|Home | About | Journals | Submit | Contact Us | Français|
To characterize the in vivo role epiphycan (Epn) has in cartilage development and/or maintenance.
Epn-deficient mice were generated by disrupting the Epn gene in mouse embryonic stem cells. Epn/biglycan (Bgn) double-deficient mice were produced by crossing Epn-deficient mice with Bgn-deficient mice. Whole knee joint histological sections were stained using van Gieson or Fast green/Safranin-O to analyze collagen or proteoglycan content, respectively. Microarray analysis was performed to detect gene expression changes within knee joints.
Epn-deficient and Epn/Bgn double-deficient mice appeared normal at birth. No significant difference in body weight or femur length was detected in any animal at one month of age. However, nine-month Epn/Bgn double-deficient mice were significantly lighter and had significantly shorter femurs than wild type mice, regardless of gender. Male Epn-deficient mice also had significantly shorter femurs than wild type mice at nine months. Most of the deficient animals developed osteoarthritis (OA) with age; the onset of OA was observed earliest in Epn/Bgn-double deficient mice. Message RNA isolated from Epn/Bgn double-deficient knee joints displayed increased matrix protein expression compared with wild type mice, including other small leucine-rich proteoglycan (SLRP) members such as asporin, fibromodulin, and lumican.
Similar to other previously studied SLRPs, Epn plays an important role in maintaining joint integrity. However, the severity of the OA phenotype in the Epn/Bgn double-deficient mouse suggests a synergy between these two proteins. These data are the first to show a genetic interaction involving class Iand class III SLRPs in vivo.
Epiphycan, a chondroitin/dermatan sulfate proteoglycan preferentially expressed in epiphyseal cartilage during embryonic development , is a member of the small leucine-rich proteoglycan (SLRP) family. SLRPs are extracellular matrix (ECM) molecules that contain a tandem array of leucine-rich repeat (LRR) motifs flanked by N- and C- terminal cysteines. Many SLRP core proteins are also substituted with glycosaminoglycan (GAG) chain(s), although the presence, nature, and size of the GAGs appear to be tissue-specific and age-dependent .
SLRPs are divided into three classes based upon the spacing of their N-terminal cysteines, the number of LRRs they contain, and their corresponding gene structures. Class I (decorin, biglycan, and asporin) and class II SLRPs (fibromodulin, PRELP, keratocan, lumican, and osteoadherin/osteo-modulin) contain 10–12 LRRs, whereas class III SLRPs (epiphycan/PG-Lb/DSPG3, opticin, and osteoglycin/mimecan) contain 6–8 LRRs. Chondroadherin, NYX/nyctalopin, ECM2, podocan, and nephrocan contain both the N-terminal cysteine cluster and the LRR domain characteristics of SLRPs but are not assigned to a class.
SLRPs interact with a variety of ECM proteins. Reported binding partners include fibronectin , thrombospondin , fibrinogen , heparin cofactor II , C1q [7–9], mannose-binding lectin , collectin 43 , conglutinin , pulmonary surfactant D , and TGF-β [12, 13]. The most commonly reported binding partner for SLRP proteins is collagen . Moreover, the targeted disruption of individual SLRPs can lead to obvious collagen fibril formation defects in vivo. However, single SLRP-deficient mice display relatively mild phenotypes when considering their tissue distributions; an observation that suggests some SLRPs may compensate for the loss of others. For example, lumican-null mice exhibit corneal clouding and skin laxity, but do not show significant alterations in tendon stiffness . However, lumican loss-of-function enhances the reduced tendon stiffness seen in fibromodulin-null mice, suggesting that lumican compensates for the loss of fibromodulin . This putative genetic interaction is particularly interesting in light of the fact that fibromodulin and lumican bind to a similar site on collagen .
Many different SLRPs are expressed in cartilage, including asporin, biglycan, chondroadherin, decorin, epiphycan, fibromodulin, keratocan, lumican, opticin, and PRELP [18–23]. However, the ability of SLRPs to influence OA progression is not understood. Furthermore, increased SLRP degradation has been detected in human osteoarthritic cartilage samples . Such degradation presumably leads to cartilage matrix destabilization. Recent reports have implicated one particular SLRP, asporin, with hip and knee OA in Asian populations [25, 26]. A specific human polymorphism in the asporin gene, resulting in an N-terminal region that contains 14 aspartic acid residues rather than 13 aspartic acid residues, is significantly associated with individuals that have OA.
Very little is reported about class III SLRPs, particularly epiphycan. Epiphycan (Epn) expression appears to be restricted to cartilage and testis . The unique tissue expression of Epn in chick  and mouse  cartilage suggests that this SLRP may play a key role in chondrocyte differentiation and, hence, cartilage stability. To further investigate this possibility, we generated and characterized Epn-deficient mice. These mice exhibit a very mild skeletal phenotype. Since biglycan (Bgn) is also expressed in mouse cartilage, and may compensate for the loss of Epn, Epn/Bgn double-deficient animals were also generated and characterized.
An 8 kb fragment of the 129S6 mouse Epn gene was identified with a rat Epn probe and used to construct the targeting vector, which contained a LacZ expression cassette followed by a neomycin (neo)-resistance gene flanked by loxP sites. A MC-1 thymidine kinase cassette was included immediately outside the homologous mouse sequence of the targeting vector.
Mouse AB2.2 embryonic stem cells (ESC; a gift from Allan Bradley) were electroporated with the linearized targeting vector described above using a BioRad Gene Pulser and grown in M15 media containing G418 to select for cells containing the neo-resistance gene. The targeting vector was designed to insert the LacZ cassette six base pairs downstream of and in frame with the starting ATG codon. ES clones containing the correct insertion were identified by Southern blot hybridization and injected into C57BL/6J albino blastocysts according to standard procedures. High percentage coat-color-contribution chimaeras were crossed with C57BL/6J albino mice to obtain germline transmission of the targeted allele. Southern blot and PCR analyses were used to genotype the mice.
Double heterozygous Epn/Bgn mice (F1 generation) were produced by intercrossing homozygous Bgn-deficient females (Epn+/+Bgn−/−) with homozygous Epn-deficient males (Epn−/−Bgn+/0). The production and characterization of Bgn-deficient mice was previously described . Epn/Bgn knockout mice (F2) were obtained by intercrossing F1 double heterozygous Epn/Bgn mice. All of the mice were maintained in a hybrid 129S6:C57BL/6J albino genetic background. Each animal experiment was approved by the Institute of Biosciences and Technology Institutional Animal Care and Use Committee.
Mouse ESC genomic DNA was digested with ScaI or BglII. Southern blot hybridization was performed according to the procedure of Ramírez-Solis et al.  with radiolabeled 5’ or 3’ probes that hybridized to regions of the Epn gene lying outside of the targeting vector homology.
The PCR scheme for the Bgn allele was performed as described previously . To identify the Epn genotype of each mouse, the following primers were used in separate PCR reactions: a forward primer (Epn1fw) corresponding to sequences in Epn exon II (5′-GGTCAGGGGCAAATACCAAGGACTCT), a reverse primer (Epn2rv) corresponding to sequences in Epn exon III (5′-CTCTACATGG TTGTCAGGAATGTG), and a second forward primer (Bpa) corresponding to a sequence contained within the bovine growth hormone polyadenylation signal sequence of the neo cassette (5′-GCTTCTGAGGCGGAAAGAACCAGCTA). The amplified PCR product for the wild type Epn allele with primer pair Epn1fw/Epn2rv was 2.6 kb, whereas the product for the disrupted Epn allele with primer pair Bpa/Epn2rv was 0.4 kb.
The body weights and femur lengths of each mouse was measured at one and nine months of age. Femur length was measured under a dissecting microscope using calibers. The data from 13–40 different mice per genotype of each gender were analyzed using the Kruskal-Wallis test followed by a Dunn’s multiple comparison post-test.
All tissues were fixed in either 4% paraformaldehyde or 10% formalin, decalcified at room temperature in 0.5M EDTA for ≥ one week, dehydrated in ascending concentrations of alcohol, and paraffin-embedded. Serial sections were cut from the paraffin block using a microtome and thereafter dewaxed and rehydrated. For histological analysis of collagen content, tissue sections were stained with hematoxylin for 2 min followed by immersion in van Gieson stain for 5 min. Remaining sections were stained with hematoxylin for 2 min, briefly differentiated in acidified alcohol, and stained with 0.2% Fast green for 2 min. After washing with 1% acetic acid, the sections were stained with Safranin-O for 3 min to examine tissue morphology and proteoglycan content.
Immunohistochemical staining was performed using polyclonal antibodies that detect either mouse Bgn (LF-159) or mouse Epn (R561). The rabbit polyclonal anti-Bgn antibody, LF-159 , was a gift from Dr. Larry Fisher (NIH/NIDCR, Bethesda, MD). The polyclonal anti-Epn antibody, R561, was raised in rabbits against purified recombinant Epn protein (Ala18-Thr111 fused at the C-terminus to glutathione S-transferase; Alpha Diagnostics International, San Antonio, TX). R561 antibody characterization demonstrated that it is specific for Epn. Tissue sections were digested with 4 mg/ml hyaluronidase in phosphate buffered saline (PBS) at 37°C for 30 min prior to immunostaining. The sections were then incubated with R561 (1:1000) or LF-159 (1:2000) overnight at 4°C. Antibody-staining was visualized with the Vectastain Elite ABC immunoperoxidase kit (Vector Laboratories, Hercules, CA). Hematoxylin was used as a counterstain.
Serial sagittal histological sections of individual mouse knee joints were examined for signs of OA in a double-blind fashion. The slides were scored for OA severity used a grading system based upon the criteria of Maier and Wilhelmi . This system uses the following scale to quantify OA severity: grade 0 indicates no apparent changes, grade 1 indicates slight fibrillation of the articular cartilage, grade 2 specifies defects limited to the uncalcified cartilage, grade 3 specifies defects penetrating into the calcified cartilage, and grade 4 indicates subchondral bone exposure at the joint surface. Scores for each genotype were averaged and analyzed using an unpaired Student’s t-test with Welch’s and Bonferroni corrections.
Males were chosen for these experiments because they displayed increased joint disease severity as compared to females. Epn/Bgn double-deficient and wild type mice were sacrificed at three months of age (n = 3 for each genotype). The tibial-femoral joints of each male mouse were harvested by dissecting the joint immediately above the femoral condyles and immediately below the tibial plateau, measuring approximately 1 cm in length. Skin, tendons, and ligaments around the joint were removed for reproducibility of the procedure. The joint tissue was then immediately frozen in liquid nitrogen and pooled for RNA isolation (for a total of 6 joints per genotype). Total RNA was isolated using TRIzol®-mediated organic extraction (Invitrogen, Carlsbad, CA) followed by purification with RNeasy mini columns (Qiagen, Valencia, CA) according to the manufacturer’s instructions. RNA integrity was assessed using formaldehyde gel electrophoresis.
Second-strand cDNA was generated from 5 μg of joint tissue RNA using an oligo(dT)-T7 promoter primer and SuperScriptTM II reverse transcriptase (Invitrogen, Carlsbad, CA). cRNA was amplified from the cDNA and biotinylated using the BioArrayTM High YieldTM RNA Transcript Labeling Kit (ENZO, Farmingdale, NY) according to the manufacturer’s instructions. After purification, labeled cRNA was quantified using a spectrophotometer and fragmented in 200 mM Tris/acetate (pH 8.1), 500 mM potassium acetate, and 150 mM magnesium acetate for 35 min at 94°C. The fragmented cRNA was then hybridized overnight to the Mouse Genome 430A 2.0 GeneChip (Affymetrix, Santa Clara, CA). The resultant signal was detected according to the manufacturer’s instructions. This procedure was repeated three times for each genotype; a different cRNA preparation was used for each hybridization event. All six arrays were analyzed using the dChip software package (www.dchip.org). Each array had a different level of overall brightness; the dChip program was therefore used to normalize arrays to the median intensity level using an invariant-set normalization method . After normalization, expression values were calculated using a perfect match model . The program parameters were set such that expression values at least 1.25-fold and an absolute difference of more than 100 between wild type and double-deficient samples were detected.
The Epn locus was disrupted in mouse embryonic stem cells by inserting a targeting vector into exon III downstream of the endogenous Epn translation start codon (Fig. 1A). Southern blot analysis verified that the Epn locus was correctly targeted, generating bands of the predicted fragment size (Fig. 1B and C). The genotypes of the mice were confirmed using PCR (Fig. 1D). Homozygous Epn-deficient mice were born with the expected Mendelian ratio, appeared healthy at birth, and were fertile. LacZ activity was not detected in the mutant embryos for unknown reasons.
Epn protein was undetectable in Epn-deficient mice at any age (Fig. 2A). In contrast, Epn protein was observed in wild type newborn mice within the ECM of all cartilaginous zones except the cartilage immediately adjacent to the joint cavitation space and the perichondrium (data not shown; ). At two weeks of age, Epn was detected in the deep layer of articular cartilage of wild type mice, but not within the superficial layer (Fig. 2B). Epn was also detected in growth plate cartilage. Very weak Epn staining was observed in the growth plate cartilage of two month old wild type mice (Fig. 2C). Epn protein was not detected in wild type mice older than two months (data not shown). The distribution of Epn protein in Bgn-deficient mice was similar to that described above for wild type mice (data not shown). Control sections from wild type mice where the anti-Epn antibody (R561) was omitted from the staining procedure showed no background (data not shown).
The localization of Epn protein within the tibial-femoral joint strongly resembles the localization of another SLRP, biglycan (Fig. 2D). Since biglycan genetically interacts with another SLRP molecule, fibromodulin , Epn-deficient mice were bred with biglycan (Bgn)-deficient mice to test whether biglycan genetically interacts with epiphycan. Similar to Epn-deficient mice, Epn/Bgn-deficient mice were born in the expected Mendelian ratio, appeared healthy at birth, and were fertile.
Due to the fact that Epn and Bgn are expressed in the cartilage growth plate of long bones, we examined the body weights and femur lengths of our SLRP-deficient mice. No significant difference in body weight was detected between the different genotypes at one month of age, regardless of gender (Fig. 3). At nine months, the body weights of male Epn-and Bgn-deficient mice were also similar to their wild type littermates. However, nine month old male Epn/Bgn double-deficient mice were significantly lighter by 17.1% (p < 0.001). No significant difference in body weight was detected between nine month old wild type and Epn-deficient females. However, the body weights of both Bgn- and Epn/Bgn-deficient females were 16.3% lighter than their wild type littermates at nine months (p < 0.001).
Next, the ability of Epn and/or Bgn to influence the growth of long bones was examined. The average femur length of one month old Epn-, Bgn-, or Epn/Bgn double-deficient mice was not significantly different than the average femur length of wild type mice, regardless of the gender. However, the femur lengths of nine month old male Epn- and Epn/Bgn double-deficient mice were significantly shorter than their wild type littermates by 3.9% (p<0.05) and 7.1% (p<0.001), respectively. Male Bgn-deficient femurs were not significantly shorter than wild type at this age. For females, only Epn/Bgn double-deficient femurs were significantly shorter than wild type at nine months (by 7%, p< 0.001).
To further assess joint tissue morphology and proteoglycan content, histological sections of tibial-femoral joints from four nine month old animals per genotype were characterized by hematoxylin/eosin and Safranin O staining. Histological evidence of premature OA was not observed in joint sections from wild type mice (Fig. 4A). However, osteoarthritic signs were observed in two of the four Epn-deficient mice, including articular surface fibrillations and significant proteoglycan loss (Fig. 4B). Similar observations were made in three of the four Bgn-deficient mice examined (Fig. 4C). The Bgn-deficient mice also showed evidence of ligament and tendon ossification (data not shown), in agreement with Ameye et al. . The most severe OA-like phenotype was observed in Epn/Bgn double-deficient animals (Fig. 4D); joint sections from all four animals showed damaged articular cartilage that was often completely eroded to the underlying bone, which was typically sclerotic. Epn/Bgn double-deficient mice showed increased osteophyte formation within and around the joint space. Increased ligament and tendon ossification was also detected in these mice (data not shown).
To quantify the observed OA-like changes detected in our SLRP-deficient mice, histological sections of tibial-femoral joints from male mice of each genotype at various ages were examined in a double-blind fashion and scored using the OA grading system devised by Maier and Wilhelmi . Compared to wild type mice, a number of the Epn- and Bgn-deficient mice began to display OA-like changes at six and four months of age, respectively (Fig. 5). Epn/Bgn double-deficient mice appeared to develop OA-like characteristics even sooner, as early as three months. Additionally, the OA grade in double-deficient mice was highly significant at six months, three months earlier than Bgn-deficient mice. Epn-deficient mice displayed OA-like changes that were only of borderline significance at nine months of age. Significant OA-like changes were also detected in female Epn/Bgn double-deficient mice, although these were less severe (data not shown).
Since OA is associated not only with aggrecan but also collagen loss, we examined histological sections of male tibia-femoral joints for total collagen content using van Gieson stain at two, three, and six months. A difference in collagen staining when compared to wild type mice was detected within the superficial zone of articular cartilage in Bgn-deficient and Epn/Bgn-deficient mice as early as three months (Fig. 6). Collagen loss consistently appeared to be more pronounced at the articular surface of Epn/Bgn double-deficient mice when compared to Bgn-deficient mice (Fig. 6 inset). This was accompanied by a deficiency of superficial chondrocytes. Epn-deficient mice showed similar collagen staining when compared to wild type mice data not shown).
To examine what happens at the molecular level before significant OA-like morphological changes occur, we compared the mRNA profile of whole knee joints from three month old wild type and Epn/Bgn double-deficient male mice. We chose to analyze changes in the whole joint for two reasons: (1) OA is a disease of the whole joint and (2) In our experience, it is extremely difficult to obtain mouse cartilage samples that are not contaminated with other tissues, such as subchondral bone.
We detected a significant increase in the mRNA of 49 separate genes within Epn/Bgn double-deficient mice as compared to wild type mice. A total of 7 specific mRNAs were significantly decreased at this time point in the disease progression; the largest decrease detected was for Bgn (further confirmation that the mice were Bgn-deficient; Epn is not expressed at this age in wild type mice). The majority of mRNAs that were increased in double-deficient mice corresponded to secreted proteins. Of the genes encoding ECM proteins, the α1 chain of type III procollagen showed the largest increase (Table 1). The metalloproteinases MMP-2 and MMP-14, as well as the serine protease HtrA1, were also increased in Epn/Bgn double-deficient mice. Interestingly, three other SLRPs (asporin, fibromodulin, and lumican) were significantly up-regulated in double-deficient mice. A complete description of all the gene changes can be found in the NCBI Gene Expression Omnibus (GSE9099).
We previously showed that Epn has a highly restricted expression pattern within cartilage , suggesting that it plays a role in skeletal formation and/or maintenance. The generation of Epn-deficient mice demonstrated that although Epn is not essential for embryonic skeletal development, it is critical for proper long bone growth and joint maintenance. For example, Epn significantly influences femur length in male mice. This effect is presumably due to subtle defects in the growth plate since Epn is not expressed in bone. Furthermore, Epn is necessary for retaining cartilage integrity with age because Epn loss, in the absence of Bgn, significantly accelerates the rate of OA-like lesion development when compared to their single SLRP-deficient littermates. The rapid progression of OA in the double-deficient mouse suggests one of two possibilities: (1) Epn and Bgn have some overlapping functions within cartilage or (2) Epn and Bgn have independent functions that, when deleted, exacerbate the progression of OA. Future work will be necessary to demonstrate which possibility is correct. Nevertheless, our data are the first to show an in vivo genetic interaction involving class I and class III SLRPs.
Our collagen content analysis in the tibia-femoral joints of SLRP-deficient animals demonstrated that collagen loss occurred at the articular cartilage surface in both Bgn-deficient and Epn/Bgn double-deficient mice. However, collagen loss appeared to be greater in Epn/Bgn double-deficient animals. These data imply that Epn influences the stability of the collagen matrix in a Bgn- deficient background, again suggesting that Bgn and Epn share a function in maintaining joint integrity. Collagen loss in mutant animals preceded significant OA-like changes as measured by the Maier and Whilhelmi OA grading scale and may therefore be responsible for increasing their susceptibility to develop OA. In human osteoarthritic cartilage samples, collage degradation  and loss of collagen tensile strength  is detected early in the disease. Interestingly, increased collagen degradation has been observed either before  or at a similar time as proteoglycan loss . Future studies to investigate whether this apparent loss of collagen staining is due to increased collagen degradation are underway.
Epn/Bgn double-deficient mice are not the only SLRP-deficient mice that develop OA-like changes. Previous work has shown that Bgn-deficient , fibromodulin-deficient , Bgn/fibromodulin double-deficient , and lumican/fibromodulin double-deficient mice  also develop an OA-like disease. In all of the cases examined thus far, the double-deficient mice develop OA earlier than their single SLRP-deficient counterparts, demonstrating the necessity for multiple SLRPs to maintain joint integrity. The SLRP-mutants described above also show tendon abnormalities that likely contribute toward OA development and/or progression.
We currently do not know why the absence of a protein that is apparently expressed only early in life, such as Epn, would show effects later in life. However, there are several possible explanations. Firstly, Epn could be required for stabilizing the cartilage matrix through its ability to influence collagen fibrillogenesis, matrix assembly, or chondrocyte-ECM interactions. The fact that other SLRPs are known to influence collagen fibril formation and our collagen content analysis support this possibility. Secondly, Epn may influence one or more currently undefined cell signaling events that are necessary for proper chondrocyte/cartilage maintenance. This idea is supported by the recent work of Schaefer et al. , where Bgn was shown to signal through TLR-2 and TLR-4. Thirdly, Epn may be required for repairing localized joint damage. In this scenario, a currently undetectable, localized expression of Epn by chondrocytes possibly helps to repair matrix defects that accumulate throughout life. Fourthly, Epn may directly or indirectly influence the structure of the subchondral bone matrix. Epn loss would therefore lead to cartilage degeneration because of the altered subchondral bone structure. The answer may also be a combination of the possibilities listed above.
We also do not know the reason behind the gender differences reported herein. For example, femurs are significantly shorter in male but not female Epn-deficient mice when compared to wild type mice. Sex hormone and/or body weight differences may play a role. However, future work is needed to fully understand these effects. Generally speaking, gender differences in genetically-manipulated mice are not well studied, although there a few examples reported in the literature (e.g. Bgn/Fibromodulin-deficient mice  and estrogen receptor-β knockout mice ).
Interestingly, our microarray analysis of Epn/Bgn double-deficient mice detected increased ECM gene expression that can also be seen in human OA samples [43–46]. These include the mRNAs of the type III collagen α chain [43, 45], matrix metalloproteinase 2 [43, 45], fibronectin [43, 45, 46], lumican , and fibromodulin . Furthermore, we detected increases in mRNAs that are known to be increased at the protein level in human osteoarthritic cartilage, including type VI collagen protein , annexin VIII protein , and insulin-like growth factor 1 protein . However, our microarray data also suggests that superoxide dismutase 3 is increased in Epn/Bgn double-deficient tissue; this protein has actually been shown to be decreased in human osteoarthritic cartilage and in a naturally occurring OA mouse model, STR/ort . This discrepancy may be due to the fact that the tissues are not identical (whole joints vs. cartilage alone).
Of interest, the mRNA levels of three different SLRPs (asporin, fibromodulin, and lumican) are increased in Epn/Bgn double-deficient mice as compared with wild type mice. These data suggest that the expression of various SLRPs are somehow interrelated and may be further proof of the ability of SLRP molecules to compensate for each other.
Taken together, the data presented herein suggests that Epn/Bgn double-deficient mice may be a useful tool for the study of OA. The tibia-femoral joints from Epn/Bgn double-deficient mice show similar morphological changes to human OA joints, including articular cartilage erosions, proteoglycan loss, subchondral bone sclerosis, and osteophyte growth. While our deficient mice are not the only OA murine model to be proposed, their usefulness lies in the fact that significant characteristics of the disease are detected fairly quickly (by 6 months) with complete penetrance. Furthermore, our model has the potential to shed light on how SLRPs influence the molecular events that lead to OA. The fact that many different genetic and physical manipulations in mice cause an OA-like phenotype suggests that joint health is maintained by a variety of factors. Since human OA is thought to be a multifactorial disease, we believe that the use of multiple murine models, including SLRP-deficient mice, is necessary to best understand the disease progression.
This work was supported by NIH grants P01AR042919 and R01AR047433.
CONFLICT OF INTEREST STATEMENT
No conflicts of interests are associated with this manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.