To examine synovial fluid as a site for generating citrullinated antigens, including the candidate autoantigen citrullinated α-enolase, in rheumatoid arthritis (RA).
To examine synovial fluid as a site for generating citrullinated antigens, including the candidate autoantigen citrullinated α-enolase, in rheumatoid arthritis (RA).
Synovial fluid was obtained from 20 patients with RA, 20 patients with spondylarthritides (SpA), and 20 patients with osteoarthritis (OA). Samples were resolved using sodium dodecyl sulfate–polyacrylamide gel electrophoresis, followed by staining with Coomassie blue and immunoblotting for citrullinated proteins, α-enolase, and the deiminating enzymes peptidylarginine deiminase type 2 (PAD-2) and PAD-4. Proteins from an RA synovial fluid sample were separated by 2-dimensional electrophoresis, and each protein was identified by immunoblotting and mass spectrometry. Antibodies to citrullinated α-enolase peptide 1 (CEP-1) and cyclic citrullinated peptide 2 were measured by enzyme-linked immunosorbent assay.
Citrullinated polypeptides were detected in the synovial fluid from patients with RA and patients with SpA, but not in OA samples. Alpha-enolase was detected in all of the samples, with mean levels of 6.4 ng/μl in RA samples, 4.3 ng/μl in SpA samples, and <0.9 ng/μl in OA samples. Two-dimensional electrophoresis provided evidence that the α-enolase was citrullinated in RA synovial fluid. The citrullinating enzyme PAD-4 was detected in samples from all 3 disease groups. PAD-2 was detected in 18 of the RA samples, in 16 of the SpA samples, and in none of the OA samples. Antibodies to CEP-1 were found in 12 of the RA samples (60%), in none of the SpA samples, and in 1 OA sample.
These results highlight the importance of synovial fluid for the expression of citrullinated autoantigens in inflammatory arthritis. Whereas the expression of citrullinated proteins is a product of inflammation, the antibody response remains specific for RA.
Antibodies to citrullinated proteins (ACPAs) are highly specific for rheumatoid arthritis (RA) and are a powerful tool for its diagnosis and for prediction of disease severity. In addition, the study of ACPAs enables incorporation of risk factors, including smoking and presence of the shared epitope, into a common etiopathogenic model in which citrullination and the generation of antibodies play an intimate role in the pathogenesis of RA (for review, see ref.1).
Citrullinated proteins are formed when arginine residues are deiminated by peptidylarginine deiminase (PAD). Five PADs have been identified in humans (2), of which PAD-2 and PAD-4 are found in rheumatoid synovial fluid cells (3) and in synovial membrane (4, 5). PADs are calcium dependent (2), and are thus more likely to be active in the extracellular compartment, where calcium concentrations are higher. Cell death may result in increased PAD activity, because the loss of membrane integrity will both increase intracellular calcium concentration and enable extracellular leakage of PAD enzymes.
Citrullinated proteins have been detected in the synovial membrane of patients with various forms of arthritis (6) and in other inflamed tissues (7), suggesting that, whereas citrullination is associated with inflammation in general, the development of ACPAs is specific to RA. In patients with RA, ACPA-producing plasma cells have been detected in the synovial membrane (8), and higher concentrations of ACPAs in the rheumatoid joint compared with the serum have also been reported (8, 9). This suggests that synovial citrullinated proteins are driving a local production of antibodies, and that the resulting immune complexes contribute to the chronic inflammation in the rheumatoid joint.
Previous studies of citrullinated antigens have tended to focus on synovial membrane (10–12). In the present study we investigated synovial fluid as a site of extracellular deimination, focusing on the candidate autoantigen citrullinated α-enolase (13). We recently mapped the autoantibody response to an immunodominant peptide, which we have termed citrullinated α-enolase peptide 1 (CEP-1). Antibodies to CEP-1 are closely correlated with antibodies to cyclic citrullinated peptide 2 (CCP-2), are highly specific for RA, and have a diagnostic sensitivity of ∼50% depending on the cohort of patients studied (14). Having identified a major epitope, we now have reagents to study the distribution of native and citrullinated α-enolase and its antibodies in clinical samples. In addition, CEP-1 is derived from a real protein, as opposed to the purely artificial peptides that comprise the CCP-2 assay, and, as such, may reflect the pathologic processes occurring in the joints of patients with RA.
Synovial fluid was removed from the knees of patients at the time of therapeutic arthrocentesis. Samples were obtained, following the patients' provision of informed consent and approval from the local ethics committee, from 20 patients with RA and 20 patients with spondylarthritides (SpA). These patients were attending the Rheumatology Clinic at Karolinska University Hospital (Stockholm, Sweden). Synovial fluid samples from 20 patients with osteoarthritis (OA), selected for the absence of cellularity and thus serving as noninflammation controls, were obtained from the knees of patients attending the Department of Rheumatology at Lund University Hospital (Lund, Sweden), following the provision of informed consent and approval from the local ethics committee. For 2-dimensional electrophoresis, an additional rheumatoid synovial fluid sample was obtained from the knee of a patient attending Charing Cross Hospital in London, UK.
After centrifugation to remove cells, the synovial fluid samples were stored at −70°C until used. After thawing, the samples were digested with hyaluronidase type IV-S (50 μg/ml; Sigma, St. Louis, MO), vortexed, and passed through a 0.2-μm Ministart filter unit (Sartorius, Hannover, Germany). Protease inhibitor cocktail (Sigma) and EDTA were added to final concentrations of 10 μl/ml and 50 mM, respectively, and the samples were stored at −20°C.
A rabbit polyclonal anti–CEP-1 antibody was generated at Cambridge Research Biochemicals (Ely, UK). Briefly, 2 rabbits were immunized subcutaneously every 2 weeks, for 10 weeks, with 200 μg keyhole limpet hemocyanin (KLH)–conjugated CEP-1 (peptide:KLH ratio 1:1) in Freund's incomplete adjuvant per booster. Blood was collected 7 days after each injection, and sera were analyzed for the presence of anti–CEP-1 antibodies by enzyme-linked immunosorbent assay (ELISA) (14).
When antibody titers had reached significant levels, animals were killed and blood samples were harvested. The crude antisera were depleted of cross-reactive antibodies by chromatography on a thiopropyl-Sepharose column conjugated to a control peptide in which the citrulline residues had been replaced with arginine residues. The unbound fraction was affinity purified on a second thiopropyl-Sepharose column, conjugated to CEP-1. Anti–CEP-1–specific antibodies were eluted and further depleted of nonspecific antibodies in 3 subsequent passages through the depleting column. Bound antibodies were eluted from the column with 27 ml of 0.1M glycine/HCl, pH 2.5, and the flow-through was collected as the unbound fraction.
Both antibody fractions demonstrated similar preferential reactivity for citrullinated α-enolase, with the glycine eluate having a greater sensitivity for citrullinated α-enolase (as demonstrated by immunoblotting of uncitrullinated and in vitro citrullinated α-enolase at the same dilution, 1:400 [data not shown]). Both fractions were stored at −20°C until further used.
The full-length human α-enolase coding sequence was amplified, by polymerase chain reaction (PCR), from the complementary DNA of HL60 cells differentiated with vitamin D3 (13). The PCR forward and reverse primers contained the Bam HI and Xho I restriction sites, respectively (AGTTGGATCCTCTATTCTCAAGATCCATGCCA and ATCCTCGAGTTACTTGGCCAAGGGRTTTCTGAAGTTCCTG, respectively). The PCR product was ligated in-frame into the multiple cloning site of the plasmid expression vector pGEX 6P3 (GE Healthcare, Bucks, UK), 3′ to the glutathione S-transferase (GST) coding site.
Expression of GST α-enolase was induced in the protease-deficient BL21 strain of Escherichia coli. Briefly, the bacteria were grown at 28°C. Once an optical density at 600 nm (OD600 nm) of 0.6–0.8 had been reached, IPTG was added to the culture to a final concentration of 0.1 μM to induce GST α-enolase expression. Cultures were incubated for a further 18 hours, and fusion protein was purified from bacterial lysates according to the plasmid manufacturer's instructions. PreScission Protease (GE Healthcare), used to cleave the GST from the α-enolase, was pelleted with the excised GST, using glutathione–Sepharose 4B (GE Healthcare). Purified α-enolase, as determined by Coomassie blue staining and tandem electrospray mass spectrometry (MS), was dialyzed against phosphate buffered saline (PBS), and the protein concentration was measured by bicinchoninic acid assay (Pierce, Rockford, IL).
Citrullinated proteins were identified using an Anti–Modified Citrulline (AMC) detection kit (Upstate Biotechnology, Lake Placid, NY) according to the manufacturer's instructions. Rabbit anti–α-enolase (H300; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti–PAD-2 (Abcam, Cambs, UK), and goat anti–PAD-4 (Abcam) were used at 1:200 in incubations overnight at 4°C.
Synovial fluid proteins were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) using 10-well 4–12% Bis-Tris precast gels (Invitrogen, Paisley, UK). Gels were either stained with Coomassie blue or transferred to nitrocellulose membranes for immunoblotting. Following incubations with primary and secondary antibodies, the membranes were blocked in 5% milk in PBS/0.1% Tween, blotted with antibodies diluted in blocking buffer, and washed in PBS/0.1% Tween. The preparations were then developed using enhanced chemiluminescence. Anti-goat and anti-rabbit secondary antibodies conjugated to horseradish peroxidase (Dako, Glostrup, Denmark) were diluted to 1:5,000 and 1:3,000, respectively, and reacted with membranes for 1 hour at room temperature.
To prevent masking by abundant comigrating proteins, synovial fluid samples were depleted of albumin and IgG, using the Proteoextract Albumin/IgG Removal Kit (Calbiochem, Notts, UK), prior to immunoblotting for the quantification of α-enolase and the detection of PAD-2 and PAD-4. Briefly, synovial fluid, diluted 1:9 in the supplied binding buffer, was added at 560 μl per column. Proteins in the flow-through were immunoblotted for α-enolase without concentration or were immunoblotted at a 10-fold concentration, attained using a 0.5-ml concentrator with a molecular weight cutoff of 10,000 kd (Vivascience, Stonehouse, UK). Proteins in the flow-through were also immunoblotted for PAD-2 or PAD-4. Poly-His–tagged recombinant human PAD-2 and PAD-4 (each 50 ng/well) (3) were used as positive controls. In addition, nondepleted synovial fluid samples (diluted 1:30 in PBS) from 20 patients with RA, 20 patients with SpA, and 20 patients with OA were applied directly (2 μl/dot) onto a single nitrocellulose membrane for detection of citrullinated proteins. Immunoblotting was performed using the AMC detection kit in accordance with the manufacturer's instructions.
Synovial fluid samples depleted of IgG and albumin and boiled in 2× Laemmli buffer were loaded in duplicate onto 15-well 4–12% Bis-Tris precast gels (Invitrogen). On each gel, a range of purified α-enolase standards was also loaded. Following SDS-PAGE and electrotransferesis, nitrocellulose membranes were blocked in 5% milk/PBS/0.1% Tween overnight, incubated in rabbit anti–α-enolase, diluted 1:200 in blocking buffer, washed in PBS/0.1% Tween, incubated with anti-rabbit secondary antibodies (Dako), washed in PBS/0.1% Tween, and developed using enhanced chemiluminescence. Concentrations of α-enolase were calculated by densitometry.
Films (Amersham Hyperfilm; GE Healthcare) were scanned using a scanning densitometer (GS-710 Calibrated Imaging Densitometer; Bio-Rad, Hercules, CA). Fifty-kilodalton bands were quantified for total pixel volumes (corresponding to total amounts of α-enolase) using Phoretix software (Nonlinear Dynamics, Ltd., Newcastle, UK). Pixel volumes of bands at 50 kd in lanes containing recombinant α-enolase were used to create a standard curve of total α-enolase levels relative to pixel volume, achieved using Prism software (GraphPad Software, San Diego, CA). From this curve, total α-enolase levels in each of the wells loaded with synovial fluid samples were deduced, and a mean α-enolase level for each patient sample was calculated from duplicate lanes. This assay was performed for each individual 15-well gel, to avoid generation of artefacts resulting from intergel and interfilm variation. Each gel was loaded with 4 samples, added in duplicate. Because of the dilutions and volumes of the synovial fluid samples, the concentration of α-enolase in the fluid was multiplied by 1.5. Thus, the densitometric value representing the lowest value on the standard curve, 0.6 ng/μl, equated to 0.9 ng/μl of α-enolase in the synovial fluid.
To characterize citrullinated proteins by tandem MS, an RA synovial fluid sample with an abundance of citrullinated proteins was fractionated by liquid-phase IEF (Zoom IEF; Invitrogen). Briefly, synovial fluid was mixed with 40 μl carrier ampholytes (Invitrogen) and 80 μl of 1M dithiothreitol, and made up to a volume of 4 ml in IEF buffer, giving a final concentration of 0.5 mg/ml protein. Polyacrylamide Zoom disks (Invitrogen), each with the appropriate pH, were added to the assembly, to yield 5 liquid-phase protein fractions (pH 3.0–4.5, 4.6–5.3, 5.4–6.1, 6.2–6.9, and 7.0–10.0). Anode (pH 2.5–2.9) and cathode (pH 3.0–10.0) buffers were added to their respective electrodes at a volume of 17.5 ml. Samples were added to each of the chambers (650 μl/chamber), and IEF was performed according to the manufacturer's guidelines, consisting of 100V for 20 minutes, 200V for 80 minutes, and 800V for 80 minutes. Each fraction was further separated by SDS-PAGE, followed by either staining with Coomassie blue or immunoblotting.
Coomassie blue–stained proteins were excised, digested with trypsin, and characterized by tandem MS as previously described (13). For these experiments, semipurified α-enolase (Hytest, Turku, Finland), with or without prior digestion with PAD, was used as the positive control.
Ninety-six–well plates (Maxisorp; Nunc, Roskilde, Denmark) were coated with CEP-1 (CKIHA-X-EIFDS-X-GNPTVEC, where X represents citrulline) or the arginine-containing control peptide (CKIHAREIFDSRGNPTVEC) at 10 μg/ml (diluted in a 50-mM carbonate buffer, pH 9.6), or with 2% bovine serum albumin (BSA) in carbonate buffer, and incubated overnight at 4°C. Wells were washed with PBS/0.05% Tween and blocked with 2% BSA (diluted in PBS) for 3 hours at room temperature. Sera were diluted 1:100 in radioimmunoassay (RIA) buffer (10 mM Tris, 1% BSA, 350 mM NaCl, 1% Triton-X, 0.5% sodium deoxycholate, 0.1% SDS), added in duplicate, and incubated for 1.5 hours at room temperature. Plates were washed as described above and incubated for 1 hour at room temperature with peroxidase-conjugated mouse anti-human IgG (Hybridoma Reagent Laboratory, Baltimore, MD), diluted 1:1,000 in RIA buffer.
After a final wash (PBS/0.05% Tween), bound antibodies were detected with tetramethylbenzidine substrate (KPL, Gaithersburg, MD). The reaction was stopped by the addition of 1M H2SO4, and absorbance was measured at 450 nm in a Multiscan Ascent microplate reader (ThermoLabsystems, Helsinki, Finland). A control serum was included on all plates to correct for plate-to-plate variation.
For each serum tested, background OD450 nm values (i.e., in wells coated with 2% BSA in carbonate buffer) were subtracted from peptide OD450 nm values. OD values higher than 0.2 were considered to be peptide positive. Anti-CCP antibody status was analyzed using the CCP-2 kit (Eurodiagnostica, Malmo, Sweden) according to the manufacturer's instructions.
Immunoblotting with the AMC detection kit demonstrated that both the RA samples and the SpA samples contained citrullinated proteins across the entire range of molecular mass, although none were detected in the OA samples (individual immunoblotting results are available upon request from the corresponding author). These findings were also confirmed by dot-blot analysis of synovial fluid samples from all patients (Figure 1), assessed on a single membrane to eliminate possible variability between separate membranes. Citrullinated proteins were detected by dot-blot in 19 of 20 RA samples, in 17 of 20 SpA samples, and in only 2 OA specimens.
Immunoblotting demonstrated the presence of α-enolase in both the RA samples and the SpA samples; however, α-enolase was barely detectable in the OA samples (Figure 2). We used densitometry and a standard curve of recombinant protein to quantify the α-enolase in all 60 samples. As shown in Figures 2 and 3, higher levels of α-enolase were detected in samples from both the patients with RA (mean 6.4 ng/μl, range 1.2–15.4) and the patients with SpA (mean 4.3 ng/μl, range 1.1–9.2) as compared with the OA controls (<0.9 ng/μl in all samples).
Antibodies to CEP-1 were detected in 12 of the 20 RA synovial fluid samples (60%) but in only 1 OA sample (5%) and in none of the SpA samples (Figure 4). None of the RA patients had antibodies to the arginine-containing control peptide, while 1 OA synovial fluid sample and 1 from a patient with SpA were positive for the control peptide. All 12 RA patients who were anti–CEP-1 positive were also anti–CCP-2 positive.
To characterize the citrullinated proteins present in rheumatoid synovial fluid, we selected an RA synovial fluid sample containing heavily citrullinated proteins. This sample was resolved by liquid-phase IEF, followed by SDS-PAGE. Bands from Coomassie blue–stained gels were identified by tandem MS (Figure 5a). Immunoblotting with the AMC antibody demonstrated the presence of citrullinated proteins with molecular masses between 35 kd and 250 kd (Figure 5b). The concentration of these proteins was greatest in the more acidic fractions, particularly between pH 3.0 and pH 4.6, whereas Coomassie blue–stained material was most abundant in fractions with a pH higher than 5.5. This is consistent with the acidic shift associated with deimination of arginine.
We identified α-enolase with 2 different antibodies: the rabbit anti–α-enolase H300, which reacted with purified α-enolase equally in its uncitrullinated and citrullinated forms, and the anti–CEP-1 antibody, which preferentially reacted with the in vitro citrullinated form. The H300 antibody reacted with a 50-kd polypeptide, which was most abundant in the IEF fractions with pH of 6.2–6.9 and 7.0–10.0 (Figure 5c). This is consistent with the characteristics of native α-enolase, which has a calculated pH of 6.9. The reactivity of the anti–CEP-1 antibody showed a shift toward more acidic forms of the molecule, including a stronger reaction with the 5.4–6.1 fraction and a weaker reaction with the 7.0–10.0 fraction (Figure 5d), compared with the reactions with the H300 antibody. This is consistent with the presence of some partially citrullinated α-enolase.
All bands visible by Coomassie blue staining were excised and digested in gel with trypsin, and the resulting peptides were sequenced by tandem MS; the proteins identified are shown in Figure 5a. No citrulline residues were observed with the use of tandem MS in peptides derived from any of the synovial fluid proteins, whereas peptides containing deiminated arginine residues were readily detected in a PAD-treated sample of α-enolase analyzed in parallel, suggesting that the stoichiometry of in vivo citrullination is relatively low.
We investigated the presence of enzymes capable of deimination of arginine, using antibodies against PAD-2 and PAD-4 in immunoblots of synovial fluid cultures that had been depleted of IgG and albumin and subsequently concentrated. Prominent bands were seen when blotting for PAD-4 in all of the synovial fluid samples. However, reactivity was generally lower in the OA samples. In comparison, reactivity with the anti–PAD-2 antibody was detectable in 18 of 20 RA samples, in 16 of 20 SpA samples, and in none of the OA samples. Both PAD-2 and PAD-4 migrated at a molecular weight similar to the 75-kd molecular weight standard, just below the recombinant PAD-2– and PAD-4–positive controls, the difference in mass being accounted for by the poly-His tag (Figure 6).
Although the antibodies showed good specificity for their respective controls, there were widespread cross-reactions with other proteins in the synovial fluid. Some of the samples (see OA sample 3 in Figure 6) showed a reaction of the PAD-2 antibody with a polypeptide that migrated just above the 75-kd marker and the positive control. This was interpreted as not being PAD-2, because of the increased molecular mass. There was no reaction with the conjugate controls (results not shown). These findings are consistent with the notion that the extracellular compartment of synovial fluid in patients with inflammatory arthritis is a site for generation of citrullinated proteins by either or both of these enzymes.
In this study we have demonstrated that synovial fluid from patients with RA and from patients with SpA is characterized by an abundance of citrullinated proteins. The presence of abundant citrullinated proteins in synovial fluid from patients with SpA and patients with RA, compared with a lack of citrullinated proteins in OA synovial fluid, suggests that, similar to that in the synovial membrane, the presence of elevated levels of citrullinated proteins may be characteristic of inflammation, and not restricted to RA.
Our identification of α-enolase as a candidate citrullinated antigen in synovial fluid suggests that this previously neglected compartment of the rheumatoid joint is a site of expression of autoantigens. This observation is consistent with the findings from previous studies, which have demonstrated the presence of citrullinated fibrin in synovial fluid from patients with RA, but not in OA synovial fluid (15, 16), and is consistent with a recent description of mutated and citrullinated vimentin in RA synovial fluid (17), although in that report, the material included lysed cells. We also found that the mean α-enolase levels in rheumatoid synovial fluid were increased at least 6-fold in the samples from RA patients, and increased at least 4-fold in the SpA synovial fluid. Because the levels in the OA samples fell below the lowest concentration of recombinant α-enolase on the standard curve, it may be that the ratio of concentrations in the 2 types of inflammatory arthritis tested herein may be even higher in comparison with OA.
The higher levels of anti–CEP-1 antibodies found in RA patients (60% of the RA synovial fluid samples) compared with the controls supports the concept that expression of citrullinated proteins is a product of inflammation, whereas the antibody response remains specific to RA. In this study we did not examine the relative concentration of anti–CEP-1 antibodies in synovial fluid compared with that in the serum. A highly significant enrichment has already been demonstrated for anti–CEP-1 and other ACPAs in more than 300 paired serum and synovial fluid samples in a separate study (Snir O, et al: unpublished observations).
We demonstrated reactivity with the anti–CEP-1 antibody in acidic fractions containing proteins with pH below the calculated pH of unmodified α-enolase. Reactivity with the AMC detection antibody was also evident in the more acidic fractions. This is consistent with some of the α-enolase being citrullinated in vivo, and therefore gaining acidity. However, we were unable to demonstrate specific citrullinated residues in synovial fluid α-enolase by tandem MS.
Similar findings were obtained for other abundant proteins, mainly well-documented serum proteins, in which citrullination was demonstrated in comigrating polypeptides, as revealed by staining with the AMC antibody. The only sample in which deiminated arginines were demonstrable by MS was the in vitro citrullinated α-enolase, in which every arginine was deiminated, as observed both in this study and in our previous report (13). The failure to detect citrulline residues in the deiminated synovial fluid proteins is probably due to partial deimination, to the low stoichiometry of in vivo citrullination, and to the distribution of the modification over different arginines in different molecules. This is entirely consistent with the findings from another recent study, in which deimination was detected in synovial membrane proteins by both electrophoretic shift assay and staining with the AMC antibody, but not by MS, even in relatively abundant proteins such as citrullinated fibrinogen (12).
This is the first report demonstrating the presence of extracellular PAD-2 and PAD-4 in the synovial fluid, although it required removal of the abundant serum proteins, and subsequent concentration, to demonstrate the enzymes clearly. The anti–PAD-4 antibody reacted unequivocally at the appropriate molecular weight with all of the synovial fluid samples, with fainter bands visible in the OA samples, which again supports the hypothesis that synovial fluid is a site of active deimination in the presence of inflammation. In comparison, reactivity with the anti–PAD-2 antibody was only detectable in the RA and SpA samples. Therefore, it cannot be discounted that PAD-2 is also present in the OA samples, but was below the levels of detection used in this study. If both PAD-2 and citrullinated proteins are indeed absent in OA samples, the hypothesis can justly be made that it is the PAD-2 that is responsible for the deimination of the extracellular citrullination of synovial fluid proteins.
The presence of polypeptides of other molecular weights that was detected with both the anti–PAD-2 and anti–PAD-4 antibodies raises questions regarding the specificity of these reagents. Therefore, caution should be taken when using these reagents in assay systems such as immunohistochemistry, since there is no control for molecular weight, and therefore, there is the possibility of false-positive results.
In the present study we have provided evidence that the synovial fluid from patients with inflammatory arthritis is a site of expression of citrullinated proteins and up-regulation of α-enolase. The presence of PAD-2 and PAD-4, in which the extracellular environment would favor activation of the enzymes, supports the hypothesis that the synovial fluid may be an important site of expression of citrullinated proteins in inflammatory arthritis. The restriction of the immune response to patients with RA may help explain the chronic autoimmune response in the rheumatoid joint, characteristic of this disease.
Dr. Venables 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. Kinloch, Lundberg, Lim, Venables.
Acquisition of data. Kinloch, Lundberg, Wait, Wegner, Lim, Zendman, Saxne, Malmström, Venables.
Analysis and interpretation of data. Kinloch, Lundberg, Wait, Wegner, Lim, Saxne, Malmström, Venables.
Manuscript preparation. Kinloch, Lundberg, Wait, Saxne, Malmström, Venables.
We thank Ms Alex Martin (Imperial College, London) for her expert technical support, and the European Union's Sixth Framework Programme project AutoCure for providing a forum for discussion and collaboration.