Shox2 regulates progression through chondrogenesis in the mouse proximal limb

The vertebrate limb can be divided into three discrete domains along the proximodistal (PD) axis, each distinguishable by one or more morphologically distinctive skeletal elements. The stylopod, characterized by the humerus in the forelimb and the femur in the hindlimb, is the most proximal limb domain, situated adjacent to the body wall. The zeugopod is located medially and consists of the radius and ulna in the forelimb and the tibia and fibula in the hindlimb. The autopod, more commonly known as the hand or foot in humans, is the distalmost limb segment and is comprised of numerous carpal, metacarpal, tarsal, metatarsal and phalangeal skeletal elements. Members of paralogous groups 9 to 13 of the HoxA, C and D clusters play important roles regulating the formation and specifying the unique PD identities of these three limb compartments (reviewed in Zakany and Duboule, 2007; Zeller et al., 2009). Interestingly, targeted disruptions of limb Hox genes have resulted in the compartment-specific genetic ablation of all limb skeletal elements except one, the humerus. In fact, mice lacking all HoxA and D gene function in their forelimbs fail to develop autopodial and zeugopodial skeletal elements, but still generate the proximal third of the humerus (Kmita et al., 2005), a finding indicative of the involvement of an ancillary non-Hox factor in forelimb stylopodial patterning.

Even though HoxA, C and D paralogous group members can account for the formation and identity specification of the three limb segments, it is endochondral ossification that is the principal engine driving elongation of each individual limb compartment. All vertebrate limb skeletal elements are generated via endochondral ossification – the multistep process by which cartilage templates are established, enlarged, and progressively replaced by bone. More explicitly, somatopleuric mesenchyme migrates into the emerging limb bud where it undergoes condensation into compact cellular aggregates that prefigure the future skeletal elements. Inside these condensations, undifferentiated mesenchymal cells upregulate expression of the transcription factor Sox9 and commence chondrogenesis. Newly differentiated chondrocytes produce the characteristic cartilage extracellular matrix (ECM) components aggrecan (Acan) and collagen type II (Col2a1) and proliferate to expand the size of the nascent cartilage anlage. Prompted by Runx2 expression, proliferating chondrocytes at the center of the template exit the cell cycle and initiate hypertrophic differentiation. Hypertrophic chondrocytes express the unique marker collagen type X (Col10a1) and experience a massive increase in cytoplasmic volume. Perichondrial cells and chondrocytes at the ends of the cartilage template express Pthlp, which stimulates chondrocytes to proliferate. As these proliferating growth plate chondrocytes leave the domain of PTHLP influence, they mature into prehypertrophic chondrocytes and express Ihh. It is the enormous increase in hypertrophic chondrocyte volume and the columns of rapidly proliferating chondrocytes that are responsible for driving longitudinal bone growth. Finally, hypertrophic chondrocytes express genes like Bglap, Ibsp and Spp1 in order to mineralize their ECM before undergoing apoptosis (reviewed in Kronenberg, 2006; Solomon et al., 2008; Wuelling and Vortkamp, 2010).

In humans, functional loss of the pseudoautosomal SHOX gene is causally linked to several osteochondrodysplasias typified by short stature. Specifically, SHOX haploinsufficiency causes the majority of Leri-Weill dyschondrosteosis cases, a small percentage of idiopathic short stature cases, and the skeletal phenotype associated with Turner syndrome. SHOX nullizygosity is responsible for a rare skeletal dysplasia known as Langer mesomelic dysplasia. Although the short stature phenotypes caused by SHOX deficiency are highly variable, their hallmark feature is mesomelia, or disproportionate shortening of the fore- and hindlimb zeugopods (reviewed in Binder, 2011; Blaschke and Rappold, 2006; Marchini et al., 2007). Importantly, this limb compartment-specific effect spatially corresponds with the zeugopodial expression domain of SHOX in embryonic human limb buds (Clement-Jones et al., 2000). Histological analyses of disorganized proliferative zone chondrocytes in the growth plates of individuals affected by Leri-Weill dyschondrosteosis (Munns et al., 2001) and Langer mesomelic dysplasia (Evans et al., 1988) have led to the hypotheses that loss of SHOX function might result in premature maturation of resting zone chondrocytes into proliferating chondrocytes, precocious differentiation of proliferating chondrocytes into hypertrophic chondrocytes, and accelerated growth plate fusion (Munns et al., 2004).

Although mice lack a SHOX ortholog, they do possess the autosomal paralogous gene Shox2 (Blaschke et al., 1998; Rao et al., 1997). In a mouse model of SHOX deficiency, Prrx1-Cre-driven conditional deletion of Shox2 in embryonic limb mesenchyme results in near complete absence of the humerus and femur. Mechanistically, a significant delay in chondrocyte hypertrophy, as evidenced by depressed expression of Runx2 and Col10a1 in the developing stylopodial skeleton, is responsible for the drastic rhizomelic limb shortening (Cobb et al., 2006), an effect reminiscent of the delayed hypertrophy underlying Hoxa11/d11 disruption-induced mesomelia (Boulet and Capecchi, 2004). It has been hypothesized that aberrant paracrine signaling by non-chondrogenic mesenchyme surrounding the developing stylopodial skeletal elements might cause the delay in Runx2 expression and subsequent chondrocyte hypertrophy (Yu et al., 2007). We wished to determine whether Shox2 expressed by chondrocytes plays a role in limb skeletal development and patterning. We found that Col2a1-Cre-driven conditional deletion of Shox2 in developing chondrocytes results in shortening of the fore- and hindlimb stylopods and that this reduction in humeral and femoral length is the result of precocious chondrocyte hypertrophy. Although chondrocyte-specific deletion of Shox2 results in stylopodial rather than zeugopodial shortening, we contend that Col2a1-Cre-driven Shox2 deletion represents a potential mouse model of SHOX deficiency that accurately mimics a current mechanistic hypothesis by which loss of SHOX gene function results in human short stature syndromes.

In order to determine the role of Shox2 expressed by chondrocytes in endochondral ossification, we employed a Col2a1-Cre driver to delete Shox2 in developing chondrocytes concomitant with the expression of collagen type II. We analyzed the sizes and shapes of endochondral skeletal elements in newborn and four-week-old animals and found that conditional deletion of Shox2 in chondrocytes results in significant shortening of the humerus and femur (Fig. 1). Although these elements were shorter, the morphologies of Col2a1-Cre Shox2 mutant humeri and femora were normal, indicating that while Shox2 expressed by chondrocytes affects longitudinal bone growth in the stylopod, it does not affect patterning of the stylopodial skeleton. The lengths and morphologies of fore- and hindlimb zeugopodial and autopodial skeletal elements were unaffected by chondrocyte-specific deletion of Shox2 (Fig. 1). Moreover, we could not detect conditional Shox2 deletion-induced aberrations in the sizes or shapes of any other endochondral skeletal elements, nor in mouse weight or crown-rump length (data not shown). Col2a1-Cre Shox2 mutant mice are able to procreate and exhibit mobility comparable with littermate controls, though a close examination of four-week-old live animals reveals noticeable stylopodial length reduction (supplementary material Movie 1).

We employed the embryonic limb bud micromass culture model system in order to determine the mechanism by which conditional Shox2 deletion in chondrocytes causes stylopodial shortening. We initially generated micromass cultures from E12.5 limb buds because at this gestational time point, mutant stylopods have yet to experience any length reduction (Fig. 2A). After 7 days of culture, E12.5 Col2a1-Cre Shox2 mutant limb bud micromasses developed fewer, smaller cartilaginous nodules than controls. These mutant cultures stained weakly with Alcian blue and were only faintly collagen type II immunopositive. In contrast, mutant cultures were strongly immunopositive for collagen type X (Fig. 2B). Quantitative real time PCR showed that relative to controls, Col2a1-Cre Shox2 mutant cultures express significantly lower levels of genes involved in early chondrogenesis, like Sox9, Sox6, Acan and Col2a1, and chondrocyte maturation, like Runx2 and Ihh, but significantly higher levels of the hypertrophic marker Col10a1 (Fig. 2C; supplementary material Fig. S1). Based on micromass morphology and gene expression, we hypothesized that Col2a1-Cre Shox2 mutant cultures might be accelerating the early chondrogenic and chondrocyte maturation phases of cartilage formation in order to undergo premature hypertrophic differentiation.

Runx2 is known to positively regulate expression of both Ihh and Col10a1 (Kim et al., 1999; Yoshida et al., 2004). Thus, we wished to determine whether at an earlier developmental time point, Col2a1-Cre Shox2 mutant cultures would show elevated transcript levels of Runx2 and Ihh, in addition to Col10a1. Real time PCR revealed that by E12.5, mutant forelimb buds already express ∼2.4-times more Col10a1 transcripts than controls (supplementary material Fig. S2), a gestational time point at which Col10a1 is detected only in the stylopod of the developing forelimb (St-Jacques et al., 1999), if it can be detected at all (Maye et al., 2011). So, we prepared micromass cultures from E11.5 buds in order to monitor chondrocyte maturation from an earlier time point. We found that by culture day 3, E11.5 Col2a1-Cre Shox2 mutant limb bud micromasses had already downregulated expression of Sox9, Sox6, Acan and Col2a1 while significantly upregulating expression of Runx2, Ihh and Col10a1 (Fig. 2D,E; supplementary material Fig. S3). Finally, we used real time PCR to analyze the expression of genes involved in mineralization of the ECM by hypertrophic chondrocytes. We discovered that at culture day 7, E11.5 Col2a1-Cre Shox2 mutant micromasses express significantly higher levels of Bglap, Ibsp and Spp1 mRNA transcripts than controls (Fig. 2F). Taken together, these data demonstrate that chondrocytes from mutant limbs progress through the normal steps of chondrogenic differentiation, but do so in a significantly accelerated manner.

We used real time PCR to compare transcript levels of a variety of chondroregulatory growth factors between E11.5 Col2a1-Cre Shox2 mutant and control micromass cultures (Fig. 3A and data not shown). We found that chondrocyte-specific Shox2 deletion significantly increased levels of Bmp2 and 4 mRNAs, but had no effect on any of the other growth factors analyzed, including Bmp7, Gdf5, Wnt5a and Fgf10. We hypothesized that BMP signaling might be responsible for the precocious chondrocyte maturation, hypertrophy and ECM mineralization observed in Col2a1-Cre Shox2 mutant micromasses. To test this hypothesis, we treated E11.5 Col2a1-Cre Shox2 mutant micromass cultures with the BMP signaling antagonist Noggin. BMPs are required for the mesenchymal condensation step that precedes overt chondrogenesis (Capdevila and Johnson, 1998; Pizette and Niswander, 2000), so we delayed the onset of Noggin treatment 36 hours from initial micromass spotting. Moreover, we employed Noggin at a concentration of 10 ng/ml in order to avoid any inhibitory effects on basal chondrogenic gene expression (supplementary material Fig. S4). 10 ng/ml Noggin treatment completely blocked the premature chondrocyte maturation, hypertrophy and ECM mineralization observed in E11.5 Col2a1-Cre Shox2 mutant cultures, as indicated by collagen type X and IBSP immunohistochemistry, western blotting for RUNX2, and real time PCR for Runx2, Ihh, Col10a1, Bglap, Ibsp and Spp1 (Fig. 3B–D).

We have used the micromass culture model system to show that chondrocyte-specific deletion of Shox2 results in precocious chondrocyte maturation and hypertrophy. We next sought to determine whether our in vitro mechanism could be responsible for stylopodial shortening in vivo, so we performed in situ hybridization (ISH) on E12.5, E13.5 and E14.5 control and Col2a1-Cre Shox2 mutant limbs. As expected (Isshiki et al., 2011; Yu et al., 2007), we detected endogenous expression of Shox2 mRNA throughout the humeral anlage, except in the hypertrophic zone, where expression of Shox2 transcripts is noticeably downregulated (Fig. 3E). As previously reported (Cobb et al., 2006; Yu et al., 2007), we also observed lower levels of Shox2 transcripts in humeral chondrocytes relative to surrounding mesenchyme (Fig. 3E,F), indicating that a decrease in Shox2 expression might be associated with the onset of early chondrogenesis, in addition to later stages of chondrocyte differentiation. Importantly, we found that Col2a1-Cre Shox2 mutant humeri exhibit precocious formation of a central Col10a1-positive zone at E12.5 and both expansion and premature central separation of the Col10a1-positive region at E14.5, thereby demonstrating that Col2a1-Cre Shox2 mutant stylopodial chondrocytes do, in fact, undergo precocious maturation and hypertrophy in vivo (Fig. 3F). Accordingly, the Alizarin red-positive mineralized zone comprises a greater proportion of the E16.5 humerus in Col2a1-Cre Shox2 mutants than in controls (supplementary material Fig. S5A). We also detected Shox2 transcripts in the zeugopodial cartilage anlagen of both fore- and hindlimbs in an expression pattern similar to that observed in the stylopodial elements (supplementary material Fig. S5B). However, in this limb segment, Col2a1-Cre-driven deletion of Shox2 did not lead to precocious upregulation of Col10a1 expression (data not shown). This result was not unexpected as Col2a1-Cre Shox2 mutants do not exhibit zeugopodial shortening (see Fig. 1).

Among the growth factors we tested, Bmp4 is the most highly upregulated in micromass cultures of Col2a1-Cre Shox2 mutant limb bud cells (see Fig. 3A), and increased Bmp4 expression has been detected in the stylopodial mesenchyme of Shox2 knockout mouse limbs (Yu et al., 2007). Therefore, we examined the humeral expression pattern of Bmp4 using ISH. In accordance with earlier reports of Bmp4 expression in the growth plate (Isshiki et al., 2011; Shu et al., 2011), we found that Bmp4 is expressed weakly in the prehypertrophic zone and more strongly in the hypertrophic region of the E13.5 humerus (Fig. 3E). Col2a1-Cre-driven Shox2 deletion resulted in strong Bmp4 expression throughout the E12.5 mutant humerus, which contrasted starkly with the Bmp4 expression domain in control humeri where transcripts were detected faintly and only in the prehypertrophic zone (Fig. 3F). Bmp2 is known to be strongly expressed by hypertrophic chondrocytes (Isshiki et al., 2011; Shu et al., 2011). Therefore, it was not surprising when Col2a1-Cre-driven conditional Shox2 deletion resulted in strong expression of Bmp2 in the precociously formed hypertrophic zone of E12.5 mutant humeri, which contrasted with control humeri that lacked both strong Bmp2 expression and any indication of hypertrophic differentiation (supplementary material Fig. S5C).

Prrx1-Cre-driven conditional deletion of Shox2 in embryonic limb mesenchyme causes more severe shortening of the stylopodial skeletal elements than Col2a1-Cre-driven Shox2 deletion (compare Fig. 4A,B with Fig. 1A,B; see supplementary material Movie 1). In contrast to Col2a1-Cre Shox2 mutants, rhizomelia in Prrx1-Cre mutant mice is due to significantly delayed expression of Runx2 and Col10a1 in the stylopodial skeletal anlagen (Cobb et al., 2006). When we generated micromass cultures from E12.5 Prrx1-Cre Shox2 mutant limb buds, we found, expectedly, that mutant cultures express high levels of genes involved in early chondrogenesis and low levels of genes involved in chondrocyte maturation, hypertrophy, and ECM mineralization (Fig. 4C,D; supplementary material Fig. S6). Unexpectedly, we discovered that unlike E12.5 limb bud control cultures that are characterized by cartilage nodules separated by Alcian blue-, collagen II-negative non-cartilaginous tissue, Prrx1-Cre mutant cultures form a more uniform sheet of cartilage with very few internodular spaces (Fig. 4C). We analyzed expression of transcription factors involved in tenogenesis and myogenesis in order to determine if the ectopic internodular cartilage that forms in Prrx1-Cre mutant cultures does so at the expense of tendon and muscle cells normally present in these internodular regions (Asou et al., 2002; Swalla and Solursh, 1986). Real time PCR revealed that relative to controls, E12.5 Prrx1-Cre Shox2 mutant cultures express significantly lower levels of the tenogenic marker gene Scx and the myogenic markers Myf5, Myod1 and Myog (Fig. 4E).

At E12.5, the cartilage anlagen of the developing limb already stain strongly with Alcian blue, indicating production of cartilage ECM proteoglycans by differentiated chondrocytes (see Fig. 2A). Moreover, the most obvious chondrostimulatory effect of Prrx1-Cre-driven conditional Shox2 deletion involved normally non-chondrogenic internodular cells and not nodular chondrocytes. We wished to determine if Shox2 deletion could also enhance chondrogenesis of undifferentiated limb bud mesenchymal progenitor cells (MPCs), so we generated micromass cultures from the distal subridge region of E11.5 limb buds (Zhang et al., 2004). Following 7 days of culture, Prrx1-Cre Shox2 mutant MPC micromasses developed significantly larger cartilaginous nodules that stained more intensely with Alcian blue and expressed noticeably more collagen type II than controls. Though, Shox2-deficient MPC cultures still failed to progress through early chondrogenesis to hypertrophic differentiation, as indicated by almost undetectable collagen type X accumulation (Fig. 4F).

We have shown that E12.5 Prrx1-Cre Shox2 mutant limb bud cultures form sheets of cartilage and that these elevated levels of cartilage tissue production occur at the expense of internodular tenogenesis and myogenesis. Yu et al. (Yu et al., 2007) have reported elevated Bmp4 levels in non-chondrogenic stylopodial mesenchyme of Shox2 knockout mouse limbs. In vivo, this ectopic Bmp4 expression does not result in chondrogenic differentiation of these normally non-chondrogenic cells. However, recent work by Cooper et al. (Cooper et al., 2011) demonstrates that when embryonic chick limb bud cells are cultured in micromass, limb positional information is gradually lost unless the cells are exposed to PD patterning signals, like retinoic acid (RA), FGF8, and WNT3A. We hypothesized that increased Bmp4 expression in Prrx1-Cre Shox2 mutant micromass internodular regions might induce these normally non-chondrogenic cells to undergo chondrogenesis when limb positional information is diminished.

We first performed real time PCR to determine whether E12.5 Prrx1-Cre Shox2 mutant limb bud micromass cultures express higher levels of Bmp4 transcripts than controls. We found that like Col2a1-Cre mutant cultures, Prrx1-Cre mutant micromasses express significantly elevated levels of both Bmp2 and 4 (supplementary material Fig. S7). Whole mount ISH revealed that E12.5 limb bud control cultures display strong nodular expression of Bmp2 and 4, whereas Prrx1-Cre Shox2 mutant micromasses show strong expression of these Bmp transcripts in both nodular and internodular regions (Fig. 5A and data not shown). We then employed real time PCR to demonstrate that compared to intact E12.5 limb buds, mRNA transcript levels of limb patterning genes are significantly reduced in E12.5 limb bud micromasses over a 7-day culture period (Fig. 5B; supplementary material Fig. S8). Moreover, we found that Prrx1-Cre Shox2 mutant micromasses did not start accumulating Alcian blue-, collagen II-positive cartilage tissue in their internodular spaces until culture day 5 (supplementary material Fig. S9), the time point at which expression of Meis1 and 2 finally tapers off in E12.5 limb bud micromass cultures (supplementary material Fig. S8B).

We wished to determine whether increased BMP levels were, in fact, responsible for internodular chondrogenesis in mutant cultures. We treated mutant micromasses with 10 ng/ml Noggin for the last 3 days of a 7-day culture period and found that Noggin completely blocked the internodular chondrogenesis normally observed in Prrx1-Cre Shox2 mutant cultures (Fig. 5C). We also found that we could mimic the internodular chondrogenesis observed in these cultures by treating wild-type micromasses with 25 ng/ml BMP4 for the last 3 days of a 7-day culture period (Fig. 5D). Next, we attempted to determine whether maintenance of proximal limb patterning information in micromass culture could prevent Shox2 deletion-induced internodular chondrogenesis. Like Cooper et al. (Cooper et al., 2011), we used RA to proximalize our limb bud micromasses. We found that a low 5 nM RA dose delivered throughout the final 6 days of a 7-day micromass culture period had no effect on nodular chondrogenesis or Col2a1 expression, but was able to significantly increase transcript levels of the proximal limb marker genes Meis1 and 2 (Fig. 5E,F; supplementary material Fig. S10). Importantly, treatment with 5 nM RA completely abrogated internodular chondrogenesis in Prrx1-Cre mutant cultures (Fig. 5F). Moreover, transfection of Prrx1-Cre Shox2 mutant limb bud cells with a Meis1 expression plasmid prior to micromass spotting also blocked the internodular chondrogenesis normally observed in mutant cultures (supplementary material Fig. S11).

We noticed that both enhancement of early chondrogenesis in Prrx1-Cre mutant micromass cultures (see Fig. 4) and precocious chondrocyte maturation in Col2a1-Cre mutant cultures (see Figs 2, 3) were dependent on BMP activity. We hypothesized that moderate increases in BMP signaling caused by decreased Shox2 levels trigger early chondrogenesis, whereas for chondrocyte maturation, hypertrophy and ECM mineralization to occur, even higher levels of BMP activity are required. So, we theorized that exogenous BMP application should overcome the block to chondrocyte maturation observed in Prrx1-Cre mutant cultures. As shown in Fig. 5G, BMP2 treatment overcame the depressed chondrocyte maturation, hypertrophy and ECM mineralization characteristic of Prrx1-Cre mutant cultures. Furthermore, Runx2, Ihh, Col10a1 and Bglap were significantly more responsive to BMP treatment in Shox2-deficient cultures than in controls.

We considered the possibility that elevated Noggin levels in Prrx1-Cre Shox2 mutant chondrocytes might contribute to the delay in maturation and hypertrophy observed in these cells. Real time PCR revealed that E12.5 Prrx1-Cre Shox2 mutant limb buds and micromasses express significantly higher levels of Nog mRNA transcripts than controls (supplementary material Fig. S12A). Importantly, the magnitude of Nog transcript increase over control levels was greater in Prrx1-Cre than Col2a1-Cre mutants (supplementary material Fig. S12A,B), even though Prrx1-Cre mutants show more moderate relative increases in Bmp2 and 4 transcripts than Col2a1-Cre mutants (see Fig. 3A; supplementary material Fig. S7). Therefore, it is possible that elevated Noggin levels might counteract the modest increase in BMP signaling observed in Prrx1-Cre Shox2 mutants, thereby blocking the transition from early chondrogenesis to chondrocyte maturation and hypertrophy. Conversely, in Col2a1-Cre mutants, the comparably smaller increase in Nog expression may be insufficient to dampen the greater increase in BMP activity, resulting in accelerated maturation, hypertrophy and ECM mineralization.

Nog mRNA is expressed throughout the nascent humeral anlage (Brunet et al., 1998; Nifuji et al., 2001). We employed ISH to determine whether conditional Shox2 deletion results in changes to the spatiotemporal expression pattern of Nog, in addition to changes in Nog expression levels. However, we were unable to detect any deviation from the normal Nog expression pattern in E12.5 and 13.5 Prrx1-Cre and Col2a1-Cre Shox2 mutant limb buds (data not shown).

We have shown that Shox2 deletion in embryonic limb mesenchyme enhances early chondrogenesis and that deletion of Shox2 in chondrocytes triggers rapid maturation and hypertrophy. These findings prompted us to examine the effects of Shox2 knockdown on the chondrogenic differentiation of multipotent T1/2Cs and MSCs. When cultured in micromass in the presence of BMP2, T1/2Cs undergo early chondrogenesis, maturation and hypertrophic differentiation (Carlberg et al., 2001). We first analyzed endogenous expression levels of Shox2 mRNA transcripts during a 13-day BMP2-stimulated T1/2C chondrogenic time course. As expected, Shox2 expression was inversely correlated with chondrogenic gene expression and decreased to ∼60% of day 1 levels by day 7, the culture time point at which expression of Sox9 and Col10a1 peaked (Fig. 6A). We next stably transfected T1/2Cs with Shox2 shRNA expression plasmids. Surprisingly, during routine expansion, we found that our stable Shox2 knockdown T1/2C lines had spontaneously commenced early chondrogenesis in the absence of any external chondrostimulation. We detected no upregulation of chondrocyte maturation or hypertrophy genes, nor any increased expression of genes associated with adipogenesis or osteogenesis in this cell type, like Pparg, Runx2 and Spp1 (Fig. 6B). In the absence of chondrostimulatory growth factor, T1/2C micromass cultures normally undergo extremely low levels of chondrogenesis. However, when we cultured Shox2 knockdown T1/2Cs in micromass, these cells accumulated copious amounts of Alcian blue-, collagen II-positive ECM, even in the absence of BMP2 (Fig. 6C). Importantly, these Shox2 knockdown-induced increases in Alcian blue-, collagen II-positive ECM production were completely abrogated by 10 ng/ml Noggin treatment (data not shown). Like cultured Prrx1-Cre Shox2 mutant limb bud cells, Shox2 knockdown T1/2Cs failed to transition from early chondrogenesis to hypertrophy. Again, we tried to overcome this block to chondrocyte maturation and hypertrophy with BMP supplementation. Strikingly, after only 24 hours of BMP2 treatment, Shox2 knockdown T1/2C micromass cultures exhibited massive increases in the expression of genes regulating early chondrogenesis, chondrocyte maturation, and hypertrophic differentiation (Fig. 6D–F).

We wished to confirm the role of Shox2 as a repressor of multipotent mesenchymal cell chondrogenesis in primary cells, so we generated Shox2 knockdown lines of mouse bone marrow MSCs. Like micromasses of Shox2 knockdown T1/2Cs, high-density pellet cultures of Shox2-deficient MSCs underwent enhanced early chondrogenesis, but failed to upregulate expression of Col10a1 in the absence of chondrostimulatory growth factor. Interestingly, pellets of Shox2 knockdown MSCs cultured in the absence of BMP2 generated a collagen type II-rich ECM and did not upregulate expression of the Acan gene (Fig. 7A,B). Finally, BMP2 exposure significantly elevated expression of both early chondrogenic and hypertrophic genes in Shox2 knockdown MSC pellets relative to controls (Fig. 7C,D).

In humans, loss of SHOX2 gene function has yet to be associated with any known clinical syndrome. However, in mice, Shox2 deletion causes a wide range of developmental defects. Most obviously, mouse embryos nullizygous for Shox2 die between E11.5 and E17.5 due to aberrant formation of the sinoatrial node (Blaschke et al., 2007; Espinoza-Lewis et al., 2009; Yu et al., 2007). Shox2 mutants also exhibit a unique incomplete clefting phenotype of the future hard palate (Yu et al., 2005), as well as defects in temporomandibular joint formation (Gu et al., 2008) and in the differentiation of Ntrk2-positive dorsal root ganglion mechanosensory neurons (Scott et al., 2011). Finally, loss of Shox2 function in embryonic limb mesenchyme causes significant neural and muscular patterning defects in the forelimb stylopod (Vickerman et al., 2011), in addition to severe rhizomelia (Cobb et al., 2006; Yu et al., 2007). In this study, we show that Shox2 is an important repressor of chondrogenesis at two separate stages – the initial differentiation of chondroprogenitor cells into early chondrocytes and the maturation of early chondrocytes into hypertrophic chondrocytes (see Fig. 8).

We have shown that Col2a1-Cre-driven conditional Shox2 deletion causes significant shortening of the humerus and femur and that this rhizomelia is caused by precocious hypertrophic differentiation of stylopodial chondrocytes. Munns et al. (Munns et al., 2004) hypothesized that human SHOX might function as a repressor of chondrocyte maturation and that SHOX deficiency might result in precocious hypertrophic differentiation and accelerated growth plate fusion postnatally. Experiments in chicks, which like humans possess both Shox and Shox2 genes, provide evidentiary support for Shox as a repressor of chondrocyte maturation. For example, Shox misexpression throughout the developing chick limb bud results in an ossification delay in the coracoid of the scapula and the ischium of the hip. Furthermore, similar to the expression pattern of Shox2 in the developing mouse stylopod, expression of Shox is downregulated in chondrogenic condensations of the chick zeugopod (Tiecke et al., 2006). And finally, Shox overexpression in chick limb bud micromass cultures results in significantly depressed expression of the chondrocyte maturation genes Ihh (Tiecke et al., 2006) and Fgfr3 (Decker et al., 2011).

Human SHOX deficiency results in mesomelic shortening of the limb, whereas deletion of Shox2 causes rhizomelia in mice. Although human SHOX and mouse Shox2 specify different limb compartments in their respective species, these regional regulatory functions correspond with precise spatiotemporal expression patterns of SHOX and Shox2 in the developing human and mouse limb bud, respectively (Clement-Jones et al., 2000; Cobb et al., 2006). That said, there is conserved protein function between human SHOX and mouse Shox2 as human SHOX can rescue both aberrant sinoatrial node differentiation and forelimb stylopodial shortening in Shox2 knockout mice (Liu et al., 2011).

Interestingly, we detected Shox2 mRNA transcripts in the developing zeugopodial cartilage anlagen, in addition to the humerus and femur. This expression pattern is intriguing as neither Col2a1-Cre nor Prrx1-Cre conditional deletion of Shox2 manifests as zeugopodial shortening in newborn mice. However, it is not unexpected as strong expression of Shox2 is detectable from the flank to the handplate in the early limb bud (Cobb et al., 2006). Importantly, Hoxa11, c11 and d11 can entirely account for the formation and specification of the zeugopodial skeletal elements (Davis et al., 1995; Wellik and Capecchi, 2003), whereas limb Hox genes are not wholly responsible for forelimb stylopodial formation (Kmita et al., 2005). Thus, it will be important to delete Shox2 in a Hoxa11 or d11 null background in order to determine whether any functional redundancy exists between Shox2 and the limb Hox11 genes. It is possible that in these compound mutants, Shox2 deletion will result in zeugopodial shortening, a phenotype that is not observed in either Hoxa11 or d11 single mutants (Davis and Capecchi, 1994; Small and Potter, 1993).

We used the micromass culture model system to show that Noggin-sensitive members of the TGF-β superfamily are responsible for the precocious chondrocyte maturation and hypertrophy observed in Col2a1-Cre Shox2 mutants. We also found that Bmp2 and 4 seem to be involved in this premature hypertrophic differentiation both in vitro and in vivo. Bmp4 is a particularly strong candidate for the molecular link between Col2a1-Cre-driven conditional Shox2 deletion and precocious stylopodial hypertrophy, as increased Bmp4 levels have been detected in Shox2 knockout limb buds (Yu et al., 2007). Additionally, Puskaric et al. (Puskaric et al., 2010) have shown that SHOX2 can directly bind the Bmp4 promoter and activate its transcription in Cos-7 and HEK-293 cells. Importantly, SHOX2 is known to possess both transcriptional activation and repression capabilities, depending on the cell type (Yu et al., 2007). If Bmp4 is the in vivo effector of precocious chondrocyte hypertrophy driving stylopodial shortening, this mechanism would not be unprecedented. Tsumaki et al. (Tsumaki et al., 2002) generated transgenic mice with the Bmp4 gene under the control of Col11a2 promoter and enhancer sequences. This chondrocyte-specific Bmp4 overexpression resulted in considerable expansion of growth plate hypertrophic zones and significant shortening of the stylopodial skeletal elements. Although the expression pattern of Bmp2 in Col2a1-Cre Shox2 mutant humeri does not support a role for this Bmp as the principal effector of precocious stylopodial hypertrophy in mutants, Bmp2 and 4 are known to regulate each other’s expression and Bmp2 is a strong stimulator of hypertrophy as tamoxifen-induced conditional deletion of Bmp2 in chondrocytes results in significantly delayed chondrocyte maturation and severe chondrodysplasia (Shu et al., 2011).

Col2a1-Cre Shox2 mutants exhibit shortened stylopodial skeletal elements characterized by precocious chondrocyte hypertrophy, whereas chondrocytes in the shortened stylopodial skeleton of Prrx1-Cre mutants display delayed hypertrophy. When Shox2 is deleted from developing chondrocytes (i.e. Col2a1-Cre mutants), the result is a considerable increase in Bmp expression levels, driving chondrocyte maturation and hypertrophy (see Fig. 8A). However, when Shox2 is deleted from undifferentiated mesenchymal cells (i.e. Prrx1-Cre mutants), a more modest increase in Bmp expression occurs, enough to trigger an enhancement of early chondrogenesis, but not enough to surpass the critical threshold of BMP activity required to drive chondrocyte maturation and hypertrophy. These Shox2-deficient chondrocytes stuck in early differentiation are, however, extremely sensitive to BMPs and upon exposure, undergo both rapid and elevated levels of hypertrophy (see Fig. 8B). It is striking that this progression through in vitro chondrogenesis regulated by stepwise increases in BMP activity parallels the stepwise decreases in Shox2 expression that occur during endochondral ossification in vivo – first, during initial differentiation of the stylopodial cartilage anlagen and second, during chondrocyte maturation and hypertrophy (see Fig. 8C). It follows that the delayed hypertrophy in Prrx1-Cre Shox2 mutant stylopods might be due to the inability of these chondrocytes, which no longer express Shox2, to implement the second reduction in Shox2 levels that is normally associated with chondrocyte maturation. Currently, it is not clear why a stepwise reduction in Shox2 expression would be required to elevate BMP levels to an extent adequate for promoting chondrocyte maturation. In this regard, our model does not exclude the possibility that Shox2-dependent signals from mesenchyme surrounding the chondrogenic condensations might also be required for progression through hypertrophy and that these signals are anomalous in Prrx1-Cre Shox2 mutants. Oppositely, these unidentified chondrocyte maturation-promoting signals would be unaffected in Col2a1-Cre mutants as Shox2 is not deleted from non-chondrogenic limb mesenchyme in these animals.

Unlike persistent articular chondrocytes, which are phenotypically stable in non-disease states, early chondrogenesis is transient in MSCs. Yet, MSCs continue to represent a promising candidate cell source for generating repair and replacement tissue that replicates the molecular composition and organization, as well as the biomechanical properties, of articular cartilage (reviewed in Bobick et al., 2009). So, a significant obstacle impeding the engineering of biomimetic articular cartilage from MSCs is the continued maturation and hypertrophy of MSC-derived chondrocytes (reviewed in Steinert et al., 2007). Here, we have shown that Shox2 knockdown induces MSC chondrogenesis without hypertrophy, as long as Shox2-deficient MSC pellets are kept BMP-free. Another important difference between native articular cartilage and cartilage engineered from MSCs is the collagen-deficiency of MSC-generated ECM. Importantly, engineered cartilage with lower than native levels of collagen is characterized by inferior tensile properties (Huang et al., 2008; Lima et al., 2007). We have shown that Shox2 knockdown induces MSCs to generate a collagen type II-rich ECM, again as long as Shox2-deficient MSC pellets remain BMP-free. That said, it will be imperative to analyze the chondrogenic effects of Shox2 knockdown in MSCs from species that possess both Shox and Shox2, like humans, in case functional redundancy exists between these genes in the regulation of multipotent mesenchymal cell chondrogenesis.

To generate conditional Shox2 mutant mice, males harboring a Col2a1-Cre or Prrx1-Cre transgene, as well as one wild-type and one deleted Shox2 allele, were crossed to females carrying two floxed Shox2 alleles. Mutants were positive for Cre and carried one deleted and one floxed Shox2 allele. Mutants were compared to littermate controls that carried either one or two functional copies of Shox2, as abnormalities have not been observed in heterozygotes. Noon on the day of plugging was designated E0.5. Mice were maintained on a mixed C57BL/6-129/Sv background. All animal experiments were approved by the Life and Environmental Sciences Animal Care Committee at the University of Calgary. The deleted and floxed Shox2 alleles, as well as the Prrx1-Cre transgene, were described previously (Cobb et al., 2006; Logan et al., 2002). Mice harboring Col2a1-Cre were from the Jackson Laboratory (Bar Harbor, ME). Alcian blue and Alizarin red (Sigma-Aldrich, Oakville, ON) skeletal staining were performed according to standard protocols.

Micromass cultures of whole E11.5 and E12.5 limb buds, as well as the distal subridge region of E11.5 limbs, were prepared as previously described (Hoffman et al., 2006; Woods et al., 2005), with minor modifications. Limb tissue was dissociated into a single cell suspension at ∼1.5×10⁷ cells/ml in medium consisting of 3∶2 F12 Nutrient Mixture/high-glucose DMEM containing GlutaMAX and sodium pyruvate (Life Technologies, Burlington, ON), supplemented with 10% Qualified FBS (Life Technologies) and 100 µg/ml normocin (InvivoGen, San Diego, CA). 10-µl drops of cell suspension were spotted onto Nunc tissue culture plastic (Thermo Fisher Scientific, Rockford, IL). Micromasses were flooded with medium supplemented with 50 µg/ml ascorbate (Sigma-Aldrich) 60 minutes later. Medium was replenished at culture day 4. Recombinant human BMP2 and 4, as well as recombinant mouse Noggin, were purchased from R&D Systems (Minneapolis, MN). All-trans-retinoic acid was from Sigma-Aldrich.

Transient transfection of limb bud cells with a modified Effectene (Qiagen, Toronto, ON) protocol was performed as mentioned in Karamboulas et al. (Karamboulas et al., 2010). Briefly, 1 µg of plasmid DNA was combined with 15 µl of EC buffer, supplemented with 0.4 M trehalose (Sigma-Aldrich), and 1 µl of Enhancer. Following a 10-minute incubation, 5 µl of Effectene was added. 7.5 µl of Effectene/DNA was triturated with 40 µl of limb bud cell suspension and 10-µl micromass drops were spotted. Meis1a-MSCV was plasmid 21018 from Addgene (Cambridge, MA).

T1/2Cs (American Type Culture Collection, Manassas, VA) were expanded in monolayer and induced to undergo chondrogenesis in micromass according to standard protocols (Haas and Tuan, 2000), with minor modifications. T1/2Cs were expanded in medium consisting of high-glucose DMEM containing GlutaMAX and sodium pyruvate, supplemented with 10% Qualified FBS and 100 µg/ml normocin. T1/2C micromasses were established by spotting ∼1.0×10⁵ cells in a 10-µl drop on Nunc tissue culture plastic. Micromasses were maintained in medium supplemented with 50 µg/ml ascorbate and 100 ng/ml BMP2 for up to 13 days. Medium was replenished every four days.

Primary bone marrow MSCs from three adult mouse strains, C57BL/6J, C57BL/6-TgN(ACTbEGFP)1Osb, and C57BL/6-Tg(UBC-GFP)30Scha/J, were acquired from the Texas A&M Health Science Center Institute for Regenerative Medicine (Temple, TX). MSC expansion was carried out in medium consisting of high-glucose IMDM (Life Technologies) containing GlutaMAX and sodium pyruvate, supplemented with 10% Qualified FBS, 10% horse serum (Life Technologies) and 100 µg/ml normocin. Chondrogenic pellets were prepared as previously detailed (Peister et al., 2004), with minor modifications. Pellets containing ∼2.0×10⁵ MSCs were maintained in 15-ml polypropylene tubes (BD Biosciences, Mississauga, ON) for 15 days in culture medium consisting of high-glucose DMEM containing GlutaMAX and sodium pyruvate, supplemented with 10% Qualified FBS, 10% horse serum, 0.1 µM dexamethasone (Sigma-Aldrich), 50 µg/ml ascorbate, 40 µg/ml L-proline (Sigma-Aldrich), 1∶100 (v/v) ITS+ Universal Culture Supplement Premix (BD Biosciences), 100 µg/ml normocin and 100 ng/ml BMP2. Medium was refreshed every five days.

HuSH Shox2 shRNA and scramble sequence control plasmids were purchased from OriGene Technologies (Rockville, MD). T1/2Cs and MSCs were transfected with plasmid DNA at P10 and P6, respectively, using SuperFect Transfection Reagent (Qiagen), according to the manufacturer’s protocol. Selection of stably transfected T1/2Cs and MSCs was carried out in 1 µg/ml puromycin (Life Technologies) over two passages. T1/2C micromasses and MSC pellets were prepared from stably transfected polyclonal cells at P12 and P8, respectively.

Alcian blue staining and immunohistochemistry in limb bud and T1/2C micromasses were performed as previously detailed (Bobick and Kulyk, 2004). Alcian blue and H&E (Thermo Fisher Scientific) staining, as well as immunohistochemistry, of MSC pellet 7-µm paraffin sections were performed as described (Bobick et al., 2010). Primary antibodies against collagen types II and X, as well as IBSP, were from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA). Detection of primary antibodies was performed with the SuperPicTure Polymer Detection Kit (Life Technologies) employing an HRP-conjugated secondary antibody and DAB as a chromogen.

RNA isolation and quantitative real time PCR were performed as previously described (Bobick et al., 2010), with minor modifications. Total RNA was isolated from intact limb buds and cells in culture using E.Z.N.A. Total RNA Kit I (Omega Bio-Tek, Norcross, GA). Following first strand cDNA synthesis using qScript cDNA SuperMix (Quanta BioSciences, Gaithersburg, MD), levels of specific transcripts were determined by real time PCR using PerfeCTa SYBR Green FastMix for iQ (Quanta Biosciences) and an iCycler iQ Real Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA). Relative gene transcript levels were calculated using individual standard curves and normalized to relative mRNA levels of the constitutively expressed Actb Tbp and B2m genes. All forward and reverse primer sets have been previously described and are listed in supplementary material Table S1.

Western blotting was performed as previously described (Bobick et al., 2010). Representative blots shown in figures are from experiments performed at least three times with limb bud micromass culture sets, each set prepared from a single embryo, or T1/2C micromass sets, each prepared from an independently generated and propagated T1/2C line. Primary antibodies specifically detected SHOX2 (Sigma-Aldrich), SOX9 and RUNX2 (Abcam, Cambridge, MA), and actin (Santa Cruz Biotechnology, Santa Cruz, CA). Primary antibodies were detected with appropriate HRP-conjugated secondary antibodies (Santa Cruz Biotechnology) and Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific).

Whole mount micromass (Cash et al., 1997) and 10-µm cryosection (Cobb et al., 2006) ISH were performed as previously detailed. The Shox2 (Cobb and Duboule, 2005), Ihh and Bmp4 (Bitgood and McMahon, 1995), and Col10a1 (Jacenko et al., 1993) riboprobes have been described elsewhere. PCR primers used for generating the Bmp2 and Nog probes were CGAAGAAAAGCAACAGAAGCCC/CCCCACATCACTGAAGTCCACATAC and TCGAACATCCAGACCCTATCTT/CAGACTTGGATGGCTTACACAC, respectively. DIG labeled probes were detected with an AP-conjugated anti-DIG antibody (Roche, Laval, QC) and NBT/BCIP (Roche) was the color development substrate. Representative serial cryosections are shown. ISH was carried out on at least three embryos of a particular genotype at each gestational time point.

Quantitative data regarding skeletal element length and mRNA transcript levels were analyzed by two-tailed unpaired t-test using InStat 3 for Macintosh (GraphPad Software, La Jolla, CA).

The authors thank Dr Frank Jirik (Calgary) for mice, Dr Dave Hansen (Calgary) for use of equipment, Mr Stanley Neufeld (Calgary) and Ms Kristen Hui (Calgary) for generating the Bmp2 and Nog ISH probes, respectively, Ms Jessica Rosin (Calgary) for cloning assistance, and all members of the Cobb laboratory for critical discussion. Fig. 8 was prepared with Servier Medical Art.