbxo45 with Cul1 and Rbx1.
Given that Fbxo45 is not able to associate with Cul1 and Rbx1, we hypothesized that it might interact with another RING-finger protein. We thus adopted a proteomics approach to identify proteins that bind to Fbxo45. An expression plasmid encoding 3×FLAG-tagged Fbxo45 was introduced into HEK293T cells, and immunoprecipitates prepared from cell lysates with anti-FLAG were subjected to SDS-PAGE. An ~500-kDa protein was thus found to interact specifically with Fbxo45 (Fig. (Fig.2A)2A) and was identified as PAM, a RING finger-type ubiquitin ligase, by LC-MS/MS. In addition, Skp1, but not Cul1 and Rbx1, was identified in the immunoprecipitates by LC-MS/MS analysis (data not shown), consistent with the results of coimmunoprecipitation experiments.
To determine the region of PAM that is responsible for binding to Fbxo45, we generated a series of deletion mutants (R1 to R6) of PAM (Fig. (Fig.2B)2B) and tested their ability to bind to HA-tagged Fbxo45 by coimmunoprecipitation analysis in HEK293T cells. Only the R4 deletion mutant, which contains the Myc-binding region, was found to interact with Fbxo45 (Fig. (Fig.2C).2C). We next generated a mutant comprising the Myc-binding region alone (R7) and its reciprocal R4-based mutant (R8). The R7 mutant interacted with Fbxo45 to a great extent, whereas the R8 mutant showed no detectable association, indicating that the Myc-binding region of PAM is sufficient for binding to Fbxo45. The interaction of the Myc-binding region of PAM with Fbxo45 was shown to be specific in that no such interaction was apparent with Fbxw2 or Skp2 (Fig. (Fig.2D).2D). We also examined which domain of Fbxo45 is responsible for binding to PAM (Fig. (Fig.2E).2E). An NH2-terminal deletion mutant lacking the F-box domain retained the ability to interact with PAM, whereas a COOH-terminal deletion mutant lacking the SPRY domain did not. These data thus suggested that the SPRY domain of Fbxo45 mediates the interaction with PAM.
We observed that the abundance of full-length Fbxo45 and the NH2-terminal deletion mutant of Fbxo45 but not that of the COOH-terminal deletion mutant was greater when these proteins were coexpressed with the Myc-binding region of PAM (Fig. (Fig.2E).2E). To determine whether this phenomenon might be attributable to stabilization of Fbxo45 by PAM, we transfected HEK293T cells with a vector for HA-tagged Fbxo45 or HA-tagged Skp2 (negative control) in the absence or presence of a vector for the 3×FLAG-tagged R7 deletion mutant of PAM. A cycloheximide chase experiment revealed that Fbxo45 was more stable in cells coexpressing the Myc-binding region of PAM than in those expressing Fbxo45 alone whereas the stability of Skp2 was unaffected by coexpression of the Myc-binding region of PAM (Fig. (Fig.2F).2F). The stability of PAM did not appear to be affected by Fbxo45 (data not shown). These results thus support the notion that Fbxo45 and PAM form a novel complex that is independent of Cul1 and Rbx1 (Fig. (Fig.2G2G).
To elucidate the physiological functions of Fbxo45, we generated mice deficient in this protein. The Fbxo45 gene was disrupted in mouse ES cells by replacement of exon 1 with IRES-lacZ and PGK-neo-poly(A)-loxP cassettes (Fig. 3A and B). Mice heterozygous for the Fbxo45 mutant allele were healthy, fertile, and phenotypically indistinguishable from wild-type littermates, although the abundance of Fbxo45 in tissues of the mutant mice was about half of that in those of wild-type animals (Fig. (Fig.3C).3C). In contrast, no adult homozygous mutants were obtained from heterozygote crosses (see Table S1 in the supplemental material). Fbxo45−/− embryos were recovered at the expected Mendelian ratio until E17.5 to E18.5, suggesting that the homozygous mutation is not embryonic lethal. Immunoblot analysis revealed that Fbxo45 was exclusively expressed in neuronal tissues, including the cerebrum and medulla oblongata in wild-type mice, whereas it was not detectable in Fbxo45−/− mice (Fig. (Fig.3D).3D). These data suggested that Fbxo45 might play an important role in neural development.
Homozygous mutant newborns were found to die within 30 min after birth and were readily distinguished by their characteristic tucked position, pale color (Fig. (Fig.3E),3E), and absence of spontaneous movement. The Fbxo45−/− mice also exhibited skeletal deformities, such as kyphosis (Fig. (Fig.3F).3F). At E16.5, the homozygous mutants had a beating heart, and all internal organs appeared intact and appropriately positioned (see Fig. S1 in the supplemental material). Pulmonary alveoli of Fbxo45−/− newborns did not expand, although all components of lung tissue appeared normal (Fig. 3G and H). Together with the neuropathological anomalies described below, these observations suggested that Fbxo45 mutant mice die as a result of respiratory distress.
Given the lack of respiratory movement in neonatal Fbxo45−/− mice, we focused on the NMJs of respiratory muscles, especially those of the diaphragm. In wild-type mice, axons of the phrenic nerve exit the cervical spinal cord and contact the developing diaphragm at E12. The nerve and muscle develop together, with the nerve having formed three main branches—one extending dorsally to the crus, one innervating the dorsal portion of the diaphragm, and one innervating the ventral portion of the diaphragm—by E13 (9). Whole-mount immunostaining of the diaphragm at E14.5 or E15.5 revealed that motor innervation of the diaphragm was incomplete in Fbxo45−/− mice (Fig. 4A and B). Whereas the entire diaphragm is contacted by axons of the phrenic nerve in wild-type mice, branching and projection of the phrenic nerve in the mutant mice were substantially impaired, with axons consistently failing to reach the most ventral region of the diaphragm. Furthermore, branching of axons in the mutant animals was sometimes misoriented (Fig. (Fig.4B,4B, panel f). In wild-type mice, right and left phrenic nerves innervate the right and left portions of the diaphragm, respectively, and the regions innervated by the two nerves do not overlap (Fig. (Fig.4C).4C). In contrast, in the diaphragm of Fbxo45−/− mice, right and left phrenic nerves sometimes extended across the endplate band of the muscle and became entangled (Fig. (Fig.4D4D).
We next examined whether peripheral sensory projections were similarly affected by Fbxo45 deletion. Whole-mount immunostaining revealed excessive branching of cutaneous sensory nerves in Fbxo45−/− mice at E14.5 (see Fig. S2 in the supplemental material). Axons appeared overgrown, and the nerves were defasciculated. This pattern of innervation resulted in a decrease in the areas surrounded by the sensory nerves in Fbxo45−/− mice compared with those in wild-type mice.
We next analyzed the pattern of developing neuromuscular synapses in Fbxo45 mutants at E16.5 and E18.5. To examine the distribution of nerves and clusters of AChRs, we stained presynaptic nerve terminals and postsynaptic AChRs with antisynaptophysin and α-bungarotoxin, respectively. In wild-type mice, NMJs are confined to a central region of the muscle fiber, the endplate band, which corresponds to the region in which intramuscular nerves are located (33, 41). Although AChRs were largely clustered within the endplate band of the muscle in both wild-type (Fig. 5A to C) and mutant (Fig. 5D to F) embryos, the band of AChR clusters in Fbxo45−/− mice was much narrower than that in wild-type mice at E16.5 (Fig. 5A to F; quantified in Fig. Fig.5G)5G) and was partially aneural at E18.5 (Fig. 5N to S). Furthermore, an ectopic endplate band was formed in the region of the pars sternalis of the diaphragm in some mutant embryos. In wild-type embryos, synapses between muscle and the left phrenic nerve are formed in the left portion of the pars sternalis, with muscle fibers manifesting only one endplate band in their central portion (Fig. 5H to J). However, in the diaphragm of mutant embryos, synapses were formed not only between muscle and the left phrenic nerve but also between muscle and the right phrenic nerve extending across the ventral midline in the left portion of the pars sternalis (Fig. 5K to M). Phrenic nerves never extend across the ventral midline in wild-type mice. These results suggested that Fbxo45 plays an indispensable role in normal innervation and synaptogenesis by peripheral nerves during development.
The widespread expression of Fbxo45 in the central nervous system, together with the innervation defects in the periphery of Fbxo45−/− mice, led us to examine the major axon tracts in the brains of mutant embryos. The brains of Fbxo45−/− embryos at E18.5 were smaller than those of wild-type mice (see Fig. S3A in the supplemental material). Histological sections revealed that the major axon tracts in the brain differed markedly between the two genotypes at E18.5. The anterior commissure was absent in Fbxo45−/− mice (Fig. 6A to D). The internal capsule was also absent, as were corticofugal and thalamocortical projections in the subcortical telencephalon (Fig. 6E to H). In contrast, the corpus callosum of mutant embryos was indistinguishable from that of wild-type embryos (see Fig. S3B and C in the supplemental material). Primary cultures of neurons prepared from the brains of Fbxo45−/− embryos at E18.5 manifested outgrowth of projections (see Fig. S4 in the supplemental material), suggesting that the loss of major axon tracts in the brains of Fbxo45−/− mice does not result from impairment of axon outgrowth. These results thus suggested that Fbxo45 is necessary for the formation of certain axonal tracts in the brain.
Development of the cerebral cortex normally proceeded in mutant embryos at an early stage (E14.5), as the cortical plate was formed in the preplate, splitting it into the marginal zone and the subplate. The laminated cortical structure was partly formed at postnatal day 1, and layer-specific differences in staining and cell density were prominent in the cortical plate of wild-type embryos (Fig. (Fig.7A).7A). However, in mutant embryos, such differential patterns of different types of cells were not clearly observed (Fig. (Fig.7B).7B). We next examined neuronal migration in Fbxo45 mutant mice by pulse labeling with BrdU at E14.5. A large subset (65%) of labeled nuclei was located in the outermost layer of the cortical plate adjacent to the marginal zone (bin 1) in the cortex of wild-type embryos at E18.5 (Fig. 7C and E). In contrast, BrdU-labeled nuclei were more dispersed in the cortical plate of mutant embryos, with a smaller subset (38%) of these nuclei being detected in the outermost layer (Fig. 7D and E). These observations indicated that radial migration of neurons was impaired in the mutant embryos. We also examined brain sections (E18.5) stained for calretinin, a marker of cortical interneurons. Although not all interneurons express calretinin, this marker has been used routinely to reveal a large subpopulation of γ-aminobutyric acid-containing interneurons that migrate tangentially to the cortex or hippocampus (22, 45). Calretinin-positive cells accumulated abnormally in the basal telencephalon—the area in which migrating interneurons arise (23)—in the brains of Fbxo45 mutant embryos (Fig. 7G to J). In contrast, few calretinin-positive cells were apparent in the basal telencephalon of the wild-type brain as a result of the migration of interneurons to the cortex or hippocampus (Fig. (Fig.7F7F).
Furthermore, development of the entire spinal cord was markedly affected by the loss of Fbxo45, as evidenced by lateral expansion of gray matter (Fig. 8A to D) and aberrant distribution of motor neurons (Fig. 8E to H). Neurons arise in the ventricular zone, located within the medial portion of the spinal cord, and migrate to their final position at the lateral edge of the cord. At brachial and lumbar levels, motor neurons are located in two regions: those that innervate limb muscle form the lateral motor column (LMC), and those that innervate the body wall form the medial motor column (MMC) (19, 42). The LMC was poorly formed and the MMC was aberrantly distributed in the spinal cords of Fbxo45−/− embryos (Fig. 8E to H). These results suggested that Fbxo45 is important for cell migration in the brain and spinal cord.
Evolutionary considerations suggest that the Fbxo45-PAM complex might target the receptor tyrosine kinase ALK, the mitogen-activated protein kinase kinase kinase DLK, or tuberin (TSC2) for ubiquitylation-dependent proteolysis (5, 12, 21, 26, 27). We therefore examined whether these candidate proteins accumulate in the brains of Fbxo45−/− embryos. Immunoblot analysis revealed that neither the abundance of ALK nor the phosphorylation level of ERK, which functions downstream of ALK (25), differed between wild-type and Fbxo45−/− embryos (see Fig. S5A and B in the supplemental material). We also observed no difference in the amounts of DLK or in the phosphorylation levels of p38 mitogen-activated protein kinase, JNK, or c-Jun, all of which function downstream of DLK (8, 15), between the two genotypes (see Fig. S5C to E in the supplemental material). In addition, the abundance of TSC2 did not appear to be affected by Fbxo45 deletion (see Fig. S5F in the supplemental material). Finally, the abundance of LZK, the other closest mammalian homolog of DLK (32), did not appear to differ between the two genotypes (see Fig. S5G in the supplemental material). These results thus suggest that ALK, DLK, TSC2, and LZK are not targets of Fbxo45 in the brain or that other ubiquitin ligases may participate in the degradation of these proteins.
Whereas eukaryotic genomes encode hundreds of F-box proteins (10, 17), only a single F-box protein that contains a SPRY domain has been identified in worms, flies, mice, or humans (21, 43, 47). Moreover, all of these last proteins are more closely related to each other by sequence than to other F-box proteins (21, 43). We have now shown that the mammalian F box- and SPRY domain-containing protein Fbxo45 forms an atypical ubiquitin ligase complex with the RING-finger protein PAM and that Fbxo45 is required for normal development of central and peripheral neurons. Although the physical association of Fbxo45 with PAM (H. sapiens) and of FSN-1 with RPM-1 (C. elegans) is evolutionarily conserved, the composition of the mammalian Fbxo45-PAM complex appears to differ from that of the C. elegans FSN-1-RPM-1 complex. Our results show that mammalian Fbxo45 contains a substitution in the F-box domain for an amino acid that is conserved in other F-box proteins and that this amino acid substitution interferes with the binding of Fbxo45 to Cul1. Fbxo45 thus retains the ability to associate with Skp1 but not with Cul1 and Rbx1 (the RING finger-containing subunit of conventional SCF complexes). Instead of Rbx1, Fbxo45 interacts directly or indirectly through its SPRY domain with PAM in a manner independent both of the F-box domain of Fbxo45 and of Skp1. Fbxo45 thus appears to bind to Skp1 via its NH2-terminal F-box domain and to PAM via its COOH-terminal SPRY domain. In contrast, C. elegans FSN-1 forms an SCF-like complex with SKR-1, CUL-1, and RPM-1 (21).
Our present results also reveal that the role of Fbxo45 in neural development has been conserved during evolution. However, mammalian Fbxo45 not only regulates the fine structure of synaptic connections, as does its invertebrate orthologs (21, 43), but also is required for proper neuronal migration and large-scale patterning of neuronal connectivity in the brain. The impaired branching and projection of the phrenic nerve and the partial innervation of the diaphragm apparent in Fbxo45−/− mice are indicative of functions for Fbxo45 beyond the regulation of nerve terminal morphology, such as control of the guidance of axons to their targets. Indeed, we detected a pronounced and specific loss of axon tracts throughout the central nervous system of the mutant mice, with the anterior commissure being completely absent and both thalamic and cortical axons failing to enter the subcortical telencephalon to form the internal capsule. Furthermore, neuronal migration was found to be impaired in Fbxo45-deficient embryos. These results thus suggest that Fbxo45 performs additional functions in neural development compared with its invertebrate orthologs. Consistent with this notion, the abundance of potential substrates including those targeted by invertebrate Fbxo45 orthologs—including ALK, DLK, and LZK (21, 27)—was not affected by the loss of Fbxo45 in mice, suggesting that other proteins may be targeted by Fbxo45 in mammals and that deregulation of the abundance of these proteins might be responsible for the defects in neural development observed in Fbxo45−/− mice.
Given that PAM and its orthologs Phr1 and RPM-1 are large proteins consisting of multiple functional domains (2, 11, 35, 48), these molecules are thought to perform more functions than Fbxo45 and its orthologs. In C. elegans, worms lacking functional FSN-1 exhibit a phenotype that is highly similar to that of rpm-1 mutants with regard to γ-aminobutyric acid-ergic synapses but less pronounced with regard to DD motoneurons and sensory neurons (21). These differences in phenotype suggest that RPM-1 might either interact with other F-box proteins in addition to FSN-1, act as a ubiquitin ligase without an F-box partner, or perform ubiquitin-independent functions. We also found marked similarity between the characteristics of Fbxo45-deficient mice and those of Phr1-deficient mice. Both Fbxo45- and Phr1-deficient mice die soon after birth as a result of respiratory distress and show both defective outgrowth of the phrenic nerve and impaired synaptic activity at NMJs (1, 2). Furthermore, both types of mutant mice manifest a pronounced and specific loss of axon tracts such as the anterior commissure and projections that form the internal capsule (1, 20). These observations indicate that Fbxo45 and Phr1 function in the same pathway and that not only the physical association of PAM or RPM-1 with Fbxo45 or FSN-1, respectively, but also the functional collaboration of these proteins is evolutionarily conserved.
Our study suggests that although the cellular processes that underlie axon guidance, synaptogenesis, and cell migration are fundamentally different, similar molecules may be involved. We have thus shown that Fbxo45 contributes to axon projection, synapse formation, and neuronal migration. Although the molecular mechanisms responsible for the coordination of axon growth, synaptogenesis, and cell migration remain obscure, they may include regulation by protein kinases that modify the function of microtubule modulators. Such kinases include Cdk5 (38), glycogen synthase kinase 3β (16), MARK1 (MAP/microtubule affinity-regulating kinase 1), protein kinase A (6, 34), and JNK (39). Our findings that disruption of Fbxo45 affects axon guidance, synaptogenesis, and neuronal migration support the existence of such a coordinating mechanism and provide clues for determination of its molecular basis.
We thank M. Shirane, K. Yumimoto, R. Tsunematsu, and M. Nishiyama for valuable discussions; N. Nishimura, Y. Yamada, K. Takeda, and other laboratory members for technical assistance; and A. Ohta and M. Kimura for help in preparation of the manuscript.
This work was supported in part by a grant from the Ministry of Education, Culture, Sports, Science, and Technology of Japan and by a research grant from the Takeda Science Foundation.
Published ahead of print on 27 April 2009.
†Supplemental material for this article may be found at http://mcb.asm.org/.