CXCL8(-2-77), CXCL8(1-77), CXCL8(2-77), CXCL8(3-77), CXCL8(6-77), CXCL8(7-77), CXCL8(8-77), and CXCL8(9-77) were chemically synthesized on an Activotec P11 automated peptide synthesizer (Activotec, Cambridge, U.K.) based on N-(9-fluorenyl) methoxycarbonyl (Fmoc) chemistry as described (28). RP – high performance liquid chromatography (HPLC) was used to purify synthesized proteins to homogeneity (Proto 300 C4 column; 150 ×4.6 mm, Higgins Analytical Inc., Mountain View, CA). Elution was performed with an acetonitrile gradient in 0.1% (v/v) trifluoroacetic acid (TFA), with two percent of the effluent being used for analysis with online electrospray ionization – ion trap mass spectrometry (AmaZon SL mass spectrometer; Bruker Daltonics, Bremen, Germany). Homogeneous CXCL8 proteins were folded into their correct configurations as reported previously (28). The purity and concentration of synthesized and correctly folded CXCL8 forms was confirmed by ion trap mass spectrometry and specific sandwich ELISAs, respectively.

Patients were recruited at the University Hospital Leuven after informed consent according to the ethical guidelines of the Declaration of Helsinki. Parents or legal guardians signed the informed consent on behalf of children. Synovial fluid was collected only if joint aspiration was required for treatment. Synovial fluids were collected in BD vacutainer tubes containing ethylenediaminetetraacetic acid (EDTA) (BD Biosciences, East Rutherford, NJ). The Ethics Committee of the University Hospital Leuven approved experiments involving human samples (S59874 and ML1814).

CXCL8 forms were immobilized on 96-well plates through overnight incubation at 4°C. Three washing steps with [PBS + 0.05% (v/v) Tween-20] were performed and plates were blocked with 0.1% (w/v) casein during 1 h at 37°C. After three washing steps, captured CXCL8 forms were detected using biotinylated polyclonal rabbit anti-human CXCL8 (500-P28BT; PeproTech, Rocky Hill, NJ), raised against recombinant human CXCL8(6-77), combined with peroxidase-conjugated streptavidine (R&D Systems, Minneapolis, MN). Finally, detection was obtained with a peroxidase substrate solution composed of 0.1 M citrate (pH 4.9) and 0.004% (v/v) H₂O₂, and containing 0.42 mM 3,3′,5,5′-tetramethylbenzidine (TMB; Sigma Aldrich, Saint Louis, MO). Conversion of TMB was quantified via optical density (OD) measurements at 450 nm.

Polyclonal rabbit anti-human CXCL8 (60 μg/ml; vide supra) was immobilized on a carboxyl sensor (Nicoya, Kitchener, ON, Canada) using an OpenSPR benchtop surface plasmon resonance (SPR) instrument (Nicoya). The sensor was blocked with 0.1 M ethanolamine (pH 8.5). Serial dilutions of CXCL8(1-77) and CXCL9(9-77) were loaded in [10 mM HEPES + 500 mM NaCl + 0.5% (w/v) bovine serum albumine + 0.25% (v/v) Tween-20] (pH 7.4) at a flow rate of 20 μL/min. Regeneration was performed with 20 mM Gly (pH 2.6). Results were analyzed with TraceDrawer software (TraceDrawer, Uppsala, Sweden).

For each sample, 5 μg biotinylated polyclonal rabbit anti-human CXCL8 (vide supra) was coupled to 25 μl streptavidin-coated magnetic particles (Dynabeads™ M-280 Streptavidin, Thermo Scientific, Waltham, MA) during 30 min at room temperature (RT) under continuous rotation. Antibody-labeled beads were washed four times with 0.5 ml PBS using a DynaMag 2 magnet (Thermo-Scientific) and incubated with 0.5 ml synovial fluid or with 0.5 ml human plasma enriched with CXCL8(1-77), CXCL8(6-77) and CXCL8(9-77) (5 ng of each form). After 30 min of incubation at room temperature under rotation, beads were washed four times with 0.5 ml PBS and CXCL8 proteoforms were eluted with 0.1 M glycine pH 2.8 (elution volume of 20 μl). Samples were directly loaded in a cooled autosampler (5°C) and analyzed by nano-LC-MS/MS. To investigate the kinetics of CXCL8 processing in the presence of synovial fluids, exogenous CXCL8(1-77) (500 ng) was subjected to human synovial fluid (20 μL) for a period of 0, 3, 6, 12 or 20 h and the same procedure was followed.

The osteosarcoma cell line MG-63 was grown in Eagle's minimum essential medium and stimulated with the synthetic double stranded RNA poly rI:rC (50 μg/ml; P-L Biochemicals, Milwaukee, WI) to produce CXCL8 as described (29). To isolate CXCL8 variants, MG-63 cell culture supernatant was concentrated and partially purified using controlled pore glass (CPG) and heparin affinity chromatography (GE Healthcare; Chicago, IL). Protein elution was achieved using a NaCl gradient of 0.05–2.0 M in 50 mM Tris-HCl (pH 7.4). Heparin-Sepharose fractions containing CXCL8 immunoreactivity, demonstrated by a specific CXCL8 ELISA developed in our laboratory (20), were further purified on a C8 Aquapore RP-300 column (220 ×2.1 mm, PerkinElmer Life Sciences; Waltham, MA) by RP-HPLC (Waters 600 HPLC System: controller and solvent delivery system; Milford, MS). CXCL8 elution was achieved through an acetonitrile gradient (0–80%) in 0.1% (v/v) TFA/ultrapure water (pH 2.0). UV absorbance was measured at 214 nm reflecting protein concentrations. CXCL8 proteoform identification was performed by Edman degradation on the purified fractions using a PPSQ-51A protein sequencer (Shimadzu, Kioto, Japan) and via nano-LC-MS/MS.

Synthesized CXCL8 proteoforms (used as reference molecules for optimization of nano-LC-MS/MS parameters) or partially purified cell culture supernatants from MG-63 osteosarcoma cells were diluted in 0.1% (v/v) TFA prior to analysis by nano-LC-MS/MS. Elution fractions from immunomagnetic isolation of CXCL8 proteoforms from synovial fluids or plasma were analyzed as such. Samples (5 μL) were injected on an UltiMate 3000 nano-scale RP-UPLC (Thermo Scientific) equipped with an autosampler. Physical separation of proteins was realized using a PepMap 300 C4 pre-column (5 ×0.3 mm; Thermo Scientific) combined with a Proto 300 C4 column (50 ×0.15 mm; Higgins Analytical Inc.). Samples were loaded on the pre-column in 4% (v/v) acetonitrile in 0.1% (v/v) TFA and elution was performed with an acetonitrile gradient in 0.1% (v/v) formic acid (flow rate of 0.5 μL/min). The eluate was directly injected into an Amazon Speed ETD mass spectrometer (Bruker Daltonics) provided with captive spray ionization – ion trap technology. Collision-induced dissociation (CID) was exploited for low-energy fragmentation (MS/MS fragmentation amplitude of 1.0) of preselected precursor ions with specific m/z values ± 2 in multiple reaction monitoring (MRM) mode. Hystar 3.2 and Trap Control 8.0 software (Bruker Daltonics) was used for data collection. Results were analyzed with Compass Data Analysis 5.0 software (Bruker Daltonics).

Tandem mass spectrometry (MS/MS) has become the standard technique for protein identification in complex biological samples. Given its precision, accuracy and selectivity, we selected nano-LC-MS/MS as the method of choice for development of a novel tool that allows for identification and quantification of CXCL8 proteoforms in complex small-sized samples without the use of proteases (e.g., trypsin) during sample preparation. Authentic CXCL8(1-77), elongated CXCL8(-2-77) and all six naturally occurring truncated CXCL8 proteoforms i.e., CXCL8(2-77), CXCL8(3-77), CXCL8(6-77), CXCL8(7-77), CXCL8(8-77), and CXCL8(9-77) were chemically synthesized and used as reference molecules (protein sequences of CXCL8 forms are provided in Supplementary Figure 1). For optimization of nano-LC-MS/MS parameters, we took advantage of the fact that CXCL8 proteins contain one Asp-Pro (-D-P-) bond, i.e., the peptide bond with the highest sensitivity to acidic hydrolysis (Figure 2A; Supplementary Figure 1). To guarantee maximal sensitivity of the method, MS/MS parameters were optimized to ensure that only the D-P bond breaks during top-down CID (Figure 2A). This results in the generation of two signature fragment ions and, importantly, a minor loss in intensity. The rationale for this approach, as exemplified for full-length CXCL8(1-77), is shown in Figure 2A.

For each CXCL8 form, a precursor ion with specific mass-to-charge (m/z) value most suitable for fragmentation (usually the precursor ion with the highest intensity) was selected by single mass spectrometry. Mass- and fragmentation spectra of three major CXCL8 forms, i.e., full-length CXCL8(1-77) and NH₂-terminally truncated forms CXCL8(6-77) and CXCL8(9-77), are shown in Figures 2B,C. An overview of the selected precursor ions for CXCL8(-2-77), CXCL8(1-77), CXCL8(2-77), CXCL8(3-77), CXCL8(6-77), CXCL8(7-77), CXCL8(8-77), and CXCL8(9-77), the resulting fragment ions and corresponding amino acids in the sequence of CXCL8 is provided in Table 1. The different CXCL8 proteoforms share a common, COOH-terminal fragment ion (m/z value 615.3 with 4 charges)—that can be used as an internal control—and have specific NH₂-terminal fragment ions. The combination of detection of signature fragment ions, generated by fragmentation of a pre-selected precursor ion, at the expected elution time point, strongly indicates the presence of a specific CXCL8 proteoform. To explore whether nano-LC-MS/MS can be used for quantification of CXCL8 proteoforms, mixtures were prepared containing equal amounts of CXCL8(1-77), CXCL8(6-77), and CXCL8(9-77). We found that CXCL8 proteins can be quantified by determining the intensity of their fragment ions in the extracted ion chromatogram (EIC), that shows the detection of signature fragment ions generated by fragmentation of a specific precursor ion during protein elution. Dose-response curves demonstrating the simultaneous quantification of 3.1 pg−12.5 ng CXCL8(1-77), CXCL8(6-77), and CXCL8(9-77) in MRM mode are shown in Figure 2D. Up to five CXCL8 proteoforms can be measured simultaneously in MRM mode without loss of sensitivity. Analysis of a mixture containing CXCL8(1-77) and CXCL8(6-77) (ratio 1:4) confirmed succesful quantification if samples contain unequal amounts of CXCL8 proteoforms (Supplementary Figure 2). Finally, the dynamic range of detection was at least four orders of magnitude (Figure 2D).

Large-scale purification of natural CXCL8 from diverse cellular sources including osteosarcoma cells (MG-63), fibroblasts, monocytes, endothelial and epithelial cells, revealed NH₂-terminal sequence heterogeneity (15, 19, 30–34). To validate our nano-LC-MS/MS-based approach, we examined the presence of natural CXCL8 proteoforms in cell culture supernatant from osteosarcoma cells (MG-63) after partial purification by adsorption to CPG, heparin affinity and RP chromatography. As reported previously, osteosarcoma cell-derived CXCL8 was found to display NH₂-terminal sequence heterogeneity (33, 34). Edman degradation was performed on 740 ng RP-HPLC purified CXCL8. Proteins corresponding to CXCL8(1-77) and CXCL8(3-77) with NH₂-terminal sequences AVLPRSAKELRXQXIKTY and LPRSXKE were identified (X refers to an unidentified amino acid). A very weak signal of SX(K)(E)XRXQ, indicating the presence of CXCL8(6-77), was detected in which signals for Lys (K) and Glu (E) were close to the background. Analysis of EIC and fragmentation spectra, obtained after exposing a hundred-fold smaller amount of CXCL8 (7.4 ng) to nano-LC-MS/MS analysis, confirmed the occurrence of elongated CXCL8(-2-77), authentic CXCL8(1-77) and the truncated variants CXCL8(3-77), CXCL8(6-77), CXCL8(7-77), and CXCL8(8-77) (Figures 3A,B). Based on the intensity of their signature fragment ions, we found that CXCL8(1-77), representing 71.6% of the total amount of CXCL8, was the most abundant CXCL8 proteoform in supernatant of cultured MG-63 cells, followed by CXCL8(3-77) (12.2%), CXCL8(-2-77) (8.5%), and the highly potent truncated forms CXCL8(6-77) (5.7%), CXCL8(8-77) (1.5%), and CXCL8(7-77) (0.5%) (Figure 3C).

Mounting evidence suggests a prominent role for neutrophils and neutrophil-attracting chemokines in chronic joint inflammation including RA and JIA (35–39). We collected synovial fluids from fourteen adult RA patients [median age (range) of 42 (22–79) years and female/male ratio of 8/6] and from twelve pediatric patients with oligoarticular (n = 9) or polyarticular (n = 3) JIA [median age (range) of 11 (4–17) years and female/male ratio of 9/3] to explore the potential role of CXCL8 processing in this disease context using our newly developed ISTAMPA technology. The workflow for this approach is depicted in Figure 1. To increase the sensitivity of the method and because viscous synovial fluids cannot be loaded onto a nano-LC column as such, a pre-purification step was required. To this end, for each patient sample, total CXCL8 was isolated from 0.5 ml of synovial fluid based on immunosorbent magnetic purification using biotinylated anti-CXCL8 antibodies coupled to streptavidin-labeled magnetic beads. Captured CXCL8 was recovered by elution in 20 μL 0.1 M Gly (pH 2.8), kept at 5°C in an autosampler and subjected to nano-LC-MS/MS analysis. We validated that the biotinylated anti–human CXCL8 antibody recognized intact and truncated CXCL8 bound to ELISA plates with similar efficiency (Supplementary Figure 3A). In addition, K_D values for intact CXCL8(1-77) (45.3 nM) and the shortest form CXCL8(9-77) (39.5 nM) were comparable on SPR. Moreover, analysis of healthy human plasma containing equal amounts (5 ng) of exogeneous CXCL8(1-77), CXCL8(6-77) and CXCL8(9-77) by ISTAMPA confirmed the successful recovery and relative quantification of intact and truncated CXCL8 proteoforms (Supplementary Figures 3B,C).

Total CXCL8 levels in synovial fluids from RA and JIA patients were 19.5 ± 6.9 ng/ml (mean ± SEM) and 1.8 ± 0.7 ng/ml (mean ± SEM), respectively, as determined by ELISA. Analysis of fragmentation spectra uncovered that, on average, intact CXCL8(1-77) accounted for only 9.3 ± 2.3% (mean ± SEM) of total CXCL8 in synovial fluids from RA patients. Approximately 85% of synovial fluid-derived CXCL8 from RA patients was NH₂-terminally truncated. Strikingly, the shortest CXCL8 proteins, i.e., CXCL8(8-77) and CXCL8(9-77) which are known to possess up to 30-fold enhanced activity as compared to full-length CXCL8(1-77), were the most abundant CXCL8 proteoforms in synovial fluids from RA patients, representing 23.0 ± 3.0% (mean ± SEM) and 34.6 ± 3.6% (mean ± SEM) of total synovial CXCL8, respectively (Figure 4; Supplementary Table 1). In addition, CXCL8(6-77) was detected in most RA synovial fluids [13.2 ± 2.5% (mean ± SEM)] (Figures 4A–C; Supplementary Table 1). A few RA synovial fluids contained detectable levels of CXCL8(-2-77), CXCL8(2-77), CXCL8(3-77) and/or CXCL8(7-77) (Supplementary Table 1). As compared to RA, synovial fluids from JIA patients contained higher percentages of intact CXCL8. The relative abundance of intact CXCL8(1-77) and elongated CXCL8(-2-77) in JIA synovial fluids was 37.4 ± 9.1% (mean ± SEM) and 14.2 ± 4.4% (mean ± SEM), respectively. The most prominent truncated proteoform in synovial fluid from JIA patients was the most active CXCL8 protein CXCL8(9-77) [20.4 ± 6.5% (mean ± SEM)], followed by CXCL8(6-77) [9.3 ± 3.4% (mean ± SEM)], CXCL8(2-77) [8.6 ± 5.0% (mean ± SEM)], CXCL8(8-77) [7.0 ± 4.3% (mean ± SEM)] and CXCL8(7-77) [3.0 ± 1.6% (mean ± SEM)] (Figure 4C; Supplementary Table 1). Results from five independent measurements of a single RA sample confirmed the reproducibility of the method if low CXCL8 concentrations are present (Supplementary Figure 4).

To investigate the kinetics of CXCL8 truncation in the presence of synovial fluids of arthritis patients, we exposed exogeneous CXCL8(1-77) to a small volume (20 μL, which is too small to detect endogeneous CXCL8 proteoforms) of synovial fluids from JIA patients (n = 4). As expected, CXCL8(1-77) was the most abundant CXCL8 proteoform if samples were immediately purified by immunosorbent isolation and analyzed with nano-LC-MS/MS (Figures 5A,B; Supplementary Figure 5A; Supplementary Table 2). After 1-20 h of incubation, full-length CXCL8(1-77) was progressively processed to NH₂-terminally truncated, more potent proteoforms. The most frequently measured CXCL8 proteoform in incubation mixtures was CXCL8(6-77), accounting for 8.1 ± 0.6% (mean ± SEM), 18.5 ± 4.1% (mean ± SEM), 30.5 ± 4.1% (mean ± SEM), 65.1 ± 8.8% (mean ± SEM) and 71.1 ± 8.7% (mean ± SEM) of total CXCL8 after 1, 3, 6, 12, and 20 h, respectively (Figures 5A,B; Supplementary Figures 5B–D; Supplementary Table 2). In addition, minute amounts of CXCL8(8-77) and CXCL8(9-77) were detected if CXCL8(1-77) was exposed to JIA synovial fluids. The relative abundance of CXCL8(8-77) after 1, 3, 6, 12, and 20 h was 1.0 ± 0.4% (mean ± SEM), 2.6 ± 1.0% (mean ± SEM), 2.0 ± 0.8% (mean ± SEM), 2.0 ± 0.8% (mean ± SEM) and 3.0 ± 0.9% (mean ± SEM), respectively (Figures 5A,B; Supplementary Table 2). CXCL8(9-77) represented 1.2 ± 0.6% (mean ± SEM) and 1.0 ± 0.4% (mean ± SEM) of total CXCL8 after 12 and 20 h, respectively (Supplementary Figures 5E,F). The half-life of intact CXCL8(1-77) in the presence of synovial fluids from JIA patients was 8.2 h (Figure 5C).