We report the first case of lethal autosomal recessive osteopetrosis in mice caused by a spontaneous 8-bp deletion in exon 2 of the Rank gene. The phenotype, including a block in RANKL-dependent osteoclast differentiation and lymph node agenesis, copies that of Rank−/− mice, which have been produced by targeted recombination.
Introduction: Commitment of osteoclast progenitors to the osteoclast lineage requires RANKL/RANK-mediated intercellular signals. Gene-targeted defects in this signaling pathway resulted in osteoclast deficiency and severe osteopetrosis in mice, but to date, there have been no reports of spontaneous mutations in Rankl or Rank resulting in osteopetrosis.
Materials and Methods: Mice with malignant osteopetrosis and absent lymph nodes appeared spontaneously in a highly inbred colony. Appropriate crosses were analyzed to establish the pattern of inheritance. Tissues from affected pups and littermates were evaluated grossly, histopathologically, and radiographically. Osteoclast development from splenocytes was tested in vitro under a variety of conditions, including after infection with RANK-encoding retrovirus. Rank mutational analysis was performed by direct sequencing of RT-PCR products and genomic DNA.
Results: The inheritance pattern was consistent with autosomal recessive inheritance, and the phenotype resembled that of either Rankl or Rank knockout mice with the exception of as yet unexplained death of most mice 2–3 weeks after weaning. Osteoclast precursors from the spleens of affected pups failed to form osteoclasts in vitro when stimulated with macrophage-colony stimulating factor (M-CSF) and RANKL, unless they were forced to express wildtype Rank cDNA. Molecular genetic studies identified an 8-bp deletion in exon 2 of the Rank gene. The resulting allele, termed Rankdel8, encodes only a small portion of the RANK extracellular domain, which is probably nonfunctional.
Conclusions: The phenotypic similarities between Rankdel8 and mice previously described with a combined insertion and deletion in Rank confirm the role of this receptor in osteoclastogenesis and lymph node development and suggest that some forms of malignant osteopetrosis in humans could result from a similar defect.
Malignant infantile osteopetrosis (arOP) is an autosomal recessive disorder caused by deficient osteoclast-mediated bone resorption.(1) Affected individuals often seem normal at birth, but develop a generalized progressive increase in BMD that becomes evident clinically in the first year of life. Improper remodeling of the craniofacial skeleton is associated with macrocephaly, nasopharyngeal obstruction, visual and acoustic deficits, and cranial nerve compression. Nonresorbed bone displaces hematopoietic marrow in long bones, resulting in osteopetrosis and leading to pancytopenia with marked extramedullary hematopoiesis, hepatosplenomegaly, and frequent infections. Despite increased BMD, the improperly configured bones are vulnerable to pathological fractures. Without treatment, most patients die within a year from sepsis, pneumonia, hemorrhage, or severe anemia.(2)
Human arOP is genetically heterogeneous.(3) Until recently, most of the specific genes mutated in arOP were unknown. However, analysis of murine models revealed that homozygous mutations in any one of at least 11 genes cause an arOP-like phenotype.(4) Osteopetrotic mice, like humans with arOP, can broadly be subdivided into two groups based on whether osteoclasts are absent versus present, but dysfunctional.(5), (6)
Mutations that prevent formation of osteoclasts include those that disrupt components of the RANK (also referred to as TNFRSF11A)-mediated intercellular signaling pathway.(7) RANK is a transmembrane receptor expressed by osteoclast precursors in the monocyte lineage. Activation of this receptor is necessary for differentiation of macrophage-colony stimulating factor (M-CSF)-induced, CD11b+ committed osteoclast precursors into TRACP+ osteoclasts. RANK is activated by a member of the TNF family, RANKL (also referred to as osteoprotegerin-ligand, TNF-related activation-induced cytokine, osteoclast differentiation factor, and TNFSF11). Mice with targeted mutations of either Rank(8), (9) or Rankl(10) have been produced by homologous recombination (gene targeting) in embryonic stem cells. In each of the models, homozygotes showed similar phenotypic features including severe osteopetrosis, absent lymph nodes, and abnormal pregnancy-induced mammary gland differentiation.(11)Rank−/− mice have increased numbers of CD11b+ osteoclast precursors in their spleens and peripheral blood, indicating that the absence of RANK expression does not impair the precursor survival.(12) As with other congenital models of murine osteopetrosis, teeth fail to erupt in mice that lack RANK or RANKL, and the animals die shortly after weaning unless their food is ground up to permit them to swallow it.
In this paper, we characterize a new spontaneous Rank mutation that produces arOP in mice. Spontaneous mutations in RANK or RANKL resulting in osteopetrosis have not been described previously in mice or humans. The Rank alleles that were independently engineered by targeted recombination replaced portions of exons 2 and 3(9) or exons 4–7(8) (referred to here as RankneoA and RankneoB, respectively) with a PGK-neo cassette. In each case, homozygous mutants showed neither detectable RANK mRNA nor RANK protein. Spontaneous mutations that phenocopy existing multigenic disorders, like arOP, warrant molecular genetic characterization because they may reveal novel pathogenic loci or provide more subtle information about existing gene products, such as functionally important protein domains. Furthermore, description of novel mutations in a gene, the function of which has been tested by gene targeting, may validate the pre-existing knockout model.
MATERIALS AND METHODS
The Rankdel8 mutation arose spontaneously in a small colony of transgenic mice that were being bred to establish congenics with a 129 SvJ genetic background, as described elsewhere.(13) The founder animal for the congenic breeding program carried both a genetically engineered mutation in the Hox11L1 gene on chromosome 6 and the Hox11L1-nlacZ reporter gene, which is randomly integrated at an unknown locus. An osteopetrotic pup was first observed in an intercross between transgenic N5 littermates. The affected mouse carried one modified Hox11L1 allele but not the Hox11L1-lacZ transgene. The same intercross was repeated twice and gave rise to three other affected animals, including a mouse that carried neither the Hox11L1 mutation nor the Hox11L1-nlacZ transgene. Based on these findings, siblings of the affected mice were backcrossed with their parents, and several nontransgenic carriers were identified. The latter founded the colony used to collect phenotypic and genotypic data. All animal procedures were approved by the Animal Care and Use Committees of both institutions.
Bone histology and TRACP histochemistry
Long bones and skulls were collected at various ages between birth and postnatal day 35 from affected mice and their littermates. Bones were stripped of most muscle, fixed by immersion in 10% neutral buffered formalin for 4–16 h, and decalcified for 7 days in 10% EDTA in Tris buffer, pH 7.2 at 4°C. The decalcified specimens were dehydrated and embedded in paraffin. Routine histological studies were performed with 5-μm-thick, hematoxylin-and-eosin-stained sections. Osteoclasts were identified in paraffin sections that were stained histochemically for TRACP activity using a leukocyte acid phosphatase kit (Sigma) according the manufacturer's instructions.
Osteoclast differentiation in vitro
Osteoclastogenesis assays were carried out by co-culturing spleen cells with calvarial osteoblasts in the presence of 1,25-dihydroxyvitamin D3 [1,25(OH)2D3; a gift from Dr Milan Uskokovic, Hoffmann-LaRoche, Nutley, NJ, USA] or by culturing splenocytes with M-CSF and RANKL. For co-culture experiments, primary osteoblasts (0.25 × 104 cells/well), which were isolated from calvariae of 9- to 10-day-old affected mice or their unaffected littermates using a sequential collagenase/protease digestion, were co-cultured with spleen cells (0.5 × 105 cells/well in 96-well plates) from affected mice or unaffected littermates in α-MEM (GIBCO BRL) supplemented with 10% FCS (Hyclone Laboratories) in the presence of 10−8 M 1,25(OH)2D3, as described previously.(14) Cultures were maintained for 7 days at 37°C in an atmosphere of 5% CO2/air and were fed every 2 days by replacing one-half of the spent medium with fresh medium and 1,25(OH)2D3. Cells were fixed and stained for TRACP activity to identify osteoclasts, as described previously.(15) For spleen cultures, splenocytes from affected animals and their unaffected littermates were used to generate osteoclasts in the absence of osteoblast/stromal cells. These cells were cultured with 5 ng/ml of RANKL and 10 ng/ml of M-CSF (R&D Systems) under the same conditions for co-cultures, and the number of TRACP+ osteoclasts was counted at the end of the experiments. In rescue experiments, splenocytes from affected animals were first cultured with M-CSF (25 ng/ml) for 3 days to enrich the osteoclast precursors and then infected with RANK retroviral supernatant (a gift from Dr William Dougall; Amgen) in the presence of M-CSF (10 ng/ml) and polybrene (2 μg/ml; Sigma). The medium was changed every other day for 5 days, and the number of TRACP+ osteoclasts was assessed at the end of the experiments. To assess bone resorption, cells were cultured on bone slices for 9 days under the same conditions as described above. Cells were removed, and the number and area of resorption pits were counted microscopically after staining with a 0.1% toluidine blue solution, as described previously.(16)
Rank mutational analysis
The spleens from two affected and two unaffected postnatal day 10 mice were homogenized to isolate RNA (RNeasy Mini Kit; Qiagen). After DNase I treatment (DNA-free; Ambion), the mRNA was reverse transcribed to cDNA using random hexamers (SuperScript First Strand Synthesis for RT-PCR; Invitrogen). A 1.8-kb cDNA fragment was PCR amplified from each sample using primers complementary to the 5′ (forward, 5′-GTGACTCTCCAGGTCACTCCT) and 3′ (reverse, 5′-TCTGCACATTGTCCGGACCCC) ends of the Rank coding sequence. The fragments were gel purified and sequenced (BigDye, version 1.0, Terminator Cycle Sequencing Ready Reaction; Applied Biosystems) with 3 pmol of the above primers on an ABI 310 Genetic Analyzer. Sequences were compared with each other and to public databases using BLASTN software (http://www.ncbi.nlm.nih.gov/blast). To confirm the cDNA sequencing results, genomic DNA was isolated from liver samples of the above animals (DNeasy Tissue Kit; Qiagen) and PCR amplified with Rank-specific primers that correspond to intron sequences, which flank exon 2 (forward, 5′-TGATTGGTATGTTGCCTTCTCAA-3′; reverse, 5′-TCTTCACTCACTTCAGTGTGAC-3′). The 20-μl reaction volumes contained 1× buffer A, 0.5 μM of each primer, and 0.625U of FailSafe Taq Polymerase (Epicenter Technologies, Madison, WI, USA) and were run through 38 cycles for 30 s at 94°C, 59°C, and 72°C. The ∼160-bp products were purified (QIAquick PCR Purification Kit; Qiagen) and subjected to sequencing in both directions with the exon 2-specific primers, as described above.
Mutant mouse phenotype
A single affected animal was recognized among the progeny of an N5 intercross between transgenic animals that were part of a breeding program to establish 129 SvJ congenics. The underlying mutation arose spontaneously, and neither parent was dysmorphic. Although the original affected mouse carried a transgenic alteration, subsequent breeding excluded the latter from the genetic background without affecting the phenotype associated with the novel mutation.
The affected animals seemed normal at birth but exhibited typical features of severe osteopetrosis, including failed tooth eruption, which could be confidently identified by postnatal day 7 (P7; Fig. 1). In the second and third weeks after birth, the growth of the mice lagged behind that of their littermates, and other craniofacial defects became obvious, including short “compressed” snouts and occluded lachrymal ducts with consequent accumulation of crusty secretions between the eyelids. Radiographic studies showed a diffuse increase in BMD and progressive kyphosis (Figs. 1C-1F). Death occurred in most mice between 3 and 5 weeks of age, soon after weaning. The diets of three animals were changed from standard solid chow to a liquid form at weaning. These mice consumed the soft diet and survived for 2–4 weeks longer than their littermates, but eventually were killed when they seemed critically ill.
Necropsy showed gross and histopathological findings identical to those described for Rank−/− and Rankl−/− mice. In addition to craniofacial and other skeletal deformities, their spleens and livers were slightly enlarged, with extramedullary hematopoiesis. Abnormal bone microanatomy was evident in sites of membranous and endochondral bone formation (Fig. 2). The epiphyseal growth plates of long bones were enlarged due primarily to lengthening of the columns of hypertrophic chondrocytes (Fig. 2), and the transition from cartilaginous matrix to bone was impaired, such that cartilage-containing trabeculae were present far from the growth plates in the diaphysis. The density and thickness of bone trabeculae were dramatically increased, with little intervening space for hematopoietic marrow tissue, typical of severe osteopetrosis. In H&E-stained paraffin sections, multinucleate osteoclasts could not be identified, in contrast with osteoblasts and other bone marrow cells, which were easily identified (Figs. 3A and 3B). Sections of bone that were stained histochemically for TRACP activity, a marker of mature osteoclasts, were completely negative, consistent with complete absence of these cells (Figs. 3C and 3D). Other monoblast derivatives, including monocytes, macrophages, and central nervous system microglia were present and seemed cytologically normal. Analysis of the peripheral blood samples indicated that the osteopetrotic mice were anemic (mean hematocrit = 37.6 ± 5.3; n = 5) in comparison with their non-osteopetrotic littermates (mean hematocrit = 44.2 ± 3.9; n = 9; Student t-test, p = 0.03). The only other significant pathological finding was a complete absence of lymph nodes (Fig. 4), despite an intact thymus and normal distribution of small intestinal Peyer's patches. No acute infection was observed grossly or microscopically in mice that died spontaneously, and the cause of death 2–3 weeks after weaning, even on soft diet, remains to be identified.
Autosomal recessive pattern of inheritance
Analysis of 111 pups from litters that contained affected offspring showed that 28 (25.2%) had the mutant phenotype. This frequency suggested autosomal recessive inheritance of a trait with no recognizable phenotype in heterozygous carriers. As expected, a subset of phenotypically normal offspring of affected animals could be inter- or backcrossed to produce more osteopetrotic pups and asymptomatic, presumed heterozygous carriers of the mutation.
In vitro osteoclast development
Failure of osteoclast formation can occur as a result of a defect in cells in the osteoclast or osteoblast lineage. To examine these possibilities, we conducted co-culture experiments of calvarial osteoblasts from mutant mice with spleen cells from wildtype mice or vice versa in the presence of 1,25(OH)2D3. We found that mutant osteoblasts supported osteoclast formation from wildtype splenocytes, but wildtype osteoblasts could not rescue osteoclastogenesis by mutant splenocytes, suggesting that the defect in the mutant mice resides in the osteoclast lineage (Fig. 5A). Next, we cultured splenocytes, which contain undifferentiated hematopoietic precursors, from osteopetrotic pups and their phenotypically normal siblings in the presence of M-CSF and RANKL to promote osteoclast differentiation. Numerous TRACP+ osteoclasts arose in the cultures from phenotypically normal mice. However, no TRACP+ cells formed in parallel cultures from osteopetrotic animals (Fig. 5B). Because exogenous RANKL was present in the medium, these results suggested a defect in the RANK receptor or its downstream signal transduction pathway.
We tested this hypothesis by infecting dissociated splenocytes with a replication-defective retrovirus that was engineered to force expression of a wildtype RANK cDNA. Infection with the RANK retroviral construct rescued osteoclastogenesis in vitro, as evidenced by formation of numerous TRACP+ cells in splenocyte cultures from osteopetrotic pups (Figs. 5C and 5D). In this assay, some osteoclasts differentiated even in the absence of exogenous RANKL, as reported previously with this construct,(17) but the relative number of TRACP+ cells was significantly increased in the presence of exogenous ligand. When RANK virus-infected splenocytes from the mutant mice were cultured on bone slices with M-CSF, they formed numerous resorption pits, indicating that these cells were functional osteoclasts (Fig. 5D).
Rank mutational analysis
The cumulative phenotypic data, including rescue of the defect in osteoclast differentiation in vitro with forced expression of retroviral Rank, suggested that our osteopetrotic animals carried a mutation of the Rank gene. Mutational analysis was initiated by a RT-PCR to amplify Rank cDNA from splenic mRNA, while wary that some types of Rank mutation might not yield a cDNA product. Primers to the most 3′ and 5′ coding sequences produced a product of the approximate expected length for intact Rank mRNA. Sequencing of the product revealed an 8-bp deletion, located 234 bp 3′ to the translational start site (Fig. 6). Subsequent analysis of genomic DNA confirmed the 8-bp deletion in the corresponding site in the middle of exon 2. Osteopetrotic pups were homozygous for this allele, in contrast to littermates, which were either heterozygous or wildtype.
We designated this novel allele “Rankdel8.” The deletion results in a frameshift at codon 79 that changes the identity of this codon from aspartate to glutamate followed by a translational stop codon (Fig. 6). As such, the predicted protein product encoded by the Rankdel8 allele is a severely truncated version of the receptor that includes only the first and a small part of the second extracellular cysteine-rich domain, plus an appended glutamate residue at the C terminus.
Osteopetrosis is a genetically heterogeneous disorder. It occurs spontaneously in humans and other mammals and has been induced in mice often unexpectedly after deletion of ubiquitously expressed genes, such as c-src,(18)c-Fos,(19) and NF-κB,(20) and overexpression of OPG.(21) Although naturally occurring mutations in some of the genes that regulate osteoclast function, such as carbonic anhydrase,(22)Mitf,(23)CLCN7 (chloride channel 7),(24), (25)ATP6I (α3 subunit of the proton pump),(26–28) and Gl,(29) have been identified as the cause of osteopetrosis in some humans and mice, the etiology in most human cases remains unknown. Furthermore, to date, no cases of osteopetrosis have been reported to have occurred spontaneously in humans or other mammals as a consequence of mutation of c-src, c-Fos, NF-κB, OPG, RANKL, or RANK, or any of the RANK-activated downstream signaling molecules that have been identified by gene knockout studies. In contrast, activating mutations in RANK have been reported recently in patients with familial expansile osteolysis(30) and so-called early-onset Paget's disease.(31) The gross, radiographic, microscopic, and tissue culture findings from our spontaneous new model of osteopetrosis excluded many gene candidates and led directly to identification of a mutation of the Rank gene, thus validating the Rank knockout mouse models and the requirement of RANK signaling in osteoclastogenesis. These included the severe osteopetrosis, lack of lymph nodes, a thickened hypertrophic zone in the growth plate,(8) and rescue of the defect in osteoclast formation by forced expression of a Rank retroviral vector. The cause of the thickening of the growth plate in the Rank knockout and in our mice remains to be determined, and the expression of RANKL and RANK in the growth plate(32), (33) suggests that they play a regulatory role in chondrocytes as well as osteoclasts.
Numerous attempts to generate “null” mutations by gene targeting have been compromised by exon skipping, alternative transcriptional start sites, or unintended effects on expression of neighboring genes.(34–38) In such instances, a putative “knockout” animal may exhibit phenotypic features that do not reflect complete loss of gene function and/or are related to altered expression of contiguous genes. The latter may result from gene silencing or cryptic promoter activity that is conferred by PGK-neo or other selectable markers used for gene targeting. In this context, the Rankdel8 allele provides important confirmatory information. The osteopetrotic phenotype of homozygous Rankdel8 mice seem identical to mice with putative null mutations (designated herein as Rankneo) that were introduced in the Rank receptor by gene targeting.(8), (9) In each case, the affected pups develop severe osteopetrosis and lack lymph nodes. Reduced numbers of splenic B-lymphocytes, which have been reported in Rankneo/Rankneo mice,(8) are also observed in Rankdel8/Rankdel8 homozygotes (data not shown). The similarities are not surprising, given that no functional gene product is predicted from either of the targeted alleles or Rankdel8 (Fig. 6). In contrast with Rankdel8, which harbors a small deletion in exon 2, exons 2 and 3 or exons 4–7 of the Rankneo alleles were replaced by a PGK-neo cassette.(8), (9) The first or first three exons of Rankneo are intact in these mice such that an aberrant, and possibly partially functional, mRNA might be produced by alternative splicing between exon 1 or 3 and any of the exons located 3′ to the neomycin-resistance cassette. Furthermore, the neomycin-resistance cassette may elicit epigenetic modifications that suppress expression of a neighboring gene, which could contribute to one or more aspects of the Rankneo/Rankneo phenotype. In contrast, the overall gene structure of the Rankdel8 allele is preserved, and no foreign DNA is inserted, such that all of the phenotypic findings in Rankdel8/Rankdel8 mice are likely to result from severe truncation of the protein. In combination, the Rankneo and Rankdel8 models provide confidence that their identical phenotypes are caused by loss of RANK alone.
The sequence of Rankdel8 mRNA predicts production of a severely truncated receptor that consists of little more than a single cysteine-rich domain from the extracellular domain. It is conceivable that the truncated soluble peptide might bind to RANKL and function in a manner analogous to the decoy receptor osteoprotegerin (OPG). OPG is a protein with significant structural similarity to the extracellular domain of RANK.(7) OPG seems to modulate RANK activity by competing with the receptor for RANKL, and transgenic mice that overexpress OPG partially phenocopy Rank-null mice,(21) although they do form lymph nodes. We think it unlikely that a Rankdel8 gene product has similar properties to OPG for two reasons. First, no phenotypic features of RANK inhibition are observed in Rankdel8 heterozygotes. Furthermore, a critical aspartate residue, required for OPG binding,(39) corresponds to a codon that is located far downstream from the stop codon in Rankdel8 transcripts. Because this amino acid would be missing from the truncated peptide, binding to RANKL seems improbable.
When a genetic disorder arises spontaneously in a mouse colony, it is important to investigate the phenotype and etiology thoroughly, including the underlying molecular genetics. Even when well-characterized genetic alterations are known to produce an identical phenotype, the results of comprehensive studies may reveal a novel gene involved in the pathogenesis of a multigenic disorder and/or insight into regions of a specific gene product that are vital for its function. It will be important to determine if any of the forms of human malignant osteopetrosis result from mutations in RANK, such as the present Rankdel8. This mutation occurred in a highly inbred mouse colony and thus consanguinity may be an important factor in its etiology. Increased parental consanguinity and recurrence within sibships are recognized features of infantile osteopetrosis and implicate transmission as an autosomal recessive trait, as in our mice.
This study was supported by grants from the National Institutes of Health (RO1 DK56987 to RPK and RO1 AR43510 and R21 AR49305 to BFB) and the University of Washington Center for Ecogenetics and Environmental Health (NIEHS P30 ES07033).