Absence of the alpha1(IX) chain leads to a functional knock-out of the entire collagen IX protein in mice.

Cartilage fibrils contain collagen II as well as smaller amounts of collagens IX and XI. The three collagens are thought to co-assemble into cartilage-specific arrays. The precise role of collagen IX in cartilage has been addressed previously by generating mice harboring an inactivated Col9a1 gene encoding the alpha1(IX) chain, i.e. one of the three constituent chains of collagen IX (Fässler, R., Schnegelsberg, P. N. J., Dausman, J., Shinya, T., Muragaki, Y., McCarthy, M. T., Olsen, B. R., and Jaenisch, R. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 5070-5074). The animals did not produce alpha1(IX) mRNA or polypeptides and were born with no conspicuous skeletal abnormality but post-natally developed early onset osteoarthritis. Here we show that the deficiency in alpha1(IX) chains leads to a functional knock-out of all polypeptides of collagen IX, whereas the Col9a2 and Col9a3 genes were normally transcribed. Therefore, synthesis of alpha1(IX) polypeptides is essential for the assembly of heterotrimeric collagen IX molecules. Surprisingly, cartilage fibrils of all shapes and banding patterns found in normal newborn, adolescent, or adult mice were formed in transgenic animals, although they lacked collagen IX. Therefore, collagen IX is not essential, and may be functionally redundant, in fibrillogenesis in cartilage in vivo. The protein is required, however, for long term tissue stability, presumably by mediating interactions between fibrillar and extrafibrillar macromolecules.

The biomechanical properties of cartilage are intimately linked to the structure of its extracellular matrix that in essence consists of two suprastructural compartments. An extended three-dimensional fibril network penetrates an extrafibrillar matrix that appears as an amorphous mass in the electron microscope, although there is a high degree of molecular organization. Aggrecan, the major cartilage proteoglycan, is immobilized by specific interactions with high molecular mass hyaluronan. Each aggrecan molecule comprises a large core protein highly substituted by polyanionic glycosaminoglycans. This confers to the extrafibrillar matrix a very high negative charge density and, through osmotic binding of large amounts of water, generates a swelling pressure that is contained by the fibril network. Unlike the extrafibrillar matrix, cartilage fibrils exhibit a characteristic cross-striation in the electron microscope. The fibrils are macromolecular aggregates that include but are not confined to three types of collagens, i.e. collagens II, IX, and XI, as major structural and functional components. The collagen molecules are longitudinally organized into quarter staggered arrays with a repeat gap-overlap period, called D, of 67 nm. In immature cartilage, such as the chick embryo sternum, fibrils have a uniform diameter of 17 nm, are randomly oriented (1), and contain large amounts of collagens IX and XI, each representing about 10% of the total collagens (2). In postnatal mammalian hyaline cartilage, the fibrils contain much less of the minor collagens, are heterogeneous in their diameters (up to 200 nm), and form specific patterns depending on the age of the animal and the precise location within the tissue (3). Thus, cartilage at early stages of development appears to contain uniform fibrils, whereas mature cartilage is characterized by fibril populations that are morphologically and biochemically distinct. 1 The surface of most thin cartilage fibrils is populated by D-periodically arranged molecules of collagen IX. This protein is a heterotrimer of genetically distinct ␣1(IX), ␣2(IX), and ␣3(IX) chains, each incorporated into a molecule with alternating triple helical (COL1-COL3) and nontriple helical (NC1-NC4) domains. The NC2 and the NC3 domains connect the triple helical domains and confer flexibility to the molecule. The domains NC1, COL1, NC2, and COL2 are incorporated into the fibril body and may not necessarily be situated at the fibril surface. However, the globular NC4 domain attached to a stalk formed by the COL3 domain projects from the surface of the fibrils outwards and is connected to the rest of the molecule by the flexible NC3 domain (4). The NC4 domain at the aminoterminal end of the ␣1(IX) chain in cartilage is rich in basic amino acids, which has invited the speculation that this part of the molecule may mediate interactions between fibrils and their polyanionic environment. In addition, collagen IX is a proteoglycan (5, 6) with a single chondroitin-dermatan sulfate chain attached to the NC3 region of the ␣2(IX) chain (7,8). These glycosaminoglycan chains may also be involved in the contact between the fibrils and the extrafibrillar matrix.
The precise function of collagen IX in situ still is elusive. Recently, the issue has been addressed by creating transgenic mice harboring an inactivated Col9a1 gene after homologous recombination (9). Unexpectedly in view of the notions discussed above, homozygous mice showed no obvious defects in their skeletal development. However, they developed osteoarthritis with advancing age. This pointed toward collagen IX as a crucial component in the long term integrity of cartilage rather than the formation and development of cartilage matrix as such. However, the expression of collagen IX in these ani-* This work was supported by Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 310) Grant B 12. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: mals was not examined in detail in the previous study. The tissues were devoid of ␣1(IX) mRNA or the corresponding polypeptides, but it remained unclear whether ␣2(IX) and ␣3(IX) mRNA or polypeptides were synthesized. Studies on in vitro reassociation of collagen IX fragments indicated that collagen IX trimers without ␣1(IX) chains may be viable (10,11). Therefore, it remained possible that the animals produced variants of partly functional collagen IX molecules assembled from ␣2(IX) and/or ␣3(IX) chains only, which may largely substitute for normal collagen IX during endochondral bone formation.
Here, we provide this missing information. Thereby, we have also gained further insight into the molecular assembly of collagen IX in mouse cartilage as well as the consequences of its integration into cartilage matrix suprastructures.

MATERIALS AND METHODS
Experimental Animals and Cartilage Preparation-Transgenic mice with an inactivated Col9a1 gene were described elsewhere (9). DBA/2 mice (Charles River, Germany) were used as controls. Rib cages of newborn, adolescent (ϳ4 weeks old), and adult (ϳ1 year old) mice were freed from surrounding noncartilaginous tissue for further biochemical and morphological analysis.
RNA Extraction and RNA Hybridization-Day 17.5 embryos of wildtype, heterozygous, and homozygous mice were homogenized and whole embryo RNA was isolated by the LiCl/urea method described by Auffray and Rougeon (12). Approximately 20 g of RNA were electrophoretically separated on agarose gels and transferred to nylon membranes. Hybridization was performed overnight at 65°C with probes labeled by random priming using [ 32 P]dCTP. The filters were washed two times with 0.5 M sodium phosphate, 7% SDS, 1 mM EDTA, 0.1 mg/ml salmon sperm DNA, 1% bovine serum albumin, pH 7.2. Bound probes were detected by autoradiography at Ϫ70°C on Kodak X-AR x-ray films. The probes used were EcoRI/HindIII cDNA fragments either for mouse ␣2(IX) mRNA or mouse ␣3(IX) mRNA (13). ␤-Actin cDNA was employed as control probe.
Antibodies to Collagens II and IX-Triple helical pepsin fragments of collagens II, IX, and XI were isolated from transplantable (Swarm) chondrosarcoma of rats by limited digestion with pepsin (Serva) and subsequent differential salt precipitation. This procedure yields two disulfide bonded triple helical fragments of collagen IX termed HMW and LMW (14). The purity was judged by SDS-PAGE. 2 Antisera were raised by immunizing rabbits with a mixture of fragments HMW and LMW of rat collagen IX in Freund's complete and incomplete adjuvant as described (15). The specificity of the antisera was checked by immunoblotting (16) using peroxidase-conjugated secondary antibodies (Kirkegaard and Perry Laboratories, Gaithersburg, MD) and 4-chloro-1-naphtol (Sigma) as color reagent. Monoclonal antibodies CIID3 and CIIC1 to chicken collagen II cross-reacting with the mouse homologue but not other cartilage collagens (17) were a gift from Dr. R. Holmdahl (Lund, Sweden).
SDS-PAGE-Samples were prepared for electrophoresis by precipitating the collagens in solution with 3 volumes of ethanol. The protein pellets were dissolved in 0.1 M Tris-HCl, pH 6.8, containing 0.8 M urea, 10% glycerol, and 2% SDS (SDS sample buffer) and were run on 4.5-15% polyacrylamide gradient gels (18).
Extraction of Fibril Fragments, Collagens, and SDS Soluble Molecules-Rib cages from mice of different ages were homogenized with a Polytron (Kinematica, Littau, Switzerland) in 10 volumes of 2 mM sodium phosphate, pH 7.4, containing 150 mM NaCl, 100 mM 6-aminohexanoic acid, 20 mM EDTA, 5 mM benzamidine, 5 mM N-ethylmaleimide, and 0.1 mM phenylmethylsulfonyl fluoride and were subsequently centrifuged at 27,000 ϫ g for 30 min to obtain a clear supernatant. This procedure was repeated twice with fresh extraction buffer. The supernatants containing the fibril fragments were combined, and the pelleted material was divided into two portions. One portion was resuspended in 100 volumes of 0.2 M NaCl, 0.5 M acetic acid, adjusted to pH 2.5 with 12 N HCl, and subjected to digestion with pepsin (100 g/ml) for 48 h at 4°C. To complete solubilization of collagens, the digestion step was repeated. The digests were combined and centrifuged. Finally, total collagens in the neutralized supernatants were precipitated by adding solid NaCl to a final concentration of 4.5 M.
A separate portion was digested for 3 h at 37°C in 150 mM NaCl, 20 mM Tris-HCl, pH 7.4, containing 1 unit of chondroitinase ABC (Seikagaku Kyogo, Tokyo, Japan) per gram of original tissue. Five volumes of ethanol were added to stop the enzymatic reaction and to precipitate proteins. The samples were then centrifuged, and the dried residues were extracted for 3 min at 95°C with SDS sample buffer containing 2% ␤-mercaptoethanol. Immunoelectron Microscopy-All steps were performed at room temperature. Fibril fragments were adsorbed for 2 min onto Formvar/ carbon-coated copper grids (4). The grids were washed with 150 mM NaCl, 2 mM sodium phosphate, pH 7.4 (PBS), and were treated for 30 min with 2% (w/v) dried skim milk in PBS. The adsorbed material was then allowed to react for 2 h with antibodies to collagen II and/or collagen IX in 0.2% (w/v) dried skim milk in PBS. After washing five times for 2 min with PBS, the grids were incubated for 2 h with a suspension of colloidal gold particles (12 or 18 nm) coated with goat antibodies to mouse or rabbit immunoglobulins (Dianova, Hamburg, Germany) in PBS, containing 0.2% (w/v) dried skim milk. For double labeling experiments a mixture of gold particles of two different sizes was used. Finally, the grids were washed with PBS and negatively stained with 2% uranyl acetate. Electron micrographs were taken at 80 kV with a Phillips CM 10 electron microscope.
Measurement of Fibril Diameters and Labeled Fibrils-Electron micrographs were calibrated on the basis of the D ϭ 67 nm banding pattern of the fibrils. Diameters of fibril fragments were measured with a Peak scale magnifying glas 10 ϫ (Plano, Marburg, Germany) on micrographs at final magnifications of at least 6.2 ϫ 10 4 -fold. Gold particles associated with fibrils were counted. Fibrils were considered as labeled if they carried more than 1 average gold particle per 10 D periods.

Expression of Col9a2 and Col9a3 Genes in Col9a1 Knock-out
Mice-In mice with an inactivated Col9a1 gene, the expression of Col9a1 mRNA and ␣1(IX) polypeptides was reduced in heterozygotes and absent in homozygotes (9). To investigate the effect of this genetic alteration on the expression of Col9a2 and Col9a3 genes, mRNA levels in wild-type, heterozygous, and homozygous mice were examined by Northern blot analysis. Transcription of Col9a2 and Col9a3 remained unaffected by the knock-out of the Col9a1 gene because the corresponding transcripts were identical in homozygous or heterozygous transgenic animals and in control animals (Fig. 1).
Next, the presence of collagenous polypeptides in rib cartilage extracts was analyzed. Hyaline cartilage collagens are extensively cross-linked but can be solubilized quantitatively by limited digestion with pepsin to remove small, nontriple

␣1(IX) Chain Absence Leads to Collagen IX Protein Knock-out
helical regions containing the cross-linking sites. Collagen IX is thereby cleaved into two disulfide bonded fragments HMW and LMW. Reduction of HMW produces the polypeptides C2 and C5 derived from the ␣1(IX) and the ␣3(IX) chains, respectively, as well as the peptides C3 and C4, originating from the ␣2(IX) chain (19). The antibodies utilized in the present study specifically reacted in immunoblots with HMW extracted from cartilage of wild-type mice (Fig. 2A, lane 7). After reduction, C2, C3, and C5 were detected (Fig. 2A, lane 9). Extracts from transgenic animals, however, produced no bands corresponding to collagen IX fragments ( Fig. 2A, lanes 8 and 10). By contrast, staining of parallel samples by Coomassie Blue revealed that wild-type and transgenic mice yielded similar quantities of collagens II and XI, irrespective of the age of the animals ( Fig.  2A, lanes 1-6).
Although no polypeptides derived from the ␣2(IX) or ␣3(IX) chains were detected in cartilage of ␣1(IX)-null mice after pepsin extraction, the possibility still remained that such chains were synthesized but not incorporated into pepsin-resistant, triple helical collagen IX molecules. Therefore, rib cages of newborn, adolescent, or adult mice were extracted by boiling in SDS sample buffer containing 2% ␤-mercaptoethanol. This procedure yielded intact collagen IX polypeptides from chick cartilage (20). The crude extracts were subjected to SDS-PAGE and immunoblotting with the antiserum to collagen IX. Polypeptides with apparent molecular masses of 84, 72, and 67 kDa corresponding to ␣1(IX), ␣2(IX), and ␣3(IX) chains, respectively, were detected in samples from normal (Fig. 2B, lane 7) but not ␣1(IX)-null mice (Fig. 2B, lane 8). The crude mixtures of proteins extracted with boiling SDS sample buffer from cartilage pieces of mutant and control mice were indistinguishable by SDS-PAGE (Fig. 2B, lanes 1-6). Thus, lack of collagen IX expression did not conspicuously alter the production of other matrix constituents.
Ultrastructure of Cartilage Fibrils-The consequences of the absence of collagen IX for fibrillogenesis in cartilage were studied by ultrastructural analysis of cartilage fibrils. Fibril fragments were prepared from rib cages of mutant and wild-type mice and characterized by electron microscopy after indirect immuno-gold labeling and negative staining. To verify their cartilage origin, the fibril fragments were double-labeled with monoclonal antibodies recognizing collagen II epitopes together with our antiserum to rat collagen IX. Fibrils at different stages of cartilage development were analyzed to take into account the possibility that age-dependent differences may occur between normal and transgenic mice. Surprisingly, the shapes and ultrastructural details were not obviously different in fibril fragments isolated from wild-type and transgenic mice, respectively. The banding patterns (Fig. 3) and the diameter distributions (Fig. 4) of fibril fragments were the same in normal and ␣1(IX)-null mice. In agreement with the biochemical data, collagen II but not collagen IX could be localized on fibrils from transgenic mice. By contrast, most fibrils from control mice were labeled with both antibodies, regardless of their diameter and the age of the donor animal. DISCUSSION Here, we have demonstrated that a null mutation of the Col9a1 gene encoding ␣1(IX) chains leads to a functional knock-out of collagen IX in mice. Although the mRNAs for the ␣2(IX) and ␣3(IX) chains are normally transcribed, the corresponding polypeptides cannot be detected in cartilage of ␣1(IX)null mice. Presumably, ␣2(IX) and ␣3(IX) chains are rapidly degraded, or their production is suppressed at the translational level in the absence of ␣1(IX) chains. Preceding studies on in vitro reassociation of short polypeptides containing the presumptive interaction sites within the carboxyl-terminal COL1 domain suggested that homotrimeric molecules composed of ␣2(IX) chains or heterotrimeric molecules containing ␣2(IX) and ␣3(IX) polypeptides could be formed in the absence of ␣1(IX) chains (10,11). However, our data indicate that such molecules are not assembled in mouse cartilage in vivo.
Indirect evidence from several studies pointed toward a pivotal role of collagen IX in embryonic skeletal development. In mammals, collagen IX is most abundant in immature cartilages including areas undergoing endochondral ossification (21). The fibrils of chick embryo sternal cartilage have a uniform diameter of ϳ17 nm and contain collagens II, IX, and XI in relative proportions of 8:1:1, respectively (2). Furthermore, in vitro reconstitution experiments with isolated chick cartilage collagens suggested that as much as 10% of collagen IX was required to generate fibrils morphologically similar to  1 and 2), adolescent (lanes 3, 4, and 7-10), and adult (lanes 5 and 6) mice were digested with pepsin, and resistant fragments were separated by SDS-PAGE. Peptic fragments were stained with Coomassie (lanes 1-6) or detected by immunoblotting (lanes 7-10) with antibodies against rat collagen IX. The samples in lanes 1-8 were unreduced, whereas lanes 9 and 10 were separated under reducing conditions. B, rib cartilages from newborn (lanes 1 and 2), adolescent (lanes 3, 4, 7, and 8), and adult (lanes 5 and 6) mice were digested with chondroitinase ABC, and intact collagen polypeptides were extracted with boiling neutral buffer containing 2% SDS and 2% ␤-mercaptoethanol. The samples were subjected to SDS-PAGE and stained with Coomassie (lanes 1-6) or detected by immunoblotting (lanes 7 and 8) with antiserum against rat collagen IX. The genotype at the Col9a1 locus is indicated on top of each lane. Molecular mass markers denoted on the left side of the gels were collagenous polypeptides. The assignment of cartilage collagen polypeptides or their pepsin resistant fragments on the right side of the gels was according to Mayne et al. (14).

␣1(IX) Chain Absence Leads to Collagen IX Protein Knock-out
those of embryonic cartilage (22). Based on these observations, the prediction was that collagen IX-deficient mice fail to produce the thinnest fibrils but are capable of forming thicker fibrils of controlled diameters that are typical in mature cartilage. Therefore, embryonic cartilage that is rich in thin fibrils was expected to be most abnormal in mutant animals. Surprisingly, however, cartilage fibrils from both wild-type and mutant mouse embryos exclusively were thin and unbanded. In samples from adolescent and adult mice, the fibrils showed broader diameter distributions. Again, fibril morphology was indistinguishable between homozygotes and controls. Thus, elimination of collagen IX did not affect the ultrastructural appearance of the cartilage fibrils.
Partial disruption of collagen IX genes seem to cause more obvious phenotypes than complete inactivation. Mice homozygously harboring transgenes with a large deletion in the central portion of the ␣1(IX) chain had a phenotype comparable with that of human chondrodysplasia (23). Heterozygous mice were normal at birth but with time developed osteoarthritislike changes in their articular cartilage. Although the mutant collagen IX protein was not analyzed in this study, the authors surmised that the shortened ␣1(IX) chains competed with endogenous ␣1(IX) chains in heterotrimer formation because the phenotype severity correlated with transgene expression. An-other mouse strain, with a transgene encoding only the NC4 domain of ␣1(IX) had a similar phenotype (24). Overexpression of this transgene led to osteoarthritis, and again the severity depended on the age and level of transgene expression. Consistently with these transgenic mouse models, multiple epiphyseal dysplasia, a human dominant heritable disorder characterized by mild skeletal malformations and early onset osteoarthritis, can be caused by heterozygous mutations in the Col9a2 gene (25). These observations have prompted the hypothesis that collagen IX acts as a tissue stabilizer through interactions between the fibril surface and the extrafibrillar matrix, perhaps via its NC4 domain (26). This role of collagen IX also is consistent with the cartilage pathology of collagen IX-deficient mice.
Recent studies suggested that collagen XI, the other minor collagen of cartilage fibrils, is essential for skeletal morphogenesis, including the formation of thin cartilage fibrils. The phenotype of cho/cho mice, which includes abnormally thick cartilage fibrils, is caused by the absence of the ␣1 chain of collagen XI in extracellular matrices (27). Further, unusually thick fibrils are found in cartilage of patients with Stickler syndrome, where a genetic defect results in abnormally short ␣2(XI) chains which presumably compromises the molecular assembly of collagen XI (28). In addition, collagen XI is associated predominantly with fibrils less than 25 nm in diameter in human juvenile rib cartilage, whereas collagen IX is present on fibrils of various sizes in this tissue (29,30). However, experiments on fibril reconstitution from soluble avian collagens II and XI in vitro showed that control of lateral fibril growth was incomplete. At least one further component was required, and collagen IX was shown to satisfy this need (22). Additional surface components of cartilage fibrils include the small proteoglycans decorin and fibromodulin, which alter fibrillogenesis of collagens I and II in vitro (31,32). Our recent studies demonstrated that decorin and collagen IX coexist on some fibrils from fetal bovine epiphyseal cartilage. 1 Therefore, these FIG. 3. Collagen IX is present in cartilage fibrils from normal but not from ␣1(IX)-null mice. Immunoelectron microscopy of fibril fragments extracted from newborn (A and B), adolescent (C and D), and adult (E and F) rib cartilage of transgenic (A, C, and E) or wild-type (B, D, and F) mice. Fibrils were double-labeled with polyclonal antibodies to collagen IX and with monoclonal antibodies to collagen II. Bound antibodies were visualized by indirect immunostaining with gold-conjugated antibodies to rabbit or mouse immunoglobulins. 18-and 12-nm gold particles correspond to collagen IX and collagen II staining, respectively. Bars, 200 nm. molecules may participate in fibril formation and may thus be functionally redundant with collagen IX in controlling lateral aggregation of collagens. Similar to the mice with an ␣1(IX) knock-out, skeletal malformations were not observed in decorin-deficient mice (33). Targeted disruption of the decorin gene caused abnormal collagen fibril morphology and skin fragility. Studies on mice that are deficient in both collagen IX and decorin are in progress and may help to clarify the roles of fibril surface components in cartilage.