Induction of a CXCL8 binding site on endothelial syndecan-3 in rheumatoid synovium


  • This work was undertaken by the Robert Jones and Agnes Hunt Orthopaedic and District Hospital NHS Trust, which receives a proportion of its funding from the NHS Executive



To identify and characterize which endothelial heparan sulfate proteoglycans (HSPGs) bind the chemokine CXCL8 (interleukin-8) in human rheumatoid arthritis (RA) and nonrheumatoid synovia.


CXCL8 binding to endothelial HSPGs in RA and nonrheumatoid synovia was determined by heparinase treatment followed by an in situ binding assay and autoradiography. Endothelial HSPGs were characterized by immunohistochemical analysis and quantitative reverse transcriptase–polymerase chain reaction (RT-PCR). Phosphatidyinositol-specific phospholipase C (PI-PLC) and antibodies to HSPGs were used in in situ binding experiments to identify which HSPGs bound CXCL8.


The expression of heparan sulfate on microvascular endothelial cells was demonstrated in RA and nonrheumatoid synovia. Using antibodies to syndecan-1–4 and glypican-1, -3, and -4, the selective expression of syndecan-3 by endothelial cells was detected in RA and nonrheumatoid synovia. In addition, RT-PCR showed the presence of syndecan-3 messenger RNA in endothelial cells extracted from RA and nonrheumatoid synovia. 125I-CXCL8 bound to venular endothelial cells; treatment with heparinases I and III significantly reduced this binding in RA but not nonrheumatoid synovia. 125I-CXCL8 binding was not reduced after treatment with PI-PLC, which cleaves glycosyl phosphatidylinositol linkages, suggesting that CXCL8 did not bind to glypicans. Treatment of synovia with a syndecan-3 antibody reduced CXCL8 binding to RA but not nonrheumatoid endothelial cells; however, no reduction in binding was observed with syndecan-2 or glypican-4 antibodies.


Our results show the selective induction of a CXCL8 binding site on endothelial syndecan-3 in RA synovium. This site may be involved in leukocyte trafficking into RA synovial tissue.

Heparan sulfate (HS) is a carbohydrate chain characterized by the presence of alternating uronic acid (D-glucuronic acid or L-iduronic acid) and D-glucosamine units (1, 2). This glycosaminoglycan occurs covalently attached to specific core proteins to form heparan sulfate proteoglycan (HSPG) molecules. Two major families of cell-surface HSPGs exist: syndecans and glypicans. These 2 families of HSPGs can be distinguished by their primary amino acid sequences (3). The 4 syndecans, designated syndecan-1, -2, -3, and -4, have protein cores with characteristic structural domains (4). The variable ectodomain, which is exposed to the extracellular environment, contains 3–5 HS chains and is attached to the cell membrane via a hydrophobic transmembrane segment (5, 6). In addition, there is an intracellular domain containing peptide sequences that serve as substrates for cellular kinases, enabling syndecans to act as signaling molecules (7, 8).

The glypican family, which contains 6 members (glypican-1, -2, -3, -4, -5, and -6), is distinguished by cysteine-rich globular ectodomains and the presence of 2–3 HS chains. They are attached to the extracellular cell surface by glycosyl phosphatidylinositol linkages (9). Several minor membrane proteoglycans containing HS chains have been described, such as epican, betaglycan, and others (1, 2). HSPGs that also occur in the basement membrane include perlecan, agrin, and type XVIII collagen (10, 11). The syndecans, glypicans, and basement membrane HSPGs are expressed in a cell-, tissue-, and development-specific manner (3). HSPGs bind a wide range of protein ligands (e.g., basic fibroblast growth factor [bFGF] and platelet-derived growth factor), which interact with the HS chains (12).

Chemokines are essentially basic molecules and exhibit electrostatic interactions with polyanionic HS (13). The binding affinities have been variously reported in the nanomolar or micromolar range for CXCL8, CCL2, CCL3, and CCL5 (14–16). Chemokines display differences in their strength of interactions with glycosaminoglycans. For example, CXCL8 shows the following range: heparin > heparan sulfate > chondroitin sulfate = dermatan sulfate (14). Different chemokines show variation in their affinities for glycosaminoglycans that are not solely based on charge considerations, suggesting selectivity in these interactions (14, 17, 18). Some of the structural domains of HS that bind chemokines have been described, including those for CXCL8, CXCL4, CXCL12, and CCL3 (19–22).

Rheumatoid arthritis (RA) is characterized by chronic inflammation of the synovial membrane; the infiltration and activation of leukocytes can result in progressive destruction of cartilage and bone. There is substantial evidence showing an involvement of chemokines and their receptors in RA (for review, see ref. 23). CXCL8, the “prototypic” chemokine, is a major chemokine in RA (23–25). Significantly elevated levels of CXCL8 and other chemokines are present in human RA synovial fluid, where they are produced by cells of the synovium and cartilage. Chemokines produced by joint tissue cells are biologically active and can stimulate leukocyte migration. The functional importance of these cytokines and their receptors in joint inflammation has been shown in animal models of RA, in which antibodies to these mediators or receptor antagonists significantly reduce the number of leukocytes recruited and the severity of disease (26–28). In addition, administration of chemokines to the joints in animal models of RA results in inflammation and leukocyte recruitment (29).

An increasing body of evidence supports the functional importance of chemokine–HS interactions in the process of leukocyte migration from the blood into inflamed tissues (13, 30–32). Data suggest an involvement of HS in the transcytosis of chemokines across endothelial cells. In this way, chemokines that are produced extravascularly, by macrophages and fibroblasts for example (23), could be relayed to the luminal surface of the endothelium. At this interface HS is then involved in presenting the bound chemokines to signaling receptors on the surface of adherent blood leukocytes, as has been shown in the RA synovium (33, 34). This results in leukocyte activation and migration across the endothelium and into the tissue. The functional importance of chemokine–HS interactions has been shown in vivo, because administration of modified chemokines that exhibit reduced affinity to HS or heparin resulted in diminished leukocyte recruitment (31, 32). Other functions of glycosaminoglycan–chemokine interactions have been proposed, including the formation of immobilized chemokine gradients in the extracellular matrix and the protection of these mediators from degradation, resulting in the prolongation of their activities (15, 35, 36).

Although an involvement of chemokine–endothelial HS interactions in inflammation is relatively clear, it is not known which endothelial HSPGs are responsible for the chemokine binding in chronic inflammatory diseases. The role of the core protein of HSPG is becoming increasingly apparent (6). It not only determines when and where the HS chains are expressed but also plays a direct role in signaling. Recent studies in gene-deleted mice have presented some evidence of the involvement of syndecan-1 and -4 in inflammation (37, 38); however, little is known about the role of syndecans and glypicans in autoimmune diseases such as RA. The aim of the present study was to determine the expression of HSPGs by endothelial cells in the RA synovium and to determine which of these molecules bind the chemokine CXCL8.


Tissue samples

Samples of synovium were obtained, with informed consent, from the suprapatellar pouch and medial gutter (39) of patients with RA or from patients with noninflammatory pathologic features. RA synovium was obtained at the time of total knee replacement or synovectomy, from patients who fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for RA (40). Patients who did not have RA had knee joint symptoms suggestive of meniscal damage or osteoarthritis (OA). Tissue either was used immediately for in situ binding assays or was frozen in ice-cold isopentane for chemokine binding/inhibition assays and immunohistochemical analysis. Samples of cartilage were obtained (with informed consent) from patients who were undergoing total knee replacement for OA. Normal human skin samples were kindly provided by Drs. E. Johnston and C. Mangham at the University of Birmingham.

Chemokine binding/inhibition assay

Chemokine binding to synovia and the effects of enzymes and antibodies were determined using a previously described method (41), with minor modifications. Briefly, cryostat sections (30-μm–thick, 5 sections per tube) of synovium were cut, added to Hanks' balanced salt solution (HBSS), and pulse centrifuged for 10 seconds. The supernatant was carefully aspirated, and the pellet was gently resuspended in a 200-μl enzyme solution containing 10 units/ml heparinase I, 2 units/ml heparinase III, or 0.18 units/ml phosphatidylinositol-specific phospholipase C (PI-PLC) (42) in HBSS, or HBSS alone. All enzymes were obtained from Sigma (Poole, UK). Sections were incubated for 1.5 hours at 37°C.

After enzymatic treatment, the samples were rinsed 3 times: once with HBSS at 37°C, once with HBSS at room temperature, and once with ice-cold binding buffer. An antibody to syndecan-2 (Santa Cruz Biotechnology, Wembley, UK), syndecan-3 (Santa Cruz Biotechnology), or isotype goat IgG control (Vector, Burlingame, CA) in HBSS (either 10 μg/ml or 50 μg/ml) or an antibody to glypican-4 (a kind gift from Professor H. D. Haubeck, Freiburg, Germany) or rabbit serum control (Dako, Ely, UK) in HBSS (1:100) was added for 1 hour prior to the addition of radiolabeled chemokine. Radioactively labeled recombinant human 125I-CXCL8 (interleukin-8; molecular weight 8,400, 262 μCi/μg) (Perkin Elmer, Cambridge, UK) was diluted to a concentration of 0.3 nM (0.2 μCi) in ice-cold binding buffer (HBSS, 10 mM HEPES [pH 7.2], 0.1% bovine serum albumin), and 200 μl was added to each tube and incubated for 2 hours at 4°C. These binding conditions were identical to those described in our previous report (41), showing specific and saturable binding of 125I-CXCL8 to RA and nonrheumatoid synovial endothelium. Unbound 125I-CXCL8 was removed with 3 rinses of ice-cold binding buffer. Sections were resuspended and fixed in 200 μl of Bouin's fluid (7.5 ml saturated aqueous picric acid, 2.5 ml formalin, and 100 μl acetic acid; Sigma) (43) and incubated for 30 minutes at room temperature. After fixation, sections were rinsed twice in distilled H2O before the addition of 400 μl of distilled H2O. The amount of bound 125I-CXCL8 was determined by gamma counting and was expressed as counts per minute.

Sequential sections of synovia were cut (10-μm–thick) and mounted onto 3-aminopropyltriethoxysilane–coated glass slides. Sections were rehydrated in phosphate buffered saline (PBS) for 5 minutes and then in ice-cold binding buffer for 5 minutes, prior to incubation in enzyme or antibody solutions, as described above. The tissue was rinsed first in PBS and then in ice-cold binding buffer and incubated with 125I-CXCL8, as described above. Slides were washed 3 times with ice-cold HBSS and fixed in Bouin's fixative for 30 minutes, followed by 2 rinses with distilled water. Slides were air-dried at 4oC for 2 hours, coated with LM1 emulsion (Amersham Pharmacia Biotech, Little Chalfont, UK), exposed at 4oC for 2 weeks, developed, and stained with hematoxylin and eosin.

Immunohistochemical analysis

The following antibodies were used: anti-HS (clone 10E4 and clone Hep SS-1; Seikagaku, Abingdon, UK), anti–syndecan-1 (clone B-B4; Serotec, Oxford, UK), anti–syndecan-2 (clone 10H4), anti–syndecan-3 (clone 1C7), anti–syndecan-4 (clone 8G3), anti–glypican-1 (clone S1) (all prepared and supplied by Dr. G. David, University of Leuven, Belgium), anti–glypican-3 (Santa Cruz Biotechnology), and anti–glypican-4 (a kind gift from Professor H. D. Haubeck). The method described by Roskams et al (44) was followed, using the anti-HS, anti-syndecan, and anti–glypican-1 antibodies, with minor modifications. For glypican-3, the method described by our group (24) for polyclonal antibodies was used, with minor modifications. For glypican-4, the polyclonal antibody was diluted in 10% human serum and detected with swine anti-rabbit horseradish peroxidase conjugate diluted in 10% human serum. Briefly, 10-μm–thick serial cryostat sections of synovium, skin, and cartilage were dried for 1 hour and stored at −20°C until used. Prior to immunohistochemical analysis, slides were left to equilibrate to room temperature for 30 minutes, and sections were rehydrated in PBS for 5 minutes. All primary antibodies were used at a concentration of 5 μg/ml, because initial experiments showed that this concentration gave the optimum specific staining over the range of 2.5–10 μg/ml. Antibody binding was detected using a diaminobenzidine staining kit (Vector).

Isolation of endothelial cells and RNA analysis

High endothelial venule cells were extracted from RA and OA synovia, using the method described by Lacorre et al (45), except that immunomagnetic purification involved the use of an antibody to the Duffy antigen receptor for chemokines (DARC) instead of MECA-79. DARC is a widely expressed endothelial marker that is present in the venules of normal and chronically inflamed tissues (41). Human umbilical vein endothelial cells (HUVECs) were isolated by collagenase treatment, using standard methods (46).

Total RNA was isolated from 500,000 cells, using the Absolutely RNA Microprep Kit according to the manufacturer's instructions (Stratagene, La Jolla, CA). RNA preparations (20 ng of each) were reverse transcribed and amplified for 21 cycles with the Super SMART PCR cDNA Synthesis Kit according to the manufacturer's instructions (Clontech, Palo Alto, CA). Amplified complementary DNA (cDNA) concentrations were determined from absorbance measurements by spectrophotometer.

Real-time polymerase chain reaction (PCR) was conducted on 5 ng of Super SMART cDNA and 0.4 μM of each primer, using the SYBR Green PCR Master Mix Kit (Applied Biosystems, Foster City, CA) in a final volume of 25 μl. The specific PCR primers for syndecan-3 were ACCCCAACTCCAGAGACCTT (forward) and CCCACAGCTACCACCTCATT (reverse); the specific primers for GAPDH were GAGTCAACGGATTTGGTCGT (forward) and GACAAGCTTCCCGTTCTCAG (reverse). In all samples, the housekeeping gene GAPDH was evaluated as an internal control. The cycling conditions were as follows: initial denaturation at 95°C for 10 minutes, followed by 40 cycles at 95°C for 10 seconds and 60°C for 1 minute. PCR products (10 μl) were separated on a 2% agarose/ethidium bromide gel for visualization. Reactions were analyzed with the ABI PRISM 7700 Sequence Detection System (Applied Biosystems), and the threshold cycles (Ct) for each sample, run in duplicate, were determined. The relative expression of syndecan-3 in endothelial cells was calculated using the delta delta Ct method, as previously described (47).

Statistical analysis

Student's t-tests were used to analyze the data. P values less than 0.05 were considered significant.


Binding of CXCL8 to endothelial HS. Synovial samples from patients without RA were not inflamed and showed a normal intimal layer and no leukocyte infiltration. The synovia from patients with RA demonstrated both inflamed and uninflamed characteristics. Samples showing inflammation exhibited a thickened intimal layer and mononuclear cell infiltrates. The uninflamed samples from patients with RA were similar to the nonrheumatoid samples, suggesting disease quiescence or end-stage disease. The samples were scored according to the degree of inflammation, as follows: 0 = normal intimal layer and no leukocyte infiltration, 1 = small foci/areas of leukocyte infiltration with ≥2 synovial lining cells, 2 = synovitis with moderate infiltrations and ≥3 synovial lining cells, and 3 = severe synovitis with widespread infiltration and ≥4 synovial lining cells (Table 1).

Table 1. Patient characteristics, degree of inflammation of synovial samples, and antibody reactivity*
Patient/sex/age, yearsSynovial inflammationAntibody reactivity of endothelial cells
SyndecanGlypicanHeparan sulfate
  • *

    Synovial inflammation was graded as follows: 0 = normal intimal layer and no leukocyte infiltration; 1 = small foci/areas of leukocyte infiltration with 2 or more synovial lining cells; 2 = synovitis with moderate infiltrations and 3 or more synovial lining cells; 3 = severe synovitis with widespread infiltration and 4 or more synovial lining cells. Immunoreactivity of endothelial cells was scored as follows: − = no staining; + = weak staining; ++ = strong staining; +++ = very strong staining. RA = rheumatoid arthritis.


As determined by immunohistochemistry (clone 10E4), HS was present in both RA and nonrheumatoid synovia and was localized to the endothelial cells of both venules and arterioles (Figures 1A and B) and cells in the intimal layer.

Figure 1.

Expression of heparan sulfate in human synovia and the effect of heparinase I and heparinase III on binding of 125I-CXCL8. Immunohistochemical analysis using antibody 10E4 showed that heparan sulfate localizes strongly to endothelial cells in nonrheumatoid synovia (A) and rheumatoid synovia (B). Cryosections of synovium were cut and incubated with heparinase I or III. After incubation, the sections were rinsed and incubated with 125I-CXCL8 in binding buffer. Following a further rinse the sections were fixed in Bouin's fluid and placed in a gamma counter. Heparinase treatment significantly reduced the binding of 125I-CXCL8 to rheumatoid synovia but not to nonrheumatoid synovia (C). Sections from the same synovia were mounted on slides and treated with heparinase I or left untreated, followed by autoradiography (D and E). D,125I-CXCL8 binding to synovial endothelial cells in the absence of heparinase pretreatment. E, Section serial to section in D, showing a reduction in 125I-CXCL8 binding after incubation with heparinase I. Values are the mean ± SEM percent of bound 125I-CXCL8 in 8 rheumatoid samples and 8 nonrheumatoid samples. (Original magnification × 200 in A; × 100 in B; × 1,000 in D and E.)

Cryosections were incubated with 125I-CXCL8, and binding of this chemokine to venular endothelial cells of both RA and nonrheumatoid synovia was observed; the interaction was specific, because it could be displaced by 1,000-fold excess unlabeled CXCL8, as assessed by gamma counting and autoradiography. Furthermore, the binding was multispecific, because 1,000-fold excess heterologous chemokine CCL5 displaced CXCL8 binding to the endothelium in RA and nonrheumatoid synovia, suggesting that sites other than chemokine receptors CXCR1 or CXCR2 were involved. This CXCL8 binding profile and its specificity have already been described under identical binding conditions in our previous study (41). In order to investigate whether the binding involved HS, sections were treated with and without heparinases I and III (48). Sections were then incubated with 125I-CXCL8, and gamma counting was performed. These experiments revealed that heparinase treatment significantly reduced the binding of CXCL8 to RA synovia but not to nonrheumatoid synovia (Figure 1C). The reductions in chemokine binding to RA synovia were 65% and 50% with heparinases I and III, respectively. The amounts of bound 125I-CXCL8 in non–heparinase-treated RA and nonrheumatoid synovial tissue were not significantly different (mean ± SEM 2,800 ± 354 cpm and 3,401 ± 403 cpm, respectively).

The distribution of CXCL8 binding in synovia was determined by the attachment of sequential sections to slides, which were then heparinase-treated using the same method as described above but with binding detected by autoradiography. Silver grains were detected on the endothelial cells of venules in RA synovia. Grain counting confirmed a reduction in the number of these grains after heparinase treatment (Figures 1D and E). In nonrheumatoid synovia, however, there was no difference in the number of silver grains detected on endothelial cells in heparinase-treated and untreated sections. Heparinases I and III were active in degrading endothelial HS in RA and nonrheumatoid synovia, since immunohistochemistry of sections after enzyme treatment reduced staining intensities using 10E4, an antibody that recognizes intact HS chains (data not shown).

Overall, these results indicated that HS bound CXCL8 on the venular endothelium in synovia obtained from patients with RA, but not in synovial samples from patients without RA.

Identification of HSPGs expressed by synovial endothelial cells. To investigate which HSPGs may be involved in CXCL8 binding, the presence of syndecans and glypicans on synovial endothelial cells was studied by immunohistochemistry, using a panel of antibodies to HSPGs that were available from academic and commercial sources.

Cryosections of RA and nonrheumatoid synovia, normal skin, and OA cartilage were incubated with antibodies to each of syndecan-1–4 and glypican-1, -3, and -4 (Table 1). Syndecan-3 staining was clearly demonstrated on endothelial cells in the synovia from all patients with RA (Figure 2A). Staining for syndecan-3 was weaker in nonrheumatoid compared with RA synovial endothelial cells (Table 1). Weak immunoreactivity for syndecan-2 was present in the walls of synovial blood vessels, including pericytes and smooth muscle cells (Figure 2C), but in contrast to syndecan-3 displayed a less endothelial cell–selective distribution. Glypican-4 reactivity was present in both RA and nonrheumatoid synovial venules, localizing to endothelial cells, pericytes, and smooth muscle cells (Figure 2D and Table 1). The binding of antibodies to syndecan-1 and -4 and glypican-1 and -3 to synovial endothelial cells was below the level of detection or barely detectable (Table 1).

Figure 2.

Immunohistochemical analysis of syndecans and glypicans. Sections of human synovium were incubated with antibodies and processed using immunoperoxidase and diaminobenzidine (brown staining) as substrate. Syndecan-3 immunoreactivity is present in endothelial cells of rheumatoid (A) and nonrheumatoid synovium (B). Staining for syndecan-2 is present in the wall of a blood vessel, including pericytes and smooth muscle cells (C), yet shows a less endothelial-selective distribution compared with syndecan-3 (A). Glypican-4 was present within the venule wall, including endothelial cells, pericytes, and smooth muscle cells (D). (Original magnification × 1,000 in A and B;× 500 in C and D.)

Although it was not detectable on synovial endothelial cells, expression of syndecan-1 and -4 and glypican-1 was detected in keratinocytes and blood vessels of the skin, and expression of syndecan-3 was detected in chondrocytes (data not shown). This staining pattern in skin and chondrocytes was similar to that previously reported (49, 50) and in the present study served as a positive control for the antibodies. Immunostaining was negative in all experiments in which the primary antibodies to HSPGs and HS were replaced with the same concentrations of mouse ascitic fluid (for syndecans and glypican-1), normal goat IgG (for glypican-3), normal rabbit serum (for glypican-4), and normal mouse IgM (for HS) (data not shown).

Quantitative PCR was performed to confirm the expression of syndecan-3 by synovial endothelial cells. Endothelial cells were extracted from RA and control OA synovia, RNA was reverse transcribed, and real-time PCR was carried out. The relative fold expression was calculated by the delta delta Ct method (Table 2). PCR products were also run on an agarose gel for determination of band specificity, as shown in Figure 3. Syndecan-3 messenger RNA (mRNA) was detected in synovial endothelial cells from 3 of 5 RA patients, with the cells from 2 RA patients (patients 2 and 4) showing intense expression and those from 1 RA patient (patient 5) showing weak expression (expression in cells from RA patient 2 was 78.5-fold that in cells from RA patient 5 [Table 2]). The message for syndecan-3 was also detected in synovial endothelial cells from 2 of 6 synovial samples obtained from patients with OA, with the sample from OA patient 4 showing a more intense expression than that in the sample from OA patient 3 (expression in cells from OA patient 4 was 79-fold that in cells from OA patient 3 [Table 2]). These results indicate that synovial endothelial cells express syndecan-3 mRNA, although, unlike the protein, it was not present in every sample. The lack of an exact correlation between mRNA and protein has been shown for other genes (51, 52) and suggests that syndecan-3 mRNA and protein are differentially regulated. There were no obvious differences between individual RA or OA patients, such as variation in disease activity or severity, which could explain the variation in syndecan-3 message levels between individuals.

Table 2. Real-time PCR amplification of syndecan-3 expression by synovial endothelial cells*
SampleThreshold cycleExpression of RA2/sampleExpression of OA4/sample
  • *

    Syndecan-3 expression was normalized with the housekeeping gene GAPDH, using the delta delta Ct method for calculating fold expression. The relative fold expression in synovial samples from rheumatoid arthritis (RA) patient 2 and osteoarthritis (OA) patient 4, compared with the expression in the given amplified synovial sample, is shown. PCR = polymerase chain reaction; HUVEC = human umbilical vein endothelial cell; NA = not amplified.

Figure 3.

Detection of syndecan-3 mRNA in endothelial cells using polymerase chain reaction. Syndecan-3 mRNA was detected in synovial endothelial cells from some patients with rheumatoid arthritis (RA) (patients 2, 4, and 5) and control patients with osteoarthritis (OA) (patients 3 and 4). Expression of syndecan-3 mRNA was also demonstrated in cultured human umbilical vein endothelial cells (HUVECs) but not in HeLa cells. M = molecular weight marker; NC = negative control.

To investigate whether endothelial cells other than those from synovia express syndecan-3 mRNA, PCR was performed on RNA extracts from cultured HUVECs. We observed that syndecan-3 was also produced by HUVECs, with a high expression level similar to that observed in cells from RA patients 2 and 4 and OA patient 4. Expression of syndecan-3 was not demonstrated in HeLa cells.

Syndecan-3 binding of CXCL8 in the RA synovium. Results of immunohistochemical analysis indicated glypican-4 expression but a lack of detectable glypican-1 and glypican-3 expression by synovial endothelial cells. This suggested that the HS structures that bind CXCL8 may be distributed on glypican-4. In order to investigate further whether glypicans bind 125I-CXCL8, cryostat sections of synovia were treated with PI-PLC prior to the addition of the radiolabeled chemokine. This enzyme cleaves glycosyl phosphatidylinositol linkages, including those of glypicans, releasing these HSPGs from cell surfaces (42, 53). Overall, PI-PLC treatment of sections did not reduce the binding of 125I-CXCL8, as detected by gamma counting (Figure 4A) or autoradiography (data not shown). Only 1 of the 8 RA synovial samples showed a reduction in chemokine binding following PI-PLC treatment, and this sample was used as a positive control in experiments to indicate that the enzyme was active. Furthermore, cleavage of glypican was confirmed, because following PI-PLC treatment, synovial endothelial cells showed reduced immunostaining with anti–glypican-4 (data not shown).

Figure 4.

Endothelial syndecan-3 binds CXCL8 in rheumatoid arthritis (RA). AC, Graphs showing a lack of significant effect of phospholipase C (A), glypican-4 antibody (B), and syndecan-2 antibody (C) on binding of 125I-CXCL8 to RA and nonrheumatoid synovia. D, Following treatment with anti–syndecan-3, there was a significant reduction of CXCL8 binding to RA but not nonrheumatoid synovia in comparison with IgG control. Values are the mean ± SEM percent bound 125I-CXCL8 without treatment (set at 100%) and with treatment (n = 8 RA samples and 8 nonrheumatoid samples in A; n = 4 RA samples and 4 nonrheumatoid samples in BD). E, Autoradiogram showing CXCL8 binding to the endothelium in rheumatoid synovium in the presence of control IgG. F, Autoradiogram showing a reduction of CXCL8 binding to the endothelium of the same blood vessel as in E, in the presence of anti–syndecan-3 antibody. The autoradiographic exposure times for E and F were 3 days. (Original magnification × 1,000.)

The findings of immunohistochemical analysis and PCR suggested that syndecan-3 may be involved in the binding of CXCL8 on the synovial endothelium (Figures 2 and 3). In addition, immunohistochemical analysis indicated the potential involvement of glypican-4 and syndecan-2 in chemokine binding. In order to investigate these possibilities, sections of RA synovium were treated with an antibody to glypican-4, syndecan-2, or syndecan-3 prior to the addition of 125I-CXCL8. Treatment of sections with either anti–glypican-4 or anti–syndecan-2 did not significantly affect 125I-CXCL8 binding (Figures 4B and C); the lack of effect observed with anti–glypican-4 was consistent with the PI-PLC data (Figure 4A). However, syndecan-3 antibody treatment resulted in a significant 65% reduction of chemokine binding to RA synovia in comparison with IgG control (P < 0.0001) (Figure 4D). In nonrheumatoid synovia, there was no significant reduction in 125I-CXCL8 binding following treatment with syndecan-3 antibody. Anti–syndecan-3 also reduced binding of 125I-CXCL8 to RA but not nonrheumatoid synovial endothelial cells, as shown by autoradiography (Figures 4E and F).


This study showed that heparinases I and III significantly reduced CXCL8 binding to rheumatoid, but not nonrheumatoid, synovial endothelial cells. These results suggest that under conditions of chronic inflammation, expression of a CXCL8 binding motif is induced on the HS chains of endothelial HSPGs. We observed that endothelial cells in uninflamed synovia bound CXCL8, and immunohistochemical staining showed the presence of HS on these cells. Thus, although HSPG is present in the endothelium of uninflamed synovium, its HS chains do not appear to be decorated with CXCL8 binding structures, suggesting that molecules other than HS may be involved in binding this chemokine. In an earlier study we identified a promiscuous 125I-CXCL8 binding site on the endothelial cells of nonrheumatoid and RA synovia and provided evidence that this was the Duffy antigen rather than a chemokine receptor such as CXCR1 or CXCR2 (41). Therefore, in nonrheumatoid synovia the Duffy antigen would be involved in binding CXCL8 on endothelial cells, although a contribution from other molecules such as heparinase-insensitive glycosaminoglycans, including chondroitin sulfate, could also be involved.

The finding that CXCL8 binding structures are present in patients with RA but not in those without RA, as shown by treatment with heparinases, suggests that in chronically inflamed synovia HSPG synthesis is differentially regulated in endothelial cells. One reason for the induction of CXCL8 binding to HSPGs could involve soluble mediators, such as tumor necrosis factor. This cytokine, which is present in the RA joint (54), is known to increase HS sulfation and enhance binding of the chemokine CCL5 to cultured human microvascular endothelial cells (55).

The HSPG expression pattern of endothelial cells was studied by immunohistochemistry and RT-PCR in order to determine which of these molecules may be binding CXCL8 in rheumatoid synovium. Using antibodies to the 4 known syndecans (syndecan-1–4) and antibodies to glypican-1, -3, and -4, we showed that syndecan-3 was a major HSPG detected in synovial endothelial cells.

There have been some previous reports on the expression of syndecan-3 by human endothelial cells. The presence of this HSPG has been shown in the endothelial cells of normal liver, and intense reactivity on these cells occurs in hepatocellular carcinomas (56, 57). Human endothelial cells also express other syndecans, depending on their tissue source and stage of development. HUVECs and cultured aortic endothelial cells have been shown to express syndecan-1 and -2 (16, 58), and the present study shows that HUVECs also express syndecan-3 mRNA. In addition, endothelial cells in fetal lung tissue, but not adult lung tissue, express syndecan-4 (59). Syndecan-3 is the predominant syndecan in the nervous system, where it was first identified, and has been associated with the generation of cerebellar fibrillar plaques in Alzheimer's disease (60, 61). Interestingly, it is also an HSPG in the musculoskeletal system. The results of our study in adult human joints show that syndecan-3 is expressed in RA and nonrheumatoid synovium; it is also expressed by chondrocytes in normal and OA articular cartilage, and it is a regulator of chondrocyte proliferation (50, 62). Furthermore, syndecan-3 is involved in limb morphogenesis and skeletal development, including chondrogenesis and joint formation, and skeletal muscle regeneration (63, 64).

In some cell types, glypicans have been shown to harbor cytokine-binding sites on their HS chains. For example, PI-PLC releases bFGF from human bone marrow cells, indicating a role for glypicans in binding this cytokine (42), and the enzyme abrogates the mitogenic responses of bFGF and epidermal growth factor in breast cancer cells (53). In the current study, however, pretreatment of synovia with PI-PLC had no significant effect on 125I-CXCL8 binding to the endothelium, suggesting that the CXCL8 binding sites are not distributed on glypicans. Results of immmunohistochemical analysis showed the presence of glypican-4 in synovial endothelial cells and a lack of detectable glypican-1 and -3. However, antibodies to glypican-4 did not reduce CXCL8 binding to the RA synovium. Together with the PI-PLC data, this suggests that even though endothelial glypican-4 is present, it does not contain CXCL8 binding sites.

In contrast, the use of an antibody to syndecan-3 and the expression data indicated that CXCL8 is interacting with this HSPG on the synovial endothelium. In conjunction with the effects of heparinases, these results suggest that there is induction of a CXCL8 binding motif on the HS chains of syndecan-3 in the RA synovium. A likely explanation for the finding that an antibody to the syndecan-3 core protein inhibits CXCL8 binding is that the antibody may sterically block the interaction between this chemokine and carbohydrate structures located in the adjoining HS chains. Immunohistochemical analysis indicated that syndecan-3 is also present constitutively in the uninflamed synovium, but its HS chains do not appear to bear the CXCL8 binding structures. This suggests that HSPG synthesis by endothelial cells shows specificity and versatility, because these cells can differentially decorate syndecan-3 with a chemokine binding site depending on whether or not they are in an inflammatory environment. The synthetic specificity of chemokine binding sites on HSPGs is further illustrated by syndecan-2. This is one of the HSPGs synthesized by cultured HUVECs (58, 65) that harbor CXCL8 binding structures on their HS chains (16). Syndecan-2 is also a major HSPG expressed by monocyte-derived macrophages yet does not express binding motifs for CXCL8, and other chemokines, on its HS chains (66).

In conclusion, our finding showing the induction of a CXCL8 binding site on endothelial syndecan-3 in RA synovium has implications for leukocyte trafficking into this inflamed tissue. This endothelial site may be involved in several mechanisms, such as presenting and transcytosing chemokines and forming chemokine gradients, leading to enhanced leukocyte recruitment. It is interesting to have identified an HSPG that binds a chemokine in the setting of chronic inflammation, because emerging evidence suggests that syndecans have important functions. Their core proteins can signal, and differences between syndecans appear to exist with regard to their intracellular signaling mechanisms (6). It is also apparent that syndecan core proteins may be involved in forming membrane rafts, concentrating ligands in membrane microdomains and internalizing ligands in vesicles (6, 67–69). Chemokine transcytosis and presentation by endothelial cells involve plasmalemmal vesicles (caveolae) and membrane microdomains (the tips of endothelial microvilli) (13, 31, 70), which suggests a potential role for syndecans in these mechanisms.


We thank Professor John Gallagher (Manchester University) for helpful discussions and the following surgeons and rheumatologists for their assistance: R. Butler, J. Dixey, C. McGeoch, D. Rees, R. Spencer-Jones, R. Wade, S. White, and the Daycase unit (Robert Jones and Agnes Hunt Orthopaedic Hospital). We also acknowledge L. Americh, P. Evans, N. Harness, and M. Pritchard for their expertise in histology.