Cathepsin K is a cysteine protease expressed predominantly in osteoclasts. Activated cathepsin K cleaves key bone matrix proteins and is believed to play an important role in degrading the organic phase of bone during bone resorption. Mutations in the human cathepsin K gene have been demonstrated to be associated with a rare skeletal dysplasia, pycnodysostosis. The degree of functional activity of the mutated forms of cathepsin K in these individuals has not been elucidated, but is predicted to be low or absent. To study the role of cathepsin K in bone resorption, we have generated mice deficient in the cathepsin K gene. Histologic and radiographic analysis of the mice revealed osteopetrosis of the long bones and vertebrae, and abnormal joint morphology. X-ray microcomputerized tomography images allowed quantitation of the increase in bone volume, trabecular thickness, and trabecular number in both the primary spongiosa and the metaphysis of the proximal tibiae. Not all bones were similarly affected. Chondrocyte differentiation was normal. The mice also had abnormalities in hematopoietic compartments, particularly decreased bone marrow cellularity and splenomegaly. The heterozygous animals appeared normal. Close histologic examination of bone histology revealed fully differentiated osteoclasts apposed to small regions of demineralized bone. This strongly suggests that cathepsin K–deficient osteoclasts are capable of demineralizing the extracellular matrix but are unable to adequately remove the demineralized bone. This is entirely consistent with the proposed function of cathepsin K as a matrix-degrading proteinase in bone resorption.
Bone consists of two major compartments, an inorganic mineral phase, composed primarily of calcium phosphate, and an organic protein matrix. To degrade bone efficiently, the osteoclast must be able to solubilize both fractions. While acid secretion into the resorption microenvironment can effectively eliminate the inorganic component of bone, the organic matrix, composed principally of type I collagen, requires proteases with acidic pH optima for degradation.(1,2)
Cathepsin K, a novel cysteine protease, was first cloned from rabbit(3) and human(4,5) osteoclasts, and osteoclastomas.(6) The subsequent identification of cathepsin K as the major cysteine protease expressed in osteoclasts,(7) suggests that it may have a critical role in osteoclastic bone resorption. In vitro studies have shown that cathepsin K can efficiently degrade both telopeptide and triple helical regions of type I collagen(8) as well as osteopontin (our unpublished observations) and osteonectin.(9) This enzymatic activity makes cathepsin K unique among vertebrate proteases and suggests that it has the potential to act alone in the degradation of bone matrix. However, it remains to be proven that cathepsin K activity is essential to osteoclast function in vivo. Currently there is still considerable controversy regarding the involvement of cathepsin K and other cysteine proteases in osteoclast function.(10,11)
Recently cathepsin K was identified as the gene mutated in the human skeletal disorder pycnodysostosis.(12,13) Pycnodysostosis is an autosomal recessive disorder, which results in a unique form of skeletal dysplasia. The clinical features of this disease include osteopetrosis and bone fragility, short stature, acro-osteolysis of the distal phalanges, delayed cranial suture closure, clavicular dysplasia, and a characteristic facial appearance resulting from calvarial bossing, loss of mandibular angle, and dental abnormalities (reviewed in Maroteaux and Lamy(14) and Elmore and Cantrell.(15) Three separate mutations in the cathepsin K gene have been identified in unrelated pedigrees in which pycnodysostosis occurs.(12) These mutations generate either nonsense, missense, or stop codon mutations in the cathepsin K gene and as such could lead to loss of function of cathepsin K. Although likely, it remains to be demonstrated that the mutations observed in pycnodysostosis result in a complete loss of cathepsin K activity in osteoclasts.
While our in situ hybridization studies have demonstrated cathepsin K expression only in osteoclasts(7) and ovary (cynomolgus macaque; unpublished observations), but not numerous other tissues,(7) other mRNA localization studies have shown its expression in human heart, lung, liver, pancreas, and skeletal muscle,(5) as well as mouse muscle, and articular and nasal cartilage.(16) Hence, it is unclear whether a total or partial inhibition of cathepsin K activity will affect only developing bone or whether other organ/tissue functions may also be compromised. This issue is important since cathepsin K antagonists are being advocated as potentially effective therapies for bone disorders such as osteoporosis.(17) To clarify these issues, we have used homologous recombination in embryonic stem (ES) cells to produce mice that either carry a heterozygous or homozygous null mutation of the cathepsin K gene.
MATERIALS AND METHODS
Generation of cathepsin K knockout mice
A 1.9 kb murine cathepsin K genomic clone that spanned exon 2 through exon 4 was isolated from a 129 SVJ (Stratagene, La Jolla, CA, U.S.A.) mouse genomic library by screening with a 1.8 kb fragment of the murine cathepsin K cDNA. The targeting vector was constructed by insertion of a pMc1-neo cassette (provided by Dr. Mario Capecchi) into a unique NdeI restriction site within exon 3. An HSV-TK cassette (provided by Dr. Capecchi) was inserted at the distal end of the 3′ homology arm. The targeting construct (Fig. 1A) was electroporated into JI ES cells and clones with homologous recombination events were selected in the presence of G418 and gancyclovir.(18) DNA probes external to the 5′ and 3′ arms of the targeting vector were used for confirmation of correctly targeted events using Southern blot analysis. Targeted clones were identified (Fig. 1B) and injected into CF1 blastocysts. Chimeras were mated with C57BL/6 and BALB/c mice, respectively, and heterozygous (+/–) offspring from each strain were mated to generate homozygous (–/–) mutant mice. Genotype analysis on all mice was performed by EcoRI digestion of genomic DNA using the 5′ external DNA probe (Figs. 1A and 1C). These studies were approved by the Institutional Animal Care and Oversight Committee.
Confirmation of null mutation
The null mutation was confirmed by reverse transcriptase-polymerase chain reaction (RT-PCR) amplification from bone RNA of cathepsin K (CK)+/+ and CK–/– mice. Oligonucleotide primers (forward 5′-CACTGCTTTCCAATACGTGC-3′ and reverse 5′-GCGTTGTTCTTATTCCGAGC-3′) were designed to span a 402-bp region of the 3′ end of the cathepsin K mouse cDNA. RT-PCR carried out on cDNA generated by reverse transcription of bone RNA and on DNAse-treated RNA as a negative control. A radioactively labeled oligonucleotide (5′-GGGCCAGGATGAAAGTTGTA-3′) internal to the PCR fragment was used to probe filters.
Histology of bone and internal organs
Long bones, lumbar vertebrae, and calvariae from 8- to 9-week-old mice (CK+/+, CK+/–, CK–/–) were dissected and fixed in 4% paraformaldehyde for up to 7 days. For bones and other organs, paraffin-embedded sections (8 μm; demineralized bone) were stained using hematoxylin and eosin. Plastic embedded sections (fully mineralized) were stained using von Kossa (to reveal mineralized bone and cartilage),(19) toluidine blue, or a modified Goldners Trichrome.(20) Longitudinal sections were collected throughout each bone.
X-ray micro-computed tomography analysis of proximal tibiae
X-ray micro-computed tomography (CT) images were obtained on a Scano Medical micro-CT scanner. The samples were stored in 70% ethanol for the duration of the scanning. A total of 223 images were obtained from each bone sample using a 512 × 512 matrix, resulting in an isotropic voxel resolution of 18 × 18 × 18 μm3. An integration time of 100 ms per projection was utilized.
Trabecular bone morphometric parameters were obtained from two sites (1.4 mm and 2.5 mm from the growth plate) of the proximal tibiae of 10-week-old mice, utilizing a spherical region of interest (volumes of 0.53–0.05 mm3) and built-in mean intercept length (MIL)-based analysis software. The raw images were gauss filtered (sigma = 1.2, base = 2) and binarized using a constant threshold before the morphometric analysis. The images were also rendered for three-dimensional display and visualization.
Immunophenotyping and clonal agar assay
Mice were sacrificed and peripheral blood collected by cardiac puncture. Blood was analyzed for clinical chemistry and hematology. Spleen, thymus, and bone marrow cell suspensions were prepared and analyzed by FACS (FACStar plus cell sorter; Becton Dickinson, Mountain View, CA, U.S.A.) and sysmex K1000 hemanalyzer (Toa Medical Electronics, Tokyo, Japan). B and T lymphocytes were assessed by lineage-specific markers, B220, immunoglobulin M (IgM), and immunoglobulin D (IgD). Macrophage, granulocyte, and neutrophil populations were analyzed by surface antigen expression of Mac-1 and Gr-1 markers.(21) Hematopoietic potential of bone marrow and spleen cell suspensions were determined by clonal agar assay. Macrophage lineage–restricted colony-forming cells (CFCs) responsive to colony-stimulating factor-1 (CSF-1) alone were assayed as an index of the status of the committed progenitor cell compartment. High proliferative potential colony-forming cells (HPP-CFCs) responsive to the combined stimulus of IL-1 plus IL-3 plus CSF-1 were assayed as a surrogate index of the hematopoietic stem cell compartment exactly as previously described.(21)
Generation of cathepsin K–/–mice
The strategy for generating a targeted deletion in the mouse cathepsin K gene is shown in Fig. 1A. The insertion of the neomycin resistance gene cassette into the NdeI site within exon three (the second coding exon of the cathepsin K gene) generates a mutation more 5′ than each of the three mutations occurring in Pycnodysostosis. The genomic structure of human cathepsin K is described elsewhere by Rood et al.(22) and that of mouse cathepsin K by Gelb et al.(23)
Correctly targeted ES cell clones were identified by restriction endonuclease digestion of DNA with at least two restriction enzymes and Southern blot mapping as shown in Fig. 1B. BamHI digestion yielded a 9 kb band for the wild-type allele and a 6.2 kb band for the mutant allele, whereas an EcoRI digest gave wild-type bands of 3.6 kb and 2.5 kb and the mutant allele bands of 3.6 kb and 1.7 kb, respectively (Fig. 1B). Targeted clones were injected and chimeras bred with two different strains of mice, BALB/c and C57Bl/6J. Mice were genotyped (Fig. 1C) and frequencies expected on the basis of Mendelian ratios were obtained (CK+/+, 27; CK+/–, 45; CK–/–, 28), indicating that no fetal or early embryonic lethality was observed in the CK–/– mice.
To evaluate whether the mutation introduced by homologous recombination resulted in an RNA transcript, RT-PCR was carried out on RNA extracted from bone of both wild-type and CK–/– mice. Figure 1D shows that no cathepsin K mRNA expression was detected in the bone RNA from CK–/–mice, indicating that these mice were homozygous for a null mutation, whereas the expected 402 bp fragment was amplified in reverse transcribed mRNA from CK+/+ mice.
Cathepsin K–/– mice have osteopetrosis
Histologic evaluation of the long bones (femur, tibia, and humerus) and lumbar vertebrae (data not shown) from 2-month-old CK–/– mice revealed pronounced osteopetrosis. Representative plastic and paraffin-embedded sections from the proximal tibia (Fig. 2a) demonstrated increased trabecular and cortical bone mass compared with the wild-type and CK+/– mice. The osteopetrosis in the bones of CK–/– mice was most striking at the medullary cavity of the mid-diaphysis. The bones of CK+/– and CK+/+ mice were essentially devoid of trabecular bone in this region. Further examples of histology of heterozygotes are not included because they were indistinguishable from wild-type animals.
Three-dimensional micro-CT images of the proximal tibiae from –/–, +/–, and +/+ mice (Figs. 2b–2e) confirmed the osteopetrotic phenotype of the CK–/– mice. Quantitative analysis of the trabecular bone was made at the primary spongiosa (see Figs. 2d and 2e) and the metaphysis, and the data are summarized in Table 1. At both sites there was significantly higher bone mass, trabecular number, and thickness, and reduced trabecular spacing in the CK–/– bones. There was no difference in the trabecular bone parameters between –/+ and +/+ mice. Similarly, cortical bone was significantly thicker in the CK–/–mice (quantitation not shown).
Table Table 1. Micro-CT Generated Trabecular Bone Morphometric Parameters Measured at the Primary Spongiosa (Site 1) and Metaphysis (Site 2) of the Proximal Tibia
Osteopetrosis was markedly pronounced in the distal femur of CK–/– mice, resulting in joint deformity (Figs. 3a and 3b). Anatomical abnormalities were also noted in the growth plates of the proximal humerus of CK–/– mice (Figs. 3c and 3d), with the contours of the cartilage growth plate extending into the epiphyseal bone plate and the primary spongiosa. Islands of cartilage and primary spongiosa persisted within the medullary cavity and metaphyseal and diaphyseal cortex (Figs. 3e and 3f). This represents remnants of neonatal endochondral bone formation and is a characteristic feature of osteopetrosis.(24)
Effects of cathepsin K knockout in the growth plate
The epiphyseal growth plates from the CK–/– mice revealed signs of diminished osteoclast activity at the zone of cartilage calcification and primary spongiosa (Fig. 4). There was significant accumulation of thickened calcified cartilage septa that resulted in an extended zone of dense and disorganized primary spongiosa (Figs. 4a–4d). This was a consequence of a retardation in CK–/– mice of rapid remodeling of the calcified cartilage septa into woven bone which clearly occurred in the CK+/– and CK+/+ bones (Figs. 4c, 4d, 5a, and 5b).
The zones of resting and proliferative chondrocytes in the growth plate were identical in all three groups (Figs. 5a and b). Similarly, chondrocyte apoptosis and cartilage calcification, together with the infiltration of capillaries and mesenchyme (preosteoblasts), were similar in all groups (Figs. 5a and 5b). The transverse septae were degraded similarly in all groups, but the longitudinal septae were resorbed more slowly in the –/– group. This is consistent with the known action of osteoclasts on longitudinal but not transverse septae.
Osteoclasts from cathepsin K–/– mice do not resorb demineralized bone efficiently
Sections stained with Goldners Trichrome were analyzed at high magnification to investigate the cellular events leading to osteopetrosis in the CK–/– mice (Figs. 5c–5f). Although there was no decrease in the number of osteoclasts in CK–/– bones, analysis of the resorption sites revealed distinct zones of demineralization at all the osteoclast–bone interface (Figs. 5e and 5f) that were never apparent in the wild-type or heterozygotes (Figs. 5c and 5d). This suggests that demineralization of the bone proceeds normally via a functional osteoclast vacuolar ATPase activity but that the osteoclasts fail to adequately resorb and endocytose the bone matrix, a histologic observation previously described in pycnodysostotic bone.(25) This is consistent with the expected role of cathepsin K in degrading the organic phase of the matrix during the resorptive process.
Cathepsin K–/– mice develop splenomegaly
Enlarged spleens that weighed more relative to CK+/+ (mean weight 80.6 ± 4.0 mg) or CK+/– (mean weight 92.3 ± 7.0 mg) mice were observed in 4-month-old CK–/– mice (mean weight 166.5 ± 24.0 mg). Spleen size and weight were unaltered in 2-month-old CK–/– mice. Similarly, the cellularity of spleens from CK–/– mice were found to be significantly elevated in 4-month-old mice, but not in 2-month-old mice (Figs. 6a and 6b). Histologically, spleens from 4-month-old mice revealed a marked disorganization of the red and white pulp (data not shown). Immunophenotype analysis of the spleens at 5 months of age revealed significant increases in some but not all populations of the lymphoid and myeloid lineages examined. Analysis of B lymphocyte markers B220, IgM, and IgD revealed a significant increase in B220−IgD−, B220+IgD−, B220+IgD+, B220−IgM−, and B220+IgM+ populations in CK–/– spleens compared with wild-type controls (Fig. 7a). Similar analysis of the T lymphocyte markers CD4, CD8, and CD3 revealed a number of significant increases in the total number of certain populations in the CK–/– spleen including CD4−CD8− and CD4+CD8− populations, whereas the CD3 expressing populations were unaffected (Fig. 7b). Assessment of myeloid lineage markers Gr-1 and Mac-1 showed a significant decrease in the percentage of both Gr-1 and Mac-1 expressing populations in the CK–/–spleens, while total numbers were unchanged (Fig. 7c). Bone marrow cellularity in CK–/–mice was found to be significantly lower than that of wild-type mice (Fig. 6c). These data are consistent with our histologic data which demonstrated reduced space in the marrow cavity of CK–/– mice. Taken together, these data suggest that extramedullary hematopoiesis occurs in the spleens of CK–/– mice to compensate for the reduced bone marrow cellularity. Further, immunophenotype analysis of the cell types in the bone marrow revealed a significant decrease in absolute cell number of all subtypes, even though the percentage of each subtype in the entire population was unchanged (data not shown). In addition, a significant reduction in the number of hematopoetic progenitors (CFCs) and hematopoietic stem cells (high proliferative potential CFCs) were detected by clonal agar assay in the bone marrow of CK–/– mice (Figs. 6d and 6e). This reduction in absolute cell numbers, but not percentage of all the different subtypes, reflect the reduction in bone marrow cavity space caused by osteopetrosis rather than a deficit in hematopoietic stem cell and progenitor differentiation. In the spleen at 4 months of age, the total number of hematopoietic stem cells (HPP) is increased in CK–/– mice compared with wild-type littermates (CK+/+) (Fig. 6f).
The thymus was examined both histologically for morphological changes and by FACS analysis for lymphocyte markers (CD4 and CD8, CD3, B220, IgM and IgD). No differences were detected between –/– and +/+ mice. Thymic cellularity was also examined and no differences were detected.
Peripheral blood parameters measured were red blood cells, white blood cells, reticulocytes, hematocrit, and platelet counts (all by FACS). B and T lymphocytes were examined as for the thymus. Populations of GR-1 and Mac-1 surface antigen expressing cells were examined. No differences were detected in peripheral blood for any of the parameters measured.
We have shown that a homozygous null mutation in the mouse cathepsin K gene results in osteopetrosis, reduced bone marrow cellularity, and splenomegaly after 2 months of age. These data primarily establish a nonredundant role for cathepsin K in osteoclast function. Our findings also show that osteoclast differentiation and bone demineralization are unaffected by the absence of cathepsin K. Thus, these findings place the necessity for cathepsin K function in osteoclast development/function much later than that of genes such as CSF-1,(26) c-src,(27) PU.1,(28,29) and c-fos(30) which affect osteoclast differentiation rather than activity. Such mice display a much more severe osteopetrotic phenotype.
Intriguingly, however, this study suggests that a loss of function of cathepsin K does not universally affect the ability of osteoclasts to resorb bone, neither does it affect all bones equally. Osteoclasts from CK–/– mice were able to resorb the inorganic bone matrix adequately, but the primary defect in these mice was seen in the resorption and endocytosis of the organic phase of the matrix. Similarly, parts of the skeleton that involve rapid remodeling of bone (for example, the long bones and vertebrae) displayed the osteopetrotic phenotype while other bones such as calvariae (that are formed by intramembranous ossification) appear relatively unaffected. Similarly, the epiphysis also appeared normal and interestingly this bone is also resistant to resorption following ovariectomy.(31) The calvaria and epiphysis have in common a low bone turnover rate. Thus, it appears that rapidly remodeling bone shows the osteopetrotic phenotype, whereas bone with a low turnover rate is unaffected. This is consistent with a substantially decreased rate of resorption but not a complete absence of the process. Alternatively, it is possible that differences in cathepsin K expression at diverse skeletal sites may be contributing to the phenotypic differences observed in mice that have the cathepsin K gene null mutation. However, these results should be interpreted with caution because it is possible that these apparently normal skeletal components display some anomalies at different stages of development. This is currently under investigation.
These findings provide an excellent demonstration of the dual processes required for osteoclast activity, namely mineral dissolution and matrix degradation. In normal bone, these processes are too tightly coupled temporally to be distinguishable. In the CK–/–mice, where matrix proteolysis is dramatically slowed, it is possible to visualize recently demineralized but intact bone matrix directly apposed to the osteoclasts. In keeping with our data, Everts et al.(32,33) demonstrated in an in vitro mouse bone explant system (resorption activated by PTH) that culturing bone in the presence of nonselective cysteine protease inhibitors caused an increase in the number of osteoclasts that were in close contact with demineralized bone (12-fold over control). In the protease-inhibited osteoclasts, these authors also demonstrated the presence of cytoplasmic vacuoles containing collagen fibrils, similar to those described in osteoclasts from patients with pycnodysostosis.(25) They also described(33) a significant reduction in the relative surface density of electron-translucent vacuoles, in keeping with reduced endocytic activity. At the light microscope level, we were unable to discern collagen fibrils in the CK–/– mouse osteoclasts. Furthermore, in contrast to Everts et al.,(32,34) we were unable to observe any CK+/+ osteoclasts apposed to demineralized matrix. Similarly, in regions not lined by osteoclasts in CK–/– bones, we were unable to detect zones of demineralization. These conflicting results may reflect the smaller number of osteoclasts observed in sections of 10-week-old tibiae compared with the large number seen in the highly activated explant system.
The CK–/– mice also displayed a reduction in bone marrow cellularity (at 2 months of age) and splenomegaly (at 4 months of age). We have been unable to detect cathepsin K expression in spleen from fetal or adult normal mice.(35) The data in this study suggest that the reduced bone marrow cellularity is a consequence of the reduced bone marrow space in these osteopetrotic mice rather than a direct role of cathepsin K in hematopoiesis. The splenomegaly in turn may arise as an attempt to compensate for the reduced bone cellularity. It is possible that the decrease in the proportion of myeloid cells reflects a level of competition for myeloid precursors between osteoclasts and other myeloid lineages.
An important finding of this study is that the phenotype of the CK–/– mice resembles the human genetic disorder pycnodysostosis (due to a mutation in cathepsin K) in several respects. Osteopetrosis is a characteristic feature of pycnodysostosis(15) and this is primarily responsible for the enhanced bone fragility and/or predisposition to bone fractures observed in these individuals.(36) Individuals with pycnodysostosis also characteristically have short stature due to the reduced size of the long bones.(37) Both bone marrow hypoplasia and splenomegaly have been noted, albeit infrequently, in pycnodysostosis.(38,39) It appears that the mutations reported in the cathepsin K gene in patients with pycnodysostosis are inactivating mutations that lead to the observed phenotype.
It is surprising that the CK–/– mice did not display striking craniofacial defects since these are characteristically seen in humans with pycnodysostosis.(15) We did, however, detect some minor craniofacial anomalies—evidence of increased density of the maxilla and paranasal sinus bones in X-rays of CK–/– mice and alterations in mandibular shape were detected (data not shown).
Finally, the data from this study have relevance to the proposed use of antagonists of cathepsin K as therapeutic agents for osteoporosis. First, the CK–/– mice provide evidence that cathepsin K plays a key role in osteoclast function and bone resorption. Second, and more importantly, these data provide strong evidence for cathepsin K function predominantly (if not solely) in osteoclasts. This has been proposed previously on the basis of studies that found osteoclast-specific localization(3,5,7,35,40) or activity of cathepsin K(9,41) and studies which have shown that potent inhibitors of cathepsin K inhibit bone resorption in vitro and in vivo.(42) Since we have been unable to find anomalies in other organs/tissues in CK–/– mice, these data provide strong evidence that specific cathepsin K antagonists may preferentially (if not specifically) affect osteoclast function.
NOTE ADDED IN PROOF
A description of the phenotype of a separate cathepsin K–deficient mouse was published recently, identifying similar skeletal characteristics to those described in this paper.(43)
The authors gratefully acknowledge the editorial input of Drs. Sanjay Kumar, Michael Lark, and Larry Suva. This work was funded by SmithKline Beecham Pharmaceuticals.