To investigate the role of proteinase-activated receptor 4 (PAR-4) in mediating joint inflammation and pain in mice.
To investigate the role of proteinase-activated receptor 4 (PAR-4) in mediating joint inflammation and pain in mice.
Knee joint blood flow, edema, and pain sensitivity (as induced by thermal and mechanical stimuli) were assessed in C57BL/6 mice following intraarticular injection of either the selective PAR-4 agonist AYPGKF-NH2 or the inactive control peptide YAPGKF-NH2. The mechanism of action of AYPGKF-NH2 was examined by pretreatment of each mouse with either the PAR-4 antagonist pepducin P4pal-10 or the bradykinin antagonist HOE 140. Finally, the role of PAR-4 in mediating joint inflammation was tested by pretreating mice with acutely inflamed knees with pepducin P4pal-10.
PAR-4 activation caused a long-lasting increase in joint blood flow and edema formation, which was not seen following injection of the control peptide. The PAR-4–activating peptide was also found to be pronociceptive in the joint, where it enhanced sensitivity to a noxious thermal stimulus and caused mechanical allodynia and hyperalgesia. The proinflammatory and pronociceptive effects of AYPGKF-NH2 could be inhibited by pepducin P4pal-10 and HOE 140. Finally, pepducin P4pal-10 ameliorated the clinical and physiologic signs of acute joint inflammation.
This study demonstrates that local activation of PAR-4 leads to proinflammatory changes in the knee joint that are dependent on the kallikrein–kinin system. We also show for the first time that PARs are involved in the modulation of joint pain, with PAR-4 being pronociceptive in this tissue. Thus, blockade of articular PAR-4 may be a useful means of controlling joint inflammation and pain.
Serine proteinases are a group of proteolytic enzymes whose unique catalytic-reactive domains can hydrolyze specific peptide bonds. In mammals, serine proteinases are involved in wound healing, hemostasis, and degradation of neuropeptides following neurogenic inflammation. The levels of proteinases such as tryptase are known to be elevated in synovial fluid extracted from patients with arthritis (1, 2), in whom the enzymatic activity of proteinases is believed to lead to articular tissue destruction. In addition to their typical degradative effects, proteinases are able to regulate cell signaling via unique receptors called proteinase-activated receptors (PARs). The PARs are a superfamily of G protein–coupled receptors with 7 transmembrane-spanning domains. They are triggered by a novel mechanism in which the proteinase initially hydrolyzes a specific arginine cleavage site located on the extracellular N-terminal loop of the receptor (3). This hydrolytic event unmasks a new N-terminal sequence that then binds to a docking domain on the same receptor while remaining tethered at the other end. One of the major advances in PAR research in recent years has been the development of short synthetic peptides whose sequence mimics the tethered ligand and can therefore act as selective receptor agonists. These PAR-activating peptides are highly potent and resistant to aminopeptidases, making them useful pharmacologic tools for the in vivo study of PARs.
To date, 4 PARs have been cloned: PAR-1, PAR-2, PAR-3, and PAR-4. Different proteinases, ranging from thrombin, trypsin, cathepsins, kallikrein, or even microbial proteinases, are able to activate these receptors, inducing intracellular signaling (4). PAR-4, the most recently defined member of this family of receptors (5), can be activated by thrombin, trypsin, cathepsin G, or activated factor X of the coagulation cascade (4). In addition to its role in thrombin-induced platelet aggregation, PAR-4 activation is thought to be involved in inflammatory mechanisms, as revealed by studies that have demonstrated the effects of PAR-4 agonists on leukocyte recruitment and plasma extravasation (6–8). Although these studies have demonstrated a proinflammatory role for PAR-4 agonists, no study has yet investigated the role of PAR-4 in models of organ pathology. Studying the role of PAR-4 in debilitating and painful conditions such as arthritis could reveal novel signaling pathways related to joint pathology and symptom development. As such, we tested the hypothesis that PAR-4 activation modulates the progression of joint inflammation and alters pain sensitivity in an animal model. Further experiments were carried out to investigate the potential mechanisms of action of PAR-4 in joint inflammation and the potential role of PAR-4 in inflammatory joint pain.
C57BL/6 mice (6–8 weeks old) were purchased from Charles River Canada (Montreal, Quebec, Canada). All mice were housed under constant humidity and temperature, under a 12-hour light–12-hour dark cycle. Institutional animal care committees approved all procedures, which were in compliance with the Canadian Council for Animal Care. At the end of the experiments, the mice were humanely killed using sodium pentobarbital (200 mg/kg intraperitoneally), followed by cervical dislocation.
The PAR-4–activating peptide AYPGKF-NH2, the PAR-4–inactive control peptide YAPGKF-NH2, and the palmitoylated PAR-4 antagonist pepducin P4pal-10 (N-palmitoyl-SGRRYGHALR-NH2) were obtained from the peptide synthesis facility at the University of Calgary. The composition and purity of the peptides were confirmed by high-performance liquid chromatography analysis. All peptides were dissolved in 0.9% NaCl.
Joint blood flow, joint diameter, and thermal and mechanical nociception were determined before (basal) and after intraarticular injections of the PAR-4–activating peptide AYPGKF-NH2 (100-μg intraarticular injection; 5-μl bolus), the control inactive PAR-4 peptide YAPGKF-NH2 (100-μg intraarticular injection; 5-μl bolus), or their vehicle (0.9% NaCl; 5-μl bolus). In order to elucidate the mechanism of PAR-4 action in the joint, experiments with AYPGKF-NH2 were repeated in the presence of either the selective PAR-4 antagonist pepducin P4pal-10 (100 μg intraperitoneally) or the bradykinin antagonist HOE 140 (50 μg/kg intraperitoneally). Antagonists were administered 1 hour prior to the PAR-4–activating peptide. All experimental protocols were carried out on separate groups of mice.
The role of PAR-4 in acute inflammatory hyperemia and edema was examined in the following series of experiments. Under urethane anesthesia, the right knee joint was shaved, swabbed with 100% ethanol, and injected with 2% kaolin (10-μl bolus). The joint was then extended and flexed for 10 minutes to allow cartilage debridement and synovial irritation. Mice then received a single intraarticular injection of 2% carrageenan (10-μl bolus). Ten mice with inflamed tissue were randomly assigned to 2 groups, with 1 cohort of mice being treated with pepducin P4pal-10 (100 μg intraperitoneally) immediately following the induction of inflammation. The control group of mice received a sterile saline injection (0.9% NaCl intraperitoneally). Basal blood flow and joint edema were then measured in both groups of mice over the succeeding 3–4 hours. Six mice (3 mice randomly assigned from each treatment group) were then prepared for histopathologic analysis (see below).
A total of 8 mice were used in the studies assessing blood flow. Under deep urethane anesthesia (50 mg intraperitoneally), the right carotid artery was cannulated and attached to a pressure transducer (Stoelting, Wood Dale, IL) and then a blood pressure monitor (BP-1; World Precision Instruments, Sarasota, FL) to record continuous mean arterial pressure. A small ellipse of skin was removed from the anterior aspect of the right stifle (knee) joint, and all underlying fascia was excised. Hydration of the exposed joint capsule was attained by regular superfusion of 0.9% saline at 37°C. Each mouse was then placed supine on a thermoregulated heating pad (TR-200; Fine Science Tools, North Vancouver, British Columbia, Canada), which maintained core body temperature at 37°C as recorded by a rectal thermometer.
Knee joint blood flow was measured noninvasively by laser Doppler perfusion imaging, as previously described (9, 10). The imaging system (moorLDI2; Moor Instruments, Axminster, UK) involves directing a low-power (2-mW) red (λ = 633 nm) laser beam onto the surface of the exposed knee joint. The scanner head was positioned 15 cm above the mouse, and the gain settings were optimized for mouse joint blood flow measurement (DC gain = 0, flux gain = 2, concentration gain = 2, background threshold = 100). By means of a pivoted mirror, the laser beam scans in a continuous raster pattern over the surface of the joint, and a blood perfusion measurement is made at multiple discrete loci. Circulating erythrocytes in the articular microvasculature cause a pulsed Doppler shift in the frequency of the laser light that is proportional to the velocity of the moving blood cells (flux component). A photodetection system captures the Doppler-shifted laser light, and a central processor calculates tissue blood flow, which is displayed as a color-coded map. Joint scans were obtained before (control) and following intraarticular injection of the test compounds (5-μl bolus total volume). Blood flow images were captured every 10 minutes for up to 2 hours following drug administration.
At the end of the recording period in all blood flow experiments, the mouse was killed by anesthetic overdose (sodium pentobarbital, 80-mg intracardiac injection), and a scan of the dead mouse was obtained. This “biologic zero” (which accounts for tissue optical noise and Brownian motion) was subtracted from all captured images prior to data analysis.
All perfusion images were analyzed using proprietary software (LDI Processing Software; Moor Instruments), and the mean blood flow to the anterior joint capsule was calculated. To obviate any influence due to blood pressure fluctuations, vasomotor changes were expressed as vascular conductance, which was calculated as follows: conductance = blood flow divided by mean arterial pressure. The effects of test agents on joint vascular conductance were expressed as the percentage change in conductance between control and test images.
The knee joint diameter was measured before and at different time points after intraarticular knee injection in 8 mice and was used as an index of edema. Measurements were performed by an investigator who was blinded with regard to the experimental treatments, using a digital caliper with a resolution of 10 μm. The calipers were oriented in a mediolateral plane across the joint line, with minimal or no compression being exerted onto the joint at the time of measurement.
Pain assessment was carried out in a separate cohort of 8 mice. The latency of paw withdrawal to radiant heat stimuli was measured using a Harvard plantar test apparatus (Harvard Apparatus, South Natick, MA), essentially as described by Hargreaves et al (11). Thermal stimulation was applied to the plantar surface of the hind paw of mice that had received intraarticular injections into the knee. Thermal hyperalgesia was defined as a significant decrease in the withdrawal latency compared with basal measurements at different time points after the intraarticular injection.
Mechanical nociception was measured as follows. Mice were placed into individual plastic cages. Von Frey hair filaments with bending forces of 3.61g, 3.84g, and 4.08g were pressed perpendicularly against the plantar skin of the ipsilateral hind paw and held for 5 seconds. This stimulation was repeated 3 times in each hind paw tested, at intervals of several seconds. The responses to these stimuli were ranked as follows: 0 = no response, 1 = move away from the filament, and 2 = immediate flinching or licking of the hind paw. The nociception score was calculated as follows:
The nociception score was measured before and at different time points after the intraarticular injection of test agents (5-μl bolus total volume).
Following blood flow measurement, 6 of the mice with acute inflammation were prepared for histopathologic analysis. Three of the mice had been treated with pepducin P4pal-10 (100 μg intraperitoneally) immediately following inflammation induction, while the remaining 3 mice served as saline-injected controls (0.9% NaCl injected intraperitoneally). The ipsilateral hind limbs were removed and fixed in 10% neutral buffered formalin for 5 days. Tissues were then rinsed in distilled water, and excess muscle was removed. Limbs were then slowly decalcified in 10% formic acid for ∼10 days and embedded in paraffin. Sagittal sections of the whole joint were cut at a thickness of 15 μm, mounted on glass microscope slides, and dried overnight at 50°C. Sections were deparaffinized in xylene and then dehydrated in serial dilutions of ethanol before being rehydrated in water for 1 minute. Sections were then stained with Weigert's hematoxylin and eosin and dehydrated in ethanol, coverslipped, and finally left to dry at room temperature. Sections were viewed under bright-field microscopy and graded by an observer who was blinded with regard to the degree of joint inflammation (i.e., the degree of cellular infiltration and level of pannus formation). Whole-number scoring was used for each parameter, which ranged from 0 (normal) to 3 (severe), for a possible total maximum score of 6. A total of 30 sections from the saline-treated mice and 48 sections from the pepducin P4pal-10 group were scored.
Three mice were anesthetized with 2% isoflurane prior to perfusion through the left ventricle with 10 ml 0.9% saline and then 10 ml 4% paraformaldehyde (PFA). Knee joints were removed immediately and post-fixed in 4% PFA overnight prior to decalcification with Cal-Ex (Fisher Scientific, Fairlawn, NJ). Tissues were cryopreserved in 3 changes of 30% sucrose/phosphate buffered saline (PBS), embedded in OCT compound (Sakura Finetek USA, Torrance, CA), and sectioned to 10 μm on a Microm HM 500 cryostat (Microm, Waldorf, Germany). Sections were washed once for 5 minutes in PBS (pH 7.2) and blocked for 60 minutes in 10% normal donkey serum (Sigma-Aldrich, St. Louis, MO), followed by incubation in a humidity chamber overnight at 4°C with goat polyclonal anti–PAR-4 (1:50 dilution; Santa Cruz Biotechnology, Santa Cruz, CA). Slides were washed 3 times for 5 minutes in PBS, and specific staining was detected using Cy3-labeled donkey anti-goat IgG (Jackson ImmunoResearch, West Grove, PA) or anti–PAR-4 antibody together with the blocking peptide (Santa Cruz Biotechnology). Slides were washed for a final time with PBS and coverslipped with FluorSave (Calbiochem-Novabiochem, San Diego, CA). Imaging was analyzed using OpenLab 3.0 software (Quorum Technologies, Guelph, Ontario, Canada) on a Leica DM6000B (Leica, Wetzlar, Germany). The following controls were carried out to test the specificity of the primary antibody: 1) immunohistochemistry was performed in the absence of the primary antibody, and 2) PAR-4–blocking peptide (1:100 dilution; Santa Cruz Biotechnology) was used together with the primary antibody. No immunofluorescence was detected in either of these control experiments. Positive immunostaining with PAR-4 was confirmed in sections of rat colon, as previously described (12) (data not shown).
All data conformed to a Gaussian distribution and were analyzed using parametric statistics, i.e., Student's t-test, one-way analysis of variance (ANOVA), and two-way ANOVA. A Bonferroni post hoc test was used to determine differences at individual time points. P values less than 0.05 were considered significant. All data points are expressed as the mean ± SEM.
Immunopositive staining for PAR-4 was detected in chondrocytes and subchondral bone associated with the femur and tibia of the mouse knee (Figure 1B). PAR-4 was also detected in the menisci of the normal knee joint, with positive staining appearing throughout the depth of the tissue (Figure 1B). Finally, pronounced PAR-4 staining was also observed in the synovium (Figure 1D). No staining for PAR-4 was observed in slices exposed to both anti–PAR-4 antibody and the blocking peptide (results not shown).
Intraarticular injection of the selective PAR-4 agonist AYPGKF-NH2 caused a gradual increase in knee joint blood flow, which reached a maximal response ∼2 hours after administration (Figure 2A). This hyperemic response to the PAR-4 agonist was statistically significantly different from the response to vehicle control and the PAR-4–inactive peptide YAPGKF-NH2 (P < 0.0001 by two-way ANOVA; n = 8). Similarly, the intraarticular injection of AYPGKF-NH2 caused an increase in joint diameter, characteristic of edema (Figure 2B). This increase was maximal 2 hours after the injection, and at ∼4 hours was still significantly higher than the effect of the saline injection or injection of the control peptide YAPGKF-NH2 (P < 0.05 by two-way ANOVA; n = 8) (Figure 2B).
The AYPGKF-NH2–mediated increase in joint vascular conductance and edema in the mice was significantly attenuated by treatment with the selective PAR-4 antagonist pepducin P4pal-10 (Figures 2C and D), suggesting that PAR-4 activation is responsible for the AYPGKF-NH2–induced inflammatory response. Pepducin P4pal-10 given alone had no effect on joint edema.
Intraarticular injection of the selective PAR-4 agonist AYPGKF-NH2 caused a significant decrease in paw withdrawal latency in response to a plantar thermal stimulus, compared with the effects of the control peptide YAPGKF-NH2 or vehicle (saline) (Figure 3A). This decreased withdrawal latency is characteristic of thermal hyperalgesia. In response to plantar mechanical stimulation with a small-size von Frey hair filament, mice that had received a knee joint injection of saline or the control peptide YAPGKF-NH2 did not show a nociceptive response (Figure 3B). In contrast, mice that had received an intraarticular injection of the PAR-4–activating peptide AYPGKF-NH2 showed a strong nociceptive response to this small-size von Frey filament (Figure 3B). Such a response is characteristic of allodynia. When stimulated with larger-size von Frey filaments, mice injected with saline or control peptide YAPGKF-NH2 showed a nociceptive response proportional to the size of the filament, and this response did not vary significantly by time after injection (Figures 3C and D). However, mice injected with the PAR-4–activating peptide AYPGKF-NH2 showed a significant increase in nociceptive response to middle-size (3.84g) (Figure 3C) and large-size (4.08g) filaments (Figure 3D) compared with basal values. This increased nociceptive response is characteristic of mechanical hyperalgesia.
AYPGKF-NH2–induced thermal hyperalgesia (decrease in withdrawal latency) was not observed in mice that received systemic treatment with the PAR-4 antagonist pepducin P4pal-10 (Figure 4A). Similarly, AYPGKF-NH2–induced mechanical allodynia (increased nociception score in response to small-size von Frey filament) and hyperalgesia (increased nociception score in response to medium- and large-size von Frey filaments) were not observed in mice treated with pepducin P4pal-10 (Figures 4B–D). When given alone, pepducin P4pal-10 had no effect on thermal sensitivity or mechanosensitivity (see Supplementary Figure 1, available on the Arthritis & Rheumatism Web Site at http://www3.interscience.wiley.com/journal/76509746/home). Taken together, these results show that PAR-4 activation is responsible for AYPGKF-NH2–induced allodynia and hyperalgesia.
An AYPGKF-NH2–mediated increase in joint vascular conductance and edema was completely inhibited by treatment of mice with the bradykinin B2 receptor antagonist HOE 140 (Figures 2C and D). Similarly, thermal and mechanical hyperalgesia, characterized by decreased withdrawal latency and an increased nociception score, respectively, were fully inhibited in mice treated systemically with HOE 140 (Figure 4). HOE 140 by itself had no significant effect on joint edema, thermal sensitivity, or mechanosensitivity in these normal mice.
Three hours after intraarticular injection of kaolin/carrageenan, there was a pronounced inflammatory reaction in the joint, characterized by synovial hyperplasia and an influx of inflammatory cells (see Supplementary Figure 2, available on the Arthritis & Rheumatism Web site at http://www3.interscience.wiley.com/journal/76509764/home). In contrast, pretreatment of mice with acute tissue inflammation with pepducin P4pal-10 ameliorated the severity of tissue inflammation. Histologic scoring of whole joint sections under blinded conditions revealed that the levels of cellular infiltration and synovial hyperplasia were significantly reduced (P < 0.0001 by Student's unpaired t-test; n = 30–48) in mice pretreated with the PAR-4 antagonist (Table 1).
Blood flow studies showed that treatment of kaolin/carrageenan-injected knees with the PAR-4 antagonist pepducin P4pal-10 reduced the hyperemia associated with acute joint inflammation (Figure 5A). Joint edema was also significantly reduced at 2 hours and 3 hours after kaolin/carrageenan injection in mice treated with pepducin P4pal-10 (Figure 5B).
One of the pronounced hallmarks of arthritis is the extensive destruction and overall abnormal remodeling of joint tissues. High levels of proteinases released into diseased joints are likely the primary driving force responsible for these structural changes. In addition to the enzymatic activity of proteinases, it has been eloquently demonstrated that these agents can act as signaling molecules by triggering a specialized family of receptors called PARs (for review, see ref. 3). This study provides the first evidence that PAR activation modulates pain sensitivity in joints and reveals a vital role of PAR-4 in promoting joint inflammation.
The results of the immunolocalization experiments presented here clearly show for the first time that PAR-4 is expressed extensively throughout the mouse knee joint in tissues such as the cartilage, subchondral bone, menisci, and synovium. Articular activation of PAR-4 with the selective peptide agonist AYPGKF-NH2 caused a gradual rise in synovial blood flow with concomitant edema formation, which was maximal ∼2 hours after intraarticular injection. These inflammatory reactions to the peptide could be blocked by pretreatment with the PAR-4 antagonist pepducin P4pal-10. These results are consistent with observations in the hind paw and the mesenteric circulation, in which PAR-4 activation leads to edema and granulocyte infiltration (6–8). Considering the strong expression of PAR-4 in endothelial cells (8), it could be hypothesized that in the joint, PAR-4 activation–induced edema is attributable to endothelial cell activation and subsequent increased vascular permeability; however, the expression of PAR-4 in the various cell types described here indicates that there are multiple targets in which PAR-4 agonists could induce inflammation.
Other PARs have similarly been implicated in the pathogenesis of arthritis. PAR-4, for example, has been detected in the synovial tissue of patients with rheumatoid arthritis and patients with osteoarthritis (13), and thrombin-induced PAR-1 activation led to increased inflammatory chemokine expression by human synovial fibroblasts (14). In other studies, antigen-induced arthritis was found to be less severe in PAR-1–knockout mice compared with wild-type controls (15). Similarly, PAR-2 has been implicated in joint disease, as evidenced by its isolation from the joints of patients with arthritis (16, 17) and by the observation that PAR-2 activation leads to proinflammatory changes in the knee joints in animal models (18, 19). Thus, the data presented here extend the role of PARs in arthritis by showing that PAR-4 is also proinflammatory in joints.
PAR activation is known to modulate pain in animal models (20–22); however, the role of PARs in joint pain has not previously been explored. In this study, it was demonstrated that intraarticular injection of a PAR-4 agonist caused a heightened pain response to both thermal and mechanical stimuli in mice. This allodynic and hyperalgesic effect of AYPGKF-NH2 was inhibited by treating the mice with the selective PAR-4 antagonist pepducin P4pal-10. These pain-causing effects of PAR-4 activation are in contrast to the previously described antinociceptive effects of AYPGKF-NH2 in the rat hind paw (20). In addition to the obvious differences in species and injection sites between the studies, the most likely explanation for these divergent results is that in the present study, a higher dose of the PAR-4 agonist was used. Thus, it appears that low-dose AYPGKF-NH2 has the ability to reduce nociceptive pain, whereas higher concentrations of the peptide cause hyperalgesia and allodynia.
The mechanism of action of PAR-4 in causing joint inflammation and pain appears to involve the kallikrein–kinin system, because the bradykinin antagonist HOE 140 attenuated edema, hyperemia, and nocifensive responses to AYPGKF-NH2. This observation is consistent with findings of a previous study showing that PAR-4–mediated paw edema could be inhibited by pretreatment with a bradykinin B2 antagonist (7). Because PAR-4 has been detected on the surface of leukocytes, endothelial cells, and vascular smooth muscle cells (8), it is hypothesized that PAR-4–activating peptides stimulate these cells to release kallikreins, which then cleave kininogens to produce active kinins, which ultimately bind to bradykinin receptors, resulting in vasodilatation and increased vascular permeability. Indeed, bradykinin is known to be vasoactive in knee joints, where it causes synovial vasodilatation and protein extravasation (23–25). The role of the kallikrein–kinin system in promoting joint pain is also corroborated by other studies showing that bradykinin causes peripheral sensitization of knee joint afferent nerves, leading to enhanced pain sensation (26, 27). Other PARs such as PAR-1 and PAR-2 produce pain and inflammation by causing the secondary release of proinflammatory neuropeptides from sensory neurones and by activating connective tissue mast cells (28–31). In contrast, however, the inflammatory actions of PAR-4 are not neurogenically driven, nor do they involve mast cell degranulation (6).
Experiments were also undertaken to determine whether blockade of PAR-4 with a selective antagonist could inhibit disease severity in a model of acute synovitis. Pepducin P4pal-10 treatment of mice with kaolin/carrageenan–induced inflammation significantly ameliorated the severity of joint inflammation, as evidenced by a reduction in pannus formation and inflammatory cell infiltration. The antiarthritic effects of PAR-4 antagonism were consistent throughout the whole depth of the joint as well as in both lateral and medial compartments. In other experiments, it was shown that pretreating acutely inflamed knees with pepducin P4pal-10 attenuated the hyperemia and edema associated with this model. These results clearly indicate that blockade of PAR-4 has a beneficial effect on the pathogenesis of acute synovitis.
In summary, this study has shown that the triggering of PAR-4 caused a proinflammatory and painful reaction in the mouse knee joint. Blockade of bradykinin receptors attenuated the physiologic activity of AYPGKF-NH2, suggesting that the kallikrein–kinin system is involved in mediating the nocifensive and inflammatory responses to PAR-4 activation. Finally, blockade of PAR-4 with a selective antagonist ameliorated pannus formation and leukocyte infiltration in response to an acute inflammatory insult, implicating PAR-4 as one of the major proinflammatory signals in this model. As such, PAR-4 may be a useful target for the treatment of inflammatory joint disease and associated pain.
Dr. McDougall had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. McDougall, Vergnolle.
Acquisition of data. McDougall, Zhang, Cellars, Joubert, Dixon.
Analysis and interpretation of data. McDougall, Zhang, Joubert, Vergnolle.
Manuscript preparation. McDougall, Dixon, Vergnolle.
Statistical analysis. McDougall, Zhang, Joubert.