COL6A3 Protein Deficiency in Mice Leads to Muscle and Tendon Defects Similar to Human Collagen VI Congenital Muscular Dystrophy*

Background: The collagen VI COL6A3 subunit may serve critical biological functions independent of the COL6A1 and COL6A2 subunits. Results: A Col6a3 mouse mutant displays muscle and tendon abnormalities resembling the Col6a1-null mouse. Conclusion: The primary role of the COL6A3 subunit is in assembling collagen VI microfibrils. Significance: The Col6a3 mutant mouse represents an animal model for collagen VI disorders. Collagen VI is a ubiquitously expressed extracellular microfibrillar protein. Its most common molecular form is composed of the α1(VI), α2(VI), and α3(VI) collagen α chains encoded by the COL6A1, COL6A2, and COL6A3 genes, respectively. Mutations in any of the three collagen VI genes cause congenital muscular dystrophy types Bethlem and Ullrich as well as intermediate phenotypes characterized by muscle weakness and connective tissue abnormalities. The α3(VI) collagen α chain has much larger N- and C-globular domains than the other two chains. Its most C-terminal domain can be cleaved off after assembly into microfibrils, and the cleavage product has been implicated in tumor angiogenesis and progression. Here we characterize a Col6a3 mutant mouse that expresses a very low level of a non-functional α3(VI) collagen chain. The mutant mice are deficient in extracellular collagen VI microfibrils and exhibit myopathic features, including decreased muscle mass and contractile force. Ultrastructurally abnormal collagen fibrils were observed in tendon, but not cornea, of the mutant mice, indicating a distinct tissue-specific effect of collagen VI on collagen I fibrillogenesis. Overall, the mice lacking normal α3(VI) collagen chains displayed mild musculoskeletal phenotypes similar to mice deficient in the α1(VI) collagen α chain, suggesting that the cleavage product of the α3(VI) collagen does not elicit essential functions in normal growth and development. The Col6a3 mouse mutant lacking functional α3(VI) collagen chains thus serves as an animal model for COL6A3-related muscular dystrophy.

␣2(VI), and ␣3(VI) ␣ chains, and these collagen VI molecules are assembled into microfibrils. The three ␣ chains each contain a central triple-helical domain of 335-336 amino acids flanked by N-and C-globular domains that are primarily made up of ϳ200-amino acid modules homologous to the von Willebrand factor type A (VWA) 2 domains (1-3). The ␣1(VI) and ␣2(VI) collagen chains are similar in size (140 kDa), having one VWA module in the N-globular domain and two such modules in the C-globular region. The ␣3(VI) collagen chain is 3 times their size due to substantially larger N-and C-globular domains. Its N-globular domain consists of up to 10 VWA modules, and the C-globular domain contains two VWA and three other protein modules (3,4). The three ␣ chains fold together into a triple-helical collagen VI molecule (monomer) that further assembles into dimers and tetramers (1,5). The tetramers are secreted into the extracellular space where they assemble end to end into collagen VI microfibrils (5). Several lines of evidence suggest that the C-terminal end of the ␣3(VI) collagen chain is cleaved off after microfibrillar assembly (6 -8).
The cleaved C-terminal domain may serve functions independent of collagen VI microfibrils. For instance, the C-terminal C5 domain of the ␣3(VI) collagen has been shown to interact with tumor endothelial marker 8 and thereby may play a role in tumor angiogenesis (9). Very recently, the C-terminal cleavage product of the ␣3(VI) collagen, newly named as "endotrophin," has been shown to promote malignant tumor progression in mice (10).
Mutations of the COL6A1, COL6A2, and COL6A3 genes encoding the ␣1(VI), ␣2(VI), and ␣3(VI) collagen chains underlie Bethlem myopathy, Ullrich congenital muscular dystrophy, and phenotypes intermediate between Ullrich congenital muscular dystrophy and Bethlem myopathy (14,15). The signature features of collagen VI-related muscle disorders are early onset muscle weakness in conjunction with connective tissue abnormalities. Ullrich congenital muscular dystrophy is a severe disease characterized by significant muscle weakness, joint contractures, and distal joint hypermobility resembling the Ehlers-Danlos syndrome. Independent ambulation may never be achieved or may be achieved but lost during the first two decades of life due to progressive muscle weakness and joint contractures. Bethlem myopathy is characterized by mild to moderate muscle weakness and multiple progressive joint contractures, mainly affecting the fingers, wrists, elbows, and ankles. It is a slowly progressive disorder and about a half of the Bethlem myopathy patients need aids for mobility after age 50 years (16).
Genotype-phenotype correlations of collagen VI-related muscle disorders are emerging but still incompletely understood. Notably, Allamand et al. (14) recently reported that most of the over 200 collagen VI mutations detected to date are distributed in the COL6A1 and COL6A2 genes. The finding seems to imply that the ␣3(VI) collagen ␣ chain, because of its extended N-and C-globular domains, may serve an indispensable function that is distinct from assembling collagen VI microfibrils and to suggest possible phenotypic differences associated with some mutations in COL6A3. To investigate whether the phenotype resulting from deficiency of the ␣3(VI) collagen differs from that of the other two subunits, we set out to generate a mouse mutant with the Col6a3 gene ablated by gene targeting. However, due to aberrant splicing and insertion of a neomycin gene, the targeted allele transcribed a low level of mutant Col6a3 mRNA that produced a non-functional protein.
We show that the homozygous mutant mice display myopathic and connective tissue symptoms similar to the Col6a1-null mice and patients with collagen VI muscle disorders. The results indicate that assembly of collagen VI microfibrils is the main function of the COL6A3 gene.

EXPERIMENTAL PROCEDURES
Construction of the Targeting Vector-A cosmid clone containing the mouse Col6a3 gene was isolated from a 129/Sv genomic library by screening with a mouse Col6a3 cDNA clone (17). A gene-targeting vector was prepared by replacing a 0.65-kb HindIII fragment spanning from exon 15 to intron 16 with a neomycin resistance gene driven by the phosphoglycerol kinase promoter (Pgk-Neo) (see Fig. 1A). The Pgk-Neo gene was inserted in the opposite transcription orientation of the Col6a3 gene. The targeting vector consisted of a 5.4-kb HindIII frag-ment as a long arm, a 2.5-kb HindIII fragment as a short arm, and a diphtheria toxin A gene for negative selection.
Generation of the Col6a3 Mutant Mouse-The targeting vector was linearized with NotI and electroporated into mouse 129/Sv embryonic stem (ES) cells. Cells were selected with 250 g/ml G418. Correctly targeted ES clones were identified by Southern blotting using a DNA probe located externally to the targeted region (see Fig. 1A). Targeted clones were injected into C57BL/6 blastocysts and implanted into pseudopregnant females to obtain chimeric mice. The chimeric founders were mated with C57BL/6 mice to generate heterozygous mice, which were then intercrossed to produce wild type, heterozygous, and homozygous mice. Initial phenotype analyses were performed with littermates on a mixed 129/Sv and C57BL/6 background. Mice were backcrossed onto the C57BL/6 background for eight generations and used for further analysis. All animal procedures were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals, and all experimental protocols were approved by the Institutional Animal Care and Use Committee of Thomas Jefferson University.
Genotyping of ES Cells and Mice-Genomic DNA isolated from ES cells or mouse tails was digested with BglI, and Southern blot hybridization was performed using a 0.7-kb [ 32 P]dCTP-labeled external probe (see Fig. 1A). The probe was generated by PCR amplification of the genomic clone using primers GAGAGGAGATGTCGGGATTCG and GTGAG-GCAGCAATGATGTAGAG. Genotyping of the mice was also carried out by PCR amplification of mouse tail DNA. A 521-bp PCR product from the wild type allele was detected using forward primer GCTGACCAATATCCCTCGTG and reverse primer CCAGACAGGCTACAAACTCCA. A product of ϳ200 bp was amplified from the mutant allele with the same reverse primer and primer GCCAGAGGCCACTTGTGTAG located in the Pgk-Neo gene.
Dermal Fibroblasts, Northern Blot, Western Blot, RT-PCR, and Collagen VI Deposition-Dermal fibroblasts were prepared from skin of 3-5-day-old littermates obtained from intercrossing heterozygous animals. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (Invitrogen). Northern blotting was performed with total RNA extracted from cells using the RNeasy Mini kit (Qiagen, Valencia, CA) and hybridized with [ 32 P]dCTP-labeled mouse Col6a1, Col6a2, and Col6a3 cDNA probes by the procedure described previously (18). RT-PCR analysis of the Col6a3 mRNA surrounding the gene-targeted region was performed with total RNA isolated from fibroblasts using forward primer TCTT-GAACGTGTGGCTAACC and reverse primer CTTTTCTC-CAGAAGAACCAGG. For the analysis of collagen VI deposition, fibroblasts were grown in DMEM with 10% FBS in the presence of 50 mg/ml ascorbate phosphate until 5 days after confluence, fixed with 3.7% formaldehyde, and incubated with polyclonal antibodies against ␣1(VI) or ␣3(VI) collagens (19,20) for 1 h. Immunostaining was detected with Cy3-conjugated goat anti-rabbit IgG (1:1000 dilution; Jackson Immuno-Research Laboratories, West Grove, PA) and observed using a Zeiss Axioskop epifluorescence microscope with a Toshiba 3CCD camera and ImagePro software (Media Cybernetics, Silver Spring, MD).
RT-PCR Analysis of Markers for ER Stress, Inflammation, and Apoptosis-Total RNA was isolated from quadriceps muscles and dermal fibroblasts using the RNeasy Mini kit (Qiagen). Muscle tissues were dissected from Col6a hm/hm and Col6a3 ϩ/ϩ littermates at three different ages (5 days, 1 month, and 3 months). Dermal fibroblasts were established from three newborn litters of the Col6a hm/hm and Col6a3 ϩ/ϩ mice. The RNA preparation was treated with DNase I to remove any contaminating genomic DNA. Single-stranded cDNA was synthesized from 0.5 g of total RNA using RNA to cDNA EcoDry Premix (Clontech) at 42°C for 1 h. Quantitative real time RT-PCR was performed using SYBR Green PCR Master Mix (Invitrogen/ Applied Biosystems (ABI)) on an ABI StepOne Real-Time PCR system. The reaction was initiated at 95°C for 10 min (to activate the AmpliTaq polymerase) followed by a 40-cycle amplification phase (denaturation at 95°C for 15 s and annealing/ extension at 60°C for 60 s). A melting curve analysis was performed for each gene to ensure a single PCR product was being amplified with no primer-dimer formation. The amount of PCR product was estimated using a relative standard curve quantification method. The mRNA expression levels of target genes were normalized by hypoxanthine phosphoribosyltransferase 1 (Hprt1) mRNA. All the primers used were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA). Statistical analysis was performed using Sigma Plot 11.2 statistical software (Systat Software, Inc.). Primer sequences are shown in Table 1.
Histology and Immunostaining of Tissues-Muscles were dissected, mounted on 10% gum tragacanth, and frozen in liquid nitrogen-cooled isopentane. Sections of 8 m thick were stained with hematoxylin-eosin. Single fiber areas and the percentage of fibers with central nuclei were determined from images of four different gastrocnemius muscles per group. Approximately 100 fibers from each image taken at 20ϫ magnification were outlined and measured using ImageJ software. For immunostaining, 6 -8-m cryosections were fixed in methanol and incubated with the primary antibodies for 2 h at room temperature. The primary antibodies used were polyclonal antibodies specific for the ␣1(VI) collagen (20), ␣3(VI) collagen (19), decorin (LF-113; from Dr. Larry Fisher), tenascin-X (M11; from Dr. Ken-Ichi Matsumoto), NG2 (from Dr. William Stallcup), integrin ␣7B (U31ϩ; from Dr. Ulrike Mayer), matrilin-2 and -4 (from Dr. Mats Paulsson), and type I collagen (Fitzgerald Industries International, Acton, MA) and a monoclonal antibody for laminin ␥1 (Chemicon MAB1914).
Physiological Measurements-Ex vivo physiological measurements were performed on dissected extensor digitorum longus (EDL) muscle from 4-month-old male mice as described (21,22). Muscles were dissected, weighed, and then placed in a custom-built organ bath in oxygenated Ringer's solution (pH 7.4) at 24°C for the duration of the experiment. The muscle was attached using 5-0 nylon to a Grass FT03 force transducer (Astro-Med, West Warwick, RI) at one end and length controller (precision, 2.5 m) containing a 1.8°step motor-driven micrometer at the other end. The muscle was stimulated at 25 V using a Grass S48 stimulator to develop an electrical field between two platinum electrodes. Muscle length was adjusted to achieve maximal twitch and tetanic response, and optimal length (Lo) was measured. Muscles were subject to a series of three twitches and three tetanic contractions as well as five eccentric contractions (ECCs) for mechanical property evaluation. Muscles were stimulated using a supramaximal stimulus of 700 ms (500-ms isometric phase and 200-ms eccentric phase) with total lengthening of Lo/10 at the lengthening velocity of 0.5Lo/s. The time between two ECCs was 5 min. ECC force drop was calculated by using the difference of isometric phase of the first and fifth ECCs. Data were digitized and acquired by using a PowerLab/8SP A/D converter and software (AD Instruments, Colorado Springs, CO). Cross-sectional area was determined using the Brooks-Faulkner approximation (23).
Transmission Electron Microscopy-Flexor digitorum longus tendons and corneas were dissected from postnatal day 30 homozygous mutant mice and wild type littermates. Transmission electron microscopy was performed as described previously (24,25).
Statistical Analysis-Data were analyzed using a paired Student's t test and one-way analysis of variance. p Ͻ 0.05 is considered statistically significant. All data are expressed as mean Ϯ S.E. unless otherwise noted.

RESULTS
Generation of the Col6a3 Mouse Mutant-A gene-targeting vector was designed to disrupt the Col6a3 gene by replacing the genomic region between the HindIII sites in exon 15 and intron 16 with the Pgk-Neo cassette (Fig. 1A). Exon 15 encodes a cysteine-rich linker region and the beginning of the triple-helical domain. Insertion of the Pgk-Neo cassette within exon 15 was predicted to yield a premature translational termination codon after the N-globular domain-coding region. Electroporation of the construct into mouse ES cells resulted in ϳ350 neomycinresistant clones of which eight were correctly targeted clones based on Southern blot analysis. Three of the positive clones were injected into mouse blastocysts, and 14 chimeric mice were obtained. The targeted allele was transmitted through the germ line to the next generation, and intercrossing of the heterozygous F1 mice from each ES cell clone generated offspring of the three genotypes at the expected Mendelian frequencies (Fig. 1B). The homozygous Col6a3 mutant mice developed normally, were fertile, and did not show overt phenotypic abnormalities. They have normal life spans and have survived beyond 2 years of age.
Unexpectedly, Northern blot analysis revealed that the homozygous mutant synthesized a very low level of a slightly smaller ␣3(VI) collagen mRNA (Fig. 1C). RT-PCR analysis using primers flanking the gene-targeted region confirmed that the ␣3(VI) collagen mRNA produced by the homozygous mouse was indeed smaller than the normal Col6a3 mRNA (Fig.  1D). Analysis of the heterozygous mouse showed that the mutant transcript was present at a substantially lower level compared with the normal Col6a3 mRNA (Fig. 1D). Quantitative RT-PCR analysis indicated that the mutant mRNA was expressed at a level ϳ5% of the wild type mRNA (data not shown). DNA sequencing identified an in-frame deletion of 147 bp in the mutant mRNA that corresponded to the sequences encoded by exons 15 and 16. The results indicated that the splice acceptor site in exon 15 was not utilized and thereby led to the deletion of exons 15 and 16 (Fig. 1E). Thus, contrary to our prediction, the gene targeting event generated a hypomorphic mutation in Col6a3. The mutant allele is designated Col6a3 hm . The resultant mutant ␣3(VI) collagen chain contained an in-frame deletion encompassing a cysteine-rich linker region and the beginning of the triple-helical domain. As the cysteine-rich sequence is critical for stabilizing the triplehelical collagen VI molecule, its deletion would prevent the mutant chain from assembling into a stable collagen VI triplehelical monomer.
Deficiency of Collagen VI in Col6a3 hm/hm Mice-Collagen VI expression was examined in the wild type (Col6a3 ϩ/ϩ ), heterozygous (Col6a3 ϩ/hm ), and homozygous (Col6a3 hm/hm ) littermates. Immunostaining of gastrocnemius muscles with the ␣3(VI) collagen-specific antibody revealed that this chain was barely detectable in Col6a3 hm/hm and was substantially reduced in Col6a3 ϩ/hm (Fig. 2A). The ␣1(VI) collagen expression was similarly reduced in the Col6a3 ϩ/hm muscle, but a low level could be detected in the Col6a3 hm/hm muscle ( Fig. 2A). Dermal fibroblasts were established from littermates of all three genotypes and analyzed for the deposition of collagen VI microfibrils. Immunostaining with the ␣3(VI) collagen-specific antibody showed that extracellular collagen VI microfibrils were abundant in the wild type fibroblasts, significantly reduced in Col6a3 ϩ/hm , and barely detectable in the Col6a3 hm/hm cells (Fig. 2B). The decreases in extracellular col-FIGURE 2. Collagen VI deficiency in skeletal muscle and dermal fibroblasts of the Col6a3 mutant mice. A, cryosections of quadriceps muscles from 1month-old Col6a3 ϩ/ϩ , Col6a3 ϩ/hm , and Col6a3 hm/hm littermates immunostained with polyclonal antibodies specific for the ␣3(VI) and ␣1(VI) collagens. B, immunostaining of collagen VI deposited by fibroblasts from the Col6a3 ϩ/ϩ , Col6a3 ϩ/hm , and Col6a3 hm/hm littermates. Cells were grown in the presence of 50 g/ml ascorbate phosphate for 4 days postconfluence and then stained with polyclonal antibodies specific for the ␣3(VI) and ␣1(VI) collagens. Note the intracellular accumulation of the ␣1(VI) collagen in the Col6a3 ϩ/hm and Col6a3 hm/hm fibroblasts. Magnification bars, 50 m. C, Western blot analysis of culture medium and cell extracts from dermal fibroblasts of Col6a3 hm/hm and Col6a3 ϩ/ϩ mice using antibodies specific for the ␣3(VI) collagen (left panels) and ␣1(VI) collagen (right panels). Very low levels of the ␣3(VI) collagen were found in both culture medium and cell extracts of the Col6a3 hm/hm fibroblasts. The ␣1(VI) collagen was barely detectable in the culture medium the Col6a3 hm/hm fibroblasts, but its amount was significantly higher in the cell extract of the Col6a3 hm/hm fibroblasts than in the Col6a3 ϩ/ϩ counterpart.
lagen VI microfibrils were similarly observed when the ␣1(VI) collagen-specific antibody was used except that substantial intracellular reactivity was found in the Col6a3 ϩ/hm and Col6a3 hm/hm fibroblasts. Western blot detected much lower levels of the ␣3(VI) collagen in the culture medium and cell extract of dermal fibroblasts from the Col6a3 hm/hm mice compared with the wild type counterparts (Fig. 2C). The ␣1(VI) collagen level was also very low in the culture medium of the Col6a3 hm/hm fibroblasts but was much higher in the cell extract of the Col6a3 hm/hm fibroblasts compared with the wild type cells. Together, the results demonstrate that the Col6a3 hm/hm mice are deficient in collagen VI protein in the extracellular matrix and display intracellular retention of the ␣1(VI) collagen.
To determine whether intracellular accumulation of collagen VI chains in the Col6a3 hm/hm mice would induce ER stress and the unfolded protein response, the expression of selected molecular markers for the ER stress pathway (26) was examined using RNA isolated from quadriceps muscles and dermal fibroblasts. By RT-PCR, splicing of X-box-binding protein 1 (Xbp-1) mRNA induced by ER stress was not detected in skeletal muscles and dermal fibroblasts of Col6a3 hm/hm mice (Fig. 3A). Quantitative RT-PCR analysis did not show a significant difference between the Col6a3 hm/hm and Col6a3 ϩ/ϩ mice in the expression of mRNAs for two additional ER stress markers, chaperone-binding protein (Bip) and CCAAT/enhancer-binding protein homologous protein (Chop) (Fig. 3B). In addition, no significant difference was found in mRNA expression of FIGURE 3. RT-PCR analysis of markers for ER stress, inflammation, and apoptosis in Col6a3 hm/hm and Col6a3 ؉/؉ mice. A, RT-PCR analysis of total RNA from quadriceps muscles of Col6a3 ϩ/ϩ (wt) and Col6a3 hm/hm (hm) mice at ages 5 days (lanes 1 and 2), 1 month (lanes 3 and 4), and 3 months (lanes 5 and 6) and total RNA from dermal fibroblasts of two separate litters of Col6a3 ϩ/ϩ (wt) and Col6a3 hm/hm (hm) mice. PCR products were run on a 2% agarose gel. Unspliced Xbp-1 mRNA (uXbp-1; 170 bp) was detected in all samples. Spliced Xbp-1 mRNA (sXbp-1; 144 bp), which is induced during ER stress, was absent. B, C, and D, real time RT-PCR analysis of ER stress markers Bip and Chop; inflammation markers Il-1␤, Il-6, and Tnf-␣, and apoptosis markers Casp-3, Casp-9, and Bcl-2. Muscle analyses include total RNA isolated from quadriceps muscles of 5-day-, 1-month-, and 3-month-old Col6a3 ϩ/ϩ (wt) and Col6a3 hm/hm (hm) mice. Fibroblast analyses include total RNA prepared from dermal fibroblasts of three pairs of newborn Col6a3 ϩ/ϩ (wt) and Col6a3 hm/hm (hm) littermates. PCRs were done in duplicate for each RNA samples, and mRNA expression of each gene was normalized to that of the Hprt1 gene. Values of the Col6a3 hm/hm (hm) are expressed as -fold over Col6a3 ϩ/ϩ (wt) and mean Ϯ S.D. (error bars). *, p Ͻ 0.05 between Col6a3 hm/hm and Col6a3 ϩ/ϩ ; ns, not significant. molecular markers of inflammation (Il-1␤, Il-6, and Tnf-␣) and apoptosis (Casp-3, Casp-9, and Bcl-2) between the Col6a3 hm/hm and Col6a3 ϩ/ϩ mice (Fig. 3, C and D).
Myopathy and Compromised Muscle Function in Col6a3 hm/hm Mice-Analysis of adult mice at 4 months of age showed that the body and organ weights of the Col6a3 hm/hm mice were somewhat reduced compared with the Col6a3 ϩ/ϩ controls, although a statistical significance was not reached ( Table 2). Weights of EDL, soleus, and tibialis anterior muscles in the mutant mice were significantly lower than the corresponding muscles from the wild type controls (p Ͻ 0.05). Gastrocnemius, quadriceps, and diaphragm muscles in the mutant mice also showed decreases in weight, but statistical significance was not reached. The difference in the body weight between the Col6a3 hm/hm mice and wild type mice was more pronounced with age. The body weights of 1.5-2.0-year-old Col6a3 hm/hm and their wild type littermates were 23.9 Ϯ 3.9 and 32 Ϯ 2.4 g, respectively (n ϭ 4 for each group; p Ͻ 0.05). No gross skeletal abnormalities were seen in aged mutant mice.
Mild myopathic symptoms could be detected by histology. A wide variation in fiber sizes with many small sized fibers and an increase in the interstitial connective tissue between muscle fibers were observed in gastrocnemius, quadriceps, and diaphragm muscles of the Col6a3 hm/hm mice compared with the Col6a3 ϩ/ϩ controls (Fig. 4A). Also evident was the presence of regenerative fibers in the Col6a3 hm/hm mice, indicated by centrally localized nuclei, in contrast to the wild type controls (Fig.  4A). The percentage of the central nucleation and distribution of single fiber areas in gastrocnemius muscle sections of the two genotypes were quantified (Fig. 4, B and C). The gastrocnemius muscle of 6-month-old Col6a3 hm/hm mice showed increased immunoreactivity with an antibody against collagen type I (Fig.  4, D and E), suggesting accelerated collagen I deposition and fibrosis progression.
To evaluate muscle function, physiological measurements were performed using dissected EDL muscles. The analyses showed that the EDL muscles from mutant mice were weaker than those of the wild type controls. Absolute and specific forces of both twitch and tetanic contractions were significantly reduced in Col6a3 hm/hm compared with the wild type mice ( Table 3). The decline in absolute force was more pronounced compared with the change in specific force, reflecting the fact that the EDL in Col6a3 hm/hm was smaller. Notably, there was little difference in ECC force drop generated by the first and fifth tetani under eccentric conditions in Col6a3 hm/hm when compared with the wild type animals.

Expression of Matrix Proteins in Col6a3 Mutant
Muscle-To determine whether there are changes in the expression of selected extracellular matrix proteins and receptors that are known to or likely interact with collagen VI, quadriceps muscles from Col6a3 ϩ/ϩ , Col6a3 ϩ/hm , and Col6a3 hm/hm littermates were analyzed by immunostaining (Fig. 5). Laminin ␤1 was expressed in the muscle basement membranes and was comparable among the three genotypes. The basement membrane staining pattern demonstrated wide variations in muscle diameters in the Col6a3 ϩ/hm and Col6a3 hm/hm mice compared with the wild type animal. The expression of tenascin-X was increased in the Col6a3 ϩ/hm and Col6a3 hm/hm mice compared with the wild type mice (total fluorescence measurements of four non-overlapping images were 2.3 Ϯ 0.5-fold in Col6a3 hm/hm and 1.9 Ϯ 0.6-fold in Col6a3 ϩ/hm versus Col6a3 ϩ/ϩ ). However, there was no significant difference among the three genotypes in the expression of other matrix proteins and receptors examined, including perlecan, decorin, biglycan, matrilin-2 and -4, integrin ␣7B, and NG2 receptor (data not shown).
Abnormal Collagen Fibril Structure in Tendon, but Not Cornea, of the Col6a3 hm/hm Mice-Collagen VI is a major component of both tendon and cornea; therefore, to determine the role in non-muscle tissues, an analysis of matrix structure was performed using transmission electron microscopy and 1-month-old Col6a3 hm/hm and Col6a3 ϩ/ϩ littermates. In flexor digitorum longus tendons of the wild type mice, a mixture of large and small collagen fibrils was seen (Fig. 6A, panels a, b, and  c). In contrast, the collagen fibrils in the Col6a3 hm/hm tendons exhibited a substantial increase in the number of small diameter fibrils throughout the tendon (Fig. 6A, panels d, e, and f). In addition, aberrant fibrils with highly irregular profiles, including irregular fibril contours, were observed. There was a regional difference in that the number of small diameter fibrils, aberrant fibrils, and abnormally large diameter fibrils were concentrated in the pericellular region close to the tendon cells (Fig. 6A, panel e) and less abundant in areas distant from the cells (Fig. 6A, panel f). Contrary to the observations in tendons, the collagen fibril structure in the cornea of 1-month-old Col6a3 hm/hm was comparable with that of the wild type littermates (Fig. 6B).

DISCUSSION
We have generated and characterized a mutant mouse deficient in the normal ␣3(VI) collagen chain. First, our gene targeting approach yielded a mouse model with significantly reduced expression of the mutant Col6a3 gene. The gene-targeting vector was designed to cause a translational frameshift in exon 15 of the Col6a3 gene and thereby prevent the synthesis of the ␣3(VI) collagen chain. However, the insertion of the Pgk-Neo gene in a small exon of 93 bp inactivated the 5Ј splice acceptor site of exon 15, resulting in splicing from exon 14 to exon 17 (Fig. 1E). Analysis of the heterozygous mutant mouse indicates a substantial reduction in the expression of the mutant allele compared with the wild type counterpart (Fig.  1D). The reduced expression observed is consistent with previ-

Col6a3 Mutant Mouse Displays Muscle and Tendon Defects
ous findings that insertion of a selectable marker in a gene suppresses its expression (27,28). Second, in the mutant allele, skipping of exons 15 and 16 results in an in-frame deletion of 49 amino acids in the mutant ␣3(VI) collagen chain. The mutant ␣3(VI) collagen chain likely can be assembled into a triple-helical collagen VI molecule as biochemical and genetic evidence suggests that collagen VI chain assembly proceeds in the direction of C to N terminus (29,30). However, the assembled triplehelical collagen VI molecule would be unstable because the mutant ␣3(VI) collagen chain lacks the cysteine-rich linker seg-ment that stabilizes the triple-helical monomer. This prevents the formation of collagen VI microfibrils extracellularly. Thus, the Col6a3 hm allele should be equivalent to a functional null allele. This interpretation is supported by the observation that Col6a3 hm/hm fibroblasts assemble few collagen VI microfibrils, and the Col6a3 ϩ/hm cells deposit approximately half the amount of collagen VI microfibrils compared with the wild type fibroblasts (Fig. 2B).
Intracellular accumulation of the ␣1(VI) collagen chain was found in dermal fibroblasts from both Col6a3 ϩ/hm and FIGURE 4. Analysis of limb and diaphragm muscles of the Col6a3 mouse mutant. A, cryosections of gastrocnemius (Gas), quadriceps (Qua), and diaphragm (Dia) muscles of Col6a3 ϩ/ϩ and Col6a3 hm/hm mice were stained with hematoxylin and eosin. Centrally localized nuclei (arrows) are abundant in the Col6a3 hm/hm muscles but barely detectable in the Col6a3 ϩ/ϩ controls. An increase in the interstitial connective tissue between muscle fibers is evident in the Col6a3 hm/hm mice compared with the Col6a3 ϩ/ϩ control. Magnification bars, 50 m. B, quantification of the percentage of gastrocnemius muscles with centrally located nuclei in Col6a3 ϩ/ϩ and Col6a3 hm/hm mice. C, distribution of single fiber areas in gastrocnemius muscles from Col6a3 ϩ/ϩ and Col6a3 hm/hm mice. The Col6a3 hm/hm muscles show a wide variation in fiber size compared with the Col6a3 ϩ/ϩ control. Data in B and C were from four muscles per group and are expressed as mean Ϯ S.E. (error bars). D, increased immunoreactivity of type I collagen in Col6a3 hm/hm muscle. Cryosections of quadriceps muscles from 6-month-old Col6a3 ϩ/ϩ and Col6a3 hm/hm mice were immunostained with affinity-purified antibody against type I collagen. A representative image is shown. E, quantification of increased immunoreactivity with type I collagen in Col6a3 hm/hm quadriceps muscles. The percent area positive for type I collagen in four muscles per group was analyzed using ImageJ software. Values are expressed as -fold change over Col6a3 ϩ/ϩ and as mean Ϯ S.E. (error bars). MAY 17, 2013 • VOLUME 288 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 14327
Col6a3 hm/hm mice (Fig. 2B). This indicates that synthesis of the individual collagen VI chain is independently regulated. In the absence of a functional ␣3(VI) collagen chain, the excess ␣1(VI) chain is accumulated in the cytoplasm, which also may contribute to the abnormal phenotype. However, the molecular markers of the ER stress pathway were not increased in the Col6a3 hm/hm muscles and dermal fibroblasts (Fig. 3, A and B). Our result is consistent with the finding in patients with collagen VI mutations that intracellular retention of collagen VI ␣ chains does not increase ER stress markers (31).
Analysis of the Col6a3 hm/hm mouse mutant demonstrates that a deficiency in the normal ␣3(VI) collagen chain results in a relatively mild gross and muscle phenotype similar to that of the previously reported Col6a1-null mice (32). Both mouse models have normal life spans unlike patients with severe deficiency of collagen VI who have a high risk of early death. Similar to the Col6a1-null mice, the Col6a3 hm/hm mice exhibit mild myopathic pathology and compromised muscle contractile forces ( Fig. 4 and Table 3) (32,33). Although the Col6a1-null muscles show spontaneous apoptosis (33), the Col6a3 hm/hm muscles do not exhibit a significant increase in the expression of molecular makers for apoptosis compared with wild type controls (Fig. 3D). It should be noted that mice from ages 5 days to 3 months were used in our analysis to mimic the childhood onset of collagen VI muscular dystrophy. The lack of a significant increase in markers for apoptosis in the Col6a3 hm/hm mice could be because the degree of apoptosis in young animals is usually low. The Col6a3 hm/hm muscles do not display significantly increased inflammation (Fig. 3C), and they show a slight increase in fibrosis (Fig. 3, D and E). This is likely due to the mild myopathic phenotype of mice deficient in collagen VI. Whether the Col6a3 hm/hm muscles display mitochondrial dysfunction and defective autophagy as demonstrated in the Col6a1-null mice (33,34) awaits a comprehensive analysis of the Col6a3 hm/hm mice in the future.
Previous studies suggest that the C-terminal C5 domain of the ␣3(VI) collagen is proteolytically processed after assembling into microfibrils and that the cleaved C5 domain can be found in tumor endothelium where it interacts with tumor endothelial marker 8 (6,8,9). Recently, the C-terminal cleavage product produced by the adipose tissue has been shown to pro- FIGURE 5. Immunofluorescence staining of skeletal muscles with antibodies for selected extracellular matrix proteins. Cryosections of quadriceps muscles from 1-month-old Col6a3 ϩ/ϩ , Col6a3 ϩ/hm , and Col6a3 hm/hm mice were stained with polyclonal antibodies against ␣3(VI) collagen (Col6a3), laminin ␤1 (Lamb1), and tenascin-X (Tnx). The expression of tenascin-X was increased in the Col6a3 ϩ/hm and Col6a3 hm/hm mice. Magnification bar, 100 m.

Col6a3 Mutant Mouse Displays Muscle and Tendon Defects
mote cancer progression in mouse models of breast cancer (10).
The lack of a severe phenotype in the ␣3(VI) collagen mutant mouse implies that the large N-and C-globular domains of the ␣3(VI) ␣ chain do not serve indispensable functions independent of collagen VI microfibril assembly during normal growth and development. Whether the marked reduction of the C5 domain affects tumor formation or progression in the homozygous mouse remains to be investigated.
Physiological measurements of muscle function demonstrate that there are significant differences in both absolute and specific force generated by a single twitch or tetanic stimulation (Table 3). This is consistent with the muscle weakness in mutant animals. In both twitch and tetanic stimulation, the force decrease was more remarkable and significant for the absolute force, whereas the drop in specific force was less pronounced. This difference is explained by the reduced EDL muscle weight noted in the mutant animals ( Table 2) so that the decrease in specific force is amplified when calculated for the whole muscle. Interestingly, we did not record any significant drop in ECC comparing the first and fifth eccentric tetanic contractions in the Col6a3 mouse mutant (Table 3). ECC force drop is a physiological hallmark of the mdx mouse model of Duchenne muscular dystrophy due to a deficiency in dystrophin (35,36). This is thought to be due to the inherent instability of the plasma membrane and lack of costameric stability resulting from the lack of dystrophin in the mdx mouse model and in patients with Duchenne muscular dystrophy, leading to active degeneration and regeneration in the muscle. Severe collagen VI deficiency causing the clinical phenotype of Ullrich congenital muscular dystrophy is a much less dystrophic myopathy at least early in the disease course (37), indicating that mechanical fragility of the plasma membrane is not as important in this muscle disease compared with the conditions caused by deficiencies of dystrophin or of the dystrophin-associated glycoprotein complex. In keeping with this hypothesis is that serum creatine kinase levels in patients with collagen VI deficiency are frequently normal or only mildly elevated. Thus, the mechanisms leading to muscle disease in collagen VI deficiency are different from those of the classic muscular dystrophies.
Although a decrease in laminin ␤1 expression was previously found in patients with collagen VI congenital muscular dystrophy (38), we found no difference in laminin ␤1 reactivity in collagen VI-deficient mice compared with the wild type controls (Fig. 5). Our results show that tenascin-X protein expression is increased in muscle deficient in collagen VI (Fig. 5). In humans, the absence of tenascin-X leads to a recessive form of Ehlers-Danlos syndrome characterized by joint hypermobility and skin hyperextensibility (39). The connective tissue abnormalities partially overlap with those of collagen VI muscular dystrophy (40,41). Additionally, a mild myopathic phenotype has been described recently in mice lacking tenascin-X (42). These findings suggest that there may be a functional overlap between collagen VI and tenascin-X so that there is a compensatory increase in tenascin-X in the endomysium of collagen VI-deficient muscle. To this end, both collagen VI and tenascin-X are able to modulate collagen I fibrillogenesis in vitro, and deficiency in either protein leads to abnormal collagen fibrils (43,44). On the other hand, contrary to expectation based on the compensatory hypothesis, collagen VI mRNA and protein expression is decreased in fibroblasts from tenascin-X-null mice (43).
Our ultrastructural analysis of the Col6a3 hm/hm tendon is consistent with the finding in Col6a1-null mice indicating an association between deficiency in collagen VI and abnormal tendon collagen fibril assembly/structure (25). Collagen VI is a  MAY 17, 2013 • VOLUME 288 • NUMBER 20 key component in the pericellular matrix of tendons and ligaments (45). Our studies demonstrate that the absence of collagen VI alters organization of the tendon fibroblasts and the pericellular collagen fibril-forming microdomains defined by the fibroblast surface. The precise molecular mechanisms by which collagen VI modulates collagen I fibril assembly are not known but are likely to be indirect, mediated through changes in the expression of molecules that regulate collagen fibrillogenesis by tendon fibroblasts. For instance, a consistent small increase in MMP-2 activity has been found in the Col6a1-null tendon (25). In addition, collagen VI binds a number of extracellular molecules such as small leucine-rich repeat proteoglycans and fibril-associated collagens with interrupted triple helices that have been implicated in regulation of the later stages in fibril assembly (46). It is possible that lack of collagen VI provides additional interactions with the collagen fibrils, thus altering fibril structure in the mature tendon. This would be consistent with the maintenance of small diameter fibrils observed in the mutant tendons. Collagen VI is also a major component of cornea and is closely associated with cornea fibroblasts (47). However, we found that collagen fibrillogenesis in the cornea is not affected by the absence of collagen VI. Collagen fibrils in the cornea are homogenously small diameter as they do not undergo lateral growth during fibril maturation unlike collagen fibrils in most other connective tissues (48). In addition, higher concentrations of the regulatory molecules are found in the corneal stroma compared with the tendon so that the absence of interactions with collagen VI would not be as significant as in the tendon. Our demonstration of a tendinopathy in the Col6a3 hm/hm mouse has direct implications for the human collagen VI-related myopathies as the joint hyperlaxity and more importantly the striking progressive joint contractures are a significant contributor to the morbidity and disease progression. Because the contractures typically are out of proportion to the muscle weakness, the tendon abnormalities identified in our mouse model likely contribute significantly to the disease in patients.

Col6a3 Mutant Mouse Displays Muscle and Tendon Defects
In conclusion, our studies demonstrated that mice lacking a functional ␣3(VI) collagen ␣ chain display muscle and tendon features reminiscent of human patients albeit with significantly milder phenotypic manifestation. Thus, the mouse mutant represents a second model for recessive collagen VI muscular dystrophy and can be utilized to further understand the consequence of COL6A3 dysfunction.