Temporomandibular joint (TMJ) is described as a synovial sliding-ginglymoid, load-bearing joint where the mandibular condyle is capable of not only rotary (hinge) but also translatory (sliding) movements during the daily tasks including chewing, speaking, swallowing, and yawning1,2,3. Temporomandibular joint disorder (TMD) is a class of degenerative musculoskeletal conditions manifested as deformities in the morphology and function of the TMJ3,4. It includes abnormal position and/or structure of the TMJ disc and dysfunction of the associated musculature of the face (orofacial pain)3,5. TMD together with other conditions such as arthritis, congenital anomalies, and injuries/truama to the TMJ are significant causes of morbidity that can negatively affect the quality of life of human beings.

Clinical management starts with noninvasive treatment modalities (physical therapy, occlusal splints, and prescription of pharmacologic agents), but for some patients other minimally invasive strategies (injections of sodium hyaluronate and/or corticosteroids, arthrocentesis, and arthroscopy) are considered3. For those patients who show no improvement with the nonsurgical treatment modalities, open joint surgery may be carried out for discectomy, reshaping or reconstructing the articulating surfaces, and total joint replacement which is the most invasive option3. Recently, there is an emphasis to apply cell-based regenerative therapies5,6, but one common problem is the unfamiliarity with and lack of thorough understanding of native tissue characteristics3.

The advent of microarray technology has enabled the researchers to analyze the expression of thousands of genes simultaneously in a single experiment7, and provided a huge amount of information about gene expression for cells8. However, the reliability and validity of such molecular studies is totally dependent on the procurement of pure cell populations with relatively high abundance9. Hence, the two impediments are the heterogeneity of the native tissues and the abundance of the biomolecules extracted from cells. The larger the number of “contaminating/unwanted” cells upon procurement, the greater the chances of getting false results and inaccurate interpretation of the data9. Based on the histology, typical articular cartilage can be divided into three distinct layers: the superficial, middle, and deep zones, with variation in the morphology and density of the chondrocytes10,11,12. The mandibular condylar cartilage (MCC), a fibrocartilage layer that covers the mandibular condyle of TMJ, is also multizonal in structure but it is unique in terms of cell phenotypes. From the articular surface of MCC to the underlying bone, the following four zones are identifiable: fibrous (FZ), proliferative (PZ), mature (MZ), and hypertrophic (HZ) zones13,14. Cells have traditionally been harvested from MCC by mincing the tissue with a scalpel, followed by enzymatic digestion; consequently, the outcome includes various cell types from the four different zones15. Gross sampling may conceal the individual gene expression profiling of zone-specific cell populations16. On the other hand, Laser-Capture microdissection (LCM) technology, allows precise procurement of cells of interest from a heterogeneous tissue in a relatively rapid and practical manner17. Moreover, DNA (Deoxyribonucleic acid), RNA (Ribonucleic acid), and proteins can be extracted from the homogenous isolates of cell population18, and that could be followed by an array of analytical applications, which allows reliable studying of in vivo genomic and proteomic profiling of the tissue of concern. Generating microarray data from LCM samples is feasible9, however, such a combination should be capable of coping with several challenges19,20.

Beside its unique histomorphological zonal organization, MCC is distinguished from growth and articular cartilages in the embryonic origin, biochemical and biomechanical properties14,21,22,23. Functionally, MCC has a dual role; one as an articular fibrocartilage responsible for load distribution and disc articulation, and the other one as a major site responsible for mandibular growth24. Additional importance of the MCC stems from its role in life-long bone remodeling process. In light of these distinctive features, it would not be surprising if the molecular and genetic regulation of the biological processes of MCC were different from those of other articular hyaline cartilages and epiphyseal growth cartilages. In addition, the MCC is a multi-zonal fibrocartilage containing different types of cells which are well characterized histomorphologically6,25 but the factors governing their morphological transition across the zones have not been fully understood. Therefore, it is logical to speculate that unique genetic profiles in vivo might exist across the four MCC zones.

A crucial prerequisite for functional replacement of affected/diseased MCC is to generate accurate knowledge at cellular and molecular level. Combined LCM and MAA (microarray analysis) enable large-scale in situ studies that could clarify many hidden or masked diagnostic and therapeutic aspects which were not previously identified16. However, very little is reported on the regional or zonal genetic profiling and molecular phenotypes of MCC cells in the literature6,26,27. The aim of our study is to perform a comprehensive gene expression profile analysis with a specific focus on pathways analysis for each zone of the MCC from 5-week-old rats using LCM and MAA, and to formulate a well-supported hypothesis to identify genes which could potentially distinguish the cells of MCC zones from each other and from the articular chondrocytes.

Materials and methods

Preparation of LCM samples and RNA extraction

Animal use protocol was approved by the Committee on the Use of Live Animals in Teaching and Research of the University of Hong Kong (CULATR 2311-11), and the procedures were carried out in accordance with the institutional guidelines and in accordance with ARRIVE guidelines ( Sprague–Dawley 5-week-old male rat (Rattus norvegicus) was sacrificed by intraperitoneal injection using 20% Dorminal (200 mg pentobarbital sodium, Alfasan, Woerden-Holland, Netherlands) with a dose of 100 mg per 100 g of body weight. MCC and FCC specimens were harvested, and then our optimized LCM protocol described earlier16 was applied to collect RNA from the four zones of MCC individually thereby four groups: FZ, PZ, MZ, and HZ were prepared (Supplementary Fig. 1a). Similarly LCM sample was prepared by microdissecting the chondrocytes from the middle and deep zones of FCC tissue (group C) as a control (Supplementary Fig. 1b). Briefly, the freshly dissected specimens were embedded in optimal cutting temperature (OCT), frozen on precooled isopentane (Sigma-Aldrich, USA). Frozen sections were cut at 7 μm thickness using a cryostat set to − 24 °C to − 30 °C, and were mounted on glass microscope slide (HistoBond + adhesive microscope slides, Marienfeld laboratory glassware, Germany). Slides were processed using Arcturus HistoGene LCM Frozen Section Staining Kit (CA, USA) was used according to the manufacturer’s instructions, but with a few modifications. The slides were individually placed onto the microscope platform of the Arcturus PixCell II Laser Capture Microdissection System (CA, USA). Under 10× objective, the unstained tissue sections provided adequate morphology to distinguish MCC zones. LCM was performed and the collected dissected cells were immediately lysed in 50 μL of lysis buffer (Arcturus PicoPure RNA Isolation Kit, Applied Biosystems, CA, USA) as recommended by the manufacturer. The cell lysate of 30 microdissected tissue sections from each zone were and pooled to create single samples, except for FZ group where the lysates of 60 microdissected sections were pooled. RNA was isolated from the pooled samples using Arcturus PicoPure RNA Isolation Kit (Applied Biosystems). To assess RNA quality, 1 μL was analyzed using RNA 6000 Pico kit and Bio-Analyzer 2100 (Agilent Biotechnologies). RNA integrity number (RIN) values were > 5 which is within the accepted range for LCM samples28.

Amplification, labeling, fragmentation, and hybridization of RNA samples

Eleven µl of each sample was used as total RNA input material for a two-cycle linear amplification process in accordance to the protocol provided with the Arcturus RiboAmp HS Plus amplification Kit (Applied Biosystems, CA, USA) ( After the first strand cDNA synthesis reaction of the first amplification round, 2 μl of each sample was removed to assess the integrity of the starting mRNA by evaluating the 3′/m ratio of GAPDH and β-Actin housekeeping genes via the qRT-PCR assay as described previously16. Concurrently, RNA control provided with the kit was also amplified to calculate the amplification efficiency by dividing the RNA yield after amplification over the initial RNA input (500 pg). Following the amplification process, quality control measurements were employed to determine the mRNA transcript length using Agilent 2100 Bioanalyzer with a total RNA 6000 Nano-lab-Chip (Agilent technologies, CA, USA). RNA was quantified using a Nanodrop 2000c spectrophotometer (Thermo Scientific) and assessed for purity where a ratio of two optical densities; A260 and A280 (OD 260/280) was calculated. Labeling of aRNA transcripts was performed using the Arcturus Turbo Labeling Biotin Reagent Kit (Applied Biosystems, Amsterdam, The Netherlands) and according to the manufacturer’s instructions ( After quantification, 16.8 µg of the labeled aRNA samples were hydrolyzed in a fragmentation buffer (5×) solution of the GeneChip 3′ IVT Express Kit (Affymetrix, Santa Clara, CA, USA). The protocol followed was in accordance with the Arcturus Turbo Labeling Kit with Biotin manufacturer’s instructions ( Fragment size of the products was confirmed using the Agilent 2100 Bioanalyzer. Fragmented and labeled aRNA samples (12.5 μg in 25 μl each) were sent to the Center for Genomic Sciences/The University of Hong Kong (CGS/HKU) facility where microarray hybridization, scanning and image analysis procedures were carried out using GeneChip Rat Genome 230 2.0 Array (Affymetrix, Santa Clara, CA, USA), which contained 31,099 probesets, according to the Affymetrix GeneChip Expression Analysis Technical Manual ( Each probe set in the array is represented by 11 pairs consisting of 25mer oligonucleotides, and each pair includes a perfect match oligo (complementary to the aRNA target) and a mismatch oligo to calculate the nonspecific and background hybridization.

Microarray data have been submitted to the National Centre for Biotechnology Information (NCBI) and was deposited in gene expression omnibus (GEO) (accession number: GSE162823).

Microarray data analysis

Following hybridization, the array was washed, stained, and scanned, and then resultant data were analyzed. Microarray Suite version 5 (MAS5.0 Affymetrix, Santa Clara, CA, USA) software enables both probe set summarization, as well as initial quality examination of data. MAS5.0 probe summarization algorithms included: background correction, probe summarization (to convert probe level values to probeset expression values), and normalization. Background correction was achieved by subtracting the signal of the nonspecific binding (mismatch) probe from that of the perfect match probe for each of the 11 pairs of a probe set, then the 11 intensities were condensed to one value per gene (probe set). To identify the differentially expressed genes among the MCC zones as well as FCC, MAA data were further analyzed using GeneSpring GX version 12 software (Agilent Technologies Inc., Santa Clara, CA, USA) where the Biological Significance Analysis workflow was selected. In this workflow type, default parameters for data processing are applied, and according to the RMA (Robust Multichip Averaging) normalization approach, absolute expression values were baselined to a median of all samples expression value, normalized using the quantile scheme, and then transformed into log2 based value of relative intensity for each probe set. The data sets were then filtered by removing the lowest 20% of intensity values, leaving data from 26,121 probe sets. For fold change (FC) analysis, the ratio between the normalized intensities of a probe set belonging to two groups of the samples is calculated. The default cut off for the FC, which is log2 value of 2, was used to identify genes with expression ratios outside of this value in any of the ten pair conditions (FZ/C, PZ/C, MZ/C, HZ/C, FZ/PZ, FZ/MZ, FZ/HZ, PZ/MZ, PZ/HZ, and MZ/HZ). Any fold-change value that is less than one was replaced by the negative of its inverse.

To asses the biological functions of the differentially expressed genes at FC >  ± 1.4 (a total of 4634 genes, Supplementary Table  1), data were analyzed through the use of IPA (QIAGEN Inc., Core analysis that identifies top canonical pathways, upstream regulators, biological and diseases function, and toxicity function for each pairwise comparisons was generated. The significance of IPA core analysis was measured in two ways: (1) a Fischer’s exact test was used to calculate a p-value determining the probability that the association between the differentially expressed genes and the canonical pathway is explained by chance alone; (2) Z-score to provide predictions (increased function/ decreased function/no effect) about upstream or downstream processes.

Quantitative real-time PCR (qRT-PCR)

RNA samples were used as a template to synthesize first-strand cDNA according to the manufacturer's instructions using Superscript III Reverse Transcriptase (Invitrogen, CA, USA) and oligo(dT)12–18 (Invitrogen, CA, USA). Ten μL of LCM-RNA was used per 20 μL cDNA synthesis reaction using Veriti 96 well thermal cycler (Applied Biosystems, CA, USA). Ten genes were randomly selected from the list of differentially expressed genes of the microarray data namely; Car9, Cmtm5, Ctsz, Drd4, Dusp27, Fam180a, Gdf10, Itgbl1, RGD 1311447, and Ucma. qRT-PCR was performed using a Step One Plus RT-PCR system (Applied Biosystems, CA, USA) and Power SYBR® Green PCR master mix (Applied Biosystems, Warrington, UK). Relative quantities for the tested genes were determined utilizing the corresponding standard curves generated in the same experiment and GAPDH as the endogenous control as previously described16.


Gene expression signature

We were able to create zone-specific gene expression profiles of MCC by combining LCM and MAA technologies.

High and low absolute expression levels of genes

The top 10 genes with greatest/least absolute expression levels in each group are listed in (Table 1, Supplementary Figs. 211). Most of the highly expressed genes in group C had similar profiles that show high abundance in the FCC tissue with much lower expressed values in all MCC zones. The most interesting finding among the profiles of genes with high/low absolute expression values was the distinctive profiles of the 10 genes that showed the least abundance in PZ. These genes were uniquely downregulated at PZ but with much higher abundance levels in all other zones. In addition to this unizonal characteristic, analysis of the absolute expression values also revealed a bizonal pattern. Considerable overlap was found between FZ and PZ, for instance, Crapb1, Tnmd, Dpt, Bcl11b, Plxdc1, P4ha, Aspn, Hs3st6, Pcdh20 and Fndc1 genes tend to be highly expressed in both FZ and PZ while reduced in other zones. Similarly, the top genes identified as the greatest (or least) expressed at MZ were also increased (or reduced) for HZ. In the subset of the lowest 10 genes expressed at HZ, Fndc1 was found to have a profile different from the other 9 members where it was markedly high at both FZ & PZ zones and extremely low at HZ. For the remaining 9 genes, the expression levels at MZ were close to those of HZ especially for Angptl1, Col14a1, and Cpxm2.

Table 1 The top ten annotated genes with the greatest/lowest absolute expression levels in the FCC and four zones of MCC tissues.

Differential expression of genes by fold change analysis

The most commonly used application of MAA is identification of the differentially expressed genes rather than absolute quantification of RNA transcript abundance. Gene expression profile of each MCC zones was compared with group C (FCC tissue), in addition, MCC zones were compared against each other. Thereby, ten pairwise comparisons were established to assess the relative gene expressions; FZ versus C, PZ versus C, MZ versus C, HZ versus C, FZ versus PZ, FZ versus MZ, FZ versus HZ, PZ versus MZ, PZ versus HZ, and MZ versus HZ. Fold change (FC) analysis of a minimum log2 value of 2 in at least one of the 10 pairwise comparisons indicated that 2022 transcripts were differentially expressed (Supplementary Table 2). When MCC zones were compared with FCC (FZ/C, PZ/C, MZ/C, HZ/C comparisons), the total number of differentially expressed transcripts was 1670, of which 833 were downregulated in MCC zones relative to to FCC and 837 genes were upregulated in MCC (Supplementary Table 3). Comparisons of MCC zones with each other (FZ/PZ, FZ/MZ, FZ/HZ, PZ/MZ, PZ/HZ, and MZ/HZ comparisons) revealed 874 differentially expressed genes (Supplementary Table 4).

To focus on the strongly up- and down-regulated genes, only genes with ≥ 50 FC (or ≥ 5.64 log2 FC) were selected to compare MCC zones to FCC. The created gene subset consisted of 30 genes categorized either as transporters (Crabp1, Atp1a2), cytokines (Cmtm5), peptidases (Capn6), growth factor (Gdf10, Wisp3), transcription regulators (Hoxc9, Hoxc10), G-protein coupled receptors (Agtr2, Casr), enzymes (Ptgds), or others. The genes of this subset were extremely downregulated in MCC zones as compared to FCC except for Crabp1 (Table 2). This gene was among the highly expressed ones at FZ and PZ but with relatively reduced levels at MZ, HZ, and FCC (Table 1, Supplementary Figs. 4 and 6), indicating that Crabp1 upregulation could be a characteristic feature for FZ and PZ. Extreme downregulation of many genes (e.g. Lect1, RGD1311447, Mfi2, Mia, RGD1566401, Chad, Vit, Tpd52l1, Gdf10, A1i3/Mug1, Hoxc9, Agtr2, Hoxc10, and Ptgds) in MCC zones could be attributed to the greatly increased absolute expression of these genes in FCC tissue (Table 1, Supplementary Fig. 2). On the other side, Clec3a and Matn3 which are the most downregulated genes with − 8.92 and − 7.71 FC respectively, showed reduced absolute expression at FZ which might be a unique characteristic for this zone (Table 1, Supplementary Fig. 5).

Table 2 Top 30 differentially expressed genes that showed strong modulation in at least one of the four pairwise comparisons where each MCC zone was compared to FCC (the control, C) at ≥ 50 -fold change cut-off value (log2 50 FC = 5.64, bold and italics indicate gene upregulation and downregulation respectively above the cut-off value).

Upon comparing MCC zones against each other, genes of the six pairwise comparisons were moderately modulated, unlike the strong modulation demonstrated above when MCC zones were compared to FCC. Additionally, the predominance of genes downregulation shown in Table 2 was not as evident in Table 3. In fact, FC analysis at ≥ 20 FC (or ≥ 4.32 log2 FC) cut-off value revealed 30 differentially expressed genes among MCC zones; 5 downregulated at FZ and 25 upregulated at FZ & PZ, and can be classified as transporters (Crabp1), kinases (Ephb3), peptidases (Cpxm2), growth factor (Igf2), transcription regulators (Bcl11b, Foxa2), G-protein coupled receptors (Mrgprf), and others. Table 3 also illustrates that 24 and 7 genes were upregulated in FZ/HZ and FZ/MZ comparisons respectively, similarly 10 genes showed upregulation at PZ relative to HZ. Such modulation finding at relatively high cut-off FC value would indicate the drastic differences between the phenotypes of corresponding cells of MCC zones, particularly the non-adjacent zones. In general, most of this subset genes were moderately upregulated at FZ and PZ and downregulated at MZ and HZ. For instance, the abundance of Crabp1, Fndc1, and Dpt genes was noticeably high at FZ and PZ, and relatively low at MZ and HZ, leading to extremely high modulations especially for FZ/HZ and PZ/HZ comparisons (Table 1). On the other hand, Clec3a, Col9a1, Hils1, Foxa2, and Matn3 genes, which showed the least absolute expression at FZ, were uniquely downregulated when comparing FZ to both PZ and MZ.

Table 3 Top 30 differentially expressed genes that showed strong modulation in at least one of the six pairwise comparisons where MCC zones were compared against each other at ≥ 20 -fold change cut-off value (log2 20 FC = 4.32, bold and ilatlics indicate gene upregulation and downregulation respectively above the cut-off value).

Altered canonical pathways

When analyzing genes with FC >  ± 1.4 at threshold p-value < 0.05 for the ten comparisons, ingenuity pathway analysis identified large numbers of altered pathways, ranging from 477 to 548, (the full list of pathways for each comparison is provided in Supplementary Tables 514). The top 10 canonical pathways of each MCC zones in comparison to FCC (the control), and the top 27 pathways from the analysis of MCC zones against eachother are shown in Fig. 1a–f. To illustrate the possible changes in biological processes across MCC zones, we used the composite summary tool of IPA, and based on the z-score activation state, there was an overrepresentation of pathways with predicted inhibition at both superficial zones (FZ&PZ) as compared to MCC deeper layers (MZ&HZ) except for two cholesterol biosynthesis pathways (Fig. 1e). Conversely, most of the identified pathways showed predicted activation upon comparing MCC zones against FCC (Supplementary Fig. 12). Proliferative zone (PZ) was found to have more unique differentially expressed pathways, whilst MZvsHZ comparison showed the least significance values and activation scores (Fig. 1e,f, Supplementary Fig. 12). Leukocyte extravasation signaling (LES) pathway was predicted to be activated at all MCC zones in comparison to FCC with increased cell movement of blood cells, interaction of blood cells, invasion of cells, migration of cells, vasculogenesis, cell movement of smooth muscle cells, and migration of muscle cells. Within MCC, cell mobility, cell polarity, tail retraction, and actin cytoskeleton contraction biologic processes were predicted to be inhibitted at superficial zones as compared the deeper ones (Figs. 1c,d,f, 2). On the other hand, Superpathway of Cholesterol Biosynthesis showed positive z-score (predicted activation) in both FZ and PZ comparisons against FCC and deep MCC zones with predicted decrease in 1,25 dihydroxyvitamin D3 biosynthesis (Fig. 1a,f), Conversely, Protein Kinase A (PKA) Signaling was predicted to be inhibited in all MCC zones in comparison with FCC, and remained significantly inhibited at superficial MCC zones as compared the deeper ones (Fig. 1d, Supplementary Fig. 12). PKA is an enzyme regulates other proteins by phosphorylation, thus it is essential for many processes such as metabolic energy, cell survival, cell proliferation, muscle contraction, membrane transport and gene expression. Hepatic Fibrosis/Hepatic Stellate Cell Activation Pathway was also one of the most significant identified pathways in all comparisons, however, its state of activation could not be predicted by IPA (Fig. 1a–d). Similarly, the activation state of other significant pathways e.g. Axonal Guidance Signaling, Hepatic Fibrosis Signaling, and Osteoarthritis Pathway was undetermined.

Figure 1
figure 1

Canonical pathways identified by IPA analysis of the differentially expressed genes. (ad) The top ten significant pathways for four pairwise comparisons; each MCC zone compared with FCC (the control). (e,f) Composite summary showing the identified canonical pathways when comparing MCC zones against each other (six comparisons). The pathways are ranked by the negative log of the p-value of the enrichment score and colored according to the z-score. A positive score indicates a significantly increased function (orange), a negative z-score indicates a significantly decreased function (blue), and an undetermined prediction is shown in gray color. Data were analyzed through the use of IPA (QIAGEN Inc.,

Upstream regulators

To provide biological insight to the reported expression changes, we performed upstream regulator analysis (URA). The analysis revealed hundreds of upstream molecules that could explain the differential gene expressions observed in MAA data (Supplementary Table 15). Upstream regulators can be any gene, transcription facor or small molecule that could affect gene expression. The three most activated/inhibited upstream regulators identified by IPA analysis for each of the ten comparisons (total of 35 molecules) are listed in Table 4. These regulators affect various biological processes such as development of body trunk, development of connective tissue cells, development of hematopoietic progenitor cells, metabolism of protein, phosphorylation of protein, inflammation, cellular homeostasis, cell cycle progression, cell–cell contact, activation of connective tissue cells, cell growth and proliferation, transformation, differentiation, movement, migration, as well as cell death and survival. Regulators determined with IPA are sometimes dependent on each other, mechanistic networks are usually constructed to indicate possible signaling mechanisms (Supplementary Fig. 13).

Table 4 The three most activated/inhibited upstream regulators identified by IPA analysis.
Figure 2
figure 2

Diagram of Leukocyte extravasation signaling (LES) canonical pathways with overlaid molecular activity prediction as identified by Ingenuity Pathway Analysis (IPA) analysis of the differentially expressed genes in mandibular condylar cartilage zones and femoral condylar cartilage (FCC). The upregulated (red color) and downregulated (green color) genes identified in six comparisons, MZvC (a), HZvC (b), FZvMZ (c), FZvHZ (d), PZvMZ (e), PZvHZ (f) are shown along with predictions on biological function. Cell mobility, cell polarity, tail retraction, and actin cytoskeleton contraction biologic processes were predicted to be activated (orange color) at the deep zones (MZ & HZ) as compared FCC. These processes are predicted to be decreased (blue color) at MCC superficial layer (FZ & PZ) when compared to deeper zones. Data were analyzed through the use of IPA (QIAGEN Inc.,

Validation of microarray data by qRT-PCR

Ten genes from FCC and each MCC zones were selected to validate the MAA data by qRT-PCR. Seven genes out of 10 showed expression patterns in real-time PCR comparable to those of the MAA (Supplementary Figs. 1423).


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 514). 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 115). 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.