Search tips
Search criteria 


Cell Rep. 2016 June 21; 15(12): 2577–2587.
Published online 2016 June 9. doi:  10.1016/j.celrep.2016.05.055
PMCID: PMC4920976

Joint Development Involves a Continuous Influx of Gdf5-Positive Cells


Synovial joints comprise several tissue types, including articular cartilage, the capsule, and ligaments. All of these compartments are commonly assumed to originate from an early set of Gdf5-expressing progenitors populating the interzone domain. Here, we provide evidence that joints develop through a continuous influx of cells into the interzone, where they contribute differentially to forming joint tissues. Using a knockin Gdf5-CreERT2 mouse, we show that early labeling of Gdf5-positive interzone cells failed to mark the entire organ. Conversely, multiple Cre activation steps indicated a contribution of these cells to various joint compartments later in development. Spatiotemporal differences between Gdf5 and tdTomato reporter expression support the notion of a continuous recruitment process. Finally, differential contribution of Gdf5-positive cells to various tissues suggests that the spatiotemporal dynamics of Gdf5 expression may instruct lineage divergence. This work supports the influx model of joint development, which may apply to other organogenic processes.

Keywords: Gdf5, synovial joint, skeletogenesis, interzone, progenitor cell, Sox9, lineage tracing, mouse

Graphical Abstract

An external file that holds a picture, illustration, etc.
Object name is fx1.jpg


According to the prevailing model, organogenesis is initiated with the specification of a subset of cells as progenitors that are then organized to form a primordium. As development proceeds, the primordium undergoes extensive growth, differentiation, and morphogenesis until the organ obtains its final form (Biehs et al., 2013, Garrison et al., 2009; Göttgens, 2013; Gu et al., 2002, Little, 2011, Stanger et al., 2007, Wilkinson and Göttgens, 2013). Synovial joints are structures that separate adjacent skeletal elements and thus allow locomotion. The mature joint is composed of articular cartilage, synovial fluid, ligaments, and a fibrous capsule, which together function to transmit biomechanical loads during skeletal motion (Decker et al., 2014).

The first histological indication for joint formation is the appearance of a higher cell density domain called the interzone at the site of the future joint. The interzone consists of three layers, the middle of which is termed the intermediate zone (Mitrovic, 1977). Molecularly, interzone cells lose the expression of chondrocyte-specific genes such as collagen type II. Instead, they express a new set of genes that includes growth and differentiation factor 5 (Gdf5), Wnt4, and Wnt9a (Guo et al., 2004, Hartmann and Tabin, 2001, Später et al., 2006, Storm et al., 1994).

Over the years, it has been suggested that interzone cells serve as joint progenitors. Removal of the interzone was shown to lead to joint loss and bone fusion (Holder, 1977). Light and electron microscopy studies have suggested that interzone cells from the two outside layers contribute secondarily to the adjacent epiphyses, whereas intermediate zone cells form the external layers of the joint (Ito and Kida, 2000). Finally, lineage studies utilizing Gdf5-Cre mice showed that cells of all mature joint components originate from Gdf5-expressing lineages, further promoting the view that interzone cells are joint progenitors (Koyama et al., 2007, Koyama et al., 2008, Rountree et al., 2004).

These works support a model in which, during the initial stages of skeletogenesis, a subset of mesenchymal cells is specified as interzone cells to serve as joint progenitors. However, evidence accumulated in recent years indicates that, during development, cells from the vicinity are integrated into the forming joint (Hyde et al., 2008, Koyama et al., 2007, Pacifici et al., 2006, Ray et al., 2015). These findings suggest a more elaborate mode of development, reopening the question of whether joint development follows the widely accepted scheme of organogenesis. In this study, we use a new Gdf5-CreERT2 knockin mouse line to propose a revised model for joint development. We show that joint development occurs through a continuous influx of new cells that are integrated into the interzone, where they contribute to the forming joint, hence the term influx model. Finally, we show that this mechanism facilitates lineage divergence and, thereby, joint development.


Establishment of the Gdf5-CreER Knockin Mouse as a Model System

The current consensus that Gdf5-positive interzone cells are joint progenitors has been based on lineage tracing experiments using a Gdf5-Cre transgenic mouse. This line was produced by integrating the Cre recombinase sequence into a 140-kb bacterial artificial chromosome (BAC) containing the Gdf5 locus (Rountree et al., 2004). This transgene might not contain all of the potential regulatory elements that control endogenous Gdf5 expression and might also be influenced by the genomic integration site. Therefore, to directly test that view, we sought to generate a Gdf5-CreER mouse that would allow us to temporally label Gdf5-expressing cells. Taking a knockin approach, we inserted a cassette encoding for CreER into the translation start site of the Gdf5 gene using homologous recombination (Figure S1Aa). To demonstrate both the accuracy and the efficiency of the new Gdf5-CreER line, we crossed this mouse with a Rosa26-tdTomato reporter line and induced recombination by a single tamoxifen dose administered on embryonic day (E) 13.5. Whole-mount preparations of E18.5 embryos showed that joints in the vertebrae, forelimb, and hindlimb were marked (Figures S1Ab–S1Ad’’), suggesting that Cre recombinase activity was indeed restricted to joints. Importantly, no reporter activity was evident without tamoxifen administration (Figures S1Ba–S1Bc).

Next, we crossed mice heterozygous for Gdf5-CreER and analyzed the homozygous offspring. As expected, the phenotype recapitulated that of brachypodism (bp) mice (Harada et al., 2007, Mikic, 2004, Storm et al., 1994); namely, malformation in the autopods and fibula, failure of cruciate ligament development, and expansion of the Gdf5 expression domain in the autopod (Figures S1Ca–S1Ce’). Together, these results indicate that the Gdf5-CreER line is a suitable system for lineage analysis of Gdf5-positive interzone cells.

Joint-Forming Cells Are Not Specified in a Single Early Event

According to the interzone progenitor model, once the interzone has formed, the cells inside it are specified as joint progenitors, as indicated by Gdf5 expression. This hypothesis predicts that administration of tamoxifen at that stage would “capture” most of the population of joint progenitors, resulting in cell labeling throughout the joint. To directly examine this hypothesis, we performed a series of pulse-chase experiments by crossing Gdf5-CreER mice with the Rosa26-tdTomato reporter. As seen in Figure 1A, on E11.5, Gdf5 expression was clearly observed in the presumptive elbow, knee, and autopod joints. Assuming a 24-hr delay in maximum Cre activity, a single dose of tamoxifen was administered on E10.5 to label Gdf5-positive interzone cells. Interestingly, analysis of sections through the knee, elbow, and metacarpophalangeal joints on E18.5 showed only restricted labeling, mostly of epiphyseal cells (Figure 1B). This surprising result suggests that the Gdf5-positive cells that populate the interzone at the time of its formation are not the main source of joint progenitors.

Figure 1
Limited Proliferation of Gdf5+ Lineage Cells Contradicts the Interzone Progenitor Hypothesis

An alternative hypothesis that could reconcile our observations with those obtained previously using the Gdf5-Cre line is that there is a constant inflow of new cells into the forming joint. The “influx” model predicts that a single administration of tamoxifen will mark only a partial population of joint cells and that only several administrations at different developmental stages will induce extensive labeling. To test this hypothesis, we administered three doses of tamoxifen on E11.5, E13.5, and E15.5. Examination of sections through several E18.5 joints showed that tdTomato-positive cells occupied most of the joint area (Figure 1C), thus reinforcing the influx hypothesis.

The interzone progenitor model implies a requirement for continuous proliferation of the progeny of early-specified Gdf5-positive progenitors throughout development to sustain rapid joint growth. To further distinguish between the two hypotheses, we examined the proliferative potential of these cells. We analyzed, by bromodeoxyuridine (BrdU) assay, the descendants of Gdf5-positive cells labeled by a single tamoxifen administration on E11.5. Sections through the elbow and knee joints showed that approximately 20% of the descendants, marked by tdTomato, were proliferating on E13.5. Moreover, the level of proliferation dramatically decreased through E15.5 and E18.5 (Figures 1D and 1E). Taken together, these results imply a continuous influx of cells into the forming joint.

Joint Development Involves a Continuous Influx of New Gdf5-Expressing Cells

Next, to directly demonstrate the recruitment process, we monitored the spatial distribution of Gdf5 and tdTomato expression in the interzone during development. The probe for the tdTomato gene was expected to mark cells that have already been specified as Gdf5-positive at the time of tamoxifen administration and their descendants, whereas the probe for Gdf5 would mark cells that expressed Gdf5 at the time of harvesting. Thus, expression of both markers would indicate cells that were recruited early and maintained Gdf5 expression, whereas newly recruited, non-pulsed cells would express only Gdf5. First, to verify the sensitivity of in situ hybridization (ISH) for tdTomato, we compared the results of that assay (red) with immunofluorescent staining for tdTomato protein (green) (Figure 2A). The co-localization of labeled cells showed that ISH for tdTomato is a reliable tool in cell lineage studies.

Figure 2
Joint Development Involves Continuous Steps of Recruitment and Specification of New Cells

According to our model, immediately after Cre activation, interzone cells should express both Gdf5 and tdTomato. To examine this assumption, we performed double ISH for both genes on sections from Gdf5-CreER, Rosa26-tdTomato embryos following administration of a single dose of tamoxifen at different time points during joint development (E11.5–E13.5). Embryos were harvested 28 hr after tamoxifen administration because we had expected a delayed activation of the Rosa26 reporter. In agreement with our hypothesis, sections through the knee showed a high overlap between Gdf5- and tdTomato-positive cells pulsed at all the time points (Figures 2B and S2).

To assess the efficiency of Gdf5-CreER recombination quantitatively, we measured the level of co-localization of Gdf5- and tdTomato-expressing cells. The results showed that 60%–67% of the Gdf5 expression domain was also positive for tdTomato (Figure 2C). Surprisingly, 28 hr after tamoxifen administration, we already observed a population of tdTomato-positive, Gdf5-negative cells (Figure 2B, arrows). Assessment of the proportion of that population relative to the entire tdTomato expression domain showed that, at all examined stages, approximately 30% of tdTomato-positive cells had already lost Gdf5 expression. This observation indicates a very dynamic cellular behavior and suggests that the actual recombination efficiency is higher than calculated (Figures 2C and 2D). Moreover, the coexpression of Gdf5 and tdTomato by most cells after that interval demonstrated our ability to effectively detect the expression of both markers as well as the high level of Cre activity following a single tamoxifen administration.

Another prediction of our model is that longer pulse-chase intervals would reduce the overlap between the Gdf5 and tdTomato expression domains as a result of Gdf5 expression by cells recruited after pulsing. Indeed, examination of E13.5–E15.5 knee joints of embryos pulsed on E11.5 showed that, in addition to cells that co-expressed both markers, a high number of cells expressed only Gdf5 but not tdTomato (Figure 2E, arrows). As in the previous experiment (Figure 2B), a third population of cells that solely expressed tdTomato was also identified, presumably having lost Gdf5 expression. These results reveal a highly dynamic behavior of joint-forming cells; new cells turn Gdf5 expression on, whereas others turn it off.

Because, in that experiment, tamoxifen was administered only on E11.5, the observed dynamic behavior could be unique to that developmental stage. To address that issue, we repeated the pulse-chase experiments, administering single doses of tamoxifen at different stages (E10.5–E13.5), and examined tdTomato and Gdf5 expression in the knee joint on E15.5. In all experiments, the three abovementioned populations were detected (Figure 2F), indicating a continuous influx of Gdf5-positive cells into the joint.

Finally, our finding that cells were constantly leaving the Gdf5 expression domain suggested that the contribution of the Gdf5 lineage to the developing epiphyses is more intricate and fine-tuned than has been appreciated previously (Koyama et al., 2007, Rountree et al., 2004, Storm and Kingsley, 1999). To better understand the temporal contribution of the Gdf5 lineage to the epiphysis, we performed pulse-chase experiments using Gdf5-CreER mice crossed with Rosa26-tdTomato mice. Following administration of a single dose of tamoxifen on E11.5–E13.5, E15.5 embryos were stained for both tdTomato and collagen type II α 1 (COL2A1). Activation on E11.5 resulted in a substantial population of tdTomato, COL2A1 double-positive cells. Later marking of Gdf5-expressing populations by tamoxifen administration on E12.5–E13.5 (Figure 2G) resulted in a gradual decrease in labeling of epiphyseal chondrocytes and labeled chondrocytes more proximal to the joint, indicating a spatial shift toward the joint area. These results indicate that the cells lost their Gdf5 expression (Figure 2F), joined the growing cartilage, and differentiated into chondrocytes (Figure 2G).

This set of experiments shows that the Gdf5 expression domain is highly dynamic, with a constant influx of new cells into the joint area as well as out of it into the epiphyses.

The Influx of Cells into the Joint Facilitates Lineage Divergence

A possible implication of the influx model is that Gdf5-positive cells recruited at different time points may give rise to specific joint tissues. To address this hypothesis, we analyzed the spatiotemporal contribution of cell lineages marked at different stages to various joints of E18.5 embryos. For that, we administered single doses of tamoxifen between E10.5 and E16.5 to embryos heterozygous for Gdf5-CreER and Rosa26-tdTomato. As seen in Figure 3, descendants of Gdf5-positive cells marked at different time points populated different regions of the developing joints.

Figure 3
Differential Contribution of Cell Subpopulations to Various Joint Tissues

In the knee, Gdf5-positive cells were detected in the epiphyses, articular cartilage, meniscus, and intra-articular ligaments. Early tamoxifen administration (on E10.5) resulted in tdTomato-positive cells located mostly at the epiphyses. However, when tamoxifen was administered at E11.5, tdTomato-positive cells were also observed in the articular cartilage. Upon activation on E12.5–E13.5, the number of tdTomato-positive cells in the epiphyses decreased, whereas their relative number in the articular cartilage increased. Finally, following activation on E14.5–E15.5, the expression was restricted to the most superficial layer of the articular cartilage (Figures 3A and 3B). Tamoxifen administration at all stages led to a contribution of tdTomato-positive cells to the meniscus and intra-articular ligaments. However, whereas pulsing at early and intermediate stages led to a massive contribution throughout the tissues, at later stages it resulted in an exclusive contribution of these cells to the edges of the meniscus (Figure 3A) and intra-articular ligament (Figure 3C).

In the elbow joint, a similar trend was observed, except that a contribution of Gdf5-positive cells to the articular cartilage was observed following activation on E12.5 (Figures 3D and 3E). The metacarpophalangeal joint also exhibited a similar spatiotemporal pattern. However, administration on E13.5 (Figure 3F) resulted in prominent labeling of the joint capsule. Finally, the different contribution of labeled cells to epiphyseal chondrocytes seen in knee sections (Figures 3A and 3C) prompted us to further examine the spatial distribution of Gdf5- and tdTomato-expressing cells in E15.5 embryos pulsed on E11.5 (Figure 3G) or E12.5 (Figure 3H). Mid-sagittal sections through the knee, where the cruciate ligaments were visualized, revealed significantly more cells that expressed both genes (Figures 3G and 3H, bottom) and a clear reduction in tdTomato-positive, Gdf5-negative cells compared with medial sections showing the menisci. Presumably, the latter cell population highly contributes to the epiphysis (Figures 3G and 3H, top).

Importantly, in all joints we observed asymmetry in the contribution of Gdf5-positive cells to the skeletal elements at the two sides of the joint. For example, the femur and the humerus were more extensively marked than the opposing bones. This observation suggests that asymmetric contribution may be part of the morphogenetic process (Figure 1C).

Overall, the spatiotemporal differences in Gdf5 expression patterns show the dynamics of the establishment of the different joint tissues. Although some elements, like the cruciate ligaments and menisci, are formed by constant recruitment of new Gdf5-expressing cells that maintain its expression, others, like the metacarpophalangeal joint capsule, originate from a temporally specific group of cells. It is tempting to speculate that this mode of development may facilitate lineage divergence into different cell fates to allow the formation of the various tissues that compose the joint.

A Continuous Influx of Sox9-Positive Cells May Contribute to Joint Morphogenesis

Previously, it was shown that joint cells originate from Sox9-positive chondroprogenitor cells (Soeda et al., 2010). To confirm this, we crossed Sox9-CreER mice with the Rosa26-tdTomato reporter. Examination of E15.5 knee and elbow joints following administration of a single tamoxifen dose at E10.5 showed that all structures of the joints were tdTomato-positive, thus originating from an early population of Sox9-positive cells (Figures S3Aa–S3Ad). To further establish that Gdf5-positive cells are derived from that early pool, double ISH for tdTomato and Gdf5 was performed on sections through the knee of Sox9-CreER, Rosa26-tdTomato heterozygous embryos. The co-expression of both markers confirmed that conclusion (Figure S3Ae).

Several studies have suggested that cells from the vicinity may migrate into the developing joint (Hyde et al., 2008, Koyama et al., 2007, Pacifici et al., 2006, Ray et al., 2015). To locate the source that may supply new cells to the developing joint, we next analyzed the expression of the two abovementioned genes alongside COL2A1, a marker of differentiating chondrocytes, throughout joint development. To visualize all three markers simultaneously, double ISH for Sox9 and Gdf5 was performed together with immunofluorescence staining for COL2A1. Examination of the elbow and knee joints at E12 revealed, as expected, a population of Sox9-Gdf5 double-positive cells at the presumptive joint area. This population was later observed at all examined stages (Figures 4A–4F and S3Ca–S3Cf). However, around the developing joint, we observed a cloud of Sox9-positive cells that did not express Col2a1 or Gdf5 (Figures 4A and S3Ca). That population was also prominent in double immunofluorescence staining for SOX9 and COL2A1 (Figure S3B).

Figure 4
Gene Expression Dynamics Suggests that Influx of Sox9-Positive Cells Contributes to Joint Formation

As development proceeded, two concurrent events were observed: massive growth and morphogenesis of the joint and a gradual reduction in the population of Sox9-positive, Col2a1, Gdf5-negative cells around the joint (Figures 4A–4F and S3Ca–S3Cf; arrows), which was also confirmed by immunofluorescent staining (Figure S3B). Interestingly, in the interzone, between the Col2a1-positive cells that flanked it and the Gdf5 expression domain, another population of cells that was Sox9-positive and Col2a1, Gdf5-negative was observed (Figures 4C, 4D, and S3C; dashed arrows). These cells were likely descendants of early Gdf5-positive cells that had lost Gdf5 expression, as we demonstrated previously (Figure 2), and were in the process of differentiation to chondrocytes while integrating into the epiphyses (Figure S3D).

Together, these results suggest that Sox9-positive, Col2a1, Gdf5-negative cells from the vicinity of the joint migrate into the forming joint, turn on Gdf5 expression, and contribute to joint development.


Over the years, a general scheme of organogenesis has been widely adopted. This scheme has guided efforts to trace the lineages of organ-forming cells to identify their progenitors. In accordance with this view, Gdf5-positive interzone cells have been suggested as joint progenitors. However, the data we present here are inconsistent with the progenitor model sensu stricto and favor an alternative strategy for joint development that involves a constant influx of new cells that are integrated into the forming organ to sustain its development.

Several lines of evidence in our study are inconsistent with the view that joint progenitor cells are all specified at an early stage of limb development. None of our attempts, using a knockin Gdf5-CreER line, to capture the entire progenitor population by a single Cre activation succeeded; each time, only a certain subpopulation of joint cells was marked. Only several doses of tamoxifen at different developmental stages resulted in extensive labeling throughout the joint. Moreover, cell proliferation assays revealed a reduction in the proliferation of the Gdf5 lineage through development that could not accommodate the demands of the developing joint. Finally, our finding of a time-dependent increase in the number of new Gdf5-positive cells in the interzone may also be incompatible with the progenitor model.

A possible limitation of our approach would be low recombination activity of the new Gdf5-CreER line, which would lead to low labeling efficacy. However, comparable levels of tdTomato and endogenous Gdf5 expression observed and computed shortly after Cre activation suggest that our new line acts efficiently on the Rosa26-tdTomato reporter line.

The concept of the addition of cells to a developing organ instead of relying exclusively on proliferation of progenitor cells is interesting. Previously, we showed that cells that form bone eminences and the patella are added secondarily to the already established bone shaft anlage (Blitz et al., 2009, Blitz et al., 2013, Eyal et al., 2015). Although these studies fall short of providing direct evidence for the continuity of this process, the different timing suggests that the influx model may be a broader phenomenon in skeletal development.

Intriguingly, our findings may appear to disagree with Holder’s seminal experiment in which removal of the interzone led to failure in joint development (Holder, 1977). There are several possible ways for reconciling these seemingly opposing conclusions. For example, Holder might have removed with the interzone the surrounding tissues, including the potential source for migrating Gdf5-positive cells, leading to failure in joint development. Also, several lines of evidence support the notion that the interzone is a vital signaling center in joint development (Decker et al., 2014, Longobardi et al., 2015). Thus, removal of the interzone could have eliminated the regulatory signals mediating the recruitment process. A third possibility is that Holder also removed muscles, which were shown previously by us and others to be essential for joint development (Kahn et al., 2009, Nowlan et al., 2010, Pitsillides and Ashhurst, 2008, Shwartz et al., 2013).

A central question in organ development is the mechanism by which lineage divergence occurs to form different tissues. This question is particularly interesting in the case of joint development because all joint tissues were shown to originate from Gdf5-expressing cells (Koyama et al., 2007, Koyama et al., 2008, Rountree et al., 2004). Previous studies have suggested two possible mechanisms for the divergence into epiphyseal and articular chondrocytes based on either temporal or spatial differences. The temporal hypothesis is that cells that express Gdf5 early will contribute to the epiphysis (Storm and Kingsley, 1999), whereas the spatial hypothesis is that the location of the progenitors in the interzone determines the ultimate lineage destination (Ito and Kida, 2000). We propose the high cellular dynamics of Gdf5 expression as the underlying factor of tissue divergence.

Although the mechanism that translates Gdf5 expression dynamics to lineage divergence is yet to be deciphered, we identify two important parameters, the duration and time of onset of Gdf5 expression, whose combination may determine lineage divergence (Figure 5). For example, cells that would eventually contribute to the epiphysis lost Gdf5 expression, whereas cells that contributed to the meniscus and knee cruciate ligaments exhibited completely different dynamics and maintained Gdf5 expression. Interestingly, articular cartilage was also found to express Gdf5 continually. However, the recruitment of most of these cells started 1 day later than in the ligaments and menisci. This differential behavior can serve as a mechanism by which cells contributing to different joint tissues acquire their fate (Figure 5). This notion is supported by previous studies showing that Gdf5 is essential for the development of intra-articular ligaments (Harada et al., 2007), which were absent in bp mice, and that human mesenchymal stem cells (hMSCs) and murine adipose tissue-derived mesenchymal stem cells treated with GDF5 showed an elevation in tendon-specific markers such as scleraxis (Scx) (Park et al., 2010, Tan et al., 2012). An example for the onset of Gdf5 expression as a cue for lineage divergence is the case of the metacarpophalangeal joint capsule, where the contributing cells start to express this gene only on E13.5.

Figure 5
A Revised Model for Joint Development

The influx of new cells and the dynamic behavior of joint-forming, Gdf5-expressing cells may provide flexibility both in development and during evolution. Various joint structures support specific functions and mechanical needs. Development through multiple steps, some of which are lineage-specific, may enable faster and more efficient changes in joint structure by manipulating only a subpopulation of joint-forming cells.

A key question that remains to be solved is the source of the recruited cells. Previous studies have shown that all joint structures originate from matrilin 1-negative (Hyde et al., 2007) and Sox9-positive cells (Soeda et al., 2010). Other works indicate that cells are recruited into the developing joint (Hyde et al., 2008, Koyama et al., 2007, Pacifici et al., 2006, Ray et al., 2015). Our study identifies a population of Sox9-positive, Col2a1, Gdf5-negative cells outside of the joint area that may serve as an external source. However, this supposition is yet to be conclusively confirmed. Another possibility is an internal origin correlating to the Sox9-positive, Col2a1, Gdf5-negative cells we observed in the interzone. However, we believe that there are several indications for recruitment from an outside source.

First, the Gdf5 expression domain is initially very broad, occupying the entire interzone and leaving no room for Sox9+, Gdf5−, Col2a1− cells to contribute to the Gdf5-expressing population. Moreover, as we demonstrate, the CreERT2 system is highly efficient and is thus likely to mark the majority of Gdf5-positive interzone cells. Nonetheless, we cannot rule out the possibility that a unique subpopulation of unlabeled Gdf5-positive interzone cells contributes to joint growth and, as an internal source, to cell recruitment. Last, we demonstrate a clear decrease in the number of Sox9+, Gdf5− cells flanking the forming joint during development, supporting the argument that this population is the source for joint cells.

Interestingly, a recent study by Ray et al. (2015) suggests that articular cartilage originates from proliferating chondrocytes, proposing another possible source outside of the interzone. However, because a different system was used in that study, the Col2a1-CreERT2 mouse, the exact spatial and temporal relationships between the process identified by Ray et al. (2015) and the one described here are yet to be determined. A plausible explanation for the apparent differences is that both these processes contribute to joint formation in varying proportions. Thus, examination of different joints at different time points may yield inconsistent results.

From a medical perspective, joints exhibit a very limited capacity for regeneration (Iwamoto et al., 2013, Sandell, 2012). Our results suggest that, to improve regeneration, it may be necessary to search for a mechanism that can enhance or activate cell recruitment.

In conclusion, our findings indicate that joints do not form from a single pool of early-specified Gdf5-positive cells but, rather, through a continuous influx of Gdf5-positive cells, hence the term influx model. Moreover, we show highly dynamic spatiotemporal patterns of Gdf5 expression by interzone cells, which may serve as differentiation cues. An additional differentiation signal may be the temporal specification of cells that will contribute differentially to the various tissues of the mature joint, thus facilitating lineage divergence of Gdf5-positive cells.

Experimental Procedures


All experiments involving mice were approved by the Institutional Animal Care and Use Committee (IACUC) of the Weizmann Institute. Rosa26-tdTomato mice (B6;129S6-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J) have been described previously (Madisen et al., 2010, Soeda et al., 2010). The generation of Gdf5-CreER knockin mice is described in the Supplemental Experimental Procedures. To create Gdf5-CreER, Rosa26-tdTomato mice, floxed Rosa26-tdTomato mice were mated with Gdf5-CreER mice.

For harvesting of embryos, timed pregnant females were sacrificed by cervical dislocation.

Tamoxifen Administration

Pregnant female mice were administered 0.12–0.18 mg/g tamoxifen/body weight in corn oil by oral gavage.


Embryos were fixed in 4% paraformaldehyde (PFA)/PBS for 3 hr at 4°C, washed overnight in 30% sucrose/PBS at 4°C, embedded in O.C.T., and sectioned at a thickness of 10 μm.

Immunofluorescent Staining

For paraffin section immunofluorescence, embryos were fixed overnight in 4% PFA/PBS, dehydrated to 100% ethanol, embedded in paraffin, and sectioned at a thickness of 7 μm. 10 mM sodium citrate (pH 6.0) was used for antigen retrieval. Slides were incubated overnight at 4°C with primary antibody anti-BrdU (AbD) diluted 1:200 and with biotinylated anti-RFP (Abcam) diluted 1:100 in blocking solution to visualize tdTomato. For COL2A1 and tdTomato double immunostaining, after staining for tdTomato, slides were subjected to additional digestion by Proteinase K (Sigma). Slides were incubated overnight at 4°C with the primary antibody anti-COL2 (Developmental Studies Hybridoma Bank [DSHB], University of Iowa) 1:50 in blocking solution.

Following fluorescence ISH, slides were incubated overnight at 4°C with the primary antibody anti-COL2 (DSHB) 1:50 in blocking solution or with anti-RFP (Abcam) diluted 1:100 in blocking solution to visualize tdTomato.

In Situ Hybridization

Non-fluorescent ISH on paraffin sections was performed using a digoxigenin (DIG)-labeled probe for Gdf5. Double fluorescence ISH on paraffin sections was performed using fluorescein- and DIG-labeled probes. The extended protocol was described by Shwartz and Zelzer (2014).

BrdU Administration

Mice were injected intraperitoneally with 100 mg/kg of body weight BrdU labeling reagent (Sigma) and sacrificed 4 hr later. Forelimbs and hindlimbs were fixed in 4% PFA overnight at 4°C, dehydrated in ethanol, and embedded in paraffin.

Cell Count

For calculation of the percentage of tdTomato-positive cells contributing to the articular cartilage, 10-μm thick cryosections were counterstained with DAPI. tdTomato-positive cells were manually counted at the distal femoral and humeral heads as well as at the articular cartilage area, identified histologically as the external four to five cell layers. The percentage was calculated for each photographed section and averaged first for each embryo and then for each analyzed stage of tamoxifen administration.

To test the proliferation of tdTomato-positive cells, tdTomato-positive and tdTomato-BrdU-double positive cells were counted manually in BrdU-, RFP-, and DAPI-stained paraffin sections. In both experiments, 3-12 sections of each joint from each embryo were used, and three embryos from different litters were analyzed for each embryonic stage. Sections were photographed and examined by an LSM 710 or 780 confocal microscope. Data are represented as mean ± SD.

Co-localization Measurements

Section double ISH for tdTomato and Gdf5 was imaged with an LSM 710 or 780 confocal microscope. The co-localization analysis was performed using the Imaris 8.2 colocalization module. For each analyzed picture, a region of interest covering the entire area of the knee joint containing tdTomato- and Gdf5-positive cells was defined and used for all measurements. Two to eleven sections of each joint from each embryo were used, and two to three embryos from two to three different litters were analyzed for each embryonic stage. The results represent all measured values ± SD.

Author Contributions

Y.S., S.V., and S.K. conducted the experiments. E.Z. and Y.S. designed the study and wrote the manuscript.


We thank N. Konstantin for expert editorial assistance. We thank Dr. Reinat Nevo for help with the co-localization analysis and all members of the E.Z. laboratory for advice and suggestions. This study was funded by the European Research Council (ERC; grant no. 310098), the Jeanne and Joseph Nissim Foundation for Life Sciences Research, the Y. Leon Benoziyo Institute for Molecular Medicine, Beth Rom-Rymer, the Estate of David Levinson, the Jaffe Bernard and Audrey Foundation, Georges Lustgarten, and the David and Fela Shapell Family Foundation.


Published: June 9, 2016


Supplemental Information includes Supplemental Experimental Procedures and three figures and can be found with this article online at

Supplemental Information

Document S1. Supplemental Experimental Procedures and Figures S1–S3:
Document S2. Article plus Supplemental Information:


Biehs B., Hu J.K., Strauli N.B., Sangiorgi E., Jung H., Heber R.P., Ho S., Goodwin A.F., Dasen J.S., Capecchi M.R., Klein O.D. BMI1 represses Ink4a/Arf and Hox genes to regulate stem cells in the rodent incisor. Nat. Cell Biol. 2013;15:846–852. [PubMed]
Blitz E., Viukov S., Sharir A., Shwartz Y., Galloway J.L., Pryce B.A., Johnson R.L., Tabin C.J., Schweitzer R., Zelzer E. Bone ridge patterning during musculoskeletal assembly is mediated through SCX regulation of Bmp4 at the tendon-skeleton junction. Dev. Cell. 2009;17:861–873. [PubMed]
Blitz E., Sharir A., Akiyama H., Zelzer E. Tendon-bone attachment unit is formed modularly by a distinct pool of Scx- and Sox9-positive progenitors. Development. 2013;140:2680–2690. [PubMed]
Decker R.S., Koyama E., Pacifici M. Genesis and morphogenesis of limb synovial joints and articular cartilage. Matrix Biol. 2014;39:5–10. [PubMed]
Eyal S., Blitz E., Shwartz Y., Akiyama H., Schweitzer R., Zelzer E. On the development of the patella. Development. 2015;142:1831–1839. [PubMed]
Garrison A.P., Helmrath M.A., Dekaney C.M. Intestinal stem cells. J. Pediatr. Gastroenterol. Nutr. 2009;49:2–7. [PubMed]
Gu G., Dubauskaite J., Melton D.A. Direct evidence for the pancreatic lineage: NGN3+ cells are islet progenitors and are distinct from duct progenitors. Development. 2002;129:2447–2457. [PubMed]
Guo X., Day T.F., Jiang X., Garrett-Beal L., Topol L., Yang Y. Wnt/beta-catenin signaling is sufficient and necessary for synovial joint formation. Genes Dev. 2004;18:2404–2417. [PubMed]
Harada M., Takahara M., Zhe P., Otsuji M., Iuchi Y., Takagi M., Ogino T. Developmental failure of the intra-articular ligaments in mice with absence of growth differentiation factor 5. Osteoarthritis Cartilage. 2007;15:468–474. [PubMed]
Hartmann C., Tabin C.J. Wnt-14 plays a pivotal role in inducing synovial joint formation in the developing appendicular skeleton. Cell. 2001;104:341–351. [PubMed]
Holder N. An experimental investigation into the early development of the chick elbow joint. J. Embryol. Exp. Morphol. 1977;39:115–127. [PubMed]
Hyde G., Dover S., Aszodi A., Wallis G.A., Boot-Handford R.P. Lineage tracing using matrilin-1 gene expression reveals that articular chondrocytes exist as the joint interzone forms. Dev. Biol. 2007;304:825–833. [PubMed]
Hyde G., Boot-Handford R.P., Wallis G.A. Col2a1 lineage tracing reveals that the meniscus of the knee joint has a complex cellular origin. J. Anat. 2008;213:531–538. [PubMed]
Ito M.M., Kida M.Y. Morphological and biochemical re-evaluation of the process of cavitation in the rat knee joint: cellular and cell strata alterations in the interzone. J. Anat. 2000;197:659–679. [PubMed]
Iwamoto T., Matsumura N., Sato K., Momohara S., Toyama Y., Nakamura T. An obliquely placed headless compression screw for distal interphalangeal joint arthrodesis. J. Hand Surg. Am. 2013;38:2360–2364. [PubMed]
Kahn J., Shwartz Y., Blitz E., Krief S., Sharir A., Breitel D.A., Rattenbach R., Relaix F., Maire P., Rountree R.B. Muscle contraction is necessary to maintain joint progenitor cell fate. Dev. Cell. 2009;16:734–743. [PubMed]
Koyama E., Ochiai T., Rountree R.B., Kingsley D.M., Enomoto-Iwamoto M., Iwamoto M., Pacifici M. Synovial joint formation during mouse limb skeletogenesis: roles of Indian hedgehog signaling. Ann. N Y Acad. Sci. 2007;1116:100–112. [PubMed]
Koyama E., Shibukawa Y., Nagayama M., Sugito H., Young B., Yuasa T., Okabe T., Ochiai T., Kamiya N., Rountree R.B. A distinct cohort of progenitor cells participates in synovial joint and articular cartilage formation during mouse limb skeletogenesis. Dev. Biol. 2008;316:62–73. [PubMed]
Little M.H. Renal organogenesis: what can it tell us about renal repair and regeneration? Organogenesis. 2011;7:229–241. [PubMed]
Longobardi L., Li T., Tagliafierro L., Temple J.D., Willcockson H.H., Ye P., Esposito A., Xu F., Spagnoli A. Synovial joints: from development to homeostasis. Curr. Osteoporos. Rep. 2015;13:41–51. [PubMed]
Madisen L., Zwingman T.A., Sunkin S.M., Oh S.W., Zariwala H.A., Gu H., Ng L.L., Palmiter R.D., Hawrylycz M.J., Jones A.R. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 2010;13:133–140. [PubMed]
Mikic B. Multiple effects of GDF-5 deficiency on skeletal tissues: implications for therapeutic bioengineering. Ann. Biomed. Eng. 2004;32:466–476. [PubMed]
Mitrovic D.R. Development of the metatarsophalangeal joint of the chick embryo: morphological, ultrastructural and histochemical studies. Am. J. Anat. 1977;150:333–347. [PubMed]
Nowlan N.C., Sharpe J., Roddy K.A., Prendergast P.J., Murphy P. Mechanobiology of embryonic skeletal development: Insights from animal models. Birth Defects Res. C Embryo Today. 2010;90:203–213. [PubMed]
Pacifici M., Koyama E., Shibukawa Y., Wu C., Tamamura Y., Enomoto-Iwamoto M., Iwamoto M. Cellular and molecular mechanisms of synovial joint and articular cartilage formation. Ann. N Y Acad. Sci. 2006;1068:74–86. [PubMed]
Park A., Hogan M.V., Kesturu G.S., James R., Balian G., Chhabra A.B. Adipose-derived mesenchymal stem cells treated with growth differentiation factor-5 express tendon-specific markers. Tissue Eng. Part A. 2010;16:2941–2951. [PubMed]
Pitsillides A.A., Ashhurst D.E. A critical evaluation of specific aspects of joint development. Dev. Dyn. 2008;237:2284–2294. [PubMed]
Ray A., Singh P.N., Sohaskey M.L., Harland R.M., Bandyopadhyay A. Precise spatial restriction of BMP signaling is essential for articular cartilage differentiation. Development. 2015;142:1169–1179. [PubMed]
Rountree R.B., Schoor M., Chen H., Marks M.E., Harley V., Mishina Y., Kingsley D.M. BMP receptor signaling is required for postnatal maintenance of articular cartilage. PLoS Biol. 2004;2:e355. [PubMed]
Sandell L.J. Etiology of osteoarthritis: genetics and synovial joint development. Nat. Rev. Rheumatol. 2012;8:77–89. [PubMed]
Shwartz Y., Zelzer E. Nonradioactive in situ hybridization on skeletal tissue sections. Methods Mol. Biol. 2014;1130:203–215. [PubMed]
Shwartz Y., Blitz E., Zelzer E. One load to rule them all: mechanical control of the musculoskeletal system in development and aging. Differentiation. 2013;86:104–111. [PubMed]
Soeda T., Deng J.M., de Crombrugghe B., Behringer R.R., Nakamura T., Akiyama H. Sox9-expressing precursors are the cellular origin of the cruciate ligament of the knee joint and the limb tendons. Genesis. 2010;48:635–644. [PubMed]
Später D., Hill T.P., Gruber M., Hartmann C. Role of canonical Wnt-signalling in joint formation. Eur. Cell. Mater. 2006;12:71–80. [PubMed]
Stanger B.Z., Tanaka A.J., Melton D.A. Organ size is limited by the number of embryonic progenitor cells in the pancreas but not the liver. Nature. 2007;445:886–891. [PubMed]
Storm E.E., Kingsley D.M. GDF5 coordinates bone and joint formation during digit development. Dev. Biol. 1999;209:11–27. [PubMed]
Storm E.E., Huynh T.V., Copeland N.G., Jenkins N.A., Kingsley D.M., Lee S.J. Limb alterations in brachypodism mice due to mutations in a new member of the TGF beta-superfamily. Nature. 1994;368:639–643. [PubMed]
Tan S.L., Ahmad R.E., Ahmad T.S., Merican A.M., Abbas A.A., Ng W.M., Kamarul T. Effect of growth differentiation factor 5 on the proliferation and tenogenic differentiation potential of human mesenchymal stem cells in vitro. Cells Tissues Organs (Print) 2012;196:325–338. [PubMed]
Wilkinson A.C., Göttgens B. Transcriptional regulation of haematopoietic stem cells. In: Hime G., Abud H., editors. Transcriptional and Translational Regulation of Stem Cells: Advances in Experimental Medicine and Biology. Springer; 2013. pp. 187–212.