C-type natriuretic peptide regulates endochondral bone growth through p38 MAP kinase-dependent and – independent pathways

Background C-type natriuretic peptide (CNP) has recently been identified as an important anabolic regulator of endochondral bone growth, but the molecular mechanisms mediating its effects are not completely understood. Results We demonstrate in a tibia organ culture system that pharmacological inhibition of p38 blocks the anabolic effects of CNP. We further show that CNP stimulates endochondral bone growth largely through expansion of the hypertrophic zone of the growth plate, while delaying mineralization. Both effects are reversed by p38 inhibition. We also performed Affymetrix microarray analyses on micro-dissected tibiae to identify CNP target genes. These studies confirmed that hypertrophic chondrocytes are the main targets of CNP signaling in the growth plate, since many more genes were regulated by CNP in this zone than in the others. While CNP receptors are expressed at similar levels in all three zones, cGMP-dependent kinases I and II, important transducers of CNP signaling, are expressed at much higher levels in hypertrophic cells than in other areas of the tibia, providing a potential explanation for the spatial distribution of CNP effects. In addition, our data show that CNP induces the expression of NPR3, a decoy receptor for natriuretic peptides, suggesting the existence of a feedback loop to limit CNP signaling. Finally, detailed analyses of our microarray data showed that CNP regulates numerous genes involved in BMP signaling and cell adhesion. Conclusion Our data identify novel target genes of CNP and demonstrate that the p38 pathway is a novel, essential mediator of CNP effects on endochondral bone growth, with potential implications for understanding and treatment of numerous skeletal diseases.


Background
Bone formation occurs through the related, but distinct processes of intramembranous and endochondral ossification [1,2]. While the former is responsible for the for-mation of bones directly from precursor cells, such as the majority of the skull, the latter is responsible for the development of long bones, ribs, and vertebrae through a cartilage intermediate. In endochondral ossification, mesenchymal cells condense and begin to differentiate into chondrocytes, some of which later form the growth plate that controls longitudinal growth of endochondral bones [3]. The growth plate zones consists of resting, proliferating, and terminally differentiated hypertrophic chondrocytes, each of which are characterized by the expression of specific markers [4,5]. This organization of the growth plate and the coordinated proliferation and hypertrophy of chondrocytes are responsible for elongation of bones and eventually determine final bone length. Hypertrophic chondrocytes are thought to undergo apoptosis, and simultaneously their surrounding cartilaginous matrix is degraded and replaced by bony tissue, produced by cells entering through vascularization of hypertrophic cartilage. The intricate control mechanisms regulating the proliferation, differentiation and apoptosis of chondrocytes as well as the subsequent vascular invasion are not completely understood. However, disturbances of these processes can result in numerous diseases such as chondrodysplasias and other growth disorders, demonstrating the need for a better understanding of the pathways involved [4,[6][7][8][9].
C-type natriuretic peptide (CNP) has recently been shown to be an important regulator of endochondral ossification. The dominant phenotype of CNP-deficient mice is dwarfism, as demonstrated by shortened long bones primarily due to reduced heights of proliferating and hypertrophic zones of the growth plate [10,11]. CNP also increases growth in mouse bone organ cultures [12,13]. More recently, loss-of-function mutations in NPR2, the gene encoding the CNP receptor, have been identified as cause of acromesomelic dysplasia type Maroteaux, an autosomal recessive chondrodysplasia in humans [14]. This CNP receptor is also known as guanylyl cyclase-B (GC-B) or NPR-B. Binding of CNP to GC-B results in increased intracellular cGMP, which can further activate downstream factors, such as cGKI and cGKII (cGMPdependent protein kinase I and II) as well as phosphodiesterases (PDEs) that break down cGMP and camp and specific ion channels [15][16][17][18]. In addition, CNP can also bind to a different receptor, NPR-3 (natriuretic peptide receptor 3) that is thought to act as a decoy/clearance receptor serving to limit the effects of natriuretic peptides. Interestingly, mice deficient for cGKII show a similar, although not identical phenotype to CNP-deficient mice [19], and cGKII has been shown to be required for the effects of CNP overexpression in transgenic mice [20]. These data clearly identify cGKII as an essential mediator of CNP effects, but the signaling pathways downstream of cGKII, potential parallel pathways and the target genes conferring cartilage responses to CNP are not completely understood. However, recent studies showed that overexpression of CNP results in inhibition of the MEK1/2-ERK1/2 MAP kinase pathway and rescues the effects of an activating fibroblast growth factor receptor 3 mutation on endochondral bone growth [21][22][23].
MAP kinases are central signaling molecules in most eukaryotic cells that integrate extracellular signals leading to altered cell proliferation, differentiation, and transcription in many cell types, including chondrocytes [24,25]. For example, both the ERK and the p38 MAPK families have been shown to play important roles in controlling chondrocyte differentiation in vitro and in vivo [3,26,27]. In the current study we demonstrate, for the first time, an essential role for the p38 MAP kinase pathway in CNP signaling in cartilage and identify target genes of CNP in chondrocytes using genome-wide microarrays.

CNP signaling enhances endochondral bone growth
We used an organ culture system of embryonic day 15.5 (E15.5) mouse tibiae to examine the effects of CNP on endochondral bone growth. Tibiae were cultured for six days in the presence of BSA (control) or different concentrations of CNP. 10 -8 , 10 -7 and 10 -6 M concentrations of CNP caused a 31%, 40%, and 42% increase, respectively, in longitudinal growth of tibiae (Fig. 1A,B). Treatment with 1 μM CNP almost doubled tibia weight relative to controls (Fig. 1C). Incubation of tibiae with 10 -4 M (8-(4chlorophenylthio) cGMP stimulated tibia growth in a similar or stronger manner (55%) as CNP (Fig. 1D). A general inhibitor of phosphodiesterases (PDEs), 3-Isobutyl-1-methylxanthine (IBMX) at 10 -4 M, was used to study the role of PDEs in bone growth in the organ cultures. PDE inhibition stimulated longitudinal growth by 30% when compared to the control. In contrast, specific inhibition of PDE 1 by 8-methoxymethyl IBMX did not alter bone growth significantly, indicating that this enzyme is either not involved in regulating bone growth or can be functionally replaced by other proteins. These data demonstrate that CNP/cGMP signaling stimulates endochondral bone growth, while PDEs inhibit this process. Removal of the perichondrium by enzymatic digestion and/or manual dissection did not alter the response to CNP, demonstrating that the anabolic effects of CNP are independent of the perichondrium (Fig. 1E).
At the histological level, the most significant effect of CNP was a marked expansion of the hypertrophic zone ( Fig.  2A). This enlargement of the hypertrophic zone was accomplished by increases in both the number and maximal size of hypertrophic chondrocytes (Fig. 2B), in agreement with earlier studies [23].

CNP-induced endochondral bone growth requires p38
MAP kinase signaling MAP kinases play multiple roles in chondrocyte differentiation and cartilage development [3]. We therefore exam-CNP enhances endochondral bone growth Figure 1 CNP enhances endochondral bone growth. Mouse E15.5 tibiae were harvested and cultured for six days in the presence of vehicle, CNP at the indicated concentrations, membrane-permeable 8-(4-cpt) cGMP (0.1 mM), the non-specific PDE inhibitor IBMX (0.1 mM), or a selective inhibitor of PDE I, 8-methooxymethyl, IBMX (10 μM). After six days in culture, vehicle and CNP-treated (1 μM) bones were stained with Alcian Blue and Alizarin Red and representative images are shown, in comparison to a freshly isolated tibia (A). Growth of tibiae over the culture period at indicated concentrations of CNP and treatments was measured (B, D), and the weight of bones was determined (C). CNP, 8-(4-cpt) cGMP and IBMX stimulated tibia growth, when compared to control conditions. E15.5 tibiae were isolated under three different conditions: perichondrium was left intact with very loose dissection, perichondrium was removed with dispase, and perichondrium was removed mechanically (E). Bones were then incubated with or without CNP (1 μM) for six days and bone growth was determined as change in bone length relative to day 1. Removal of the perichondrium did not influence the stimulatory effect of CNP on bone growth. All data represent means ± SD of three or four independent trials (p < 0.05).   3A). While simultaneous addition of MEK inhibitors and CNP had maximal effects on bone growth, these effects were not statistically different from treatment with CNP alone. These data suggest that CNP and MEK1/2 act through a common pathway and are in agreement with recent studies demonstrating an inhibitory role of the MEK/ERK cascade in endochondral bone growth [26] and down-regulation of MEK/ERK activity by CNP [21][22][23].
We next examined whether p38 is involved in the effects of CNP on cartilage growth. Inhibition of p38 activity by two different compounds, PD169316 or SB202190 (10 μM each), did not affect basal endochondral bone growth, when compared to the inactive control compound SB202474 (10 μM) (Fig. 3B). In contrast, inhibition of p38 by SB202190 or PD169316 blocked CNP-induced growth (Fig. 3B). This effect was obvious by day 6 of culture and maintained by day 8 (Fig. 3C). Moreover, p38 inhibition completely reversed CNP effects on tibia weight (Fig. 3D), further demonstrating a requirement for p38 activity in CNP-induced endochondral bone growth. Next we examined whether CNP regulates the p38 pathway by investigating the phosphorylation of the kinases MKK3 (MAP kinase kinase3) and MKK6, direct and specific activators of p38. Western blotting with phosphospecific antibodies revealed that CNP and cGMP increase the phosphorylation of MKK3/6 in primary chondrocytes after 10 minutes of incubation (Fig. 3E), demonstrating that CNP signaling activates the p38 pathway in chondrocytes. Immunohistochemistry for phosphorylated (active) p38 showed little staining under control conditions, but demosntrated a strong increase in p38 phosphorylation in CNP-treated tibiae (Fig. 3F).

CNP delays tibia mineralization in a p38-dependent manner
To examine the effects of p38 inhibition on growth plate organization, we performed histological analyses of organ culture sections. As above, CNP stimulation caused an expansion of the hypertrophic zone of the growth plate, while SB202190 by itself did not have any marked effects (Fig. 4A). However, p38 inhibition suppressed the enlargement of the hypertrophic zone in response to CNP, providing further evidence for a requirement for p38 activity for the anabolic effects of CNP.
During dissections and analyses of organ cultures, we also noticed that CNP-treated bones were more fragile and appeared less mineralized. Alcian Blue/Alizarin Red staining of tibiae confirmed that the mineralized area was smaller in CNP-treated bones and displayed weaker Alizarin Red staining (Fig. 4B). We quantified the area of the mineralized (red) and cartilaginous (blue) regions of tibiae using digital image analyses. CNP treatment did increase the Alcian Blue-stained area considerably, without effects on the Alizarin Red-stained area (Fig. 4C). This resulted in a reduction of the mineralized area relative to the total area of the bone by about 30%. These data suggests that CNP-induced growth of cartilage is not matched by a corresponding expansion of the mineralized area and that CNP treatment delays the remodeling of hypertrophic cartilage. Inhibition of p38 activity by SB202190 resulted in a slight, but significant increase of the mineralized area (relative to total area) and reversed the effects of CNP on the Alcian Blue-stained area completely (Fig. 4C).

Microarray analyses identify hypertrophic chondrocytes as main targets of CNP signaling
We next performed microarray analyses to identify target genes of CNP in chondrocytes. Tibiae were cultured for six days in the absence or presence of CNP and then micro-CNP induces expansion of the hypertrophic zone  dissected into three distinct zones: the resting/proliferative (RP), the hypertrophic (H), and the mineralized (M) zones (Fig. 5A). RNA was isolated directly from tibiae from three independent trials for each zone and both treatments were analyzed using Affymetrix Mouse 2.0 arrays in the London Regional Genomics Center as described (London, Ontario, Canada) [28]. Real-time PCR analyses of collagen II (Col2a1) and collagen X (Col10a1), known markers of cartilage development, confirmed that micro-dissection resulted in efficient separation of the zones (Fig. 5B). Microarray profiles of selected genes involved in endochondral bone growth are shown to further illustrate correct separation of zones (Fig. 5C).
Bioinformatics analyses of microarray results (Fig. 6A) demonstrated that the hypertrophic zone was most responsive to CNP (Fig. 6B). Only 47 probe sets in the resting/proliferative zone (35 down, 12 up) and 58 probe sets in the mineralized zone (41 down, 17 up) responded with minimum two-fold responses to CNP (see Table 1 and 2 for lists of regulated genes). In contrast, 309 probe sets in the hypertrophic area showed a two-fold or higher change in expression in response to CNP. Of these probe sets, 157 probe sets were up-regulated by CNP in the hypertrophic zone, and 152 were down-regulated by CNP ( Table 3).
One of the genes showing a strong increase in the hypertrophic zone (>6-fold) was Ptgs2, encoding cyclooxygenase 2 (Cox2), a key enzyme in the synthesis of prostaglandins. Since Ptgs2 and its products (such as prostaglandin E2) are known to play important roles in chondrocyte differentiation and skeletal remodeling [29][30][31], we selected this gene for validation experiments. Induction of Cox2 mRNA expression in the hypertrophic zone by CNP was confirmed by real-time PCR, which showed a 10-fold increase in transcript levels ( Fig. 6C). p38 inhibition did reduce the basal levels of Cox2 mRNA, but surprisingly did not affect the induction by CNP. Among the genes down-regulated by CNP was the Tnfsf11 gene, encoding RANKL, a known activator of osteoclastic resorption of bone and cartilage [32,33]. Tnfsf11 displayed a 3.8-fold reduction in expression according to microarray analyses. Because down-regulation of Tnfsf11 could provide a molecular mechanism for the observed delay in mineralization and cartilage remodeling in response to CNP, we chose to validate its expression. Realtime PCR analysis confirmed down-regulation of Tnfsf11 mRNA levels in the hypertrophic zone by CNP (Fig 6D).
To answer the question why the hypertrophic zone is much more responsive to CNP treatment than other zones, we examined expression of key genes in the CNP signaling pathway. Analyses of our microarray data demonstrated that the genes encoding CNP (Nppc), its signal-ing receptor GC-B (Npr2) and the decoy receptor (Npr3) are expressed at similar levels in all three zones of microdissected tibiae under control conditions (Fig. 7A). However, Prkg1 (encoding cGMP-dependent kinase I) expression is 5.9 fold higher in hypertrophic chondrocytes than in the resting/proliferative cells, and seven-fold higher in the hypertrophic versus the mineralized zone (Fig. 7A).
Similarly, Prkg2 expression is 4.4-fold and 2.5-fold higher in the hypertrophic zone versus the resting/proliferative and mineralized zones, respectively (Fig. 7A). This expression pattern of key mediators of CNP signaling can explain the strong responsiveness of hypertrophic chondrocytes to CNP. In addition, our microarray data on expression of the decoy receptor Npr3 in the hypertrophic zone, while variable and thus not statistically significant, suggested that CNP strongly activates the expression of Npr3. We therefore decided to analyze its expression by real-time PCR which demonstrated a statistically significant 16-fold induction of Npr3 expression in the hypertrophic zone by CNP that was not altered by p38 inhibition (Fig. 7B). CNP did not affect Npr3 expression in the other growth plate zones.

Annotation of microarray data identifies CNP-regulated pathways
To gain insight into biological processes regulated by CNP, we employed KEGG annotation [34] on genes showing at least two-fold changes in response to CNP in the hypertrophic area (Fig. 8A). Numerous pathways were affected by CNP, most of them comprised approximately proportionally by up-and down-regulated genes. However, genes related to cell adhesion were strongly enriched in up-regulated genes, suggesting that CNP promotes cell adhesion. Most notably, CNP induced expression of genes involved in cell-cell interactions such as Icam2 (intercellular adhesion molecule 2), Cdh5 (Cadherin 5), and Esam1 (endothelial cell-specific adhesion molecule). In contrast, down-regulated genes included many genes encoding extracellular matrix molecules, for example Matn1 and Matn3 (Matrilin 1 and 3), Col9a2 (procollagen type IX, alpha 2) and Col14a1 (procollagen type XIV, alpha 1) ( Table 3). In addition, up-regulated genes included three members of the TGFβ superfamily, Gdf5, Inhbb and Inhba, as well as the BMP antagonist Grem1 (Fig. 8B). Besides the TGFβ family, members of the Wnt and hedgehog signaling pathways are important regulators of cartilage differentiation, and components of these pathways were regulated by CNP. Other categories in which up-regulated genes were over-represented included tight junctions and calcium signaling, whereas pantothenate and CoA biosynthesis was one example for a pathway dominated by down-regulated genes. Finally, more transcription factorencoding genes were up-regulated than down-regulated by CNP (Fig. 8B).
Inhibition of the MEK1/2-ERK1/2 pathway stimulates tibia growth, while p38 MAPK is required for CNP-induced bone growth Figure 3 Inhibition of the MEK1/2-ERK1/2 pathway stimulates tibia growth, while p38 MAPK is required for CNPinduced bone growth. Mouse E15.5 tibiae were harvested and cultured for six days in the presence of control or CNP (1 μM) and vehicle (DMSO) or MEK1/2-ERK1/2 pathway inhibitors PD98059 (10 μM) and U0126 (10 μM) (A). Though both PD98059 and U0126 stimulated basal bone growth, inhibition of the MEK1/2-ERK1/2 pathway did not further enhance CNPinduced bone growth (*: p < 0.05 when comparing control/inhibitors to control/vehicle; #: p < 0.05 when comparing CNP/ vehicle to control/vehicle; p > 0.05 when comparing CNP/vehicle to CNP/inhibitors). Tibiae were incubated with control or CNP and pharmacological inhibitors of the p38 MAPK pathway (SB202190 or PD169316, 10 μM each) or an inactive analog (SB202474, 10 μM) (B). p38 inhibition did not effect basal bone growth significantly, but did suppress CNP-induced bone growth (*: p < 0.05 when comparing CNP/inhibitors to CNP/SB202474; #: p < 0.05 when comparing CNP/SB202474 to control/SB202474). Bone growth was measured over an extended time course of eight days, showing that CNP continued to significantly influence growth on day 8, while SB202190 reversed these effects (C). Bones from each treatment were weighed under different conditions, and it was found that p38 inhibition reversed the effects of CNP on weight (D). Protein extracts from primary chondrocytes cultured with control, CNP (10 -6 M), or 8-(4-cpt) cGMP (0.1 mM) for 10 minutes were examined for phosphorylation of the p38 activators MKK3/6 by western blot analysis (E). Both treatments increased phosphorylation of MKK3/6, supporting the stimulation of p38 MAP kinase activity by CNP signaling. Immunohistochemistry with an antibody against phosphorylated p38 demonstrates markedly higher signal in CNP-treated tibiae when compared to control bones (F). p38 MAPK activity is required for CNP-induced hypertrophy Figure 4 p38 MAPK activity is required for CNP-induced hypertrophy. E15.5 tibiae were isolated and incubated with or without CNP (1 μM) and DMSO or SB202190 (10 μM). Hematoxylin and Eosin staining of tibia sections after six days of culture show that p38 inhibition reversed CNP-induced expansion of the hypertrophic zone (A). Tibiae were stained with Alizarin Red and Alcian Blue, and representative images demonstrate increased bone growth by CNP and the reversal of these effects upon p38 inhibition (B). The area of the mineralized zone (red) was measured as absolute area (C, bottom) and as a percentage of total area (C, top), demonstrating that CNP-treated bones displayed significantly smaller mineralized area in relation to the whole bone area. This was reversed upon p38 inhibition. Representative images are shown, while all data represent means ± SD of four independent trials, each with six bones (p < 0.05). Microarray analyses identify the hypertrophic area as the main target of CNP treatment. E15.5 tibiae were isolated, incubated with or without CNP (1 μM) and DMSO or SB202190 (10 μM) and micro-dissected into the resting/proliferating, hypertrophic, and mineralized regions prior to RNA extraction and microarray analyses. Analyses of microarray results from three independent trials using Genespring 7.2 (A) illustrated that the hypertrophic zone was most significantly responsive to CNP treatment, when compared to control conditions (B). Six times as many probe sets showed at least 2-fold expression changes in the hypertrophic zone when compared to either resting/proliferating or mineralized regions. Real-time PCR analyses on micro-dissected tibiae were used to validate selected microarray patterns. CNP induction of Ptgs2, the gene encoding cyclooxygenase-2, was confirmed (C). SB202190 treatment did reduce basal Cox2 mRNA levels, but did not interfere with CNP induction of Cox2. Tnfsf11, the gene encoding RANKL, was confirmed to be down-regulated in response to CNP treatment. Data represent means ± SD of three independent trials (p < 0.05).

*
We also performed KEGG analyses of CNP-regulated genes in the other zones. In the proliferative zone, genes involved in cytokine receptor interactions were most prominent with four genes, three of each were upregulated by CNP ( Table 1). The categories of cell adhesion and focal adhesion molecules were represented by three genes each (data not shown). In contrast, no category was represented by more than two genes in the mineralized zone (data not shown).

Discussion
Gene disruption and other studies have identified the CNP pathway as one of the most important anabolic regulators of endochondral bone growth. However, the molecular and cellular mechanisms involved are not completely understood. Here we provide multiple novel insights into these mechanisms. Most importantly, we show that the p38 MAP kinase pathway is an essential mediator of CNP effects on endochondral bone growth. Second, we identify the hypertrophic zone of the growth plate as the main target of CNP signaling, likely because of the high levels of cGMP-dependent kinase I and II expression in this zone. Third, we used genome-wide microarray analyses to identify multiple target genes potentially involved in CNP effects in cartilage.
Earlier studies have demonstrated that CNP stimulates bone growth through enhanced proliferation, mineraliza-       data show that expression of RANKL, a potent activator of osteoclastic bone resorption, in the hypertrophic zone is down-regulated by CNP. RANKL is expressed in hypertrophic cartilage [35][36][37], where it likely stimulates the removal of hypertrophic cartilage by osteoclasts and facilitates vascular invasion and ossification. Repression of RANKL expression by CNP could thus delay these remodeling events. Experiments are under way in our laboratory to examine whether osteoclast activity is indeed reduced in CNP-treated organ cultures.
We and others have shown important roles of p38 in hypertrophic chondrocyte differentiation in vitro and in vivo [3,27,[38][39][40][41]. Thus, it is not surprising that p38 inhibition reverses CNP effects on longitudinal growth and the expansion of the hypertrophic zone. Moreover, our data show that p38 activity is required for the repression of mineralization by CNP. These data are in agreement with a recent study showing delayed primary and secondary ossification in transgenic mice overexpressing an activated form of MKK6, an upstream activator of p38, in cartilage [27]. It should be noted, however, that other phenotypes of these mice (such as reduced proliferation and delayed hypertrophy) are not recapitulated in our studies, potentially due to altered patterns and/or levels of p38 activation in the two studies (e.g. transgenic expression of activated MKK6 under the collagen II promoter versus activation of p38 through the endogenous NPR2/cGMP signaling cascade) or because CNP acts through additional pathways besides p38. Independent of these complications, our studies provide strong evidence for a novel function of p38 signaling in maintaining hypertrophic cartilage and delaying the replacement of cartilage by bone.
However, our data also show that p38 signaling is not required for all effects of CNP on hypertrophic cartilage. While p38 inhibition results in lower basal levels of Cox2 mRNA in chondrocytes, in agreement with observations by other studies [42][43][44], CNP still causes a strong increase in Cox2 expression in the presence of SB202190. Similarly, Npr3 induction by CNP is independent of p38 activity. Therefore, it appears likely that p38 signaling is required to achieve and/or maintain the expanded hypertrophic zone in CNP-treated bones, but not for induction of some target genes. Studies to identify additional signaling pathways connecting CNP to Cox2 gene expression are underway in our laboratory.
Our studies also demonstrate antagonistic roles of p38 and another MAP kinase pathway, the MEK1/2-ERK1/2 pathway. Inhibition of MEK1/2 activity results in enhanced growth of endochondral bones, with no additive or synergistic effect with CNP. While our studies were in progress, other groups showed that the MEK1/2-ERK1/ 2 indeed reduces endochondral bone growth in vivo [26] and that CNP inhibits ERK1/2 activity [21], in agreement with our studies. Additional studies confirmed the close and reciprocal interactions between CNP-cGMP and FGF-MEK1/2-ERK signaling [22,23]. For example, CNP was shown to repress FGF-induced growth arrest and extracellular matrix degradation by counteracting MEK1/2 activation, while FGFs 2 and 18 suppress CNP-stimulated cGMP production [22,23]. However, none of these studies evaluated a potential role of p38 signaling in this context. Since p38 has also been implicated in FGF signal transduction in chondrocytes [45,46], it will be interesting to investigate whether this MAP kinase is involved in the antagonistic effects of CNP and FGF in endochondral Expression patterns from microarray analyses demonstrate up-regulation of cGMP-dependent kinase genes in the hypertrophic zone Figure 7 Expression patterns from microarray analyses demonstrate up-regulation of cGMP-dependent kinase genes in the hypertrophic zone. Microarray analyses of the principal players in the CNP pathway in micro-dissected tibiae cultured with and without CNP (1 μM) are shown (A). Prkg1 and Prgk2, encoding cGMP-dependent kinases I and II, were strongly upregulated in the hypertrophic zone, irrespectively of exogenous CNP. In addition, CNP strongly stimulated expression of Npr3, the natriuretic peptide clearance receptor, in the hypertrophic zone. Real-time analysis confirmed induction of Npr3 by CNP, which primarily occurs through a p38-independent manner. Data represent means ± SD of three independent trials (p < 0.05).   bone growth. However, both CNP and FGFs activate p38 in chondrocytes, but they have opposing effects on the growth of endochondral bones. The role of p38 in FGF effects on chondrocytes has, to our knowledge, only been studied in cell culture, not in a three-dimensional model that allows direct assessment of bone growth. Based on our data, we don't expect that p38 activation contributes to the growth-repressing activities of FGF, but this prediction needs to be experimentally verified. Nevertheless, the fact that both FGFs and CNP activate p38 despite their opposing effects on bone growth makes it unlikely that p38 contributes to crosstalk between the two signaling systems, in contrast to ERK1/2.
Similarly, it will be important to examine whether regulation of the different MAP kinases by CNP occurs through independent, parallel pathways, or whether they regulate each other. In addition, the pathways connecting CNP to the MAP kinase modules have not been completely resolved. For example, while it has been shown that repression of ERK activity by CNP occurs at the level of the upstream kinase Raf1 and requires cGMP-dependent kinase activity [22], the exact molecular mechanism involved has not been described.
This is, to our knowledge, one of the first studies to use micro-dissection of mammalian endochondral bones for genome-wide expression analyses by microarrays. We chose to perform these studies after 6 days of CNP stimulation, as opposed to a short term treatment. While this approach does not allow us to distinguish direct and indirect target genes of CNP, it mimics the in vivo situation where cells are exposed to auto-/paracrine CNP signaling for extended periods. Our study should therefore identify genes that are regulated by long term exposure to CNP and are thus likely to be involved in the physiological activities of CNP in the growing skeleton. Our microarray data were confirmed by real-time PCR analyses for selected genes; these data and our earlier studies [28,47,48] strongly suggest that the vast majority of the expression profiles detected by our microarrays correspond to the authentic gene expression patterns. Analyses of array data as well as confirmatory real-time PCRs (e.g. for type X collagen) also demonstrated that our micro-dissection protocol results in efficient separation of different zones of the cartilage and can be used for identification of novel hypertrophy-specific genes. Moreover, these data clearly show that the hypertrophic zone of the growth plate is by far the most responsive to CNP. This responsiveness does not correlate to altered levels of mRNAs for CNP itself, its signaling receptor, or the decoy receptor in control conditions. Instead, our data suggest that the expression of cGMP-dependent kinases I and II (cGKI, II) is much higher in this zone than in the other ones, providing further evidence for a crucial role of these enzymes in CNP signal transduction. Interestingly, the expression of these two genes, as well as the Nppc and Npr2 genes, in the hypertrophic zone is not altered in response to CNP. In contrast, Npr3 expression is strongly induced by CNP in the hypertrophic zone. While this induction was not identified as significant in the microarray analyses, subsequent real-time PCR confirmed the existence of this previously unknown feedback loop that likely limits CNP effects in growing cartilage.
Our expression data suggest that both cGKI and II are involved in mediating CNP effects on cartilage development. Studies with genetically altered mice and naturally occuring rat mutants demonstrate that cGKII is the dominating protein in chondrocytes [19,49]; however, the cartilage phenotypes of cGKII-and CNP-deficient mice are not identical, suggesting the possibility of an additional role of cGKI in CNP signaling. Double knockout mice for both cGKI and II will be required to resolve this issue.
Thus, our data in conjunction with published studies support a model where basal CNP signaling promotes proliferation and extracellular matrix synthesis in growth plate chondrocytes. Once cells start to differentiate, they increase their expression of cGMP-dependent kinases and their responsiveness to CNP. This results in an extension of hypertrophic chondrocyte life and a delay in osteoclast and potentially vascular invasion, thus promoting maximal growth of hypertrophic chondrocytes and endochondral bone growth. At the same time, high levels of CNP signaling induce expression of Npr3 that ultimately limits CNP effects, allowing for expression of RANKL and for remodeling of the metaphysis. Experiments are under way to examine whether this model accurately describes cellular mechanisms of CNP signaling in endochondral ossification and to identify the molecular mechanisms involved.
Detailed analyses of our microarray data provided novel insights into biological processes regulated by CNP. CNP treatment induced the expression of several genes for cellcell interactions in the hypertrophic area (as well as the resting/proliferating zone), while at the same time repressing genes for ECM proteins. Another process regulated by CNP is signaling by TGFβ family members. Most interestingly, CNP induces expression of Gdf5 and Grem1, both of which have been implicated in skeletal development. Loss-of-function mutations of Gdf5 have been identified as cause of reduced skeletal growth in human chondrodysplasias and brachypodism mice [50]. Interestingly, GDF5 has been shown to stimulate cell adhesion in chondrocytes [51], in agreement with our data showing increased expression of both Gdf5 and cell adhesion molecules in response to CNP. Therefore, GDF5 is an excellent candidate for mediating the anabolic effects of CNP. In contrast, Grem1 encodes a BMP antagonist that is required for limb development and controls chondrogenesis [52][53][54][55][56]. Moreover, Grem1 expression is induced by BMP/GDF signaling [57][58][59], suggesting that its stimulation could be secondary to increased expression of Gdf5 and/or related factors (e.g. Inhbb and Inhba) in response to CNP.
In summary, our results identify several novel components and characteristics of CNP signaling during endochondral bone growth. Collectively, these studies lead to the novel concept that CNP acts, at least in part, by delaying the terminal steps of endochondral ossification, i.e. the replacement of hypertrophic cartilage by bone. Further tests of this model in vivo and elucidation of the mechanisms involved will not only result in improved understanding of endochondral bone development, but will also be crucial for the development of potential therapeutic applications.

Materials
Timed-pregnant CD1 mice were purchased from Charles River Canada. CNP, 8-(4-cpt) cGMP, and pharmacological inhibitors were obtained from Sigma and Calbiochem. Cell culture reagents were from Invitrogen and general chemicals from VWR. All real-time PCR probes and reagents were purchased from Applied Biosystems. The phospho-MKK3/6 (Cat. number 9231) and phospho-p38 (9216) antibodies were from Cell Signaling Technologies, and the β-actin antibody was from Sigma.

Organ Culture
Tibiae were isolated from embryonic day 15.5 (E15.5) embryos from CD1 timed-pregnant mice (Charles River Canada) using the Stemi DV4 Stereomicroscope (Zeiss). Dissection day was considered to be day 0 and tibiae were allowed to recover from dissection overnight in serumfree α-MEM media containing 0.2% Bovine Serum Albumin (BSA), 0.5 mM L-glutamine, 40 units penicillin/mL and 40 μg streptomycin/mL as described [60]. The following morning, bones in 24-well Falcon plates were measured using an eyepiece in the Stemi DV4 Stereomicroscope and treated with CNP (0.01 to 1 μM) or BSA/HCl (1 mM) vehicle, and DMSO or U0126, PD98059, PD169316, SB202190 or SB202474 (10 μM each). Media was changed every 48 hrs beginning on day 1, and bones measured on days 1, 3, 6, and 8. Results are expressed as change in length relative to day 1. Experiments were repeated at least three times, with 4-6 bones per treatment for each trial.
For weight determination and Alizarin Red/Alcian Blue staining, 6 bones per treatment were weighed at day 6 of culture and then placed in 4% Paraformaldehyde (PFA) in DEPC-treated PBS for overnight fixation. Subsequently, tibiae were placed in staining solution for 45-60 minutes (0.05% Alizarin Red, 0.015% Alcian Blue, 5% acetic acid in 70% ethanol). Images of stained bones were taken using a Nikon SMZ1500 dissecting microscope with Photometric CoolSNAP colour digital camera (Nikon Canada) and PTI Image Master 5 program. Stained areas in images were measured using Openlab 4.0.4 software program.
For experiments requiring perichondrium removal, tibiae were isolated from embryonic 15.5 day embryos under three different modes: very loose dissection ensuring that perichondrium was intact, very careful dissection in which perichondrium was removed mechanically, and treatment of tibiae with dispase (3 mg/mL in PBS) for 3-5 minutes with concurrent mechanical removal of perichondrium [61,62]. Media was changed every two days beginning on day 1 and bone lengths measured on days 1 and 6, with change in length expressed relative to day 1.

Histology and Immunohistochemistry
After experiment completion, tibiae were rinsed with PBS and fixed in 4% PFA overnight. Bones were then stained with mercurochrome for visualization, placed in 10% formalin solution, and sent for embedding and sectioning in the Pathology lab at University of Western Ontario Hospital or the Molecular Pathology Core Facility at the Robarts Research Institute (London, Ontario, Canada). Following sectioning, bones were stained with hematoxylin and eosin using standard protocols. For immunohistochemistry, sections were incubated with primary anti-phospho-p38 antibody (1:50 dilution) over night at 4°C. Bound antibody was visualized using the UltraVision LPValue detection system (Lab Vision) with AEC chromogen substrate (Lab Vision).

RNA isolation from organ cultures and microarray analyses
For experiments requiring RNA isolation from organ cultures, E15.5 tibiae were harvested and treated as described above with or without CNP and SB202190. On day 6 of treatment, tibiae were separated under a dissecting microscope into the resting/proliferative, hypertrophic, and mineralized areas. Same areas from approximately 24 bones were pooled per trial, in each of three independent trials. RNA was isolated following the RNeasy ® Lipid Tissue Extraction protocol from Qiagen (Mississauga) and RNA integrity verified using the Agilent 2100 Bioanalyzer. Microarray analyses from three trials were performed at the London Regional Genomics Centre (London, Ontario, Canada) using MOE430 2.0 Affymetrix arrays consisting of 45,000 probe sets (covering the entire mouse genome). Results were analyzed using GeneSpring 7.2 software as described [28]. Microarray data were independently filtered using GeneSpring Bioscripts quality fil-ters for noise and one-way ANOVA testing, to eliminate genes that were not expressed or showed great variability between replicates. The remaining 5199 probes sets common to both filtering methods were used for all subsequent analyses. Lists of genes undergoing at least two-fold changes were analyzed using the Babelomics suite [63] and in particular the KEGG pathways module in the Fatig-oPlus tool.

Real-Time PCR
Real-Time PCR analysis was performed as described using Applied Biosystems 7900 HT Real-Time PCR System and TaqMan ® Gene Expression Assays [28,40,64]. All probes (Npr3, Col2a1, Col10a1, Ptgs2, Tnfsf11 and Gapdh)were purchased from Applied Biosystems. Gene expression levels were determined using the Standard Curve quantitative method with Gapdh levels as the basis of comparison.

Statistical Analyses
All experiments were performed in at least three independent trials. Two-Way ANOVA (parametric) test with Bonferroni post-test were performed using the Graph Pad/ Prism software. One-way ANOVA with Bonferroni posttest and paired t-tests were used when appropriate.