Mice, rats, and clinical samples

LoxP-flanked PDGFR-β and Cdh5-Cre were obtained from Cyagen (Stock No. 017986, China). First, mice carrying loxP-flanked PDGFR-β alleles (PDGFR-βlox/lox) and Cdh5-Cre transgenic mice were crossed to obtain Cdh5-Cre::PDGFR-βlox/− mice, which were then mated with PDGFR-βlox/lox mice to obtain Cdh5-Cre:PDGFR-βlox/lox mice (PDGFR-β−/−). For the endothelium-specific PDGFR-β knockout experiment, 3-month-old male PDGFR-βlox/lox mice and PDGFR-β−/−mice were used to perform anterior cruciate ligament transection surgery (ACLT) to observe post-traumatic OA in each independent experiment (n = 8 per group). Two-month-old and 15-month-old male PDGFR-βlox/lox mice and PDGFR-β−/−mice were used to observe age-related OA for each independent experiment (n = 8 per group). PCR analyses of genomic DNA were used to determine the genotypes of the mice using the following primers:

Table 3 The primer sequences used to determine the genotypes of the mice

All 1-month-old (n = 10), 3-month-old (n = 10), 6-month-old (n = 10), and 15-month-old (n = 10) male C57BL/6J (WT) mice and 3-month-old male SD (Sprague Dawley) rats (n = 5 in each group) were purchased from the animal center of Southern Medical University, Guangzhou, China.

ACLT was introduced to establish a post-traumatic OA model as previously described.6,18,62 In the OA group, ACLT surgery was carried out on the left knee joint under the supervision of a surgical loupe. For rodents in the sham group, the joint capsule was opened and then sutured in the left knee. After surgery, the rodents were randomized to plastic cages based on body weight and the study plan, allowing them to move around freely in cages. The mice were euthanized at 4 or 8 weeks after surgery, and the rats were sacrificed at 8 weeks post-operation. The experimental protocols were approved by the Institutional Animal Care and Use Committee of Southern Medical University.

Human specimens were collected from OA patients who underwent total knee arthroplasty (TKA). The patients were aged 65–75 years with symptomatic radiographic tibial knee OA and no other systemic diseases (n = 5, 3 female and 2 male). The medial compartment of specimens with severe cartilage injury was used for the OA group, while the lateral compartment of specimens with intact cartilage coverage was used for the RN group. CT scanning and histological examination were performed on the specimens. The personal information was anonymized. This study was approved by the ethical medical committee of Nanfang Hospital.

Local injection of AAV in subchondral bone

Recombinant AAV for silencing endothelial PDGFR-β was locally injected into subchondral bone in vivo. The pdgfrb shRNA virus (AAV2/9-TIE1p-EGFP-miR30shRNA(pdgfrb)) and the controls (AAV2/9-TIE1p-EGFP) were constructed by Obio Technology (China). The target sequence of pdgfrb shRNA was CAGGTGGTGTTTGAGGCTTAT, and the control was AAV2/9-TIE1p-EGFP. The pdgfrb shRNA sequence was inserted into the intron of the TIE1p promoter to drive the expression of EGFP, which was utilized to detect AAV-induced pdgfrb shRNA expression. ACLT surgery was carried out on the left knee of 3-month SD rats to induce OA. Ten days later, 5 × 109 AAV particles in a 30 µl volume were injected into the subchondral bone of the left ACLT-treated knee in the SD rats.

Histochemical, immunofluorescence, and histomorphometric analyses

The rodents were anesthetized, and then, the knees were harvested after heart perfusion, fixed for 24 h in 4% paraformaldehyde, and decalcified for 3 weeks in 10% EDTA (pH 7.4). The samples were fixed in OCT (optimal cutting temperature) compound (Sakura Finetek) or paraffin. Then, the medial compartment of the samples was longitudinally oriented and cut to 4 μm. Hematoxylin and eosin (H&E) staining was performed to calculate the ratio of the thickness of calcified cartilage (CC) and hyaline cartilage (HC) in cartilage. The tidemark line shows the separation of CC and CC by H&E staining. Safranin O/Fast Green (SOFG) staining was performed to examine proteoglycans in cartilage. The tissue samples were mounted after being dehydrated (ethanol). Images were acquired by microscopy (Olympus). OARSI (Osteoarthritis Research Society International-modified Mankin criteria) scores were measured as previously described. Immunostaining staining was performed using the following primary antibodies: PDGFR-β (Santa Cruz, 1:50, sc-374573), endomucin (Santa Cruz, 1:50, sc-65495), CD31 (Abcam, 1:200, ab222783), nestin (Santa Cruz, 1:50, sc-58813), lepR (Santa Cruz, 1:50, sc-8391), MMP13 (Santa Cruz, 1:50, sc-30073), ATAMDTS (Santa Cruz, 1:200, #C04789), SOX9 (Santa Cruz, 1:50, sc-166505), Aggrecan (Santa Cruz, 1:50, sc-166951), and COL II (Santa Cruz, 1:50, sc-52658) at 4 °C overnight. Secondary antibodies conjugated with fluorescence tags were incubated at room temperature (RT) for 1 h in the dark. Nuclei were labeled with DAPI (DAPI; Servicebio) before imaging. The entire subchondral bone area of the slices was then microphotographed for histomorphometric measurements. Olympus confocal microscopy was employed to calculate positive subchondral cells. The number of entire subchondral positive cells was counted for each specimen, and five sequential samples were measured per rodent in each group. Quantitative analysis was performed in a blinded manner using ImageJ software.

Microcomputed tomography analysis

After soft tissue dissection, the harvested knee joints were fixed in 70% ethanol overnight. We then scanned and reconstructed the samples with high-resolution microcomputed tomography (CT) (SkyScan 1172) and CT reconstruction software (NRecon v1.6). Then, CTAn v1.9 and μCTVol v2.0 were used for three-dimensional model visualization and further data analysis. The parameters of the scanner were as follows: 50 kVp voltage, 200 μA current, and 9 μm per pixel resolution. We selected the region of interest as the entire subchondral bone of the specimens. Tb.Pf, BV/TV, Tb.N, and SBP.Th were measured.

Gait analysis

We used a highly sensitive, automated computer-assisted method to determine the impact of endothelium-specific PDGFR-β knockout on gait coordination according to a previously described protocol63 (Rodent Gait Behavior Analyzer (GAT-RGBA); Shenzhen Giant (Ju’An) Technologies Co., Ltd.). The CatWalk system is a complex device that quantitatively measures footfall and motor performance. First, we trained animals to willingly go across the illuminated glass platform. Next, a high-speed high-resolution camera captured the animal paw location when the animals walked on the illuminated glass walkway. For a representative number of completed runs per animal, up to 20 runs were recorded for each animal. Next, gait analysis was performed. Paw area, cadence, stride duration, and stride length were all calculated as the mean gait characteristics. When passing through the glasswalk, the mouse’s paw area refers to the area of the mouse’s foot that touches the ground. Cadence is a unit of measurement for stride frequency, measured in steps per second. Stride duration (s) refers to the duration of one paw to complete a stride. The distance (cm) between successive placements of the same paw is known as the stride length. Behavioral characterization was approved by the ethical medical committee of Nanfang Hospital.

Von Frey test

Mechanical allodynia was assessed using a digital electronic von Frey anesthesiometer, as previously described (IITC Life Science, CA, USA).64,65 This exam was performed five times with a 15 min inter-test delay between each repetition. Each mouse was hung above a wire grid in a clear plastic container. Before the trial, the animals were exposed to the testing environment for at least 15 min. The polypropylene non-Frey filament was placed perpendicular to the midplantar surface of a selected left hind paw. At the threshold, the mouse responded by flicking its paw away from the stimulus. The intensity of stimulation was determined when the mouse lifted its foot or added a paw.

Flow cytometry

Following the dissection of soft tissue, the knee joints and blood were harvested. For the analysis of subchondral endothelial cells and MSCs, 10 subchondral bone specimens from 5 knee joints (including tibial and femoral subchondral bone in each knee joint) were used for cell isolation and analysis in each sample. We first removed the outer surface of the knee joint by immersing the specimens in protease solution (2.5 mg·mL−1 trypsin and 2 mg·mL−1 collagenase A) for 20 min. Then, the specimens were digested for 60 min to obtain the desired cells. Following the lysis of red blood cells (BD FACS™; BD Biosciences, San Jose, CA), we harvested cells within the supernatant, which were then used to detect the number of total endothelial cells (ECs), CD31hiEmcnhi cells, and CD31loEmcnlo cells or to detect the alteration of MSCs. FACS was conducted with antibodies against CD31-APC (RD, FAB3628A-025), CD45-Per-CP (BioLegend, 103133), Ter119-Brilliant Violet 421 (BioLegend, 116233), Endomucin-FITC (Santa Cruz, sc-65495), Sca-1-PE (BioLegend, 108107), and CD24-Brilliant Violet 421 (BioLegend, 101825). CD31hiEmcnhi cells were plotted and sorted by first setting standard quadrant gates. Then, as we previously described,36 in quadrant 2, gates were set arbitrarily at >103 log Fl-2 (endomucin-FITC) and >103 log Fl-4 (CD31-APC) fluorescence to discriminate CD31hiEmcnhi cells from the total double-positive cells. After negative selection of the leukocyte common antigens CD119 and Ter45 at <102 log Fl-3 (Ter119-Brilliant Violet 421) and <102 log Fl-1 (CD45-Brilliant Violet 421), CD31+CD45−Ter119−, referred to as BMECs, were sorted using side scatter and CD31-APC fluorescence at >102 log Fl-4 (CD31-APC). Cells were then resuspended in staining buffer and counted on an LSR II flow cytometer (BD Biosciences) before flow cytometry. CellQuest software was used to collect data on a FACSCalibur flow cytometer (Becton Dickinson). We used FlowJo software to analyze the data and create all flow cytometry contour plots (with outliers) (TreeStar).

Quantitative real-time polymerase chain reaction analysis

The subchondral bones dissected from the proximal tibias, chilled in liquid nitrogen, were cut into pieces. We used TRIzol reagent (Invitrogen, Carlsbad, CA) for the homogenization of the specimens, and then, the total RNA was harvested. One milligram of total RNA was employed to synthesize cDNA using a cDNA Synthesis kit (Fermentas, Burlington, Canada). After that, FastStart Universal SYBR Premix ExTaqTM II (TaKaRa Biotechnology, Japan) was employed to conduct qRT‒PCR. The 2−△△CT approach was employed to measure relative gene expression.

Cell line and cell transfection

BMECs were cultured in DMEM (Gibco, USA) with 100 mg·mL−1 streptomycin sulfate (Life Technologies, USA), 15% fetal bovine serum (FBS; Gibco, USA), and 100 U·mL−1 penicillin. GENE provided the lentivirus used to construct PDGFR-β knockdown or overexpression constructs (Shanghai, China). PDGFR-β overexpression lentivirus (termed oePDGFR-β), a negative control (termed NC), PDGFR knockdown lentivirus (termed shPDGFR-β-7, shPDGFR-β-8, shPDGFR-β-9), or a scramble control (termed shNC) was used to infect BMECs plated in 6-well dishes at 50% confluence. Pools of stable transductions were formed using puromycin (7 μg·mL−1) and mycoplasma elimination reagent (0.1%, Yeasen, Shanghai) for 2 weeks. We carried out transfections employing the GENE transfection kit (GENE, China).

LC–MS/MS analysis

BMECs were treated with PBS and PDGF-BB for 24 h. Then, we collected total cell lysates and incubated them with an antibody against PDGFR-β for immunoprecipitation. PDGFR-β-binding immunoprecipitates were subjected to LC–MS/MS analysis (PTM Bio, China). We converted the raw data (.wiff) into peak lists (.mgf) employing Protein Pilot software v4.0 (Applied Biosystems). Protein Pilot was employed to measure P values, the average relative expression, upper confidence interval, lower confidence interval, and error factors. Differential expression was defined as a fold change ≥ 1.5 or ≤0.5 and P values ≤ 0.05.

Immunoprecipitation

Cells were washed two times with ice-cold PBS after treatment and lysed with ice-cold lysis buffer (0.3% CHAPS, 2 mmol·L−1 EDTA, 10 mmol·L−1 pyrophosphate, 10 mmol·L−1 glycerophosphate, 40 mmol·L−1 HEPES [pH 7.4], one tablet of EDTA-free protease inhibitors (Roche, USA) per 25 mL for 15 min. Then, we acquired the supernatant by centrifugation at 12 000 × g for 10 min. The supernatant was then incubated with primary antibody overnight at 4 °C. After that, a 50% slurry of Protein A + G Sepharose was added and incubated at 4 °C for another 2 h. The immunoprecipitates were washed 3 times with ice-cold PBS. Then, 50 µL of 1x SDS sample loading buffer was added, boiled for 10 min, and analyzed by immunoblotting. The following antibodies were used: PDGFR-β (Santa Cruz, 1:50, sc-374573), TLN1 (Abcam, 1:1 000, ab108480), FAK (CST, 1:1 000, 3285), goat anti-mouse IgG HRP (Invitrogen, 1:10 000, 31430), and goat anti-rabbit IgG HRP (Invitrogen, 1:10 000, 31460).

Western blot assay

PDGFR-β proteins were collected from subchondral bone tissue. The bone tissue was ground in liquid nitrogen until it became a powder. The cell proteins were collected from BMECs. Cell lysis buffer was added to extract the subchondral bone or BMEC protein. The PierceTM BCA Protein Assay Kit was employed to determine the protein levels (Thermo Fisher Scientific, Waltham, MA). Next, 1x SDS sample loading buffer was added to the cell lysates and boiled for 10 min. After protein extraction, the supernatants were separated by sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (Yamei, PG110-114) and blotted onto polyvinylidene fluoride membranes. After the membranes were blocked with skim milk, the antibodies were added to the primary antibody dilution and incubated with the membranes overnight at 4 °C. We washed the membranes with TBST three times (10 min each time). Then, goat anti-mouse IgG (1:10 000) or goat anti-rabbit IgG (1:10 000) was added and incubated at room temperature for 60 min. After that, we washed the membranes three times again (10 min each time). Enhanced chemiluminescence (ECL Kit; Amersham Biosciences) was used for protein visualization. The following Western blotting antibodies and inhibitors were used: PDGFR-β (Santa Cruz, 1:1 000, sc-374573), phospho-FAK (Tyr925) (CST, 1:1 000, 3284), phospho-FAK (Tyr397) (CST, 1:1 000, 3283), FAK (CST, 1:1 000, 3285), TLN1 (CST, 1:1 000, 4021), phospho-TLN1 (Ser425) (CST, 1:1 000, 5426), VEGF (Proteintech, 1:1 000, 19003-1-AP), goat anti-mouse IgG HRP (Invitrogen, 1:10 000, 31430), goat anti-rabbit IgG HRP (Invitrogen, 1:10 000, 31460) and PF-573228 (MCE, 50 nmol·L−1, HY-10461).

Enzyme-linked immunosorbent assay (ELISA)

We cultured cells in DMEM with 15% FBS for 24 h and collected the cell culture medium by centrifugation at 200 × g for 15 min. All samples were stored at −80 °C before analyses. Next, the supernatant from various treatment conditions was collected to quantify VEGF levels employing a commercial ELISA kit (ab100786; Abcam) (five per group). The absorbance of the ELISAs listed above was measured by employing a microplate reader at 450 nm (Bio-Rad 680, Hercules, USA). The wavelength was corrected to 570 nm. The protein concentration in each sample was calculated using the standard curve.

Tube formation assay

BMECs at a density of 2 × 104 cells per well were seeded in 96-well plates that were already coated with 50 μL of Matrigel in each seed well (BD Biosciences). The plates were then incubated at 37 °C under different treatment conditions. After culture for 3 h, the cells were imaged and analyzed using an inverted microscope (Leica) and Image-Pro Plus 6 software. Total loops, total branching points, and total tube length were measured.

Cell proliferation assay

A Cell Counting Kit-8 was employed to detect BMEC proliferation (CCK-8; Dojindo, Kumamoto, Japan). In brief, BMECs at a density of 3 000 cells per well were seeded in a 96-well plate and incubated in DMEM under different treatments for 24 h at 37 °C. The wells treated with complete culture medium (without cells) were set as blanks. We added 10 μL of CCK-8 solution to each well on Days 1, 3, and 5. In addition, the wells were incubated at 37 °C for 2 h. The optical density (OD) was detected at 450 nm based on a microplate reader (Bio-Rad 680). The cell proliferation rate was evaluated by drawing a growth curve.

Migration assay

The wound scratch assay was used to analyze the migration of BMECs under different conditions. In brief, 5.0 × 105 cells per well were plated in a 35 mm culture plate and incubated for 24 h at 37 °C until confluence. Next, a scratch was made with a sterile p200 pipette tip. Cell fragments were washed 3 times with PBS and then incubated in DMEM with 5% fetal bovine serum (FBS; Gibco, USA) under different treatments. Images of the wounds were taken immediately, 12 h, and 24 h after scratching. We used ImageJ software to detect the change in the width of the scratched areas. The rate of migration was measured as follows: migration area (%) = (A0 – An)/A0 × 100, where An and A0 represent the residual area and initial area of the wound, respectively.

Statistical analysis

Data are presented as the mean ± s.d. For two-group comparisons, data were analyzed by a two-tailed Student’s t-test. For multiple group comparisons, one-way analysis of variance (ANOVA) was used. We first examined homogeneity of variance and then evaluated the differences between groups using post hoc multiple comparisons. Specifically, we used Dunnett’s T3 to evaluate the group differences if heterogeneity existed. However, the Bonferroni test was adopted if there was no heterogeneity. Significant differences were defined at P < 0.05. SPSS 22.0 analysis software (SPSS, Inc.) was employed for all data analyses.