Disruption of the Fbxw8 Gene Results in Pre- and Postnatal Growth Retardation in Mice^{▿ †}

Generation of Fbxw8 gene trap mice.

Male chimeric mice generated from the embryonic stem (ES) cell line RRT057 were obtained from BayGenomics (http://www.genetrap.org) (27). The ES cell line RRT057 was generated by using a gene trap protocol with the trapping construct pGTOLxf, containing a splice-acceptor sequence subcloned toward the 5′ end from the β-geo reporter cassette encoding the β-galactosidase (β-Gal)-neomycin resistance fusion protein. The founder chimera was mated with C57BL/6 females. Mice were genotyped by PCR using primers that identified the β-geo insertion and the wild-type Fbxw8 genomic sequence. Primer sequences for Fbxw8 genotyping will be provided upon request.

Cell culture.

MEFs were established from embryos isolated at embryonic day 14.5 (E14.5) and cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. For growth curves, 10⁵ cells were plated in duplicate into 60-mm dishes and cells were counted twice at 24-h intervals by using a hemocytometer. Serial passage growth over a 3-day interval (expressed as the ratio of the number of cells on day 3 [N3] to the number of cells on day 0 [N0], or the N3/N0 ratio) was determined according to the 3T3 protocol (30). Human lymphoblastoid B-cell lines derived from a 3-M syndrome patient (FY0211) and the patient's parents (FY0212 and FY0213) were obtained from the European Collection of Cell Cultures.

Histological analysis.

Embryos and placentas from timed matings of Fbxw8^+/− mice were fixed in Bouin's fixative (Sigma) and embedded in paraffin. For β-Gal staining, frozen sections of embryos and placentas were prepared and stained according to the protocols of the β-Gal stain manufacturer (Chemicon).

Western blotting and immunoprecipitation.

MEFs were lysed in NET-N buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1 mM EDTA, 0.5% NP-40) containing protease inhibitor cocktail set I (Calbiochem), and lysates were subjected to immunoblot analysis. For immunoprecipitation, cell lysates were incubated with the corresponding antibody and protein A-Sepharose beads (Amersham) for 3 h and then washed vigorously three times with the lysis buffer before boiling in sodium dodecyl sulfate (SDS) sample buffer. For the determination of tyrosine phosphorylation, MEFs were serum starved for 14 h and treated with a 30-ng/ml concentration of recombinant mouse insulin-like growth factor 1 (IGF1; R&D Systems) for 15 min and then lysed in buffer containing 50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1% Triton X-100, 10 mM NaF, and 5 mM Na₃VO₄. The rabbit anti-FBXW8 and anti-CUL7 antibodies were generated by Bethyl Laboratories (Montgomery, TX). Additional antibodies used were vinculin (hVIN1; Sigma), SKP1 (Neomarkers), IGFBP2 (R&D Systems), IGF1 receptor (IGF1R [C-20; Santa Cruz]), and phosphotyrosine (PY99; Santa Cruz).

siRNA transfection.

Short interfering RNA (siRNA) oligonucleotides were obtained from Dharmacon. Sequences of CUL7, FBXW8, and luciferase gene siRNA oligonucleotides were as follows: CUL7 A, AAU CAA CUG CCA UGU CUA CAA; CUL7 B, AAG AUA CUC CAA CUU CUA CAA; FBXW8 N1, AAG AUG UGC ACA GGU GAG CAA; FBXW8 N2, AAG ACG UGG AAG GUG AUU GCA; and luciferase gene siRNA, AAC UUA CGC UGA GUA CUU CGA. Cells in a 6-well plate were transfected with 10 μl of 20 μM oligonucleotides by using Oligofectamine (Invitrogen). Cells were harvested 72 h after transfection and used for immunoblotting.

Quantitative PCR.

Total RNA was extracted using an RNeasy mini kit (QIAGEN). Three micrograms of RNA was reverse transcribed using the SuperScript III first-strand synthesis system (Invitrogen). Specific cDNA was quantified by using TaqMan gene expression assays, specific primers, probes for mouse Cul7 and Fbxw8 and the mouse β-actin gene, and the real-time PCR 7500 system (Applied Biosystems).

Analysis of secreted proteins.

Early-passage MEFs were incubated in 100-mm dishes in Opti-MEM (Invitrogen) for 24 h, and supernatants were harvested for trichloroacetic acid precipitation according to the procedure reported recently (19). The precipitate from trichloroacetic acid was subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and stained with Coomassie blue. The band reproducibly differing in abundance was excised and identified using mass spectrometry.

Transcriptome profiling and Affymetrix data analysis.

RNAs were extracted from E18.5 whole embryos by using TRIzol (Invitrogen), subjected to homogenization followed by RNA cleanup with a QIAGEN system, labeled, and hybridized to an Affymetrix GeneChip Mouse Genome 430 2.0 array. Expression data were processed using the R-Bioconductor package Affy (www.bioconductor.org). The preferred analysis methods (5) were used to generate expression values for each probe set. The details of the analysis will be provided upon request.

Generation of Fbxw8^−/− mice.

Mice were generated from ES cells (RRT057) containing an insertion mutation, or gene trap, in the third intron of the Fbxw8 gene. The gene trap vector generates a spliced fusion transcript comprising Fbxw8 and the β-geo cassette encoding β-Gal and neomycin resistance. We determined that the trapping construct was inserted 2.6 kb from the 3′ end of exon 3 of Fbxw8 in the RRT057 cells by sequencing of genomic DNA (Fig. 1A). A PCR assay was established to distinguish the gene-trapped allele from the wild-type allele (Fig. 1B and C). In addition, Southern blotting with a probe against the β-geo cassette confirmed a single integration with the gene trap (Fig. 1D).

To determine if the gene trap reduced FBXW8 protein expression, lysates prepared from MEFs were Western blotted with an antibody against FBXW8. As shown in Fig. 1E, reduced FBXW8 levels in Fbxw8^+/− MEFs compared to those in wild-type MEFs were detected, and almost no signal was detected in Fbxw8^−/− MEFs (Fig. 1E, upper panel). In darker exposures, Fbxw8^−/− MEFs were found to express a slight amount of FBXW8 relative to that in wild-type or heterozygote MEFs (Fig. 1E, middle panel). Real-time PCR revealed that the level of Fbxw8 mRNA expression in heterozygote MEFs was 57% and that in homozygous mutant MEFs was 0.8% of the level in wild-type MEFs (Fig. 1F).

Growth retardation of Fbxw8^−/− mice.

Similar to Cul7^+/− mice, heterozygous Fbxw8^+/− mice were fertile and indistinguishable in appearance and behavior from their wild-type littermates. An analysis of embryos derived from heterozygote intercrosses revealed that the percentage of Fbxw8^−/− mice varied between 22 and 24% at E12.5 to E18.5, indicating that Fbxw8^−/− mice were viable until late in gestation, although some fetal loss cannot be excluded since the numbers of Fbxw8^−/− as well as Fbxw8^+/− embryos were smaller than expected relative to the number of Fbxw8^+/+ embryos. Relative to heterozygote littermates, Fbxw8^−/− embryos were reduced in size (to 83% of littermate size) at E12.5 and became progressively smaller (to 67% of littermate size) by E18.5 but were otherwise normal in appearance. At each gestational stage, Fbxw8^−/− placentas were also significantly reduced in size. The weight of Fbxw8^−/− placentas was 60 to 67% that of wild-type placentas, whereas the weights of heterozygous and wild-type placentas were similar to each other (Fig. 2A).

All E18.5 embryos recovered from the uteri of Fbxw8^+/− intercross mice were alive, as indicated by spontaneous movement and beating hearts. However, all Fbxw8^−/− E18.5 embryos (n = 15) were unable to initiate or sustain strong and regular breathing and quickly turned cyanotic. In contrast, nearly all Fbxw8^+/+ and Fbxw8^+/− E18.5 embryos began spontaneous breathing. A histological examination of lungs obtained from E18.5 embryos revealed reduced alveolar space in Fbxw8^−/− lungs (Fig. 3B) compared to that in wild-type lungs (Fig. 3A), indicating a failure to inflate. These results are similar to those observed in Cul7^−/− mice, which also showed small embryonic and placental sizes, with neonatal lethality resulting from respiratory failure (2). Although neonatal lethality was completely penetrant for Cul7 null mice, some Fbxw8 null mice survived birth and lived until weaning. Of 203 offspring from Fbxw8^+/− intercrosses weaned at 3 weeks of age, only 17 (8%) were Fbxw8 null, while 58% were heterozygous and 34% were of the wild type (Fig. 2A). This ratio indicated that approximately 70% of Fbxw8^−/− mice died soon after birth.

We closely examined all surviving mice from several litters over a number of months. As shown in Fig. 2B, both male and female Fbxw8^−/− mice were smaller than Fbxw8^+/+ or Fbxw8^+/− mice. Fbxw8^−/− mice were approximately 70% of the weight of the wild-type and heterozygous mice of the same sex at 1 month of age. Serial examination revealed that although Fbxw8^−/− mice gained weight, they grew at a lower rate than their heterozygous and wild-type littermates and remained runted in comparison. Fbxw8^−/− mice appeared normal except for their small size.

Several pairs of Fbxw8^−/− and Fbxw8^+/+ mice were sacrificed at 3 months of age. FBXW8 protein expression was readily detected in most organs, except for the brains, of wild-type but not Fbxw8^−/− mice (Fig. 2C). In addition, the organs in Fbxw8^−/− mice were proportionally smaller, except for the brains, the sizes of which were similar to those of the brains of wild-type mice (Fig. 2D). Detailed histopathological examination, however, revealed no significant abnormalities in Fbxw8^−/− mice (data not shown). Since bone abnormalities in 3-M patients with mutations of CUL7 have been observed frequently, we examined the vertebral bodies (lumbar vertebra 5) and femurs of Fbxw8^−/− mice. Micro-computed tomography scanning of the vertebral bodies and femurs revealed no specific abnormalities in the overall shape (data not shown).

FBXW8 expression in embryos and placentas.

The gene trap vector enabled the determination of the tissue-specific expression pattern of Fbxw8, since the β-geo fusion protein derived from the insertion was under the control of the endogenous Fbxw8 promoter. We assayed for β-Gal expression in histologic sections of Fbxw8^−/− embryos and placentas. In E18.5 embryos, β-Gal was highly expressed in skeletal muscle, with intense staining observed in the diaphragms, intercostals, and abdominal walls (Fig. 3C). Muscles in the extremities were also stained. Smooth muscle cells had slight, if any, β-Gal expression. In addition, β-Gal staining in developing bone was observed (Fig. 3E). Under higher magnification, we observed that β-Gal was expressed in the epiphyses of tubular bones, corresponding to cartilage, but not in the diaphyses, where the formation of bone in E18.5 embryos had already occurred (Fig. 3F). We also detected β-Gal expression in the lungs, specifically in the bronchial epithelia (Fig. 3G). Weak β-Gal expression was detected in kidneys, eyes, and skin, and little, if any, was detected in the brains. Heterozygote E18.5 embryos showed a similar pattern of expression, although at reduced levels compared to those in the homozygote mutant embryos (data not shown). In addition, E14.5 mutant embryos showed a pattern of expression similar to that in E18.5 mutant embryos, with no significant differences (data not shown).

When Fbxw8^−/− mutant placentas were examined, intense β-Gal staining was detected in the labyrinthine layer (Fig. 3H). Moderate expression was also detected in the spongiotrophoblast layer, but not in the decidual layer. We observed a thin and relatively disorganized spongiotrophoblast layer in Fbxw8^−/− placentas compared to that in the wild-type and heterozygote placentas, similar to the spongiotrophoblast layer in Cul7^−/− placentas (Fig. 3I) (2, 32).

FBXW8 protein expression and CUL7 protein expression are mutually dependent.

We have previously detected the specific interaction of CUL7 with FBXW8 by using epitope tag constructs and have reported that the expression of FBXW8 protein was reduced in Cul7^−/− cells compared to that in wild-type MEFs (2, 9). To determine if the endogenous CUL7, FBXW8, and SKP1 proteins interact, immunoprecipitations from MEFs as well as cells from human cell lines were performed with monospecific antibodies against CUL7 and FBXW8 and immunoprecipitates were analyzed by Western blotting. Immunoprecipitation analysis of CUL7 and FBXW8 in mouse and human cells revealed the coprecipitation of these proteins as well as SKP1 (Fig. 4A).

We have reported that in Cul7^−/− MEFs, FBXW8 protein expression was reduced due to a significantly shorter half-life than that in wild-type MEFs (2). To determine if CUL7 expression was affected by the disruption of the Fbxw8 gene, lysates were prepared from wild-type and Fbxw8 null MEFs, as well as from whole E18.5 embryos and placentas. As shown in Fig. 4B, the expression of CUL7 was decreased in Fbxw8^−/− cells and tissues relative to that in wild-type cells and tissues, and conversely, FBXW8 was reduced in Cul7^−/− cells and tissues. To determine if CUL7 protein expression in Fbxw8^−/− MEFs was decreased due to a shorter half-life, protein synthesis was inhibited with cycloheximide. In wild-type MEFs, the level of CUL7 was stable, with no detectable decrease in signal after 8 h of cycloheximide treatment (Fig. 4C). In contrast, CUL7 levels in Fbxw8^−/− MEFs began to decrease after 2 h of cycloheximide treatment and were nearly undetectable at 8 h. We examined the expression of Cul7 mRNA by reverse transcription-quantitative PCR and found similar expression levels in wild-type and Fbxw8^−/− MEFs (data not shown), indicating that the decrease in protein stability accounted for the decreased CUL7 expression in Fbxw8^−/− MEFs. Similar results were confirmed with Cul7^−/− MEFs and tissues, where we observed decreased FBXW8 expression due to a decrease in protein stability (Fig. 4b).

To determine if CUL7 and FBXW8 expression levels were mutually dependent in human cells, we transfected HeLa cells with siRNA against CUL7 or FBXW8. Using two different siRNAs specific for each gene, we observed decreased expression of FBXW8 when CUL7 levels were reduced and reduced levels of CUL7 when FBXW8 was knocked down relative to levels in mock- or control-transfected siRNA-treated cells (Fig. 4D).

Individuals with 3-M syndrome were previously reported to have mutations in both alleles of CUL7 (12). We determined that the FY211 lymphoblastoid B-cell line derived from an individual with 3-M syndrome contains the mutation 263delT, causing a valine-to-alanine substitution followed by a short nonsense sequence (V88A plus 25 residues) resulting in a truncated protein (see Fig. S1 in the supplemental material). We examined the cell line from the 3-M syndrome patient as well as the heterozygote lines from the patient's parents for CUL7 and FBXW8 protein expression. Full-length CUL7 protein was not detected in lysates prepared from the 3-M cells but could readily be detected in the parental cells (Fig. 4E). Notably, the levels of FBXW8 expression in the 3-M syndrome patient cells were also decreased compared to those in the parental cell lines (Fig. 4E). These results strongly support the model in which CUL7 protein expression and FBXW8 protein expression are dependent on each other.

Increased IGFBP1 expression in Fbxw8^−/− embryos.

To explore the phenotype of Fbxw8^−/− embryos, transcriptome profiling of E18.5 Fbxw8^−/− and Fbxw8^+/− whole embryos was carried out. We selected Fbxw8^+/− embryos as controls since they also contained the β-geo integrated gene trap that might affect the expression of some genes. Five Fbxw8^−/− and five Fbxw8^+/− embryos were profiled from an initial batch of three each and a second batch of two each to increase detection power. Though both batches had a large number of genes differentially expressed, when all 10 expression profiles were analyzed together, only Fbxw8, the trapped gene, showed significant differential expression. When results from the two batches of Fbxw8^−/− and Fbxw8^+/− embryos were compared, only nine genes consistently showed differential expression in the same direction, with a q value (false discovery rate if the P value is significant) cutoff of 0.2 (Table 1). In addition to Fbxw8, most of the down-regulated genes were muscle related. The only gene up-regulated in Fbxw8^−/− embryos relative to heterozygous embryos was the IGFBP1 gene. Quantitative PCR analysis confirmed the elevated expression of IGFBP1 mRNA in Fbxw8^−/− embryos compared to that in their heterozygous littermates (data not shown).

Increased IGFBP2 expression in Cul7^−/− and Fbxw8^−/− MEFs.

We had previously noted that fibroblasts from Cul7^−/− embryos grew more slowly than MEFs from their heterozygous or wild-type littermates (2). To determine if Fbxw8^−/− MEFs also grew slowly, we prepared samples of Fbxw8^+/+, Fbxw8^+/−, and Fbxw8^−/− MEFs from littermate embryos and determined their growth rates. Equal numbers of passage 2 MEFs were seeded into replicate plates, and the MEFs were fed every 2 days and counted. Fbxw8^−/− MEFs grew more slowly than Fbxw8^+/+ and Fbxw8^+/− MEFs (Fig. 5A). Similar results were obtained with knockout MEFs from two different litters (data not shown). We attempted to generate immortalized versions of Fbxw8^−/− MEFs by using serial passage in a 3T3 protocol (30). Growth over a 3-day interval (expressed as the N3/N0 ratio) was examined (Fig. 5B). Wild-type MEFs grew relatively well for 10 passages, but Fbxw8^−/− MEFs underwent a severe growth arrest within five or six passages without evidence of crisis. A similar growth arrest of Cul7^−/− MEFs after six passages has been observed previously (data not shown). MEFs were generated from embryos derived from mating Trp53 null mice with Fbxw8^+/− mice. Fbxw8^−/− Trp53^−/− MEFs grew as rapidly as Fbxw8^+/+ Trp53^−/− MEFs and showed no evidence of growth arrest after serial passage (data not shown). In contrast, the Fbxw8 null phenotype, including a small embryo, a dysmorphic placenta, a high incidence of neonatal lethality, and a small adult size, was not suppressed by the loss of Trp53 (data not shown). Therefore, it appears that the poor growth phenotype of Fbxw8^−/− MEFs was strongly dependent on wild-type p53 but that the significant developmental defects observed in Fbxw8^−/− mice were not eliminated by the loss of p53.

Given the growth defect observed in both Cul7^−/− and Fbxw8^−/− MEFs, we asked if any secreted proteins were differentially expressed in the tissue culture supernatant. Cul7^−/− and wild-type MEFs were incubated in reduced-serum medium for 24 h, and supernatants were harvested for analysis by SDS-PAGE. Coomassie blue staining revealed a specific band at 32 kDa that showed consistent up-regulation in Cul7^−/− MEFs compared to the level in wild-type MEFs (Fig. 5C). The analysis of this band by mass spectrometry yielded eight unique peptides specific to mouse IGFBP2. The presence of increased levels of IGFBP2 in the tissue culture supernatant from Cul7^−/− MEFs compared to those in the supernatant from wild-type MEFs was confirmed using a specific antibody (Fig. 5D). The concentration of IGFBP2 in the tissue culture supernatant was determined by enzyme-linked immunosorbent assay under a wide variety of growth conditions. We observed that supernatant from Cul7^−/− MEFs consistently contained two- to threefold-higher levels of IGFBP2 than that from Cul7^+/+ MEFs under a variety of growth conditions (data not shown).

To determine if Fbxw8^−/− MEFs also expressed higher levels of IGFBP2, lysates from wild-type, Cul7^−/−, and Fbxw8^−/− MEFs were blotted with an IGFBP2 antibody. Increased expression of intracellular IGFBP2 was detected in Fbxw8^−/− as well as Cul7^−/− MEFs compared to that in wild-type MEFs from littermate controls (FIG. 5E). Although we consistently detected elevated levels of IGFBP2 in Fbxw8^−/− and Cul7^−/− MEFs, we could not detect the expression of IGFBP1 in MEFs, and there were no differences in the levels of expression of IGFBP3, IGFBP4, or IGFBP5 between wild-type and mutant MEFs (data not shown).

Given the elevated levels of IGFBP2 in Cul7 and Fbxw8 mutant MEFs, we tested whether IGFBP2 could be targeted for ubiquitination and subsequent proteasomal destruction by the SKP1-CUL7-FBXW8 CRL complex. We did not observe any specific binding of IGFBP2 to CUL7 or FBXW8 (data not shown). In addition, the levels of IGFBP2 did not change in either mutant or wild-type MEFs when cells were treated with proteasome inhibitors (data not shown). Furthermore, when Fbxw8^−/− and wild-type MEFS were treated with cycloheximide, we observed that the half-lives of IGFBP2 in wild-type and Fbxw8 null cells were similar (data not shown). Despite the repeated observation that IGFBP2 levels were elevated in Fbxw8^−/− and Cul7^−/− MEFs, we were unable to detect any evidence either that IGFBP2 was a direct substrate for ubiquitination and degradation or that its stability was affected by the CUL7-FBXW8 complex.

The IGFBPs serve to modulate the availability of IGFs to activate the IGF1R tyrosine kinase and downstream signal transducers. To determine if the increased levels of IGFBP2 affected IGF1R signaling, MEFs were serum starved before stimulation with IGF1. In the absence of IGF1, the tyrosine phosphorylation of IGF1R was undetectable in wild-type as well as Cul7^−/− and Fbxw8^−/− MEFs. After stimulation with IGF1 for 15 min, tyrosine-phosphorylated IGF1R was readily detected in wild-type MEFs but was detected at reduced levels in Cul7^−/− and Fbxw8^−/− MEFs (Fig. 5F). The levels of expression of IGF1R in all cells were similar regardless of IGF1 stimulation. The addition of neutralizing antibodies to IGFBP2 to the tissue culture medium failed to affect the growth rate of knockout MEFs. In addition, the RNA interference-mediated knockdown of IGFBP2 did not affect the proliferation index or the cell cycle profile of Fbxw8^−/− MEFs. Although elevated levels of IGFBP2 in Fbxw8^−/− and Cul7^−/− MEFs were consistently observed, we were unable to demonstrate that IGFBP2 contributed to significant growth delay.