Elevated levels of fibrinogen-derived endogenous citrullinated peptides in synovial fluid of rheumatoid arthritis patients

Synovial fluid samples were obtained from 11 RA patients and 10 patients with other inflammatory diseases (four with psoriatic arthritis, two with ankylosing spondylitis, two with gout, one with osteoarthritis and one with ochronosis). Synovial fluid was obtained by joint punctures (arthrocentesis), using needles with a diameter from 1.6 to 2.2 mm. A sample was immediately centrifuged at 2,500 × g for 10 minutes at 4°C. Supernatant and pellet were separately stored at -80°C. All procedures were performed within two hours. All samples were collected for diagnostic purposes and all patients gave informed consent that the samples could also be used anonymously for research purposes, in accordance with the ethical guidelines as established by the ethics committee of the University Medical Centre Ljubljana (Slovenia).

Peptides were isolated from synovial fluid based on the protocol recently described by Kamphorst et al. [15]. First, intact proteins were removed from the synovial fluid by ultracentrifugation. Spin filter columns with a 10 kD cutoff were first washed with 100 μl ethanol by centrifugation for 10 minutes at 11,000 × g at 4°C and then with 100 μl 50 mM NH₄HCO₃ pH 7, also for 10 minutes at 11,000 × g at 4°C. Next, 50 μl of synovial fluid was mixed with 235 ul of 50 mM NH₄HCO₃ pH 7 and 15 μl DMSO and was passed through the spin filter by centrifugation for 60 minutes at 11,000 × g at 4°C. The spin filter was then washed again with 100 μl 50 mM NH₄HCO₃ pH 7 for 10 minutes at 11,000 × g at 4°C and the filtrates of the last two steps were combined and acidified with 5 μL formic acid (FA). Next, Sep-Pak C18 (Waters Corp., Milford, MA, USA) disposable cartridges with a capacity of 50 mg were activated with 1 ml acetonitrile (ACN) and conditioned with 1 ml 2.5% ACN/1% FA. After passing the filtrates over the cartridges, they were washed with 1 ml 2.5% ACN/1% FA. Finally, peptides were eluted with 150 μL 80% ACN/1% FA and dried by vacuum centrifugation. For mass spectrometry (MS) analysis, dried peptide samples were reconstituted in 50 μl of 5% FA. Five μl of each sample was spiked with 1 μl of internal standard, containing three standard peptides with mass-to-charge ratio (m/z) of 556.28, 477.78, and 497.60.

Isolated synovial fluid peptides were analyzed by nanoscale liquid C18 chromatography using an Agilent 1100 series LC system coupled to an LTQ-FT-ICR mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). The mass spectrometer was operated in the data-dependent mode to automatically switch between MS and MS/MS acquisition and the five most intense ions were fragmented in the linear ion trap using collisionally induced dissociation.

Obtained MS/MS spectra were analyzed by database searching against all human proteins in the SwissProt database using the Mascot software platform version 2.1.0 (Matrix Science Ltd., London, UK). The search criteria included methionine oxidation, arginine citrullination and asparagine and glutamine deamidation as variable modifications. The peptide tolerance was set to 5 ppm and the MS/MS tolerance to 0.6 Da. All searches were done without enzyme restrictions and peptides were, unless described differently, identified with a minimum Mascot score of 35. For spectral counting, peptide spectrum matches with a Mascot score of at least 20 were included. Technical details of the LC-MS/MS analysis and database searching can be found in Additional file 1.

The synovial fluid-derived peptides were analyzed by LC-MS/MS on a high-resolution LTQ-FT-ICR mass spectrometer and the resulting spectra were searched against all human proteins in the SwissProt database, allowing for citrullination, deamidation and oxidation as variable modifications. Because citrullination leads to an additional mass of nominally one Dalton compared to regular arginine residues, such searches are prone to false positives stemming from the misidentification of ¹³C isotopes or modifications that lead to similar nominal mass increases (for example, deamidation of asparagines or glutamine). Such false, randomly matched, identifications are expected to have lower identification scores, because the identification score itself is calculated using the probability that an observed match between experimental data in the MS/MS spectra and a database sequence is random. We plotted the average relative number of peptides containing citrulline residues at various cutoff scores for patient and control samples (Figure 1). This showed that, at lower scores, the percentage of identified citrullinated peptides was almost 25% in patient samples and that this value dropped when higher cutoff scores were used leveling off at approximately 10% at cutoff scores of 35 and higher. Therefore, we chose to include only peptides identified with a score of at least 35 for further analysis. This analysis also showed that, irrespective of score cutoff, the average number of citrullinated peptides identified in RA patients was higher than in control patients. Because we were mainly interested in citrullinated peptides elevated in the RA patients, we generated a first list of candidate peptides based on their spectral counts in the two patient groups. We selected all peptides that were identified in at least two more RA patients than control patients and had an overall spectral count which was 2.5 times higher in the RA patients compared to controls. This resulted in a list of 17 citrullinated peptides, 15 of which we identified as being derived from the alpha chain of fibrinogen (Table 1). This is similar to previous reports [15] and is not unexpected, as (citrullinated) fibrinogen alpha is known to be highly abundant in the synovial fluid proteome [3] and, unlike many other proteins, undergoes efficient proteolysis as part of the blood clotting cascade. While other proteins might be even more abundantly present in the synovial fluid (for example, albumin or antibodies) [17], these would not be detected in this study, because they are not as commonly proteolytically processed as fibrinogen. Figure 2A shows the LC-MS chromatogram of the synovial fluid of patient S13, with all major peaks corresponding to fibrinogen-derived peptides indicated to illustrate that these peptides dominate the chromatogram.

Given the difficulties with the reliable identification of citrullinated peptides described above, further analyses were performed to unambiguously confirm the citrullination of each of these 17 peptides. Misidentification of ¹³C isotopes was prevented by choosing a stringent precursor tolerance (5 ppm) for the initial searches, taking into account that ¹³C leads to a larger mass difference (1.0036 Da) than citrullination (0.9802 Da). Deamidation, on the other hand, cannot be discriminated from citrullination based upon mass alone. However, deamidation does not change the number of charged residues in the peptide (under the acidic conditions used for LC-MS), whereas the conversion of arginine to citrulline neutralizes a positive charge. Therefore, we anticipated citrullination to have a pronounced effect on ionization and fragmentation of the peptide, and on the elution time of the peptide in the reversed phase separation preceding our MS analysis. We analyzed the retention times of all peptides in Table 1 and compared them with the corresponding noncitrullinated peptides. This revealed that, on average, citrullinated peptides eluted five minutes later from the C18 column than their arginine-containing counterparts (Figure 2B and Additional file 1, Table S2), in agreement with their increased hydrophobicity. We also analyzed the precursor mass spectra and fragmentation spectra for all of these peptides and found that citrullination caused the dominant precursor charge state to be lower for peptides with a charge of three and higher (Figure 3 and Additional file 1, Table S2), whereas deamidation had no such effect (data not shown). Also the MS/MS spectra changed upon citrullination, as anticipated, with more prominent b-ions showing up when the converted arginine was on the C-terminal part of the peptide and increased levels of y-ions when it was located more N-terminally (Figure 3).

Having robustly established the citrullinated identity of these peptides, we set out to analyze their abundance in the synovial fluid. From the sequence of the peptides, it was apparent that mostly two regions of fibrinogen were identified with multiple peptides, with different termini, probably due to endogenous proteolytic activity. These two regions span amino acids 20-35 (sequence ADSGEGDFLAEGGGVR) and 414-433 (sequence EEVSGNVSPGTRREYHTEKL) of the fibrinogen alpha chain. The first sequence corresponds to the so-called fibrinopeptide A, a signal peptide generated from the N-terminus of fibrinogen alpha during blood clotting. Because of their prominent presence, we decided to focus on these regions, and to include other peptides covering these sequences in further analyses. We determined the relative abundance of all selected peptides in all samples by calculating the area of their extracted ion chromatograms (XIC) in the LC-MS analyses (as is shown for peptide DSGEGDFLAEGGGVR in Figure 2B), normalized using spiked (citrullinated) synthetic peptides. Figure 4A shows a heat map displaying the abundance of each of these citrullinated peptides for fibrinogen region 20-35 in the synovial fluid samples, suggesting that some of them are more often present in high amounts in RA patients compared to controls. This effect can be attributed both to an increase in the total amount of peptide present and to an increased level of citrullination. Therefore, we also calculated the extract ion chromatogram areas for all corresponding noncitrullinated peptides (Figure 4B). This revealed that the increased levels observed for the citrullinated peptides were not seen for their noncitrullinated counterparts, suggesting an increased level of citrullination. A similar analysis of region 414-433 is shown in Additional file 1, Figure S1 and showed a higher abundance of both the citrullinated and the noncitrullinated peptides in the RA patients.

To visualize the changed levels of citrullination for the peptides covering amino acids 20-35 of fibrinogen alpha, we also calculated the average relative amount of selected citrullinated peptides versus their noncitrullinated counterpart by calculating the ratio between the observed areas for these peptides. This revealed that the noncitrullinated peptides are detected in roughly 100-fold higher amounts (shown in Figure 5 by the approximate average ratio of 0.01), but this does not reflect actual abundances because the arginine-containing peptides will be detected much more efficiently due to their higher basicity and thus ionization efficiency [18]. Still, the data can be interpreted relative to each other and although the difference between individual patients is large, it is clear that a significant increase of the citrullinated variants of the peptide was observed in the RA patients (Figure 5). For both combinations of corresponding citrullinated and noncitrullinated peptide sequences shown in Figure 5, a t test on the log-transformed areas revealed significant differences between RA patients and control patients with P values of < 0.05. In addition, for both sets also the categorical Fisher's exact test (using cutoff 0.01) confirmed the significant differences with P < 0.05. The ratios for all peptides corresponding to amino acids 20-35 of fibrinogen alpha are shown in Additional file 1, Figure S2.

Interestingly, fibrinopeptide A also contains a known phosphorylation site on fibrinogen alpha (S22) [19], which is most likely generated by a member of the casein kinase II (CK2) family [20]. To determine if the presence or absence of phosphorylation was in any way linked to the citrullination of R35, we extended our database search to include this possibility and were able to identify two peptides (with sequences ADpSGEGDFLAEGGGVcR and DpSGEGDFLAEGGGVcR) containing both a phosphorylated S22 and a citrullinated R35 in a number of the patients (Additional file 1, Figure S3A). We then also determined the XIC areas of these peptides in all patients, showing that the peptides containing both modifications showed a similar behavior to the peptides containing only the citrullination site (for example, more present in RA patients), whereas peptides containing only the phosphorylation site were equally abundant in all patients (Additional file 1, Figure S3B). A final fibrinopeptide A-derived phosphopeptide containing a citrulline residue with sequence pSGEGDFLAEGGGVcR was identified with low identification scores in some of the samples, but was found to be present in only very low amounts and was, therefore, not included in this analysis (data not shown).