Targeted disruption of Col11a2 produces a mild cartilage phenotype in transgenic mice: Comparison with the human disorder otospondylomegaepiphyseal dysplasia (OSMED)



Transgenic mice were prepared by homologous recombination with a Col11a2 targeting gene in which an inverted neomycin-resistant gene was inserted between restriction sites in exons 27 and 28. The targeted allele was transcribed in shortened mRNAs, which could be detected by Northern blotting. However, translation of the full-length Col11a2 chain was unable to occur because of the presence of premature termination codons within the inverted neomycin-resistant gene. Analysis of pepsin-resistant collagen chains from rib cartilage of homozygous mice demonstrated the lack of synthesis of intact α2(XI) chains. However, pepsin-resistant collagen chains of either α1(XI) or α1(V) were still detected on sodium dodecyl sulfate polyacrylamide gel electrophoresis. Therefore, α2(XI) chains are not essential for the assembly of some molecular forms of triple-helical type V/XI collagen. The phenotype was milder than in the cho/cho mouse in which, as the result of mutation, translation of the full-length α1(XI) chain fails to occur and the mice die at birth (Li et al., 1995). Homozygous mice without expression of an α2(XI) chain had a smaller body size, receding snouts, and deafness. Nasal bones in the homozygous transgenic mice were specifically shorter and dimpled on their external surfaces. Chondrocytes in growth plates of all long bones were markedly disorganized and failed to align in columns. Analysis of growth plates from transgenic mice by in situ hybridization showed expression of α1(II) and α1(XI) but not of α1(I) or α1(V) which, in contrast, were expressed in the developing bone and in the bone collar. Expression of α1(X) specifically in the hypertrophic cartilage was observed in normal and transgenic mice. No obvious osteoarthritis was observed throughout the life of homozygous mice up to 1 year of age, although minor morphologic anomalies in the articular cartilages were discernible. The mild phenotype is consistent with similar mutations in the COL11A2 gene seen in patients with nonocular Stickler syndrome and some patients with otospondylomegaepiphyseal dysplasia (OSMED), as well as in patients with a nonsyndromic form of deafness called DFNA13. © 2001 Wiley-Liss, Inc.


Type XI collagen is a member of the growing family of collagens, each of which has a unique biological function (Mayne and Burgeson, 1987; Vuorio and de Crombrugghe, 1990; Prockop and Kivirikko, 1995). Type XI collagen was originally identified in cartilage as a minor fibrillar collagen that is similar to the more abundant type II collagen. The protein is assembled from three α-chains designated as α1(XI), α2(XI), and α3(XI) (Burgeson and Hollister, 1979; Reese and Mayne, 1981; Burgeson et al., 1982). However, the α3(XI) chain is derived from the same gene as the α1(II) chain of type II collagen which, by an unknown mechanism, becomes assembled with the α1(XI) and α2(XI) chains to form a unique procollagen molecule (Burgeson and Hollister, 1979; Reese and Mayne, 1981; Burgeson et al., 1982; Eyre and Wu, 1987; Morris and Bächinger, 1987; Morris et al., 1990; Thom and Morris, 1991). In addition, it was established that the structure of type XI collagen is closely related to the structure of type V collagen. Both type V and XI collagens are found in small amounts in cultured cells and in both cartilaginous and noncartilaginous connective tissues (Bernard et al., 1988; Brown et al., 1991; Kleman et al., 1992). During the maturation of bovine bone, isolated fractions rich in type V collagen contain increasing proportions of the α1(XI) chain and a decreasing proportion of the α1(V) chain (Niyibizi and Eyre, 1989). Analyses of collagen fragments from mammalian vitreous humor demonstrated that fibrils were assembled from molecules containing both α1(XI) and α2(V) chains but not the α2(XI) chain (Mayne et al., 1993). Therefore, from the above results, it was proposed that type V and type XI collagens are a heterogeneous class of collagen molecules assembled from five or six different chains (Mayne et al., 1993). Collectively, these molecules are called type V/XI collagen.

The biological role of type XI collagen has, in part, been demonstrated by examination of genetic mutations. The cho/cho mouse, which is homozygous for a premature termination codon in the Col11a1 gene, dies at birth. Death is most likely caused by an incorrect assembly of cartilage matrices in the lungs, trachea, and ribs, leading to an inability to inflate the lungs (Li et al., 1995). In man, heterozygous mutations in the COL11A1 gene cause a variant of chondrodysplasia with associated eye involvement (Richards et al., 1996; Annunen et al., 1999; Martin et al., 1999), recently designated (OMIM #604841) as Stickler syndrome type II or Marshall syndrome (OMIM #154780) (Griffith et al., 1998). Mutations in the gene for COL11A2 leads to a spectrum of clinical conditions ranging from (1) heterozygotes, who present with nonsyndromic hearing loss called DFNA13 (McGuirt et al., 1999); (2) heterozygotes or homozygotes with another variant of the Stickler syndrome but without eye involvement called Stickler syndrome type III (OMIM #184840), or the related syndrome of otospondylomegaepiphyseal dysplasia (OSMED) (Vikkula et al., 1995; van Steensel et al., 1997; Sirko-Osadsa et al., 1998; Pihlajamaa et al., 1998); and (3) homozygotes or compound heterozygotes with a more severe form of OSMED in which the mutations give rise to premature stop codons so that probands are predicted to be unable to synthesize full-length α2(XI) chains (Melkoniemi et al., 2000). Here, we describe the phenotypic properties of transgenic mice with a targeted disruption of the Col11a2 gene and compare these mice with OSMED patients with mutations in COL11A2.


Preparation of Transgenic Mice

A line of transgenic mice with a targeted allele for Col11a2 (Liu et al., 1996) was prepared by homologous recombination in ES cells, by using a construct that spanned intron 8 to 41 of the Col11a2 gene (Fig. 1). The neomycin gene (neo) driven by the phosphoglycerate kinase promoter was inserted in the reverse orientation between two restriction sites in exon 27 and 28 so that it joined sequences from the two exons. Heterozygous mice for the null allele were bred to generate homozygous mice as shown by Southern blot analysis (Fig. 2A). Northern blot analysis of mRNA from chondrocytes isolated from the xiphoid cartilage of homozygous mice demonstrated the absence of full-length mRNA for Col11a2 (Fig. 2B). However, at least two short RNA transcripts of 4.2 kb and 3.8 kb hybridized with a probe containing the sequences of exon 62 to exon 65 of the Col11a2 gene. Transcription, therefore, can proceed through the invented neo, but the mRNA is unstable, probably due to the inclusion of the inverted neo sequence within the transcripts (Culbertson, 1999). Reverse transcriptase-polymerase chain reaction (RT-PCR) was performed on total RNA from the femoral epiphyseal cartilage with the forward primer targeted to exon 23 and the reverse primer targeted to exon 29. Four bands, ranging in size from 1.0 kb to 1.4 kb, were obtained after PCR rather than a single band of anticipated size approximately 1.5 kb. All four bands were cloned and completely sequenced. The results showed that the 3′-end of exon 26 was consistently spliced to the same A/B boundary located 71-bp downstream of the 3′-end of the neomycin-resistant gene (labeled A in Fig. 3). However, three of the four transcripts were found to have undergone additional internal splicing of sequences within the inverted neomycin-resistant gene (designated as H2, H3, and H4; Fig. 3). All splicing events were located between GT/AG consensus dinucleotides. All four transcripts contained premature termination codons in all three frames either in the sequence labeled G or in the downstream polylinker sequence. Therefore, translation is unable to proceed beyond the inverted neo insertion. We found no evidence that the neo sequence could be completely spliced from the transcripts or for other splicing events as described previously for neo sequences (Forsberg et al., 1996).

Figure 1.

Schematic of the relevant region of the Col11a2 gene, the targeting vector, and the mutant Col11a2 gene after homologous recombination. The genomic fragment for the targeting vector spanned intron 8 to intron 41. TK, herpes simplex thymidine kinase gene; neo, neomycin-resistant gene; E, X, and B, restriction sites for EcoRI, Xhol, and BamHI.

Figure 2.

A: Southern blot analysis identifying heterozygous and homozygous transgenic mice. The probe was an EcoRI fragment that spanned intron 41 to intron 49 (see Fig. 1 and text). B: Northern blot analysis of cartilage RNA from the wild-type and homozygous mice. Xiphoid cartilage was dissected from mice and chondrocytes were obtained by enzymatic digestion of tissue. Total RNA was extracted and analyzed by Northern blot analysis by using a HindIII/BamHI cDNA fragment encoding exons 62 to 65.

Figure 3.

Schematic representation of the results obtained by reverse transcriptase polymerase chain reaction on cartilage mRNA from homozygous mice by using primers located in exons 23 and 29. Four bands were detected from 1.0 to 1.4 kb, and all of these bands were subcloned and sequenced. In all the Col11a2 transcripts from the homozygous mice, the 3′-end of exon 26 was found to be spliced to an acceptor site approximately 71 nucleotides downstream of the 3′-end of the inverted neomycin-resistant gene (designated H1–H4). In three of the transcripts, there was also internal splicing of coding sequences of the inverted neomycin-resistant gene (H2, H3, and H4).

To confirm that the Col11a2 allele was inactivated, rib cartilage was dissected from 1-month-old wild-type and homozygous transgenic mice, the tissue was digested with pepsin and the collagenous components were assayed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) for the presence of type XI collagen. As shown in Figure 4, lane B, the α1(XI) or α1(V) chain of type V/XI collagen was still present. However, the α2(XI) chain was absent. The α3(XI) chain was not separated from the abundant α1(II) chain that appeared to be synthesized in normal amounts.

Figure 4.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis of pepsin-resistant collagen from the rib cartilage of 1-month-old wild-type (left) and homozygous mice (right). Equal amounts of sample were applied to the gel. The protein bands were visualized by Coomassie blue staining to identify α-chains of type II and type XI collagens. A band of α2(XI) chain cannot be detected in the homozygous rib extract.

Phenotype of the Mice

Heterozygous mice at birth were indistinguishable from normal littermates. Homozygous mice were approximately 25% smaller than normal littermates (8.51 g ± 1.87 SD [n = 14] vs. 11.1 g ± 0.85 SD [n = 6]; P = 0.005). Homozygous mice were also recognizable from control and heterozygous littermates in that they had a shorter snout and a slightly bulged forehead (Fig. 5A). They continued to be of slightly smaller body size up to 2 years of age and could be readily identified by their more triangular face. These distinguishing features were most apparent in stained skeletons (Fig. 5B). However, none of over 200 newborn homozygote mice that were examined had a cleft palate as is commonly observed in OSMED patients (Melkoniemi et al., 2000). When examined at 4 months of age, homozygous mice did not respond to loud noise, an observation suggesting they had a hearing deficiency. Subsequently, click-evoked auditory brainstem response testing confirmed this, and the structural basis of the deficiency seemed to arise from the disorganized collagen fibrils of the tectorial membrane (McGuirt et al., 1999). There were no obvious gross abnormalities of the eye or other major organs.

Figure 5.

A: Photographs of normal and homozygous mice to show the facial differences. B: Stained skeletons of newborn wild-type (+/+), heterozygous (+/-) and homozygous mice (-/-). Note the facial differences in the homozygous mouse. C: Appearance of the skulls of wild-type and homozygous mice. Note the steeper slope of the skull and the indentations in the nasal bones (arrow) in the homozygous mouse. D: Comparison of length-to-width ratios of bones on the dorsal aspects of skulls from 1-year-old wild-type (+/+) and homozygous (-/-) mice. Each data point represents the average ± SEM from seven skulls. Bones are arranged in a posterior (INT) to anterior (NAS) order. The difference in ratio for the nasal bone is significant at P ≤ 0.001 and arises from the shorter length of the bone in the homozygote skull. INT, interparietal bone; PAR, parietal bone; FRO, frontal bone; PRE, premaxillae; NAS, nasal bone.

Craniofacial Morphology

Gross anatomic observation of skulls from 1-year-old mice revealed that the nasal bones of the homozygous mice were more sloped and their medial surfaces were often depressed (Fig. 5C arrow). Thus, in situ, the two adjoining nasal bones produced a noticeable depression. The length to width ratios of the nasal bone of homozygous mice showed a significant difference when compared with wild-type, but other bones of the skull (interparietal, parietal, frontal, and premaxillae bones) were not significantly affected (Fig. 5D).

In coronal sections of the skull of 1-year-old mice, the difference in indentation and shape of the nasal bones were clearly recognizable (Fig. 6A,C). In coronal sections of homozygote neonates (Fig. 6D), the nasal bones were caved deeply toward the Y-shaped cartilage arising from the developing nasal septum especially when compared with sections from wild-type mice (Fig. 6B). The nasal bone develops by intramembranous ossification, but its shape is determined by the adjoining cartilage (Fig. 6B,D). The differences in the length, slope, and indentation of the nasal bone in homozygous mice, therefore, may be accounted for by differences in the shape of the cartilage, which acts as a template for the formation of the nasal bones (Fig. 6B,D).

Figure 6.

A,C: Coronal section of the anterior skull of wild-type and homozygous mice at 1 year of age. A pronounced depression formed by the two nasal bones (N) is apparent in the homozygous mouse, whereas the premaxilla bone (P) appears unaffected. B,D: Development in neonatal mice of the nasal bones (N) in which the septal cartilage (S) acts as a template. Note that the nasal bones project more inferiorly in the homozygous mouse. NA, nasal cavity.

Morphology of Long Bones

Paraffin-embedded sections of knee joints from wild-type and homozygous mice at 1-month postpartum were prepared. Examination of these sections showed a profoundly disorganized growth plate in the homozygote with the chondrocytes failing to align in columns (Fig. 7B,D). Similar results were found for all other growth plates that were investigated. In the wild-type, growth plates were well organized with chondrocytes being aligned in longitudinal columns so that individual resting, proliferating, and hypertrophic zones could be easily recognized (Fig. 7A,C). In contrast, the growth plate of homozygous mice lacking α2(XI) appeared completely disorganized and, on closer examination, the nuclei of some chondrocytes appeared pyknotic. However, hypertrophic chondrocytes could often be observed just below the proliferating layer, indicating that terminal differentiation of chondrocytes was able to proceed normally. The tibial articular cartilage was on average thinner than the wild-type (compare Fig. 7E,F). The same was observed for all articular cartilages at 1 month of age (see Fig. 9). In addition, chondrocytes of the articular cartilage of the homozygous mice were more randomly organized (compare Fig. 7E and 7F). Transmission electron microscopy of the growth plate showed the presence of fine collagen fibers often aligned between the columns of cells in both wild-type and homozygous transgenic mice (Fig. 8A,B). The collagen fibrils of the homozygote mouse often appeared less organized in lateral arrays than in the wild-type mouse as previously observed in the tectorial membrane (McGuirt et al., 1999). In some areas of the proliferative zone, the fibrils became closely associated with each other and formed electron-dense bundles (Fig. 8C). However, the formation of large aggregates of fibrillar collagen as described for the cho/cho mouse did not occur (Li et al., 1995).

Figure 7.

Histologic analysis of 1-month-old postpartum tibial epiphyses from wild-type (A,C,E) and homozygous mice (B,D,F). In the wild-type mouse (+/+), the well-organized growth plate contains resting, proliferating, and hypertrophic layers. In the homozygous (-/-) mouse, the columns of chondrocytes fail to align properly and the cells appear disorganized often forming random clusters. E and F: Note the difference in organization of the chondrocytes in the articular cartilage. The wild-type forms a well-organized matrix, whereas in the homozygous animal, the articular cartilage is thinner and the cells appear mildly disorganized. Scale bars = 125 μm in B (applies to A,B), 50 μm in D (applies to C,D), 25 μm in E (applies to E,F).

Figure 8.

Transmission electron microscopy of the femoral growth plate (1-month postpartum) of a wild-type (+/+) (A) and homozygous (-/-) transgenic mouse (B,C). A and B show comparable regions located in the proliferative zone of the growth plate between two columns of chondrocytes. Note the parallel alignment of the fibrils in A and the slightly less ordered fibrils in B. C is a representative area of the proliferative zone where the chondrocytes are highly disorganized. Note that the fibrils are often closely associated laterally to form electron dense bundles. Scale bars = 500 nm

We additionally examined sections of the femurs of mice at 15.5 days, 17.5 days, neonate, and at 1 week, 2 weeks, and 4 weeks after birth. Before the formation of a distinct growth plate, the chondrocytes do not align in columns and were not markedly disorganized when the wild-type and the homozygote were compared. However, at 1, 2, and 4 weeks postpartum, the lack of organization of the chondrocytes became more clearly recognizable and was most prominent after the growth plate had fully formed at 4 weeks (results not presented). Within a limb, all growth plates showed a similar disorganization in the homozygotes. Therefore, it seems that the α2(XI) chain is required for the correct organization of the collagen fibrils of each growth plate but may not play a critical role at early stages of chondrogenesis.

To evaluate chondrocyte differentiation, we performed in situ hybridization on the neonatal tibia for Col1a1, Col2a1, Col10a1, Col11a1, and Col5a1. All examined probes for these collagen genes showed the same expression patterns in the tibia of wild-type or homozygote mice, with the α1(V) collagen chain being expressed in the developing bone and bone collar and not in the cartilage (Fig. 9H,J). Type X collagen was expressed only in the hypertrophic cartilage (Fig. 9C,F). Additionally, we performed immunohistochemical staining with antibodies against type II collagen and type X collagen. Type II collagen was detected throughout the cartilaginous region of the growth plate, but type X collagen was detected only in the hypertrophic layer in both wild-type and homozygous mice (data not presented). Thus, although the growth plate is markedly disorganized, the chondrocytes are, nevertheless, proceeding to hypertrophy.

Figure 9.

In situ hybridization of several collagen mRNAs in the tibial growth plate. Col1a1 (A,D) and Col5a1 (H,J) are only detected in the bone collar and in the developing bone. Col2a1 (B,E) and Col11a1 (G,I) are detected both in the resting and proliferating layers of the developing cartilage but not in the hypertrophic cartilage. Col10a1 (C,F) is detected only in the hypertrophic chondrocytes. No difference in the expression patterns could be recognized for the five collagen mRNAs in both wild-type and homozygous mice. Scale bars = 250 μm in F (applies to A–F), in J (applies to G–J).

Articular Cartilage

We measured the thickness of epiphyses from several long bones, because initial observations suggested that they were thinner in homozygotes (compare Fig. 7E and 7F). All examined articular cartilages (proximal and distal femur, acetabulum, proximal and distal humerus, glenoid fossa, proximal ulna, and proximal radius) were on average significantly thinner at 1 month postpartum in the homozygous mice (Fig. 10). There was no sign of significant erosion of the articular surfaces in either wild-type or homozygous mice. However, the articular cartilage of the homozygous mice could often be distinguished by failure of the chondrocytes to align in a columnar manner (Fig. 7E,F). In addition, chondrocytes of homozygotes often had distorted shapes. Joints were examined up to 1 year of age, but no evidence could be found for increased osteoarthritis in the homozygotes compared with the wild-type, this agreeing with a previous study (Lapveteläinen et al., 2001).

Figure 10.

Comparison of the thickness of joint cartilages in 1-month-old mice. All the articular cartilages from homozygous (-/-) mice were thinner than for wild-type (+/+) joints. Average thickness ± SD of articular cartilages at five knee joints, two hip joints, five shoulder joints, and three elbow joints of four pairs of wild-type (+/+) and homozygous (-/-) mice. Data points represent from 14 to 70 measurements per joint for a total of 420. At the shoulder joint, the thickness of cartilage was not significantly different at the glenoid fossa (SG), and only significant at the 90% level (P = 0.10) at the humeral articular surface (SH). KF, knee joint, femur; HF, hip joint, femur; HA, hip joint, acetabulum; SH, shoulder joint, humerus; SG, shoulder joint, glenoid fossa; EH, elbow joint, humerus; EU, elbow joint, ulna; ER, elbow joint, radius


As shown by Northern blotting, targeting of the Col11a2 gene in the present transgenic mice did not prevent transcription of the gene but generated shortened mRNAs. These arise most likely from instability in the mRNA introduced by the presence of the inverted neomycin-resistant gene. By bridging PCR reactions across the neo insertion, several splicing variants were found in which the normal acceptor site at the 5′-end of exon 27 was never used. Instead, an acceptor site within the inverted neo sequence was preferred. In addition, three additional internal splicing events also occurred in the inverted neo sequence during transcriptional processing events. However, the sequence labeled G in Figure 3 of the neo together with the polylinker sequence was found to contain stop codons in all three reading frames. Translation, therefore, is unable to proceed beyond the inverted neo sequence. This was also demonstrated by direct isolation of pepsin-resistant collagen from rib cartilage, which on SDS-PAGE showed no α2(XI) chains (Fig. 4). Therefore, the homozygous mice described in this study possess nonfunctional Col11a2 alleles and are directly comparable to OSMED patients, who are predicted to be unable to synthesize full-length α2(XI) chains (Melkoniemi et al., 2000).

Previous observations in mice and humans suggested that mutations in the α1(XI) and the α2(XI) chains of type XI collagen would produce significant phenotypes in both heterozygotes and homozygotes. The predicted premature termination codon in the Col11a1 gene of cho/cho mice produces a lethal phenotype in homozygotes with defects in the cartilage of ribs, mandible, and trachea (Li et al., 1995). These mice also have shorter long bones and wide metaphyses. Therefore, the phenotype of the cho/cho mouse shows that the α1(XI) chain of type XI collagen is essential for the formation of normal cartilage. In humans, heterozygous mutations in the COL11A1 gene resulted in Stickler syndrome with ocular abnormalities frequently causing vitreous degeneration and retinal detachment, together with midfacial hypoplasia, hearing loss and vertebral dysplasia (Richards et al., 1996; Griffith et al., 1998; Annunen et al., 1999). Heterozygous mutations in the COL11A2 gene produced similar but milder phenotypes (Vikkula et al., 1995; van Steensel et al., 1997; Sirko-Osadsa et al., 1998; Pihlajamaa et al., 1998; McGuirt et al., 1999) that have been variously classified as DFNA13, nonocular Stickler syndrome, Weissenbacher-Zweymüller syndrome, and heterozygous OSMED. A mutation of the COL11A2 gene that produced an in-frame splicing out of one exon was found in a variant of the Stickler syndrome (or heterozygous OSMED). It lacked the ocular changes seen in classic Stickler syndrome, because the α2(XI) chain is not present in the vitreous collagen fibrils and is replaced by the α2(V) chain (Mayne et al., 1993). Two substitutions for obligate glycine residues in the α2(XI) chain produced a similar phenotype of nonocular Stickler syndrome in heterozygotes. In homozygotes, one of the same substitutions for obligate glycine residues produced a more severe phenotype of OSMED that included generalized epiphyseal dysplasia, sensorineural deafness, and midface hypoplasia (Vikkula et al., 1995; van Steensel et al., 1997). Some of the homozygous patients with mutations in COL11A2 are predicted to be null for the α2(XI) chain (Melkoniemi et al., 2000). In the transgenic mice prepared here, heterozygotes had no discernible abnormal phenotype. The phenotype of the homozygous mice consisted of smaller body size and shorter snouts. Also, as shown previously, collagen fibers in the tectorial membrane of the cochlea were not properly aligned, presumably resulting in nonsyndromic deafness (McGuirt et al., 1999). Microscopic observation revealed that the growth plates were disorganized such that chondrocytes in them were not aligned in columns compared with the wild-type mice. The disorganization of the growth plate was seen in most long bones, including femur, tibia, fibula, humerus, radius, and digits. Marked disorganization of bone development was also previously seen in some patients with mutations in the COL11A2 gene (Melkoniemi et al., 2000). On the other hand, in situ hybridization for expression of mRNAs for several other collagen molecules revealed no differences in location in the growth plate of the wild-type and homozygous mice, suggesting chondrocyte differentiation is occurring normally. In mice, Col11a2 may be required for the formation of an organized collagen network within the growth plate. Chondrocytes can then continue to remain correctly aligned during development.

One interesting observation was that a band that migrates to the location of either α1(XI) or α1(V) chains was present after SDS-PAGE on collagen isolated from the rib cartilage of homozygous mice after pepsin digestion (Fig. 4). However, in situ hybridization subsequently showed that the α1(V) chain was not expressed in the cartilage of the homozygotes (Fig. 9H,J). Therefore, the present results provide direct evidence that in certain conditions the α1(XI) chain can form triple-helical molecules in cartilage that do not contain α2(XI) chains. At present, we do not know if the α1(XI) chain forms a homotrimer or will form heterotrimers with the α3(XI) chain. However, it seems likely that the type XI molecules synthesized by the homozygotes are partially compensating for the lack of an α2(XI) chain. Interestingly, it was reported that α1(V) chains can be expressed in human embryonic kidney cells and will form an [α1(V)]3 trimer (Fichard et al., 1997). Formation of [α1(XI)]3 in the transgenic mice, therefore, is a distinct possibility. Somewhat similar results were found for a targeted deletion of the α1(II) chain where continued synthesis of pepsin-resistant forms of the α1(XI) and α2(XI) chains could be detected (Aszódi et al., 1998).

Especially relevant to the present results are OSMED patients for which mutations have been identified in both alleles that result in stop codons and, therefore, are expected to be functional nulls for α2(XI) (Melkoniemi et al., 2000). These patients consistently are of short stature, have profound hearing loss, marked facial hypoplasia, but no ocular deficiency. This phenotype seems directly comparable to the transgenic mice described in this study. The only difference is that this group of OSMED patients consistently presents with a cleft palate, whereas we have never observed a cleft palate in over 200 transgenic mice.

The current results are in sharp contrast to previous results with mice that are knockouts for the α1(IX) chain of type IX collagen (Fässler et al., 1994). In mice lacking an α1(IX) chain, the α2(IX) and α3(IX) chain continue to be synthesized but are not incorporated into a pepsin-resistant protein (Hagg et al., 1997). Therefore, the α1(IX) chain is essential for the assembly of triple-helical type IX collagen, whereas the α2(XI) chain is not essential for the assembly of some forms of triple-helical type V/XI collagen.

Previous work showed type XI collagen is expressed in the growth plate (Sandell et al., 1994; Balmain et al., 1995; Vornehm et al., 1996) where it may play an important role in its continued development (Wardale and Duance, 1993; Keene et al., 1995). In the growth plate, the collagen fibrils between the columns of chondrocytes are known to lie parallel to the columns (Hunziker, 1998). It may be that loss of this organization results in the disorganized growth plate observed in the transgenic mice described here. The present results and the recent publication on the changes in the tectorial membrane (McGuirt et al., 1999) suggest that the α2(XI) chain may be required for the correct assembly of lateral associations between individual collagen fibrils.


Gene Construct

The Col11a2 gene was isolated from a cosmid library prepared with a commercial vector (pWE15; Stratagene) and DNA from a 129/Sv mouse. The targeting vector (PVII) was prepared so that two genomic fragments spanning intron 8 to exon 27 and exon 28 and intron 41 were interrupted by a neomycin-resistance gene in the reverse orientation (Fig. 1). Also, a herpes simplex virus thymidine kinase gene was included at both ends of the targeting vector. Construction of the vector was initiated with a vector (pPNT) containing both a gene for neomycin-resistance and a gene for herpes simplex virus thymidine kinase (Tybulewicz et al., 1991). Because the neomycin-resistance gene in the vector proved inactive, it was replaced by a neomycin-resistance gene from a plasmid (pMC1-Neo-Oikt/a; Stratagene). The first step was to insert between the EcoRI/KpnI sites in pPNT a 4.8-kb EcoRI/KpnI fragment from a subclone of the Col11a2 gene containing sequences spanning intron 8 to exon 27. The second step was to insert between the KpnI/NotI sites in pPNT a 4.3-kb KpnI/NotI fragment from a subclone of the Col11a2 gene that contained sequences spanning exon 28 to intron 41. (During this step with the KpnI/NotI digestion the inactive PGK-neo was removed.) The third step was to insert into the BamHI site contained in the 3′-end of the 4.3-kb KpnI/NotI fragment (exon 28 to intron 41) a BamHI/BamHI fragment containing the herpes simplex virus thymidine kinase gene from a plasmid (pHSV106; GIBCO/BRL). Finally, the TK-neo derived from pMCI-neo by digestion with XhoI and SalI was passaged through SalI cut pBluescript, so that it was now flanked by AccI and ClaI sites. The TK-neo was liberated with AccI and ClaI digestion and cloned in the reverse orientation into a ClaI site now located in the polylinker sequence between exons 27 and 28.

To prepare transgenic mice, the PV11 construct was linearized with NotI and electroporated into embryonic stem cells from 129/Sv mice with 50 μg of DNA and 107 cells at 830 V and 3.0 μF. The cells were plated onto a feeder layer of mouse fibroblasts that had been transfected with the neomycin-resistance gene and selected for resistance with G418. The embryonic stem cells were selected for 7–10 days with 400 μg/ml of G418 and 2 μM gancyclovir (Syntex, Inc). Resistant clones were expanded for Southern blot analysis. Cells from one of 20 correctly targeted clones were injected into 2.5-day blastocysts removed from pregnant mice of the FVB/N line, and the blastocysts were implanted into the uterine horns of pseudopregnant mice of the CD-1 line. Mice with chimeric coat color were bred into wild-type mice of the FVB/N strain for at least eight generations. Transgenic offspring were identified by Southern blot analysis for germline transmission of the mutated allele.

Southern Blot Analyses

DNA was obtained from the tail by digestion with proteinase-K, extracted with phenol/chloroform, and precipitated with ethanol. The DNA was digested with BamHI, electrophoresed on an agarose gel, and transferred to a nylon membrane (Biotrans; ICN). The membrane was probed with a 1.8-kb EcoRI fragment that spanned introns 41 to 49 of the Col11a2 gene and that was labeled with [32P]dCTP by random primer extension.

Northern Blot Analyses

For extraction of mRNA, costal xiphoid cartilage was dissected from mice. The cartilage was digested for 2 hr at 37°C with 2 mg/ml of collagenase and 0.025% trypsin, and the chondrocytes were isolated by centrifugation. The cells were cultured overnight in DMEM containing 10% fetal calf serum, and mRNA was extracted (Micropoly(A) Pure Kit; Ambion, Inc) The mRNA was separated on a 1% agarose gel and blotted onto a nylon filter. The filter was probed with a mouse cDNA fragment that was 32P-labeled (Vandenberg et al., 1996). The cDNA fragment was prepared by RT-PCR of total RNA from mouse chondrocytes. The primers were designed to span the codons in exons 62 to 63 (63: 5′ GTT CCG TGG ACG GAA GCA ACC 3′) and in exon 65 (65: 5′ AGG GCT CAG CTC ATC CTC GTT 3′). The PCR product was cloned into the EcoV site of a plasmid vector (pT7Blue Novagen) and isolated from the amplified plasmid by HindIII/BamHI digestion.

RT-PCR Assays

Initially, RT-PCR was carried out to amplify fragments from exon 23 (XI-1: 5′ CAG GAC CTC CTG GAC AAC A 3′) to exon 29 (XI-5: 5′ AGG AAA GCC GTC CTC GCC CT 3′) by using total RNA from the epiphyseal cartilage of both +/+ and -/- animals. PCR was performed for 35 cycles with the following conditions: denaturation for 30 sec at 94°C, annealing for 30 sec at 60°C, and extension for 1 min at 72°C. The PCR products were run on a 1% agarose gel. The PCR products were directly subcloned into pGEM-T easy vector (Promega Corp.). DNA sequences were determined manually by the dideoxy-chain-termination method by using a Thermosequenase radiolabeled terminator cycle sequencing system (Amersham Pharmacia Biotech).

Analysis of type XI Collagen in Cartilage

Rib cartilages from 1-month-old mice with the targeted deletion or control wild-type mice were dissected free of visible muscle and connective tissues under a dissecting microscope followed by extraction in 0.02 M phosphate buffer, pH 7.2, containing 2 mM EDTA, 10 mM NEM, and 1 mM PMSF at 4°C overnight. The pellet was collected by centrifugation at 30,000 × g for 30 min and resuspended in 0.5 M acetic acid containing 0.2 M NaCl. Collagenous molecules were digested by pepsin (1 mg/ml) at 4°C for 24 hr. To complete solubilization of collagen, this step was repeated three times. After neutralization by dialysis against 0.1 M Tris-HCl, pH 8.0, the supernatant was dialyzed against 0.7 M NaCl in 0.5 M acetic acid. The pellet was removed by centrifugation (30,000 × g; 30 min), and then the supernatant was dialyzed against 2.0 M NaCl in 0.5 M acetic acid for 2 days to precipitate the collagen. The final pellet was dissolved in 0.5 M acetic acid, dialyzed against 0.1 M acetic acid extensively, and then lyophilized. Each extract from +/+ and -/- mice was analyzed by 6% SDS-PAGE under reducing conditions.

Staining of Skeletons

The skin and internal organs were removed and samples were fixed in 95% ethanol for 2 days followed by staining in 0.15% Alcian blue dissolved in 75% ethanol and 20% acetic acid for 2 days. The samples were dehydrated under 100% ethanol for 2 days and immersed in 1% KOH for 2 days. The samples were then stained with 0.001% Alizarin red S in KOH for 2 days before being dehydrated in graded solutions of glycerin and stored in 100% glycerin.

Gross Anatomic Observation

To prepare dry skulls, heads from the wild-type and homozygous mice were skinned and fixed in 4% buffered formaldehyde for 3 days. The specimens were then washed for several hours in running tap water and immersed for several days in 1% KOH. After further washing soft tissues were carefully removed and skulls were dried. Length and width of the dorsal aspects of interparietal, parietal, frontal, premaxillae, and nasal bones were measured with a caliper that was read on a millimeter scale.

Coronal Sections of Skulls

Skulls of 1-year-old mice fixed in 4% formaldehyde were prepared as above then decalcified in Cal-Ex (Fisher Scientific), and coronal sections were prepared and stained with hematoxylin and eosin. For neonatal mice, coronal sections were prepared in the same manner as 1-year-old mice but without decalcification.

Light and Electron Microscopy of Joint Cartilages

For light microscopy, joints were fixed in Bouin's solution, and then decalcified in 10% EDTA solution for 2 days. The joints were dehydrated with ethanol and embedded in paraffin. Sections through each joint were stained with hematoxylin-eosin and observed by light microscopy. Measurements were made of the width of articular cartilages from several joints.

For transmission electron microscopy, femoral epiphyses were dissected from the hind leg, and fixed in 2.5% glutaraldehyde and 1% osmium tetroxide. After dehydration by ethanol, epiphyses containing the growth plate were embedded in Spurr's resin. Ultra-thin sections through the growth plate were cut longitudinally and stained with uranyl acetate and lead citrate. Sections were observed on an H7000 electron microscope (Hitachi, Japan).

In Situ Hybridization and Immunohistochemistry

The following probes for several collagen genes were used for in situ hybridization: for Col5a1: 5A1-F: 5′-GGC TAC CAG AAG ACG GTG CT-3′, 5A1-R: 5′-AAT CCA TCG GAA AGG CAC GT-3′ (622 bp); for Col10a1: 10A1-F: 5′-GGC TTC ATA AAG GCA GGC CA-3′, 10A1-R: 5′-CAT GGG AGC CAC TAG GAA TC-3′ (469 bp); FOR Col1a1: 1A1-F: 5′-AGA AGG CCC TGC TCC TCC AG-3′, 1A1-R: 5′-GGC AGG GCC AAT GTC TAG TC-3′ (220 bp); for Col2a1: 2A1-F: 5′-ACA CAC TGG TAA GTG GGG CA-3′, 2A1-R: 5′-TGG GGC TGG GAA CAG TCA CT-3′ (435 bp); for Col11a1: 11A1-F: 5′-AGT CGG TCC AGC TTG CTT TC-3′, 11A1-R: 5′-TTA TTG CCA TCA CAG TCC AC-3′ (429 bp). In situ hybridization was processed essentially as described by McLaughlin and Margolskee (1993) with probes labeled with 33P. PCR products were subcloned into pBluescript SK-II (Stratagene). The plasmids containing the cDNA fragment for each collagen type were linearized with restriction digestion and then RNA probes were labeled by a Riboprobe Transcription System (Promega) in the presence of [33P]UTP (1,000 μCi/mmol). The labeled probes were used immediately for in situ hybridization.

Immunohistochemistry was performed as described by Mayne et al. (1994) for type II collagen and Johnson et al. (1999) for type X collagen. In brief, for type II collagen, cryosections were incubated with a monoclonal antibody against mouse type II collagen and detected by FITC-conjugated secondary antibody. For type X collagen, paraffin-embedded sections were deparaffinized and incubated with a polyclonal antibody against type X collagen (kindly provided by Dr. Bjorn Olsen, Harvard University) after digestion with hyaluronidase and then detected by FITC-conjugated secondary antibody.


The authors thank Dr. Jaspal S. Khillan for expert advice and criticisms, and Dr. Bjorn Olsen for the type X collagen antibody. The construct used in these experiments was prepared by Dr. Philipp Vandenberg. D.J.P. and R.M. received support from the National Institutes of Health.