P-selectin glycoprotein ligand 1 therapy ameliorates established collagen-induced arthritis in DBA/1 mice partly through the suppression of tumour necrosis factor

Abstract

We investigated the therapeutic potential of P-selectin glycoprotein ligand (PSGL)-1 in established collagen-induced arthritis (CIA) in DBA/1 mice. PSGL-1 is the high-affinity specific ligand for P-selectin and is thus important in cell recruitment to inflammatory sites. I-316 PSGL-1 or rPSGL-1Ig fusion protein were administered to mice after the onset of clinical arthritis for 10 days, and the effect of treatment on both clinical and histopathological progression of disease was studied. It was found that both PSGL-1 biologicals effectively suppressed progression of clinical arthritis, and this was accompanied by protection against damage of joint tissues. We sought to investigate a mechanism underlying the effect of rPSGL-1Ig on the reduction of clinical arthritis. Blockade of PSGL-1/P–selectin interaction blocks recruitment of leucocytes, thus we observed a notable reduction in viable cell numbers of synoviocytes from rPSGL-1Ig treated mice. In view of this finding we suspected an effect of treatment on the production of pro-inflammatory mediators such as bioactive tumour necrosis factor-α (TNF) in synovial membrane ex vivo cell cultures. Production of TNF was reduced in arthritic mice that had been treated with rPSGL-1Ig. To further investigate the mechanism of rPSGL-1Ig, we explored the possibility that PSGL-1 might also have a direct signalling effect on TNF release from inflammatory cells. Thus synoviocyte cultures from arthritic mice were incubated with rPSGL-1Ig. A significant reduction in the spontaneous bioactive TNF release from these cultures was noted. We therefore confirmed these surprising findings using cultures of a mouse macrophage like cell line RAW 264·7, stimulated by LPS. Our results indicate that both forms of PSGL-1 have significant therapeutic effects in CIA murine model of RA. The mechanism of action involves reduced cellularity of synovium as anticipated, along with a reduction in TNF production from inflammatory cells in the synovium. The latter mechanism needs further mechanistic analysis.

Introduction

Rheumatoid arthritis (RA) is a chronic polyarticular inflammatory disease, characterized by progressive joint destruction. This destruction is initiated by the hyperplastic synovium, which is infiltrated by inflammatory cells that secrete proinflammatory cytokines such as tumour necrosis factor (TNF)-α and interleukin-1 (IL-1) [1].

Arthritis induced in rodents by the systemic administration of type II collagen (CII) is an experimental model with many resemblances to RA, and is widely used for studying the disease processes (reviewed in [2,3]). Upon immunization with purified CII in complete Freund's adjuvant, genetically susceptible strains of mice develop arthritis, which is characterized by cellular and humoral immune responses to CII [4–6]. The cellular response has been shown to be Th1 driven [7]. Like RA, collagen-induced arthritis (CIA) is characterized by the rapid onset of inflammation of joints followed by the erosion of cartilage and bone. Inflammatory cytokines such as TNF and IL-1 are considered to be the key mediators, and anti-TNF therapies markedly attenuate the clinical severity of CIA [8,9], as they do in RA [10], and blocking IL-1 has also been shown to be effective in controlling disease.

The synovium in CIA is hyperplastic and heavily infiltrated by inflammatory cells, which are the main sources of proinflammatory mediators [3], with polymorphs particularly abundant. As there is a notable reduction in the influx of cells in RA after treatment with anti-TNF [10], we and others hypothesized that treatments aimed at reducing the influx of inflammatory cells into the synovium may be therapeutically beneficial, and this hypothesis could be most readily tested by blocking selectin–ligand interactions in an animal model of RA, namely the collagen-induced arthritis model.

Macrophages, neutrophils, T and B lymphocytes are drawn to the sites of local infection, inflammation and tissue injury by a complex process of extravasation, which involves first tethering, firm adhering and then penetration through cell junctions to the affected tissue site (reviewed in [11]). A number of molecules have been elucidated to play a key role in inflammatory cell recruitment process.

The selectin family of adhesion molecules is crucially responsible for establishing the initial interaction of rapidly moving leucocytes with the vascular endothelium [12,13], initiating ‘rolling’ of the cells over the endothelium. The selectin family has three members. l-selectin is constitutively expressed on neutrophils, monocytes, eosinophils and some subsets of T and B lymphocytes [14]. E-selectin is expressed on endothelial cells, in response to inflammatory signals such as IL-1β, TNF and interferon-gamma (IFN-γ) [12]. P-selectin (CD62) is constitutively expressed on the surface of α-granules of unstimulated platelets and gets rapidly redistributed to the surface upon activation [14,15]. Megakaryocytes and endothelial cells also express P-selectin [16]. All selectins are highly glycosylated proteins containing an N-terminal extracellular Ca⁺⁺ dependent lectin-like domain, a transmembrane domain and a short cytoplasmic tail. They are capable of binding to leucocytes through a restricted number of glycoprotein ligands possessing a mucin-like extracellular domain [17]. P-selectin glycoprotein ligand 1 (PSGL-1, CD 162) is a specific mucin-like ligand for P-selectin [18–20]. PSGL-1 has binding ability to both l-selectin and E-selectin [21,22], but it is the main high-affinity specific ligand for P-selectin [18]. PSGL-1 is distributed on various cell types such as B cells, T cells (γ/δ subset), residential and circulating dendritic cells, CD34⁺ stem cells and various myeloid cells from the bone marrow show a variable level of expression [23–25]. The binding between PSGL-1 and the selectins has been disrupted by the use of antibodies (Ab) raised against the specific selectins or to the ligand [18, 21, 23, 26–29]. It has been well documented in various models of inflammation that blocking the interaction between P-selectin and PSGL-1 inhibits leucocytes from entering the site of an inflammatory immune response (reviewed in [30]). Moreover, it has been suggested that the expression of active PSGL-1 is crucially relevant for extravasation of Th1 cells from the peripheral circulation to sites of inflammation [31]. This prompted us to undertake the present study where we initially investigated the therapeutic potential of blocking PSGl−1 selectin interactions in CIA, a murine Th1 type disease model of arthritis. We have also initiated analysis of the possible mechanism of action by PSGL-1 inhibiting therapeutics in CIA.

Materials And Methods

Induction and assessment of arthritis

CII was extracted and purified from bovine articular cartilage as described [32]. Briefly, cartilage was dissected from the knee joint of a young calf, powdered in a liquid nitrogen mill and then treated with 4 m guanidinium hydrochloride to remove proteoglycans. CII was extracted from the precipitate by pepsin digestion (1 mg/ml) in 0·5 m acetic acid and then removed from solution by salt precipitation. The collagen was resuspended in a solution of Tris HCl (0·05 m, pH 7·4) containing NaCl (0·2 m) to neutralize any residual pepsin activity, exhaustively dialysed against 0·5 m acetic acid, aliquoted and stored at −70°C until use.

Male DBA/1 J mice (H-2^q) (Harlan Laboratories UK, Oxford, UK) were injected intradermally at the base of their tails with a single injection of 100 µg of CII, emulsified in complete Freund's adjuvant (Difco, Detroit, MI, USA) at 10 weeks of age.

Animals were monitored for signs of arthritis 14 days post immunization on a daily basis and a clinical scoring system was adopted wherein each limb was given a clinical score on the basis of visual signs of oedema and/or erythema as follows: 0 = normal, 1 = slight swelling and/or erythema at the base of the paw, 2 = frank oedema and erythema involving the entire paw, 2·5 = pronounced oedema and erythema leading to an incapacitated limb mobility, 3 = ankylosis, as previously described [33]. Each limb was graded in this way, giving a maximum possible score of 12 per mouse. The hind paws were also measured daily, for paw swelling with the help of a calliper (least detectable change = 0·1 mm, Kroeplin, Schluchtern, Germany). All animal studies conducted had received prior approval of the local ethical review process committee and procedures were performed under the Home Office Guidance on the Operation of the Animals (Scientific Procedures) Act 1986 (PPL: 70/5446).

Histopathology

Arthritic knees and feet from day 10 animals were dissected post mortem, fixed in 10% formalized saline and decalcified in buffered formalin containing 5·5% EDTA. The specimens were embedded in paraffin wax, sectioned on a microtome to 5 µm thickness and stained with haematoxylin and eosin stains for microscopic evaluation, which was performed by an observer blinded to the treatment received. Each section observed was given a score of normal = 0; mild = 1 (minimal synovitis, cartilage loss and bone erosions limited to discrete foci); moderate = 2 (synovitis with considerable bone and cartilage erosions); severe = 3 (loss of joint architecture), as previously described [33].

Experimental treatment regimens with PSGL-1 proteins

Both I-316 PSGL-1 and rPSGL-1Ig fusion protein were kind gifts of Dr Glenn Larsen, Dr Robert Schaub and their colleagues at Genetics Institute, Inc., Cambridge, Massachusetts, USA. Both proteins were recombinant forms of PSGL-1 produced in Chinese hamster ovary (CHO) cells. I-316 PSGL-1 was a dimer molecule containing the full length extracellular domain of native PSGL-1 molecule terminating at isoleucine 316, whereas rPSGL-1Ig fusion protein contained the first 47 amino acids of the extracellular domain of native PSGL-1 molecule fused to human IgG1 Fc portion that was mutated to reduce complement and Fc receptor binding capabilities. Both recombinant proteins were around 10⁵ kD.

Treatment commenced at the first signs of arthritis and was continued daily for a period of 10 days, which was the entire period of the acute phase of arthritis. The treatments were administered i.p. at the concentrations of 7·5 µg/mouse/day for I-316 PSGL-1 (350 µg/kg body weight, daily) and 200 µg/mouse/day for rPSGL-1Ig fusion protein (10 mg/kg body weight, daily).

Synovial cell cultures

Synovial cells were isolated from synovial membranes obtained from the inflamed knee joints of day 10 arthritic animals as previously described [33]. Briefly, knee joints were removed, and each synovial membrane was excised under a dissecting microscope and digested with collagenase A (1 mg/ml; Boehringer-Mannheim UK, Lewes, UK) and DNAse type IV (150 µg/ml; Sigma, Poole, UK) at 37°C for 60 min in the presence of polymyxin B (33 µg/ml; Sigma). The cells were washed extensively, counted and cultured at a density of 2–4 × 10⁶ cells/ml in complete RPMI-1640 (PAA Laboratories, Somerset, UK) containing penicillin (100 U/ml), streptomycin (100 µg/ml), 2-mercaptoethanol (5 × 10⁻⁵ M), l-glutamine (2 mm) and 10% heat-inactivated foetal calf serum (FCS) in the absence or presence of rPSGL-1 Ig fusion protein. After 24 h, the supernatants were harvested and stored frozen at −20°C until being assayed for levels of bioactive TNF.

RAW 264·7 cell cultures

RAW 264·7 cells, a monocytic-macrophage cell line derived from BALB/c mice [34], was cultured in DMEM (PAA Laboratories) containing 1% FCS. The cells (2 × 10⁵/100 µl) were plated in a 96-well flat-bottomed plate (Falcon Microtest 96, Becton Dickinson Labware Europe, Cedex, France) in the presence or absence of graded concentrations of rPSGL-1IgG fusion protein (100–1000 µg/ml) for 18–20 h in a humidified 5% CO₂ incubator at 37°C. The cells were then stimulated with LPS (1 µg/ml; E. coli 026:B6; Sigma) for eight hours. Supernatants were harvested, aliquoted and frozen at −20°C till measurement of TNF levels. MTT (3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazoliumbromide; Thiazolyl blue) assay (Sigma) was always performed at the end of the experiment to check for cytotoxicity. Formation of tetrazolium blue crystals, dissolved by the addition of 10% SDS in 0·01 m HCl gave a deep violet colour that was indicative of cell viability. The plates were read on a Labsystems Multiskan Bichromatic spectrophotometer (Thermo Life Sciences, Hampshire, UK) at λ_max = 574 nm.

Tumour necrosis factor bioassay

The bioactive TNF levels from synovial membrane cell cultures were assayed using a WEHI-164 cell line [35], as previously outlined by Baker et al. [36].

Measurement of murine TNF by ELISA

The murine TNF ELISA was performed as described previously [36]. The coating antibody was an in-house capture rat anti-mouse TNF monoclonal antibody purified from the supernatant of a hybridoma cell line HB10697 and was used at a concentration of 8 µg/ml at 4°C and left overnight. The plates were blocked, and then washed with washing buffer (PBS-0·05% Tween-20) followed by addition of standards and unknowns and a further overnight incubation at 4°C. The detect antibody was biotinylated rabbit anti-mouse/rat TNF (Pharmingen, USA), which was detected with the help of horseradish peroxidase conjugated streptavidin (Amersham, Life Sciences, UK). TMB substrate system (KPL, MD, USA) was used to develop the colour and the plates were read at 450 nm. TNF content in supernatants of unknown samples was extrapolated from a standard curve (10 000–4 pg/ml) generated with murine recombinant TNF protein (Pharmingen).

Statistical analysis

Statistical analyses were all performed using the Graph Pad Prism, version 3 software package (GraphPad Software, San Diego, California, USA). The nonparametric Mann–Whitney-U-test was applied for the comparison of the plots of PSGL-1 treated versus vehicle treated groups for both clinical score, as well as paw swelling. P-values of < 0·05 or 0·005 were considered to be statistically significant. The χ² test was applied to evaluate the statistical significance of all histology data. In vitro parametrically distributed data was analysed using the unpaired, two-tailed, Student's t-test with 95% confidence intervals applied.

Results

Clinical severity of established CIA is attenuated after treatment with PSGL-1

I-316 PSGL-1 was administered i.p. daily at a dose of 7·5 µg from day one of the onset of disease for a period of 10 days. The clinical progression of CIA was arrested significantly upon treatment of mice with I-316 PSGL-1 (Fig. 1a). The amelioration of disease was significant at all time points when compared to PBS-treated controls. Reduction in the clinical score was closely mirrored in the significant reduction seen in paw swelling throughout the 10-day period (Fig. 1b).

Fig. 1.

Upon onset of the first clinical signs of arthritis mice were treated with I-316 PSGL-1, intraperitoneally at 7·5 µg/mouse, daily. Each point on the graph represents the mean ± s.e.m. of six mice. (a). Clinical score over a 10-day period of acute CIA in which I-316 PSGL-1 treated mice (○) were compared to the vehicle (sterile PBS) treated control group of mice (•). (b). Paw swelling measurements (mm) over the 10-day period of acute CIA for I-316 PSGL-1 (○) versus PBS control treated mice (•). *denotes a P-value of < 0·05; **denotes a P-value of < 0·005; NS, not significant.

However, the production of I-316 PSGL-1 in large quantities was not feasible, and as further supplies were not available to us it was therefore decided to continue the study with rPSGL-1Ig fusion protein. It was found that treatment of arthritic mice with the fusion protein also resulted in a reduction in the clinical severity of the disease (Fig. 2a). The reduction seen in the clinical scores was highly significant throughout the 10-day treatment period and was mirrored in the diminished paw swelling of the hind paws of treated animals during that period (Fig. 2b).

Fig. 2.

Upon onset of the first clinical signs of arthritis mice were treated with rPSGL-1Ig fusion protein, intraperitoneally at 200 µg/mouse, daily. Each point on the graph represents the mean ± s.e.m. of 28 mice in the vehicle treated control group and for 27 mice in the fusion protein treated group. (a). Clinical score over a 10-day period of acute CIA in which fusion protein treated mice (○) were compared to the vehicle (sterile PBS) treated control group of mice (•). (b). Paw swelling measurements (mm) over the 10-day period of acute CIA for fusion protein (○) versus PBS control treated mice (•). *denotes a P-value of < 0·05; **denotes a P-value of < 0·005; ***denotes a P-value of < 0·0001; NS, not significant. Data shown in this figure represents pooled results of three individual treatment experiments. Individual treatment experiments gave comparable treatment profiles.

PSGL-1 treatment protects against pathological damage of arthritic knee and foot joints

Microscopic analysis of the paw-sections of I-316 PSGL-1 treated mice, by an observer blinded to the treatments received revealed a reduction in the percentage of severely affected joints, i.e. 8% as compared to 25% in the PBS-treated control group. Eighty-four percent of normal to mildly affected paws were reported in mice treated with I-316 PSGL-1 as compared to 50% in the control group [Table 1].

Table 1.

Histology of foot sections of I-316 PSGL-1 treated mice as compared to vehicle treated mice

Arthritic changes	Control treated (n = 12)	I-316 PSGL-1 treated (n = 12)
Normal	16% (2/12)	34% (4/12)*
Mild	34% (4/12)	50% (6/12)*
Moderate	25% (3/12)	8% (1/12)*
Severe	25% (3/12)	8% (1/12)*

Histological findings at the end of the 10-day treatment period in CIA with I-316 PSGL-1 (six mice in group) and PBS as the control treated group (six mice in group). Arthritic joint changes were scored by an observer blinded to the treatments received; and were graded as normal, mild, moderate or severe changes to the degree of destruction of the joint architecture. The results are shown as the percentage of all arthritic feet studied (n = 12) that were graded as a specific histology score as outlined in Materials and Methods.

^*P = 0·0171 (χ² test) when arthritic changes of normal and mild were compared to moderate and severe amongst control versus I-316 PSGL-1 treated mice.

Similarly, histological analysis of paw-sections from rPSGL-1Ig fusion protein treated mice showed a significant reduction in the percentage of severely affected joints from 80% in the control group to only 11% in the fusion protein treated group (Table 2). The percentage of affected joints exhibiting normal to mild and moderate structural changes was just 20% in the controls as compared to 89% in the fusion protein treated group (Table 2). Analysis of histological changes in knee joints from rPSGL-1Ig fusion protein treated mice showed an increase to 45% of joints with normal joint architecture, compared to 5% in the control group (Table 2). A reduction in the percentage of severely affected knees was also observed, from 35% in the PBS-treated group to 22% in the rPSGL-1Ig fusion protein treated group (Table 2).

Table 2.

Histology of foot and knee sections of rPSGL-1Ig fusion protein treated mice as compared to vehicle treated mice

Arthritic changes	Control treated (n = 20)	rPSGL-1Ig treated (n = 18)
Foot sections
Normal	10% (2/20)	28% (5/18)
Mild	5% (1/20)	22% (4/18)
Moderate	5% (1/20)	39% (7/18)
Severe	80% (16/20)	11% (2/18)*
Knee sections
Normal	5% (1/20)	45% (8/18)†*
Mild	10% (2/20)	5% (1/18)†
Moderate	50% (10/20)	28% (5/18)†
Severe	35% (7/20)	22% (4/18)†*

Histological findings at the end of the 10-day treatment period in CIA with rPSGL-1Ig fusion protein. Arthritic changes to the knee joint or foot were scored by an observer blinded to the treatments received; and were graded as normal, mild, moderate or severe changes to the degree of destruction of the joint architecture. The results are shown as the percentage of all arthritic knees or feet studied (n = 20 or 18).

^*P = 0·0171 (χ² test) when arthritic changes of normal, mild and moderate were compared to severe changes of feet amongst control versus rPSGL-1Ig treated mice. Arthritic changes of normal and mild were compared to moderate and severe changes of knees amongst control versus rPSGL-1Ig treated mice

^†(P = 0·0205), while normal versus severe changes amongst the two treatment groups had a ^*P-value of 0·0171 by the χ² test.

rPSGL-1Ig fusion protein treatment reduces cellular influx to arthritic joints and suppresses bioactive TNF release by synoviocytes upon ex vivo and in vitro culturing

Treatment of arthritic mice with rPSGL-1Ig fusion protein (10 mg/kg body weight, a single injection given daily for three days, i.p) resulted in a notable reduced cellular influx to arthritic joints (Table 3). As seen from the table, a high degree of biological variability was recorded in the numbers of viable cells isolated from individual knee synovial membranes of mice belonging to various treated and untreated groups. However, there was a notable trend towards reduced cellularity from membranes of rPSGL-1Ig treated mice. On average, untreated arthritic mice and vehicle (PBS) treated controls had 4·2 and 6·5 × 10⁵ cells/membrane, respectively. In contrast, rPSGL-1Ig fusion protein treated mice had 3·6 × 10⁵ cells/membrane (Table 3). Thus, an average reduction of 29·6% was noted in the cellular influx to arthritic knee joints upon rPSGL-1Ig fusion protein treatment for three days when compared to untreated and vehicle treated control groups.

Table 3.

Treatment of arthritic mice with rPSGL-1Ig fusion protein reduces joint synovial membrane cellularity

	Synovial membrane viable cell count (× 105/membrane)							Mean ± s.e.m.
Untreated	1·4	2·6	3·2	5·9	5·9	6·2	…	4·2 ± 0·8
Vehicle (PBS) treated	2·1	4·1	4·3	4·8	5·7	7·2	17·5	6·5 ± 1·9
rPSGL-1Ig treated	0·8	1·7	3·8	4·6	4·7	4·9	5·2	3·7 ± 0·7

Synovial membranes from individual mice belonging to different groups were dissected post mortem on day three of arthritis. Single cell suspensions were prepared from the pooled synovial membranes isolated from each individual mouse knees as described earlier in Materials and methods. Trypan blue dye exclusion test (Sigma) was performed to obtain the viable cell count/knee (membrane). Data shown in the table above represents the individual viable cell count per membrane of each individual mouse in the different groups, as well as the mean value ± standard error.

Reduced joint cellularity prompted us to further investigate the production of a key proinflammatory cytokine, TNF by the isolated synovial membrane cells upon ex vivo culturing in order to further define a possible mechanism of how rPSGL-1Ig fusion protein acts therapeutically in our CIA model (Fig. 2).

It has already been documented that knee synovial cells freshly isolated from arthritic mice spontaneously produce detectable levels of bioactive TNF upon culturing in vitro[33], as do human RA synovial membrane cells [1]. Synovial cells were thus isolated from the knee joints of control untreated arthritic mice (n = 6), vehicle (PBS)-treated arthritic mice (n = 7), or rPSGL-1Ig fusion protein treated arthritic mice (10 mg/kg body weight, daily, i.p., n = 7) at day 3 of clinical arthritis, and cell density equalized to 2 × 10⁵ cells/well were plated and cultured for 24 h. TNF-production was measured by WEHI bioassay at the end of the culture period from harvested supernatants. It was found that synovial cells taken from rPSGL-1Ig fusion protein treated mice produced less TNF (median = 2·950 U/ml) than synoviocytes from either untreated controls (median = 4·676 U/ml) or vehicle-treated mice (median = 5·169 U/ml) (Fig. 3). Although the decrease in TNF production in these cultures, which had been normalized to same cell density due to the variability, was not statistically significant, this finding is suggestive of a biologically significant reduction of TNF production levels in the knee joints of rPSGL-1Ig fusion protein treated mice. This is because these mice also have a reduced cellularity of the synovia of the knee joints (Table 3). Hence in situ, at the site of the inflammatory hyperplastic joint synovium the differences in bioactive TNF output/cell would be magnified.

Fig. 3.

Bioactive TNF production (U/ml) by synovial membrane cells isolated from collagen-induced arthritic mice on day 3 of clinical arthritis. Untreated controls (▪, n = 6) Vehicle (PBS) treated controls (▴, n = 7) rPSGL-1Ig fusion protein treated (200 µg, daily, i.p., •, n = 7)

We became interested in further investigating the mechanism of the noted TNF reduction from ex vivo synovial membrane cell cultures. One hypothesis is that recombinant PSGL-1 fusion protein can directly interact with selectin receptors on the surface of synoviocytes and modulate signalling of TNF production. Thus rPSGL-1Ig fusion protein (10–30 µg/ml) was added in vitro to synovial cells taken from day-10 untreated control arthritic mice. This resulted in a modest, although significant concentration-dependent suppression of bioactive TNF release (Fig. 4). As seen from Fig. 4 modest levels of bioactive TNF (0·69 U/ml) were produced by 2 × 10⁵ cells/well. This was reduced to 0·39 U/ml upon in vitro treatment with the fusion protein at the maximal concentration of 30 µg/ml. Thus a statistically significant inhibition (43·5%, P < 0·05, Student's t-test, unpaired, two-tailed, with 95% confidence intervals applied) of bioactive TNF release was achieved by direct selectin-ligand modulation in vitro.

Fig. 4.

Synovial membrane cells from mice (n = 5) with CIA were pooled together on day 10 of disease. In vitro addition of rPSGL-1Ig fusion protein (10–30 µg/ml) produced a concentration-dependent suppression in the release of bioactive TNF. Each point on the graph (○) represents the mean ± s.d. values of triplicate cell cultures. *P < 0·05, when compared to control cell cultures producing spontaneous levels of bioactive TNF. Data has been analysed by Student's t-test, unpaired, two-tailed, with 95% confidence intervals applied (Graph Pad software, version 3).

rPSGL-1Ig fusion protein blocks TNF release by RAW 264·7 cell line

In order to explore this hypothesis further and to confirm our earlier observations with synoviocytes we performed experiments with a murine monocytic-macrophage cell line (RAW 264·7). Cells were cultured and stimulated with LPS in the presence of a whole range of concentrations of rPSGL-1 Ig fusion protein (10–1000 µg/ml). Low concentrations of fusion protein (10–100 µg/ml) failed to show any inhibition of TNF output, though higher concentrations (≥100–1000 µg/ml) resulted in a concentration-dependent inhibition of LPS-induced TNF-release (Fig. 5a). Maximum inhibition (78%) was seen at a concentration of 600 µg/ml. PSGL-1 fusion protein was found not to be cytotoxic at these high concentrations, as revealed by the MTT assay (Fig. 5b).

Fig. 5.

(a). Concentration-dependent suppression in the release of TNF levels from cultured RAW 264·7 cells by rPSGL-1Ig fusion protein (100–1000 µg/ml). TNF levels were measured by a sandwich ELISA. Each point (○) represents the mean value of triplicate cultures ± s.e.m. (b). Cell viability was estimated by performing an MTT assay on the cultures. Each point (○) represents the mean value of triplicate cultures ± s.d.

Discussion

The interaction of selectins with their ligands is one of the early steps involved in leucocyte extravasation, promoting the ‘rolling’ phenomenon described by Springer and his colleagues [11,30]. Therefore, therapeutics that may block selectin–ligand interaction/function have been sought as they should inhibit inflammatory cell recruitment. The cloning of a major ligand for the selectins, PSGL-1 [20], has made it possible to generate a number of therapeutics based on soluble PSGL-1 aimed at disrupting this interaction, and these have been successfully tested in vitro and in animal models of inflammation/injury [26, 37, 38].

Two variants of soluble PSGL-1 were tested in our studies. I-316 PSGL-1 is a natural dimer, highly glycosylated, whereas rPSGL-1Ig is a less glycosylated IgG Fc dimer that is longer lived in circulation. Using the DBA/1 mouse collagen-induced arthritis model, we showed that both PSGL-1 therapeutics were effective. The amelioration seen in the clinical signs of established arthritis was more pronounced upon I-316 PSGL-1 treatment, but as this experiment was performed only once (due to operational difficulties involving synthesis of the natural dimer and hence its limited supplies), it is not possible to draw conclusions about its relative efficacy in our CIA animal model when compared to rPSGL-1Ig which was used in three discrete experiments. However, the amelioration of disease signs attained statistical significance and in light of our results these experiments support the concept that selectin interactions with their ligands are important in the arthritic process, and hence the selectin–PSGl−1 interaction is a legitimate therapeutic target in the arthritic pathology in the mouse, and thus also potentially in humans.

All aspects of the disease were ameliorated as judged by histological analysis of the joints; there was less leucocyte infiltration and less damage to cartilage and bone. Our results in this model are consistent with improvement in joint histology detected with other successful treatments, such as the use of anti-TNF biologicals [10].

We sought to analyse the mechanism by which amelioration of the disease process took place. As expected, recruitment of leucocytes was diminished, as there was a reduction in the absolute number of cells in the joint, with an average of 360 000 per knee in rPSGL-1Ig treated mice compared to 540 000 in untreated/vehicle treated mice. This reduction is only partial, but is compatible with reported data in PSGL-1 knockout mice where there is still some extravasation into inflammatory sites [39]. We believe that this is the likely major and predictable mechanism of benefit by PSGL-1 therapeutics, with less recruited cells causing less damage.

We also measured bioactive TNF production by synovial cells from knee joints of mice treated with rPSGL-1Ig and compared it to levels seen in control untreated and vehicle treated mice. There was a reduction seen in the levels of bioactive TNF production by synovial cells, adjusted to the same in vitro cell density, but due to the wide scatter seen in the median-distribution of TNF levels from individual mice, this did not reach statistical significance. The reason why there was less bioactive TNF produced per cell in rPSGL-1Ig treated mice versus untreated mice was investigated. Possibly less activated cells enter the joints, or else in the presence of PSGL-1 there is localized reduction in the activation process of cells extravasating in the inflamed joints from the surrounding vasculature. The former hypothesis was not tested by us however; the latter hypothesis is supported in part by our studies in which freshly isolated synovial cells from inflamed knee joints of untreated CIA mice were treated with rPSGL-1Ig in vitro. There was a modest reduction (about 44%) in the spontaneous TNF release, suggesting that unexpectedly, PSGL-1 may be directly involved in cell activation.

To explore further whether rPSGL-1Ig acts directly on cells we used RAW 264·7, a murine monocytic-macrophage cell line for in vitro studies. Production of TNF by RAW 264·7 cells after LPS stimulation was reduced in a concentration-dependent manner upon in vitro addition of rPSGL-1Ig. In this experiment we found that only higher concentrations of rPSGL-1Ig fusion protein inhibited TNF release, concentrations not reached in our in vivo experiments. The results thus obtained are not conclusive. There is a possibility that selectin–PSGl−1 interactions may be important activating signals, or that selectin–PSGl−1 interaction facilitates cell clustering, thereby augmenting the transmission of other cell signals mediated through costimulatory molecules, in turn leading to down-stream activation of inflammatory cytokine production. At present it is evident that blocking PSGL-1 reduces cellularity, presumably by reducing cell ingress. The mechanism of the reduced TNF production noted by us is not yet understood.

Overall, the present results in our CIA model are analogous to those previously reported where a P-selectin ligand analogue (sPSGL-1) arrested damage caused to the kidney in an ischemic/reperfusion injury model in rats. Treatment with the ligand analogue successfully arrested the inflammatory response associated with enhanced leucocyte trafficking, as well as reduced the levels of expression of class II, selectin molecules and TNF mRNA [38].

These studies emphasize the importance of blocking leucocyte trafficking to inflammatory sites, such as arthritic joints, which is one of the major ways in which anti-TNF therapy works [1]. It also suggests that blocking the selectin–PSGl−1 interaction is a useful therapeutic target, but it is also known that blocking leucocyte recruitment also involves abolishing integrin–mediated interactions [40]. Indeed, monoclonal antibodies directed against αVβ4 integrin (e.g. Natalizumab) have been shown to be beneficial in Crohn's disease and multiple sclerosis, and thus the exact method of blocking leucocyte recruitment in inflammatory disorders by interfering with either selectin mediated or integrin mediated interactions may not matter much, as long as such a recruitment is effectively diminished by therapeutic intervention.