Briefly, the fluorescently labelled peptide substrate was chemically synthesized on Rink amide resin using fluorenylmethoxycarbonyl (Fmoc) solid phase peptide synthesis (SPPS) on an Activotec P11 synthesizer (Activotec, Cambridge, U.K.). The peptide TAMRA-(GRGA)₄ consists of four repeats of the arginine-containing tetrapeptide Gly-Arg-Gly-Ala and was N-terminally coupled to the fluorescent group 5(6)-carboxytetramethylrhodamine (TAMRA). After SPPS, the 2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl (Pbf) side chain protection group on the arginines was chemically removed in a mixture containing 88.9% (v/v) trifluoroacetic acid (TFA, Biosolve, Valkenswaard, The Netherlands), 4.4% (v/v) thioanisole (Arcos Organics, Geel, Belgium), 2.2% (v/v) 1,2-ethanedithiol (Merck, Darmstadt, Germany), and 66.7 mg/mL crystalline phenol (Merck). The peptide was precipitated in 90% (v/v) cold methyl tert-butyl ether and purified through reversed-phase HPLC (Waters Corporation, Milford, MS), on a 250x8.0-mm PepMap C18 column (VDS Optilab, Berlin, Germany). Elution was performed in an increasing acetonitrile gradient in 0.1% (v/v) TFA. Peptides were detected and their relative molecular mass (Mr) confirmed via electrospray ion trap mass spectrometry (AmaZon SL, Bruker, Bremen, Germany). The purified peptides were lyophilized (Speedvac) overnight and stored at -20°C until use.

Synovial fluids of patients with RA, juvenile idiopathic arthritis (JIA), gout or Lyme’s disease were collected in ethylenediaminetetraacetic acid (EDTA)-treated BD vacutainers (BD Biosciences, Franklin Lakes, NJ, U.S.A.). Each sample was centrifuged for 10 minutes at 400 g. Cell-free supernatant were collected and stored at -80°C. Platelet-free plasma was isolated after centrifugation of whole blood (collected in EDTA-treated vacutainers) for 20 minutes at 400 g, followed by 10 minutes of centrifuging the supernatant at 17000 g. Whole blood was collected in BD Vacutainers SST and centrifuged at 800 g for 10 minutes to isolate serum. Synovial fluids were only collected upon need for joint aspiration and were provided by the rheumatology unit of the university hospital UZ Leuven. Ethical approval was granted by the local Ethics Committee of the University Hospitals Leuven (UZ Leuven) under numbers ML1814, S61878 and S65508. Patients and healthy blood donors provided informed consent according to the ethical guidelines of the Declaration of Helsinki.

For the isolation of peripheral blood mononuclear cells (PBMCs) and neutrophils, peripheral blood was taken from healthy volunteers and processed by density gradient centrifugation in a Pancoll (Pan-Biotech, Aidenbach, Germany) gradient. CD14⁺ and CD14^- cells were separated by positive selection with CD14 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). Neutrophils were purified using the EasySep™ Direct Human Neutrophil Isolation Kit (Stemcell technologies, Vancouver, Canada). All isolation methods were performed according to the manufacturer’s instructions. Cells were counted in Türk’s solution using a Bürker chamber.

After isolation, cells were centrifuged for 5 min at 400 g and resuspended in RIPA buffer (Pierce, Thermofisher, Waltham, MA, U.S.A.) with 1% protease inhibitor cocktail (P8430, Sigma-Aldrich, St. Louis, MO, U.S.A.), following the ratio of 30 μL RIPA buffer per one million cells. Cells were shaken on ice for 15 minutes and subsequently centrifuged at 17000 g for ten minutes. Supernatant was collected and kept on ice or stored at -20°C until use.

To remove cell debris from samples, synovial fluids were first centrifuged for 5 minutes at 14000 g and subsequently desalted by gel filtration chromatography using Sephadex^® G-25 (Cytiva, Waltham, United States) columns in 15 mM Tris/HCl buffer (pH 7,4) or Zeba Desalting plates (Thermofisher) according to the manufacturer’s instructions. After dialysis, the absorbance at 280 nm was measured to quantify total protein content (BioTek, VT, U.S.A.).

In all assays, in-house synthesized TAMRA-(GRGA)₄ and Evans blue (UCB, Brussels, Belgium) were used as PAD substrate and fluorescence quencher, respectively. Various concentrations of substrate (0 μM – 80 μM) and quencher (0 μM – 120 μM) were included to determine their optimal range to monitor PAD activity in cell culture supernatant, plasma, serum and synovial fluid. Human recombinant PAD4 (Sf9) (0.03 nM-300 nM, Sigma-Aldrich) and rabbit skeletal muscle PAD2 (0.1 nM-100 nM, Sigma-Aldrich) were used to optimize this assay. Enzyme inhibitors used include 3 mM EDTA, EDTA-free protease inhibitor cocktail (P8340, Sigma-Aldrich), and the PAD inhibitor BB-Cl-amidine (20-30 µM, Cayman Chemicals, Ann Arbor, MI, U.S.A.). PAD assays were initiated by the addition of 3 mM CaCl₂ and 1 mM dithiothreitol (DTT). PAD activity was monitored in black, clear flat bottom, 96-well plates (Greiner Bio One, Vilvoorde, Belgium) using a CLARIOstar plate reader (BMG Labtech, Ortenberg, Germany) at 555-580 nm. Enzyme velocity is expressed as relative fluorescence in function of time.

Statistical analyses were performed and graphs were prepared using GraphPad Prism version 8 (GraphPad Software, San Diego, CA, U.S.A.). Normality of data was assessed with the Shapiro-Wilk test prior to further statistical analysis by unpaired t-test for assessing leukocyte subpopulations and one-way ANOVA analysis or Kruskal-Wallis analysis in patient cohorts. Linear regression was performed on raw data to obtain the slope representing enzyme activity over time and this is expressed by change in relative fluorescence units per minute (ΔRFU/min). Experiments were carried out in three experimental replicates for in vitro PAD activity assays. The boxplots are presented with median as center value, errors bars indicate min to max value. Significance was considered at p ≤ 0.05.

The PAD substrate TAMRA-(GRGA)₄ served as our fluorescent substrate ( Figure 1A ) and was obtained after chemical Fmoc peptide synthesis followed by HPLC purification, and its Mr was confirmed by ion trap mass spectrometry ( Figure 1B ). The experimentally determined Mr of TAMRA-(GRGA)₄ was 1793.62 and corresponds to the theoretical average Mr value of 1793.20. The mass spectrum was devoid of contaminants, which indicated that the PAD substrate was of high purity and therefore suitable for further assay development. In a next step, we determined whether TAMRA-(GRGA)₄ was quenched upon incubation with the negatively charged dye Evans blue (12). The results clearly show that Evans blue was able to induce a dose-dependent inhibition of fluorescence emitted by TAMRA-labelled substrate ( Figure 1C ).

TAMRA-labeled, arginine-containing peptides were suggested as chemical probes for detection of PAD activity (11). We spiked different doses of recombinant PAD in cell culture medium, serum, plasma and synovial fluid to assess this method for use in experimentally and clinically relevant samples ( Figure 2 ). Spiking recombinant PAD in all samples generated fluorescent signals that were independent of the amount of active PAD ( Figures 2A–D , top panels). We concluded that the method described as such failed to measure active citrullination in protein-rich samples. Nonetheless, we managed to overcome this complication by subjecting complex samples to gel filtration chromatography and by adjusting the quencher concentration to the total protein content of the sample of interest ( Figures 2A–D , lower panels). To determine the optimal Evans blue concentration in future experiments, we made serial dilutions of EDTA-treated plasma as reference for complex samples and determined the minimal Evans blue concentration that quenches all fluorescence in the presence of 0.5 µM TAMRA-(GRGA)₄ substrate. These quencher concentrations were plotted in relation to the absorbance at 280 nm as a measure of protein content, thus creating a standard curve ( Supplementary Figure 1 ) (R2 = 0.96). This standard curve allows to determine the optimal quencher concentration for measurement of PAD activity in complex biological samples based on their total protein content.

Using the optimized method, we tested PAD2 and PAD4, the two most abundantly expressed isoforms of citrullinating enzymes (13), and determined the compatibility of this PAD activity assay with their detection. Different doses of natural PAD2 and recombinant PAD4 were tested in the activity assay and enzyme kinetics were shown by displaying PAD activity over time ( Figure 3 ). We determined the enzyme activity based on the relative TAMRA fluorescence emitted after substrate conversion over time and before substrate saturation was reached. Both enzyme isoforms were quantifiable in a dynamic range of at least 30 and suffered from very little variability between experimental repeats (n=3) ( Figures 3C, D ).

According to recent ‘omics’ data, leukocytes are expected to express significant levels of PAD (14). Therefore, we isolated fresh neutrophils and PBMCs from healthy donors to characterize their enzyme activity. In Figure 4A , active citrullination of the TAMRA-(GRGA)₄ substrate is depicted over a time course of three hours for both cell populations. Our data showed that neutrophil-derived cell lysates contained higher levels of PAD activity compared to PBMCs. To characterize the PAD activity in different mononuclear cells, PBMCs were further purified based on CD14 expression. CD14 positive monocytes expressed more active PAD than CD14 negative cells, but contained significantly less PAD compared to neutrophils ( Figure 4B ). According to these results, the amount of potentially available PAD activity is the highest in neutrophils and the activity observed in PBMCs is mostly attributable to CD14 positive monocytes.

RA, JIA, gout and Lyme’s disease are all characterized by joint inflammation and share similarities such as a high neutrophil infiltration and the release of proinflammatory cytokines in the joints. Still, citrullination seems distinctly relevant in RA pathology due to the presence of anti-citrullinated peptide antibodies (15). However, it remains uncertain whether the level of PAD activity in the joints or in circulation could be a contributing factor to the disease. Therefore, we measured active citrullination in inflamed joints of a patient cohort that comprises different arthropathies. RA and JIA patients expressed high levels of citrullination in synovial fluid whereas PAD activity was nearly absent in synovial fluids of gout and Lyme’s disease patients ( Figure 5 ). Surprisingly, JIA synovial fluid samples displayed similar levels of PAD activity compared to fluids in joints of patients with RA. This finding provides proof for extracellular citrullination in synovial fluids of patients with JIA in addition to RA. Moreover, these data suggest that RA, but also JIA, involve a citrullination pathway that remains inactive in gout and Lyme’s disease, two pathologies which are also characterized by a high influx of neutrophils in inflamed joints (16, 17). As we observed local extracellular citrullination in the joints of RA and JIA patients, we continued this study by characterizing systemic citrullinating activity in these patients ( Figure 6 ). In addition, we stratified patients with RA based on CCP serology to clarify the relevance of PAD activity in sera for the autoimmune response. Our data showed enhanced PAD activity in the sera of RA patients with circulating anti-CCP antibodies ( Figure 6A ), while a comparable amount of PAD is present in anti-CCP negative RA patients and healthy controls. Moreover, plasma samples from patients with JIA displayed low levels of PAD activity, which were similar to plasma values in healthy pediatric controls ( Figure 6B ).