CD147 induces angiogenesis through a vascular endothelial growth factor and hypoxia-inducible transcription factor 1α–mediated pathway in rheumatoid arthritis




Rheumatoid arthritis (RA) is an inflammatory and angiogenic disease. However, the molecular mechanisms that promote angiogenesis in RA have not been clearly identified. Our objective was to study the role of CD147 in angiogenesis and determine whether the strategy in which CD147 is suppressed might be useful in reducing angiogenesis in RA.


Correlations among expression levels of CD147, vascular endothelial growth factor (VEGF), and hypoxia-inducible factor 1α (HIF-1α) were determined by immunohistochemistry staining. RA fibroblast-like synoviocytes (FLS) cells were cultured under various conditions, and the production of VEGF and HIF-1α was examined by real-time polymerase chain reaction and enzyme-linked immunosorbent assay. The SCID mouse coimplantation model of RA (SCID-HuRAg) was established, mice were treated with CD147 monoclonal antibody, infliximab, or both CD147 and infliximab, and the volume of the grafts and the average vascular density were measured and analyzed. Western blot analyses were performed to examine the potential signaling pathways.


The expression levels of CD147 showed significantly positive correlations with VEGF and HIF-1α levels, as well as with vascular density, in RA synovium. After small interfering RNA transfection or after addition of specific antibodies for CD147, the production of VEGF and HIF-1α were significantly reduced. The expression of VEGF and HIF-1α decreased more after CD147 inhibition than after infliximab treatment in the engrafted tissues in SCID-HuRAg mice. The phosphatidylinositol 3-kinase/Akt pathway may be involved in this process.


CD147 induces up-regulation of VEGF and HIF-1α in RA FLS, further promotes angiogenesis, and leads to the persistence of synovitis. Inhibition of CD147 may be a promising target for novel therapeutic strategies.

Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by proliferation of synovial cells, leading to the formation of aggressive tissue called rheumatoid pannus. Angiogenesis is required to maintain the chronic inflammatory state by transporting inflammatory cells to the site of synovitis and supplying nutrients to the pannus (1). Expansion of pannus induces cartilage invasion and bone erosion, and the angiogenesis in the inflamed joints represents the net balance between angiogenic and antiangiogenic factors. A number of angiogenic factors are involved in angiogenesis, and vascular endothelial growth factor (VEGF) and hypoxia-inducible factor 1α (HIF-1α) are critical mediators among the implicated positive regulators of angiogenesis (2, 3).

CD147, also named extracellular matrix metalloproteinase inducer (EMMPRIN), is highly expressed in RA synovial tissue and triggers human synoviocytes to produce matrix metalloproteinases (MMPs), which suggests that CD147 is an important mediator in the pathogenesis of RA. However, the role of CD147 in the expression of VEGF and HIF-1α, which are important regulators of pannus formation, has not been fully identified. In this study, we investigated the effect of CD147 on the production of VEGF and HIF-1α in cultured rheumatoid fibroblast-like synoviocytes (FLS), the most important cell type in rheumatoid synovium. We also conducted a treatment study using SCID mice, into which RA synovium and normal cartilage were coimplanted (SCID-HuRAg), a well-established in vivo model of RA, to evaluate the efficacy of anti-human CD147 monoclonal antibody (mAb) in the angiogenesis process. The functional importance of CD147 during the pathogenesis of RA was thought to be mainly due to its stimulatory effects on fibroblast-derived MMP expression. The results presented here highlight the proangiogenic role of CD147 in RA.


Patients and preparation of synovial tissue.

We enrolled 58 patients with RA who satisfied the American College of Rheumatology 1987 diagnostic criteria (4). Patients were those with persistent active synovitis of the knee (characterized by pain, tenderness, and effusion) who attended the rheumatology clinic at Xijing Hospital (Xi'an, China). No patient included in the study was being treated with corticosteroids or second-line drug agents (methotrexate, sulfasalazine, or cyclosporin A). Synovial tissue samples were obtained from these patients during knee joint arthroscopy.

Clinicopathologic measurements were performed, including disease duration, composite 28-joint count Disease Activity Score (DAS28) (5), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), and general health or patient's global assessment of disease activity (using a 100-mm visual analog scale).

For comparative analysis, samples of synovium were also obtained from 34 patients with osteoarthritis (OA). These patients met the clinical and radiographic criteria of the American College of Rheumatology (6).

The study was approved by the Ethics Committee of the Fourth Military Medical University. All patients and controls gave their informed consent to participate in the study.

Immunohistologic study of CD147, VEGF, HIF-1α, and CD31.

Immunohistochemical staining of the synovium samples from the 58 RA patients and 34 OA controls was performed using a streptavidin–peroxidase kit (Zymed). The monoclonal antibodies used were anti-CD147 mAb (BioLegend), anti-VEGF mAb (Upstate Biotechnology), anti–HIF-1α mAb (Santa Cruz Biotechnology), and anti-CD31 mAb (Chemicon). Sections were reacted in turn with biotin-labeled goat anti-mouse IgG, horseradish peroxidase–labeled streptavidin, and diaminobenzidine (Zymed) before they were restained with hematoxylin for visualization of nuclei. For negative controls, irrelevant IgG was used instead of the primary antibodies.

The sections were viewed with an Olympus AX70 microscope and photographed. In positive sections, the cell membrane and/or cytoplasm was a clear brown-yellowish color. Expression of CD147, VEGF, and HIF-1α was evaluated by an investigator (HY) without any clinical information related to the sample. Four fields were selected at random, and the staining intensities, as well as the total cell numbers, were determined. The average expression levels were calculated as a percentage of the entire field of view and processed for further statistical analysis. Vascular density was evaluated by an experienced pathologist (Professor P. S. Yan) who had no clinical information related to the sample and was determined by morphometric analysis of tissue sections immunostained with anti-CD31 antibody. Four fields were selected at random, and vessels with a distinct lumen were counted to determine the number of vessels per square millimeter. The average vascular density (number of vessels/mm2) from the 4 fields for each patient was used for further statistical analysis.

Confocal laser scanning microscopy of synovium.

After fixation, frozen sections of synovial tissue were incubated first with rabbit anti-human CD147 polyclonal antibody (Zymed) or rabbit anti-human VEGF polyclonal antibody (Abcam) plus either mouse anti-human CD90 mAb (BD PharMingen) or mouse anti-human CD31 mAb (Dako) at a dilution of 1:200. Sections were then incubated with tetramethylrhodamine isothiocyanate–labeled goat anti-rabbit IgG (Sigma) or fluorescein isothiocyanate (FITC)–labeled goat anti-mouse IgG (Sigma) at a dilution of 1:500. Cell nuclei were dyed with DAPI (Invitrogen). The sections were then washed, mounted, analyzed, and photographed with an Olympus FV300 confocal laser scanning microscope. Five hundred cells were counted in every section, and distinct spots of red, green, or yellow fluorescence were observed in the membrane or cytoplasm of positive cells.

In vitro culture of FLS from synovial tissues.

FLS were isolated by enzymatic digestion from synovial tissues obtained during synovectomy, as previously described (7). The cells used for the experiments were at the third to fifth passage, because these cells were more purified than the first and second passages and had better biologic function. Cultures were washed with serum-free Dulbecco's modified Eagle's medium and then incubated for 24 hours at 37°C in 24-well plates at a density of 2 × 105 cells/well.

For experiments under hypoxic conditions, cells were placed in a sealed humidified chamber maintained at 0.5% O2, 5% CO2, and 94.5% N2, which was created using a hypoxic chamber (Astec). Anti-CD147 antibody (mouse anti-human CD147 mAb, subclass IgG1, 80 μg/ml) was added 48 hours before detection of the blockage effects (8). The irrelevant anti–Japanese encephalitis virus (anti-JEV) mAb (80 μg/ml; provided by the Department of Microbiology) was used as a negative control antibody. Anti-TNFα mAb (infliximab, 80 μg/ml) was kindly provided by Xian-Janssen Pharmaceutical. The inhibitors LY294002, PD98059, SP600125, SB203580, and U0126 (Cell Signaling Technology) were added to each well. All cultures were run in triplicate or quadruplicate.

Flow cytometry for CD14, CD68, CD90, and CD147.

The isolated RA FLS were identified by flow cytometric analysis as a homogeneous population with a phenotype of <1% CD14, <1% CD68, and >98% CD90 (9). Expression of CD147 on the surface of the FLS was determined by flow cytometry. Cells were washed 3 times with phosphate buffered saline (PBS) and then were treated with FITC-conjugated anti-CD147 antibody (BD PharMingen) or FITC-conjugated mouse IgG1 as control (R&D Systems) for 20 minutes in the dark. Cells were washed with PBS and then analyzed with a FACSCalibur flow cytometer (Becton Dickinson). Data were processed using CellQuest software.

Gene silencing of CD147 by RNA interference.

The sequences for the CD147 small interfering RNA (siRNA) were 5′-GUU-CUU-CGU-GAG-UUC-CUC-TT-3′ and 3′-DTDTCA-AGA-AGC-ACU-CAA-GGA-G-5′ (Ambion). According to the Lipofectamine 2000 (Invitrogen) manual, we transfected 50 pmoles of siCD147 with 2 μl of Lipofectamine 2000 at a density of 3 × 105 cells/well, and 5 hours after transfection, growth culture was added to the wells. Silencer negative control siRNA (Ambion) was run under similar conditions. Silencing effects of CD147 were examined by reverse transcription–polymerase chain reaction (RT-PCR) and Western blot analyses.

Real-time quantitative PCR (qPCR) assay.

Forty-eight hours after siRNA transfection, total RNA was isolated using TRIzol (Invitrogen) according to the manufacturer's instructions. The concentration and purity of RNA were determined by absorbance at 260/280 nm, and complementary DNA was synthesized using a PrimeScript RT Reagent kit (Takara). The primers for CD147 were 5′-GACTGGGCCTGGTACAAGATCAC-3′ and 5′-GCCTCCATGTTCAGGTTCTCAA-3′. The primers for VEGF were 5′-GCATGACGGACAAGTACAGGCT-3′ and 5′-AAAGTACCAGTTTGCCACGGC-3′. The primers for HIF-1α were 5′-GCTTGCTCATCAGTTGCCAC-3′ and 5′-CATAACAAAACCATCCAAGGC-3′. To control for variation in the messenger RNA (mRNA) concentration, all results were normalized to the housekeeping gene GAPDH. The primers for GAPDH were 5′-TGGCTACACTGTCCGAAAT-3′ and 5′-CATCCATGCTGCTCTAAAT-3′. Real-time qPCR was performed with the Mx3000P and Mx3005P qPCR Systems (Agilent Technologies) and SYBR Green Ex Taq (Takara). PCR amplification specificity was verified by examining the melting curve for nonspecific peaks. MxPro qPCR System software (Agilent) was used to analyze the relative quantification.

Enzyme-linked immunosorbent assay (ELISA).

Samples of serum-free conditioned medium were collected and centrifuged at 10,000g for 5 minutes to remove particulates. VEGF release from FLS into the culture medium was directly measured with an ELISA kit (R&D Systems). Total HIF-1α protein was extracted from whole cells. FLS were washed twice with ice-cold PBS and lysed with lysis buffer (pH 7.4) containing 50 mM Tris, 3 mM EDTA, 1 mM MgCl2, 20 mM β-glycerophosphate, 25 mM NaF, 300 mM NaCl, 10% (weight/volume) glycerol, 1% Triton X-100, and a protease inhibitor cocktail (Roche). Whole cell extract was obtained by centrifuging the lysates at 16,000g for 10 minutes at 4°C, followed by sonication. Quantities of HIF-1α were then determined using Quantikine ELISA kits (R&D Systems) according to the manufacturer's protocol. Optical density was determined with a Microplate reader (Bio-Rad). A standard curve for each measurement was established using known standard concentrations.

Western blot analysis.

For detection of nuclear HIF-1α expression, nuclear extracts were prepared with a NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Pierce). Protein concentrations were determined by bicinchoninic acid protein assay. Proteins were transferred to PVDF membranes, and blots were probed for 12 hours at 4°C with primary antibodies diluted in Tris buffered saline–Tween containing 5% nonfat milk (1:1,000 dilution for Akt and 1:500 dilution for HIF-1α, phospho-Akt, and phospho-p38 MAPK). The membranes were incubated with the appropriate secondary antibodies (1:2,000 dilution) for 1 hour at room temperature. Immunoreactive bands were visualized by enhanced chemiluminescence reaction. Each blot illustrated is representative of the findings from at least 3 independent experiments.

Preparation of the SCID-HuRAg mouse model.

SCID-HuRAg mice were prepared as described previously (10). Briefly, 6–8-week-old male NOD/SCID mice (SLAC) that had been bred under specific pathogen–free conditions were used for the experiments. A 1-cm incision was made in the left flank of each mouse. Normal human cartilage and rheumatoid synovial tissue were placed in the chamber in the muscle using fine forceps. The entire procedure was performed under sterile conditions. Successful implantation of human RA tissue was observed by visual assessment 4 weeks after implantation. The animal studies were approved by the local regulatory agency according to the guidelines established by the Animal Ethics Committee of the Fourth Military Medical University.

Protocol for CD147 mAb administration.

Four weeks after implantation, 30 SCID-HuRAg mice were randomly divided into 5 groups and used for the treatment part of the study. CD147 mAb (10 mg/kg; n = 6), infliximab (10 mg/kg; n = 6), a combination of CD147 mAb and infliximab (10 mg/kg; n = 6), or anti-human IgG1 mAb (10 mg/kg; n = 6) in 50 ml of 0.9% sodium chloride was administered twice a week into the implanted tissue using a microsyringe. The other 6 mice were not given any treatment. The injections were repeated 8 times over 4 weeks. Seven days after the final injection, the mice were anesthetized, euthanized, and the implanted tissue was removed.

Immunohistochemistry of paraffin-embedded SCID mouse graft sections.

RA synovium and cartilage removed from the SCID mice were fixed in 4% paraformaldehyde and decalcified with EDTA. After paraffin embedding, tissue sections (2–3 μm) were stained with hematoxylin and eosin for morphologic evaluation. VEGF expression, HIF-1α expression, and vascular density were evaluated as described above. The implanted tissues were measured with dial calipers, and the volumes were determined using the formula (length × width2)/2, as we have described previously (11). Histologic assessments were made under double-blind conditions. Two of us (C-hW and HY) and pathologists recorded the data on separate case record forms without exchanging any information until after the conclusion of the study. Finally, the data reported by all of the researchers were compiled for analysis of the results.

Statistical analysis.

Correlations between the expression levels of CD147 and VEGF expression, HIF-1α expression, or vascular density were determined by semiquantitative analysis using Spearman's nonparametric test. P values for differences between groups were calculated using Student's t-test or Mann-Whitney U test as appropriate. Wilcoxon's rank sum test was used to assess changes in the production of VEGF and HIF-1α in vivo and, the results were compared with those in the control mAb group. Changes in volumes and vascular density of the grafted tissue before and after treatment were compared with Wilcoxon's signed rank test for paired data. GraphPad Prism software (obtained from Cricket Software) was used for the above analyses. P values less than 0.05 were considered significant.


Expression of CD147, VEGF, HIF-1α, and CD31 on synovium from RA and OA patients.

As shown in Figure 1, the immunoreactivity of CD147, VEGF, and HIF-1α was more intense and more widespread in RA synovium than in OA synovium. CD147 was expressed predominantly on FLS and endothelial cells in the RA synovium. RA FLS surrounding microvessels and vascular smooth muscle cells showed high levels of VEGF expression. HIF-1α expression was observed in the lining layer and the sublining layer of RA synovium. Immunohistochemical analysis of RA tissue using antibodies to CD31 expressed in vascular endothelial cells demonstrated a large number of vascular endothelial cells, which were similar in morphology to those identified by staining with hematoxylin and eosin. Histologic study of synovial tissues from patients with active RA revealed an elevated degree of vascular density, along with mononuclear cell infiltration and proliferation of inflamed rheumatoid synovium into the joint cavity.

Figure 1.

Histologic analysis of CD147, vascular endothelial growth factor (VEGF), hypoxia-inducible factor 1α (HIF-1α), and CD31 in rheumatoid arthritis (RA) and osteoarthritis (OA) synovium. A, The expression of CD147, VEGF, and HIF-1α, and the average vascular density as determined with CD31, were more intense and more widespread in RA synovium than in OA synovium. IgG was used as an isotype control. HE = hematoxylin and eosin. Original magnification × 100. B, In RA synovium, CD147 (left) was expressed predominantly on fibroblast-like synoviocytes (FLS), as indicated by CD90 staining (top), and was partly expressed on endothelial cells, as indicated by CD31 staining (bottom). Similarly, VEGF (right) was expressed predominantly on FLS surrounding microvessels and vascular smooth muscle cells.

As shown in Table 1, the expression of CD147 was associated with significantly longer disease duration. Disease activity at the time of biopsy, which was evaluated with the DAS28, was also significantly and positively correlated with the expression of CD147. Expression of VEGF and HIF-1α and the vascular density showed significant relationships to the levels of expression of CD147.

Table 1. Correlations between clinicopathologic features and CD147 expression*
 RA patients (n = 58)rP
  • *

    The average expression levels were calculated as a percentage of the entire field of view. Values are the mean ± SD baseline data, which were recorded at the time of biopsy. RA = rheumatoid arthritis; NS = not significant; DAS28 = Disease Activity Score in 28 joints; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; VAS = visual analog scale (0–100 mm); VEGF = vascular endothelial growth factor; HIF-1α = hypoxia-inducible factor 1α.

Age, years56.4 ± 6.84NS
RA duration, years10.4 ± 5.380.8350.0043
DAS285.2 ± 1.420.8380.0029
ESR, mm/hour63.4 ± 1.650.7910.0003
CRP, mg/dl3.5 ± 0.690.6420.0014
Patient's global assessment (by VAS), mm66.2 ± 3.640.8580.0034
VEGF, %85.3 ± 4.840.5950.0027
HIF-1α, %78.5 ± 6.830.6830.0016
Vascular density,/mm2292.5 ± 10.270.6250.0037

In the synovial tissue from the OA control patients, there were few positive cells that bound antibodies against CD147, VEGF, HIF-1α, and CD31. Statistical analysis showed no correlation among these values based on the current number of OA patients.

Expression of CD147 on RA FLS and effects of silencing CD147.

As shown in Figure 2, RA FLS were identified by morphologic study and flow cytometric analysis as a homogeneous population with the phenotype <1% CD14, <1% CD68, and >98% CD90. The percentage of cells staining positive for CD147 on RA FLS (mean ± SD 98.25 ± 0.18) was significantly higher than the percentage of OA FLS (87.97 ± 0.38; P < 0.05). To investigate the role of CD147 in RA FLS, RNA interference was used to knock down the expression of CD147. The CD147-specific siRNA (siCD147) and the Silencer negative control siRNA were tested for their ability to specifically suppress CD147. RT-PCR showed that Silencer negative control siRNA was incapable of inhibiting CD147 gene expression, whereas siCD147 effectively decreased the mRNA expression of CD147. These results were confirmed by Western blotting (data not shown). The protein expression of CD147 was obviously decreased in siCD147-transfected cells, but not in Silencer negative control siRNA–transfected cells.

Figure 2.

Flow cytometry analysis of surface markers on fibroblast-like synoviocytes (FLS) from patients with rheumatoid arthritis (RA). Shown are the expression of CD14 (A), CD68 (B), CD90 (C), and CD147 (D) in RA FLS, staining of RA FLS for CD90 and CD147 (E), and anti-CD147 antibody staining of RA FLS (red line) and osteoarthritis control FLS (green line), as well as isotype control staining (black line) (F). FITC = fluorescein isothiocyanate.

Production of VEGF and HIF-1α and effects of CD147 silencing.

Results of real-time PCR, ELISA, and Western blotting (Figure 3) showed that the expression and secretion of VEGF and HIF-1α in RA FLS were significantly higher (P < 0.05) than in OA FLS. Compared with the values in untreated RA FLS, the expression of VEGF and HIF-1α decreased significantly with siCD147 treatment (by a mean ± SD of 78.59 ± 3.47% and 52.66 ± 2.19%, respectively) and with infliximab treatment (by 34.25 ± 1.27% and 30.78 ± 2.38%, respectively) (P < 0.05 for each comparison). When RA FLS were treated with siCD147 plus infliximab, the expression of VEGF and HIF-1α was also significantly reduced (by 89.25 ± 9.64% and 81.03 ± 6.39%, respectively), and significant differences as compared with the groups treated with siCD147 alone or with infliximab alone (P < 0.05 for each comparison) were noted.

Figure 3.

Effects of CD147 silencing and infliximab treatment on VEGF and HIF-1α production by RA FLS. Forty-eight hours after treatment with small interfering RNA (siRNA) for CD147, treatment with infliximab (IFX), or no treatment, the expression of VEGF and HIF-1α was found to be decreased, as determined by real-time quantitative polymerase chain reaction (A), enzyme-linked immunosorbent assay (B) and Western blotting (C) with densitometric analysis (D). β-actin was included as a loading control for the Western blots. Values in A, B, and D are the mean ± SD. # = P < 0.05 versus OA FLS; ∗ = P < 0.05 versus RA FLS; ∗∗ = P < 0.05 versus RA FLS plus infliximab. See Figure 1 for other definitions.

Effects of anti-CD147 mAb on VEGF and HIF-1α expression.

The expression of VEGF and HIF-1α in the grafts was determined by immunochemistry, as shown in Figure 4A. The mean ± SD percentages of the engrafted tissue slices that stained positive for VEGF and HIF-1α, respectively, were as follows: 71.55 ± 3.16% and 60.32 ± 2.27% in the untreated group, 70.92 ± 4.38% and 58.80 ± 3.25% in the control mAb–treated group, 23.11 ± 1.82% and 16.58 ± 2.90% in the CD147 mAb–treated group, 51.82 ± 3.52% and 32.56 ± 2.81% in the infliximab–treated group, and 12.74 ± 1.24% and 11.45 ± 1.36% in the CD147 plus infliximab–treated group. Statistical analysis showed that in comparison with the control mAb–treated group, the expression of VEGF and HIF-1α was significantly decreased in the infliximab-treated group (by 26.93 ± 2.45% and 44.63 ± 3.02%, respectively), the CD147 mAb–treated group (by 67.41 ± 3.57% and 71.80 ± 3.82%), and the CD147 plus infliximab–treated group (by 82.04 ± 3.93% and 80.53 ± 3.67%) (P < 0.05). Significant differences were observed among these 3 groups that showed decreases, and anti-CD147 mAb was found to have stronger inhibitory effects than infliximab (P < 0.05). The combination treatment group showed a similar tendency, but no inhibitory effect was found for the control mAb–treated and the untreated groups (P = 0.34 for VEGF and P = 0.15 for HIF-1α for each comparison).

Figure 4.

Histologic analysis of VEGF, HIF-1α, and CD31, as well as volumes and average vascular density of implanted tissue in the SCID mouse coimplantation model of RA (SCID-HuRAg). A, Histologic and immunohistochemical analyses of tissue from SCID-HuRAg mice treated with control monoclonal antibody (mAb), CD147 mAb, infliximab, or the combination of CD147 and infliximab. HE = hematoxylin and eosin. Original magnification × 100. B, Volume of implanted tissue in each of the treatment groups before implantation and after treatment. C, Average vascular density of implanted tissue in each of the treatment groups before implantation and after treatment. Values in B and C are the mean ± SD. # = P < 0.05 versus the corresponding group before implantation; ∗ = P < 0.05 versus the infliximab group after treatment; ∗∗ = P < 0.05 versus the CD147 mAb group after treatment. See Figure 1 for other definitions.

Effects of anti-CD147 mAb on the volume of the engrafted tissue.

Successful implantations of normal human cartilage with rheumatoid synovium were observed by visual assessment 4 weeks after implantation. Nine weeks after implantation, the engrafted tissue grew almost 1.2-fold in size in the control mAb–treated group and the untreated group. In the anti-CD147 mAb–treated group, the infliximab-treated group, and the CD147 plus infliximab–treated group, the size of the grafted tissue was reduced by a mean ± SD of 29.72 ± 1.26%, 23.05 ± 1.32%, and 34.58 ± 1.67%, respectively, as compared with its original size (P < 0.05 for each comparison), whereas no significant differences among these 3 groups was found (Figure 4B).

Effects of anti-CD147 mAb on the vascular density of the engrafted tissue.

As shown in Figure 4C, in the anti-CD147 mAb–treated group, the infliximab-treated group, and the CD147 mAb plus infliximab–treated group, the vascular density of the engrafted tissue (100.07 ± 15.23/2,500 mm2, 163.09 ± 13.42/2,500 mm2, and 65.41 ± 12.38/2,500 mm2, respectively) was lower than the original vascular density (310.67 ± 13.58/2,500 mm2, 301.51 ± 11.27/2,500 mm2, and 303.65 ± 11.20/2,500 mm2, respectively; P < 0.05 for each comparison). Moreover, histologic evaluation revealed a significant reduction in the average vascular density in the anti-CD147 mAb–treated group compared with the infliximab-treated group (P < 0.05). The combination treatment group showed a similar tendency, while no inhibitory effect was found in the control mAb group or the untreated group (P >0.05 for each comparison).

Involvement of the phosphatidylinositol 3-kinase (PI3K)/Akt pathway in CD147-induced elevated expression of VEGF and HIF-1α.

As shown in Figure 5, after 48 hours, the levels of VEGF and HIF-1α were reduced by 21.11 ± 1.38% and 6.56 ± 1.62%, 38.14 ± 2.31% and 30.49 ± 2.08%, 69.03 ± 1.36% and 66.68 ± 1.87%, 85.61 ± 1.34% and 79.90 ± 1.65%, respectively, after exposure to 1, 10, 20, and 40 μM of LY294002, a specific inhibitor of PI3K/Akt (P < 0.05 for each comparison). These results suggest that LY294002 inhibits CD147-induced VEGF and HIF-1α production in a dose-dependent manner. The maximum effect was achieved at a concentration of 40 μM. There were no significant differences among the groups following exposure to PD98059, SP600125, SB203580, and U0126, which are specific inhibitors of ERK, JNK, p38 MAPK, and MAPK, respectively (P > 0.05 for each comparison), indicating that the induction of VEGF and HIF-1α by CD147 is largely independent of these pathways. Nonspecific toxicity is not responsible for the inhibitory effects of LY294002, since the viability of FLS, as determined by morphologic analysis, MTT assay, and a double-staining method using an FITC-labeled annexin V/propidium iodide apoptosis detection kit, was not influenced by LY294002 (data not shown).

Figure 5.

Dependence of CD147-mediated VEGF and HIF-1α induction on the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway. A, Akt phosphorylation in RA FLS is sensitive to the PI3K-specific inhibitor LY294002. B, CD147 monoclonal antibody (mAb) dose-dependently inhibits the Akt phosphorylation stimulated by CD147. C, There are no significant differences in Akt phosphorylation among groups treated with CD147 mAb, LY294002, or CD147 mAb plus LY294002. D, LY294002 (left) and CD147 mAb (middle) dose-dependently inhibit VEGF production, whereas there are no significant differences in VEGF production among groups treated with CD147 mAb, LY294002, or CD147 mAb plus LY294002 (right). Values are the mean ± SD. E, LY294002 and CD147 mAb dose-dependently inhibit HIF-1α production, whereas there are no significant differences in HIF-1α production among groups treated with CD147 mAb, LY294002, or CD147 mAb plus LY294002. β-actin was used as a loading control for the Western blots. See Figure 1 for other definitions.

In the anti-CD147 mAb–treated group, the levels of VEGF and HIF-1α after 48 hours were correspondingly reduced by 62.19 ± 2.45% and 52.75 ± 2.14%, 76.02 ± 2.84% and 73.67 ± 2.83%, 92.95 ± 3.48% and 90.47 ± 3.55%, respectively, after exposure to 20, 40 or 80 μg/ml of anti-CD147 mAb (P < 0.05 for each comparison). These results suggest that anti-CD147 mAb inhibits the production of CD147-induced VEGF and HIF-1α in a dose-dependent manner. The maximum effect was achieved at a concentration of 80 μg/ml.

To determine whether Akt phosphorylation was involved in the CD147-mediated production of VEGF and HIF-1α, the expression levels of Akt and phospho-Akt were examined by Western blotting. The levels of phospho-Akt decreased by 58.63 ± 3.25%, 72.53 ± 4.36%, and 91.64 ± 3.85%, respectively, after exposure to 20, 40, or 80 μg/ml of anti-CD147 mAb (P < 0.05 for each comparison). No significant differences were observed among the groups treated with anti-CD147 mAb, LY294002, or anti-CD147 mAb plus LY294002 (P > 0.05 for each comparison). All of these results suggest that PI3K, a key downstream signaling molecule of CD147, is involved in CD147-induced VEGF and HIF-1α production in RA FLS. Similar results were observed in at least 3 independent experiments.


Angiogenesis is considered a hallmark of RA, and pannus is one of the critical elements that promote the FLS proliferation, cartilage invasion, and bone destruction. Persistent angiogenesis will lead to chronic changes in the architecture of the RA synovium via delivery of nutrients and inflammatory cells and the production of cytokines and protease activity. Angiogenesis is known to have resulted from the outgrowth of endothelial cells from preexisting capillary vessels initiated by the migration of endothelial cells away from the parental vessels.

Among the known positive regulators of angiogenesis, VEGF and HIF-1α are critical mediators of normal and abnormal angiogenesis. Several investigations have shown that the concentration of VEGF in synovial fluid is significantly higher in patients with RA than in patients with OA or other forms of arthritis (12). Furthermore, the serum concentration of VEGF is higher in RA patients than in patients with other forms of arthritis (13). HIF-1α is another important and well characterized “master regulator” of the adaptive response to alterations in oxygen tension. Activation of the HIF-1α transcription factor signaling cascade leads to extensive changes in gene expression, which triggers transcription activity in hypoxic cells, leading to the expression of genes involved in angiogenesis (14).

The presence of VEGF and HIF-1α in arthritic synovium strongly suggests their participation in synovitis; thus, the regulators of VEGF and HIF-1α production are of great interest. In our previous research, the abnormally high levels of CD147 expression in RA and its role in regulating MMP expression was extensively studied (15), but the effect of CD147 on the production of VEGF and HIF-1α by synovial cells was not addressed. In the present study, we found that the expression of CD147 was positively related to VEGF and HIF-1α expression in RA synovium, and the production of VEGF and HIF-1α were significantly reduced after transfection of RA FLS with siCD147 or after the addition of anti-CD147–specific antibodies to RA FLS in a hypoxic culture system. These findings are evidence of a role of CD147 in the angiogenesis of RA synovium. CD147 may promote pannus formation through enhancement of VEGF and HIF-1α.

To further study the proangiogenic function of CD147 in vivo, we conducted a treatment study using the SCID-HuRAg mouse model, a widely used model for the evaluation of biologic agents. We found that CD147 mAb reduced the production of both VEGF and HIF-1α as well as the formation of new vessels. The in vivo results in the SCID-HuRAg mouse model are consistent with the in vitro findings in the RA FLS system, both suggesting that CD147 up-regulation on RA FLS induces the up-regulation of VEGF and HIF-1α, which in turn, may further augment angiogenesis. Moreover, our results showed that CD147 mAb has stronger antiangiogenic effects than infliximab, a new biologic agent that has been approved by regulatory authorities for the treatment of RA.

Many drugs currently used in the treatment of RA have antiangiogenic effects, which are exerted at different levels. The rationale that TNFα plays a central role in regulating VEGF and HIF-1α was initially provided by the demonstration that anti-TNFα antibodies added to RA FLS cultures inhibited the spontaneous production of VEGF and HIF-1α (16). Paleolog et al (17) reported that anti-TNFα antibody decreased VEGF production in RA synovial tissue, suggesting that TNFα does induce VEGF in synovial tissue, whereas Hashizume et al (18) reported that TNFα failed to induce angiogenesis in a coculture system of human umbilical vein endothelial cells and synovial cells. Since many kinds of cells exist in RA synovial tissue, one reason for this discrepancy may be that TNFα induces the production of VEGF by cells other than FLS in synovial tissue. There have as yet been few reports on the relationship between TNFα and HIF-1α, but Zhou et al (19) have found that TNFα causes the accumulation of a ubiquitinated form of HIF-1α through the NF-κB pathway. The results of these studies indicate that the mechanisms by which anti-TNFα antibodies decrease angiogenesis are unknown and deserve further exploration.

The mechanism by which VEGF and HIF-1α transduction occurs via CD147 remains unclear, but the PI3K/Akt pathway would be expected to be involved. The reduction of phospho-Akt in RA FLS by CD147 mAb treatment suggests that CD147 may stimulate the PI3K/Akt pathway. The role of the PI3K/Akt pathway in CD147-induced up-regulation of VEGF and HIF-1α could have been better confirmed if an siRNA silencing method had been used. But our work was limited by the poor viability of RA FLS after siCD147 under serum-free culture conditions, which is a precondition for studies of pathways. To make up for this liability, we conducted the combination assay of LY294002 and anti-CD147 mAb. No significant difference was observed among the groups treated with LY294002, anti-CD147 mAb, and the combination of LY294002 plus anti-CD147 mAb. LY294002 and anti-CD147 mAb were found to have shared the downstream pathway, confirming that the PI3K/Akt pathway is involved in CD147-induced signaling. CD147 expressed on the surface of RA FLS may stimulate PI3K/Akt signaling pathways via homophilic interactions, which have been shown to stimulate MMP production (20). Moreover, CD147 expressed on RA FLS also induces the production of a soluble form of CD147, which in turn, further induces the expression of CD147 through a positive feedback mechanism (21).

The findings of this study extend the role of CD147 in RA, from an MMP stimulator to a proangiogenic promoter. Anti-CD147 mAb treatment has been reported to suppress the development of collagen-induced arthritis (22), and studies aimed at direct inhibition of CD147 could be a new and promising treatment for RA.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Wang had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. C.-h. Wang, Z.-h. Chen, Zhu.

Acquisition of data. C.-h. Wang, Yao, L.-n. Chen, Jia, L. Wang.

Analysis and interpretation of data. C.-h. Wang, Dai, Zheng.


The authors would like to thank experienced pathologist Professor P. S. Yan for his excellent technical assistance.