A total of 40 women with knee OA and 40 control individuals without knee OA were enrolled in a prospective study in the Thallassotherapia-Opatija Hospital, Opatija, Croatia, between the 1st of July 2018 and the 15th of December 2020. The study was approved by the Ethical Committee of the Hospital ‘Thallassotherapia-Opatija’ (approval no. 01-000-00-450/2018) and all experimental procedures were performed according to the ‘Ethical Principles for Medical Research Involving Human Subjects’ outlined by the World Medical Association Declaration of Helsinki (17). The median age and age range of the OA group and controls were 64 years old (43-75) and 62 years old (43-73), respectively.

Examinees were evaluated using the 1986 American College of Rheumatology clinical classification criteria (18). Exclusion criteria for both groups were: Infectious and autoimmune diseases, bone marrow and/or lymphatic system disorders, immune deficiency, women of reproductive age, uncontrolled blood glucose level (plasma glucose >11 mmol/l), uncontrolled hypertension (systolic >180 mmHg or diastolic >100 mmHg), chronic liver and/or renal failure, high risk of heart failure as per the New York Heart Association classification (19) III and IV, and injury to organs, blood transfusions, and/or a malignant disease within 5 years. Pain intensity was assessed using the visual analog scale from 0 to 100 mm, as judged by the patients. Radiological staging of OA in the knee was assessed using the Kellgren-Lawrence Scale (20). The clinical and laboratory characteristics of OA patients and controls are shown in Table I. Out of the 40 OA patients, hypercholesterolemia was present in all patients (100%), and glucose intolerance in 11 patients (27.5%); they were controlled with a strict diet (without drug therapy). All controls without knee OA had hypercholesterolemia and glucose intolerance was present in 10 (25%) out of the 40 women, which were regulated only with diet. Arterial hypertension was present in 33 OA patients (82.5%) and 32 controls (80%); it was controlled with β-blockers and/or angiotensin-converting enzyme inhibitors.

Peripheral venous blood (10 ml) was drawn from the patients and controls according to standard procedures; routine laboratory tests (erythrocytes, hemoglobin, leukocytes, thrombocytes, alanine aminotransferase, aspartate aminotransferase, γ-glutamyl transferase, creatinine, urate and high sensitivity C-reactive protein) were performed using the Hematology analyzer XS-1000i (Sysmex, Kobe, Japan) and Biochemical analyzer Dimension Xpand (Siemens Healthcare Diagnostics, Newark, DE, USA).

PBMCs were isolated by gradient density centrifugation (Lymphoprep solution, Nycomed Pharma; 600 x g, 20 min at 4˚C) (21). PBMCs were collected and washed (400 x g, 10 min at 4˚C) in RPMI-1640 medium (Invitrogen; Thermo Fisher Scientific, Inc.), re-suspended in tissue culture medium [RPMI-1640 supplemented with L-glutamate (2 mM), penicillin (1x10⁵ IU/l), streptomycin sulphate (0.05 g/l), and 10% FBS; Gibco; Thermo Fisher Scientific, Inc.], and counted. The cell viability exceeded 98% as measured using propidium iodide (PI; 0.5 µg/ml/1x10⁶ cells; 5 min at 22-25˚C; MerckMillipore) and a FACSCalibur flow cytometer with CellQuest Pro software version 6.0 (both from Becton Dickinson).

The simultaneous detection of surface and intracellular antigens was performed as described previously (21,22). Primary antibodies, mouse IgG1 anti-human GNLY (cat. no. D-185-3; RC8 clone; MBL International) or isotype-matched IgG1 (cat. no. bd554121; MOPC-21 clone; BD Pharmingen) were added to cell samples (1 µg/10⁶ cells, 30 min, 4˚C), which were previously washed (400 x g, 10 min at 4˚C) in FACS buffer (140 mm NaCl, 1.9 mm KH₂PO₄, 16.5 mm Na₂HPO₄, 3.75 mm KCl (all from Kemika), 0.96 mm Na₂-EDTA (Fluka), 1.5 mm NaN₃ (Difco), fixed (4% paraformaldehyde, 10 min at 22-25˚C), permeabilized in saponin buffer (0.1% saponin, Sigma, Poole, Dorset, USA), 2% fetal calf serum in PBS (NaHPO₄x12H₂O 33.9 mM, NaCl 136.8 mM, KH₂PO₄ 3 mM of distilled water; all from Kemika), 0.05% BSA (MerckMillipore), and incubated with 10% heat-inactivated fetal calf serum for 20 min at 22-25˚C. The cells were washed twice in saponin buffer (400 x g, 10 min at 4˚C) and secondary goat anti-mouse polyclonal antibodies conjugated with fluorescein isothiocyanate (FITC; cat. no. bd554001; BD Pharmingen, 1 µg/1x10⁶ cells) were added for 30 min at 4˚C. After two washes in saponin buffer, the cell membranes were restored using 1 ml FACS buffer (5 min, at 22-25˚C). Cell surface labeling was performed using a combination of phycoerythrin (PE)-conjugated anti-CD3 (cat. no. bd555332, UCHT-1 clone; BD Pharmingen; 20 µl/1x10⁶ cells) and PE-CyChrome5 (PE-Cy5)-conjugated anti-CD56 (cat. no. bd569104, B159 clone; BD Pharmingen; 20 µl/1x10⁶ cells) mAbs (30 min, 4˚C). Samples of PBMCs (1x10⁶/ml) were incubated with phorbol-myristate-acetate (10 ng/ml), ionomycin (1 µM), and monensin (3 µM) (all from MerckMillipore) for 5 h at 37˚C in a humidified incubator supplied with 5% CO₂ and intracellularly labeled for IFN-γ or IL-4 in lymphocyte subsets using PE-Cy5-conjugated anti-CD56 (20 µl/1x10⁶ cells), FITC-conjugated anti-CD3 (cat. no. bd561808, UCHT1 clone, 20 µl/1x10⁶ cells); and PE-conjugated IFN-γ (cat. no. bd554701, B27 clone, 0.25 µg/1x10⁶ cells) or PE-conjugated IL-4 (cat. no. bd551774, 8D4-8 clone; 20 µl/1x10⁶ cells) (all from BD Pharmingen). Mouse IgG1 (MOPC-21 clone) conjugated with FITC (cat. no. bd556649), PE-Cy5 (cat. no. bd555750), or PE (cat. no. bd556650) (all 20 µl/1x10⁶ cells; all from BD Pharmingen) were used as isotype controls All the incubations were performed at 4˚C for 30 min. Samples were resuspended in 2% paraformaldehyde, and processed immediately using flow cytometry, and analyzed using WinMDI 2.9 (The Scripps Research Institute).

Cytospins were prepared from freshly isolated PBMCs suspended in PBS/0.05% BSA to reach a final concentration of 8x10⁵ cells/ml. Subsequently, 100 µl of the suspension was centrifuged (Cytospin centrifuge; Shandon Inc.; 210 x g, 5 min at 4˚C) onto a glass microscope slide (Marienfeld microscope; Paul Marienfeld GmbH & Co.). The cytospin samples were air-dried (1 h) and subsequently fixed in pure acetone (5 min at 22-25˚C; Kemika). Labeling of GNLY was performed using the immunoperoxidase staining method with an LSAB™-HRP kit (Dako; Agilent Technologies, Inc.). Briefly, sections preincubated (20 min) with LSAB^™-HRP kit blocking solution were incubated (30 min) with anti-human GNLY mAb (RC8 clone) or mouse IgG1 (R312-Mouse IgG1 clone, Sigma Aldrich) diluted 1:100 in PBS/0.05% BSA. Incubation (10 min) with biotinylated anti-mouse secondary antibodies was followed by incubation with streptavidin for 10 min (both from LSAB^™-HRP kit). The reaction was developed using aminoethyl carbazole (MerckMillipore); nuclei were counterstained with hematoxylin (Gill III; MerckMillipore). All incubations were performed at 22-25˚C; samples were washed twice in PBS for 5 min between labeling steps. Subsequently, slides were mounted with Entellan (MerckMillipore) and imaged using an Olympus BX51 microscope with an Olympus DP71 camera and CellA software, v3.0 (magnification, x1,000) (Olympus Corporation).

Confocal microscopy was performed as described by Donaldson (19). Samples were fixed with 4% paraformaldehyde in PBS (10 min at 22-25˚C), permeabilized with Triton X-100 (Rohm and Haas Corporate Headquarters, 7 min), and incubated in 20% AB serum (Developed in-house) in PBS (45 min at 22-25˚C). The samples were then incubated with a combination of anti-human GNLY mAb (dilution 1:100, D185-3 clone) and rat IgG2a anti-lysosomal-associated membrane protein (LAMP)-1 mAb (dilution 1:100, 1D4B clone, BD Pharmingen) in PBS/1% BSA at 4˚C overnight. After washing, cells were labeled with secondary antibodies, Alexa Fluor 594-conjugated donkey anti-mouse (IgG H+L polyclonal; cat. no. A32744; Invitrogen; Thermo Fisher Scientific, Inc., dilution 1:500) and Alexa Fluor 488 goat anti-rat (IgG H+L polyclonal; cat. no. A11006; Invitrogen; Thermo Fisher Scientific, Inc., dilution 1:300) in PBS for 1 h at 22-25˚C, washed, and mounted with Mowiol (MerckMillipore). All washing steps were performed in PBS (three washes, 3 min each). Images were taken using a confocal imaging system (magnification x600; Olympus Fluoview FV3000) with CellA software. The cells were randomly chosen and demonstrated a well-resolved pattern.

Freshly isolated PBLs were obtained from the non-adherent fraction of PBMCs after 45 min incubation in a tissue culture Petri dish (60x15 mm; Techno Plastic Products). NK cells were purified by negative magnetic cell separation using the NK Cell Isolation Kit (cat. no. 130-092-657, Miltenyi Biotec, GmbH). The NK Cell Biotin-Antibody Cocktail (10 µl) and 40 µl ice-cold filtered FACS buffer were added to 1x10⁷ pelleted PBLs, mixed thoroughly, and incubated for 10 min at 4˚C. Subsequently, a mixture of 30 µl FACS buffer and 20 µl NK Cell MicroBead Cocktail was added per 1x10⁷ total cells and refrigerated for an additional 15 min (4˚C). The suspension was washed (300 x g, 10 min at 4˚C) and the supernatant was aspirated and replaced with 500 µl FACS buffer. The cell suspension was loaded onto a prewashed medium-sized column and placed in the magnetic field of the VarioMACS separator (both from Miltenyi Biotec, GmbH). Unlabeled cells that passed through the column were collected as NK cells. The purity and viability of the NK cells were >95%, as estimated using flow cytometry.

NK cell cytotoxicity was estimated by assessing early apoptosis of K-562 cells (23). Human erythroleukemia K-562 cells (Department of Physiology and Immunology, Faculty of Medicine, University of Rijeka, Croatia) were labeled with PKH26 (5 min at 22-25˚C) red lipophilic dye (PKH 26 Red Fluorescent Cell Linker Kit; MerckMillipore). The samples of effector NK and target K-562 cells, at different ratios in a total volume of 200 µl, were incubated in tissue culture medium for 18 h at 37˚C with 5% CO₂, and labeled with FITC-conjugated Annexin V (BD Pharmingen) (5 µg/10⁵ cells, for 15 min at 22-25˚C in the dark). PI, at a final concentration of 5 µg/ml, was added to the samples 15-20 min before performing FACS using the FACSCalibur. NK cell samples were untreated, pre-treated with mouse IgG2b anti-human perforin (cat. no. NBP1-45774, δG9 clone; 1 µg/1x10⁵ NK cells; Novus Biologicals), anti-human GNLY (cat. no. D185-3, RC8 clone; MBL International, 1 µg/1x10⁵ NK cells) or a combination of anti-perforin and anti-GNLY mAbs for 30 min at 4˚C The subset of PKH26 labeled K-562 cells (stained red), detected as PI-negative and FITC-Annexin V-positive, were considered early apoptotic cells (23). The results are expressed as the difference in the percentage of early apoptotic K-562 cells in pretreated samples at a particular effector to target cell ratio, minus the percentage of apoptotic K-562 cells cultured in the medium only.

Serum GNLY concentrations were measured using the LEGEND MAX^™ Human Granulysin ELISA Kit (cat. no. 438007; BioLegend, Inc.). The sera of OA patients and controls were prepared from peripheral venous blood samples (3 ml/sample), which were taken by venipuncture in plastic Serum Separator Tubes, 3.5 ml (Greiner Bio-One GmbH), allowed to clot for 20 min at 22-25˚C, and centrifuged (1,000 x g, 10 min, at 4˚C). The supernatant (serum) was collected in Cryo.S, 2 ml, round bottom, screw-cap vial (Greiner Bio-One GmbH) and stored at -80˚C until required for ELISA. Assays were performed following the manufacturer's instructions, and the absorbance was measured using an MRX Revelation microplate reader (Dynex Technologies Inc.).

Data are presented as the median (25-75th percentile). The difference between two groups was assessed using the non-parametric Mann-Whitney U test (TIBCO Statistica, version 13.4.0.14). Differences between multiple groups were calculated using a Kruskal-Wallis non-parametric test and a post hoc Dunn's test using MedCalc Statistical version 20.011 (MedCalc Software Ltd.). P<0.05 was considered to indicate a statistically significant difference.

Table I compares the clinical and laboratory characteristics of the group of postmenopausal women with primary OA and controls. The median OA patients' BMI (25 and 75th percentiles) of 25 kg/m² (23-29 kg/m²) was comparable to that of the control group (26; 22-27 kg/m²), as were the values for systolic and diastolic blood pressure and glucose and LDL-cholesterol concentrations. Routine hematological parameters (erythrocytes, hemoglobin, leukocytes, thrombocytes), liver parameters (alanine aminotransferase, aspartate aminotransferase, γ-glutamyl transferase), creatinine, urate and high sensitivity C-reactive protein levels did not differ between the groups. The patients were diagnosed with primary OA of the knee 6 (3.5-12) years ago. At the time of participation in the study, OA patients suffered from a pain intensity of 40/100 (20/100-50/100) mm and morning stiffness lasting 18 (10-20) min. Radiological staging of the disease in the affected joint was 2 (1-3) according to the Kellgren-Lawrence Scale. The medial knee compartment was affected in all 40 patients with OA, whereas both compartments (medial and lateral) were affected in 14/40 OA patients (35%).

Fig. 1 illustrates GNLY expression in PBLs. All PBLs were observed to be within the lymphocyte gate region (R)1 in the forward scatter/side scatter dot plot in the OA patients (Fig. 1A) and controls (Fig. 1B). The dot plot of OA patients revealed 8.9% of GNLY⁺ cells gated in R2 compared to the IgG1 isotype control (0.2%), visualized in the histogram (Fig. 1A); 3.2% GNLY⁺ cells were found in the controls when compared with the isotype control (0.9%) (Fig. 1B). Immunoperoxidase staining was used to detect higher levels of GNLY in the OA patients compared with the controls, observed as red-labeled granules (Fig. 1C). Confocal microscopy provided better insight into the colocalization of GNLY (red fluorescence) and LAMP-1, a marker of exocytosis (green fluorescence) in OA patients (Fig. 1D) and controls (Fig. 1E). The colocalization of GNLY and LAMP-1 exhibited yellow fluorescence, resulting from the overlapping of green and red fluorescence in the cytotoxic effector cells of controls, particularly in OA patients, where it accumulated on one half of the cell, forming a signet ring shape. The individual values for each of the four independent experiments per group (bar samples S1-S4) are shown in the charts (Fig. 1D and E). The median percentage (25-75th percentiles) of GNLY localized inside LAMP-1⁺ granules was 37.2% (35.4-38.5%) in the OA patients and did not significantly differ compared to the controls [51% (48-51.4%)] (Mann-Whitney U test). However, the intensity of GNLY labeling was significantly higher in OA patients than in controls (P=0.000001), as measured by flow cytometry (Fig. 1F). The frequency of GNLY⁺ PBLs, assessed by flow cytometry, was also higher [7.5% (2.8-9%)] in the OA patients compared with the controls [1.7% (0.3-3.9%)] (Fig. 1G). In the OA patients and controls, the GNLY serum concentrations, determined by ELISA were <0.3 ng/ml (Table I).

The proportion of GNLY-expressing CD3^-CD56⁺ NK cells, CD3⁺CD56^- T cells and CD3⁺CD56⁺ NKT subsets were significantly increased in the OA patients compared with the controls (P=0.01, P=0.03 and P=0.003, respectively; Fig. 2A). Dot plots illustrate the analysis of PBL subsets from OA patients (Fig. 2B) and controls (Fig. 2C) and show the frequency of CD3^-CD56⁺ NK (10 vs. 8.6%), CD3⁺CD56⁺ NKT (7 vs. 5.8%), and CD3⁺CD56^- T cells (64.6 vs. 73.4%), with respect to isotype control. These values correspond to the median (25-75th percentiles) of the percentage of NK cells [8.3% (6.8-10%)], T cells [73.2% (70.4-76.5%)] and NKT cells [2.9% (2.1-3.9%)] in OA patients, and these did not significantly differ from the values of NK cells [10.3% (7.2-14.5%)], T cells [70.6% (67.2-75%)], and NKT cells [3.5% (2.3-5.4%)] in the controls. Histograms included in Fig. 2B and C illustrate the frequency of GNLY-expressing PBL subsets in the OA patients and controls, respectively. The histograms were calculated based on the difference between the percentages of the GNLY⁺ cells minus the percentage obtained in the isotype-matched control.

Early apoptosis did not differ significantly between the OA patients and controls when obtained cells were cultured in medium only [Fig. 2D and E, Mann-Whitney U test for each, effector: target ratio (E:T)]. However, the analysis of molecular mechanisms involved in the NK cell-mediated killing of K-562 targets revealed GNLY-mediated early apoptosis only in the OA patients (Fig. 2D). The combination of anti-GNLY and anti-perforin mAbs almost completely abolished early apoptosis in cells, at E:T cell ratios of 50:1 (P=0.01), 25:1 (P=0.006) and 12.5:1 (P=0.01) (Fig. 2D), but without an additional effect compared to the cells treated with anti-GNLY antibody alone. In controls, the addition of anti-GNLY antibody or the combination of anti-GNLY antibodies with an anti-perforin mAb did not significantly alter the proportion of early apoptotic cells (Fig. 2E). Anti-perforin mAb alone did not affect the early apoptosis of K-562 cells in the OA patients (Fig. 2D) or controls (Fig. 2E).

The distribution of GNLY was analyzed in gated GNLY⁺ cells observed in the fluorescence 2/fluorescence 3 dot plots showing CD3/CD56 labeling in OA patients (Fig. 3A) and controls (Fig. 3B) in comparison to their isotype controls. Considering the samples from OA patients vs. that from controls, the GNLY distribution percentages were 57.8 vs. 68.6% in CD3^-CD56⁺ NK cells, 12.8 vs. 1% in CD3⁺CD56^- T cells, 19.7 vs. 2.2% in CD3⁺CD56⁺ NKT cells, and negligible in the isotype controls (Fig. 3A and B). The T and NKT cells of OA patients had significantly higher GNLY distribution than that in the controls (P=0.004; Fig. 3C).

Analysis of intracellular IFN-γ and IL-4 expression levels revealed differences in the balance between the OA patients (Fig. 3D) and controls (Fig. 3E). IFN-γ dominated over IL-4 in NK and T cells of OA patients (P=0.03 and P=0.01, respectively), while they were codominant in NKT cells (Fig. 3D). In the controls, IL-4 levels were higher than IFN-γ levels in NK and NKT cells (P=0.004), and it was only marginally increased in T cells (P=0.07; Fig. 3E).