Microarray experiments are considered as discovery tools that open up new avenues for research by identifying new gene targets30. Although global overview of gene expressions in the studied tissue specimens is made possible, microarrays are basically screening tools that can formulate more targeted research questions and generate well-supported hypotheses rather than proven conclusions31. In the present study, LCM, was used to selectively obtain cells from the MCC zones individually. LCM along with MAA were employed to provide new insights into characterizing the four MCC zones. The results support formulating the hypothesis that MCC cells have significantly different patterns of gene expression from those of articular chondrocytes, and more importantly, several genes were found to be expressed variably upon the transition from one zone to another whitin MCC. By demonstrating this spatial (zonal) changes in the gene expression levels, our findings support the hypothesis that cells within the MCC have different phenotypic characteristics. This hypothesis is concordant with a previous finding reported by Hinton et al. who concluded that the prechondroblastic cells (cells of both the fibrous and proliferative zones) have different gene expression profiles from the underlying chondrocytes (cells of both mature and hypertrophic zones) of the MCC27.
The difference in gene expression ratios was most obvious between the articular chondrocytes (FZ) and other MCC zones, nevertheless, strong differences were also identified within the other MCC comparisons. In a related study by Fukui et al., zonal differences between the genes of superficial fibroblastic and deep hypertrophic regions of human femoral cartilage were also found to be very pronounced12. When comparing adjacent MCC zones such as FZ with PZ, PZ with MZ, and MZ with HZ at ≥ 20 FC level, only three genes were modulated out of the 30 identified genes (Table 3). Indeed, given the strong overlap in the cellular and extracellular composition between the mature and hypertrophic chondrocytes, it is not surprising that significant expression differences are limited to relatively very few genes. On the contrary, all of the 30 differentially regulated genes at ≥ 20 FC level were identified when we compared non-adjacent zones (Table 3). We could neither support or oppose these findings by the literature as there are no similar previous zone-specific studies on the MCC, however, by the analogy with Wang et al. study, where only proliferative and hypertrophic growth plate zones were isolated and then compared, the presence of significant differential gene expression between the two studied zones could be considered supportive to our findings32. Likewise, Zhou et al. study identified 804 differentially expressed genes when the articular zone of MCC was compared with the mature zone33.
The strongest upregulated relative expression ratio was observed for the Crabp1 gene in FZ as compared to HZ and to the control (7.45- and 7.21-folds respectively) (Tables 2 and 3). Retinoic acids, the active ingredient of vitamin A, play a role in different activities including cellular growth, differentiation and development by binding to specific nuclear receptors, and then regulate gene expression34,35. Both vitamin A deficiency and excess lead to skeletal defects; large doses result in growth retardation and premature closure of the growth plate, whereas administration of retinoid antagonists prevents further differentiation of prehypertrophic chondrocytes, indicating the importance of endogenous retinoids for chondrocyte maturation36. CRABPs are carrier proteins crucially important for the transport and metabolism of retinoic acid34. The amount of the latter substance reaching the nucleus is modulated two cytoplasmic binding proteins CRABP I and II37. Overexpression of CRABP I is probably preventing retinoic acid from entering the nucleus by keeping it in the cytoplasm, and by facilitating the acid degradation38. On the contrary to our identified bizonal increase of Crabp1in the superficial zones of MCC, a study on rabbit growth plates reported much higher level of Crabp1 transcript in the maturing and hypertrophic chondrocytes than in resting and proliferating chondrocytes39. This disagreement could be attributed to the variant cell phenotypes, especially the dividing cell population (PZ) in the MCC as compared with the growth plate.
On the other hand, the most pronounced downregulated gene was Clec3a (− 8.92-fold) in FZ in relation to the femoral cartilage (Table 2). Clec3a gene is a cartilage-derived member of the C-type lectin superfamily. It requires calcium for binding, hence designated as C-type. The protein it encodes is apparently restricted to cartilage and involved in many biologic functions as it promotes cell adhesion to laminin-332 and fibronectin. While this protein has been found in nucleus pulposus, nasal cartilage and in articular cartilage, the distribution of mRNA of Clec3a in the developing rib was related to the upper hypertrophic and proliferating chondrocyte zones, suggesting a role in organizing the ECM and probably in regulating the epiphysis remodeling40. According to our MAA data, Clec3a was the most downregulated gene at FZ; a similar expression pattern was demonstrated by Grogan et al., who studied the zonal expression patterns of genes in the FCC of human and bovine. They found a significant downregulation of Clec3a gene in the superficial zone compared to the middle zone41. It is worth mentioning that some of the differentially expressed genes identified in the present study were not reported previously. This is in concordance with Hinton et al. study where novel unsuspected genes were differentially expressed in the perichondrium of the MCC27,42. Furthermore, the identification of relatively large number of unknown genes and expressed sequence tags may indicate that novel molecular pathways are not yet identified43. Intriguingly, 25.3% of the differentially expressed genes in at least one of the ten pairwise comparisons conducted at ≥ 20-fold cut-off value were unknown genes (Supplementary Table 2); thus the current study implies that several yet-to-be identified pathways may play a significant role in MCC.
In osteoarthritis, proteolytic enzymes such as matrix metalloprotease and aggrecanases degrade cartilage extracellular matrix components. This is accompanied with the expression of hypertrophic chondrocytes markers e.g. type 10 collagen (COL10A1), vascularization, and focal calcification. These features are similar to the normal endochondral ossification process that takes place in the growth plate44, where proliferating chondrocytes secrete Chondromodulin-I, Tenomodulin, and Sox to inhibit angiogenesis, while hypertrophic chondrocytes promote angiogenesis through hypoxia-inducible factor 1 (HIF) and vascular endothelial growth factor (VEGF) signaling to and recruit blood vessel invasion45. Inflammation and angiogenesis are closely correlated; angiogenesis may enable leukocyte extravasation into tissues by increasing the total endothelial surface, and several cytokines, chemokines, CAMs (cell adhesion molecules), and growth factors can also modulate neovascularization46. We predicted Leukocyte extravasation signaling (LES) pathway to be activated in MCC (Fig. 1c,d,f), in particular at deeper zones where chondrocytes hypertrophy very rapidly42 (Fig. 2a,b). Leukocyte recruitment into tissue across the endothelium requires four steps: rolling, tethering, firm adhesion, and diapedesis47, and involves the participation of different adhesion receptors such as selectins, integrins and immunoglobulin superfamilies48. In our IPA, matrix metalloproteinases (MMP3, MMP9, MMP10, MMP12, MMP14, and MMP28), chemokinases (CXCL12,CXCR4), and claudins (CLDN11,CLDN22,CLDN5) were differentially expressed in relation to LES canonical pathway (Supplementary Tables 5–14). MMPs are enzymes which can degrade collagen, proteoglycans, and other extracellular matrix components, simultaneously. These enzymes are tightly regulated by several growth factors, cytokines, specific tissue inhibitors of MMPs (TIMPs)49. The abundance of MMP9, MMP12 and TIMP1 in MZ & HZ of the MCC, alongwith the substantial downregulation of MMP3, TIMP3 and TIMP4 shown in our results further affirm the importance of balancing the expression of MMPs to TIMPs in cartilage microenvironment to maintain its integrity49. In corroboration of the crucial role of chemokines in leukocytes recruitment, we found CXCL12, CXCR4 to be substantially expressed in the deep layers of MCC. Chemokines are chemoattractant cytokines that stimulate cell movement and migration signaling events, in particular leuckocyte trafficking, they also induce many other biologic processes such as cell proliferation, survival, development, and angiogenesis under both physiological and pathological conditions50,51. In addition to matrix metalloproteinases and chemokines, our results demonstrated modulation of three members of cluadins family. Studies have shown that claudins, which are integral membrane proteins and tight junction proteins, may be involved in cell adhesion52. Claudin 11 (CLDN11), a major component of central nervous system (CNS) myelin, was abundant prenatally in developing meninges, mesoderm, and adjacent to cartilage, indicating its major role in growth and differentiation of not only oligodendrocytes but also other cells outside CNS53.
Lipids such as phospholipids, cholesterol and fatty acids in cartilage are important as source of energy for cells. They are also an essential constituent of cellular membranes, and play a role as signalling molecules54,55. High cholesterol levels are associated with osteoarthritis, whereas cholesterol synthesis inhibition reported to be associated with skeletal dysplasias; confirming the important role of cholesterol biosynthesis in chondrogenesis56. Genes-encoding proteins necessary for cholesterol biosynthesis, such as acetyl-coenzyme A acetyltransferase 1 (ACAT1), cytochrome P450 oxidase, family 51, sub-family A, polypeptide 1 (CYP51A1), 3-hydroxy-3-methylglutaryl-coenzyme A synthase 1 (HMGCS1) or 7-dehydrocholesterol reductase (DHCR7), have been detected to be highly expressed in the superficial zones of MCC (Supplementary Table 1) where Superpathway of Cholesterol Biosynthesis is also predicted to be activated according to our results (Fig. 1a,e,f),. Previously reported data has shown that cholesterol and lipid biosynthesis are crucial for regulation of differentiation, proliferation and apoptosis of undifferentiated mesenchymal cells in the growth plate, probably via regulating many other signaling pathways, such as Wnt signaling and Hedgehog (Hh) signaling56,57. Upstream regulator analysis of our microarray data identified RICTOR (rapamycin-insensitive companion of mTOR), SREBF1, SREBF2 (SREBPs are sterol regulatory element-binding proteins), and SCAP (SREBP cleavage-activating protein), which have been implicated in the process of cholesterol synthesis, among the most activated upstream regulators in the superficial FZ & PZ of MCC (Table 4 and Supplementary Fig. 13). The mammalian target of rapamycin (mTOR) is a serine/threonine protein kinase that regulates the phosphorylation of many proteins, and has two functional complexes; mTORC1 and mTORC2. RICTOR, which is a subunit of mTORC2, regulates cell metabolism, growth, proliferation and survival in response to growth factors and hormonal signals54,58. In addition to protein synthesis, mTOR is also a critical regulator of lipid biosynthesis through SREBF1/SREBP1 but little is known about mTOR lipid-induced responses in chondrocytes. SREBPs and SCAP regulate intracellular cholesterol biosynthesis, when cholesterol levels are low, SCAP/SREBP complex allows proteases to cleave SREBP and then to traffic to the nucleus where target genes for the biosynthesis of cholesterol are activated. Conversely, when intracellular levels are high, cholesterol biosynthesis is prevented by tethering the SREBP/SCAP complex to the endoplasmic reticulum membrane. Studies showed that Hedgehog signaling and intracellular cholesterol synthesis regulate each other. Activation of this signaling pathway, which regulates SCAP expression, induces cholesterol accumulation, which is crucial for chondrocytes proliferation and differentiation56,57. The predicted activation of Superpathway of Cholesterol Biosynthesis and of RICTOR, SREBF1, SREBF2 and SCAP upstream regulators at FZ and PZ of MCC (Fig. 1, Table 4, Supplementary Fig. 13), is consistent with that the undifferentiated cells of these superficial zones have high metabolism and require high levels of cholesterol and lipids, whereas the differentiated or nearly differentiated cells of the deeper zones (MZ &HZ) exhibited comparatively predicted inhibition of such regulators in our bioinformatic analysis.
It is evident that functional crosstalks exist between the signaling pathways involved in endochondral ossification process. Interestingly, studies showed that Hh signaling crosstalks with the Notch signaling, fibroblast growth factor (FGF) pathway, Wingless-related integration site (Wnt) signaling, bone morphogenetic protein (BMP) signaling, and mTOR signaling pathways44. Likewise, Wnt pathway may interact with BMP, Hh, FGF and TGF-β (transforming growth factor) signaling pathways59. Another intriguing feedback loop between PTHrP (Parathyroid-hormone-related protein) and Ihh (Indian hedgehog) signaling pathways was found to be involved in the homeostasis of articular cartilage and growth plate cells44. Furthermore, Hedgehog signaling can regulate cholesterol homeostatic genes; indicating a feedback loop in chondrocyte differentiation56,57. Unraveling the underlying mechanisms of these feedback loops and crosstalks will further provide important insights and enable better understanding of such interactions which take place in cartilaginous tissues. While numerous underlying pathways still remain unknown, IPA of zone-specific microarray data generated an abundance of data with large number of differentially expressed genes, and identified lists of activated/inhibited different signaling pathways and upstream regulators (Supplementary Tables 1–15). All of these cannot be introduced and discussed in this study but one cautionary note when interpreting bioinformatic data is to categorize the identified molecules and/or genes as either suppressors or promoters with caution. Rather than this binary assignment, it is strongly recommended to evaluate it as highly specialized and complicated balance of several bioactive molecules that is needed to maintain tissue homeostasis.
Since numerous properties are shared, rat MCC was chosen as a model of normal developmental processes taking place in the human MCC. In the present study, we selected the age of 5 weeks not only because MCC articulation function is already present in a more mature state, but also the maximum growth spurt for rats occurs at day 31.5. Accordingly, this age will allow studying and detecting genes expression profiles at a larger and broader scale in relation to both articulation and growth functions. Investigating normal conditions at different ages can be considered as baseline studies for future disease-related studies. Studying older age groups is also valuable, especially for evaluating osteoarthritic changes and cartilage degeneration. Gender is another important factor, literature showed that 80% of individuals seeking treatment for TMJ disorders are females of childbearing age. Such a high prevalence suggests a role for female hormones, particularly estrogen, in the disease process. In fact, this is the reason behind not selecting female rats as an animal model. Nevertheless, we consider this experiment as a baseline for future zone-specific studies of the mandibular condylar cartilage at which both male and female genders at different age groups can be studied and compared against each other.
One of the drawbacks of the MAA experiments is the incomplete relevance between the transcripts level determined and the corresponding proteins level. The fact that the differential expressions in mRNA do not necessarily reflect similar changes in proteins could be attributed to that the MAA technology is not related to posttranslational changes and posttranscriptional regulations60,61. Another limitation in this study was conducting the MAA experiment with no replicates. Although replication is needed to improve the data quality, the appropriate number of replicates is largely dependent on the research question to be answered. For instance, more replicates are required to confidently identify novel genes62, conversely, if the purpose of the study is to formulate a well-ground hypothesis, the issue of sample replicates is not very critical, specially if the MAA is combined with other more sensitive molecular analysis for validation such as qPCR. The limited availability of sample material in our study (very small MCC zones in size) and the relatively high cost of microarray chips and LCM kits have limited the number of biological or technical replicates. While noting that there are no firm standards on the number of replicates required in a microarray chip experiment, Bryant et al. found that the variability attributable to technical and biological variation in a typical in vitro microarray experiment in humans is low, and markedly less than the effect on gene expression of stimulation (MCC zonal architecture in our case)63. Additionally, MAA experiments designs that allow multiple independent estimates of treatment effects may allow reduced replication, or even no replication as stated by Maindonald et al.64. Such design was applied in our experiment when the four MCC zones were compared against each other64. For example, there are two estimates of the comparison between FZ and PZ: one obtained directly by comparing the two zones, and the other estimate is obtained by subtracting the FZ versus C effect from the PZ versus C effect. At the end, the results will provide an overview to allow one to claim that hypotheses can be formulated and prioritized for later work. However, for all the limitations, the current study revealed several new aspects in relation to MCC cell phenotypes, which may offer some clues to research process in this area and contribute to the future therapeutic approaches for MCC diseases and conditions.
In summary, by using a rat genome expression array with more than 31,000 probe sets, a comprehensive evaluation of genome-wide expressions was possible using LCM and MAA technologies, and robust gene expression differences were revealed, supporting the hypothesis that differential gene expression exists between articular chondrocytes of the FCC and MCC cells on the one hand, and different gene profiles exist among the four zones of the MCC on the other hand.
The current study also demonstrated that the MCC zones clearly exhibited differences in the activation/inhibition status of many canonical pathways which appear to be largely dependent on spatial (regional) expression of multiple factors that connect different signaling pathways leading to cartilage/chondrocyte development, maturation and homeostasis. Our results can undoubtedly be used in the future studies for exploring gene–gene interactions and signaling cascades which is crucial for the discovery of new therapeutic strategies for this intriguing cartilaginous tissue.