A Syndrome of Joint Laxity and Impaired Tendon Integrity in Lumican- and Fibromodulin-deficient Mice* 210

Lumican and fibromodulin regulate the assembly of collagens into higher order fibrils in connective tissues. Here, we show that mice deficient in both of these proteoglycans manifest several clinical features of Ehlers-Danlos syndrome. TheLum −/− Fmod −/− mice are smaller than their wild type littermates and display gait abnormality, joint laxity, and age-dependent osteoarthritis. Misaligned knee patella, severe knee dysmorphogenesis, and extreme tendon weakness are the likely causes for joint laxity in the double-nulls. Fibromodulin deficiency alone leads to significant reduction in tendon stiffness in theLum +/+ Fmod −/− mice, with further loss in stiffness in a Lum gene dose-dependent way. At the protein level, we show marked increase of lumican in Fmod −/− tendons, which may partially rescue the tendon phenotype in this genotype. These results establish fibromodulin as a key regulator and lumican as a modulator of tendon strength. A disproportionate increase in small diameter immature collagen fibrils and a lack of progression to mature, large diameter fibrils in the Fmod −/−background may constitute the underlying cause of tendon weakness and suggest that fibromodulin aids fibril maturation. This study demonstrates that the collagen fibril-modifying proteoglycans, lumican and fibromodulin, are candidate genes and key players in the pathogenesis of certain types of Ehlers-Danlos syndrome and other connective tissue disorders.

Marfan, and Larsen syndromes display a wide spectrum of connective tissues defects. The clinical features of these disorders include hyperextensible skin; vascular defects; ocular, bowel, and bladder fragility; joint hypermobility and dislocation; musculoskeletal defects; and articular cartilage degeneration (1)(2)(3)(4)(5)(6)(7)(8). Clearly, the structure and organization of collagens dictate the biomechanical properties and normal functions of connective tissues (9 -12). Thus, the primary defects in some of these disorders are mutations in the genes for collagens I (EDS, OI), II (Stickler), III (EDS), V (EDS), XI (Stickler) and in the genes for procollagenases (EDS) and lysyl hydroxylases (EDS), enzymes that modify collagen chains (Online Mendelian Inheritance in Man, www.ncbi.nlm.nih.gov/) (4). Although a fraction of EDS types I, II, and III are due to mutations in genes that code for collagens, the molecular defects in a large proportion of these are not known. Understandably, for decades the search for underlying molecular genetic defects has focused on the collagens and enzymes directly modifying the collagen chains (13,14). It is increasingly evident, however, that biological processes that regulate assembly or the higher order organization of the collagenous matrix are likely to contribute to these disorders. The small leucine-rich repeat proteoglycans (SLRP) have emerged as a significant multimember group, expressed at high levels in many collagenous tissues and capable of binding and regulating collagen fibril assembly and growth, likely to be important in these disorders (1,4,15,16). Along this line, Morquio syndrome or mucopolysaccharidoses IV is associated with genetic defects in enzymes involved in the degradation of keratan sulfates, glycosaminoglycan side chains of connective tissue proteoglycans (4). The SLRPs, lumican, fibromodulin, decorin, and biglycan, have recently become the focus of gene-targeting studies to understand their roles in connective tissue biology and disease (17)(18)(19)(20)(21)(22). Lumican and fibromodulin are widely expressed in collagenous connective tissues and play a significant role in defining tissue integrity. Within this family, ten structurally and functionally related members are further grouped into three classes based on amino acid sequence identity, presence of cysteine-rich clusters at the N terminus and number and spacing of the leucine-rich repeats (16,23). Lumican and fibromodulin, class II members, are related closely by structure; the core proteins contain N-and C-terminal cysteines and a central domain of 9 -10 leucine-rich motifs that are sites of protein-protein interactions. Lumican is a component of the cornea, skin, sclera, tendon, and cartilage (17, 24 -27); fibromodulin is expressed at high levels in all of these tissues except the skin and the cornea (28). Earlier studies of gene-targeted mice deficient in each of these proteoglycans have shown altered collagen fibril architecture in multiple * This work was supported by the National Institute of Health-National Eye Institute Grants EY11654 (to S. C.), AR44927 (to K. J. J.), and AR44745 (to D. E. B.) and by The Swedish Medical Research Council Grant K2002-03X-07478-17A (to A. O.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
connective tissues (17,21). Lumican-deficient mice have collagen fibrils of increased diameter forming a disorganized matrix in the cornea and skin, with consequent decrease in corneal clarity and increased skin laxity (17,29,30). Fmod Ϫ/Ϫ mice contain more immature, small diameter collagen fibrils in the tendon (19,21). Fmod Ϫ/Ϫ mice also show an increase in agedependent osteoarthritis and degenerative changes of the articular cartilage. 2 In vitro lumican and fibromodulin bind to the same region on collagen type I and affect collagen fibrillogenesis, implying that in vivo they are likely to have significant functional overlap in tissues where they are co-expressed (31). To investigate functional overlap(s) between lumican and fibromodulin, we developed mice jointly deficient in both proteoglycans. In an earlier study, we have shown mild, moderate, and severe alteration in collagen fibril structure of tendons in the Lum Ϫ/Ϫ Fmod ϩ/ϩ , Lum ϩ/ϩ Fmod Ϫ/Ϫ , and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mice, respectively (19). Although the regulatory role of lumican and fibromodulin in the assembly of tendon-collagen fibrils is clearly evident from this study, the effects of the double-null mutation on tendon mechanical properties were not elucidated nor were their influence on other organs explored. In the current study, we undertook a detailed characterization of the lumican-fibromodulin double-null phenotype. Our study reveals marked joint laxity, abnormal gait, and a severe reduction in tendon stiffness in double-null adult mice. An in-depth study of tendon ultrastructure and biomechanical strength demonstrates an essential role for fibromodulin, with lumican as its modulator, in collagen fibril structure and biomechanical strength of tendons. These studies establish a fundamental role for lumican 2 A. Oldberg, unpublished observations. FIG. 1. A, Lum ϩ/ϩ Fmod ϩ/ϩ (left) and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ (right) to demonstrate small body size and short stature. Gait abnormality in the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mouse. B, note abnormal positioning of the hind leg (arrow). C, knee laxity measurements show significantly increased laxity in the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mouse. Representative load-deflection curves are shown for one cycle during the knee laxity test. Joint laxity was defined by adding the tibial deflections in the anterior and posterior directions up to the point when the loading phase in each direction (indicated by arrow) reached a load magnitude of 0.5 newton from the initial load at zero deflection (dotted horizontal lines). Knee joint histology of transverse sections of femoral condyle of Lum ϩ/ϩ Fmod ϩ/ϩ (D, top panel) and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ (D, bottom panel). Note normal patella of Lum ϩ/ϩ Fmod ϩ/ϩ knees (arrows, top panel) located within the patella groove. In contrast, Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mice exhibited a medial misalignment of the patella (arrows, bottom panel). The distal femur exhibited a hypertrophic response, providing additional trabecular tissue to support the misaligned patella. and fibromodulin in regulating collagen fibril architecture of connective tissues and modulating tissue strength.
a p Ͻ 0.01 values represent differences relative to Lum ϩ/ϩ Fmod ϩ/ϩ (analysis of variance with a Newman-Keuls post-hoc test). b p Ͻ 0.001 values represent differences relative to Lum ϩ/ϩ Fmod ϩ/ϩ (analysis of variance with a Newman-Keuls post-hoc test). c ND, not done. d n ϭ number of tendons.
assessed by semi-quantitative Western blots in FDL tendons from 2-month-old mice of the above genotypes. Mice were handled according to policies approved by the Johns Hopkins University animal care committee.
Knee Laxity-To test for alterations in knee laxity, the right hind limbs were harvested from 6-month-old wild type (n ϭ 12) and double knockout (n ϭ 10) mice. Skin and muscle were removed from the limbs under a stereomicroscope leaving the joint capsule intact. The femur and tibia were potted in brass tubes using a quick-setting epoxy leaving the joint exposed. The limbs were placed in a custom-made loading fixture with the tibia fixed at a flexion angle of 60 degrees (32). The knee joint was preconditioned for 30 s, by translating the tibia Ϯ0.25 mm in the antero-posterior direction and then subjected to an anteroposterior translation of Ϯ0.5 mm for 60 s. Joint laxity was determined by adding the tibial translations in the anterior and posterior directions when the load magnitude reached 0.5 newton from the initial load at zero deflection. This joint laxity measure represents the deflection necessary to begin engaging the resistance offered by the passive constraints of the knee (e.g. joint capsule and ligaments).
Tendon Mechanical Properties-FDL tendons were removed from the left hind limbs, stored in Hanks' media at 4°C, and tested to failure in tension within 24 h of sacrifice. The distal and proximal ends of the FDL tendons were placed between sandpaper-covered plates leaving a 9-mm gauge region and fixed to a servohydraulic materials testing system (Instron Corp., Canton, MA). Samples were subjected to five preconditioning cycles by loading the tendons to 4.5% strain at 10%/s, holding for 20 s, and then unloading. Tendon stiffness equilibrated after three to four preconditioning cycles. The tendons were then loaded to failure at 10%/s. Load-deformation data were collected at 50 Hz using LabVIEW (National Instruments, Austin, TX). Measures of whole tendon stiffness (Table I) and maximum load (not shown) were calculated from the load-deformation curves of the failure tests. Variation in body size was corrected by regression analysis. Differences in tendon material properties were determined by taking the variation in tendon size into consideration. Tensile modulus and strength, representing tissue-level mechanical properties, independent of tendon size, were calculated by normalizing the load-deformation data by cross-sectional area and the gauge length (Table I).
Tendon and Knee Joint Morphology-To test for variations in the size of the FDL tendons, the contralateral limbs of those used in the biomechanical testing, were fixed in 10% neutral buffered formalin, rinsed, dehydrated in a series of graded alcohols, cleared in xylene, and embedded in poly-methylmethacrylate. A low speed diamond-coated saw (Buhler Inc., Lake Bluff, IL) was used to generate transverse sections serially through the area representative of the biomechanical test gauge region. The transverse sections were glued to acrylic slides, polished to 1-m diamond finish, and surface-stained with toluidine blue. The sections were digitized under transmitted light at 63ϫ using public domain image analysis software (Scion Corp., Frederick, MD). Area measurements were obtained for three transverse sections and averaged.
Transmission Electron Microscopy-Five-month-old mice were sacrificed, and three to five mice were used per genotype. The hind limbs were fixed in situ in 4% paraformaldehyde, 2.5% glutaraldehyde, 0.1 M sodium cacodylate, pH 7.4, with 8.0 mM CaCl 2 for 15 min, during which time the flexor digitorum longus (FDL) tendons were dissected. The tendons were further fixed for 2 h at 4°C and processed as previously described (19). Briefly, the tendons were post-fixed with 1% osmium tetroxide and stained en bloc with uranyl acetate/50% ethanol. After dehydration in an ethanol series, followed by propylene oxide, the tendons were infiltrated and embedded in a mixture of EMbed 812, nadic methyl anhydride, and DMP-30 (Electron Microscopy Sciences, Fort Washington, PA). Thin sections were cut using a Reichert UCT Ultramicrotome and a diamond knife and stained with 2% aqueous uranyl acetate, 1% phosphotungstic acid, pH 3.2. Sections were taken in the mid region of the tendon and were analyzed in Lum ϩ/ϩ Fmod Ϫ/Ϫ , Lum Ϫ/Ϫ Fmod ϩ/ϩ , Lum Ϫ/Ϫ fmod Ϫ/Ϫ , and Lum ϩ/ϩ fmod ϩ/ϩ mice using transmission electron microscopy. Sections were examined and photographed at 75 kV using an Hitachi 7000 transmission electron microscope.
Fibril Diameter Analyses from Transmission Electron Micrographs-For each genotype, five to six different tendons from three to five animals were analyzed. Micrographs (38 -54/group) from nonoverlapping regions of the central portion of the tendon were taken at 31,680ϫ. Micrographs were randomly chosen in a masked manner from the different groups and digitized, and diameters were measured using an image analysis system (RM Biometrics-Bioquant, Memphis, TN). A region from a single photographic negative, containing ϳ60 -290 fibrils, was measured. The frequency of different diameter fibrils was analyzed in a total of 4121 to 7960 collagen fibrils.
Western Blot Analyses-FDL tendons isolated from 2-month-old mice were weighed and cut into small pieces. T-PER tissue extraction reagent (Pierce Technology) containing Halt protease inhibitor mixture (Pierce Technology) at a final concentration of 1ϫ was added at 20 ml/g of tissue, and samples were homogenized on ice. Tissue debris was removed by centrifugation for 5 min at 10,000 rpm. The supernatant was saved, and its protein concentration was determined by the BCA protein assay (Pierce Technology). The proteoglycans were visualized by SDS-PAGE and immunoblotting using antibodies against lumican and fibromodulin (17,21). As an internal control, actin was visualized by using a polyclonal antibody (C11, Santa Cruz Biotechnology, Santa Cruz, CA).
The threshold cycle difference Ct ϭ (Ct of Lum Ϫ Ct of GAPDH) Ϫ (Ct of Lum no template control Ϫ Ct of GAPDH no template control). According to published methods (33) the relative expression ϭ 2 Ϫdifference Ct .
Statistical Analyses-Differences in mechanical and morphological properties between groups of mice were determined using a one-way analysis of variance and a Newman-Keuls post-hoc test (GraphPad Prism 3.0, San Diego, CA).
The double-knockout mice have bowed legs, exhibit severe gait abnormality, and walk on the dorsal part of the foot (Fig. 1B). Measurements of knee joint deflections show a 2-fold (p Ͻ 0.0001, t test) increase in deflection (0.64 Ϯ 0.16 mm of Lum Ϫ/Ϫ Fmod Ϫ/Ϫ compared with 0.30 Ϯ 0.12 mm of Lum ϩ/ϩ Fmod ϩ/ϩ mice) providing objective measurements of the observed joint hypermobility in the double-null mouse (Fig.  1C). Histology of the knee joints shows a severely malformed distal condyle with a medial misalignment of the patella in 65% of the double-knockout mice (Fig. 1D). This medial patellar misalignment is probably responsible for the gait abnormality of the double-null mouse. The knee shows a secondary patellar groove and compensatory overgrowth of the metaphyseal bone to support the misaligned patella. This suggests that the knee anomaly probably begins early during growth and development.
To determine if the gait anomaly and joint laxity is in part due to alterations in tendon mechanical functions, we tested the biomechanical properties of whole FDL tendons by loading to failure in tension. The different genotypes form two distinct clusters when tendon stiffness was plotted against body weight (Fig. 2). The Lum ϩ/ϩ Fmod ϩ/ϩ , Lum ϩ/Ϫ Fmod ϩ/ϩ , Lum Ϫ/Ϫ Fmod ϩ/ϩ , and Lum ϩ/Ϫ Fmod ϩ/Ϫ mice demonstrate a common linear relationship between whole tendon stiffness and body weight. The Fmod Ϫ/Ϫ genotypes show a different regression, indicating that for their body weight there is a marked reduction in tendon stiffness.
To compare whole tendon mechanical properties, the stiffness and maximum load were normalized for differences in body weight between groups. Whole tendon stiffness (Table  I) and maximum load (not shown) were unaffected in the Lum Ϫ/Ϫ Fmod ϩ/ϩ , Lum ϩ/Ϫ Fmod ϩ/ϩ , and Lum ϩ/Ϫ Fmod ϩ/Ϫ mice. However, significant reduction in tendon stiffness (Table  I) and maximum load (not shown) among the Fmod Ϫ/Ϫ genotypes (p Ͻ 0.001, ANOVA, compared with Lum ϩ/ϩ Fmod ϩ/ϩ ) indicates that fibromodulin deficiency is a prerequisite for weak tendons and that fibromodulin contributes significantly to tendon strength.
The magnitude of the reduction in stiffness associated with fibromodulin deficiency depends on the number of functioning lumican alleles (data in italic in Table I). When compared with wild type, Lum ϩ/ϩ Fmod Ϫ/Ϫ , Lum ϩ/Ϫ Fmod Ϫ/Ϫ , and the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ tendons exhibit a decrease in stiffness of 25% (p Ͻ 0.01), 45% (p Ͻ 0.001), and 61% (p Ͻ 0.001), respectively. The maximum load measurements (data not shown) demonstrate similar genotype-specific alterations. These results indicate that, in the absence of fibromodulin, the severity of the phenotype increased with the loss of each functioning lumican allele in a lumican dose-dependent manner. The absence of lumican alone causes no apparent loss in whole tendon mechanical function; apparently, lumican modulates the consequence of fibromodulin deficiency.
Because there are genotype-dependent variations in body weight, we asked whether the changes in tendon strength are an indirect effect of smaller tendon resulting from reduced body weight. In a graph of FDL cross-sectional area, quantified from plastic-embedded specimens versus body weight, all seven genotypes demonstrate a common linear regression between tendon area and weight, giving a best-fit equation: [cross-sectional area (mm 2 )] ϭ 0.004856 body weight (g) Ϫ 0.01309 (mm 2 ), r 2 ϭ 0.656, p ϭ 0.0001 (Fig. 3). Thus a smaller tendon is an outcome of a smaller body size/reduced body weight and is not a direct and independent consequence of loss of expression of either lumican or fibromodulin.
To obtain a measure of stiffness at the tissue level, the tensile modulus (whole tendon properties normalized for crosssectional geometry) was determined (data in boldface in Table  I). The tendon modulus was reduced by 29 and 49% in Lum ϩ/Ϫ Fmod Ϫ/Ϫ and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ relative to wild type controls, respectively. These results confirm that loss of fibromodulin and lumican alters tissue mechanical properties. Taken together, the reduction in whole tendon mechanical properties is not a consequence of smaller tendons but one of reduced material stiffness and strength. The reduced whole tendon stiffness (arising from the decreased tissue modulus) may alter load transfer from muscle to bone and lead to patellar misalignment as seen in a large number of the doubleknockout mice.
To determine whether joint laxity in the double-deficient genotype is associated with changes in the articular cartilage, serial sections through knees were evaluated by histology. The articular cartilage of 2-month-old Lum ϩ/ϩ Fmod ϩ/ϩ , Lum Ϫ/Ϫ Fmod ϩ/ϩ , Lum ϩ/ϩ Fmod Ϫ/Ϫ , and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mice appear normal (data not shown). However, by 5 months of age, Lum Ϫ/Ϫ Fmod Ϫ/Ϫ animals exhibit significant articular cartilage degeneration, a concomitant decrease in cell density at both the femoral and tibial surfaces and normal menisci (Fig. 4). Because the cartilage degeneration is seen in the older animals only, it appears to be largely due to abnormal usage of the joint and not an early developmental defect.
The SLRPs appear to regulate their own expression by a feedback mechanism. We examined, by Western blotting, lumican levels in FDL tendon extracts of 2-month-old animals (Fig.  5a). The results show a dramatic increase in lumican in the Lum ϩ/ϩ Fmod Ϫ/Ϫ FDL tendon. Furthermore, the presence of about half as much lumican in Lum ϩ/Ϫ Fmod Ϫ/Ϫ mice as compared with Lum ϩ/ϩ Fmod Ϫ/Ϫ littermates demonstrates a gene dosage-dependent level of the lumican protein in the Fmod Ϫ/Ϫ background (Fig. 5a). In contrast, there is no compensatory increase of fibromodulin in the absence of lumican; there is a slight decrease instead (Fig. 5b). Twice as much fibromodulin is evident in the Fmod ϩ/ϩ versus Fmod ϩ/Ϫ lumican-deficient animals revealing a gene dose-dependent expression of fibromodulin in the Lum Ϫ/Ϫ background (Fig. 5b).
To investigate regulation of gene expression at the transcriptional level, lumican and fibromodulin message levels were measured by real-time PCR on RNA extracted from FDL tendons. The data are presented as message levels of lumican (italic) and fibromodulin (boldface) relative to GAPDH (Table  II). As expected there was no detectable lumican message in the Lum Ϫ/Ϫ or fibromodulin message in the Fmod Ϫ/Ϫ mice. Lum ϩ/ϩ Fmod ϩ/Ϫ mice showed approximately twice as much relative expression of lumican as Lum ϩ/Ϫ Fmod ϩ/Ϫ mice, indicating that the amount of lumican message is dependent on the Lum gene dosage. The relative expression of lumican in the wild type and fibromodulin heterozygous mice is more or less comparable. This suggests that the dramatic increase of lumican in the Fmod Ϫ/Ϫ genotype is not due to increased transcription but rather increased protein synthesis or stability in the collagen matrix. The real-time RT-PCR results for fibromodulin message are quite different from those of lumican in two respects: 1) there is no significant difference in the relative expression of fibromodulin between the Fmod ϩ/ϩ and Fmodϩ/Ϫ genotypes, indicating an absence of gene dosage effect, 2) there is no significant difference in Fmod expression between the Lum ϩ/ϩ and Lum Ϫ/Ϫ genotypes.
Collagen fibril structure in FDL tendons was examined by transmission electron microcopy in 5-month-old animals. Compared with wild type, the morphology of the collagen fibril is different in all of the mutants (Fig. 6, A-D). Wild type tendons have fibrils with smooth, circular profiles in cross-section. In contrast, the fibril contours become increasing irregular in the Lum Ϫ/Ϫ Fmod ϩ/ϩ , Lum ϩ/ϩ Fmod Ϫ/Ϫ , and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ tendons. Fibrils with cauliflower-shaped contours, consistent with abnormal lateral growth of fibrils, are clearly more frequent in the Lum ϩ/ϩ Fmod Ϫ/Ϫ and Lum Ϫ/Ϫ Fmod Ϫ/Ϫ tendons.
Fibril diameter distributions were analyzed in FDL tendons from wild type and the null mutant mice (Fig. 6, histograms). At least two distinct populations are discernible in the fibril diameter-distribution pattern of the wild type tendons: a population of small diameter (20-to 59-nm range, peak at 30 -39 nm) fibrils, and a second broad normal distribution of larger diameter fibrils (60-to 280-nm range, peak at 150 -159 nm). These two major populations of small and large diameter fibrils comprise 25 and 28% of the total population. The very large diameter (Ͼ220 nm) makes up an additional 5%. Although the single mutant tendons maintain this general distribution of the small and large diameter fibrils, there are key digressions from this pattern. First, in the Lum Ϫ/Ϫ Fmod ϩ/ϩ tendons, there is a trend toward increased fibril diameter. The frequency of small diameter fibrils is reduced to 18%, whereas the population of large diameter fibrils has increased to 38% of total, and the Ͼ220-nm fibrils make up 10% of the total. Second, in the Lum ϩ/ϩ Fmod Ϫ/Ϫ tendons, there is a general shift toward making smaller diameter fibrils. Thus, the small diameter fibrils make up 41% of the total analyzed, and the large diameter fibril population also shows a decrease in diameter; only 10% of the total is Ͼ160 nm with no fibrils in the Ͼ220 nm range. The Lum Ϫ/Ϫ Fmod Ϫ/Ϫ tendons show an overall flattening in the distribution pattern with less definition in the transition from the small to large diameter fibrils, due to the presence of additional subpopulations of fibrils. As in the wild type, the small diameter fibrils still make up about 25% of the total in the double-null mouse. However, the large diameter population has increased to 32% of the total, with 7% in the Ͼ220-nm range. This suggests a loss in regulation of processes that control fibril diameter in the normal tendon. DISCUSSION We have developed mice deficient in lumican and fibromodulin, two small leucine-rich proteoglycans, to elucidate their role and functional overlaps in collagenous connective tissues. The Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mice share many of the clinical features of EDS: reduced body size (short stature in certain EDS types), skin hyperextensibility, easy bruisability and myopia-like increased ocular growth, 3 premature cervical softening, 4 joint hypermobility, and articular cartilage degeneration. The Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mice, in histological examinations not pre-sented here, show additional abnormalities of the heart. These include increased vasculature of the myocardium and myofiber disorganization, but no signs of increased fibrosis (data not shown). In the double-null animals, the weight of the whole heart, normalized for body weight is also significantly lower than the wild type. Investigations of the cardiac phenotype are currently underway.
To understand the pathogenic changes underlying the joint hyperextensibility phenotype, the ultrastructure and biomechanical strength of the FDL tendons of wild type, single-null, double-null, and genotypes with various combinations of heterozygosity, were analyzed. This analysis identifies, for the first time, a major role for fibromodulin in regulating tendon structure and function and a gene dosage-dependent influence for lumican on tendon functions. The broad variation in joint hypermobility in connective tissue disorders and normal individuals may arise from such variations in tendon and ligament composition with respect to lumican, fibromodulin, and other connective tissue constituents. The fragile skin phenotype of the double-null mouse was noted in the lumican-deficient mouse as well, and results of its biomechanical testing were reported previously (17). Because the skin phenotype of the double-null mouse is similar to the Lum Ϫ/Ϫ mouse, it was not further investigated in this study.
The abnormal gait in the double-null mutants appeared in very young animals. In cross-section the knee appeared abnormal, with a misaligned patella, a secondary patellar groove, and compensatory overgrowth of the metaphyseal bone. The reduced whole tendon stiffness in the double-null mice may lead to alterations in load transfer from muscle to bone, which may explain the patellar misalignment to a certain extent. Assuming ligament properties are equally affected, the reduced stiffness may reduce passive stabilization of joints and increase knee laxity. Although the abnormal gait may be due in part to the weakening of the tendon, the early appearance of the structural defects in the knee-joint suggests disruptions in developmental cues. Recently, GDF-5-deficient mice were also reported to have gait abnormality, knee dislocations, and Achilles tendon weakness in normalized strength and stiffness tests, possibly in part due to developmental anomalies (34). Both lumican and fibromodulin are present in the developing tendon as early as postnatal day 4 (19). Thus, tendon development may be affected in the double-null mouse, and a poorly developed FDL tendon may be a partial cause for the gait abnormality.
The degeneration of the articular cartilage observed in the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ may be a natural consequence of abnormal use of the knee joint, because signs of degeneration were evident in older animals only. The Fmod Ϫ/Ϫ mice show an increased incidence of osteoarthritis in the knee joints where articular cartilage changes are evident at 36 weeks and become profound by the age of 80 weeks. 2 Weak ligaments may be the primary cause of the osteoarthritis in that model. Gait impairment, ectopic ossification, and premature osteoarthritis also occur in the biglycan-fibromodulin double-null mouse (35). Ectopic ossification and osteoarthritis in the biglycan-fibromodulin-deficient mice may result from forced use of the knee joints due to structurally weak tendons as in the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ model presented here.
We measured tendon biomechanical strength in the single, double-null, and heterozygous mice and found a dramatic reduction in tendon stiffness in the Lum Ϫ/Ϫ Fmod Ϫ/Ϫ mouse. An additional novel finding of this study is that fibromodulindeficiency alone results in significant reduction in tendon strength. Fibromodulin is expressed at high levels in the tendon, and the finding that its loss alone impacts tendon function emphasizes a key role for this proteoglycan in tendons. It is somewhat of a surprise that, in the Fmod-null genotype, Lum heterozygosity yields tendon strength intermediate between the Lum ϩ/ϩ and Lum Ϫ/Ϫ genotypes. This speaks of a function for lumican that is indispensable in the absence of fibromodulin. Although the relative amount of lumican, as compared with fibromodulin, is much lower in the normal tendon, it clearly plays an essential role in tendon maturation and strength.
To determine if the functional difference between Lum ϩ/ϩ Fmod Ϫ/Ϫ and Lum ϩ/Ϫ Fmod Ϫ/Ϫ is directly related to quantitative differences in the Lum gene product, we examined levels of lumican protein and RNA in the different genotypes. Indeed, the Lum ϩ/ϩ Fmod Ϫ/Ϫ tendons contain about twice as much lumican protein as the Lum ϩ/Ϫ Fmod Ϫ/Ϫ tendons. Moreover, excess protein appears to down-regulate transcription of Lum in the fibromodulin-deficient mouse tendon. The immunoblotting data indicate excess lumican protein, whereas the realtime RT-PCR data show a down-regulation of the lumican message. Svensson and coworkers (21) also have noted a 4-fold increase in lumican protein and a 95% decrease in its message in the tail tendons of Fmod Ϫ/Ϫ mice. These findings suggest that the SLRPs regulate their levels in tissues by some poorly understood mechanism. The increased deposition of lumican in the Fmod-null background may compensate for fibromodulin deficiency and prevent further loss in tendon strength in the Lum ϩ/ϩ Fmod Ϫ/Ϫ genotype.
How does mechanical strength relate to tendon connective tissue structure? Although a clear understanding at the molecular level still eludes us, proteoglycan bridges, which separate parallel arrays of collagen fibrils in the tendon, maintain hydration to resist and transmit tensile stresses (36). The analyses of tendon collagen fibril structure in the single-null and double-null mice point to certain interesting differences in how these two proteoglycans regulate collagen fibril assembly and the resultant tendon mechanical properties. Fibromodulin aids the tendon maturation process. In its absence there is an increase in immature small diameter fibrils. This phenomenon was also noted in our earlier study on developing tendons (19). The consequence of these small diameter collagen fibrils is likely to be a weak and immature tendon. By 5 months of age, there is not only a marked increase in small diameter fibrils, but also a loss in the number of large diameter fibrils, which may explain the increased functional impairment with age in the Lum ϩ/ϩ Fmod Ϫ/Ϫ animals. The compensatory overexpression of lumican (in a Lum gene dose-dependent manner) in the Fmod Ϫ/Ϫ background, may rescue the tendon phenotype to a certain extent. In the lumican-null alone (Lum Ϫ/Ϫ Fmod ϩ/ϩ ), there is an increase in large diameter fibrils as reported earlier for collagen structure of the cornea and the skin (17,20). At the protein level we observed a slight decrease in fibromodulin in FDL tendons of Lum Ϫ/Ϫ Fmod ϩ/ϩ mice. Obviously, the decrease in fibromodulin is not significant enough to enrich for small diameter immature fibrils and develop tendon weakness. The tendon phenotype (collagen structure and tendon strength) of the double-null mouse is severe, compared to that of the fibromodulin-null alone. Yet, in the double-null mouse there is no further increase in immature fibrils as compared with the fibromodulin-deficient mouse. There is, however, a slight increase in very large diameter fibrils, possibly due to the loss of lumican-associated regulation of fibril growth. Clearly, the relationship between tendon tissue strength and its ultrastructure is far more complex than the relative proportion of large and small diameter fibrils. Although this is a factor, as suggested by our study, there are other unrecognized biochemical and spatial organizational issues that help to define tissue strength. Mice deficient in tenascin X, a model for a recessive type of EDS, have collagen fibrils within the normal range of diameter in the skin, but their density is reduced. The functional consequence of such gaps and spatial changes in fibril packing is also reduced biomechanical strength and skin laxity (37,38).
The lumican-, fibromodulin-, and lumican-fibromodulin double-deficient mice provide a series of novel mouse models for EDS and other connective tissue disorders. The previous mouse models were developed by targeted mutations of genes encoding collagens and other enzymes that directly modify collagen chains (39 -41). The current study has demonstrated the importance of proteoglycans that modify the assembly and supramolecular structure of collagen fibrils as possible players in certain types of EDS and other phenotypically similar connective tissue disorders.