Substrate stiffness regulates the differentiation profile and functions of osteoclasts via cytoskeletal arrangement

Abstract Objectives Aging and common diseases alter the stiffness of bone tissue, causing changes to the microenvironment of the mechanosensitive bone cells. Osteoclasts, the sole bone‐resorbing cells, play a vital role in bone remodeling. This study was performed to elucidate the mechanism through which osteoclasts sense and react to substrate stiffness signals. Materials and methods We fabricated polydimethylsiloxane (PDMS) substrates of different stiffness degrees for osteoclast formation progressed from osteoclast precursors including bone marrow‐derived macrophages (BMMs) and RAW264.7 monocytes. Osteoclast differentiation in response to the stiffness signals was determined by examining the cell morphology, fusion/fission activities, transcriptional profile, and resorption function. Cytoskeletal changes and mechanosensitive adhesion molecules were also assessed. Results Stiffer PDMS substrates accelerated osteoclast differentiation, firstly observed by variations in their morphology and fusion/fission activities. Upregulation of canonical osteoclast markers (Nfatc1, Acp5, Ctsk, Camk2a, Mmp9, Rela, and Traf6) and the fusion master regulator DC‐stamp were detected on stiffer substrates, with similar increases in their bone resorption functions. Additionally, the activation of cytoskeleton‐associated adhesion molecules, including fibronectin and integrin αvβ3, followed by biochemical signaling cascades of paxillin, FAK, PKC, and RhoA, was detected on the stiffer substrates. Conclusions This is the first study to provide evidence proving that extracellular substrate stiffness is a strong determinant of osteoclast differentiation and functions. Higher stiffness upregulated the differentiation profile and activity of osteoclasts, revealing the mechanical regulation of osteoclast activity in bone homeostasis and diseases.

continuous reconstruction of normal bone tissue, development of new bone, and repair of traumatic bone defects are strongly regulated by physiological stimulation and external mechanical forces. 2 Bone cells are mechanosensitive, under too little stress (eg, disuse of arms and space travel), a reduction in bone mass and the development of osteoporosis will occur, whereas under excessive mechanical force, bone hyperplasia, sclerosis, and abnormal woven bone structures can ensue. It is only under moderate mechanical stress (ie, within the physiological range) that effective bone homeostasis can be maintained and bone tissue growth and reconstruction promoted. 3,4 Osteoclasts are multinucleated giant cells derived from precursors of the monocyte/macrophage lineage. As the only bone-resorbing cells in the human body, they play a vital role in maintaining bone metabolism, with osteoclastogenesis being the starting point in every round of the bone remodeling process. 5 Osteoclast formation is regulated by two critical cytokines: macrophage colony-stimulating factor (M-CSF), which ensures the survival of osteoclast precursor cells; and receptor activator of nuclear factor-kappa B (NF-κB) ligand (RANKL), which drives the downstream signaling of transcription factors for osteoclastogenesis. 6 Excessive osteoclast differentiation can lead to pathological bone loss, such as in age-related osteoporosis, Paget's disease, and inflammatory rheumatic arthritis. Conversely, the restrained activity of these cells causes a significant increase in bone density. 7 Therefore, understanding the activity of osteoclasts under multiple stimulation modes is a prerequisite to deciphering bone physiology and pathology. Alteration of the mechanical properties of bone tissue by aging and common diseases changes the mechanical microenvironment of the bone cells. 8 One important mechanical signal from bone cell surroundings, the extracellular matrix (ECM) stiffness, was confirmed in a landmark study to be a strong determinant of the fate of mesenchymal stem cell differentiation. 9 The mechanosensitive nature of bone cells and the stimulating effects of matrix stiffness on osteoblasts, osteocytes, and chondrocytes have been extensively studied. [10][11][12] Although osteoclast differentiation has been proven to be influenced by multiple mechanical stimuli, such as tension force, 13 microgravity, 14 fluid shear stress, 15 vibration, 16 and compressive forces, 17 the activity of osteoclasts in response to different degrees of microenvironmental stiffness remains unclear.
In this study, we generated five PDMS substrates, each of a different stiffness degree, to mimic the physiological mechanical properties of the extracellular microenvironment to determine how osteoclasts sense and react to such stimuli. We provide evidence proving that extracellular substrate stiffness is a strong determi-

| Fabrication and characterization of polydimethylsiloxane substrates
The rigidity of PDMS can be regulated by changing the mass ratio of the curing agent to the liquid oligomeric base (Sylgard 184, Corning).
The substrates were processed according to a previously described method. 18 Although the rigidity of PDMS substrates is defined by the modulus of elasticity, we use the terms stiffness and Young's modulus (E) interchangeably. Mechanical tensile tests were conducted on all substrates using a universal testing machine (5967, Instron). In the linear elastic stage, E is defined as the ratio of applied stress to resultant strain according to Hooke's law, 19 E = σ/ε × σ, which is the force per area (F/S), where ε indicates the stain defined by the relative elongation (DL/L) resulting from the external force.

| In vitro osteoclastogenesis
All the animal experiments were approved by the Ethics Committee of West China Hospital of Stomatology (WCHSIRB-D-2017-029).
C57BL/6 mice were dissected to acquire femurs and tibias. Then, the bone marrow cells were flushed into a culture dish and cultivated for 24 h in complete α-MEM (HyClone) supplemented with 10% FBS and 1% penicillin-streptomycin (HyClone) at 37°C under 5% CO 2 .
Then, the cells were cultured for 72 h in complete medium containing 30 ng/ml M-CSF (Catalog#416-ML, R&D Systems), whereupon they were regarded as bone marrow-derived macrophages (BMMs).

These macrophages and RAW 264.7 monocytic cells (Shanghai Cell
Center) were subsequently seeded onto the PDMS substrates in dishes and cultured in complete α-MEM supplemented with 30 ng/ ml M-CSF and 100 ng/ml RANKL (Catalog#462-TEC, R&D Systems).
After 7 days of culture, during which the media containing inducing factors were replaced three times, osteoclastogenesis assays were performed to identify TRAP-positive multinucleated cells (nuclei number ≥3) using an acid phosphatase staining kit (387A, Sigma-Aldrich).

| Atomic force microscopy
Atomic force microscopy (AFM) (SPM9700, Shimadzu) was applied for the surface test as previously described. 20

| Scanning electron microscopy
For scanning electron microscopy (SEM) analysis, osteoclasts cultured on PDMS substrates were first fixed in 2.5% glutaraldehyde and then dehydrated with a graded series of ethyl alcohol (30%, 50%, 70%, 80%, 90%, and 100%). Then, the specimens and blank PDMS substrates were coated with gold and examined using a scanning electron microscope (HT770, Hitachi).

| Transcriptome sequencing and bioinformatics analysis
Total RNA was extracted from osteoclasts cultured on stiff (1:5) and soft (1:45) PDMS substrates (with three independent repeats), using Trizol reagent (Catalog#15596026, Invitrogen), and the quality was examined with an RNA Nano 6000 assay kit (Bioanalyzer 2100 System, Agilent Technologies). The Illumina NeoPrep system was applied to purify and fragment the mRNAs, synthesize cDNAs, and amplify the targets. Sequencing was accomplished with the Illumina NovaSeq 6000 platform, and the raw data were mapped and annotated referring to GRCm38/mm10 mouse genome from UCSC website with TopHat 2.1.0. Gene reads were counted using featureCounts (v1.5.0-p3) and normalized to FPKM values. DESeq2 in the R package (1.20.0) was applied to identify differentially ex-  Acridine orange (AO) staining of the bone slices was performed as described previously. 21 In brief, following osteoclast differentiation, the cells were stained with 1 μM AO (Catalog#HY-101879, MedChemExpress) at 37°C for 20 min, rinsed with phosphatebuffered saline, and finally imaged by confocal laser scanning microscopy (CLSM) (FV3000, Olympus).

| Immunofluorescence and confocal laser scanning microscopy
Osteoclasts were fixed with 4% paraformaldehyde for 20 min and  and FITC-labeled phalloidin (Catalog#F432, Invitrogen), the samples were sealed with 50% glycerol. All immunofluorescence images were captured by CLSM (FV3000, Olympus).

| Quantitative reverse transcription PCR
Total RNA was extracted from the osteoclasts using Trizol reagent and then purified with the RNeasyPlus Mini Kit (Qiagen).
The extracted RNA samples were quantified and then reverse transcribed to cDNA using a reverse transcriptase kit (Takara).
The quantitative real-time polymerase chain reaction (qPCR) was then performed with the cDNA, SYBR Green (Takara), and primers targeting the following genes: tumor necrosis factor receptor-

| Western blot assay
Osteoclast lysates were obtained using RIPA lysis buffer (Catalog#R0020, Solarbio) containing PMSF (Catalog#P7626, Sigma-Aldrich). After quantifying the total protein with a BCA protein assay kit, the sample was mixed with loading buffer and DTT

| Protein-protein interaction network analysis
A protein-protein interaction network was built by importing 15 target DEGs into the STRING database (v11.5) for analysis (https://strin g-db.org). The target genes were clustered into two groups using the k-means method, and connections of high confidence (cutoff edge = 0.700) were shown. 22

| Statistical analysis
All data are presented as the mean ±standard deviation and representative of three independent experiments. The Student's t-test was used to evaluate differences between groups, with a p value of less than 0.05 indicating statistical significance.

| Surface topography and elastic stiffness of the polydimethylsiloxane substrates
Five PDMS substrates, each of a different degree of stiffness, were prepared by increasing the ratio of curing agent to elastomer (1:5, 1:15, 1:30, 1:45, and 1:60). The substrate nanotopography, a key factor of biomaterials that influences cell behavior, 23 was investigated by AFM ( Figure 1A). The Ra value, representing surface roughness, was lower for the substrates than for a Petri dish ( Figure 1D). In the SEM images, the substrates had a relatively smooth surface morphology ( Figure 1B), verifying the AFM results. The various substrates were subjected to mechanical tensile tests to measure their Young's modulus (E). Upon tension loading, the substrates exhibited a stress-strain response in sequence of the elastic stage, followed by the plastic stage that ended abruptly at a fracture strain. 24 The linear region of the elastic stage of the stress-strain curve for each substrate is shown in Figure 1C. The slope was calculated from the linear regression line of the scatter points. The tensile elastic modulus of the five substrates decreased from 4.05 MPa to 1.66, 0.45, 0.10, and 0.03 MPa, respectively, in the order of stiffest to softest substrates ( Figure 1E).

| Osteoclasts displayed a distinct morphology and fusion activity on PDMS substrates of different stiffness degrees
Bone marrow precursor cells were isolated, seeded onto the differ- shown in the full video (Movie S1 (stiff) and S2 (soft)).

| Expression profile of osteoclast-specific markers was enhanced on stiffer substrates
We used immunofluorescence staining to analyze osteoclast-specific markers that control the pathways of cell fate during osteoclastogenesis. NFATc1, a master regulator of osteoclastogenesis, was significantly accumulated in the nuclear region ( Figure 3A), as confirmed quantitatively by its total fluorescence intensity ( Figure 3D) and western blotassayed level ( Figure 3E). The expression levels of NF-κB p65, which is important for the initial stimulation of NFATc1 in RANKL-induced osteoclastogenesis, 6 were higher on the stiffer substrates, as determined by western blotting ( Figure 3E, 3F). Expression and distribution of CTSK and DCST1 was also explored ( Figure 3B, 3C). DCST1 was more highly accumulated on the cell border of osteoclasts on the stiffer substrates. Western blotting showed the reduced expression of CTSK and DCST1 on the soft substrates relative to that on the stiff substrates ( Figure 3E, 3F). Additionally, qPCR analysis of the osteoclast-specific marker genes Traf6, Mmp9, Acp5, and Camk2a confirmed that the transcription levels were significantly higher on the stiffer substrates ( Figure 3G). Collectively, these results suggest that stiffer substrates enhance osteoclast differentiation. To further confirm these results, the mechanical difference between untreated stiff bone slices and decalcified relatively softer bone slices was confirmed by tensile testing ( Figure 4D). The slope value was almost identical to the Young's modulus of the bone slices, was the most abundant and also displayed the largest fold change with the highest statistically significant level ( Figure 5D).

| Substrate stiffness regulated fibronectinintegrin αvβ3 signaling and promoted the expression of downstream intercellular activators
Immunofluorescence was applied to detect changes in the distribution of integrin αvβ3 and fibronectin screened in Figure 5C and D.
Integrin αvβ3 showed much brighter intensity on the cell border on the stiffer substrates ( Figure 6A). Concordant with integrin αvβ3, the branch-structured fibronectin displayed a high level of continuous deposition along with the cytoskeleton (F-actin) in the stiffer group, whereas it appeared as fragmented filaments in the softer group ( Figure 6B). The variations in integrin αvβ3 and fibronectin protein levels were confirmed by western blot assay and quantified to be significantly reduced on the softer substrates ( Figure 6C, D).
Therefore, the levels of integrin αvβ3 and fibronectin were verified to be significantly altered in response to substrate stiffness, suggesting they are possible mechanosensors that link microenvironmental clues to the intercellular cytoskeleton.
To confirm the shift in integrin activation, downstream intercellular activators were examined. Paxillin, an important integrin-associated protein that amplifies the signal of integrininduced adhesion, 32 was confirmed by western blotting to be more highly expressed in cells on stiffer substrates ( Figure 6C). FAK, 33 another integrin-associated kinase that reinforces the activation of paxillin, was also more highly expressed on stiff substrates.
Integrin-adaptor protein intercellular activation is followed by the phosphorylation of PKCα for further reorganization of the actin cytoskeleton. 34 Moreover, integrin-mediated PKC activation regulates adhesion and podosome formation in osteoclasts through a RhoA-dependent pathway. 35,36 Rho proteins contribute to the reorganization of the actin cytoskeleton and regulate the cell shape. 37 Similarly, western blotting showed the reduced expression of p-PKC and RhoA in cells on the soft substrate relative to that on the stiff substrate ( Figure 6C, D). Therefore, cytoskeletal organization in response to substrate stiffness in osteoclasts is possibly regulated by fibronectin-integrin αvβ3 signaling pathways.

| Prediction of the network of fibronectinintegrin and cytoskeletal signaling molecules and osteoclast differentiation markers altered by substrate stiffness
To investigate the mechanism by which substrate stiffness alters integrin signaling pathways and osteoclast differentiation, we predicted out in this study (Figures 5 and 6). Osteoclast differentiation markers included those that were downregulated (verified in Figure 3). The interacting proteins were clustered into two groups (regulation of actin cytoskeleton and osteoclast differentiation), with a connection confidence higher than 0.7. 22 Critical nodes connecting osteoclast differentiation to regulation of the actin cytoskeleton were noted.
Itgb3 and Ptk2 were linked to the majority of actin cytoskeleton elements and were directly connected to Mmp9, which contributes to osteoclastic bone resorption. Prkca was indirectly linked to Nfatc1 via Rela and Traf6. Therefore, substrate stiffness-regulated integrincytoskeleton signaling was predicted to be closely related to osteoclastogenesis markers whose levels were altered under the different stiffness conditions.

| DISCUSS ION
The microenvironment of cells provides mechanical signals that ultimately translate into biochemical pathways for cell differentiation and functions. 9 One important mechanical signal of bone cell surroundings is the bone matrix stiffness. The stiffness of bone tissue can be altered by aging and common diseases that influence the architecture of mineral components (as occurs in osteoporosis, osteogenesis imperfecta, osteoarthritis, and Paget's disease), which changes the microenvironment of the cells. 8,38 In this study, we established a mechanical model to investigate how osteoclasts sense and react to matrix stiffness. PDMS substrates were selected for their good biocompatibility and wide range of mechanical properties that mimic physiological conditions highly. 39 Figure 3. The connection edges shown are of high confidence (0.700), with the thicker connection line indicating the highest confidence (0.900). The network proteins were clustered into two groups using the k-means method osteoclastogenesis profiles confirm that osteoclast differentiation reacts strongly to stiffness signals.
Notably, the architecture of cytoskeletal actin filaments was significantly altered by substrate stiffness stimulation. The rearrangement of cytoskeleton-associated adhesion molecules upon the mechanosensing of ECM stiffness has been observed in mesenchymal stem cells, 9 apical papilla-derived stem cells, 52 adipose-derived stromal cells, 53 and bone cells such as osteoblasts, 10 osteocytes, 11

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
Qingxuan Wang, Wenli Lai, and Chenchen Zhou designed the experiments. Qingxuan Wang and Jing Xie performed the experiments.
Qingxuan Wang and Chenchen Zhou analyzed and confirmed all data and edited the manuscript. All authors reviewed and approved the final paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the results of this study are available upon request from the corresponding author.