High hydrostatic pressure induces pro-osteoarthritic changes in cartilage precursor cells: A transcriptome analysis

Due to the high water content of cartilage, hydrostatic pressure is likely one of the main physical stimuli sensed by chondrocytes. Whereas, in the physiological range (0 to around 10 MPa), hydrostatic pressure exerts mostly pro-chondrogenic effects in chondrocyte models, excessive pressures have been reported to induce detrimental effects on cartilage, such as increased apoptosis and inflammation, and decreased cartilage marker expression. Though some genes modulated by high pressure have been identified, the effects of high pressure on the global gene expression pattern have still not been investigated. In this study, using microarray technology and real-time PCR validation, we analyzed the transcriptome of ATDC5 chondrocyte progenitors submitted to a continuous pressure of 25 MPa for up to 24 h. Several hundreds of genes were found to be modulated by pressure, including some not previously known to be mechano-sensitive. High pressure markedly increased the expression of stress-related genes, apoptosis-related genes and decreased that of cartilage matrix genes. Furthermore, a large set of genes involved in the progression of osteoarthritis were also induced by high pressure, suggesting that hydrostatic pressure could partly mimic in vitro some of the genetic alterations occurring in osteoarthritis.


Introduction
During normal physical activity, physiological contact pressures in human articular cartilage vary between 0 and several MPa, reaching around 20 MPa during more intense activities [1] [2]. Due to the high water content of cartilage (around 70%), most of the compressive load is sustained by the interstitial fluid [3], providing cartilage its high stiffness under compression and protecting the collagen-proteoglycan matrix from excessive deformation. As a result, it is assumed that one of the main physical stimuli sensed by chondrocytes embedded in the cartilage matrix is hydrostatic pressure (HP).
At the cellular level, moderate HP has been shown to promote cartilage differentiation by increasing the expression of cartilage matrix proteins such as aggrecan and type 2 collagen [4] PLOS ONE | https://doi.org/10.1371/journal.pone.0183226 August 16, 2017 1 / 17 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 [5], and down-regulating the expression of matrix-degrading enzymes such as MMP-1 through a mechanism involving the transcriptional regulator CITED2 [6]. In cultured bovine chondrocytes, moderate pressures have no obvious effects on the cytoskeleton [7] and cyclical pressures even reverse the adverse effects of IL-1β in human chondrocytes [8]. In vivo, moderate joint loading through physical exercise has been shown to prevent the onset and slow the progression of osteoarthritis [9]. Conversely, excessive HP has been shown to induce opposing, detrimental effects on cartilage and chondrocytes, with increased apoptosis, decreased extracellular matrix synthesis, increased expression of stress-and inflammation-related genes as well as matrix-degrading enzymes [4][5] [10]. High pressure also destabilizes the cytoskeleton [7] and induces changes in chondrocyte morphology reminiscent of those observed in osteoarthritic cartilage [11], illustrating the essential role of hydrostatic pressure mechanotransduction in the homeostasis of cartilage. In order to identify genes sensitive to high HP, Sironen et al. performed a cDNA array analysis of chondrocyte cell lines submitted to high HP (30 MPa) for up to 12 h [12]. However, this study focused on less than 600 genes and no global analysis of the effects of hydrostatic pressure on chondrocytes has been carried out so far.
In this study, we used whole genome microarrays to identify new genes modulated by high HP in the mouse ATDC5 chondrocyte precursor cell line. After 1, 4 and 24 h, pressurization under 25 MPa was found to modulate the expression of several hundreds of genes, including newly identified mechano-sensitive genes.

Cell culture
Mouse ATDC5 chondrocyte progenitor cells were obtained from the Japanese Collection of Research Bioresources Cell Bank and cultured in DMEM/F12 (Life Technologies) supplemented with 5% of FBS (Nichirei) in a humidified incubator under a 5% CO 2 atmosphere. Culture medium change was carried out every other day.

Cell pressurization
High hydrostatic pressure was applied for up to 24 h using a mechanical system (Fig 1a) consisting of a pressure chamber placed inside a water bath set at 37˚C, and connected to a highpressure cylinder operated by a manual pump. ATDC5 cells seeded at 2.10 4 cells.cm -2 in 35 mm Petri dishes and cultured for 3 days until the cells formed a confluent monolayer were sealed in polyethylene bags (Seisan Nippon Sha Ltd) filled with 20 mL of culture medium and placed in the pressure chamber where a constant 25 MPa pressure was applied. The internal pressure and temperature of the chamber were continuously monitored by a pressure sensor. Unpressurized cells were sealed in similar bags and placed in a metallic column of the same dimensions as the pressure chamber in the same water bath. For each experiment, four dishes were used. In order to obtain the control, 1 h, 4 h and 24 h pressurized samples, at the start of the experiment, one dish was pressurized and the three remaining dishes were placed in the water bath; 20 hours later one dish was moved from the water bath to the pressure chamber; 3 hours later, one dish was moved from the water bath to the pressure chamber; then, 1 hour later, all samples were collected at the same time (Fig 1c).

Cell count
Cells seeded on Petri dishes at 10 5 cells.cm -2 were left to grow for three days and pressurized, or not, for the last 6 or 24 h. After fixation for 15 min with 4% paraformaldehyde, cell nuclei were stained with DAPI. Five pictures were taken randomly and the cells counted manually. Three independent experiments were carried out.

Microarray experiments
The transcriptome of pressurized and unpressurized ATDC5 cells was analyzed using the Sur-ePrint G3 Mouse Gene Expression v2 8x60K Microarray Kit (G4852B, Agilent) containing just over 56,000 probes (representing 27,122 genes and 4,578 long non-coding RNAs). Two timecourse experiments (consisting of control, 1 h HP, 4 h HP and 24 h HP samples) were carried out, producing 8 RNA samples. The quality of the RNA was checked by agarose gel electrophoresis and found to be satisfactory, with two clear bands and no sign of degradation (Fig 1b). The RNA was then reverse-transcribed, labeled with Cy3 and hybridized to one microarray according to the manufacturer's instructions. After hybridization, fluorescence signals were measured with a G2505C microarray scanner system (Agilent).

Microarray data analysis
The raw data on each chip was normalized to the 75 th percentile of measurements taken from that chip. Genes, whose expression under pressure was more than doubled or halved compared to the control (fold change >2 or <0.5) in both experiments, were considered to be modulated. A rank product test was performed to assess the significance of the gene modulations found in the microarray. Venn diagrams were produced using Venny2.1.
Functional clustering of modulated genes was performed using DAVID 6.7 to identify over-represented groups of functionally-related genes.
Network analysis was first carried out using EGAN 1.5; genes significantly modulated with a p value<0.015 were selected to generate preliminary networks; subsequently, the genes identified by EGAN as being part of a network were then input into STRING 10.0 for confirmation. Interactions with a combined score above 0.9 were then used to generate reader-friendly networks with Cytoscape 3.4.0 [13].
Comparison between our data and previously published microarray data on surgically induced OA was carried out as follows: five microarray studies were selected on surgically induced OA, which included three mouse models (Gardiner et al. [14], Bateman et al. [15], Loeser et al. [16]) and two rat models (Wei et al. [17], Appleton et al. [18]), and genes modulated at least 2-fold were identified and compared to our list of modulated genes.

Real-time PCR
Total RNA was extracted using Trizol (Invitrogen) and cDNA was synthesized from 500 ng of RNA using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). Real-time PCR was then carried out using Thunderbird SYBR qPCR Mix (Toyobo) in a StepOnePlus real-time PCR system (Life Technologies). Rpl13a was used as the reference gene. The primer sequences are listed in Table 1.

Statistics
All graphs show the mean +/-standard deviation. The statistical significance of the differences induced in gene expression by HP was assessed using a two-tailed unpaired t-test. A rank product test was carried out for microarray results.

Results
High HP strongly affects the gene expression profile of ATDC5 cells Within 1 h of pressurization, several hundred genes were already modulated by high HP. The number of modulated genes increased after 4h and remained stable after 24 h (Fig 2a).
Comparison of the modulated gene lists over time showed that about half the early modulated genes were also modulated after 4 h. The genes modulated after 24 h were quite different from those modulated at 4 h (less than 20% overlap), though 204 genes were modulated throughout from 1 h to 24 h of pressurization (Fig 2b). Nine genes down-regulated at least 16-fold (1700019P21Rik, Alpk2, Far2os2, Gm9891, Gucy1b3, Idh3g, Lrrc4c, Rsl1, and Yipf7) were found to be stably down-regulated over several time points (Fig 3). Table 2 shows the ten most significantly up-and down-regulated genes for each time point.

High HP adversely affects the cartilage phenotype
A search for over-represented functional groups with DAVID 6.7 indicated that high pressure induced cartilage de-differentiation and cell death in ATDC5 cells (Table 3): after 1 h of pressurization, genes involved in cellular response to stress and proteolysis were up-regulated, whereas genes classified as part of the primary cilium, a cell projection with a major role in chondrogenesis [19], were down-regulated; after 4 and 24 h, genes involved in cell death/ growth arrest were induced, whereas genes involved in the extracellular matrix were downregulated after 24 h of HP.
Network analysis (Fig 4) revealed several groups of interacting proteins among the significantly modulated genes. After 4 h of pressurization, for instance, groups involved in heat shock, cell death, Wnt signaling and receptor tyrosine kinase (RTK) signaling were found among the networks; after 24 h, groups involved in response to stress, G protein-coupled receptor (GPCR) signaling and chondrogenesis could be observed. Notably, at 4 h, the groups of genes involved in heat shock and cell death were all up-regulated, whereas at 24 h, the group involved in the response to stress was up-regulated while genes implicated in chondrogenesis were down-regulated, indicating overall that, in ATDC5 cells, high pressure first induces a stressful response followed by a reversal of the chondrogenic phenotype.

Real-time PCR validates most microarray results
To check the validity of the microarray results, 17 genes related to OA and modulated after 1, 4 or 24 h of pressurization were selected and their expression was measured by real-time PCR (Fig 5). Those genes were selected based on the following criteria: 1) microarray fold change > 2; 2) microarray expression level > -5 (to omit weakly expressed genes); 3) potential implication in chondrogenesis or osteoarthritis. Apart from a few genes at certain time points (unlike the microarray results, Acan was not up-regulated after 1 h, Errfi was not induced after 4 h of HP, Arntl and Cd14 were down-regulated after 4 h of HP), the PCR results closely matched the microarray results. Notably, Adamts4 was strongly down-regulated and Adamts5 transiently up-regulated at 4 h; Acan and Col2a1, which control the synthesis of aggrecan and type 2 collagen, were strongly inhibited after 24 h; Ctgf was up-regulated after 4 and up to 24 h; Ctsk was strongly induced at 4 h; Cytl1 was markedly reduced at 24 h; Ddit3 and Gadd45a were up-regulated after 4 and up to 24 h; Cited2 was down-regulated within 1h and for up to 24 h.

High HP inhibits ATDC5 cell growth
As genes involved in apoptosis/growth arrest were found to be over-represented among upregulated genes and the up-regulation of Ddit3 and Gadd45a was confirmed by real-time PCR, we investigated whether high HP affected the growth of ATDC5 cells. A cell count after 0, 6 or 24 h of pressurization showed a significant reduction in the number of cells in pressurized dishes (Fig 6).

High HP modulates genes similarly to osteoarthritis
Seeing that several genes involved in OA were up-regulated by high pressure, we compared our microarray results with previously published data on surgically induced OA in mice and rats. After comparing genes modulated at least 2-fold in the five selected studies and the present microarray study, we found a total of 449 genes that were modulated in the same direction for at least one time point in both our array and the set of previously reported microarrays, indicating that high HP and OA-inducing surgical procedures affect similar genes (S1 Table). A list of the 30 most modulated genes among those common genes is presented in Table 4.

Discussion
The ATDC5 cell line is now a well established cell line to study chondrogenesis and other aspects of cartilage biology [20]. Using a mouse cell line also made it easier to compare our results with previous in vivo studies on mouse models of OA, the main model involving destabilization of the medial meniscus (DMM), a procedure that leads to severe osteoarthritis within 8 weeks [21]. In addition, it has been shown that stresses generated in joint cartilage is within one order of magnitude between quadruped species (i.e. mouse or cow) [22]; consequently, it is reasonable to assume that the MPa pressures that are generated in human joints may also be generated in mouse cartilage, despite the different weights. One limitation to our study was the use of constant pressures, which, at such a high intensity, are unlikely to be found in vivo. Indeed, when studying the effects of physiological pressures, cyclical pressures are usually applied; however, high pressures (25 MPa) are just as unlikely to be generated in vivo in a cyclical manner. A more physiological approach would have been to apply a single 25 MPa impact pressure to simulate a traumatic injury, which is known to promote the onset of osteoarthritis [23]. However, traumatic injuries in vivo trigger a complex response involving several cell types such as immune cells; as our goal was to identify new gene targets of high HP, we decided to use constant pressure and limit the stimulation to 24 h.
Despite the importance of HP in cartilage homeostasis, no whole genome microarray study of the effects of HP has been carried out on chondrocytes or chondrocyte precursors. Sironen et al. used a cDNA array with under 600 probes to identify genes in a human chondrocyte cell line responsive to 30 MPa for 3, 6 or 12 h [12]. Among the genes present in their array, they found 51 modulated genes, while 151 were modulated in our array. Though the overlap was not statistically significant (p = 0.064), 18 genes were found to be modulated in common and in the same direction (representation factor of 1.4), including Fosl1, Hbegf, Ddit3 and Gadd45a, the latter suggesting that growth arrest/apoptosis may have also been observed in their cells [24]. By looking at the effects of high HP on the whole mouse transcriptome after 1, 4 and 24 h of pressurization, a broad picture of the early response of the cell to high HP could be drawn: an initial stress response followed by arrested growth, increased cell death and a decrease in extra-cellular matrix synthesis, eventually starting the chondrocyte de-differentiation process. Therefore, among the modulated genes, several genes involved in the physiopathology of OA were identified: Adamts5, which was transiently but significantly induced under HP, produces an aggrecanase found responsible for aggrecan degradation in a mouse model of OA [25]. Arntl is implicated in chondrocytes' circadian rhythm, and loss of this gene leads to progressive cartilage degradation in mice [26]. CD14, an antigen usually expressed by immune cells, has also been found in human chondrocytes [27] and may play a role in lymphocytemediated cartilage degradation [28]. CITED2 is a transcriptional co-activator known to suppress the expression of the collagen-degrading enzymes MMP-1 and MMP-13 [29][6] and its strong down-regulation further indicates anti-chondrogenic effects of high HP in ATDC5 cells. Ctgf was markedly up-regulated after 24 h of pressure and is strongly expressed in osteoarthritic cartilage [30]; although CTGF may be important for cartilage differentiation and maintenance [31], its over-expression has also been shown to cause cartilage damage [32]. Ctsk, which produces the collagen-degrading protease cathepsin K, was also transiently but strongly up-regulated by high HP, and knockdown of this enzyme in mice delays the progression of the disease in surgically induced OA [33]. Cytl1, strongly down-regulated by 24 h of HP in ATDC5 cells, is a cytokine-like protein involved in chondrogenesis, whose knockdown leads to more severe cartilage destruction in a mouse model of OA [34]. Ddit3 produces a transcription factor found to participate in cartilage degradation in a mouse model of OA [35]. Errfi1 deletion in mouse cartilage leads to an OA-like phenotype in the knees, though rarely in other joints [36]. Fosl1, markedly up-regulated in our study, produces a FOS-related transcription factor that has been found to be potentially involved in the regulation of gene expression in human osteoarthritic cartilage [37]. Hbegf, also transiently up-regulated in our study, produces an EGF-related growth factor, and has been found to be over-expressed in damaged cartilage from patients suffering from OA [38]. Ptgs2, which produces cyclooxygenase 2, was also up-regulated after 24 h; its role in OA, however, is still controversial, some The list shows the 30 most up-or down-regulated genes in our microarray also found modulated in surgically induced OA in mice and rat models. The complete list of 449 genes found in common in our array and the five animal OA studies is given in S1 Table. https://doi.org/10.1371/journal.pone.0183226.t004 studies showing that cyclooxygenase 2 inhibition protects against OA [39] while others show no effect [40]. Surprisingly, Adamts4, which is known to be expressed by chondrocytes in osteoarthritic cartilage [41] and also participates in aggrecan degradation in a model of human OA [42], was significantly down-regulated for up to 24 h. However, the respective importance of Adamts4 and Adamts5 in the onset and progression of OA is still under investigation. The expression of Nfil3, a transcription factor essential to the development of certain immune cells [43], was transiently up-regulated in our study but found down-regulated in osteoarthritic cartilage [44]; the role of NFIL3 in cartilage is still poorly understood and warrants further investigation.
Overall, the microarray results seem to point to osteoarthritic-like effects of high HP on chondrogenic cells. In agreement with the PCR results, several hundred genes modulated by HP were found to be similarly modulated in cartilage from rodent OA models. ATF3, for instance, is a transcription factor, whose expression is increased in OA cartilage and which is involved in the progression of the disease in a mouse OA model [45]. Nfe2l2 produces a transcription factor, which is known to protect cartilage from degradation in mouse OA models [46]. The overlap between the present results and the previously published microarray studies used for comparison, however, was not statistically significant. This may be due to the different time points used; 24 h of pressurization may only trigger events seen at the very onset of the disease, and longer pressurization periods may be necessary to lead to a more OA-like gene expression profile. Nevertheless, among the newly identified HP-modulated genes, may lie some yet poorly known genes essential to the pathogenesis of OA, which further studies will help to uncover.
In this study, no growth factors or hormones were added to the medium to promote ATDC5 differentiation into mature chondrocytes before HP was applied. However, ATDC5 cells were seeded at a relatively high density and were fully confluent for one day before the start of the pressurization. Real-time PCR confirmed that they thus clearly expressed Sox9, Acan and Col2a1. Moreover, we found that 25 MPa strongly inhibits Acan and Col2a1 expression; this is similar to the effects of high HP (20 to 50 MPa) reviewed by Elder et al. [4], which include a decrease in collagen and aggrecan gene expression in bovine chondrocytes. In addition, one technical difficulty with using differentiated ATDC5 cells is that the cell population obtained after differentiation may not be homogeneous, making the results difficult to interpret. Though investigating the effects of pressure on primary (differentiated) chondrocytes may be an interesting future study, we believe the present study using a homogeneous population of ATDC5 cells yielded meaningful data on the mechanobiology of hydrostatic pressure.
As cells do not contain any gas phase and are considered almost incompressible, physiological HP does not induce any visible deformation of cell compartments or the cell as a whole, making the mechanotransduction mechanisms difficult to identify. However, within 1 h, several hundreds of genes responded to pressurization, and it is quite possible that several independent mechanisms could be implicated in the sensing of HP: for instance, HP may affect the conformation of a pressure-sensitive protein, though studies showing visible changes in protein conformation usually deal with pressures above 100 MPa [47]; HP may alter membrane fluidity and indirectly affect the conformation or the binding of signaling molecules [48]; HP could also alter forces within the membrane by increasing its bending rigidity [49], generating forces sufficient to trigger a mechano-chemical response.
As physiological pressures are known to be beneficial to cartilage, and high HP induces changes in cells reminiscent of those observed in OA, identifying the pressure mechanotransducer(s) in chondrocytes and their downstream pathways may be a key to better understand and prevent the progression of OA.
Supporting information S1 Table. Genes modulated similarly in the present study and in rodent OA models. List of all the 449 genes found to be modulated in both the present study and in the five studies by Gardiner