NFIA and GATA3 are crucial regulators of embryonic articular cartilage differentiation

Vertebrate limb skeletogenesis starts as a Sox9-positive mesenchymal condensation that undergoes differentiation to become Col2a1-expressing cartilage. Concomitant with cartilage differentiation, the anlagen is branched and segmented to give rise to most of the skeletal elements of the limb (Shubin and Alberch, 1986; Bi et al., 1999). Except for the cartilage on either side of the plane of segmentation, the rest of the cartilage in these elements is gradually replaced by bone. The cartilage that is replaced by bone is referred to as transient cartilage, while that which remains as cartilage is referred to as permanent or articular cartilage. The Col2a1-expressing chondrocytes present at the center of the cartilage primordium progressively differentiate into prehypertrophic chondrocytes, marked by Ihh expression, followed by maturation into collagen 10 (ColX)-expressing hypertrophic chondrocytes (Karsenty and Wagner, 2002; Archer et al., 2003; Kronenberg, 2003; Pacifici et al., 2005). Recently, our group has demonstrated that the nascent cartilage cells are bi-potential in nature and can give rise to transient or articular cartilage depending on their exposure to either BMP or Wnt signaling, respectively (Ray et al., 2015).

The interzone marks the site of the future joint. It is characterized morphologically by a region of densely packed flattened cells and molecularly by the presence of markers such as Gdf5, Wnt9a and Enpp2 [autotaxin (Atx)], and the absence of the typical cartilage marker Col2a1. The interzone acts as a signaling center and is crucial for joint morphogenesis (Holder, 1977; Archer et al., 2003; Pacifici et al., 2005). Wnt ligands, secreted from the interzone, activate Wnt/β-catenin signaling in the cells immediately adjacent to the interzone and this is necessary for embryonic articular cartilage differentiation (Hartmann and Tabin, 2001; Guo et al., 2004; Später et al., 2006a,b; Ray et al., 2015). c-Jun is a crucial transcriptional activator of Wnt ligands, namely Wnt9a and Wnt16 (Kan and Tabin, 2013). Thus, to date, c-Jun, Wnt9a and β-catenin are the only molecules known to promote embryonic articular cartilage differentiation.

C-1-1, a chicken ERG variant, is the only reported joint-specific transcription factor capable of inhibiting or blocking maturation of transient cartilage into hypertrophic chondrocytes. It should, however, be noted that even though C-1-1 can inhibit transient cartilage differentiation, it has not been reported to induce the ectopic expression of articular cartilage markers other than tenascin C (Iwamoto et al., 2000, 2001, 2007).

Compared with other chondrocyte populations, our understanding of articular cartilage differentiation remains rudimentary despite tremendous progress in recent decades. For instance, we do not have a comprehensive knowledge of the transcription factors that are crucial for articular cartilage differentiation, nor do we know how articular cartilage evades hypertrophy and persists as permanent cartilage (Karsenty and Wagner, 2002). These are key aspects, as articular cartilage is the tissue that is affected, and suspected to undergo transient cartilage differentiation, in osteoarthritis, the most prevalent skeletal disease (Pitsillides and Beier, 2011).

Previously, we reported that the transcription factor NFIA is expressed in chick embryonic articular cartilage (Singh et al., 2016). Here, we identified another transcription factor, GATA3, to be expressed in chick embryonic articular cartilage. We performed gain- and loss-of-function studies for both NFIA and GATA3. Loss-of-function of NFIA resulted in a reduction of the interzone region, while that of GATA3 abrogated articular cartilage differentiation. Gain-of-function of both these molecules blocked hypertrophic cartilage differentiation, but only GATA3 can induce the ectopic expression of several articular cartilage markers. However, our results suggest that GATA3 needs to act in collaboration with other transcription factors for proper articular cartilage differentiation.

In a previous screen, we identified the transcription factor NFIA as expressed in the outer chondrogenic layers (OCLs) of developing chicken articular cartilage from HH28 to HH38 (Singh et al., 2016). In order to investigate the necessity of NFIA in articular cartilage differentiation, we downregulated endogenous NFIA expression utilizing a well-characterized RCAS-based shRNAi (cNFIA-RNAi) against chick NFIA. The specificity and efficacy of this construct in knocking down endogenous chick NFIA have been reported previously (Deneen et al., 2006). Embryonic chick limb buds were infected at HH14 with RCAS virus particles expressing cNFIA-RNAi and harvested at HH36 (Fig. S1, Materials and Methods). Knockdown of NFIA resulted in micromelia of the infected limbs (Fig. S2C,D), as compared with the uninfected contralateral limbs (Fig. S2A,B). Skeletal preparation analysis of cNFIA-RNAi-infected limbs revealed shortening of the elements and unsegmented phalangeal elements (Fig. S2D′, arrows mark the unsegmented joints), as compared with the uninfected contralateral control (Fig. S2B,B′, arrow marks a segmented joint).

For molecular analysis, the region and extent of infection were assessed by immunohistochemistry against one of the viral gag proteins using 3C2 antibody (Fig. 1A). In the uninfected contralateral control joints the chick interzone is visualized as a COL2A1-negative, three-layered structure comprising a central intermediate layer (IL) flanked by two OCLs (Fig. 1B,B′, dotted lines). The presence of such an organized three-layered interzone could not be visualized in cNFIA-RNAi-infected joints (Fig. 1E,E′). Moreover, knockdown of NFIA resulted in a reduction in the COL2A1-negative interzone (compare red line in Fig. 1B,E).

Furthermore, analysis of layer-specific markers by RNA in situ hybridization revealed loss of both the outer and intermediate layers of the interzone following knockdown of NFIA, as demonstrated by downregulation of the IL marker PHLDA2 and OCL marker ChEST302p20 (compare Fig. 1C,C′ with F,F′, and D,D′ with G,G′). In extreme cases, complete loss of the COL2A1-negative domain across some of the infected tibia-tarsus joint was observed (Fig. S3A,A′).

While analyzing the expression of COL2A1 transcripts in cNFIA-RNAi-infected elements we noticed that the downregulation of COL2A1 expression at the center of the elements (Fig. 2A, arrows), which is typically associated with the onset of hypertrophic differentiation, did not take place (compare Fig. 2A,A′). Thus, we investigated the status of hypertrophic differentiation in cNFIA-RNAi-infected elements. RNA in situ hybridization revealed that IHH expression was not downregulated at the center of the element, the putative hypertrophic region (Fig. 2B,B′, yellow and red arrowheads). ColX immunoreactivity could not be detected in cNFIA-RNAi-infected elements (Fig. 2D,D′, yellow and red arrowheads). Since PTHrP-PTHrPR signaling plays a crucial role in hypertrophic differentiation (Kronenberg, 2003), we investigated the mRNA expression of PTHrP and PTHrPR. We could not detect any variation in PTHrP expression (data not shown) but expression of PTHrPR was clearly downregulated in cNFIA-RNAi-infected elements (Fig. 2C,C′, yellow and red arrowheads).

Next, to assess whether NFIA gain-of-function can promote an ectopic or expanded articular cartilage domain, HH14 chick limb buds were infected with RCAS retroviral particles expressing mouse Nfia cDNA (mNfia). The embryos were harvested at HH36. Expression of RCAS-mNfia resulted in micromelia of the infected limbs (Fig. S2E,G) and skeletal preparation analysis revealed a decrease in Alizarin Red staining (Fig. S2H′, asterisk), along with shortened and unsegmented skeletal elements (Fig. S2H,H′, arrows), as compared with uninfected contralateral control limb (Fig. S2F,F′).

To assess molecular changes caused by mNfia expression, we harvested RCAS-mNfia-infected as well as uninfected contralateral control limbs. The region of infection within cartilage was assessed by (1) expression of mNfia transcripts (Fig. 3A′) and (2) 3C2 immunoreactivity (Fig. 3H″). In agreement with observations of unsegmented skeletal elements (Fig. S2H′), we found that upon mNfia infection the expression of several articular cartilage/interzone-specific markers, such as GDF5, BMP4 and SFRP2, was downregulated (Fig. 3B-D′) in the putative joint region (Fig. 3B′-D′, black arrows). Thus, contrary to our expectation, we could not detect ectopic expression of any of the articular cartilage markers in the infected cells.

It should also be noted that, upon mNfia expression, the cells of the putative interzone region often assumed a rounded morphology akin to cells in the epiphysis region (Fig. 3E′, black arrow) as opposed to the densely packed flattened morphology characteristic of interzone cells (Fig. 3E). These molecular and histological changes were particularly apparent in the phalangeal joints, as opposed to more proximal joints. Moreover, these RCAS-mNfia-infected cells also expressed the early transient cartilage marker COL2A1 (Fig. 3F,F′). The cell-autonomous effect of mNfia expression is best exemplified in a joint that is partially infected. In such a joint, in the region that remained uninfected the articular chondrocytes not only maintained their densely packed flattened morphology (Fig. 3E′, red arrow) but also were COL2A1 negative (Fig. 3F′, red arrow), whereas the adjoining infected interzone cells (Fig. 3A′, black arrow) adopted transient cartilage-like morphology (Fig. 3E′, black arrow) and expressed COL2A1 (Fig. 3F′, black arrow).

The interzone or developing articular cartilage is largely non-proliferative, in contrast to adjacent transient cartilage cells (Ray et al., 2015). As expression of mNfia led to downregulation of several articular cartilage markers and to ectopic expression of COL2A1, we determined the status of BMP signaling and cell proliferation in RCAS-mNfia-infected putative interzone cells. We carried out immunohistochemistry for pSMAD1/5/8, as a readout of BMP signaling, and for the mitotic marker phosphohistone H3 (pH3). As expected, we could detect ectopic pSMAD1/5/8 immunoreactivity (Fig. 3G-G″) along with phosphohistone H3 (Fig. 3H,H′) in putative joint sites in mNfia-infected tissues (3C2 immunostaining in Fig. 3H″).

Skeletal analysis of mNfia-infected limbs revealed a decrease in Alizarin Red staining (Fig. S2H,H′). Therefore, we investigated the molecular changes associated with endochondral ossification. Expression of COL2A1 transcripts, which is normally reduced upon onset of hypertrophy, did not decrease from the epiphysis to the diaphysis of RCAS-mNfia-infected cartilage elements (Fig. 4A,A′).

In keeping with this, analysis of IHH and ColX expression revealed that expression of mNfia blocked prehypertrophic chondrocyte differentiation. There was a marked decrease in the expression of IHH transcripts (Fig. 4B′) and ColX protein (Fig. 4C′) in the infected transient cartilage cells as compared with the uninfected contralateral cartilage element (Fig. 4B-C′). Thus, although RCAS-mNfia-infected cells continued to express COL2A1 they did not progress further in the transient cartilage differentiation program and were arrested in an initial chondrogenic state.

Several lines of investigation have demonstrated that both Ihh and Pthrp block hypertrophic differentiation of cartilage and that IHH can induce expression of PTHrP in the periarticular region through a long-range negative-feedback loop (Lanske et al., 1996; Vortkamp et al., 1996; St-Jacques et al., 1999; Dentice et al., 2005; Koziel et al., 2005). Thus, we hypothesized that expression of mNfia might be affecting this IHH/PTHrP regulatory loop. We analyzed the expression pattern of PTHrP in mNfia-expressing chick limbs. As expected, loss of IHH resulted in reduced expression of PTHrP in RCAS-mNfia-infected joints (Fig. 4D,D′, black arrow).

Bonilla-Claudio et al. (2012) recently reported Gata3 to be a direct transcriptional target of BMP signaling during intramembranous ossification. While investigating whether GATA3 is also expressed during endochondral ossification in a BMP signaling-dependent manner, we discovered that GATA3 is in fact expressed specifically in the developing articular cartilage/interzone of chick (Fig. 5A-E) and mouse (Fig. S6A,A′). Further, Gata3 has been suggested to promote the expression of 16 genes that are expressed in the articular cartilage (Table S1) in a previous microarray transcriptomic profiling of whole mammary glands (Kouros-Mehr et al., 2006). Thus, GATA3 was an obvious candidate for a possible role in inducing articular cartilage fate.

In order to study the effects of loss-of-function of GATA3 with respect to chicken embryonic articular cartilage development, we created a dominant-negative version of GATA3 following the strategy of Kamei et al. (2011). Alignment of the amino acid sequences of the chick and mouse GATA3 DNA-binding domains reveals a sequence identity of 100%. Therefore, in order to generate a dominant-negative form of GATA3, we fused the DNA-binding domain of mouse Gata3 to the Engrailed transcriptional repressor domain (mGata3-EnR) (Fig. S4C).

The existing data indicating that loss of Gata3 expression is associated with the downregulation of expression of 16 articular cartilage markers (Kouros-Mehr et al., 2006) (Table S1) suggested that Gata3 plays an instructive role in articular cartilage differentiation. To investigate whether GATA3 is necessary for chick articular cartilage development we expressed mGata3-EnR in the prospective limb field by electroporation of the RCAS-mGata3-EnR construct at HH14. The efficiency of electroporation was assessed by co-electroporating a pCAG-mCherry reporter construct (Fig. S1, Materials and Methods). The embryos were harvested at HH36. Electroporation of embryonic limb buds with RCAS-mGata3-EnR consistently resulted in digit truncation and interdigital syndactyly (Fig. S5C). In extreme cases, we also observed hemangioma/blood pooling at the distal end of an infected limb (Fig. S5B). Whole-mount skeletons of RCAS-mGata3-EnR-electroporated limbs stained with Alcian Blue and Alizarin Red revealed a decrease in Alizarin Red staining (Fig. S5B′, asterisks) along with unsegmented skeletal elements (Fig. S5C′, arrows), as compared with the uninfected contralateral control (Fig. S5A,A′).

To analyze molecular and histological changes brought about by GATA3 loss-of-function, we analyzed sections of mGata3-EnR-infected limbs. The region of infection within cartilage was assessed by expression of mGata3 mRNA (Fig. 5J′) and by 3C2 immunoreactivity (Fig. 5L″,L′′′). Expression of the dominant-negative mGata3-EnR resulted in ectopic expression of the transient cartilage marker COL2A1 (Fig. 5F,F′) and the absence of joint-specific markers SFRP2, ENPP2 and GDF5 (Fig. 5G-I′) in the putative joint region cells (black arrows).

Moreover, we observed that mGata3-EnR-infected interzone cells are mitotically active, as assessed by pH3 immunoreactivity (Fig. 5L,L′), and are pSMAD1/5/8 positive (Fig. 5K,K′), thus mimicking the attributes of transient cartilage. It should be noted that in mGata3-EnR-expressing limbs, normal hypertrophic differentiation was unaffected (data not shown).

Having established that GATA3 loss-of-function results in downregulation of several articular cartilage markers, we next asked whether misexpression of GATA3 could promote an ectopic or expanded domain of articular cartilage markers. To investigate this, we expressed mGata3 cDNA in HH14 limb buds (RCAS-mGata3) and harvested the embryos at HH36. Expression of mGata3 in limb buds did not result in any gross phenotype apparent in the whole mount (data not shown). Thus, we used pCAG-mCherry as an electroporation marker and analyzed limb buds that were efficiently electroporated. We generated sections from these limbs and analyzed for molecular changes. Regions of infection within sections of limb cartilage were assessed by mGata3 RNA in situ hybridization (Fig. 6A′). Analysis of articular cartilage-specific markers revealed mild ectopic expression of c-JUN and WNT9A mRNAs within the domains of mGata3 expression (Fig. 6B-C′). We also observed weak ectopic β-catenin immunoreactivity in cells surrounding the ectopic domains of c-JUN and WNT9A expression (Fig. 6D,D′). Although expression of mGata3 resulted in the ectopic expression of c-JUN and WNT9A, it did not stimulate the ectopic expression of many other articular cartilage markers (Fig. S6B-F).

We speculated that GATA3 alone might not be sufficient to promote articular cartilage differentiation and that it needs other transcriptional collaborators. To investigate this possibility we fused the DNA-binding domain of mGata3 to the VP16 transcriptional activator domain (mGata3-VP16) (Fig. S4B) (Kamei et al., 2011). Such fusion proteins are directed to their cognate binding sites in enhancers by the DNA-binding domain, whereupon they activate transcription under the control of the enhancer irrespective of the presence of co-factors, and thus act as a constitutively active version of the transcription factor.

We introduced the RCAS-mGata3-VP16 construct into HH14 limb buds and harvested the embryos at HH36. Inspection of freshly harvested unfixed, unstained, RCAS-mGata3-VP16-infected limbs revealed micromelia (shortened limb elements), interdigital syndactyly and hematoma/hemorrhage (Fig. S5E), as compared with the uninfected contralateral control limb (Fig. S5D). Whole-mount skeletons of RCAS-mGata3-VP16-infected limbs stained with Alcian Blue and Alizarin Red revealed a decrease in Alizarin Red staining along with shortened, unsegmented skeletal elements (Fig. S5E′), compared with the uninfected contralateral control (Fig. S5D′).

The region of infection within sections of limb cartilage was assessed by 3C2 immunoreactivity (Fig. 7A″-C″). Expression of constitutively active mGata3 resulted in the ectopic expression of four joint-specific markers, namely SFRP2, c-JUN, CCNI and ENPP2 (dotted box, Fig. 6F-I′). Despite the rather broad domains of mGata3-VP16 expression, ectopic expression of joint-specific markers was present only near the putative joint site. The expression of some of the other articular cartilage markers, such as PTHrP, NFIA, PHLDA2 and ChEST302p20, remained unaltered (Fig. 6J-K′; data not shown). The ectopic expression of articular cartilage markers was accompanied with a slight downregulation of COL2A1, a transient cartilage marker, in the mGata3-VP16-infected cells near the future joint region (Fig. 6E,E′).

As stated above, previous analyses have shown that, in contrast to transient cartilage, articular cartilage cells are largely non-proliferative, β-catenin positive and pSMAD1/5/8 negative (Ray et al., 2015). We wanted to assess whether the mGata3-VP16-infected putative transient cartilage cells, which expressed ectopic articular cartilage markers, also attained these attributes of articular cartilage. To assess proliferation, we labeled the dividing cells in the embryo with EdU, a thymidine analog. Akin to articular cartilage cells in the contralateral limb (the region indicated by arrows between dotted lines, Fig. 7C), the number of EdU-positive cells was reduced in mGata3-VP16-infected cells (Fig. 7C′,C″) and, moreover, this was accompanied by ectopic induction of WNT9A (Fig. S6G-J) and β-catenin (Fig. 7A,A′) along with downregulation of pSMAD1/5/8 (Fig. 7B,B′). Overall, these lines of evidence suggest that GATA3 gain-of-function promotes articular cartilage fate at the expense of transient cartilage.

In an uninfected contralateral control cartilage anlagen, COL2A1 expression gradually decreases from the epiphysis towards the center of the cartilage element where hypertrophic differentiation takes place (Fig. 7D). Expression of constitutively active mGata3 blocks such downregulation of COL2A1 expression (Fig. 7D′). Furthermore, mGata3-VP16 arrested the chondrocytes in a prehypertrophic state, as marked by a contiguous band of IHH-expressing cells within the center of the cartilage element (Fig. 7E′, arrows) and blocked further transition to the hypertrophic state as demonstrated by the absence of ColX (Fig. 7F′, arrow); by contrast, in the control, IHH was expressed as two discrete bands (Fig. 7E, arrows) with ColX expressed in the center (Fig. 7F, arrow). These observations are somewhat similar to what was observed in NFIA loss-of-function limbs (compare Fig. 2A,B,D and A′,B′,D′). It should however be noted that although mGata3-VP16 was expressed even in the prehypertrophic region, the articular cartilage markers were not ectopically expressed in these domains.

Transcriptional regulation of articular cartilage differentiation and maintenance is not well understood. Previously, we reported the identification of NFIA as an articular cartilage-specific transcription factor (Singh et al., 2016). In the present report, we identified GATA3 as another transcription factor that is specifically expressed in chick and mouse articular cartilage, and characterized the roles of NFIA and GATA3 during chick articular cartilage differentiation. Our data suggest that although NFIA is not sufficient for articular cartilage differentiation, it is necessary to maintain the permanent fate of articular cartilage through prevention of hypertrophic differentiation. Gain- and loss-of-function experiments demonstrated that GATA3 is necessary for articular cartilage differentiation and for this requires the cooperation of other, unknown activation factors.

Knockdown of chick NFIA reduces or abolishes the COL2A1-negative interzone domain, while misexpression of NFIA prevents hypertrophic differentiation of the transient cartilage cells. Taken together, it appears that NFIA primarily aids in the maintenance of articular cartilage fate by preventing hypertrophic differentiation of interzone/articular cartilage cells.

Nfia is expressed specifically in E16.5 mouse articular cartilage (Fig. S3C,C′). However, in contrast to the effects of knockdown of NFIA in chick embryos, disruption of the Nfia gene in mice did not result in a dramatic phenotype, with only a mild disruption of the organization of articular cartilage (Fig. S3E-H′). This rather mild morphological defect in skeletal and cartilage elements might be due to redundancy (das Neves et al., 1999). Indeed, Nfia has several homologs, including Nfib, Nfic and Nfix, and we also detected expression of Nfix in mouse articular cartilage (Fig. S3D,D′). It will be interesting to analyze the effects of compound knockouts of combinations of these factors in articular cartilage differentiation.

Expression of mNfia in the developing chick limb skeleton promoted ectopic expression of COL2A1 transcripts in the putative interzone/articular cartilage cells. This is paradoxical considering that embryonic articular cartilage does not normally express Col2a1 mRNA. It should be noted that although some of the articular cartilage-specific genes, such as Wnt9a, Wnt4 and Wnt16, are anti-chondrogenic in nature, there are others, such as Gdf5 and Gdf6, which are chondrogenic. Ectopic chondrogenesis and abrogation of articular joint formation upon overexpression of NFIA is very similar to what is observed with GDF5 overexpression (Francis-West et al., 1999; Merino et al., 1999; Storm and Kingsley, 1999; Tsumaki et al., 1999). However, Gdf5 misexpression is not reported to cause a block in hypertrophy. C-1-1 was proposed to be a Gdf5 target and its misexpression causes a block in hypertrophy similar to that which we observe with NFIA misexpression, but the effect of C-1-1 misexpression on ectopic chondrogenesis (ectopic Col2a1 expression) has not been reported (Iwamoto et al., 2001, 2000). Finally, just like NFIA, both gain- and loss-of-function manipulations of GDF5 cause abrogation of joint formation. The broad similarity in articular cartilage developmental defects upon gain- or loss-of-function manipulations of GDF5, ERG or NFIA raises the question as to whether these genes functionally interact during normal articular cartilage development (Fig. 8).

Germline deletion of Gata3 in mice results in embryonic lethality by the time joint development initiates in the limbs, precluding analysis of its role in articular cartilage differentiation (Pandolfi et al., 1995). A limb-specific knockout of Gata3 needs to be generated. In the absence of such a mouse model, we took advantage of the chick embryonic system. We made several attempts to knockdown GATA3 by miRNA, as we did for NFIA, but without success. We thus generated the mGata3-EnR construct, which serves as a dominant-negative form by repressing transcription of GATA3 downstream genes. Such a strategy has previously been used for Gata4 (Kamei et al., 2011). Infection with mGata3-EnR in chicken limbs resulted in a spectrum of malformations, with some of the characteristics similar to those of Gata3 knockout mice, such as pooling of blood, hemorrhages and smaller size of the limb elements (Pandolfi et al., 1995). It is interesting that both gain- and loss-of-function of GATA3 resulted in inhibition of certain joint formation. In this context, it should be noted that Macias et al. (1993) reported that application of retinoic acid (RA) in the interdigital mesenchyme resulted in inhibition of joint formation. It is known that many of the GATA factors, including Gata4/6, are transcribed in an RA-dependent manner (Arceci et al., 1993; Mauney et al., 2010). Thus, it would be interesting to investigate if RA can induce Gata3 in early limb skeletal anlagen and whether that affects joint induction. We speculate that GATA3 loss-of-function induces an ectopic BMP signaling domain, thus eliminating a permissive environment for joint induction (Ray et al., 2015).

In the 50 kb genomic region upstream of the mouse and human Gata3 genes, five conserved regions containing ten putative TCF/Lef binding sites were identified by bioinformatics analysis (Grote et al., 2008). Therefore, Gata3 transcription is speculated to be regulated by Wnt signaling through β-catenin. Misexpression of Wnt9a or β-catenin in the limb results in ectopic expression of articular cartilage markers such as Enpp2 (Atx) and Sfrp2 (Hartmann and Tabin, 2001). Interestingly, we observed that expression of mGata3 promotes ectopic expression of c-JUN, WNT9A and β-catenin. This suggests that GATA3 is not only downstream of Wnt/β-catenin signaling, but also maintains it in a feedback loop via c-JUN. Furthermore, since c-Jun is known to promote the expression of Wnt ligands (Kan and Tabin, 2013), ectopic expression of Wnt9a and β-catenin might be an indirect effect of c-Jun activation by Gata3. Moreover, we observed that although mGata3 could induce the ectopic expression of a few articular cartilage-specific genes, more articular cartilage-specific genes were ectopically expressed when a constitutively active version of GATA3 was expressed (mGata3-VP16). Since VP16 fusion allows a transcription factor to activate transcription even in the absence of co-factors, we speculate that GATA3, in addition to activating expression of c-JUN, acts in conjunction with other transcription factors to promote articular cartilage differentiation. We have routinely observed that GATA3 gain-of-function induces the ectopic expression of articular cartilage-specific genes only in the vicinity of the interzone and not in cells at a distance from it. This suggests that only these cells are competent, whereas the cells distant from the joint site are not competent, to upregulate the expression of articular cartilage-specific genes even upon the expression of constitutively activate mGata3, further underscoring the importance of context and/or the presence of other transcription factors.

As of now, three transcription factors expressed in the joint interzone, namely the C-1-1 variant of ERG, OSR1/2 and c-JUN, have been identified to play important roles during articular cartilage differentiation (Iwamoto et al., 2000, 2001, 2007; Gao et al., 2011; Kan and Tabin, 2013). C-1-1 misexpression prevents hypertrophic differentiation but does not promote ectopic expression of articular cartilage-specific genes other than tenascin C (Iwamoto et al., 2000, 2007). However, articular cartilage-specific ablation of Erg does not affect embryonic articular cartilage differentiation but leads to osteoarthritis-like phenotypic defects in 11-month-old mice (Ohta et al., 2015). c-Jun, presumably owing to its ability to promote transcription of the Wnt ligands Wnt9a and Wnt16, is necessary for articular cartilage differentiation (Kan and Tabin, 2013). However, as no gain-of-function study has been conducted, it is unclear if c-JUN is sufficient for articular cartilage differentiation. c-JUN loss-of-function spares specification of the interzone. OSR1/2 compound knockout abrogates joint formation and promotes ectopic transient cartilage differentiation in the putative joint region (Gao et al., 2011). Nonetheless, since OSR1/2 loss-of-function reduces the expression of Wnt ligands and since β-catenin loss-of-function did not affect OSR1/2 expression, it is likely that OSR1/2 acts upstream of Wnt ligand expression and collaborates with transcriptional targets of the Wnt signaling pathway to promote articular cartilage differentiation (Gao et al., 2011). Since OSR1/2 misexpression studies have not been carried out, it is difficult to assess whether OSR1/2 can induce articular cartilage fate like GATA3. By conducting both loss- and gain-of-function analyses we have demonstrated that NFIA and GATA3 act at two different levels during articular cartilage differentiation. NFIA, similar to C-1-1, acts by preventing hypertrophic differentiation in the articular region, whereas GATA3 promotes articular cartilage fate by activating the expression of several articular cartilage-specific genes. However, to do this, GATA3 needs to collaborate with other transcription factors. It remains to be seen whether OSR1/2 and c-JUN are such transcription factors (Fig. 8).

Gain- or loss-of-function of NFIA and gain-of-function of GATA3 affected not only articular cartilage differentiation but also hypertrophic differentiation, albeit at different levels. The role of an Ihh/Pthrp feedback loop in controlling hypertrophic differentiation is well established (Kronenberg, 2003). However, it is unlikely that these are the only two players involved in the process. Ihh loss-of-function causes absence of articular joints (St-Jacques et al., 1999; Koyama et al., 2007). This is unlikely to be due to the effect of Ihh on Pthrp expression, as Pthrp is not known to promote articular cartilage differentiation. Rather, it is probable that loss-of-function of Ihh is affecting some factor(s) that is directly or indirectly responsible for articular cartilage differentiation. Unfortunately, we could not find any data in the literature that describe changes in Ihh expression upon gain or loss of ERG function but, interestingly, loss of IHH does not cause loss of ERG variant C-1-1 expression (Koyama et al., 2007). In keeping with this, gain-of-function of C-1-1 blocks hypertrophic differentiation but does not affect articular cartilage differentiation, whereas NFIA gain-of-function not only blocks hypertrophic differentiation but also affects articular cartilage differentiation. In the context of medulla blastoma, Ptch1 is reported to regulate Nfia expression (Genovesi et al., 2013). Thus, it is possible that NFIA is one of the regulators of the crosstalk between articular cartilage and hypertrophic cartilage differentiation. It is interesting to note that loss of GATA3 function did not affect hypertrophy, whereas loss of NFIA did. As NFIA and GATA3 seem to act at different stages of articular cartilage differentiation, it is possible that GATA3 loss-of-function blocks articular cartilage differentiation but does not affect expression of the factors that control hypertrophic differentiation, e.g. PTHrP and/or IHH. GATA3 gain-of-function did not affect NFIA expression (Fig. 6) but causes ectopic articular cartilage differentiation as well as block in hypertrophy. However, our data cannot rule out the possibility that these are two unconnected events. Unfortunately, we could not obtain any embryos in which mGata3-VP16 was expressed only in hypertrophic cartilage and not in articular cartilage.

In summary, it appears that a network involving GATA3, OSR1/2, c-JUN and Wnt ligands regulates articular cartilage differentiation, that another network involving NFIA, GDF5 and ERG prevents transient cartilage differentiation in the articular cartilage domain, and that a third network involving IHH, PTHrP, NFIA and C-1-1 controls the crosstalk between articular cartilage differentiation and hypertrophy. The network that controls articular cartilage differentiation is becoming clearer. However, the mechanisms of action of the network that prevents ectopic hypertrophy in the articular cartilage domain and of that which controls crosstalk between articular cartilage and hypertrophic cells remain enigmatic. Our study provides potential new avenues for investigation into these aspects.

Fertilized White Leghorn chicken eggs were obtained from the Central Avian Research Institute of India, Chandra Shekhar Azad Agricultural University (Kanpur, UP, India) and Ganesh Enterprises (Nankari, Kanpur, UP, India). Eggs were incubated at 38°C in a humidified chamber to be treated and/or harvested at specific stages of development as assessed by Hamburger and Hamilton (HH) staging criteria (Hamburger and Hamilton, 1951).

Mouse experiments were conducted as per protocol approved by the Institute Animal Ethics Committee (registration number CPCSEA-56/1999). Nfia^−/+ male and female mice were mated, embryos were harvested at E16.5 and genotyped (das Neves et al., 1999) to identify Nfia^−/− and wild type (Nfia^+/+).

Chicken embryos were harvested and eviscerated in PBS and fixed in 95% ethanol for at least 2-3 days, followed by overnight fixation in 100% acetone. Next, the tissues were stained for 2-3 days in a 1:1:1:17 volume mixture of 0.3% Alcian Blue 8GX (Sigma-Aldrich) in 95% ethanol:0.1% Alizarin Red in 70% ethanol:glacial acetic acid:70% ethanol. Post-staining, the tissues were cleared in 1% KOH and photographed under a dissection microscope.

Chick embryonic limbs were dissected and fixed overnight in 4% paraformaldehyde (PFA) at 4°C, embedded in paraffin and 5-10 µm sections cut along the parasagittal plane using a microtome.

cDNA clones used to make digoxigenin-labeled antisense riboprobes generated by in vitro transcription are detailed in Table S2. RNA in situ hybridization was performed as described previously (Singh et al., 2016). The RNA in situ hybridization signals for a given probe on control and test sections were developed on the same slide.

For pSMAD1/5/8, β-catenin, noggin and GFP immunohistochemistry, sections were processed and detected as described in Ray et al. (2015).

Sections were deparaffinized and rehydrated in PBS, followed by post-fixation in 4% PFA. Sections were incubated overnight at 4°C with the following primary antibodies: anti-pH3 (Sigma-Aldrich, H0412; 1:100), anti-β-catenin-Cy3 (Sigma-Aldrich, C7738; 1:200), anti-GAG (3C2; Potts et al., 1987), anti-pSMAD1/5/8 (Cell Signaling, 9511; 1:100) and anti-ColX (DSHB hybridoma product X-AC9, deposited by T. F Linsenmayer; 1:20). Following this step, tissues were washed in PBT (PBS containing 0.1% Tween 20) and incubated with the respective secondary antibodies, including: DyLight 549 AffiniPure goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, 111-505-003; 1:200) for anti-pSMAD1/5/8 and anti-pH3; Alexa Fluor 488-conjugated anti-mouse IgG (Jackson ImmunoResearch Laboratories, 115-545-003; 1:250) for 3C2; and DyLight 549-conjugated anti-mouse IgG (Jackson ImmunoResearch Laboratories, 115-505-003; 1:250) for anti-ColX. The tissues were counterstained with DAPI and mounted in VECTASHIELD antifade reagent (Vector Laboratories, H-1000).

EdU (E10415, C10083, Invitrogen) labeling and detection were performed as previously described (Ray et al., 2015).

DF1 (Himly et al., 1998) chicken embryo fibroblast cells were checked to ensure they were free of contamination. Cells were transfected with the RCAS-HA-mNfia construct obtained from Benjamin Deneen (Deneen et al., 2006). Virus particles were produced, concentrated and titered as described previously (Logan and Tabin, 1998). HH10 embryos were lowered within the egg by removing 2-3 ml of albumin and a window made through which a viral solution of titer 1×10⁸ IU/ml mixed with 1% Fast Green was injected into the lateral plate mesoderm (LPM), which is destined to give rise to the prospective hindlimb of the embryo.

Chick embryos were initially lowered by removing 2-3 ml of albumin and a window made to visualize the embryo under the vitelline membrane. This vitelline membrane was shorn near the hindlimb field and bathed in 100 µl sterile PBS+Pen Strep solution (Thermo Fisher Scientific, 10378016). At HH14, RCAS-HA-mNfia, RCAS-cNFIA shRNAi (target region within NFIA coding sequence: 5′-GCCATCGCCAACTGCATTAAA-3′) obtained from Benjamin Deneen (Deneen et al., 2006), RCAS-GATA3-DBD-VP16 or RCAS-GATA3-DBD-EnR mixed with 0.5 µg/µl pCAG-mCherry and 0.1% Fast Green was injected into the embryonic space between the somatic LPM and splanchnic LPM at 2 µg/µl using a microinjector. Electroporation was performed as previously described (Suzuki and Ogura, 2008).

We are grateful to Prof. Andrew Lassar (Harvard Medical School, USA) for sharing full-length mouse GATA3 construct and to Prof. Benjamin Deneen (Baylor College of Medicine, Houston, TX, USA) for providing the RCAS-mNfia and RCAS-cNFIA shRNAi constructs. We thank Prof. Richard Gronostajski for permission to obtain Nfia mutant mice from Prof. Shubha Tole.