Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The cytokine TNF induces the production of proinflammatory cytokines, prostaglandins, active oxygen species, nitric oxide, and MMPs (Tracy and Cerami, 1993; Beutler, 1995), thereby implicating TNF as a therapeutic target for treating the pain, swelling, and progressive joint destruction caused by rheumatoid arthritis. Recently, an anti-TNF antibody (Maini et al., 1998) and a soluble form of the 75-kDa TNF receptor (Moreland et al., 1997; Weinblatt et al., 1999) have been shown to decrease the symptoms and joint destruction in rheumatoid arthritis patients, thus validating TNF as a target in this disease.

An alternative to neutralization of TNF by binding proteins is to inhibit its production. TNF is made as a 26-kDa precursor form that is membrane bound. When displayed on the surface of cells, this precursor has biological activity and can bind to 55- and 75-kDa TNF receptors on nearby cells (Decoster et al., 1995). However, for TNF to be released into the extracellular space, the 26-kDa form of TNF must be specifically cleaved to its 19-kDa form. The TACE responsible for this cleavage has been purified and cloned, and an active human recombinant form has been expressed (Black et al., 1997; Moss et al., 1997). A subset of hydroxamate MMP inhibitors inhibit TACE and block the release of TNF from inflammatory cells in vitro with little or no effect on the release of other cytokines such as IL-6, IL-8, IL-1, and lymphotoxin- (Gearing et al., 1994; McGeehan et al., 1994; Mohler et al., 1994; DiMartino et al., 1997). In studies in vivo, dual MMP/TACE inhibitors block the increase in plasma TNF seen 90 min after administration of LPS and prevent endotoxic death in the mice at 24 h (Mohler et al., 1994; Moss et al., 1999). However, early TACE inhibitors have limited duration of action in vivo, making it difficult to evaluate the effect of TNF convertase inhibition in chronic inflammation models.

MMPs fall into three classes based on their substrate specificity. Interstitial collagenase (MMP-1), collagenase 3 (MMP-13), and neutrophil collagenase (MMP-8) are the only enzymes known to degrade collagen types I and II at neutral pH. Gelatinases (MMP-2 and -9) degrade collagen types IV and V and denatured protein. Gelatinases work in concert with collagenases, in that collagen type I fragmented by collagenase can unwind and become a substrate for gelatinase. Stromelysin (MMP-3) degrades fibronectin, glycoproteins, and proteoglycans, key components of cartilage. MMP inhibitors with little or no activity against TACE have been delivered to arthritic rats, resulting in significant, but not complete, inhibition of inflammation and connective tissue destruction (Conway et al., 1995; DiMartino et al., 1997; Lewis et al., 1997). The joint destruction resistant to MMP inhibitors could be due to cysteine proteases. For example, activated osteoclasts have degradative pathways using both cysteine proteinases (Esser et al., 1994; Votta et al., 1997; Everts et al., 1998) and MMPs (Hill et al., 1994).

The fact that TACE is inhibited by a subset of MMP inhibitors opened the possibility that long-acting dual inhibitors of TACE and MMPs could be discovered and that these inhibitors might dampen cell activation and the induction of degradative enzymes as well as directly inhibit MMPs, thus decreasing joint destruction in rheumatoid arthritis. In this article, we describe the ability of GW3333 [(2R,3S)-3-(formyl-hydroxyamino)-2-(2-methyl-1-propyl)-4-methylpentanoic acid [(1S,2S)-2-methyl-1-(2-pyridylcarbamoyl)-1-butyl]amide], a dual inhibitor of TACE and MMPs, to completely inhibit the release of TNF in vivo for up to 12 h after an oral dose. The efficacy of GW3333 is then compared with an anti-TNF antibody in chronic models of rat arthritis.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Animals and Compound Dosing. Male Lewis rats (Charles River Laboratories, Raleigh, NC) weighing about 220 g were used. Animals were free of pathogenic viruses as determined by a standard viral titer screen (Microbiological Associates, Bethesda, MD). The research complied with national legislation and with company policy on the care and use of animals and with related codes of practice.

Enzymes Assays. Recombinant catalytic domains of MMP-1, MMP-13, proMMP-8, and full-length active MMP-3 were expressed and purified from Escherichia coli. Enzymes were refolded in 200 mM NaCl, 50 mM Tris, 5 mM CaCl₂, 10 µM ZnSO₄, and 0.01% Brij 35, pH 7.6, for 1 h prior to the assay. Full-length proMMP-2 and the catalytic domain of proMMP-9 were purified from the media of baculovirus-infected Trichoplusia ni cells. Rat collagenase was a gift from Dr. John Jeffrey (Department of Medicine, Albany Medical College, Albany, NY). The proforms of rat collagenase, MMP-2, MMP-8, and MMP-9 were activated by incubation with 1 mM p-aminophenylmercuric acetate for 2 to 10 h. The p-aminophenylmercuric acetate was removed with a Bio-Spin 6 (Bio-Rad Laboratories, Hercules, CA) chromatography column equilibrated in assay buffer. Enzyme assays were run in a total volume of 0.180 ml of assay buffer containing 200 mM NaCl, 50 mM Tris, 5 mM CaCl₂, 10 µM ZnSO₄, and 0.01% Brij 35, pH 7.6. MMP-1, MMP-13, MMP-3, and MMP-9 concentrations were adjusted to 0.5, 0.05, 5, and 0.1 nM, respectively. MMP-2, MMP-8, and rat collagenase were assayed at 0.3 nM. Enzymes were preincubated with inhibitor for 20 min at room temperature, and the reactions were initiated with the addition of the fluorogenic substrate, Dnp-Pro-Cha-Gly-Cys(Me)-His-Ala-Lys(Nma)-NH₂ (Bickett et al., 1993). Dose responses were generated using 11 3-fold serial dilutions of the inhibitor. Product was measured using excitation and emission wavelengths of 343 and 450 nm, respectively.

Cell-Based Assays. Inhibition of lipopolysaccharide (LPS)-induced TNF release was measured in human Mono Mac-6 cells and human peripheral blood mononuclear cells (PBMC). Mono Mac-6 cells (Ziegler-Heitbrook et al., 1998) in RPMI 1640 media with 10% fetal bovine serum were preincubated for 10 min with GW3333 and then stimulated with 10 ng/ml phorbol 12-myristate 13-acetate (Sigma, #P-8139) and 30 ng/ml LPS (Sigma, # L-2630) and TNF measured in the media at 2 h by ELISA kit (#DTA50, R & D Systems, Minneapolis, MN). Human PBMCs were isolated from heparinized blood by centrifugation through cell separation media (#AN221710, Accurate Chemical and Scientific, Westbury, NY). PBMCs were suspended in RPMI 1640 medium containing 10% heat-inactivated fetal bovine serum, and 2 × 10⁵ cells were added to 96-well microliter plates in a volume of 200 µl per well. GW3333 was added to the cells 15 min before the addition of LPS (1 ng/ml, Sigma, #L-6386). After incubation for 20 h, TNF, IL-6, IL-8, and IL-1 were measured in the media using ELISA kits (R & D Systems).

Bovine Cartilage Explant Assay. Bovine nasal cartilage was cut into plugs of about 10 mg, and individual plugs were cultured overnight in 700 µl of Dulbecco's modified Eagle's medium (#11960044, Invitrogen, Carlsbad, CA) in 24-well plates. The medium was then changed to one containing GW3333, and degradation was initiated with either IL-1 (#NON0081, BioSource International, Camarillo, CA) alone or in combination with human plasminogen (#400, American Diagnostica, Greenwich, CT). GW3333 was added to media as a dimethyl sulfoxide solution, giving a final dimethyl sulfoxide concentration of 0.1%. Each plate included groups (n = 4) of medium only, degradation stimulants only, and stimulant(s) with four concentrations of GW3333. Once degradation of the plugs could be visualized, the plugs were assayed for remaining hydroxyproline (Edwards and O'Brien, 1980).

GW3333 Plasma Concentration and Plasma Protein Binding Determination. GW3333 concentrations were measured using high-performance liquid chromatography with mass spectrometric detection (HPLC-MS). Plasma was extracted with 4 volumes of acetonitrile that contained an internal standard. After centrifugation, the supernatant was evaporated to dryness and then dissolved in HPLC mobile phase. GW3333 was eluted from an HPLC column (Hypersil C₁₈, 45 × 150 mm, 5 µm) using a mobile phase consisting of 10 mM ammonium acetate (pH 4.5) and acetonitrile. GW3333 parent ion (M + H = 421 m/z) was measured using a Finnigan 700 MS after ionization by atmospheric pressure chemical ionization (ThermoFinnigan, San Jose, CA). Plasma standards were prepared by adding GW3333 to blank plasma. HPLC-MS responses were used to construct a peak area ratio (GW3333 peak area/internal standard peak area) versus concentration calibration curve.

TNF Production in Rats in Vivo. LPS (Sigma) was dissolved in PBS at 80 µg/ml and a dose of 40 µg/rat given i.v. in 0.5 ml. Plasma was prepared 90 min after LPS treatment and TNF measured by ELISA (#KRC3012, BioSource International). GW3333 was dosed 15 min before the LPS unless noted otherwise.

Peptidogylcan Polysaccharide Polymers (PGPS) Arthritis. Rats were primed with an intra-articular injection of 10 µl of PGPS at 0.5 mg/ml of rhamnose in the right ankle (Schwab et al., 1993). At 2 weeks the ankle diameters were measured with calipers and rats assigned to groups of n = 9 to get a similar distribution of initial joint diameters. Rats then received their first dose of GW3333 followed 1 h later by an i.v. injection of 0.5 ml of PGPS (0.4 mg/ml of rhamnose) in the tail vein. GW3333 was dosed b.i.d. and ankle diameter and body weights measured for 3 days. On the morning of day 3, rats (n = 3) from each group were sacrificed and plasma prepared to assess GW3333 levels 12 h after the previous dose. At the same time, rats (n = 6) from each group were dosed again with GW3333 and sacrificed (n = 3) 1 and 6 h later for plasma measurements of GW3333.

Adjuvant Arthritis. Freund's complete adjuvant was injected intradermally in the base of the tail (Conway et al., 1995) and on day 7 the rats were sorted into groups of eight by body weight. GW3333 was given orally b.i.d. from day 7 to 21. The anti-TNF antibody and the IgG control antibody were administered i.v. on days 7, 11, and 17. Body weight and the diameter of the left ankle were measured every 2 to 3 days. On the morning of day 21, rats (n = 4) from each group were sacrificed and plasma prepared to determine GW3333 levels 12 h after the last dose. Other rats (n = 4) were dosed again and sacrificed 2 h later for assay of GW3333 in the plasma. At time of sacrifice, the left ankles were fixed in 10% buffered formalin, and one section per ankle was taken and stained with hematoxylin and eosin for histological analysis. The right ankles were used for microradiographic evaluation.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Activities in Vitro. GW3333 (Fig. 1) inhibited human TACE and MMP enzyme activity in vitro at low nanomolar concentrations (Table 1). GW3333 also inhibited rat collagenase in the same concentration range. GW3333 completely inhibited LPS-induced release of TNF from human Mono Mac-6 cells and PBMCs giving IC₅₀ values of 0.17 µM (n = 2) and 0.97 ± 0.13 µM (n = 3), respectively. In two of three PBMC preparations, LPS-induced release of IL-6, IL-8, and IL-1 was also measured. Both preparations yielded similar results. As shown in Fig. 2, GW3333 had no effect on the production of these cytokines at concentrations up to 20 µM.

View larger version (9K): [in this window] [in a new window]   Fig. 1.   Structure of GW3333.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Effect of GW3333 on LPS-induced cytokine production in human PBMCs. Media was taken 20 h after LPS stimulation, and cytokines were measured by ELISA kits. In the absence of LPS, cytokines were undetectable (data not shown). Data from one preparation of PBMCs. , TNF; , IL-1; , IL-6; ×, IL-8.

View this table: [in this window] [in a new window]   TABLE 2 Inhibition of IL-1-induced degradation of bovine nasal cartilage plugs in vitro

Inhibition of TNF Production in Vivo. Injection of LPS into rats caused plasma TNF to increase in a linear manner from 30 to 90 min, followed by a decrease in TNF over time (data not shown). Therefore TNF was measured at 90 min and the plasma concentrations of GW3333 determined 30 and 90 min after LPS (Table 3). GW3333 completely inhibited TNF production in vivo, with an IC₅₀ of about 1 mg/kg. The plasma drug concentrations showed that GW3333 was active at nanomolar levels in vivo. For example, the 3.33-mg/kg dose inhibited TNF production by 88% with 451 and 146 nM GW3333 in the plasma at 30 and 90 min after LPS. Protein binding in vitro indicated that 49% of GW3333 was free in rat plasma.

Activity in PGPS Rat Arthritis Model. In the PGPS reactivation model, the i.v. reactivation dose of PGPS causes massive T-cell activation and infiltration into the previously primed joint, resulting in a peak of ankle swelling at 3 days followed by chronic synovitis, pannus formation, and marginal erosion of cartilage and bone by 30 to 40 days. This swelling response is blocked by an anti-TNF antibody (Schwab et al., 1993). We therefore compared GW3333 with a hamster anti-mouse TNF antibody known to neutralize rat TNF in vitro and in vivo (Sheehan et al., 1989; Rabinovici et al., 1993). The anti-TNF antibody administered i.v. 15 min before the PGPS reactivation completely blocked the ankle swelling, whereas the control IgG had no effect (Fig. 3). GW3333 dosed b.i.d. from day 0 partially inhibited the ankle swelling, with the 80-mg/kg dose showing a statistically significant effect at days 2 and 3 and the 27-mg/kg dose exhibiting a statistically significant effect at day 3. Prednisolone, the steroid positive control, strongly inhibited ankle swelling on days 1, 2, and 3. 

View larger version (19K): [in this window] [in a new window]   Fig. 3.   Top, effect of anti-TNF antibody on ankle swelling after PGPS reactivation. Saline vehicle, IgG control, and anti-TNF antibody given i.v. 15 min before PGPS i.v. reactivation dose on day 0. Data are mean ± S.E.M. of n = 9. Anti-TNF antibody was different from IgG control at day 2 (P < 0.05) and day 3 (P < 0.01) using Dunnett's multiple comparison test. Bottom, effect of GW3333 on ankle swelling after PGPS reactivation. Treatments were given p.o. b.i.d starting 1 h before PGPS reactivation on day 0. GW3333 b.i.d. doses are shown. GW3333 was different from vehicle with 27 mg/kg at day 3 (P < 0.01) and 80 mg/kg at day 2 (P < 0.05) and day 3 (P < 0.01). The steroid positive control (prednisolone at 6 mg/kg) was different at days 1, 2, and 3 (P < 0.01).

Activity in the Rat Adjuvant Arthritis Model. In the adjuvant arthritis model, the ankles swell from day 11 to day 21 after adjuvant administration. The anti-TNF antibody administered i.v. on days 7, 11, and 17 delayed the swelling response, but the effect was not statistically significant (Fig. 4). GW3333 b.i.d. from day 7 to 21 at the 50- and 150-mg/kg doses caused statistically significant decreases in ankle swelling at days 16, 18, and 21. 

View larger version (20K): [in this window] [in a new window]   Fig. 4.   Top, effect of anti-TNF antibody on ankle swelling during adjuvant arthritis. IgG control and anti-TNF antibody given i.v. on days 7, 11, and 17. Oral vehicle used in bottom panel is also shown. Data are mean ± S.E.M. of n = 8. Bottom, effect of GW3333 on ankle swelling during adjuvant arthritis. Treatments were given p.o. b.i.d. from days 7 to 21. GW3333 b.i.d. doses are shown. GW3333 was different from vehicle with 50 mg/kg at day 16 (P < 0.01) and day 18 (P < 0.01) and with 150 mg/kg at day 11 (P < 0.05), day 14 (P < 0.05), day 16 (P < 0.1), day 18 (P < 0.01), and day 21 (P < 0.05).

View larger version (133K): [in this window] [in a new window]   Fig. 5.   A to F, effect of GW3333 and adjuvant arthritis on tibio-tarsal joint histology. These are representative sections through the hock joint of rats showing the lesions induced by adjuvant arthritis. Sections were cut sagittally, stained with hematoxylin/eosin, and photographed at magnifications of 10× and 100×. A and B, normal rat. A, normal joint structure showing tibial tarsal bone articulation with the distal tibia. Note the relative size of the marrow cavity (M) and the amount of subchondral bone (SB) in the tibial tarsal bone, for comparison with the adjuvant arthritis animals. C and D, vehicle-treated adjuvant arthritis rat. There is marked joint destruction with peri-articular subcutaneous inflammatory edema (IE) and synovial effusion (SE). Marked bone (BD) and cartilage (CD) destruction has occurred, and there is pannus formation (arrowheads) overlying the thin rim of articular cartilage remaining. The marrow cavity of the bone has been replaced by medullary granulation tissue with osteoclasts (MG), and there is new bone formation around the bone surface (arrows). E and F, GW3333-treated adjuvant arthritis rat. There is a marked periarticular subcutaneous granulation inflammatory reaction (SG); however, there is little bone or cartilage destruction. A small amount of medullary granulation tissue with osteoclast activation and mild bone destruction is present in the distal tibia (arrow). Within the joint cavity, there is a mild inflammatory reaction with fibrin exudation (F) and a small area of pannus (arrowheads) over the distal tibial articular cartilage. The subchondral bone (SB) and overlying articular cartilage are intact. The dose of GW3333 was 150 mg/kg b.i.d.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Previous work with short-acting TACE inhibitors showed that it was necessary to inhibit LPS-induced increases in TNF at 90 min by at least 85% to save mice from endotoxic death at 24 h (Mohler et al., 1994; Moss et al., 1999). Therefore, to evaluate the role of TNF convertase inhibition in chronic inflammation models, we sought an inhibitor that would completely block LPS-induced increases in TNF over at least a 12-h dosing interval. GW3333 is the first TACE inhibitor with sufficient potency and duration of action to completely inhibit LPS-induced increases in plasma TNF for up to 12 h in the rat (Table 4). The fact that GW3333 also blocked zymosan-induced increases in TNF in the pleural cavity (Table 5) suggests that the in vivo activity of GW3333 is due to inhibition of TACE, not something specific to the LPS stimuli.

In the PGPS reactivation model, the i.v. reactivation dose of PGPS induces T-cell activation and infiltration into the previously primed joint, resulting in a peak of ankle swelling at 3 days. Anti-TNF antibodies strongly inhibited ankle swelling, confirming that TNF is critical in this model (Fig. 3; Schwab et al., 1993). GW3333 also significantly inhibited ankle swelling in the PGPS model at both 27- and 80-mg/kg b.i.d. doses (Fig. 3). Both of these doses inhibited LPS-induced TNF production in normal rats, with the 80-mg/kg dose causing nearly complete inhibition of LPS- and zymosan-stimulated TNF production over the 12-h dosing interval (Tables 4 and 5). Plasma concentrations of GW3333 were lowest (about 0.6 µM) 12 h after the 80-mg/kg dose (Table 6). Given that GW3333 is 51% plasma-bound, the minimal free plasma concentration of GW3333 after 80 mg/kg was more than 50-fold above the IC₅₀ for rat collagenase (about 6 nM), suggesting profound inhibition of MMPs in vivo. In summary, chronic inhibition of TACE and MMPs by GW3333 produced an anti-inflammatory effect in a TNF-dependent model.

The anti-TNF antibody slightly inhibited ankle swelling in the adjuvant model, with no clear protective effect on joint histology or radiological scores (Fig. 4; Tables 7 and 8). We expected a more pronounced effect of anti-TNF antibody treatment based on studies wherein an anti-TNF antibody given from day 5 to 10 of adjuvant arthritis inhibited the migration of 51Cr-labeled polymorphonuclear leucocytes and 111In-labeled T lymphocytes into rat joints (Issekutz et al., 1994). The rats might have raised antibodies to the hamster anti-murine TNF antibody over the course of 14-day antibody treatment, thus blocking the longer-term effects of the antibody treatment. In contrast to the anti-TNF antibody, GW3333 at 50 and 150 mg/kg b.i.d. inhibited ankle swelling during adjuvant arthritis (Fig. 4) as well as cartilage and bone destruction as assessed by histology (Table 7). The 150-mg/kg dose also inhibited joint damage as assessed by radiology (Table 8). The plasma drug concentrations in the 50- and 150-mg/kg groups at the end of the adjuvant arthritis experiment (Table 6) were well above the concentrations needed to inhibit LPS-induced TNF production in dose response (Table 3) and duration of action studies (Table 4) in normal rats, suggesting that TNF convertase was strongly inhibited at these doses. Given that GW3333 inhibits rat collagenase in vitro with an IC₅₀ of 6 nM and is 51% bound in plasma, we can calculate that minimal free GW3333 concentrations during the 12-h period in the 50- and 150-mg/kg groups are at least 80-fold above the IC₅₀ for rat collagenase in vitro. This data in the adjuvant model is similar to results in the PGPS model, in that combined inhibition of TNF convertase and MMPs with GW3333 resulted in an anti-inflammatory effect. The anti-inflammatory effect, in combination with direct inhibition of degradative MMPs, likely explains the preservation of cartilage and bone structure with GW3333.

The mechanism by which GW3333 suppresses ankle swelling in the PGPS and adjuvant models is probably multifaceted. It is possible that TACE inhibition lowers soluble TNF to a level that will not support continued cell infiltration and activation of inflammatory cells in the joint. This mechanism is supported by the observation that GW3333 inhibited neutrophil influx into the pleural cavity after zymosan stimulation (Table 5) and decreased subcutaneous edema and pannus in adjuvant arthritis joints (Table 7). Inhibition of MMPs and connective tissue degradation could also have an impact on local inflammation. This possibility is supported by the observation that peptide products of collagen, gelatin, and fibronectin are chemotactic for a variety of inflammatory cells (Albini and Adelmann-Grill, 1985; Laskin et al., 1986; Clark et al., 1988). Furthermore, other MMP inhibitors with little or no TNF convertase inhibitory activity also show inhibition of ankle swelling in in vivo arthritis models (Conway et al., 1995; Lewis et al., 1997).

In addition to inhibition of TNF release and MMP degradation of matrix, MMP and TACE inhibitors can also increase cell surface TNF after stimulation of inflammatory cells in vitro as detected by antibody staining and fluorescence-activated cell sorter analysis (McGeehan et al., 1994; Mohler et al., 1994; Crowe et al., 1995) as well as bioactivity toward nearby detector cells (Solomon et al., 1997). If GW3333 is increasing cell surface TNF in arthritic rats, one might expect GW3333 to have a proinflammatory component, especially given the observation that mice chronically overexpressing cell surface TNF show an arthritis-like phenotype (Georgopoulos et al., 1996). However, the anti-inflammatory effect of GW3333 in both the PGPS and adjuvant arthritis models is not suggestive of chronic increases in cell surface TNF in vivo. This is consistent with the observation that GI5402, a dual inhibitor of TACE and MMPs, inhibited LPS-induced TNF production in vivo in humans with no effect on the cell surface TNF of peripheral monocytes (Dekkers et al., 1999).

Recently it has been shown that dual TACE and MMP inhibitors decrease the ability of stimulated cells to shed p55 and p75 TNF receptors (Crowe et al., 1995; Williams et al., 1996), L-selectin (Feehan et al., 1996), Fas ligand (Kiayagaki et al., 1995), IL-6 receptor (Müllberg et al., 1995), IL-1 decoy receptor (Orando et al., 1997), and epidermal growth factor (Lanzrein et al., 1995). We do not know the effect of GW3333 on these potentially important shedding events. The fact that GW3333 did not completely inhibit inflammation in the PGPS and adjuvant arthritis models leaves open the possibility that GW3333 may have some proinflammatory effects, such as increases in cell surface TNF and/or inhibition of various sheddases, that might counteract the decreases in soluble TNF and inhibition of MMPs. Another possibility is that IL-1-mediated pathways are still operating even when TNF is inhibited. Future work is needed to more fully address the specificity of inhibitors against TACE, MMPs, and shedding events and their impact on inflammatory processes in vivo.