Title: Fibromodulin-null Mice Have Abnormal Collagen Fibrils, Tissue Organization, and Altered Lumican Deposition in Tendon
Abstract: Fibromodulin is a member of a family of connective tissue glycoproteins/proteoglycans containing leucine-rich repeat motifs. Several members of this gene family bind to fibrillar collagens and are believed to function in the assembly of the collagen network in connective tissues. Here we show that mice lacking a functional fibromodulin gene exhibit an altered morphological phenotype in tail tendon with fewer and abnormal collagen fiber bundles. In fibromodulin-null animals virtually all collagen fiber bundles are disorganized and have an abnormal morphology. Also 10–20% of the bundles in heterozygous mice are similar to the abnormal bundles in fibromodulin-null tail tendon. Ultrastructural analysis of Achilles tendon from fibromodulin-null mice show collagen fibrils with irregular and rough outlines in cross-section. Morphometric analysis show that fibromodulin-null mice have on the average thinner fibrils than wild type animals as a result of a larger preponderance of very thin fibrils in an overall similar range of fibril diameters. Protein and RNA analyses show an approximately 4-fold increase in the content of lumican in fibromodulin-null as compared with wild type tail tendon, despite a decrease in lumican mRNA. These results demonstrate a role for fibromodulin in collagen fibrillogenesis and suggest that the orchestrated action of several leucine-rich repeat glycoproteins/proteoglycans influence the architecture of collagen matrices. Fibromodulin is a member of a family of connective tissue glycoproteins/proteoglycans containing leucine-rich repeat motifs. Several members of this gene family bind to fibrillar collagens and are believed to function in the assembly of the collagen network in connective tissues. Here we show that mice lacking a functional fibromodulin gene exhibit an altered morphological phenotype in tail tendon with fewer and abnormal collagen fiber bundles. In fibromodulin-null animals virtually all collagen fiber bundles are disorganized and have an abnormal morphology. Also 10–20% of the bundles in heterozygous mice are similar to the abnormal bundles in fibromodulin-null tail tendon. Ultrastructural analysis of Achilles tendon from fibromodulin-null mice show collagen fibrils with irregular and rough outlines in cross-section. Morphometric analysis show that fibromodulin-null mice have on the average thinner fibrils than wild type animals as a result of a larger preponderance of very thin fibrils in an overall similar range of fibril diameters. Protein and RNA analyses show an approximately 4-fold increase in the content of lumican in fibromodulin-null as compared with wild type tail tendon, despite a decrease in lumican mRNA. These results demonstrate a role for fibromodulin in collagen fibrillogenesis and suggest that the orchestrated action of several leucine-rich repeat glycoproteins/proteoglycans influence the architecture of collagen matrices. leucine-rich repeat base pair(s) enzyme-linked immunosorbent assays Fibromodulin belongs to a family of extracellular matrix glycoproteins/proteoglycans sharing a leucine-rich repeat (LRR)1 structural motif (1Iozzo R.V. Crit. Rev. Biochem. Mol. Biol. 1997; 32: 141-174Crossref PubMed Scopus (460) Google Scholar). The gene family includes about 10 known members and can be divided into subfamilies based on similarities in amino acid sequences and gene organization. One subfamily includes decorin (2Krusius T. Ruoslahti E. Proc. Natl. Acad. Sci. U. S. A. 1986; 83: 7683-7687Crossref PubMed Scopus (464) Google Scholar) and biglycan (3Fisher L.W. Termine J.D. Young M.F. J. Biol. Chem. 1989; 264: 4571-4576Abstract Full Text PDF PubMed Google Scholar), which contains an N-terminal domain substituted with chondroitin/dermatan-sulfate chains. These proteoglycans show 57% protein sequence identity and are encoded by genes composed of eight exons with exon/intron junctions in conserved positions (4Fisher L.W. Heegaard A.-M. Vetter U. Vogel W. Just W. Termine J.D. Young M.F. J. Biol. Chem. 1991; 266: 14371-14377Abstract Full Text PDF PubMed Google Scholar, 5Danielson K.G. Fazzio A. Cohen I. Cannizzaro L.A. Eichstetter I. Iozzo R.V. Genomics. 1993; 15: 146-160Crossref PubMed Scopus (98) Google Scholar). Fibromodulin (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar) and lumican (7Funderburgh J.L. Funderburgh M.L. Brown S.J. Vergnes J.-P. Hassel J.R. Mann M.M. Conrad G.W. J. Biol. Chem. 1993; 268: 11874-11880Abstract Full Text PDF PubMed Google Scholar) constitute another subfamily and exhibit 48% protein sequence identity. Their genes are composed of three exons with conserved exon/intron junctions (8Antonsson P. Heinegård D. Oldberg Å. Biochim. Biophys. Acta. 1993; 1174: 204-206Crossref PubMed Scopus (49) Google Scholar, 9Grover J. Chen X.-N. Korenberg J.R. Roughley P.J. J. Biol. Chem. 1995; 270: 21942-21949Crossref PubMed Scopus (170) Google Scholar). Other members of this subfamily are keratocan (10Corpuz L.M. Funderburgh J.L. Funderburgh M.L. Bottomley G.S. Prakash S. Conrad G.W. J. Biol. Chem. 1996; 271: 9759-9763Abstract Full Text Full Text PDF PubMed Scopus (199) Google Scholar), PRELP (11Bengtsson E. Neame P.J. Heinegård D. Sommarin Y. J. Biol. Chem. 1995; 270: 25639-25644Abstract Full Text Full Text PDF PubMed Scopus (114) Google Scholar), and osteoadherin (12Wendel M. Sommarin Y. Heinegård D. J. Cell Biol. 1998; 141: 839-847Crossref PubMed Scopus (118) Google Scholar). Chondroadherin represents its own family with a different organization of the gene and a different amino acid composition (13Neame P.J. Sommarin Y. Boynton R.E. Heinegård D. J. Biol. Chem. 1994; 269: 21547-21554Abstract Full Text PDF PubMed Google Scholar). The LRR extracellular matrix glycoproteins/proteoglycans have core proteins ranging in size between 32–42 kDa, which can be divided into three main structural domains, the N-terminal, C-terminal, and central domains. The N-terminal domains are least conserved, but all members of the gene family contain four Cys residues, which form intrachain disulfide bonds (14Neame P.J. Choi H.U. Rosenberg L.C. J. Biol. Chem. 1989; 264: 8653-8661Abstract Full Text PDF PubMed Google Scholar). The glycosaminoglycan chains in decorin and biglycan are O-glycosidically linked to Ser residues in the N-terminal region, providing pronounced polyanionic properties to the proteoglycans. In analogy, the N-terminal domains in fibromodulin (15Antonsson P. Heinegård D. Oldberg Å. J. Biol. Chem. 1991; 266: 16859-16861Abstract Full Text PDF PubMed Google Scholar) and most likely also in lumican carry clusters of negatively charged Tyr sulfate residues. The C-terminal domains comprise some 50 amino acid residues and show considerable similarities among family members. This domain contains two Cys residues involved in an intrachain disulfide bond leading to the formation of a 34–41-residue loop. The common central domains constitute 60–80% of the total amino acids. In most of the members of the family, it is composed of a 10–11-fold repeat of a 20–25-residue long LRR with preferentially Asn and Leu residues in conserved positions. Such repeats are found in many intracellular proteins. Studies of one of these, i.e. the ribonuclease inhibitor, has provided important structural information. Thus its three-dimensional structure has been determined by x-ray crystallography (16Kobe B. Deisenhofer J. Nature. 1993; 366: 751-756Crossref PubMed Scopus (561) Google Scholar). It was found that the LRRs form a horseshoe-shaped coil of parallel, alternating α-helices and β-sheets stabilized by interchain hydrogen bonds. Presumably the extracellular matrix LRR glycoproteins/proteoglycans have a similar three-dimensional structure (17Weber I.T. Harrison R.W. Iozzo R.V. J. Biol. Chem. 1996; 271: 31767-31770Abstract Full Text Full Text PDF PubMed Scopus (308) Google Scholar). There are consensus site Asn residues in the central repeat domain for substitution with carbohydrates. In fibromodulin four such sites appear to serve as acceptors for keratan sulfate (15Antonsson P. Heinegård D. Oldberg Å. J. Biol. Chem. 1991; 266: 16859-16861Abstract Full Text PDF PubMed Google Scholar). However, in an individual molecule no more than two of the sites carry a keratan sulfate chain (18Plaas A.H.K. Neame P. Nivens C.M. Reiss L. J. Biol. Chem. 1990; 265: 20634-20640Abstract Full Text PDF PubMed Google Scholar). In lumican equivalent positions can be substituted with carbohydrates. This substitution is also more variable, and lumican is present as a keratan sulfate proteoglycan primarily in the cornea, while being a classical glycoprotein, without the repeat sulfated disaccharides, in tissues such as skin and cartilage (9Grover J. Chen X.-N. Korenberg J.R. Roughley P.J. J. Biol. Chem. 1995; 270: 21942-21949Crossref PubMed Scopus (170) Google Scholar). The glycosylation of fibromodulin and lumican is a process that appears to be tissue-specific, developmentally regulated (19Cornuet P.K. Blochberger T.C. Hassel J.R. Invest. Ophtalmol. Visual Sci. 1994; 35: 870-877PubMed Google Scholar), and dependent on age (20Roughley P.J. White R.J. Cszabo G. Mort J.S. Osteoarthritis Cartilage. 1996; 4: 229-255Abstract Full Text PDF Scopus (55) Google Scholar). Decorin (21Vogel K.G. Paulsson M. Heinegård D. Biochem. J. 1984; 223: 587-597Crossref PubMed Scopus (747) Google Scholar), fibromodulin (22Hedbom E. Heinegård D. J. Biol. Chem. 1989; 264: 6898-6905Abstract Full Text PDF PubMed Google Scholar), and lumican (23Rada J.A. Cornuet P.K. Hassel J.R. Exp. Eye Res. 1993; 56: 635-648Crossref PubMed Scopus (300) Google Scholar) bind to fibrillar collagens in vitro, leading to delayed fibril formation and the formation of thinner fibrils (24Vogel K.G. Trotter J.A. Collagen Relat. Res. 1987; 7: 105-114Crossref PubMed Scopus (286) Google Scholar). This is most likely caused by the binding of LRR glycoproteins/proteoglycans to the surface of the axially growing fibril (25Kadler K.E. Holmes D.F. Trotter J.A. Chapman J.A. Biochem. J. 1996; 316: 1-11Crossref PubMed Scopus (1113) Google Scholar), which inhibits the incorporation of additional triple helical collagen monomers. The binding of LRR glycoproteins/proteoglycans to collagen alters the surface properties of the fibrils and may affect the interactions between individual collagen fibrils as well as between fibrils and matrix constituents, other than LRR glycoproteins/proteoglycans. It has been proposed that competitive binding and displacement of proteoglycans regulate the growth of collagen fibrils during development (26Scott J. Biochem. J. 1988; 252: 313-323Crossref PubMed Scopus (584) Google Scholar). Interestingly, the binding of decorin to collagen is not inhibited by fibromodulin andvice versa, suggesting that these two members of the gene family bind to different sites on the collagen fibril (27Hedbom E. Heinegård D. J. Biol. Chem. 1993; 268: 27307-27312Abstract Full Text PDF PubMed Google Scholar). Ultrastructural analysis of collagen fibrils in tissues demonstrate that decorin (28Pringle G.A. Dodd C.M. J. Histochem. Cytochem. 1990; 38: 1405-1411Crossref PubMed Scopus (137) Google Scholar) and fibromodulin (29Hedlund H. Mengarelli-Widholm S. Heinegård D. Reinholt F.P. Svensson O. Matrix Biol. 1994; 14: 227-232Crossref PubMed Scopus (108) Google Scholar) bind to distinct and apparently separate sites in the gap region of the D-period of the collagen fibrilin vivo. The binding site for collagen in decorin has tentatively been localized to LRR 4–5 in the central domain (30Svensson L. Heinegård D. Oldberg Å. J. Biol. Chem. 1995; 270: 20712-20716Crossref PubMed Scopus (203) Google Scholar), and recent results suggest a critical role for a Glu residue in this region for interaction with type I collagen (31Kresse H. Liszio C. Schönherr E. Fisher L.W. J. Biol. Chem. 1997; 272: 18404-18410Abstract Full Text Full Text PDF PubMed Scopus (85) Google Scholar). The interactions between collagens and LRR glycoproteins/proteoglycans appear to be crucial for the correct assembly of collagen fibrils and the generation of a functional collagen matrix scaffold in vivo. Mice lacking decorin have fragile skin with reduced tensile strength, a phenotype consistent with a malfunctioning collagen matrix (32Danielson K.G. Baribault H. Holmes D.F. Graham H. Kadler K. Iozzo R.V. J. Cell Biol. 1997; 136: 729-743Crossref PubMed Scopus (1199) Google Scholar). Ultrastructural analysis of decorin-null mice revealed abnormal collagen fibrils in tissues such as skin and tendon, with more coarse and irregular collagen fiber outlines. Lumican-null mice show skin fragility and corneal opacity (33Chakravarti S. Magnuson T. Lass J.H. Jepsen K.J. LaMantia C. Carroll H. J. Cell Biol. 1998; 141: 1277-1286Crossref PubMed Scopus (597) Google Scholar). Electron microscopic investigations of lumican-null mice reveal a deregulated growth of collagen fibrils with a significant proportion of abnormally thick fibrils in skin and cornea. The recent evaluation of biglycan-deficient mice suggests a role for biglycan in bone homeostasis (34Xu T. et al.Nat. Genet. 1998; 20: 78-82Crossref PubMed Scopus (391) Google Scholar). We have generated fibromodulin-deficient mice to study a possible role for fibromodulin in the assembly of collagen matrices. Here we show that a null mutation in the fibromodulin gene leads to abnormal collagen fibrils in tendons. In addition, the absence of fibromodulin leads to an increase in lumican, which is associated with decreased lumican mRNA levels. A bovine fibromodulin cDNA (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar) was used to isolate several overlapping genomic clones from a cosmid mouse 129/sv library (kindly provided by J. S. Mudgett, Merck, Rahway, NJ). A 8-kilobase pair EcoRI/SalI fragment that includes exon 2 was subcloned in pBluescript KS II (Stratagene, La Jolla, CA). ABamHI site located 5 kilobase pairs from the 5′ end of the genomic fragment was used for the insertion of a phosphoglycerate kinase-neomycin resistance cassette (pGKNeo) (35McBurney M.W. Sutherland L.C. Adra C.N. Leclair B. Rudnicki M.A. Jardine K. Nucleic Acids Res. 1991; 20: 5755-5761Crossref Scopus (191) Google Scholar), which interrupts the coding sequence in exon 2 at a Trp residue located in position 8 of the mature mouse fibromodulin (GenBankTM accession numberX94998). The resistance cassette pGKNeo was blunt end ligated into theBamHI site in the opposite transcriptional orientation of the fibromodulin gene. The targeting vector also contained a herpes simplex virus thymidine kinase cassette, pIC19R/MC1-Tk (36Mansour S.L. Thomas K.R. Capecchi M.R. Nature. 1988; 336: 348-352Crossref PubMed Scopus (1438) Google Scholar), which was blunt end ligated in the NaeI site of pBluescript KSII prior to insertion of the genomic fragment. R1 embryonic stem cells (37Nagy A.J. Rossant R. Nage W. Abramow-Newerly W. Roder J.C. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 8424-8428Crossref PubMed Scopus (2011) Google Scholar) were grown and electroporated as described previously (38Fässler R. Schnegelsberg P.N.J. Dausman J. Shinya T. Muragaki Y. McCarthy M.T. Olsen B.R. Jaenisch R. Proc. Natl. Acad. Sci. U. S. A. 1994; 91: 5070-5074Crossref PubMed Scopus (265) Google Scholar). Briefly, the targeting vector (45 μg) was linearized with NotI and used in the electroporation of 3 × 107 embryonic stem cells. The cells were selected with 0.4 mg/ml G418 (Life Technologies, Inc.) and 0.2 μm FIAU (Bristol-Myers Squibb, Wallingford, CT). Individual clones were picked and isolated DNA was screened by Southern blotting after digestion withEcoRI. Filters were hybridized with a 300-bpEcoRI/SalI external probe to identify targeted clones (see Fig. 1). Two individually targeted ES cell clones were used to generate chimeric mice as described (38Fässler R. Schnegelsberg P.N.J. Dausman J. Shinya T. Muragaki Y. McCarthy M.T. Olsen B.R. Jaenisch R. Proc. Natl. Acad. Sci. U. S. A. 1994; 91: 5070-5074Crossref PubMed Scopus (265) Google Scholar). Chimeric males were mated with C57/B6 mice, and males with germ line transmission were further bred with 129/sv females to establish an inbred strain of fibromodulin-null mice. Total RNA was isolated from mouse tails or tendons using guanidine isothiocyanate. After CsCl gradient centrifugation, RNA was subjected to agarose gel electrophoresis in formaldehyde, and fibromodulin mRNA was detected on the filters as described previously (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar). Proteins were extracted from tails using 4 m guanidine HCl, 0.05 m sodium acetate, pH 5.8, containing the protease inhibitor Complete (Boehringer Mannheim) as described (39Larsson T. Sommarin Y. Paulsson M. Antonsson P. Hedbom E. Wendel M. Heinegård D. J. Biol. Chem. 1991; 266: 20428-20433Abstract Full Text PDF PubMed Google Scholar). After 48 h at 4 °C, the extracts were cleared by centrifugation, and the supernatant was subjected to precipitation with 9 volumes of 95% ethanol. The protein precipitates were reprecipitated twice with 9 volumes of ethanol after resuspension in 0.05 m sodium acetate, pH 5.8. Finally the precipitates were dissolved in SDS-polyacrylamide gel electrophoresis sample buffer and electrophoresed on a 10% SDS-polyacrylamide gel. Proteins were transferred to nitrocellulose filters by diffusion and detected using a rabbit antiserum against bovine fibromodulin as described (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar). Bound antibodies were visualized with an ECL kit (Amersham Pharmacia Biotech) or 125I-labeled goat anti-rabbit IgG, a generous gift from Dr. B. Åkerström (Department of Cell and Molecular Biology, Lund University, Lund, Sweden) (40Björck L. Cigén R. Berggård B. Löw B. Berggård I. Scand. J. Immunol. 1977; 6: 1063-1069Crossref PubMed Scopus (36) Google Scholar). Antisera against intact rat decorin (41Oldberg Å. Antonsson P. Moses J. Fransson L.-Å. FEBS Lett. 1996; 386: 29-32Crossref PubMed Scopus (29) Google Scholar) and bovine fibromodulin (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar) were used in protein transfer blots, ELISA, and immunostainings of mouse tissues. An antiserum against mouse lumican was prepared in rabbits using a bacterially produced lumican/glutatione-thio-transferase fusion protein as described (41Oldberg Å. Antonsson P. Moses J. Fransson L.-Å. FEBS Lett. 1996; 386: 29-32Crossref PubMed Scopus (29) Google Scholar). AMscI/StuI fragment from exon 2 of the mouse lumican gene was ligated in pGEX-3X (Amersham Pharmacia Biotech). This DNA encodes a 66-residue-long peptide, corresponding to residues 19–84 in the mouse sequence (GenBankTM accession numberAF013262). Fusion protein was produced and purified as recommended by the manufacturer. ELISA was performed as described previously (39Larsson T. Sommarin Y. Paulsson M. Antonsson P. Hedbom E. Wendel M. Heinegård D. J. Biol. Chem. 1991; 266: 20428-20433Abstract Full Text PDF PubMed Google Scholar). Microtiter plates were coated with recombinant mouse decorin, mouse fibromodulin, and mouse lumican fusion proteins, which were produced in bacteria and purified as recommended by the manufacturers. A mouse lumican/maltose-receptor fusion protein was prepared by ligating a MscI/StuI genomic fragment (see above under “Immunological Methods”) in pMALp (New England Biolabs). Recombinant mouse lumican fusion protein was isolated on amylose-Sepharose. A full-length mouse decorin cDNA was prepared using mRNA from mouse tail. The mRNA was converted to cDNA with oligo(dT) primers and reverse transcriptase (Boehringer Mannheim). The single-stranded cDNA was amplified using the primers 5′-CCGGATCCATGAAGGCAACTCTCATCTTC and 5′-GGGGTACCTTGTAGTTTCCAAGTTGAATG (42Scholzen T. Solursh M. Suzuki S. Reiter R. Morgan J.L. Buchberg A.M. Siracusa L.D. Iozzo R.V. J. Biol. Chem. 1994; 269: 28270-28281Abstract Full Text PDF PubMed Google Scholar) and the DNA polymerase Dynazyme (Finnzymes Oy, Espoo, Finland). Amplification conditions were 30 cycles of 95 °C for 30 s, 55 °C for 45 s, and 72 °C for 2 min. The polymerase chain reaction product was ligated in the BamHI/KpnI site of pQE 30 (Qiagen, Hilden, Germany), expressed as a His-tagged protein in bacteria and purified on nickel-agarose. Mouse fibromodulin antigen was also produced as a His-tagged protein. Single-stranded cDNA derived from mouse tail mRNA (see above under “RNA and Protein Transfer Blot Analysis”) was amplified using the primers 5′-TCCCCCGGGGATGCAGTGGGCCTCCGTC and 5′-CCCAAGCTTCAGATCCGATGAGGTTGG (GenBankTM accession numberX94998) and polymerase chain reaction conditions as described above for the production of mouse decorin. A BglII/HindIII fragment, representing a 189-amino acid-residue C-terminal fragment was ligated in pQE30 (Qiagen). Recombinant protein was expressed in bacteria and purified as described above. Protein extracts from tails or tail tendons were prepared as described (39Larsson T. Sommarin Y. Paulsson M. Antonsson P. Hedbom E. Wendel M. Heinegård D. J. Biol. Chem. 1991; 266: 20428-20433Abstract Full Text PDF PubMed Google Scholar), precipitated with ethanol and redissolved in 0.14 mNaCl, 8 mm Na2HPO4, 2.7 mm KCl, 1.5 mm KH2PO4, 3 mm NaN3, pH 7.4, containing 0.8% SDS. Protein concentrations in extracts were determined according to Bradford (43Bradford M.M. Anal. Biochem. 1976; 72: 248-254Crossref PubMed Scopus (225016) Google Scholar) using a Coomassie protein assay reagent kit (Pierce). Immunolocalization of proteins in mouse tissues was performed on paraffin sections of mouse tails and Achilles tendon as described (44Aszodi A. Modis A. Paldi A. Rencendorj I. Kiss I. Bösze Z. Matrix Biol. 1994; 14: 181-190Crossref PubMed Scopus (29) Google Scholar). Tails were fixed in 95% ethanol, 1% acetic acid and decalcified in 6% EDTA in phosphate-buffered saline for 1 week prior to dehydration and embedding. Achilles tendons were fixed in 4% paraformaldehyde in phosphate-buffered saline. For histological analysis, paraffin sections of tail, sternum, knee-joint, muscle, and inner organs were stained with hematoxylin and Chromotrope 2R (Sigma). Ribonuclease protection assays were used to determine amounts of mRNA. Total RNA was isolated from tails or tail tendons after extraction with guanidine isothiocyanate and centrifugation through a CsCl cushion as described (6Oldberg Å. Antonsson P. Lindblom K. Heinegård D. EMBO J. 1989; 8: 2601-2604Crossref PubMed Scopus (258) Google Scholar). Poly(A)+ RNA was isolated from the total RNA using an oligo(dT) matrix kit (Qiagen). Antisense mouse lumican mRNA was prepared using a 258-bp MscI/XbaI fragment (GenBankTM accession number AFO13262) of the mouse lumican gene ligated in pBluescript KSII, using T7 RNA polymerase and a RNA transcription kit (Ambion, Austin, TX). Antisense mouse actin RNA was transcribed from pTRI-β-actin-mouse (Ambion). The protection assay was performed using a kit according to the manufacturer′s recommendations (Ambion). Protected double-stranded RNA was separated on a 6% polyacrylamide sequencing gel, and the radioactivity was determined using a Bio Imaging Analyzer (Fuji Photo Film Co., Japan). Fibromodulin, lumican and decorin mRNA were localized by in situ hybridization on longitudinally cut paraffin sections of tail as described (45Apte S.S. Fukai N. Beir D.R. Olsen B.R. J. Biol. Chem. 1997; 272: 25511-25517Abstract Full Text Full Text PDF PubMed Scopus (123) Google Scholar) after fixation in 4% paraformaldehyde and decalcification in 6% EDTA, 0.5% paraformaldehyde in phosphate-buffered saline. RNA probes were labeled with digoxigenin-UTP by in vitro transcription with SP6, T7, or T3 RNA polymerases using a digoxigenin RNA labeling kit (Boehringer Mannheim) according to the manufacturer′s recommendations. Antisense mouse fibromodulin probe was transcribed from a 267-bpBamHI/BglII cDNA fragment in pBluescript KSII using T7 RNA polymerase. Mouse lumican probes were transcribed from a 258-bp MscI/XbaI fragment ligated in pBluescript, where T7 RNA polymerase was used to produce an antisense probe and T3 RNA polymerase was used to produce a sense probe. A 324-bpBamHI/HincII mouse decorin cDNA fragment was ligated in pGEM7Z (Promega, Madison, Wisconsin), and T7 polymerase was used to produce an antisense probe, whereas SP6 was used to produce a sense probe. Tissues from littermates representing fibromodulin-null, heterozygous, and wild type mice of the same sex at 7 and 20 weeks of age were subjected to electron microscopy. Immediately after sacrifice, Achilles tendon samples were fixed by immersion in 0.1 m sodium cacodylate-buffered 2% glutaraldehyde containing 0.1 m sucrose. The specimens were postfixed in 0.1 m s-collidine-buffered 2% osmium, dehydrated in graded ethanol, and embedded in an epoxy resin according to standard procedure (46Reinholt F.P. Engfelt B. Hjerpe A. Jansson K. J. Ultrastruct. Res. 1982; 80: 270-279Crossref PubMed Scopus (45) Google Scholar). Ultrathin sections were contrasted with uranyl acetate and lead citrate. Samples cut longitudinally or transversely with respect to the main axis of collagen fibers were studied. Micrographs from transverse sections of the tendon were sampled by systematic random sampling and subjected to morphometry. Thus, from each animal the diameter of each of the transversely cut collagen fibrils of the tendon was measured on 20 micrographs from two tissue blocks with an interactive optomechanical particle size analyzer and presented in a histogram. The fibromodulin gene was inactivated by homologous recombination in embryonic stem cells with a targeting vector interrupting the coding sequence in exon 2 with a neomycin resistance cassette (Fig. 1). After transformation and selection with G418 and FIAU, we identified three correctly targeted clones out of 150 screened. Two of these were used to produce chimeric mice, which were mated with C57/B6 females. Chimeric males with germ line transmission were mated with 129/sv females to generate an inbred strain. Breeding of heterozygous mice resulted in the recovery of homozygous and heterozygous mice in the proportions expected for a single copy gene mutation. The fibromodulin-null mice did not exhibit any gross anatomical abnormalities, grew to normal size, were fertile, and had a normal life span. Light microscopic investigations of heart, liver, lung, kidney, skin, and cartilage did not reveal any abnormalities. Transfer blot analysis of proteins extracted from tails showed the absence of fibromodulin in the null mice (Fig.2 A). Northern blot analysis showed an absence of fibromodulin mRNA in fibromodulin-null mice and about half the amount in heterozygous mice as compared with wild type mice (Fig. 2 B). Hematoxylin/chromotrope stained cross-sections of tails revealed abnormal tendon collagen fiber bundles in fibromodulin-null and heterozygous mice. In comparison the tendon fiber bundles in wild type animals are highly organized with evenly distributed cells (Fig. 3,A–C). In fibromodulin-null animals most fiber bundles have a different appearance and show an abnormal morphology (Fig. 3,G–I), as compared with those in wild type mice. Also in heterozygous animals, 10–20% of the collagen fiber bundles are similar to the abnormal bundles in fibromodulin-null mice (Fig. 3,D–F). The abnormal fiber bundles appear less organized with unevenly distributed cells. The transverse sections suggest fewer cells in fibromodulin-null as compared with wild type tendon. However, longitudinal sections indicate that a similar number of fibroblasts are present in fibromodulin-null and wild type tendon fibers. The abnormal fiber bundles also appear to have reduced endotenon tissue, which is composed of cells and loose connective tissue surrounding these bundles (47Blevins F.T. Djurasovic M. Flatow E.L. Vogel K.G. Orthop. Clin. North Am. 1997; 28: 1-16Abstract Full Text Full Text PDF PubMed Scopus (123) Google Scholar). In association the fibromodulin-null collagen bundles appear to shrink more during the fixation and also detach from the surrounding tissues. Furthermore, the number of fiber bundles, surrounded by epitenon (47Blevins F.T. Djurasovic M. Flatow E.L. Vogel K.G. Orthop. Clin. North Am. 1997; 28: 1-16Abstract Full Text Full Text PDF PubMed Scopus (123) Google Scholar), is reduced in fibromodulin-null and to a lesser degree also in heterozygous animals. Counting fiber bundles in four groups of littermates (1, 1.5, 3.5, and 20 months old) shows a 25–55% reduction in fibromodulin-null as compared with wild type littermates. Heterozygous animals had 0–32% less bundles than wild type. The distribution of fibromodulin, lumican, and decorin was investigated in cross-sections of tails using immunohistochemistry. In wild type animals fibromodulin is preferentially present in the collagen bundles of tendons, but also a weaker staining of the dermis was observed (Fig.4 A). Lumican was preferentially detected in dermis and in peritenon tissues immediately surrounding tendons in wild type mice (Fig. 4 B), whereas the decorin antiserum showed staining in tendons and skin (Fig.4 C). The fibromodulin, lumican, and decorin staining patterns in heterozygous and wild type tail samples showed differences in the collagen fiber bundles. Fiber bundles with an abnormal morphology stained weakly with the lumican antiserum (Fig.4 E) and showed an uneven, patchy staining pattern with fibromodulin (Fig. 4 D) and decorin (Fig. 4 F) antisera. Fibromodulin was absent in fibromodulin-null mice (Fig.4 G), where we observed a change in the amount and distribution of lumican. The anti-lumican staining in fib