Synovial joint formation initiates with the appearance of the interzone, a layer of densely packed mensenchymal cells that delineate the anatomical boundaries between adjacent cartilaginous skeletal elements. Removal of the interzone leads to joint ablation and fusion, consistent with the idea that cells residing within the interzone provide the progenitors for joint morphogenesis (Holder, 1977). Lineage tracing studies have confirmed that cells within the interzone give rise to tissues common among synovial joints including the articular cartilage surfaces of individual bones, the intra-joint ligaments that connect bones to each other, and the synovial membrane, a thin layer of tissue lining the joint space that controls the environment within a joint (Koyama et al., 2008).
Research devoted to deciphering the molecular mechanisms controlling interzone formation have identified the bone morphogenetic protein (BMP), transforming growth factor-beta (TGFβ), and Wnt signaling pathways as mediators of joint development. Regulated BMP signaling appears to be of fundamental importance, as both too much and too little BMP activity result in synovial joint anomalies. One of the earliest interzone markers is Gdf5 (growth differentiation factor 5), a secreted signaling molecule of the BMP family. Mice carrying an inactivating mutation in Gdf5 (brachypodism mice) display joint fusions in the forelimb and hindfeet (Storm and Kingsley, 1996). Noggin, a secreted BMP antagonist, is required for joint formation upstream of Gdf5. Noggin knockout mice fail to form joints, and lack Gdf5 expression in the interzone (Brunet et al., 1998). TGFβ activity is also necessary for joint morphogenesis and for the induction of tendon and ligaments within the joint, possibly by regulating GDF5 production by interzone cells (Seo and Serra, 2007; Spagnoli et al., 2007; Pryce et al., 2009). Ectopic expression of Wnt9a (previously called Wnt14), another early marker of the interzone, can direct ectopic joint formation in developing mouse and chick limbs (Hartmann and Tabin, 2001; Guo et al., 2004). However, Wnt signaling is not required for the induction but for the subsequent maintenance of the fate of joint interzone cells by actively suppressing their chondrogenic potential to allow joint formation (Spater et al., 2006). The interaction of these signaling pathways appears to be coordinated within the interzone by heparin sulfate proteoglycans to assure proper joint morphogenesis (Mundy et al., 2011). Finally, muscle movement and the biophysical stimuli generated by muscle contractions also play a fundamental role in joint development, critical for cavitation and maintenance of the joint space (Kahn et al., 2009).
While the early events of synovial joint formation have been thoroughly examined, information about later events such as how individual synovial joints acquire the unique structures crucial for their specific functions is currently unavailable. For example, the knee contains a meniscus and cruciate ligaments, highly specialized structures that are not found in the elbow or other synovial joints. While development of the meniscus and cruciate ligaments requires interzone formation, the presence of these elements only in the knee suggests that the knee interzone is fundamentally distinct from that of other synovial joints. To identify factors involved in the formation of individual joints, gene expression of the knee was compared with that of the elbow at embryonic day (E) 15 and E16 using microarray analysis. We chose the elbow because it forms at the same time as the knee but does not support meniscal or cruciate ligament morphogenesis. We reasoned that by subtracting genes expressed in these two highly specialized joints, we would eliminate genes that are common for the development of all limb joints and would be able to identify genes specific to the elbow and the knee. Our microarray results demonstrate that the knee joint has a unique gene expression signature by E15, with substantial enrichment for TGFβ signaling in knee joint development. Localization of the knee-enriched gene Tgfbi confirms that this TGFβ pathway gene is expressed in the joint structures found only in the knee. In comparison, our data suggests that morphogenesis of the elbow joint is regulated in part by muscle movement as evidenced by the high levels of expression of muscle differentiation genes in the tissue surrounding the elbow, and the phenotype of the splotch-delayed (muscleless) mouse. Our results provide a valuable resource for further characterization of genes and signaling pathways that determine the morphogenesis of individual joints and may be useful in analyzing the pathogenesis of disease and injury in specific joint tissues.
To visualize differences in how the elbow and knee joints form, we carefully examined the development of these joints from E14 to E16 (Fig. 1). At E14, despite the relative advance in forelimb development compared with hindlimb development, the elbow and the knee joints appear similar histologically. Both have defined interzones that mark the future site of each joint (Fig. 1A,D). Between E14 and E15, each joint undergoes a distinct transformation. First cavitation occurs, creating the synovial cavities that give each joint its unique shape (Fig. 1B,E,G). In the knee, a mesenchymal cushion (the presumptive meniscus) develops between the femur and tibia, and by E15 cells have clustered to form triangular meniscal condensations (Fig. 1E). At this stage, the meniscus is populated by cells that have a Col2a1+ expression history similar to the skeletal progenitor population, and also contains a Col2a1− population of cells that migrate into the knee joint (Hyde et al., 2008). At E15.5, the cruciate ligaments, derived from Sox9-expressing precursors originating in the mesenchymal condensations (Soeda et al., 2010), form in the center of the knee joint, connecting the femur to the tibia (Fig. 1G). Then at E16, the meniscus becomes regionalized into distinct inner and outer zones (Fig. 1F). While these specialized structures are forming in the knee, the elbow joint refines its specific shape, but does not develop additional elbow-specific tissues (Fig. 1B,C). Based on these histological observations, we chose E15 and E16 as the best time points to compare gene expression in the knee and elbow for the purpose of identifying factors involved in joint specific morphogenic events.
To identify genes that may distinguish knee joint formation from that of the elbow, we chose to compare gene expression profiles of each joint at E15 and E16. These stages were chosen because they appear to be the critical time window when the structure and organization of these two joints diverge (see Fig. 1). Knee and elbow joint regions including the epiphysis and intervening tissue were dissected from E15 and E16 mouse limbs (Fig. 2), and RNA isolated from these tissues was analyzed using Illumina WG-6 microarrays (Fig. 2). Genes with fold changes of approximately 0.5–1.5 were discarded as they represented transcripts common to both joints including the known chondrocyte markers Col2a1, Sox5, Matn1 (data not shown; Supp. Table S1, which is available online).
At E15, 23 genes showed a greater than two-fold enrichment in the knee (Table 1) while at E16, 9 genes fit this criteria (Table 2). Five genes were found to be enriched in the knee at both time points; Hoxc10, Tbx4, Meox1, Rspo2 and Sox9. Quantitative reverse transcriptase-polymerase chain reaction (qPCR) analysis of these genes verified their differential expression in knee compared with elbow, confirming our microarray results (data not shown; Fig. S1). Genes enriched in the knee at E15 and E16 fall into several functional categories, including patterning genes and signaling components of the TGFβ/BMP and Wnt pathways (Tables 1 and 2). Tbx4, Hoxc10, and Pitx1 are transcription factors known to be involved in hindlimb patterning (Peterson et al., 1994; Nelson et al., 1996; Minguillon et al., 2005), while other genes such as Rspo2 and Sfrp2, or Bmper and Tgfbi modulate Wnt signaling and BMP/TGFβ signaling, respectively. While the data set of E15 knee enriched genes contained Gdf5 and Sox9, factors known to play important roles in joint development, and the cell surface proteins Tlr2 and Itga11, many of the genes we found enriched in the E15 and E16 knee have not been previously described as having a role in joint development.
Table 1. Genes With >2-Fold Expression Difference in the Knee vs. Elbow at Embryonic Day 15
Growth factors and regulators
Functional classification (GO Term)
Multicellular organismal process
Transcription factor activity
Skeletal system development
Transpath pathway analysis
Inhibitory phosphorylation of TGFBRII
TGF-beta – Smad2/3
Table 2. Genes With >2-Fold Expression Difference in the Knee vs. Elbow at Embryonic Day 16
Growth factors and regulators
To give context to our gene list, we performed bioinformatics analysis using BioBase Explain 3.0 software and identified the functional categories and signaling pathways that were enriched in the list of knee-specific genes (Table 1). Through this approach, we found enrichment for TGFβ signaling in the E15 knee joint compared with the elbow joint. Consistent with this finding, Pryce et al. (2009) reported that knee joint development in Tgfbr2Prx1cre cKO mice is abnormal and that the knee lacks cruciate ligaments. Our further examination of these knees showed that they also lack menisci, suggesting that in addition to its role in cruciate ligament development, TGFβ signaling likely plays a role in meniscus specification and formation (Fig. 3). The elbow joints of Tgfbr2Prx1cre cKO mice develop normally (Seo and Serra, 2007), further supporting our finding that knee, but not elbow, development requires TGFβ signaling.
Next, we performed expression analysis of the knee-enriched gene set identified from our microarray data. We focused on three categories: genes with the greatest differential expression, genes known to be involved in limb patterning and skeletal development, and genes whose expression pattern had not been previously described in the knee joint. These criteria allowed us to shorten our list to focus on Hoxc10, Tbx4, Sfrp2, Rspo2, and Tgfbi. To distinguish the components of the joint, we localized the expression of known musculoskeletal markers at E16. As shown in Figure 4A, the chondrocyte marker Col2a1 is present in the developing cartilage of both the elbow and knee but is not found in either of the knee-specific structures, the meniscus or cruciate ligaments. We observed strong expression of the ECM glycoprotein Tenascin-C in the articular cartilage and tendons of the elbow and knee at E16, and also detected a significant level of Tenascin-C in the meniscus and cruciate ligaments (Fig. 4B). The hindlimb-specific genes Tbx4 (Minguillon et al., 2005; Hasson et al., 2010), and Hoxc10 (Peterson et al., 1994; Nelson et al., 1996) were then examined. Tbx4 was detected in proliferating, prehypertrophic, and hypertrophic chondrocytes of the femur and tibia but was not seen in the articular cartilage, meniscus, or cruciate ligaments (Fig. 4C). We detected Hoxc10 expression in the connective tissue of the hindlimb including the cruciate ligaments and the patellar tendon, but not in the meniscus or cartilage (Fig. 4D).
Next, we examined the expression pattern of knee-enriched genes that are involved in canonical signaling during skeletal development. We first looked at Tgfbi, a TGFβ-inducible gene that is detected in many embryonic tissues including E14.5 cartilage but whose role in development is unknown (Skonier et al., 1992; Schorderet et al., 2000). In the E16 knee, expression of Tgfbi was found in prehypertrophic chondrocytes, articular cartilage, inner meniscus, and ligaments (Fig. 4E) consistent with the known requirement for TGFβ signaling in knee development (Pryce et al., 2009) (Fig. 3). Expression of two modulators of the Wnt signaling pathway, that is, Sfrp2, a Wnt signaling inhibitor (Leimeister et al., 1998; Morello et al., 2008), and Rspo2, a Wnt signaling activator (Kazanskaya et al., 2004; Nam et al., 2007), was analyzed next. In the E16 knee, Sfrp2 was expressed in the articular cartilage, meniscus, and cruciate ligaments. In the meniscus, Sfrp2 localized to the inner and middle regions (Fig. 4F). Rspo2 was also found in the cruciate ligaments and in the middle-inner region of the meniscus but was not detected in the articular cartilage (Fig. 4G). Expression of each of the knee-enriched genes was also examined in the E16 elbow and found to be absent (Fig. 4).
When we examined comparable microarray data from the elbow, the majority of genes that showed a greater than two-fold enrichment at E15 and E16 were muscle-specific genes. At E15, 32 genes showed a greater than two-fold enrichment in the elbow (Table 3), while at E16, there were 37 genes meeting this criteria (Table 4). Sixteen genes were common to both lists. Our bioinformatics analysis revealed an over-representation of genes with the functional classification of myofibrils, contractile fibrils, and muscle organ development as well as enrichment for the creatine synthesis and degradation pathway (Table 3). Unexpectedly at E16, both the epidermal; growth factor (EGF) and platelet-derived growth factor (PDGF) signaling pathways are enriched (Table 4). Although there is no reported role for EGF signaling in muscle development or function (Threadgill et al., 1995), PDGF signaling is required for proper myotome formation (Soriano, 1997).
Table 3. Genes With >2-Fold Expression Difference in the Elbow vs. Knee at Embryonic Day 15
Functional classification (GO Term)
Muscle organ development
Transpath pathway analysis
Creatine biosynthesis and degradation
Creatine, phosphocreatinine ---> creatinine
Table 4. Genes With >2-Fold Expression Difference in the Elbow vs. Knee at Embryonic Day 16
Growth factors and regulators
Functional classification (GO Term)
Muscle organ development
Transpath pathway analysis
EGF ---> ERK
PDGF A/B ---> ERK
We used in situ hybridization to verify the preferential expression of muscle genes in the developing elbow joint. Our analysis of these genes revealed that they were expressed in the muscle tissue surrounding the joint. The slow muscle marker Car3 (Kim et al., 2004) and Csrp3, a marker of differentiated striated muscle (Arber et al., 1997) were found in the triceps and pronator muscles that insert into the ulna and are necessary for motion of the forelimb (Fig. 5C,D). Expression of the enzyme Ckm, a key regulator of energy metabolism in muscle (van Deursen et al., 1993), was also found in these structures (Fig. 5E). In the E16 knee, Car3 was not detected; however, Ckm and Csrp3 expression could be seen in the gastrocnemius muscle that attaches to the femur and is involved in movement of the hindlimb (Fig. 5C–E). When we examined the localization of the homeobox transcription factors Pitx2, a marker of muscle progenitors required for activation of MyoD (Shih et al., 2007; L'Honore et al., 2010), and Hoxa7, a gene involved in patterning of the upper thoracic region (Chen et al., 1998), we found these to be expressed in the muscle cells surrounding the elbow as well as in the cartilage of the elbow joint itself (Fig. 5F,G). This localization pattern was not seen in similar structures in the knee (Fig. 5F,G). When we looked for genes unrelated to muscle specification, we found that Dusp26, a phosphatase thought to regulate p53 action (Shang et al., 2010), was restricted to the cartilaginous elements of the elbow joint, whereas its expression was not detected in the hindlimb cartilage of the knee (Fig. 5H).
Given that our microarray analysis suggested that myogenic factors may be necessary for elbow joint formation but not required for knee development, we analyzed the elbow and knee joints of a muscleless mouse mutant, Spd (splotch-delayed; Tremblay et al., 1998). Mice homozygous for the Spd mutation fail to form joints at several sites in the body, including the elbow at the humeroradial and humeroulnar junctions (Kahn et al., 2009). Histological analysis of Spd mutant knees revealed that the meniscus and cruciate ligaments developed normally (Fig. 6), supporting our microarray findings.
The question of how individual joints attain the unique organization and structure required to support their specific functions remains largely unresolved. Here, we compare formation of the knee with that of the elbow, and using data from comparative microarray analysis at key embryonic stages, we identified the period E15 to E16 as the time when molecular evidence of elbow- and knee-specific identities could be detected. At this time, elbow joints showed enrichment for a set of factors involved in muscle specification and development, consistent with the requirement for muscle contraction in elbow development and supported by experiments showing that the elbow joint fails to form properly in muscleless mutants (Kahn et al., 2009). The knee joint appears to rely instead upon instruction from the TGFβ signaling pathway for proper morphogenesis. Knees from mice where TGFβ signaling is absent during joint development (Tgfbr2Prx1cre) lack menisci and cruciate ligaments, knee-specific structures absolutely required for proper knee function. These structures are often injured while performing normal activities and have very limited intrinsic capacity for healing, identifying the molecular events that lead to their formation may be a first step in finding new therapies for joint injury and disease.
Histological examination of mouse forelimb and hindlimb at E14 showed very similar patterns of joint morphogenesis. Distinct interzones were present that marked the site of the future elbow joint in the forelimb and knee joint in the hindlimb. Localization of known joint marker genes did not provide additional clues on how the developing elbow and knee might differ from each other at this time. At E15, elbow and knee joints had acquired their distinctive shapes, and during the next 24 hr, the meniscus and cruciate ligaments could be found in the knee. Based on this clear distinction between elbow and knee, we performed gene microarrays using knee and elbow libraries from E15 and E16 mouse limbs as a way to uncover the regulatory signals necessary for establishing and maintaining the identity of individual joints. Because the formation of individual joints is a dynamic process, we will need to look at earlier time points to uncover the causative molecular and cellular events responsible for sculpting their distinct morphologies. Follow-up studies will focus on using laser capture microdissection to isolate interzone cells from E12–E14 prospective elbow and knee joints and comparing their gene expression profiles.
The elbow gene enrichment signature provided by our array analyses consisted of mostly muscle-specific genes. While this finding could be due, in part, to our dissection procedure, it more likely reflects the complex muscle patterning required for the extensive range of motion that describes forelimb function. Our finding was borne out by expression of these muscle genes surrounding the developing elbow joint but mostly absent from the musculature surrounding the knee. Further support for our array findings came from analysis of the elbow and knee joints of the splotch-delayed mouse (Fig. 6). This muscleless mouse mutant fails to form a functional elbow joint (Kahn et al., 2009) but undergoes normal knee joint morphogenesis, consistent with the influence of muscle development in distinguishing elbow and knee formation. Although we did not observe a defect in knee joint development in muscle-deficient mutant mice, experiments performed in chicks using decamethonium bromide to induce paralysis resulted in knee joint fusion (Osborne et al., 2002). The disparity in these data may be due to differences in pharmacological versus genetic induction of paralysis and warrants further investigation of the specific role muscle plays in the formation of individual synovial joints.
Examination of knee joint enriched genes found that, as would be predicted, the hindlimb-specific transcription factors Hoxc10 and Tbx4, distinguished knee joint from elbow, validating our screen and identifying a potential role for these genes in knee morphogenesis that has not been previously described. Our subsequent bioinformatics analysis identified TGFβ signaling pathway components in the knee joint, consistent with the previous findings of Pryce et al. (2009) who found that knee morphogenesis is greatly impaired in the absence of Tgfbr2. Our examination of the knees of Tgfbr2Prx1cre mice showed that along with lack of cruciate ligaments described by Pryce et al. (2009), these knees also lacked menisci, establishing for the first time that TGFβ signaling is required for meniscal morphogenesis. As we find enhanced expression of Tgfbi in meniscus and cruciate ligaments, these knee-specific structures appear to be primary targets for TGFβ signaling in the developing knee. Because activation of canonical TGFβ signaling pathway is known to inhibit muscle formation (McPherron et al., 1997); down-regulation of this pathway in the elbow would allow for the musculature to form, and we speculate that a decrease in TGFβ signaling may be a key regulatory event in elbow morphogenesis. Of interest, it has been reported that the development of the intervertebral disc, a fibrocartilagenous structure similar to the meniscus also requires TGFβ signaling (Sohn et al., 2010), suggesting formation of these cushion-like tissues occurs by a similar mechanism.
Our findings suggest that Wnt activity has a complex role in the knee joint. Sfrp2 and Rspo2, modulators of canonical Wnt signaling, are highly expressed in the developing knee and appear to have unique but overlapping patterns of expression in the meniscus and cruciate ligaments. As Sfrp2 dampens both canonical and noncanonical Wnt signaling while Rspo2 activates it, it is likely that the requirement for these factors will be tissue-specific and temporally dependent, suggesting a role for Wnts in regionalization of the knee joint during its development. Regional differences are known to be an important factor in meniscal regenerative capacity in adults (Makris et al., 2011), but little information is available on how the stem cells in particular areas of the meniscus may differ. It is also important to note that as neither Sfrp2 or Rspo2 are found in the elbow joint, and Wnt signaling was not determined to be enriched using pathway analysis, our data provide another specific example of the fundamental differences in how the knee and elbow are formed.
Timed pregnant CD-1 and C57BL/6 mice were obtained from Charles River labs. Knee samples from Tgfbr2Prx1Cre cKO mice were a gift of Dr R. Schweitzer (Shriners Hospital for Children, Portland, OR), and knee samples from Spd mutant mice were a gift of Dr. E. Zelzer (Weizmann Institute, Israel).
Histology, In Situ Hybridization, and Immunohistochemistry
In vitro transcription to generate riboprobes was performed using standard protocols and reagents (Promega, Madison, WI). The plasmid used for generating the Tbx4 probe was a gift of Dr. V. Papaioannou (Columbia University, New York, NY); Rspo2, Dr. J. Yoon (Maine Medical Center, Portland, ME). Probes for Col2a1, Hoxc10, and Tgfbi were generated as previously described (Storm and Kingsley, 1996; Ferguson et al., 2003; Choe et al., 2006; Tsuji et al., 2006). Probe for Sfrp2 was generated using the following primers 5′ACGAGACCATGAAGGAGGTG-3′ and 5′-GGAGATGCGCTTGAACTCTC-3′; for Car3 5′-AGC CCA TGA CTG TGA GCT CAG-3′ and 5′-GTGGTTATTAAGAATGTTAGG-3′; for Myl3 5′GATGCTGACACCATGTCTGG-3′ and 5′-TAAGGCCACAGGGTGGATAC-3′; for Pitx2 5′-AAGTCGGCTCCCTAAAGAGG-3′ and 5′-GCGGTTTCTCTGGAAAGTGG-3′; for Ckm 5′-GAGTCCTACACGGTCTTCAAG-3′and 5′-TTCCAGCTTCTTCTCCATCTCC-3′; for Csrp3 5′-AGAGTCTTCACCATGCCAAAC-3′ and 5′-ACATTCCCAGTGGTCTCTTGT-3′. Forelimbs and hindlimbs from E16 mice were dissected in phosphate buffered saline (PBS), fixed overnight in 4% paraformaldehyde/PBS, and embedded in paraffin. Sections were stained with Toluidine Blue using standard methods. In situ hybridization was carried out as previously described (Gamer et al., 2009).
For Tenascin-C immunohistochemistry, forelimbs and hindlimbs from E16 CD-1 mice were fixed overnight in 4% paraformaldehyde/PBS, washed for 3 hours in 15% sucrose and overnight in 30% sucrose at 4°C before embedding in OCT. Rabbit anti-Tenascin-C antibody was a gift of Dr. H. Erickson (Duke University, Durham, NC) and immunohistochemistry for Tenascin-C was carried out as described previously (Pryce et al., 2009).
Knee and elbow joints including the epiphysis and intervening tissue were isolated from E15 and E16 CD-1 mice. RNA was isolated using the RNeasy kit (Qiagen, Gaithersburg, MD) and RNA quality was assessed by Agilent Bioanalyzer analysis. Microarray analysis was performed at the Partners HealthCare Center for Personalized Genetic Medicine (Cambridge, MA). Labeled cDNAs were hybridized to Illumina Mouse WG-6 microarrays. Three independent biological replicates were run to account for biological variability. Microarray data were analyzed using BeadStudio (Illumina, Inc., San Diego, CA). Array data were subjected to quantile normalization and a t-test was performed on the replicates followed by multiple hypothesis correction using the Benjami Hochberg test. Only fold changes with a P value of less than 0.05 and a false discovery rate of less than 0.05 are reported. BioBase ExPlain 3.0 was used to for bioinformatics analysis.