Ribosomal protein L24 defect in Belly spot and tail (Bst), a mouse Minute

The Minutes are an intriguing group of eukaryotic mutations that have been valuable in elucidating concepts of cell autonomy, compartmental development and cell competition(Garcia-Bellido et al., 1973; Morata and Ripoll, 1975; Moreno et al., 2002; Stern and Tokunaga, 1971). First described in Drosophila melanogaster more than 80 years ago,this dominantly inherited class of mutants comprises over 50 loci scattered across the Drosophila genome(Lambertsson, 1998; Schultz, 1929). Despite the large degree of locus heterogeneity, Minutes are remarkably similar,sharing three principal phenotypes – a 2-3 day delay in larval development, short thin bristles and recessive lethality. Variable phenotypes include small body size, female sterility, large or rough eyes, and abnormal wings. The haploinsufficiency and nonadditivity of these mutations, and their phenotypic similarity, led Schultz(Schultz, 1929) to propose that each Minute represents a unique member of a common metabolic or biochemical pathway. This prescient prediction was confirmed, as all of the molecularly cloned Minute loci to date encode ribosomal proteins(Lambertsson, 1998).

Although riboprotein mutations have also been identified in Escherichia coli and Saccharomyces cerevisiae(Warner, 1999), little is known about riboprotein mutations in mammals. The mammalian ribosome comprises 79 ribosomal proteins and four rRNAs, which combine in equimolar ratios to form the small (40S) and large (60S) subunits. Each riboprotein is encoded by a single gene. Despite the large number of riboprotein genes(Uechi et al., 2001), only one is known to be mutated in mammals. RPS19, a component of the small ribosomal subunit, is mutated in a quarter of humans with Diamond–Blackfan anemia (DBA, MIM205900), a congenital red blood cell hypoplasia (Draptchinskaia et al.,1999). In addition to anemia, these patients commonly have low birth weight, small stature, and craniofacial and skeletal defects. Aside from reduced somatic growth, it is unclear how these tissue-specific phenotypes result from mutations in a gene that is universally required for protein synthesis.

Belly spot and tail (Bst) is a semidominant, homozygous lethal mutation. Heterozygotes have decreased pigmentation and a kinked tail(Southard and Eicher, 1977). Bst/+ mice also have a significant retinal abnormality, characterized by delayed closure of the choroid fissure, marked deficiency of ganglion cells and subretinal neovascularization (Rice et al., 1995; Rice et al.,1997; Smith et al.,2000; Tang et al.,1999). Accordingly, Bst has been proposed as a model for human optic nerve atrophy and age-related macular degeneration.

Here, we report that the riboprotein gene Rpl24 is mutated in Bst/+ mice. The mutation impairs Rpl24 mRNA splicing and L24 production, which in turn affects ribosome biogenesis, protein synthesis and the cell cycle. Bst/+ cells have a significant growth disadvantage in chimeras that is similar to the classical cell competition effect observed in Drosophila Minute somatic mosaics. These findings establish Bst as a mouse Minute.

C57BLKS Bst/+ and +/+ mice were weighed daily from P5 to P21 and at 8 weeks of age. Skeletal preparations were stained with Alcian Blue and Alizarin Red (McLeod, 1980)and examined at P0 and P5. Paraffin sections (3 μm) of adult eyes were immunostained with a neurofilament (NF160) monoclonal antibody (NN18, 1:500,Sigma) as described (Brown et al.,2001).

Mouse strains were obtained from The Jackson Laboratory (Bar Harbor, ME). An intersubspecific backcross was performed between (C57BLKS Bst/+× CAST/Ei) F1 and C57BLKS +/+ mice. N2 progeny were scored at weaning for white spotting, white hind feet and tail kinks, and were genotyped using MIT markers (Dietrich et al.,1992) and four new microsatellites:

• D16Ero1

• Forward: 5′-CTTTCTCTAATCATGGCAAAAACA-3′

• Reverse: 5′-TGAGCAAGACAATCACAGAGTATG-3′

• D16Ero2

• Forward: 5′-GAGCAAAATTTACGATGTGATTTG-3′

• Reverse: 5′-ATCTCTATTGCACAAAGATGGACA-3′

• D16Ero4

• Forward: 5′-ACATAGTCAACAGACCGGACCT-3′

• Reverse: 5′-ATGTTCTACTCGGTCTCTCCATCT-3′

• D16Ero7

• Forward: 5′-GCATTTAGTCAACACCATCAGAAT-3′

• Reverse: 5′-CAAGCAAGCATTTCATACAAAAGT-3′.

PCR products were separated using 6% polyacrylamide sequencing or 4%Metaphor agarose gels. To confirm Bst genotypes, critical recombinants and N2 mice with low expressivity were test-crossed to C57BLKS. Opa1 was genotyped using PCR primers 5′-GAGCAGAAGAGCCCTGGAC-3′ and 5′-CAATGGAGGGTGGATGTTTC-3′ and a BlpI polymorphism in the 3′UTR. To type Hes1, a genomic segment corresponding to the terminus of bacterial artificial chromosome (BAC) clone RP24-110I11 was amplified using primers 5′-TTATTTGTTTCTACGACCCAGCTT-3′ and 5′-ACTGGTCTTGAACTCCTAGACTCC-3′ and scored for a single-strand conformational polymorphism (SSCP) by MDE gel electrophoresis (FMC Bioproducts).

The Bst allele was typed by PCR, using upstream primer 5′-TTTGCAGCGCACATACGAG-3′ which overlaps the first intron branchpoint deletion and downstream primer 5′-GCTGACTCACATTTTGCATTAAGA-3′ within the third exon. The wild-type allele was amplified using upstream primer 5′-CTCTTTGCAGCGCACATACTAAC-3′, which overlaps the intact first intron branchpoint and downstream primer 5′-GGAAAACCTGCAGTTAACAAATTC-3′ within the second intron. PCR parameters were 3 minutes at 95°C, followed by 40 cycles (30 seconds at 95°C, 1 minute at 56°C, 1 minute at 72°C) and 7 minutes extension at 72°C.

Total RNA was extracted using Trizol reagent (Invitrogen) from livers of C57BLKS+/+, C57BLKS Bst/+ and SPRET/Ei adult mice; livers of P21 weanlings and embryonic day (E) 18.5 embryos derived from a (C57BLKS Bst/+ × CAST/Ei) F1 × SPRET/Ei cross; and heads of co-isogenic C57BLKS Bst/+ and +/+ E13.5 embryos. First strand cDNA synthesis was primed using random hexamers. To evaluate Bst candidate genes, 40 cycles of RT-PCR were performed on E13.5 head and body RNA using gene-specific primers. To evaluate Bst effects on Rpl24splicing, three RT-PCRs were performed (designated A, B and C in Fig. 3D,E). These reactions share a common downstream primer (5′-CTGTCTTCTTTGATGCCTGCTTAG-3′),which was ³²P or HEX end-labeled. The upstream primers were 5′-GCAGGCCGACATCTATCAC-3′ (A),5′-GGATGGCTCCTCTTTGCAG-3′ (B) and 5′-ACATCTATCACCATGAAGGTCGAG-3′ (C). The upstream primer in reaction C spans the exon 1-2 junction and only amplifies from correctly spliced Rpl24 mRNA. PCR parameters were 3 minutes at 95°C,followed by 35-40 cycles (30 seconds at 95°C, 1 minute at 56°C, 1 minute at 72°C) and 7 minutes at 72°C. To eliminate heteroduplex products in reaction C, a final 10-minute extension step was performed after adding fresh PCR reagents. Samples were then digested overnight with AluI at 37°C. ³²P-labeled products were denatured at 95°C, separated on 6% polyacrylamide sequencing gels and analyzed using a Phosphorimager (Amersham Biosciences). HEX-labeled products were separated on an ABI sequencer.

BAC clone RP23-10L10 (accession no. AZ117821) contains the 5.2 kb Rpl24 transcription unit flanked by 93 kb 5′ and 83 kb 3′genomic DNA. Purified BAC DNA was injected into pronuclei of (C57BL/6 ×SJL) F2 eggs, generating 17 transgenic founders. The integrity of the BAC ends was assessed in founders by PCR, using primer pairs that span the vector-insert junctions:

• 10L10 Prox-2

• Forward: 5′-TCGAGCTTGACATTGTAGGACTAT-3′

• Reverse: 5′-AGGGTGGTTTCAGTGTGTCG-3′

• 10L10 Dist-1

• Forward: 5′-GATGTTCATGTTCATGTCTCCTTC-3′

• Reverse: 5′-CCCTCAATCTCATTAACCTTCTGT-3′.

To confirm the presence of Rpl24 within each transgene, we crossed founders to SJL mice and compared the ratio of SJL and C57 alleles in offspring using a Sau3A1 RFLP within intron 4 or a microsatellite marker (D16Ero9) located 1 kb upstream of Rpl24. D16Ero9 was typed using primers 5′-CTGGCTAGCATAGCTTTTCTTTTT-3′ and 5′-GGAAAGGTGAGCAGGTTAAGAATA-3′. To evaluate transgene correction,founders were crossed to C57BLKS Bst/+ mice. The resulting progeny were examined for Bst phenotypes and assessed using BAC-specific and Rpl24 allele-specific PCRs.

The human RPL24 cDNA was amplified by PCR from IMAGE clone 5609137(Open Biosystems) using an upstream primer with an engineered BamHI site (5′-CGTGGATCCGTCGCCATGAAGGTCGAGCTGTG-3′) and a downstream primer with an engineered EcoRI site(5′-TTTAGAATTCTAATCTGCCAGTTTAGCGTTTTCC-3′). Following digestion, the resulting 0.5 kb BamHI-EcoRI fragment was subcloned into BglII- and EcoRI-digested pBROAD3(Invivogen). The final construct (R26-huL24) contains the entire 471 bp RPL24 coding sequence, is transcribed from the 1.9 kb murine ROSA26 promoter, and terminates in a β-globin polyadenylation site(Fig. 4D). The 3.0 kb transgene was excised with PacI, gel purified and injected into pronuclei of coisogenic C57BLKS Bst/+ and +/+ eggs. Transgenic mice were identified by PCR using primers 5′-ACAGGTGTGAAACAGGAAGAGAAC-3′ and 5′-GGGAACAAAGGAACCTTTAATAGA-3′, which selectively amplify the human cDNA construct.

Sex-matched adult littermates were fasted for 48 hours and re-fed for 2 hours. This fasting protocol decreases liver mass twofold(Hutson and Mortimore, 1982; Volarevic et al., 2000). Immediately before refeeding, each mouse was given an intraperitoneal injection of ³²P-orthophosphate (30 μCi per gram body weight). Livers were snap frozen after re-feeding. Total liver RNA (1-10 μg per lane) was separated by formaldehyde agarose gel electrophoresis, transferred to a nitrocellulose membrane and exposed to X-ray film (Biomax, Kodak). Density tracings were analyzed using NIH Image software (Scion). The membrane was stained with Methylene Blue to confirm equal RNA loading.

Livers from sex-matched adult littermates were homogenized in ice-cold hypotonic lysis buffer (1.5 mM KCl, 2.5 mM MgCl₂, 5 mM Tris-HCl pH 7.4, 1% Na deoxycholate, 1% Triton X-100) using 1 ml buffer per 10 mg liver. After centrifugation for 15 minutes at 2500 g, heparin (1 mg/ml) and cycloheximide (100 μg/ml) were added, and 700 μl of the supernatants were applied to 13 ml 7-47% (w/v) sucrose gradients. Samples were centrifuged in a Beckman SW-41 rotor at 288,000 g for 2 hours at 4°C. Gradients were collected using a Biocomp fractionator with UV photometer.

Murine embryonic fibroblast (MEF) cultures were generated at E13.5 from co-isogenic C57BLKS Bst/+ and +/+ embryos and from (CAST/Ei ×C57BLKS Bst/+) F2 embryos as described(Robertson, 1987) and tested at equivalent early passages (<P8). Experiments were performed in triplicate unless otherwise noted.

To evaluate protein synthesis, MEFs were plated at a density of 5×10⁴ cells per 35 mm dish. After 24 hours, cells were incubated for 2 hours in 2 ml fresh media containing 10 μCi[³H]leucine, washed with ice-cold phosphate buffered saline (PBS),and treated with 5% trichloroacetic acid (TCA) for at least 2 hours at 4°C. [³H]leucine incorporation was measured after lysis in 0.5 N NaOH, 0.5% sodium dodecyl sulphate (SDS) and was normalized to mitochondrial dehydrogenase activity in parallel cultures, measured using a water soluble tetrazolium (WST) assay (Roche).

To measure cell-doubling time, parallel cultures were plated in complete media at a density of 5×10⁴ cells per 35 mm dish. Starting 36 hours after plating, adherent cells were harvested at 12-hour intervals and counted using a Coulter Z2 counter. The growth rate was determined between 36 and 60 hours after plating.

Bivariate cell cycle analysis was performed on asynchronous cultures using a BrdU Flow Kit (BD Pharmingen). Cultures were labeled with 10 μM BrdU for 75 minutes, stained with an FITC-conjugated anti-BrdU antibody (new DNA synthesis) and 7-AAD (total DNA content), and sorted using a Coulter Elite ESP Fluorescence Cell Sorter. The fraction of cells in G1, S and G2/M is directly related to the duration of each phase in the cell cycle(Gray et al., 1990). Flow cytometry data were analyzed using WinMDI v.2.8 software.

To evaluate S-phase entry after serum starvation, parallel MEF cultures were plated in complete media at 5×10⁴ cells per 35 mm dish. After plating, cells were washed twice with serum-free media and refed Dulbecco's modified Eagle's medium (DMEM) containing 0.5% fetal bovine serum(FBS) for 48 hours. Cultures were released from serum starvation(Herrera et al., 1996) by the addition of complete DMEM containing 15% FBS. DNA synthesis was measured periodically by adding 0.8 μCi [³H]thymidine per dish for 1 hour. These cultures were washed twice with ice-cold PBS and treated with 5%trichloroacetic acid (TCA) for at least 3 hours at 4°C. The cells were washed, lysed in 0.5 N NaOH, 0.5% SDS and counted for [³H]thymidine incorporation. This experiment was repeated using MEFs derived from three separate litters.

E3.5 blastocysts were harvested as described(Robertson, 1987) from 24- to 30-day-old C57BLKS +/+ females that had been superovulated and mated to C57BLKS Bst/+ males. Blastocysts were microinjected with exactly five or ten ROSA26 embryonic stem (ES) cells, which were between passage 15 and 18(kindly provided by Liz Robertson and Phil Soriano). After reimplantation into pseudopregnant (C57BL/6 × DBA) F2 females, chimeras were analyzed at E12.5 or after birth.

The percent agouti coat color was estimated independently by our lab and the University of Michigan transgenic core at the time of weaning. Rpl24 genotypes were determined afterward from tail DNA by allele-specific PCR. To further evaluate the extent of chimerism, adult tissues were fixed at 4°C for 2-3 hours in 2% paraformaldehyde, 0.2%glutaraldehyde 0.1 M NaPO₄ (pH 7.3), 0.02% NP40, 0.02% Na deoxycholate; cryopreserved through 10-30% sucrose in PBS with 5 mM MgCl₂; and embedded in optimal cutting temperature (OCT) medium(TissueTek). Cryosections (10 μm) were stained for β-galactosidase in 100 mM NaPO₄ (pH 7.3), 3 mM K₃Fe(CN)₆, 3 mM K₄Fe(CN)₆, 1.3 mM MgCl₂, 0.1% X-gal for 7 hours at 37°C, counterstained with neutral red, dehydrated and mounted in Permount.

Genomic DNA was extracted from E12.5 embryos, amplified by PCR, and tested using an XbaI polymorphism in the βh₁ locus that distinguishes 129/SvEv (ROSA26) and C57BLKS alleles(Fiering et al., 1995). The relative abundance of alleles was determined using a Phosphorimager.

Bst/+ mice have a pleiotropic phenotype that includes a white ventral midline spot, white hind feet and a kinked tail(Fig. 1A). Bsthomozygotes die before E9.5 (data not shown). Abnormal retinal histogenesis in Bst/+ mice causes a 30-100% reduction in the number of retinal ganglion cells (RGCs), with some laminar disorganization, hypoplastic or absent optic nerves, and occasional optic nerve colobomas(Fig. 1B,C)(Rice et al., 1997). In addition to tail kinks, there are other skeletal abnormalities. Bst/+mice frequently have a triphalangeal first digit and an extra preaxial digit(Fig. 1D). They also have six lumbar vertebrae, compared with the five in co-isogenic C57BLKS controls (data not shown). Skeletal stains show fewer caudal vertebrae in Bst/+ mice and fused or wedge-shaped hemivertebrae at the sites of tail kinks (data not shown).

Bst/+ mice appear smaller than wild-type littermates. To determine the onset and magnitude of this size difference, we compared growth curves and adult weights of co-isogenic cohorts. The Bst/+ mice were 20% smaller than wild-type littermates through the first 3 postnatal weeks and remained so into adulthood (Fig. 1E,F).

Bst arose spontaneously in the inbred C57BLKS strain(Southard and Eicher, 1977)and was mapped to a 10 cM interval on mouse chromosome 16(Rice et al., 1995). To further localize Bst, we performed an intersubspecific backcross between (C57BLKS Bst/+ × CAST/Ei) F1 and C57BLKS+/+ mice. N2 progeny were scored for Bst pigmentation and tail phenotypes and genotyped with informative microsatellite DNA markers. Among 828 N2 progeny,386 were Bst/+. This segregation ratio (386:442) is less than 1:1(P=0.092), consistent with decreased viability of Bst/+mice. Bst maps within a 0.5 cM interval, delimited by D16Mit199 and D16Ero1(Fig. 2A). Our data formally exclude Opa1, the mouse homolog of the human optic nerve atrophy gene(MIM605290), and the Hes1 basic helix-loop-helix transcription factor, which has roles in retinal and somite development(Saga and Takeda, 2001; Tomita et al., 1996). Both genes lie more than 5cM centromeric to Bst(Fig. 2A).

The non-recombinant interval spans 1.5 Mb(Fig. 2B) and contains at least 18 known or predicted genes (Ensembl mouse and UC Santa Cruz human databases,April 2002). Eleven are expressed in E13.5 head and E15.5 eye (data not shown)and were evaluated as Bst candidates by comparing genomic DNA sequences from C57BLKS Bst/+ and +/+ mice. These studies revealed only a single change, within Rpl24(Fig. 3A,B).

Rpl24 encodes a protein component of the large (60S) ribosomal subunit. L24 is a highly basic protein of 157 amino acids and is conserved among eukaryotes and archaebacteria. Its position near the polypeptide exit and elongation factor binding sites (Ban et al., 2000), its contacts with the 40S subunit(Spahn et al., 2001) and the properties of L24 mutant yeast (Dresios et al., 2000) suggest that L24 helps catalyze the peptidyl transferase reaction. Rpl24 spans six exons, extends over 5.2 kb and contains a characteristic 5′ terminal oligopyrimidine (TOP) initiator element (Hariharan et al.,1989). More than 20 processed Rpl24 pseudogenes are interspersed throughout the genome (data not shown), similar to other riboprotein genes (Zhang et al.,2002). However, only one of these loci, within the Bstcritical interval, contains introns and is therefore the bona fide Rpl24 gene.

Bst destroys the first intron splice branchpoint of Rpl24, which matches the consensus sequence (UACUAAC) exactly. The deletion removes four nucleotides, including the critical adenosine(underlined) that reacts to form the lariat intermediate(Fig. 3A,B). No other obvious branchpoints are present within this 108 bp intron. Using an allele-specific PCR assay, we determined that the deletion in Rpl24 is unique to the Bst mice (Fig. 3C). This branchpoint is intact in the parental C57BLKS strain, highly related C57 strains, different Mus species (M. spretus, M. castaneus and M. molossinus) and rat Rpl24 (accession no. X78443).

To determine how Bst affects Rpl24 splicing, we performed RT-PCR experiments (Fig. 3D). Using primers in exons 1 and 5, we amplified a novel 500 bp product from Bst/+ RNA, which reflects inclusion of intron 1. No other products were detected, including a 320 bp product that would result if exon 2 were skipped. This result was confirmed by direct PCR sequencing and a second RT-PCR using primers in intron 1 and exon 5(Fig. 3D). Rpl24translation begins in the first exon (Fig. 3A). Retention of intron 1 would cause premature termination of the L24 polypeptide after the first two codons. The first intron contains two other stop codons and the next AUG codon (M91) is in the fourth exon.

Activation of a cryptic branchpoint could restore proper splicing(Ruskin et al., 1985). Because Bst homozygotes are not available, we evaluated Rpl24splicing in interspecies hybrid mice, derived from a (C57BLKS Bst/+× CAST/Ei) F1 × SPRET/Ei cross, by a PCR strategy, using a polymorphic AluI restriction site in exon 4 and an upstream primer spanning the exon 1-2 junction. This allowed us to resolve and compare the amount of correctly spliced Rpl24 transcripts from wild-type and Bst alleles (Fig. 3E-G). These results show that 20-25% of Bst transcripts are correctly spliced. Bst is therefore hypomorphic. Inbred C57BLKS Bst heterozygotes and homozygotes thus have approximately 60% and 20%of normal Rpl24 transcript levels, respectively.

Like the Drosophila Minutes, Bst has general effects on somatic growth and specific effects on different tissues(Fig. 1). To confirm that the Rpl24 mutation is solely responsible for the pleiotropic Bstphenotypes, we created transgenic mouse lines expressing wild-type Rpl24 from a 181 kb mouse BAC (RP23-10L10) or a human cDNA expression cassette (Fig. 4A,D). Four BAC transgenic lines were bred to C57BLKS Bst/+ mice(Fig. 4C). Founders Tg321 and Tg326 have intact transgenes that completely correct the Bstphenotypes (Fig. 4B,C). Founders Tg898 and Tg901 have partial BAC transgenes lacking Rpl24and these fail to correct the Bst phenotype. We generated cDNA transgenics using human RPL24 and the ubiquitously active murine ROSA26 promoter (Zambrowicz et al.,1997). Human and mouse L24 proteins are identical. We injected co-isogenic C57BLKS Bst/+ and +/+ eggs with the R26-huL24 construct(Fig. 4D) and recovered seven transgenic founders. Three of the four Bst/+ founders were indistinguishable from wild-type littermates (data not shown), proving that RPL24 alone is sufficient to correct the Bst phenotype and that Rpl24 is Bst. Taken together, our data show Bst acts via a partial loss-of-function mechanism and should be considered a mouse Minute.

Ribosome biogenesis requires highly coordinated synthesis of ribosomal proteins and rRNAs (Warner,1999). These are assembled in a tightly ordered process involving more than 200 accessory proteins and small nucleolar ribonucleoproteins(Fatica and Tollervey, 2002). Decreased availability of individual components can impair these events(Saveanu et al., 2001; Volarevic et al., 2000; Zhao et al., 2003).

Although large and small subunits assemble independently, transcription and processing of the 45S rRNA precursor into 18S and 28S rRNAs couples formation of both subunits (Fatica and Tollervey,2002). As such, rRNA processing is a sensitive indicator of ribosome biogenesis. We therefore examined rRNA synthesis in a pulse labeling experiment. Sex-matched co-isogenic C57BLKS Bst/+ and wild-type littermates were fasted for 48 hours, injected with ³²P-orthophosphate and re-fed for 2 hours. Profiles of unlabeled(steady state) rRNA were similar after re-feeding(Fig. 5A, Methylene Blue), but the amount of newly processed rRNA in the Bst/+ mouse was dramatically decreased (Fig. 5A, autoradiogram). The 28S rRNA was reduced to a greater extent than the 18S rRNA, consistent with a defect in large subunit assembly(Fig. 5A, densitometry). In yeast, mutations specific to the 60S subunit can affect 18S rRNA processing indirectly, most probably through the exosome(Zanchin et al., 1997). This nuclease complex degrades improperly spliced mRNAs, excess rRNA precursors and unused rRNA intermediates, and is required for rRNA processing(Allmang et al., 2000; Butler, 2002; Mitrovich and Anderson, 2000). Likewise, ribosome subunits are stable once assembled and most unincorporated riboproteins are rapidly eliminated(Warner, 1977). Bstis therefore unlikely to affect the abundance of L24 at steady state. The labeling of the 34S and 45S pre-rRNAs was similar in both mice(Fig. 5A, autoradiogram),suggesting that the decrease in 28S rRNA is due to increased turnover rather than a specific bottleneck in rRNA processing. This cellular response is consistent with a dynamic shortage of L24 polypeptide.

There were no changes in the ratio of small to large subunits or in the polyribosome fraction of Bst/+ liver homogenates(Fig. 5C). However, a prominent shoulder between the 60S subunit and 80S monosome peaks was consistently observed in sucrose gradients from Bst/+ mice, compatible with an alteration in 60S subunit structure. This change was observed under fed,fasted and re-fed conditions (data not shown).

In spite of its apparent tissue specificity, L24 deficiency should affect every cell in the Bst/+ mouse, in the same way that cell growth is universally slower in Minute flies(Simpson and Morata, 1981). To test this hypothesis, we examined cell size, protein synthesis, proliferation rates and cycle kinetics in exponentially growing mouse embryonic fibroblast(MEF) cultures. We found no difference in size between Bst/+ and wild-type MEFs by flow cytometry (forward light scatter, data not shown). This is consistent with recent studies in Drosophila showing no size difference between Minute and wild-type imaginal disc cells(Montagne et al., 1999). We examined overall protein synthesis in MEFs by [³H]leucine incorporation, normalized to cell viability(Fig. 6A). Bst/+cultures showed a 31.2±5.8% reduction in the rate of protein synthesis,similar to the 30% reduction described for M(3)i⁵⁵/+Drosophila Minute embryos (Boring et al., 1989).

Bst/+ MEFs proliferate significantly more slowly than wild-type cultures. The population doubling time (T_d) is 25.8±0.4 hours for Bst/+ MEFs compared with 21.6±0.2 hours for wild-type MEFs (Fig. 6B). To determine what part of the cell cycle is elongated, we performed bivariate flow cytometry on asynchronous MEF cultures using BrdU incorporation and DNA content as sorting parameters. The ratio of G1 to S-phase cells was 1.45 (±0.05) for Bst/+ MEFs and 1.04(±0.01) for control MEFs (Fig. 6C). This suggests that Bst causes actively growing cells to remain longer in G1 phase. A small increase in the G2/M fraction was also observed. If the absolute length of S phase is constant (9.1 hours) and the cell cycle is increased by 20% overall(Fig. 6B), then G1 phase is increased by 40% in Bst/+ MEFs(Fig. 6E).

To further investigate the prolongation of G1 by Bst, we compared the time required for Bst/+ MEF cultures to enter S phase after serum stimulation. Parallel cultures were arrested in G1 by serum starvation,released by the addition of serum-rich media, and assayed periodically for DNA synthesis by [³H]thymidine incorporation. We observed a 2.4±0.5 hour increase in the time required to reach half-maximal incorporation in Bst/+ cultures compared with wild type(Fig. 6D). These results confirm the increased length of G1 in Bst/+ MEFs and are consistent with classical studies in which the protein synthesis inhibitor cycloheximide was shown to specifically lengthen G1 phase following serum stimulation(Brooks, 1977).

Among the most striking features of Drosophila Minutes is the competition that occurs between M/+ and wild-type cells in somatic mosaics (Garcia-Bellido et al.,1973; Morata and Ripoll,1975; Simpson,1979). The increased doubling time of Bst/+ MEFs suggests that a similar competition effect may occur between mutant and wild-type cells in developing mice. To test this prediction, we generated chimeras by injecting exactly five or ten wild-type ROSA26 (R26) embryonic stem (ES) cells into a/a (non-agouti) C57BLKS Bst/+ and +/+ E3.5 blastocysts (Fig. 7A). The R26 ES cells are derived from the A/A (agouti) 129/SvEv strain and contain a ubiquitously expressed lacZ enhancer trap(Zambrowicz et al., 1997). In every experiment, we found that a greater fraction of mice derived from Bst/+ blastocysts were detectably chimeric (≥5% agouti coloration). These chimeras had a greater agouti coat content(Fig. 7B,C) and more extensive lacZ staining throughout the body(Fig. 7D). Some tissues, such as brain, have a greater R26 contribution than others, such as liver,regardless of the recipient blastocyst genotype. This is most likely due to strain differences between C57 and the 129 ES cells. A similar strain effect was described in neural tissues of interspecific aggregation chimeras(Goldowitz, 1989).

To assess competition during embryogenesis, we determined the relative amount of C57 single and 129 diffuse βh₁ globin alleles in total E12.5 genomic DNA. The fractional R26 contribution was 0.43±0.06 for +/+ chimeras and 0.63±0.12 for Bst/+chimeras injected with 10 ES cells (P=0.18, n=11). While these values suggest that some differential growth has occurred by E12.5, most of the competition appears to take place between late gestation and weaning. About half of embryos and adults had no detectable ES cell contribution(Fig. 7B and data not shown),in keeping with the efficiency of chimera recovery reported elsewhere(Gossler et al., 1986). These combined results show that Bst/+ cells have a significant growth and survival disadvantage in the presence of wild-type cells, and this extends throughout the animal.

The Rpl24 mutation causes a kinetic defect in ribosome biogenesis. Bst/+ mice have features remarkably similar to Drosophila Minutes. Our chimera studies show that the cell competition first described in Drosophila also occurs in mice. In Minute fly larvae, wild-type somatic clones outcompete surrounding M/+ cells,filling and so defining entire compartments of the wing, while respecting developmental boundaries (Garcia-Bellido et al., 1973; Morata and Ripoll, 1975; Simpson,1979). Conversely, M/+ clones arising in wild-type flies are selectively eliminated, by a TGFβ signaling mechanism that recognizes cells with slower proliferation rates(Moreno et al., 2002). This active process occurs in some, but not all, cell types and serves to maximize tissue fitness (Abrams, 2002). Future studies are needed to show whether cell competition in mice is an active process similar to that in the fly wing.

If wild-type cells outcompete Bst/+ cells in the germline as expected, our results may have practical implications for generating high level ES cell chimeras, as a useful alternative to tetraploid recombination methods (Nagy et al., 1993; Wang et al., 1997). Bst/+ blastocysts may represent ideal hosts, as the introduced ES cells should readily populate the embryo.

Riboprotein mutations appear to be rare in mammals, in contrast to bacteria, yeast and Drosophila. The paucity of mutations can be explained if heterozygous phenotypes are too mild, too severe, heterogeneous or otherwise poorly ascertained. The fact that Bst is hypomorphic suggests that a threshold level of gene expression might be required for survival. Null mutations may be lethal or subvital. Likewise, ribosomes may tolerate mutations in few of their components. L24 does not exist in eubacteria and is one of seven dispensable riboproteins in yeast(Baronas-Lowell and Warner,1990). Yeast lacking L24 have an increased doubling time and a reduced peptidyl transferase rate(Baronas-Lowell and Warner,1990; Dresios et al.,2000). These observations suggest that L24 enhances translation efficiency but does not have a unique structural role. Indeed, L24 is added to the 60S subunit relatively late, most probably during export through the nuclear pore or in perinuclear cytoplasm(Harnpicharnchai et al., 2001; Saveanu et al., 2001). A special case of redundancy might also exist for L24. An evolutionarily conserved paralog, Rlp24 (ribosomal-like protein 24), associates with pre-60S complexes (Saveanu et al.,2001). Mouse Rlp24 and L24 are 31% identical and equally similar to the archaean L24 homolog. It has been proposed that Rlp24 plays an essential role in ribosome biogenesis and that Rlp24 and L24 sequentially occupy the same docking site on the 60S subunit, in the nucleus and cytoplasm,respectively (Saveanu et al.,2001). If Rlp24 can reach the cytoplasm, it might in principle partially substitute for L24 in the mature ribosome. Alternatively, a large number of unrecognized riboprotein mutations may exist in mice as they do in Drosophila, with semidominant spotting, tail defects and small size as core phenotypes, and homozygous lethality(Hrabe de Angelis and Balling,1998; Morgan,1950; Tease and Fisher,1993).

Since Rpl24 is essential for protein translation, Bsthomozygotes are likely to die at or before the blastocyst stage, when the pool of maternal ribosomes is exhausted (Copp,1995), or slightly later, owing to low-level Rpl24expression from the Bst allele(Fig. 4E-G). In a normal mouse blastocyst, ≤20% of the 2.5×10⁸ ribosomes originate from the oocyte (Piko and Clegg,1982) and ribosomal proteins account for 8% of total protein synthesis (LaMarca and Wassarman,1979). By contrast, 0-nu Xenopus mutants lacking rRNA genes can survive to the tadpole stage using maternal ribosomes only(Brown and Gurdon, 1964).

The specific defects in Bst/+ mice(Fig. 1) are difficult to reconcile with a global reduction in protein synthesis. Developing retinas,somites and melanocytes may be hypersensitive because mRNAs critical for these tissues are translated at or near a required threshold. Indeed, different mRNAs are translated with widely varying rates of initiation and elongation(Dever, 2002; Lodish, 1976). However, Bst phenotypes may be more easily explained by tissue differences in(1) the extent of coupling between cell division and organogenesis, (2) levels of riboprotein gene expression or (3) extraribosomal functions.

The cell cycle alterations in Bst/+ mice(Fig. 6) provide the basis for the first model. In Drosophila, mutations in the promitotic dmyc gene are phenotypically similar to Minutes(Johnston et al., 1999). Mice with graded c-myc alleles are likewise small, have short tails and prolonged cell cycles (Trumpp et al.,2001). Bst phenotypes could arise in tissues in which the timecourse of organogenesis is relatively uncoupled from cell proliferation. During the life cycle of Minute flies, delayed eclosion allows for complete development. In mice, however, gestation cannot be prolonged; Bst/+ pups are born at the same time as wild-type littermates. Organogenesis must proceed within the time allotted to ensure viability and must be relatively uncoupled from cell proliferation. In principle, defects could result from situations in which development outpaces cell proliferation. Such a dyschronic mechanism may explain the phenotypes in the retina and vertebral column, where timing of developmental events is particularly critical.

In the case of the developing retina, a single progenitor population gives rise to seven major cell types in a characteristic order(Turner et al., 1990). The progenitors are thought to proceed through a sequence of competency states,each of which favors the differentiation of one or a few particular cell types(Livesey and Cepko, 2001). RGCs are the first-born retinal neurons in all vertebrate species(Altshuler et al., 1991). Although histogenesis can occur when cell division is blocked(Harris and Hartenstein,1991), the timing of cell cycle exit greatly influences retinal development and progenitor cell fate (Dyer and Cepko, 2001). Furthermore, cell cycle length is significantly modulated during the normal course of retinal histogenesis(Alexiades and Cepko, 1996; Li et al., 2000). In the Bst/+ retina, there is no change in the fraction of cells undergoing proliferation or apoptosis, but the total number of neuroblasts is reduced by roughly 60% (E10.5) to 20% (E13.5), and the onset of terminal mitoses is delayed from E10.5 to E12.5 (Tang et al.,1999). These findings can now be considered in light of the general increase in Bst/+ cell-cycle length(Fig. 6). Assuming our MEF results apply in vivo, Bst is likely to have a disproportionately greater effect during the early stages of retinal histogenesis, when the cycle time is shorter (Alexiades and Cepko,1996), protein synthetic demands are greatest, and RGCs are born(Livesey and Cepko, 2001). The subretinal neovascularization (Smith et al., 2000) is probably a secondary consequence of RGC deficiency,similar to that observed in Math5 mutant mice and humans with optic nerve aplasia (Brown et al.,2001; Lee et al.,1996; Brzezinski et al.,2003).

Timing is similarly critical during somite development. Somites are added iteratively to the caudal end of the neural tube, in response to periodic waves of gene expression (Saga and Takeda,2001). This process is controlled by an intrinsic clock that is sensitive to cycloheximide (Hirata et al.,2002; Palmeirim et al.,1997). The vertebral defects in Bst/+ mice are more severe posteriorly, consistent with a progressive deterioration in periodicity and phase coherence of the intrinsic somite clock(Saga and Takeda, 2001). This pattern could occur if high levels of the protein synthesis were needed to propagate successive cycles of somitogenesis.

In the second model, pleiotropic Bst phenotypes may reflect tissue-specific differences in Rpl24 expression. To achieve unimolar stoichiometry, dispersed riboprotein genes are coordinately expressed via 5′TOP elements, which define transcriptional start sites and control translation of riboprotein mRNAs(Hariharan et al., 1989; Jefferies et al., 1997). In spite of this convergently evolved feature, 5′TOP sequences may vary in potency across tissues. Differential regulation of riboprotein mRNAs has been described in pluripotent embryonal carcinoma cells, various cancer cell lines and in RNA profiling studies (Blackshaw et al., 2001; Panda et al.,2002). If the developing retina, somites and melanoblasts express relatively low levels of Rpl24 mRNA, they may be more sensitive to quantitative reduction caused by the Bst mutation.

In the third model, Bst phenotypes may occur because L24 has unique extraribosomal functions, apart from protein translation(Wool, 1996). For example, the small subunit riboprotein S3 functions as an endonuclease during DNA repair,while S0 acts on the cell surface, as the primary laminin receptor(Ardini et al., 1998). This explanation would be supported if mice with mutations in other riboprotein genes have phenotypes unlike Bst. The high incidence of RPS19 mutations among DBA patients(Draptchinskaia et al., 1999)suggests that S19 has extraribosomal functions in erythropoiesis. Recent studies show that L24 from a variety of species interacts directly with the transactivator (TAV) protein of cauliflower mosaic virus to promote translation re-initiation on polycistronic mRNAs(Park et al., 2001). In principle, the unique Bst phenotypes could reflect tissue specificity of cellular mRNAs that utilize internal ribosome entry(Hinnebusch, 1997; Johannes et al., 1999; Morris and Geballe, 2000).

Human RPS19 and mouse Rpl24 mutations are both associated with decreased growth and skeletal defects. Curiously, triphalangeal first digits are frequent in Bst/+ mice and are the single most common extramedullary findings in DBA patients. This suggests a shared mechanism in which limb malformations arise through impairment of ribosome biogenesis. However, unlike DBA patients, Bst/+ mice appear hematologically normal in the absence of erythroid stress (data not shown). Conversely, RPS19 mutations do not cause optic nerve disease or obvious hypopigmentation. These findings might reflect differences between humans and mice, or specific roles for S19 and L24 in hematopoiesis and retinogenesis,respectively. Accordingly, if the retinal phenotype is unique to Bst/+ mice, then RPL24 is a good candidate for dominant human optic nerve disorders associated with small stature, including colobomas of the posterior segment (Onwochei et al.,2000) and septo-optic dysplasia (De Morsier syndrome, MIM182230). Affected children have small optic nerves and growth retardation, similar to Bst/+ mice. A few patients have mutations in the HESX1homeobox gene (Dattani et al.,1998), but in most cases the cause is unknown. Growth hormone levels are often decreased, but pituitary anatomy and endocrine profiles are usually normal.

Riboprotein gene mutations may also contribute more broadly to common human malformations, such as the CHARGE (MIM214800) and VACTERL (MIM192350)syndromes, which involve a constellation of developmental defects but as yet have no clear genetic basis (Lalani et al., 2003). Eighty percent of CHARGE syndrome patients have retinal colobomas, anatomically equivalent to those in Bst/+ mice(Fig. 1B,C), and growth retardation as core diagnostic features(Russell-Eggitt et al., 1990; Tellier et al., 1998). Conversely, 11% of patients with ocular colobomas have diagnostic features of CHARGE syndrome (Chestler and France,1988).

Amsterdam et al. (Amsterdam et al.,2004) recently reported that zebrafish with heterozygous ribosomal protein gene mutations have a high incidence of malignant nerve sheath tumors(11 out of 16 riboprotein gene insertions identified in a recessive lethal screen). However, tumor formation was not elevated in the L24 mutant line.

We thank Margaret Van Keuren, Elizabeth Hughes, Donalyn Scheuner, Mark Adams, Deanna Kulpa, Stephanie Schmoll, Bob Lyons, Dave Siemieniak, Sean Morrison, Ömer Yilmaz, Phil Andrews, John Strahler, Sinisa Volarevic, Joe Brzezinski and Sara Schulz for technical advice; Elizabeth Robertson and Philippe Soriano for providing ES cells; and David Ginsburg, Sally Camper,Dave Turner, John Moran, Miriam Meisler, Randy Kaufman and Steve Weiss for helpful discussions and critically reading the manuscript. E.R.O. was supported by MSTP and UM Organogenesis NIH training grants (T32 GM07863 and HD07505). This work was supported by NIH grants to T.G. (EY14259, EY12994 and GM067840) and the UM Diabetes Research and Training Center (DK20572).