A mouse model of osteochondromagenesis from clonal inactivation of Ext1 in chondrocytes

Strategy to Model Loss of Heterozygosity.

Exon 2 of Ext1 codes for highly conserved amino acids of the catalytic center (100% amino acid homology between mouse and human). Missense mutations in exon 2, which lead to MO in humans, have been demonstrated to completely abrogate the enzymatic function of EXT1 in vitro (7, 19). To investigate if low-prevalence clonal inactivation of Ext1 would result in osteochondroma formation in mice, we generated mice carrying an inducible loss-of-function allele of Ext1 (Ext1^e2neofl), in which exon 2 was flanked by head-to-head loxP sites (Fig. 1A). Both heterozygous and homozygous carriers of the Ext1^e2neofl allele were healthy, displaying no obvious phenotype.

Although prolonged exposure of head-to-head loxP sites to Cre-recombinase has been demonstrated to cause chromosomal loss and apoptosis due to interchromatid translocations (20–23), limited exposure to Cre-recombinase yields a distribution of forward and inverted flanked fragments (24). Homozygosity for a head-to-head flanked allele was therefore hypothezised to generate chimeric tissues containing a low prevalence of inversion of both alleles.

To determine if the previously described chromosomal loss and apoptosis from inverted loxP sites was applicable to our allele, we crossed homozygous Ext1^{e2neofl/e2neofl} mutants to mice carrying Cre-recombinase under the control of the collagen IIa1 promoter (Col2-Cre) (25). No double-heterozygous Col2-cre;Ext1^e2neofl/+ offspring (0 of 113) were produced. Similarly, Ext1^{e2neofl/e2neofl} mutants crossed to mice expressing Cre-recombinase from the Hprt locus produced only very small litters. No double-heterozygous Hprt-Cre;Ext1^e2neofl/+ mice were retrieved even at E11.5 (zero of four matings) (Table 1).

To test if chromosomal loss and apoptosis could be avoided by limited exposures to Cre-recombinase, we exposed E14.5 mouse embryonic fibroblasts (MEFs) isolated from heterozygous Ext1^e2neofl/+ and homozygous Ext1^{e2neofl/e2neofl} mutants to TATCre (26). The cells remained healthy over at least five passages and had normal cytogenetic profiles on standard chromosomal spreads. Fluorescent in situ hybridization (FISH) with a chromosome 15 paint demonstrated the presence of the expected complement of two chromosomes 15 in homozygous Ext1^{e2neofl/e2neofl} cells after TATCre exposure (Fig. 1D). Genotyping of the Ext1 allele detected all possible Cre-mediated recombination products predictable from the design of the _targeting construct (Fig. 1 B and C).

Accordingly, introduction of a transgenic Cre-recombinase under the control of a doxycycline-inducible Col2 promoter (Col2-rtTA-Cre) (27) generated viable Col2-rtTA-Cre;Ext1^e2neofl/+ and Col2-rtTA-Cre;Ext1^{e2neofl/e2neofl} double mutants. Furthermore, induction of Cre activity during gestation after two matings between Col2-rtTA-Cre;Ext1^e2neofl/+ mice resulted in one Col2-rtTA-Cre;Ext1^e2neofl/+ and one Col2-rtTA-Cre;Ext1^{e2neofl/e2neofl} offspring. PCR-based genotyping of tail and other cartilaginous tissues revealed the presence of all predictable recombinations. Sequencing of the individual PCR products confirmed the expected identity of the rearranged fragments.

Exon 2 Inversion Disrupts Ext1 Function.

To simplify recombination products of the allele, we crossed Ext1^{e2neofl/e2neofl} males to Prx1-cre females (28), which express Cre-recombinase in the female germ line. By PCR we screened for mutants that had lost the neo cassette and retained exon 2 in either wild-type (Ext1^e2fl) or inverted (Ext1^e2flinv) orientation (Fig. 1B). Stable germ-line transmission of the generated alleles was confirmed by crossing founders (F₀) to C57B6 females. F₁ males that produced offspring (F₂) consistently carrying the allele in the expected orientation were used for further breeding. Both heterozygous Ext1^e2fl/+ and Ext1^e2flinv/+ mice were born at Mendelian ratios.

To verify that exon 2 inversion generates a functionally null allele, we isolated embryos of Ext1^e2flinv/+ matings at E7.0. We detected 4 of 10 embryos that were small and had failed to undergo gastrulation (Fig. 2B), similar to the phenotypes described for Ext1^−/− and Ext2^−/− mutants (17, 18). PCR analysis of microdissected embryonic tissue confirmed that these were homozygous Ext1^{e2flinv/e2flinv} mutants. To further support the inactivity of the inverted allele we investigated the production of HS by immunohistochemistry using the 10E4 (Seikagaku) antibody. In contrast to wild-type and heterozygous embryos, which showed solid staining in both the embryonic and extraembryonic tissues and basal membranes, we detected HS only in the placenta, but not in the basal membranes of the Ext1^{e2flinv/e2flinv} embryos proper (Fig. 2).

Postnatal Inactivation of Ext1 Generates Osteochondromas.

To investigate whether clonal inactivation of Ext1 function postnatally would lead to exostosis formation in mice, we induced expression of Cre-recombinase in chondrocytes from postnatal day 8 (P8) to P15 by administration of doxycycline in the drinking water.

Analysis of hematoxylin and eosin (H&E) and Safranin-O stained serial sections of knees and rib cages did not detect osteochondromas in 4-, 6-, or 10-week-old mutants, which lacked Cre-recombinase and/or homozygosity for the Ext1^e2fl or Ext1^e2neofl allele (Table 1). In contrast, Col2-rtTA-Cre;Ext1^{e2neofl/e2neofl} and Col2-rtTA-Cre;Ext1^e2fl/e2fl mice formed multiple osteochondromas with 100% penetrance after induction with doxycyline (Table 1 and Fig. 3). No differences could be detected in mutants with and without the neo cassette, excluding a nonspecific effect of the cassette itself on phenotype (29). To more precisely specify the induction time, we switched to administration of doxycycline by i.p. injection to the lactating dam. One i.p. administration of doxycycline to the mother at P8 was sufficient to induce osteochondromas with 100% frequency. No difference in osteochondroma morphology was observed between the two administration strategies.

As in human patients, the induced osteochondromas developed on most endochondral bones and were mainly located near the major growth centers. Both described osteochondroma morphologies, pedunculated and sessile, were observed (Fig. 3F). Planar reconstructions from microCT imaging revealed cortical and medullary continuity of each osteochondroma with the host bone, consistent with the radiographic diagnostic criteria in humans (Fig. 3 B and C).

To investigate the tissue composition of the osteochondromas, we analyzed sections through knees and rib cages at 4, 5, 6, and 8 weeks of age on a morphological and molecular level. In every mouse ≥6 weeks of age, safranin-O and H&E staining demonstrated cartilage-capped bone protuberances with cortical and medullary continuity and morphologic recapitulation of the physeal zones (Fig. 3). The morphological observations were supported by in situ hybridization analysis (Fig. S1), revealing a region of Col-II expression in cells at the peripheral cap of the osteochondroma and in columnar chondrocytes, which was followed by Ihh-expressing prehypertrophic chondrocytes and Col-X-expressing hypertrophic chondrocytes. As in osteochondromas arising in human MO patients, the osteochondromas induced in Col2-rtTA-Cre;Ext1^e2fl/e2fl mice mimic the organization of the growth plate on a molecular and morphological level.

Osteochondromas Originate in Proliferating Chondrocytes.

The cellular origin of the osteochondroma has been disputed to be either a peripheral physeal chondrocyte or a cell in the adjacent perichondrium/periosteum. As the Col2 promoter has previously been shown to drive expression in both tissues embryonically (30), we analyzed Col2-rtTA-Cre;Rosa26^LacZ/+ mutants that had received doxycycline by drinking water at P8–P12. β-Galactosidase (β-gal) staining at P16 demonstrated expression of the allele in ≈50% of proliferating growth plate chondrocytes and rarely in cells of the periosteum and primary spongiosa (n = 5). Although showing infrequent Cre-mediated recombination, these periosteal cells could not be thus excluded as the origin for osteochondromas (Fig. 4A).

We next introduced a tamoxifen-inducible Cre-recombinase under the control of the Osterix promoter (Osx-CreERT) (31) to Rosa26^LacZ/+ and Ext1^{e2neofl/e2neofl} mutants. β-Gal staining of P16 knees from Osx-CreERT;Rosa26^LacZ/+ mice that had received i.p. tamoxifen at P8 revealed expression in the osteoblasts of the perichondrium/periosteum and the primary spongiosa (n = 3). Expression was also detected in hypertrophic chondrocytes of the physis, but was excluded from the proliferative zone (Fig. 4B). MicroCT analysis of the lower extremities of Osx-CreERT;Ext1^{e2neofl/e2neofl} mice induced with tamoxifen at P8 did not detect any signs of osteochondroma development at 9 weeks of age (n = 4) (Fig. 4E). PCR on genomic DNA derived from bone tissues confirmed that recombination had taken place. Furthermore, safranin-O staining of serial sections of knees demonstrated mild aberrations in hypertrophic chondrocytes, but no osteochondromas (Fig. 4H), strongly indicating that the likely origin of the osteochondroma is a proliferating growth plate chondrocyte, the only localized cell type expressing Col2-rtTA-Cre, but not Osx-CreERT.

To confirm the proliferating chondrocyte as the cell of origin, we investigated sections through knees of doxycycline-induced Col2-rtTA-Cre;Ext1^e2fl/e2fl mutants at 4 and 5 weeks of age. Small clusters of cells were noted, disrupting the columnar organization at the periphery of the growth plate. At slightly later stages, expanding peripheral clones of chondrocytes were found at the level of the primary spongiosa, often still in continuity with the proliferating zone of the physis. In situ hybridization with chondrocyte-specific markers revealed expression of Col-II and Ucma, but not Ihh and Col-X, confirming an early differentiation state in the chondrocytes of these early osteochondromas (Fig. S2).

Homozygous Loss of Ext1 Is Required for Osteochondromagenesis.

To confirm that osteochondromas are generated from homozygous mutant cells, we isolated clones of chondrocytes from sections of Col2-rtTA-Cre;Ext1^e2fl/e2fl exostoses at different stages of development by laser capture microdissection (LCM) (Fig. S3). Genomic PCR using nested primer pairs to detect the Ext1^e2fl and the Ext1^e2flinv allele, respectively, revealed a mixture of all possible genotypes. About 70% of captured cell clusters produced neither the forward orientation nor the inverted PCR product, fitting the technical success rate from other studies using sections of paraformaldehyde-fixed, paraffin-embedded specimens for LCM (32, 33). Morphologically distinct, small clones of cells frequently displayed a homozygous Ext1^e2fl/e2fl or Ext1^{e2flinv/e2flinv} genotype, indicating that the clustered appearance reflects a genetically clonal origin and that recombination is rare after doxycycline clearance. Of 55 clones successfully analyzed, 38 were homozygous for the inverted Ext1^{e2flinv/e2flinv} allele, 14 homozygous for the Ext1^e2fl/e2fl allele, and 3 heterozygous Ext1^e2fl/e2flinv (Fig. 5).

To understand the spatial distribution of genotypes within the osteochondromas, we tested parallel sections with LCM genotyping and 10E4 immunohistochemistry, confirming that HS-negative areas are consistently Ext1^{e2flinv/e2flinv} and that osteochondromas containing Ext1^e2fl/e2fl chondrocytes displayed small, discrete regions of HS synthesis (Fig. 5). Screening of 10 osteochondromas with 10E4 immunohistochemistry demonstrated 6 with no detectable HS in the cap chondrocytes and 4 with a small minority of cells staining for HS, confirming that clonal dysfunction of Ext1 is present in the majority of chondrocytes in all osteochondromas, but that cells with functional Ext1 can be integrated into the mutant tissue during osteochondromagenesis.

Transgenic Mice.

A total of 3.6 kb of genomic DNA including exon 2 of Ext1 (42) [chromosome (chr) 15: 53,135,004–53,138,589] was amplified by PCR from the murine embryonic stem (ES) cell line 129 SvJ and cloned into the XhoI site of pMyr (Stratagene). An inverted loxP site was introduced upstream to exon 2 into a SacI site generated by site-directed mutagenesis using Quick Change (Stratagene). The fragment was then subcloned into the XhoI site of pOSDUPDEL upstream of the forward-facing loxP-flanked neomycin resistance (Neo) cassette. A total of 5.3 kb of DNA (chr 15: 53,138,682–53,144,025) beginning in intron 2 were amplified by PCR and subcloned into the EcoRI site downstream from the floxed Neo cassette in pOSDUPDEL (pOSDUPDEL-Ext1^e2neofl; Fig. 1A).

The linearized _targeting vector was electroporated into R1 ES cells (129 × 1/SvJ3 129S1/Sv). G418-resistant clones were screened by long-range PCR to identify Ext1-_targeted ES cell lines. Ends of PCR products were sequenced to confirm identity. One positive cell line was injected into blastocysts. Germ-line transmission and further genotyping were initially confirmed by long-range PCR from tail-tip-derived DNA. Routine genotyping was done with primers A and B of a panel of primers used to detect recombination (Fig. 1 B and C).

Rosa26^LacZ (43), Col2-rtTA-Cre (27), Hprt-Cre (44), Col2-Cre (25), and Osx-CreERT (31) mice were genotyped as described. In addition, presence or absence of Cre-recombinase was confirmed by PCR in each mouse.

Expression of Cre-recombinase in Col2-rtTA-Cre mice was induced either by administration of 4 mg/mL doxycycline in 5% sucrose through the drinking water of the lactating mother at P8 for 8 days or by i.p. injection of doxycycline (80 μg/g body mass) in PBS into lactating mothers at P8. Nuclear translocation of CreERT was induced in Osx-CreERT mice by i.p. administration of tamoxifen (0.2 mg/g body mass) in corn oil directly to the mutants.

Cell Culture.

E14.5 embryos from Ext1^e2neofl/+ matings were harvested, disemboweled, dissociated in trypsin, and plated in Dulbecco’s modified Eagle medium (DMEM) with 10% FBS. Nonadherent cells were discarded. Washed cells were treated with TATCre (4 mM in DMEM) for 3 h before adding serum-containing media. For cytogenetics, cultures were arrested with methotrexate, released with thymidine, treated with colcemid (Invitrogen), swelled in hypo-osmolar potassium chloride, fixed, and stained with Sytox Green (MBL International) or DAPI (Vectashield). Single whole mouse chromosome paint probe 15 (Applied Spectral Imaging) was applied following the manufacturer’s recommended protocol.

μCT.

Mice were killed at 9 or 10 weeks of age. No contrast was used. A General Electric Medical Systems EVS-RS9 μCT scanned the lower half of the body. Planar reconstructions and three-dimensional rendering were achieved with Microview software (General Electric).

Tissue Analysis.

Knees and rib cages were harvested, fixed in 10% formalin in PBS, and decalcified in 5% formic acid for histology or fixed in 4% paraformaldehyde in PBS and decalcified in 25% EDTA (pH 7.4) at 37°C for histology, in situ hybridization, immunohistochemistry, β-gal staining, and LCM. Five- to 10-μm sections from paraffin-embedded tissue were stained with H&E or safranin-O. Six-micrometer sections were processed for radioactive in situ hybridization using [P33]-UTP-labeled antisense riboprobes as described (45). Probes for in situ hybridization were as follows: Col2 (46), ColX (47), Ihh (48), Ptch1 (49), and Ucma (50). β-Gal staining was performed on frozen sections as previously described (51) and counterstained with nuclear fast red.

For immunohistochemistry, 6-μm paraffin sections were pretreated with 10,000 units/mL hyaluronidase I (Sigma) for 30 min at 37°C, quenched with 3% H₂O₂ in PBS for 30 min, blocked in 10% goat serum in PBS for 30 min, and incubated overnight at 4°C with anti-10E4 antibody (Seikagaku), diluted 1:100 in blocking solution. Goat biotinylated secondary antibody against mouse IgM (Vector Laboratories) at a dilution of 1:100 was applied for 30 min. Detection was performed with Streptavidin-HRP (Perkin-Elmer) diluted 1:100 in PBS for 30 min followed by incubation with DAB for 5 min and then counterstaining with Methyl Green (Sigma).

LCM.

Six-micrometer paraffin sections were transferred to PALM MembraneSlides, dried, deparaffinized, and stained with Mayer’s hematoxylin and eosin. Clones of 1–20 cells were then captured with the PALM Microbeam System (Carl Zeiss MicroImaging) and digested in Proteinase K. PCR was performed directly on the heat-inactivated cell lysate with outer primers. Outer primer PCR products were then used as templates for subsequent nested PCRs using inner primers.