RA and PsA patients’ and healthy donors’ blood was collected in the rheumatology outpatient department of the Semmelweis University in the Hospital of Hospitaller Brothers of St. John of God (Budapest, Hungary). Patients were diagnosed with RA according to the 2010 ACR (American College of Rheumatology)/EULAR (European League Against Rheumatism) classification criteria (26), or with polyarthritic PsA according to the CASPAR (Classification Criteria for Psoriatic Arthritis) classification criteria (27). Based on their DAS-28 (disease activity score calculator for RA) or DAPSA (disease activity in PsA score) disease activity scores, patients had low, moderate, or high disease activity, or were in remission. The national ethics committee approved the study and informed consent was obtained from each individual (approval number: 8490-2/2018/EÜIG). This work was accomplished following the Helsinki Declaration (28).

Ficoll, MEM α, L-glutamine, and Penicillin–Streptomycin were obtained from Sigma (St. Louis, MO, USA). FBS, PBS, and trypsin-EDTA solutions were purchased from Gibco Laboratories (Gaithersburg, MD, USA). EasySep was obtained from StemCell Technologies (Vancouver, Canada), rh M-CSF, and rh RANKL were purchased from PeproTech (London, UK).

Acetonitrile, water, formic acid, ammonium-bicarbonate, and sodium deoxycholate were obtained from Merck (Darmstadt, Germany). Methanol was purchased from VWR International (Debrecen, Hungary). Dithiothreitol, iodoacetamide, and TFA were purchased from Thermo Scientific (Waltham, MA, USA), and RapiGest SF surfactant was obtained from Waters (Milford, MA, USA). Trypsin/Lys-C Mix and Trypsin Gold were purchased from Promega Corporation (Madison​, WI, USA). All chemicals, reagents, and solvents were HPLC-MS grade.

Human blood samples were collected into vacutainer EDTA tubes (Greiner Bio-One, Mosonmagyarovar, Hungary). PBMCs (peripheral blood mononuclear cells) were isolated via a Ficoll density gradient (Sigma, St. Louis, MO, USA). CD14+ cells were isolated from PBMCs using a positive magnetic selection method (EasySep, StemCell Technologies, Vancouver, Canada) (29). After isolation, a third of the monocytes were immediately frozen in liquid nitrogen. Two-thirds of monocytes were cultured (1x10⁵ monocytes per well) in MEM α (Minimum Essential Medium α) (Sigma, St. Louis, MO, USA) with 10% FBS (Gibco Laboratories, Gaithersburg, MD, USA), 1% L-glutamine (Sigma, St. Louis, MO, USA), and 1% Penicillin–Streptomycin (Sigma, St. Louis, MO, USA) at 37°C and 5% CO₂ in a humidified atmosphere. Monocytes were incubated with 50 ng/mL rh M-CSF (macrophage colony-stimulating factor) for 24 h (PeproTech, London, UK). Thereafter the samples were stimulated with 50 ng/mL of both rh M-CSF and rh RANKL (receptor activator of nuclear factor κB ligand) (PeproTech, London, UK). From then the media was replaced every 3-4 days.

On day 1, we froze the monocytes in liquid nitrogen. On days 5 and 9 of culture, differentiating pre-osteoclasts and osteoclast cells were washed in phosphate-buffered saline (PBS, Gibco Laboratories, Gaithersburg, MD, USA) and lysed directly in the culture plate by adding trypsin-EDTA solution (Gibco Laboratories, Gaithersburg, MD, USA), then frozen in liquid nitrogen. Samples were stored at - 80°C for further investigation.

Cells stored in the medium were centrifuged, and pellets were washed using an ammonium-bicarbonate buffer. Cells were lysed and proteins were reduced in 50 μL lysis buffer containing 2% sodium deoxycholate, 0.5% RapiGest SF surfactant, and 100 mM dithiothreitol, heated 60°C for 30 min and sonicated as well as vortexed for 10 cycles (10 x 1 min). Protein concentration of the supernatant was measured using NanoDrop™ 1000 (Thermo Scientific, Waltham, MA, USA) as 5-15 mg/mL for most of the samples. If the concentration was >15 mg/mL, an additional 50 μL of lysis buffer was added to the samples.

Samples containing 2.5 nmol protein were first alkylated and digested with LysC and trypsin for 4 hours, then they were analyzed using a Dionex UltiMate 3000 RSLCnano system (Sunnyvale, CA, USA) coupled with a Maxis II QTOF mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany). Protein identification, quantitation was performed using the Byonic v3.8.18 (https://proteinmetrics.org) and MaxQuant v1.6.17 (https://maxquant.org) software respectively. Detailed methodology – including missing data imputation and differential expression statistics – is described in the methodological supplementary.

To investigate the dynamic proteomic changes during osteoclast differentiation, we used the methodological process shown in Figure 1. First, blood samples of six healthy donors and six RA and six PsA patients were collected (Table 1). Monocytes were isolated from PBMCs by using a positive magnetic separation, then were cultured as described (31). Cells were differentiated into pre-osteoclasts and osteoclasts, which were collected on days 5 and 9. A total of 18 individual proteomic analyses were performed corresponding to the day 1, 5, and 9 of differentiation using 6-6-6 biological replicates of RA, PsA, and control samples. We identified a total of 1435 proteins (Supplementary Table 1) across 18 samples. For further analysis, we used the log2 transformed imputed LFQ values (Supplementary Table 2).

First, we compared the samples using principal component analysis (PCA) dimensionality reduction to study which parameters have the most pronounced effect on the expression changes during differentiation. As shown in Figure 2, the clustering of the samples overlaps largely with the different stages of the differentiation. To quantify the effect of the differentiation on the clustering, we clustered the samples using the k-means clustering algorithm and the results with the sample labels using the adjusted Rand score index (32). Differentiation stage-based classification overlapped with the k-means clusters with a score of 0.89. On the other hand, neither the date of experiment, donor’s age, laboratory parameters (RF, anti-CCP, CRP, ESR), duration of disease, or DAS-28 score (Supplementary Figure 1) had any significant effect on the clustering of the samples. Our data show that the samples clearly separate from each other based on their cell type. The PCA analysis also suggests that there is a substantially bigger difference between the pre-osteoclasts (differentiation day 5) and monocytes (day 1) than between the pre-osteoclasts and osteoclasts (Figure 2A). This is also shown in Figures 2D-F, where the expression changes are plotted against the log p values. These findings are consistent with the osteoclast differentiation process, in which monocytes are gradually transformed first to preosteoclasts, and then to osteoclasts.

Next, we compared the localization information of the detected proteins of all samples using the Uniprot database. The proteins that were detected were localized primarily to the cytoplasm, cell membrane, endoplasmic reticulum membrane, nucleus, and mitochondria (Figure 3A). To gain an overall insight into the proteomic changes during the osteoclast development, we compared the changes in expression in three pairs of samples (osteoclasts vs. monocytes, pre-osteoclasts vs. monocytes, and osteoclasts vs. pre-osteoclasts). There are robust changes in the expression of proteins during the differentiation (Figure 3). Similar to the previously reported changes on the RAW264.7 cell line (33), there is increased expression of mitochondrial proteins, and decreased expression of nuclear proteins of osteoclasts compared to the monocytes (left lane). As expected from the transition of the non-polarized monocytes to polarized osteocytes, the expression of the apical proteins is enhanced (Figure 3B) (34, 35).

Next, we applied enrichment analysis to the differential expression data of all samples. Specifically, we compared the enrichment of the proteins based on their localizations (GO cellular compartment, GO-CC) and biological processes (GO-BP). As shown in Figure 3 and Supplementary Table 3, expression of the different secretory vesicle proteins was decreased, while the basolateral membrane proteins were enriched in the osteoclast/preosteoclast group. Moreover, the proteins localized to the ribosomes were enriched during the later phase of the differentiation. In line with the transition from an immune cell to a bone-resorbing cell type, the expression of proteins in the MHC cellular compartment is also decreased. Furthermore, if we compare the changes in biological processes, the osteoclasts show enrichment of proteins involved in osteoclast differentiation and carbohydrate metabolism, while the processes involved in the immune functions of the monocytes, e.g. migration, cytokine production, and various proteins in immune responses are downregulated (Figure 3 and Supplementary Table 3) (33, 36). The expression of basolateral proteins (besides apical proteins in Figure 3B) is enhanced as well (Figure 3). (30, 31).

To have a deeper look at the main differences between osteoclast and monocyte samples, we have selected two biological processes, one of which is enhanced and one of which is decreased), and depicted the involved proteins in Figures 4, 5 (Supplementary Table 4). Monosaccharide metabolic processes in osteoclast samples are increased in line with the elevated metabolic demands of these secretory cells (Figures 4A). On the other hand, proteins involved in the defense response to bacterium pathway are decreased in line with the transition during differentiation from monocytes into osteoclasts (Figures 5A). These include mainly secreted proteins, like cathepsin G and haptoglobin, but also some vesicular ones like myeloperoxidase.

Looking at the individual proteins, the biggest expression changes in the osteoclasts (over 10 fold differential expression changes) are connected to their transition from an immune cell to a cell type with a high level of metabolism and secretion and downregulated immune-function related proteins (Supplementary Table 5). The proteins with the highest differential expression in the osteoclasts compared to monocytes are mostly the ones that have been previously reported in the literature as some of the characteristic proteins in these cells. The increased metabolism is characterized by the overexpression of the transferrin receptor protein 1 (required for mitochondrial respiration) (37). Creatine kinase B-type plays a central role in tissues with large energy, as it catalyzes the transfer of phosphate between phosphagens and ATP, and was found to be markedly induced in mature osteoclasts (19) and has a key role in osteoclast-mediated bone resorption (19, 38). The increased osteoclast-specific secretion is supported by the increased expression of vacuolar-type H⁺-ATPase (V-ATPase), sodium/hydrogen exchanger 9B2 (a Na⁺/H⁺ antiporter) (39, 40), cathepsin K and B (proteolytic enzymes) (41, 42), and lysosomal acid phosphatase (43). Integrin alpha-V crossing the cell membrane and anchored to the bone surface is essential, with which osteoclasts come into close contact with bone and promote the formation of the resorption lacunae (44). These results show, that the proteins with increased expression in osteoclasts detected in our experiments fit well with the previously reported characteristics of these cells.

Most of the proteins which are strongly downregulated in healthy osteoclasts play an important role in immune responses (e. g. complement receptor type 1), or in inflammatory processes (e. g. protein S100-A8), or have an antiviral (e. g. neutrophil defensin 1) or an antimicrobial (e. g. lysozyme C) effect. Platelet glycoprotein Ib beta chain is important in blood coagulation, while Annexin A3 is an anticoagulant. Some proteins (e. g. intercellular adhesion molecule 3) play a role in cell adhesion and phospholipase B-like 1 in lipid metabolism. Myeloid cell nuclear differentiation antigen is a transcription regulator. Some proteins (e. g. interferon-induced transmembrane protein 2) have a role in innate immunity, some in adaptive immunity (e. g. immunoglobulin kappa constant), while ficolin-1 in both (Supplementary Table 5).

In certain autoimmune diseases, like RA and PsA, activation of osteoclast cells is increased as a result of the inflammation, which promotes the production of TNF and RANK-L. Increased osteoclast activation fosters pathological bone destruction, such as the erosion of the bone and cartilage, leading to focal osteolysis of the inflamed joints. Therefore, we next investigated if there are differences in protein expression in osteoclasts originating from healthy subjects and those from patients with RA or PsA. As shown in Figure 2C, diagnoses did not have a significant impact on the clustering of the samples, when all cell types were included in the analysis. Therefore, we tested whether the differences are more pronounced if the samples are clustered within cell types. Surprisingly, as shown in Figure 6, we could detect separation of the cells with PCA analysis only when the cells were fully differentiated, but not when the monocytes were compared, suggesting that these diseases had overall more significant effects on the protein expressions of the osteoclasts than in monocytes.

The Rand index score generated by the cluster strength algorithm was increased during cell differentiation, reaching a 0.43 value in the case of osteoclast samples. Surprisingly, these findings suggest that when monocytes are isolated from the blood of healthy donors and RA or PsA patients and differentiated in vitro under the same circumstances, differences between healthy and unhealthy samples still manifest in more mature stages of cell development.

Since the largest differences between healthy and unhealthy samples were found in osteoclasts, we focused on these cells during the enrichment analysis of the RA and PsA samples. Figure 7 and Supplementary Table 6 show the enriched cellular compartments and biological processes, respectively, in which the RA samples significantly differ from the healthy ones. Interestingly, the RA and healthy osteoclasts show marked differences regarding to protein cellular compartments and biological processes. Regarding the compartments, most changes are related to different membrane compartments, most likely pointing to changes in vesicular trafficking (Figure 7). Interestingly, some of the proteins involved in the ATP biosynthetic processes were decreased, while proteins playing role in immunological processes were increased, like the antigen-presenting MHC-I and MHC-II proteins (Figures 7, 8). These data suggest, that osteoclasts differentiated from the RA patients’ monocytes fail to fully adapt to the increased metabolic demand found in healthy osteoclasts and retain some of their immunological functions. One explanation of these proteome characteristics in RA osteoclast could be a different rate of differentiation, resulting in a higher proportion of undifferentiated cells at the end of the experiments. To explore this possibility, we have created a series of weighted protein level averages of osteoclasts and either monocytes or preosteoclasts to simulate a mixture of cells. When we have averaged these proteins from non-osteoclast samples with weights between 0 and 1 with those from osteoclasts (with a weight of 1), the clustering of the healthy and RA samples still clustered in different groups when checked with PCA analysis (Figure 2). This suggests that even if some of the differences might stem from the slower rate of differentiation in either of the samples, this does not fully explain the detected phenotypes.

In the case of the PsA samples, proteins localized to Golgi, endosomes, several types of endocytic and phagocytic vesicles and granules, endoplasmatic reticulum, and MHC protein complexes (Figure 7 and Supplementary Table 6). In addition, several processes involved in immunological functions, like antigen presentation, response to bacterium, positive regulation of adaptive immune response, and positive regulation of lymphocyte responses were increased in PsA samples (Figures 7, 9). On the other hand, the expression of ribosomal, nuclear, and chromosome proteins, connected to organelle organization, metabolic processes, cytoskeleton, transcription, and translation was decreased (Figures 7, 9).