Identification of Human Cathepsin G As a Functional Target of Boswellic Acids from the Anti-Inflammatory Remedy Frankincense¹

Frankincense, the gum resin derived from Boswellia species, is frequently used in folk medicine to cure inflammatory diseases. Data from animal models of inflammation and from clinical trials suggest a therapeutic value of frankincense in the treatment of acute and chronic inflammatory and allergic disorders (1, 2). The pentacyclic triterpenes boswellic acids (BAs)³ (see Fig. 1 A) are major ingredients of frankincense. Whereas 11-keto-β-BA (KBA) and 3-O-acetyl-11-keto-β-BA (AKBA) were shown to exhibit numerous biological activities in vitro, the corresponding BAs lacking the 11-oxo moiety (i.e., β-BA and Aβ-BA) are considered as less relevant. In particular, AKBA is thought to be mainly responsible for the pharmacological actions of frankincense (2). In search for a molecular basis of the anti-inflammatory effectiveness, 5- and 12-lipoxygenase (1), cyclooxygenase-1 (3), human leukocyte elastase (HLE) (4), and IκB kinases (5) have been identified as possible targets of AKBA in biochemical and cellular in vitro models. However, due to the high concentrations required to interfere with these targets (IC₅₀ of 1–50 μM) and the low maximal plasma levels of AKBA (<0.1 μM) obtained after oral administration of standard doses of frankincense extracts (6), the pharmacological relevance of the proposed targets and mechanisms remain questionable (7).

An elegant technique for the identification of a high-affinity small molecule (ligand)-protein (target) interaction is the protein-fishing approach, using an affinity matrix composed of the small molecule covalently linked to an insoluble resin (8). Application of this strategy revealed human histone deacetylase as target for trapoxin (9) or mTOR as target for the immunosuppressant rapamycin (10). Using a BA-affinity Sepharose matrix and neutrophil lysates as source of targets, we identified cathepsin G (catG) as a selected, high-affinity target for BAs. CatG, a neutral serine protease, is mainly expressed in neutrophils, stored in azurophilic granules, and released upon degranulation (11, 12). After release into the plasma, it cleaves extracellular matrix proteins, including laminin, proteoglycans, collagen, fibronectin, and elastin (13, 14), implying a role in local destruction of connective tissue at sites of injury. CatG also processes chemokines (11, 15, 16), functioning as chemoattractant for T cells and other leukocytes (17), and modulates integrin clustering on neutrophils (18). Moreover, catG stimulates platelets via the protease-activated receptor-4 for aggregation and secretion (19) and acts as chemotactic agonist for the formyl peptide receptor on phagocytic leukocytes (20). Accordingly, catG inhibitors have been proposed to exhibit potential in treating certain inflammatory disorders such as asthma, chronic obstructive pulmonary disease, emphysema, reperfusion injury, psoriasis, and rheumatoid arthritis (21). Since frankincense extracts showed beneficial effects in several of these disorders, interference of BAs with catG may possess pharmacological relevance.

BAs were prepared as described previously (22). α-Amyrin was from Extrasynthèse; Boswellia serrata gum extract PS0201Bo was standardized to at least 1% KBA and 1% AKBA (quantified by reversed phase HPLC) and was provided by Pharmasan; EAH-Sepharose 4B, GE Healthcare Bio-Sciences; Abs against catG, BIOMOL; human catG, human tryptase, human chymase, human chymotrypsin, elastase inhibitor IV (N-(2-(4-(2,2-dimethylpropionyloxy)phenylsulfonylamino)benzoyl)aminoacetic acid) and 2-[3-{methyl[1-(2-naphthoyl)piperidin-4-yl]amino}carbonyl)-2-naphthyl]-1-(1-naphthyl)-2-oxoethylphosphonic acid (JNJ-10311795), Calbiochem; human proteinase-3, Elastin Products; matrigel, BD Biosciences; trypsin (sequencing grade), Roche; all other fine chemicals were obtained by Sigma-Aldrich, unless stated otherwise.

For isolation of platelets and neutrophils venous blood was taken from healthy adult donors, with consent. The subjects had no apparent inflammatory conditions and had not taken anti-inflammatory drugs for at least 10 days before blood collection. Leukocyte concentrates were prepared by centrifugation at 4000 × g for 20 min at 20°C. Neutrophils were immediately isolated by dextran sedimentation, centrifugation on Nycoprep cushions, and hypotonic lysis of erythrocytes. Neutrophils (7.5 × 10⁶ cells/ml; purity >96–97%) were finally resuspended in PBS (pH 7.4) plus 1 mg/ml glucose. Platelets were isolated from supernatants (800 × g, 10 min, room temperature) after centrifugation of leukocyte concentrates on Nycoprep cushions (see above) to obtain platelet-rich plasma. Platelet-rich plasma was then mixed with PBS (pH 5.9; 3/2, v/v), centrifuged (2000 × g, 15 min, room temperature), and the pelleted platelets were resuspended in PBS (pH 5.9)/0.9% NaCl (1/1, v/v), washed by centrifugation (2000 × g, 10 min, room temperature) and finally resuspended in PBS (pH 7.4). Test compounds were solubilized in ethanol or DMSO, never exceeding 1% (v/v). To exclude toxic effects of BAs or vehicle (ethanol, 1%) during preincubation, neutrophil viability was analyzed by light microscopy and trypan blue exclusion. Incubation with 30 μM of any of the BAs or ethanol (1%) at 37°C for up to 30 min caused no significant change in cell viability.

KBA was treated with glutaric anhydride to form the half-ester glut-KBA, and analyzed by ¹H and ¹³C nuclear magnetic resonance as well as by mass spectrometry. Glut-KBA was linked to EAH Sepharose 4B by standard amide coupling procedures. The carboxylic acid of the KBA core was unlikely to react due to steric crowding. The success of the coupling reaction was determined by two methods. First, glut-KBA was used in defined excess (2 μmol of the glut-KBA per 1 μmol of NH₂ groups of the EAH Sepharose 4B). After the coupling reaction, the hypothetical excess of glut-KBA (1 μmol) could be indeed recovered. Second, treatment of glut-KBA with KOH in isopropanol under reflux for ∼3 h cleaved the ester bond and gave KBA, analyzed by thin layer chromatography.

For protein pull-down experiments, neutrophils were lysed by 1% Triton X-100 and lysates were incubated with the Sepharose slurries in lysis buffer (50 mM HEPES (pH 7.4), 200 mM NaCl, 1 mM EDTA, 1% Triton X-100, 2 mM PMSF, 10 μg/ml leupeptin, 120 μg/ml soybean trypsin inhibitor). Alternatively, isolated catG (1 μg) was solubilized in 425 μl of lysis buffer containing 0.02% BSA and incubated with the Sepharose slurries. The Sepharose beads were intensively washed, and precipitated proteins were finally separated by SDS-PAGE and visualized by silver staining or by Western blotting, respectively, as described (23). For identification of selected proteins of interest, bands were cut out from the gel and proteins were digested and subjected to MALDI-TOF mass spectrometry (MS).

Protein bands were manually cut out and dissected gel pieces were subjected to in-gel digestion (24, 25), adapted for use on a Microlab STAR digestion robot (26). Samples were reduced, alkylated, digested overnight by trypsin, extracted, and the extracts were dried in a vacuum centrifuge. MALDI-TOF-MS experiments were performed on an Ultraflex TOF/TOF mass spectrometer (Bruker Daltonics). The samples were dissolved in 5 μl of water/acetonitrile/TFA (29.5/70/0.5, v/v/v), and α-cyano-4-hydroxycinnamic acid (3 mg/ml) was used as matrix. Analyte and matrix were spotted consecutively in a 1:1 ratio on a stainless steel target and dried under ambient conditions. The dried sample was washed with ice-cold 5% formic acid to reduce salt contamination. Spectra were externally calibrated with a Sequazyme peptide mass standards kit (Applied Biosystems) and internally calibrated on a tryptic auto digestion peptide (m/z of 2163.0564). The spectra were processed in flexAnalysis version 2.2 (Bruker Daltonics) using the SNAP algorithm (signal to noise threshold, 3; maximal number of peaks, 150; quality factor threshold, 40). Proteins were identified by Mascot (Matrix Science) (peptide mass tolerance, 100 ppm; maximum missed cleavages using the NCBInr database: 2,314,886 sequences, 1,066,605,192 total letters, date Jan. 26, 2005). Proteins with a score of 76 or higher were considered significant (p < 0.05).

Human neutrophils (2 × 10⁹) were suspended in ice-cold 0.15 M NaCl and sonicated (five times for 30 s, 65%). Insoluble material was removed by centrifugation (600 × g, 10 min). The resulting supernatant was centrifuged (16,000 × g, 30 min) and the granular fraction (pellet) was resuspended in 1 M NaCl plus 0.005% Triton X-100. The suspension was centrifuged at 16,000 × g for 30 min, and four volumes of water were added to restore isotonicity. Proteins were precipitated by ammonium sulfate (60% saturation) and then resuspended in 40 ml of 0.05 M Tris-HCl (pH 8.0). After centrifugation (16,000 × g, 30 min), the supernatant was subjected to an elastin-Sepharose affinity chromatography column (2.5 × 20 cm) and equilibrated with 0.05 M Tris buffer (pH 8.0). The column was eluted with the equilibration buffer until the OD₂₈₀ returned to baseline and then washed with 2 vol of 0.05 M Na acetate, 1 M NaCl (pH 5.0), and fractions containing catG activity were eluted with 0.05 M Na acetate, 1 M NaCl, 20% DMSO (pH 5.0). Active fractions were pooled, dialyzed in Vivaspin cut-off columns (5000 MWCO) against 20 mM Na acetate, 0.15 M NaCl (pH 5.5), and subjected to ion-exchange chromatography (CM Sephadex C-50) column, equilibrated in the same buffer. The sample was applied, washed with equilibration buffer, and bound material was eluted by a linear NaCl gradient from 0.15 to 1 M. The total elution volume was 300 ml, and fractions of 6 ml were collected at a flow rate of 30 ml/h. The homogeneity of the purified material was assessed by SDS-PAGE and catG activity assays.

The proteases were mixed with test compounds or DMSO (vehicle control, <0.5%) in the respective assay buffer in a 96-well plate (total volume 200 μl) and preincubated for 20 min at 25°C. The respective chromogenic protease substrate was added and the proteolysis was monitored at 410 nm by spectrophotometric measurement using a Victor² plate reader (PerkinElmer). The enzymatic activity was determined by the progress curve method. The protease activity in the presence of inhibitor was compared with an uninhibited control (DMSO as vehicle), and inhibition of the protease is given as the percentage of the control without inhibitor.

For analysis of isolated catG, the enzyme was either purified from neutrophils as described, or commercially available purified catG from neutrophils (Calbiochem) was used. Purified enzyme (0.2 μg), diluted in 200 μl of HEPES 0.1 M, NaCl 0.5 M (pH 7.4), 10% DMSO, was incubated with N-Suc-Ala-Ala-Pro-Phe-pNA (Suc-AAPF-pNA, 1 mM) as substrate and the absorbance was measured at 410 nm at 25°C. Kinetic studies were performed at substrate concentrations ranging from 0.1 to 4 mM using the Lineweaver-Burk method.

Inhibition of related proteases was performed in analogy to catG. The amounts of protease, the assay buffer, and the substrate were individually adjusted to each type of protease as follows: tryptase, 0.5 μg of purified enzyme, Tris-HCl 0.1 M (pH 8.3), 10% DMSO as assay buffer, and N-α-benzoyl-dl-Arg-pNA (1 mM) as substrate; chymase, 0.1 μg of purified enzyme, Tris-HCl 0.45 M, NaCl 1.8 M (pH 8.0), 10% DMSO, and 0.5 mM Suc-AAPF-pNA as substrate; chymotrypsin, 0.1 μg of purified enzyme, Tris-HCl 0.1 M, CaCl₂ 25 mM (pH 8.3), 10% DMSO, and 0.2 mM Suc-AAPF-pNA as substrate; human leukocyte elastase, 0.15 μg of purified enzyme, HEPES 0.1 M, NaCl 0.5 M (pH 7.4), 10% DMSO, and 0.2 mM N-methoxysuc-Ala-Ala-Pro-Val-pNA as substrate; proteinase-3, 0.5 μg of purified enzyme, MOPS 0.1 M, NaCl 0.5 M, 5,5′-dithiobis-(2-nitro-benzoic acid) 0.1 mM (pH 7.5), 10% DMSO and 1 mM Boc-Ala-Ala-Nva-SBzl as substrate.

Rigid automated molecular docking was performed using GOLD v4.0 (Cambridge Crystallographic Data Centre, Cambridge, U.K.; www.ccdc.cam.ac.uk), which relies on a genetic algorithm (27). We used a known crystal structure of catG (identifier: 1t32, 1.85 Å resolution) (28) from the Protein Data Bank (29). Hydrogens were added, and then energy minimized using the AMBER99 force field (30) within the software package MOE v2008.10 (Chemical Computing Group). For the co-crystallized catG inhibitor JNJ-10311795, hydrogen atoms were added, and energy minimization was performed using the MMFF94x force field (31). The three-dimensional structure of AKBA has recently been determined by our group (N. Kather, L. Tausch, D. Poeckel, O, Werz, E. Herdtweck, and J. Jauch, manuscript in preparation:). GOLD parameter settings for the genetic algorithm were: number of operations, 100,000; population size, 100; selection pressure, 1.1; number of islands, 5; niche size, 2; migrate, 10; mutate, 95; crossover, 95. A 20-Å radius around the active site defined the binding pocket for automated docking. The ChemScore fitness function (32) was employed for scoring the predicted receptor-ligand complexes. It estimates the total free energy change that occurs on ligand binding and was trained by regression against binding affinity data. As the fitness scores are dimensionless, they cannot be used explicitly as values for binding energy or binding affinity. However, in each case, the scale of the score gives a guide as to how good the pose is: the higher the score, the better the docking result is likely to be. The ChemScore function takes into account factors such as protein-ligand hydrogen bond energy, lipophilic interactions, the entropic loss that occurs when single, acyclic bonds in the ligand become nonrotatable upon binding, and clashes between protein and ligand atoms as well as ligand internal torsional strain energy. Each docking run was repeated 10 times to obtain average score values with SE. The same method was used for redocking of the co-crystallized inhibitor. Root mean square deviation (RMSD) values between the x-ray co-crystal coordinates and the docking solutions were computed, and a mean value with SE was calculated. PyMOL was used for visualization of docking poses (33).

Experiments were conducted on a Biacore X device. Purified catG (100 μg/ml) was immobilized onto CM5 biosensor chip flow cells using the standard amine coupling method according to the manufacturer’s instructions. Flow cell 1 was not altered (reference), and catG (170 fmol/mm²) was immobilized on flow cell 2 corresponding to 4500 resonance units. Equilibration of the baseline was obtained by a constant flow of HBS-P buffer (10 mM HEPES, 150 mM NaCl, 0.01% P20, 1% DMSO (pH 7.4)) for 4 h. Stock solutions of BAs, α-amyrin, and JNJ-10311795 were diluted into assay buffer. Measurements were performed at 25°C and a flow rate of 30 μl/min. After recording association, the liquid phase was replaced by assay buffer and the dissociation was monitored. The binding profiles were obtained after subtracting the response signal of the untreated reference cell 1. Sensorgrams were processed by using automatic correction for nonspecific bulk refractive index effects using Biacore evaluation version 3.1 software.

Platelets (6 × 10⁸/ml PBS plus 1 mg/ml glucose) were incubated with 2 μM Fura-2-AM for 30 min at 37°C. After washing, 10⁸ platelets were resuspended in 1 ml of PBS plus 1 mg/ml glucose, preincubated with inhibitors for 10 min at room temperature, and incubated in a thermally controlled (37°C) fluorometer cuvette in a spectrofluorometer (Aminco-Bowman series 2; Thermo Spectronic) with continuous stirring. CaCl₂ (1 mM) was added 30 s before the addition of catG (0.1 nmol/ml), and the fluorescence emission at 510 nm was measured after excitation at 340 and 380 nm, respectively. Intracellular Ca²⁺ levels were calculated according to the method of Grynkiewicz et al. (34). In other experiments, 10⁸ Fura-2-AM-loaded platelets were mixed with 10⁷ unloaded neutrophils in 1 ml of PBS plus 1 mg/ml glucose and 0.1 mM EDTA, and preincubated with the inhibitors for 10 min at room temperature. Five min prior to stimulation with 100 nM fMLP, 10 μM cytochalasin B was added, and CaCl₂ (1.1 mM) was added 1 min prior to fMLP. The fluorescence was measured and intracellular Ca²⁺ concentration was calculated as described above.

The migration of neutrophils along a chemotactic (fMLP) gradient through Matrigel was performed according to Steadman et al. (35) with some slight modifications. In brief, freshly isolated neutrophils (2 × 10⁶) were resuspended in 1 ml of HEPES-buffered RPMI 1640 medium with 10% (v/v) FCS and preincubated with test compounds or with vehicle (DMSO, <0.5%). Cell suspension (150 μl) was then placed on the upper chamber of a two-compartment Boyden chamber (5-μm pore size filters) in a 24-well format. Cells were then allowed to migrate through Matrigel-coated pore size filters for 40 min into the lower chamber containing buffer (negative control) or fMLP (0.1 μM) as chemoattractant. Cells on the bottom of the wells were fixed with 3.7% formaldehyde, stained with Gram’s Violet, washed, and the stain was solubilized using acetic acid. The absorption of the eluted stain was measured at 570 nm/620 nm.

Freshly isolated neutrophils (2 × 10⁵) were resuspended in 100 μl of HEPES-buffered RPMI 1640 medium with 10% (v/v) FCS and preincubated with test compounds or with vehicle (1% DMSO) for 20 min at 37°C. The cell suspension was placed on the upper chamber of a two-compartment microchemotaxis chamber (8-μm pore size filters) in a 24-well format. Cells were allowed to migrate into filters for 60 min in contact to the lower chamber containing HEPES-buffered RPMI 1640 medium with 10% (v/v) FCS and DMSO (negative control) or fMLP (0.1 μM) as chemoattractant. Cells in the filters were fixed with 4% formaldehyde and stained with Harris hematoxylin solution. The total numbers of cells in three high-power microscopic fields (×400) were counted.

In a randomized, double-blind, placebo-controlled, multicenter trial the tolerability and efficacy of orally administered B. serrata gum extract PS0201Bo was investigated for maintaining the remission of Crohn’s disease. In the study addressing the tolerability (phase I), two soft gelatin capsules of PS0201Bo each containing 400 mg (total of 800 mg of PS0201Bo) were taken by healthy adult male volunteers in the morning immediately after a defined meal. Venous blood samples were collected from the volunteers after 0, 2, 4, 8, and 24 h, citrated, and immediately stimulated with 10 μM cytochalasin B and 2.5 μM fMLP for 5 min. Plasma was prepared and catG activity was measured (see above).

In the study addressing the efficacy of PS0201Bo (phase II–III), the patients received two soft gelatin capsules of PS0201Bo each three times daily (total dose of 2400 mg of PS0201Bo/d) or two capsules of placebo, respectively, during or immediately after a meal over a period of 52 wk. Venous blood samples from the patients were collected before study medication application and after 4 wk of treatment and catG activity in the plasma was assessed as described above. In parallel, aliquots of the plasma were used for analysis of BAs (see below).

The study protocol and a sample patient information/consent form were reviewed by the competent Ethics Committee (IECs) of the University of Frankfurt, Germany, and a favorable opinion was issued on July 6, 2006 (Geschäfts no. 158/06); regulatory authority approvals were obtained from the competent Higher Federal Authority, the Bundesinstitut für Arzneimittel und Medizinprodukte, Bonn, Germany (EudraCT no. 2006-002939-24, Vorlage no. 4031905 of July 18, 2006). This trial was conducted in compliance with the protocol and principles of the Declaration of Helsinki (1996) and the International Conference on Harmonisation/World Health Organization Good Clinical Practice guidelines.

The content of BAs in human plasma was determined by liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS). BAs were separated using a Gemini C18 column (150 mm × 2 mm internal diameter, 5 μm particle size, and 110 Å pore size; Phenomenex) and determined with an API 4000 triple quadrupole mass spectrometer (Applied Biosystems) equipped with a Turbo-V-source operating in negative electrospray ionization mode. High-purity nitrogen for the mass spectrometer was produced by a NGM 22-LC/MS nitrogen generator (cmc Instruments). A linear gradient was employed at a flow rate of 0.6 ml/min mobile phase with a total run time of 10 min (mobile phase A, water/ammonia (100/0.05, v/v); mobile phase B, acetonitrile/ammonia (100/0.05, v/v)). The gradient started from 90% phase A to 10% within 3.5 min. This was held for 1.5 min at 10% phase A. Within 0.5 min the mobile phase shifted back to 90% phase A and was held for 4.5 min to equilibrate the column for the next sample. The injection volume of samples was 20 μl. Retention times of KBA, AKBA, β-BA, and Aβ-BA, and internal standard (2,5-dimethyl-celecoxib (DMC)) were 3.70, 4.13, 4.19, 4.34, and 5.24 min, respectively. The quantification of BAs was performed with Analyst software V. 1.4.2 (Applied Biosystems) employing the internal standard (DMC). Ratios of analyte peak area and internal standard area (y-axis) were plotted against concentrations (x-axis), and calibration curves were calculated by least squares regression with 1/concentration² weighting.

BA stock solutions (500 μg/ml) were prepared in methanol and further diluted with methanol to obtain working standards with the following concentration ranges: β-BA from 1.25 to 12,500 ng/ml and Aβ-BA, KBA, and AKBA from 0.125 to 1250 ng/ml. The following working standards were used for plasma to obtain calibration standards with the concentrations ranges: β-BA from 0.125 to 1250 ng/ml, and Aβ-BA, KBA, and AKBA from 0.013 to 125 ng/ml. The lower limit of quantification belongs to the standard curve and is the first point that has (1) accuracy between 80 and 120%, and (2) five times the peak-to-noise ratio (according the Food and Drug Administration 2001 Guidance for Industry: Bioanalytical Method Validation) and was 0.125 ng/ml for β-BA and Aβ-BA and 0.038 ng/ml for KBA and AKBA, respectively.

Samples for standard curve and quality controls were prepared with 200 μl of blank human plasma (Blutspendedienst Hessen, Deutsches Rotes Kreuz, Frankfurt, Germany), 50 μl of ammonia solution 25%, 20 μl of working standards, and 20 μl of internal standard solution (25 ng/ml DMC). Plasma from patients was prepared similarly, but instead of 200 μl of blank human plasma and 20 μl of working standard, 200 μl of sample and 20 μl of methanol were added. BAs and internal standard were extracted by liquid-liquid extraction. The prepared samples were extracted twice with ethyl acetate. The organic phase was removed at 45°C under a gentle stream of nitrogen. The residues were reconstituted with 50 μl of acetonitrile/water/ammonia (60/40/0.1, v/v), centrifuged for 2 min at 10,000 × g, and then transferred to glass vials (Macherey-Nagel) before injection in the LC/MS-MS system.

Statistical evaluation of the data was performed by one-way ANOVAs for independent or correlated samples followed by Bonferroni post hoc tests. Where appropriate, Student’s t test for paired observations was applied. A p value of <0.05 was considered significant.

KBA was linked at the C3-OH moiety using glutaric anhydride yielding the half ester 3-O-glutaroyl-KBA (glut-KBA). The free OH moiety at C3 is unlikely to be part of a pharmacophore since esterification (e.g., in AKBA) frequently improves the potency in pharmacological assays (1). The de novo formed free carboxylic group of glut-KBA was amide-coupled with the primary amine of EAH Sepharose 4B, yielding KBA-Seph (Fig. 1 B). EAH Sepharose 4B (Seph) without glut-KBA as ligand was used as negative control to discriminate unspecific protein binding. After incubation of neutrophil lysates with Sepharose beads, bound proteins were separated by gel electrophoresis and visualized by silver staining.

Several neutrophil proteins were precipitated by both KBA-Seph and Seph. However, a 26-kDa protein was significantly enriched in KBA-Seph precipitates (Fig. 1,C). This protein band was excised, in-gel digested by trypsin, and subjected to MALDI-TOF-MS. Analysis of the obtained peptide fragments using a peptide sequence data base revealed that the peptides matched (sequence coverage of 63%) to the primary amino acid sequence of the human neutral serine protease catG (25.8 kDa, Fig. 1,D). Immunological analysis by Western blot confirmed the identity of catG, and also isolated catG bound to KBA-Seph (Fig. 1, E and F), which excludes coprecipitation via a linker molecule and thus confirms a direct interaction of BAs with catG.

To further study the binding of BAs to catG, we used automated molecular docking. The co-crystallized synthetic catG inhibitor JNJ-10311795 (K_i for catG of 38 nM) (28), designed and developed based on structural information from the x-ray structure of catG, was successfully redocked into Protein Data Bank structure 1T32 (reference for catG). The acquired binding mode into the S1 and S2 pocket was identical to the x-ray structure with an RMSD of 0.22 ± 0.03 Å and a ChemScore of 52.97 ± 0.08. The protein-ligand interaction seems to be mainly driven by lipophilic/pi-pi interactions and a tight H-bonding network formed by the phosphonate group, which is also reflected in the ChemScore (Table I). The 2-naphthyl group occupies S1 and the 1-naphthyl group S2, with the latter being involved in an aromatic stacking interaction with the imidazole of H57. One phosphonate oxygen is H-bonded with Nε of H57, another oxygen is H-bonded to Nε of K192, and the third is H-bonded to the backbone NH of G193 and Oγ of S195. The carboxamido-N-(naphthalene-2-carboxyl)piperidine segment of JNJ-10311795 occupies the hydrophobic S3/S4 cavity, which is defined by the side chains of Y215, I99, and F172 (Fig. 2). Docking of BAs into the same box resulted in average docking scores for AKBA of 26.12 ± 0.32, KBA of 28.38 ± 0.25, Aβ-BA of 24.83 ± 0.36, and β-BA of 26.51 ± 0.21, with an apparently identical binding mode occupying the same part (S1) of the binding site as JNJ-10311795. For all BAs, we observed only one binding mode with RMSD values between 0.44 ± 0.06 and 0.55 ± 0.08 Å over all docking solutions (Table I). From the ligand poses, one can speculate that the contributions to the scores are mainly based on lipophilic and H-bonding interactions (Table I). Aβ-BA forms an H bond between the acetyl oxygen and the side chain hydroxyl of S40, and β-BA can be H-bonded with its hydroxyl group and the backbone carbonyl oxygen of H57, similar to KBA, with the latter forming also an H bond between the ketone oxygen and the backbone NH of G193 and the side chain hydroxyl of S195. In our model, AKBA forms H bonds between its acetyl oxygen and the side chain hydroxyl of S40, and between the ketone oxygen and the side chain hydroxyl of S195. Docking of α-amyrin as a negative, nonbinding control yielded a ChemScore of 27.36 ± 0.25, with two different docking poses detected with RMSD values of 2.36 ± 0.35 Å over all solutions (Fig. 2).

To gain more insights into the specificity of the BA-binding to catG, we tested whether an excess of either soluble β-BA or JNJ-10311795 could prevent binding of the isolated protease to immobilized KBA. As shown in Fig. 3 A, JNJ-10311795 (10 μM) clearly antagonized catG-binding from KBA-Seph, supporting the hypothesis that BAs bind to the same pocket of catG as JNJ-10311795. Also, soluble β-BA at 1 mM reduced catG-binding to immobilized BAs. Application of higher concentrations (>1 mM) of BAs or analysis of α-amyrin was infeasible due to insufficient aqueous solubility of the compounds.

Next, we attempted to characterize the interaction of BAs with catG by SPR spectroscopy using a Biacore X device. The data obtained indicate a reversible interaction between catG (immobilized) and BAs or JNJ-10311795 (used as analytes), whereas the α-amyrin (negative control) does not interact with catG (Fig. 3,B). However, at higher BA concentrations (>8–10 μM), only undefined and inconsistent binding patterns were obtained (Fig. 3 C), presumably originating from concentration-dependent aggregation or superstoichiometric binding of Bas, thus precluding determination of detailed binding kinetics.

To investigate if the interaction of BAs with catG may have a functional consequence, we first analyzed whether BAs modulate the proteolytic activity of catG. All BAs investigated potently inhibited isolated catG, whereas the related α-amyrin and ursolic acid showed no significant inhibition (Fig. 4,A). Also glut-KBA was effective, indicating that esterification of the C3-OH by glutaric acid is not detrimental. Concentration-response studies (Fig. 4,B) revealed IC₅₀ values of 0.6 μM for AKBA, 0.8 μM for β-BA, 1.2 μM for Aβ-BA, and 3.7 μM for KBA. The potency of BAs was reduced by increasing the peptide substrate concentration, and Lineweaver-Burk blots imply a competitive inhibition of catG by Aβ-BA (Fig. 4,C). Moreover, wash-out experiments support a reversible inhibition of catG by BAs (Fig. 4 D).

The effects of BAs on the activity of the closely related serine proteases chymase, tryptase, HLE, proteinase-3, and chymotrypsin were determined under individually optimized, but still comparable, experimental conditions. AKBA and Aβ-BA (10 μM, each) failed to significantly inhibit chymase and tryptase, but slightly suppressed HLE and proteinase-3 activity (IC₅₀ values of AKBA and Aβ-BA ≥30 μM) (Fig. 5,A). Aβ-BA inhibited chymotrypsin with an IC₅₀ of 4.8 μM (Fig. 5 B). Also for JNJ-10311795 it was found before that among various serine proteases, chymotrypsin is rather sensitive against this compound (K_i of 490 nM) (28). Together, apart from the inhibitory effect of Aβ-BA on chymotrypsin, BAs seem to be rather selective for catG.

Once released into the plasma, catG may contribute to migration/invasion of neutrophils along chemoattractant gradients by degrading extracellular matrix proteins (13). The chemoattractant fMLP increased (4.1-fold) chemoinvasion of human neutrophils through Matrigel (assessed by modified Boyden chamber assay), which was efficiently suppressed by JNJ-10311795 (0.1 μM) as well as by AKBA or Aβ-BA (Fig. 6,A), with IC₅₀ values of 2.7 and 4.5 μM, respectively (Fig. 6,C). In contrast, the elastase inhibitor IV (0.2 μM) failed in this respect (data not shown). On the other hand, in the absence of Matrigel, neither JNJ-10311795 nor AKBA or Aβ-BA reduced the chemotactic response of neutrophils toward fMLP (Fig. 6 A). Furthermore, preincubation of neutrophils with BAs did not result in an increased chemokinetic activity in the absence of fMLP (data not shown).

CatG, released from challenged neutrophils, also stimulates platelets for aggregation and secretion by cleavage of the protease-activated receptor-4, accompanied by mobilization of Ca²⁺ (19). Addition of isolated catG (0.1 nmol/ml) to human washed platelets caused a marked but transient Ca²⁺ mobilization that was essentially prevented by preincubation of platelets with AKBA (5 μM) or JNJ-10311795 (0.2 μM) to the same extent (Fig. 6,D). In contrast, AKBA (5 μM) failed to inhibit Ca²⁺ mobilization in platelets activated by 0.5 U/ml thrombin (not shown), excluding general suppression of Ca²⁺ mobilization by AKBA. In another type of experiment, fMLP caused no Ca²⁺ mobilization in Fura-2-loaded platelets unless unloaded neutrophils were coincubated. This Ca²⁺ mobilization was strongly suppressed by 0.1 μM JNJ-10311795 or 3 μM AKBA (Fig. 6 E). Together, BAs are able to block cellular functions by inhibiting catG.

In a clinical trial (i.e., on the safety and effectiveness of frankincense extract in healthy volunteers and patients with Crohn’s disease), a single dose of 800 mg orally administered B. serrata gum extract to healthy volunteers (phase I) transiently reduced catG activity ex vivo in cytochalasin B/fMLP-stimulated blood (Fig. 7,A). In the phase II–III study, catG activity was first measured in plasma prepared from venous blood of Crohn’s disease patients before medication. Four weeks after oral intake of 3 × 800 mg/day extract (representing the steady-state plasma levels of BA, Table II) or of placebo, respectively, catG activity was analyzed in plasma again. CatG activity of patients receiving placebo was not significantly different during treatment, whereas a clear and significant reduction of catG activity in the plasma of patients that received frankincense extract was evident (Fig. 7 B). No reduced activity of HLE or proteinase-3 in the plasma was immediately apparent (our unpublished data). Note that neutrophil numbers in the blood of verum-treated patients did not significantly change during treatment (6.9 ± 1.1 cells/nl before treatment and 6.7 ± 0.6 cells/nl after treatment).

Herein we provide evidence that catG is a functional target of BAs with pharmacological relevance. First, catG was identified as a selective BA-binding protein from neutrophils in a target-fishing assay using a BA-affinity Sepharose matrix. Because isolated catG bound to this affinity matrix, BAs may bind directly to catG and not via a linker protein. Moreover, BAs docked into the active site of catG in a similar fashion as for the synthetic catG inhibitor JNJ-10311795 (28), and JNJ-10311795 antagonized catG-binding to immobilized BAs. Also, SPR studies indicate a reversible interaction between BAs and catG. Second, BAs potently inhibit catG activity with partially submicromolar IC₅₀ values in a competitive and reversible manner. Third, BAs suppress catG-mediated cellular responses such as fMLP-induced chemoinvasion of neutrophils, as well as Ca²⁺ mobilization in platelets evoked by isolated catG or by fMLP-stimulated neutrophils. Fourth, studies with human subjects demonstrate a significant suppression of catG activity ex vivo after oral administration of B. serrata gum extracts vs placebo. Finally, catG may play roles in inflammatory diseases (i.e., rheumatoide arthritis, bronchial asthma) (11, 21, 36), where frankincense extracts showed beneficial effects in several studies (2).

Inhibition of a serine protease, that is, HLE by AKBA, was shown before, but significantly higher concentrations (IC₅₀ of 15 μM) were required (4) as compared with catG. We found that BAs inhibited all three neutrophil serine proteases from azurophilic granules (catG, HLE, and proteinase-3) that have an overlapping range of extracellular substrates, with catG being the most affine target. The successful identification of catG as BA target by the affinity fishing approach is probably the result of the combination of several fortunate circumstances. Thus, the relatively high affinity of catG to KBA-Seph enabled binding under stringent conditions and endured the thorough washing of the precipitates. Moreover, the C3-OH moiety chosen to link KBA is no pharmacophore since esterification with glutaric acid did not hamper the interference with catG or docking into the active center. In fact, the esterified analogs of KBA (i.e., AKBA and Glut-KBA) were even somewhat superior in inhibiting catG, and also for other targets, AKBA is frequently more potent than KBA (37, 38, 39). Finally, BAs are highly rigid molecules, thus favoring tight binding to catG. Note that besides catG, no other protein from neutrophils was obviously enriched in KBA-Seph pull-downs vs Seph precipitates.

In the docking studies, BAs bound to the active center in catG as JNJ-10311795 with a similar binding mode blocking the S1 pocket. It is reasonable to assume that the interactions are mainly hydrophobic together with H bonds formed between the ketone oxygen and/or the hydroxyl group of the BAs and several amino acids of catG. Notably, these are identical with the essential interactions of JNJ-10311795. We showed experimentally that α-amyrin, which differs from β-BA only in the carboxyl group and the configuration of the C3-OH, does not bind to catG or inhibit its activity. Even though this is not reflected in the docking scores, one explanation might be that the carboxyl group of the BAs shields the ligand from the solvent and this entropic contribution might be crucial for binding. Entropic contributions are not included in the scoring function of the docking program. This hypothesis is supported by the fact that the carboxyl group of all BAs always points toward the solvent in all docking solutions, never into the protein binding pocket.

Although AKBA and KBA have been considered as the pharmacologically relevant BAs (1, 2), we find that BAs lacking the C11-oxo moiety are about equipotent to 11-keto-BAs to interfere with catG. In view of the high plasma levels of β-BA/Aβ-BA (e.g., 2.4 - 10.1 μM) as compared with the fairly low levels of KBA and AKBA (≤0.34 and 0.1 μM, respectively) achieved after oral administration of frankincense preparations (Ref. 6 and this study), the interaction of β-BA/Aβ-BA with catG might be of pharmacological relevance.

We showed that BAs and JNJ-10311795 bound to catG with a similar binding mode, and in neutrophil/platelet coincubation experiments, AKBA as well as JNJ-10311795 suppressed the increase in intracellular Ca²⁺ concentration of platelets evoked by catG or by catG released from fMLP-challenged neutrophils. Similarly, BAs as well as JNJ-10311795 blocked chemoinvasion of neutrophils in response to fMLP (mediated by catG), but both compounds failed to affect chemotaxis. Comparisons of pharmacological actions of BAs and of JNJ-10311795 in vivo reported in the literature indicate that the substances exhibit related properties. Thus, JNJ-10311795 efficiently inhibited glycogen-induced rat peritonitis and blocked LPS-induced rat airway neutrophilia (28). Similarly, mixtures of BAs or frankincense extracts impaired carrageenan-induced pleurisy in rats accompanied by reduced neutrophil infiltration (40). Moreover, the increases in the pro-inflammatory cytokines TNF-α and IL-1β in inflamed rat tissues were reversed by JNJ-10311795 (28) as well as by frankincense (1).

CatG may play a role in the pathogenesis of various inflammatory diseases (11, 28, 41), and catG inhibitors were suggested to have potential for treating asthma, psoriasis, and rheumatoid arthritis (11, 21), for which frankincense extracts were proposed to have therapeutic benefit (1, 2). Nevertheless, experiments with catG-deficient mice led to controversial results regarding the importance of catG in chemoinvasion and in inflammatory processes. For example, MacIvor et al. showed that neutrophils from catG^−/− mice displayed normal phagocytosis, superoxide production, and normal chemotactic responses to C5a, fMLP, and IL-8 (42). In contrast, others found reduced neutrophil infiltration, myeloperoxidase, and chemoattractants (CXCL1 and CXCL2) in ischemic kidneys of catG-deficient mice (43). Also, mice lacking catG and HLE failed to initiate cytoskeletal reorganization and cell spreading, and they exhibited severe defects in the release of MIP-2 and reactive oxygen intermediates, and exogenously added, proteolytically active catG largely restored these defects (18). However, no catG inhibitor is available for therapeutic use, and data from clinical trials with respective candidates are still missing.

In conclusion, we provide evidence for catG as a functional target of BAs in vivo. The potent interference of β-BA and Aβ-BA with catG in relationship to their high achievable plasma levels favors this interaction as a possible molecular basis underlying certain beneficial effects of frankincense observed in animal models of inflammation, as well as in human subjects suffering from inflammatory disorders. However, the importance and value of catG as a therapeutic target in chronic inflammatory diseases has not been entirely assessed and is controversially discussed. On the other hand, the knowledge about the functionality of BAs as catG inhibitors may offer new possibilities for use of frankincense in additional inflammatory diseases, for example, in chronic obstructive pulmonary disease, where catG may contribute, but also in acute inflammatory states. Finally, investigations with defined BAs or with structurally optimized derivatives in animals and humans may reveal the therapeutic potential and suitability of single substances in vivo.

We thank Bianca Jazzar for expert technical assistance.

The authors have no financial conflicts of interest.