Fatty Acid Fingerprints and Hyaluronic Acid in Extracellular Vesicles from Proliferating Human Fibroblast-like Synoviocytes

Extracellular vesicles (EVs) function as conveyors of fatty acids (FAs) and other bioactive lipids and can modulate the gene expression and behavior of target cells. EV lipid composition influences the fluidity and stability of EV membranes and reflects the availability of lipid mediator precursors. Fibroblast-like synoviocytes (FLSs) secrete EVs that transport hyaluronic acid (HA). FLSs play a central role in inflammation, pannus formation, and cartilage degradation in joint diseases, and EVs have recently emerged as potential mediators of these effects. The aim of the present study was to follow temporal changes in HA and EV secretion by normal FLSs, and to characterize the FA profiles of FLSs and EVs during proliferation. The methods used included nanoparticle tracking analysis, confocal laser scanning microscopy, sandwich-type enzyme-linked sorbent assay, quantitative PCR, and gas chromatography. The expression of hyaluronan synthases 1–3 in FLSs and HA concentrations in conditioned media decreased during cell proliferation. This was associated with elevated proportions of 20:4n-6 and total n-6 polyunsaturated FAs (PUFAs) in high-density cells, reductions in n-3/n-6 PUFA ratios, and up-regulation of cluster of differentiation 44, tumor necrosis factor α, peroxisome proliferator-activated receptor (PPAR)-α, and PPAR-γ. Compared to the parent FLSs, 16:0, 18:0, and 18:1n-9 were enriched in the EV fraction. EV counts decreased during cell growth, and 18:2n-6 in EVs correlated with the cell count. To conclude, FLS proliferation was featured by increased 20:4n-6 proportions and reduced n-3/n-6 PUFA ratios, and FAs with a low degree of unsaturation were selectively transferred from FLSs into EVs. These FA modifications have the potential to affect membrane fluidity, biosynthesis of lipid mediators, and inflammatory processes in joints, and could eventually provide tools for translational studies to counteract cartilage degradation in inflammatory joint diseases.


Introduction
Extracellular vesicles (EVs) are nano-and micro-sized membrane-bound particles released by virtually all cell types [1,2]. They are often classified as exosomes (EXOs), to its pathogenesis, it would be important to characterize the FA composition of these particles. FAs have the potential to induce a plethora of beneficial and detrimental actions on different joint tissues [24], and their composition in EVs can influence the stability of EV membranes as well as reflect the availability of precursors for transcellular biosynthesis of LMs [23]. The aims of the present study were (i) to examine the HA and EV secretion by normal FLSs, (ii) to determine the expression of related genes, and (iii) to characterize the FA composition of FLS EVs, all in a temporal pattern during cell proliferation. It was hypothesized that (i) HA synthesis and secretion would decrease during cell proliferation, (ii) EV release would be stimulated during cell growth, and that (iii) FA transfer from FLSs into EVs would be selective.

Imaging of FLSs and Their HA and EV Release
FLSs were cultured for up to eight passages, and they exhibited homogenous fibroblastic (spindle-shaped) morphology ( Figure 1A) with no change in cell viability. A thick pericellular HA coat (5-15 µm) was observed around cells ( Figure 1A,E,I). The release of individual EVs from the plasma membrane was visualized from live FLSs with confocal laser scanning microscopy (CLSM) ( Figure 1A-I). Also, EVs that were left behind by migrating cells could be observed attached to the bottom of ibidi chambers ( Figure 1H).
To the best of our knowledge, the secretion of HA-containing EVs (HA-EVs) by FLSs has not been a focus of intensive research. Neither have the FA profiles of FLS EVs been examined in detail, while it is known that EVs from dermal fibroblasts can be enriched with SFAs and PUFAs compared to parent cells [23]. As FLS EVs may alleviate OA or contribute to its pathogenesis, it would be important to characterize the FA composition of these particles. FAs have the potential to induce a plethora of beneficial and detrimental actions on different joint tissues [24], and their composition in EVs can influence the stability of EV membranes as well as reflect the availability of precursors for transcellular biosynthesis of LMs [23]. The aims of the present study were (i) to examine the HA and EV secretion by normal FLSs, (ii) to determine the expression of related genes, and (iii) to characterize the FA composition of FLS EVs, all in a temporal pattern during cell proliferation. It was hypothesized that (i) HA synthesis and secretion would decrease during cell proliferation, (ii) EV release would be stimulated during cell growth, and that (iii) FA transfer from FLSs into EVs would be selective.

Imaging of FLSs and their HA and EV Release
FLSs were cultured for up to eight passages, and they exhibited homogenous fibroblastic (spindle-shaped) morphology ( Figure 1A) with no change in cell viability. A thick pericellular HA coat (5-15 μm) was observed around cells ( Figure 1A,E,I). The release of individual EVs from the plasma membrane was visualized from live FLSs with confocal laser scanning microscopy (CLSM) ( Figure 1A-I). Also, EVs that were left behind by migrating cells could be observed attached to the bottom of ibidi chambers ( Figure 1H). ). Fixed and immunostained cells with CD44 antibody (pseudocolored green) as maximum intensity projection from a stack of images (panel (B)) and the same culture imaged with super-resolution to indicate extracellular vesicle (EV) budding from the membrane of a single cell (panel (C) (maximum intensity projection from a stack of images). Live cells labeled with CellMask Deep Red plasma membrane stain (pseudocolored green, panel (D) and merged with Alexa Fluor-labeled HABC to detect HA (pseudocolored red, panel (E) as maximum intensity projections from a stack of images. The corresponding side views of panels (D) and (E) are shown in panels (F) and (G), respectively. Super-resolution (Airyscan ® ) optical sections from the same cultures shown in panels D-G indicate single EVs of variable size that are attached to the bottom of the culture plate (pseudocolored green, panel (H)) and a more detailed structure of the HA coat around the plasma membrane with budding EVs (pseudocolored red, panel (I)). Arrows in all panels indicate EVs. Confocal optical section of live fibroblast-like synoviocytes stained with Alexa Fluor 568-labeled hyaluronan (HA) binding complex (HABC) and NucBlue (pseudocolored red and blue, respectively, panel (A)). Fixed and immunostained cells with CD44 antibody (pseudocolored green) as maximum intensity projection from a stack of images (panel (B)) and the same culture imaged with super-resolution to indicate extracellular vesicle (EV) budding from the membrane of a single cell (panel (C) (maximum intensity projection from a stack of images). Live cells labeled with CellMask Deep Red plasma membrane stain (pseudocolored green, panel (D) and merged with Alexa Fluorlabeled HABC to detect HA (pseudocolored red, panel (E) as maximum intensity projections from a stack of images. The corresponding side views of panels (D) and (E) are shown in panels (F) and (G), respectively. Super-resolution (Airyscan ® ) optical sections from the same cultures shown in panels (D-G) indicate single EVs of variable size that are attached to the bottom of the culture plate (pseudocolored green, panel (H)) and a more detailed structure of the HA coat around the plasma membrane with budding EVs (pseudocolored red, panel (I)). Arrows in all panels indicate EVs.
After EV collection from FLS-conditioned media, nanoparticle tracking analysis (NTA) and CLSM were utilized to observe the number and size distribution of EVs. The main population was <200 nm in diameter as determined by NTA ( Figure 2). With CLSM, the average EV diameter was higher, 1530 ± 48 nm, indicating that they consisted of a different subpopulation when visualized with this method (Figure 3). The count of larger EVs was higher at timepoints 1-2 compared to timepoint 5, while the EV area and diameter increased during cell growth ( Table 1). The temporal reduction in the EV count was also evident when expressed per cell. The medium HA concentrations reduced between timepoints 1 and 2 and remained low at timepoints 3-5 ( Figure 4). According to the ComDet analysis, most EVs contained HA-fluorescence, i.e., they were transporting HA ( Figure 1H-I, Figure 3G-I), and there was also a temporal decrease in HA-EV counts (Table 1). After EV collection from FLS-conditioned media, nanoparticle tracking analysis (NTA) and CLSM were utilized to observe the number and size distribution of EVs. The main population was <200 nm in diameter as determined by NTA ( Figure 2). With CLSM, the average EV diameter was higher, 1530 ± 48 nm, indicating that they consisted of a different subpopulation when visualized with this method (Figure 3). The count of larger EVs was higher at timepoints 1-2 compared to timepoint 5, while the EV area and diameter increased during cell growth ( Table 1). The temporal reduction in the EV count was also evident when expressed per cell. The medium HA concentrations reduced between timepoints 1 and 2 and remained low at timepoints 3-5 ( Figure 4). According to the ComDet analysis, most EVs contained HA-fluorescence, i.e., they were transporting HA ( Figure 1H-I, Figure 3G-I), and there was also a temporal decrease in HA-EV counts (Table 1).   After EV collection from FLS-conditioned media, nanoparticle tracking analysis (NTA) and CLSM were utilized to observe the number and size distribution of EVs. The main population was <200 nm in diameter as determined by NTA ( Figure 2). With CLSM, the average EV diameter was higher, 1530 ± 48 nm, indicating that they consisted of a different subpopulation when visualized with this method (Figure 3). The count of larger EVs was higher at timepoints 1-2 compared to timepoint 5, while the EV area and diameter increased during cell growth ( Table 1). The temporal reduction in the EV count was also evident when expressed per cell. The medium HA concentrations reduced between timepoints 1 and 2 and remained low at timepoints 3-5 ( Figure 4). According to the ComDet analysis, most EVs contained HA-fluorescence, i.e., they were transporting HA ( Figure 1H-I, Figure 3G-I), and there was also a temporal decrease in HA-EV counts (Table 1).
together, FLSs, and EV fraction) that were mostly separated from each other ( Figure 6B). The discriminant function 1 depicted on the horizontal axis separated the media from the cells and EVs, and the variables mainly responsible for this included DMA 18:0, DMA 16:0, 15:0ai, 18:3n-3, and 14:0. The cells and EVs showed separation by the discriminant function 2 on the vertical axis by FAs, such as 20:3n-6, 16:0, 15:0i, and 24:0. The analysis classified 95.8% of samples correctly into their respective groups.

Temporal Changes in the FASs
The proportions of 20:4n-6 in FLSs were lower at timepoint 1 compared to timepoints 2-5 ( Figure 7A), and this was also reflected in the total n-6 PUFA sums ( Figure S2). The n-
other and from the other FLS samples ( Figure 8A). Function 1, which explained 45.1% of the variance in the dataset, separated timepoints 2 and 4 from the others, and the FAs mainly responsible for this were 24:1n-9 and 16:1n-5. In particular, timepoint 5 was separated by function 2, explaining 29.7% of the variance, and the FAs with the highest separation power were 17:0ai, 14:1n-5, and 18:1n-7. The analysis classified 100% of samples correctly into their respective timepoints.  There were no statistically significant temporal alterations in the FA profiles of the EV fraction according to the Kruskal-Wallis analysis of variance (ANOVA). However, timepoints 3-4 clustered together in the discriminant analysis, whereas the other timepoints were clearly separate from each other ( Figure 8B). Functions 1-2 explained 92.1% of the variance in the dataset. Function 1 separated timepoints 1 and 5 from the others, and 22:1n-7 had the largest separation power. In particular, timepoint 2 was separated by function 2 with 18:2n-6, 18:3n-6, 17:0ai, and 20:5n-3 as the most significant separating FAs. The analysis classified 100% of samples into the correct timepoint. Proportions of 18:2n-6 in the EV fraction correlated positively with confluence (r s = 0.412, p = 0.005) and cell count (r s = 0.342, p = 0.021; Figure 7D). There were no significant associations between the EV counts and the proportions of the main FAs or FA sums in the EV fraction. There were no statistically significant temporal alterations in the FA profiles of the EV fraction according to the Kruskal-Wallis analysis of variance (ANOVA). However, timepoints 3-4 clustered together in the discriminant analysis, whereas the other timepoints were clearly separate from each other ( Figure 8B). Functions 1-2 explained 92.1% of the variance in the dataset. Function 1 separated timepoints 1 and 5 from the others, and 22:1n-7 had the largest separation power. In particular, timepoint 2 was separated by function 2 with 18:2n-6, 18:3n-6, 17:0ai, and 20:5n-3 as the most significant separating FAs. The analysis classified 100% of samples into the correct timepoint. Proportions of 18:2n-6 in the EV fraction correlated positively with confluence (rs = 0.412, p = 0.005) and cell count (rs = 0.342, p = 0.021; Figure 7D). There were no significant associations between the EV counts and the proportions of the main FAs or FA sums in the EV fraction.

EV and HA Release and FLS Gene Expression
EVs were isolated from FLS-conditioned medium by ultracentrifugation and their characteristics validated by NTA and CLSM. They were positive for surface antigens CD44 and CD63, and for phalloidin that binds to filamentous actin, which is a cytosolic protein recovered in EVs [25]. These positive stainings were co-localized with CellMask plasma membrane stain and HA binding complex (HABC), verifying that FLSs secrete HA-transporting EVs [3]. EV size was variable in FLS-conditioned medium, and the subpopulations consisted of smaller EVs (most likely EXOs and MVs) by NTA and of larger EVs (MVs and ABs) when visualized by CLSM. Larger EVs were observed to directly bud from the plasma membrane, and they also attached to the bottom of ibidi chambers when the cells migrated. Similar to other methods for EV visualization [25], CLSM has limitations, but it can still be considered a useful supplement to EV studies. While it does not necessarily detect the smallest particles, CLSM can show phenomena, such as the association of EV membranes and HA-particles, and assess their potential significance in both the progress and amelioration of OA and related diseases.
The HA coat around FLSs is known to affect EV uptake to cells, which may be partially mediated by CD44 on the EV surface [6]. The present study documented a thick HA coat up to 15 μm around FLSs, and increased expression of CD44 in high-density cells. All HAS1-3 genes were expressed by FLSs, but HAS2 was the most expressed isoform. Previously, HAS1 was determined to be the predominant isoform in human synovial fibroblasts

EV and HA Release and FLS Gene Expression
EVs were isolated from FLS-conditioned medium by ultracentrifugation and their characteristics validated by NTA and CLSM. They were positive for surface antigens CD44 and CD63, and for phalloidin that binds to filamentous actin, which is a cytosolic protein recovered in EVs [25]. These positive stainings were co-localized with CellMask plasma membrane stain and HA binding complex (HABC), verifying that FLSs secrete HA-transporting EVs [3]. EV size was variable in FLS-conditioned medium, and the subpopulations consisted of smaller EVs (most likely EXOs and MVs) by NTA and of larger EVs (MVs and ABs) when visualized by CLSM. Larger EVs were observed to directly bud from the plasma membrane, and they also attached to the bottom of ibidi chambers when the cells migrated. Similar to other methods for EV visualization [25], CLSM has limitations, but it can still be considered a useful supplement to EV studies. While it does not necessarily detect the smallest particles, CLSM can show phenomena, such as the association of EV membranes and HA-particles, and assess their potential significance in both the progress and amelioration of OA and related diseases.
The HA coat around FLSs is known to affect EV uptake to cells, which may be partially mediated by CD44 on the EV surface [6]. The present study documented a thick HA coat up to 15 µm around FLSs, and increased expression of CD44 in high-density cells. All HAS1-3 genes were expressed by FLSs, but HAS2 was the most expressed isoform. Previously, HAS1 was determined to be the predominant isoform in human synovial fibroblasts followed by HAS2 and HAS3 [26]. HAS3 is known to produce HA polymers of smaller size compared to the other isoforms [27]. It can be hypothesized that the thick HA coat could be related to the finding of HAS2 as the dominant isoform, as it has previously been observed that the HA coat formed by HAS2 transfectants was significantly larger than that by HAS1 transfectants [27].
The temporal reduction in the medium HA concentration likely derived from the documented lower expression of HAS1-3 during FLS proliferation. Decreased HA synthesis with increasing cell density has previously been observed for different fibroblast cell cultures [28]. IL-1β and TNF-α are among factors stimulating HA secretion by OA and RA synoviocytes [29], but in the present study, the expressions of these cytokines changed to opposite directions when the cell density increased. HA is known to suppress the proliferation of synoviocytes, but only at high concentrations of high-molecular-weight HA, such as those present in normal SF [30]. The HA levels in our culture medium were much lower, approximately 0.1 µg/mL, and highly unlikely to affect proliferation.
Cell type, density, and detachment of cells are among factors that can potentially affect the secretion of EVs in cultures [31]. The present study observed a temporal reduction in the EV and HA-EV counts in FLS-conditioned medium during cell proliferation. It was previously reported by Patel et al. [32] using several cell lines that EV production per cell was higher at lower seeding densities. We hypothesize that the stress caused by passaging could lead to increased EV release, and EV secretion could also be a compensatory mechanism for intercellular communication in low-density cells with fewer cell-to-cell contacts [32]. On the other hand, the decreased EV release by high-density cells may result from inhibition caused by the subsequently high number of cell-to-cell contacts [31]. EVs have been documented to promote cellular adhesion [33] and to affect proliferation [34,35]. FLS-derived EXOs stimulated chondrocyte proliferation and migration in association with cartilage repair [36], while bone marrow MSC-derived EVs inhibited RA-FLS proliferation and viability [37] and may, thus, influence pannus formation. These findings suggest that EVs of different origin may also induce beneficial effects on joint tissues. EVs could transport bioactive lipids and other molecules between cells and their FA cargo could participate in intercellular communication on inflammatory and resolution processes.

FLS FASs during Cell Proliferation
Knowledge about the molecular changes that occur during the culture of primary FLSs is limited [38]. While inflammatory factors can stimulate the proliferation of OA and RA synoviocytes [29], and synovial hyperplasia is considered essential for joint destruction in RA [15], the understanding of the mechanisms of normal FLS proliferation remains incomplete. Cell density can affect, for instance, lipid composition and intracellular trafficking, and it was previously noted in HEp-2 cells that the levels and species distribution of several lipids, such as diacylglycerols, phosphatidic acids, cholesterol esters, and lyso-PEs (phosphatidylethanolamines), were altered during culture for one to three days [39]. Different cancer cell lines have an increased demand for MUFAs during growth [40] and, regarding hepatocytes, MUFA-containing PCs were proposed as markers of cell proliferation [41]. In addition, FA profiles were documented to change during the differentiation of pre-adipocytes to mature adipocytes [42]. In this case, 16:0 and 16:1n-9 clearly increased, while 18:0, 20:4n-6, and 22:6n-3 decreased in proportion.
Similar to studies on EV release by FLSs, research on the temporal changes in FLS lipid composition during cell proliferation is in the early stages. The FA modifications of cells can affect, for instance, membrane fluidity, properties of membrane-bound receptors, eicosanoid production, and growth [43]. The present study documented elevated proportions of 20:4n-6, 24:0, and total n-6 PUFAs in proliferating cells and a simultaneous decrease in n-3/n-6 PUFA ratios. 20:4n-6 is known to modulate the proliferation of different cell types [44,45] and to affect EV release and cargo delivery [46,47], and its metabolites can play a role in EV-mediated cancer invasiveness [48]. Increased proportions of 20:4n-6 and its precursor 20:3n-6 were observed in RA compared to OA synovium [Mustonen et al., unpubl. data]. 20:4n-6 levels in glycero-PLs of bone marrow MSCs have been documented to increase during long-term cultivation at the expense of n-3 PUFAs [49]. These changes could be associated with altered inflammatory signaling and immunomodulatory capacity, and with senescence. Regarding other n-6 PUFAs, 18:2n-6 proportions decreased from control media to FLS-conditioned media, and it was presumably taken up by the cells [50] to promote growth and long-chain n-6 PUFA synthesis. 18:2n-6 is an essential PUFA, the amount of which can be insufficient in culture media [51]. It is capable of influencing EV release [52] and proliferation of different cell types [53,54], and is converted via 18:3n-6 and 20:3n-6 to 20:4n-6, the derivative of which is PGE 2 . PGE 2 can induce deleterious and beneficial effects on synovial fibroblasts. In RA, it has been demonstrated to both stimulate and inhibit the overgrowth of synovial tissue [24].
FLS proliferation was associated with the up-regulation of CD44, TNF-α, PPAR-α, and PPAR-γ expression, while the other measured genes were either down-regulated or did not show significant temporal changes. TNF-α is a potent mediator of inflammatory functions that regulate cell death, survival, differentiation, proliferation, and migration [55]. It is overproduced in RA joints and plays a role in the establishment of synovitis, pannus formation, and joint destruction [56]. TNF-α is also known to stimulate the biosynthesis of different PL classes by OA FLSs [22]. Based on the present results, we tentatively suggest that, in addition to OA/RA synoviocytes [29], TNF-α could participate in the regulation of proliferation of normal FLSs. Moreover, its levels correlated positively with total n-3 and n-6 PUFA percentages and inversely with those of total MUFAs. PPAR-α is a transcription factor that functions as a master regulator of hepatic lipid metabolism governing, for instance, FA uptake, binding, and oxidation; ketogenesis; and triacylglycerol (TAG) turnover [57]. In the present study, PPAR-α correlated positively with FLS count, but was not associated with FA proportions. FAs and FA-derived compounds, such as eicosanoids, are natural ligands for both PPAR-α and PPAR-γ [58]. PPAR-γ modulates adipogenesis in fat tissues as well as whole-body lipid metabolism and insulin sensitivity. It also induces apoptosis in RA synoviocytes [59] and inhibits the production of inflammatory cytokines in RA and OA synoviocytes [60]. The potential roles of PPAR-α and PPAR-γ in FLS lipid metabolism remain scarcely investigated. In the present study, PPAR-γ was noted to correlate positively with the number of days from passage/confluence and with the proportions of 20:4n-6 and total n-6 PUFAs, while the relationship was negative for total MUFAs.
EVs were previously shown to be enriched with cholesterol, SM, glycosphingolipids, and PS, whereas PC and PI were often more abundant in parent cells [7,8,10,11]. PL species with two saturated fatty acyl groups (e.g., PC 16:0/16:0) can also show increases in EVs. Regarding several lipid classes, species containing 18:0/18:1, 16:0/18:1, and 18:1/18:1 were clearly elevated in EVs [10,11], which supports the present results that showed the enrichment of 16:0, 18:0, and 18:1n-9 in FLS-derived EVs. To the best of our knowledge, this is the first time the FA composition of human FLS EVs has been characterized in detail. This is of importance due to the central role of FLSs in OA and RA and the potential part their EVs play in joint diseases [2]. Previously, dermal fibroblasts were studied in this respect, and their EVs were enriched with lyso-PLs, hydroxylated SM, and ether-linked PLs, while PS and PE reduced compared to parent cells [61]. In addition, total SFA levels increased from cells to EVs [23], similar to the present study. This could lead to increased membrane rigidity and stability of EVs. The present study also observed decreased UFA/SFA ratios in EVs compared to parent FLSs, confirming earlier literature with reduced unsaturation in EVs [7]. Regarding EVs from the metastatic prostate cancer cell line PC-3, fatty acyl chain lengths in PC and PE could be shortened in EXOs [11], also supporting the present results with decreased TACLs in EV total lipids.
The low proportion of PUFAs, such as 20:4n-6, in FLS EVs is intriguing ( Figure 6A). The comparison to earlier studies is not always straightforward due to different experimental designs and different ways FA data are expressed. A previous study documented increased amounts of di-PUFA species in specific PL fractions of EVs compared to parent bone marrow MSCs [12]. These cells were treated with PUFA supplements and could potentially have removed excess PUFAs by producing EVs, which may not necessarily represent a natural phenomenon. In dermal fibroblast-derived EVs, the amount of total PUFAs relative to protein content was also increased in respect to parent cells [23]. The lack of enrichment of 24:0 and 24:1n-9 in the present study ( Figure 6A) is another finding that calls for explanation. Their proportions were expected to increase due to the importance of sphingolipids in the EV membrane [10], but for instance, in dermal fibroblast-derived EVs, the most common molecular species among SMs was that with 16:0 [61] and, in Sagini et al. [23], 24:1 was below detection. Total DMA proportions were also decreased from FLSs to EVs in the present study, even though plasmalogens can be enriched in EV membranes [62]. However, our FLS EV results show resemblance to EV data from the bronchoalveolar lavage fluid of horses, which also lacked enrichment of PUFAs, 24:0, and 24:1n-9 when EVs were isolated by size-exclusion chromatography [Höglund et al., unpubl. data].
Serum products used for cell cultures contain lipid particles of similar size to EVs [8]. In the present study, fetal bovine serum (FBS) was ultracentrifuged overnight to remove as many EVs and lipid particles as possible before the serum was used on cells. This ultracentrifuged FBS was used until sampling to avoid serum-starvation, which inhibits the proliferation and migration of synoviocytes [63]. Alterations in medium lipid composition would have consequences for the FASs of FLSs and EVs they release [8, 10,12], and an abrupt change to serum-free medium could have resulted in reduced EV secretion [31]. We harvested EVs by ultracentrifugation, and this method can result in co-isolation of EVs, lipoproteins, and lipid droplets containing TAGs and cholesteryl esters instead of membrane lipids [10]. The FA analysis of the EV preparation could, thus, have been affected by remnant lipids precipitated by ultracentrifugation. This is known to be a problem especially for plasma samples, but its potential effect cannot be wholly excluded regarding the present results.
In conclusion, the EV release and HA synthesis by FLSs decreased with increasing cell densities. FLS proliferation was featured by elevated 20:4n-6 proportions and reduced n-3/n-6 PUFA ratios, and selective transfer of FAs was documented from FLSs into EVs. While synoviocytes are integral parts of joint anatomy and health, their overgrowth leads to pathological conditions. In the present study, we observed both potentially beneficial (IL-1β, COX-2, PPAR-α, PPAR-γ) and detrimental effects (HA, TNF-α, 20:4n-6, n-3/n-6 PUFA ratio) of cell growth on the measured variables in normal FLSs. Inflammatory joint diseases, principally OA and RA, display a complex combination of both pro-and anti-inflammatory phenomena, and to influence this balance in a manner beneficial to the patients, basic data on these molecular effects are a prerequisite for the development of novel treatment options. The results offer an attractive starting point for translational studies to enhance the synthesis of molecules beneficial for joint homeostasis and to control those inducing runaway proliferation of FLSs.
The media from the culture flasks were sterile-filtered (5 µm; Sartorius) and stored as aliquots at −80 • C. The cells were washed with phosphate buffered saline (PBS; Corning, Manassas, VA, USA), detached with Trypsin-EDTA solution (T3924; Merck), followed by a treatment with trypsin inhibitor (T6414; Merck) and centrifugation at 220× g for 5 min. The pellet was diluted in sterile-filtered PBS (0.22 µm) and the total cell count was estimated. The cells were divided into two Eppendorf tubes and centrifuged at 2300× g for 5 min. One of the precipitates was diluted in 100 µL of sterile-filtered PBS for FA analysis and the other in 200 µL of TRI Reagent (T9424; Merck) for RNA isolation, and both were stored at −80 • C. Altogether, 9 media and cell samples were obtained per timepoint, but all analyses could not be conducted on every sample due to their limited volume.

FA Determination
Harvested sterile-filtered media (300 µL) and FLSs were transmethylated in methanolic H 2 SO 4 under nitrogen atmosphere, and the formed FA methyl esters were extracted with hexane and analyzed by the Shimadzu GC-2010 Plus gas chromatograph (Shimadzu, Kyoto, Japan) [65,66]. Another subsample of sterile-filtered media (3 mL) was centrifuged at 1000× g for 10 min at 4 • C, and the supernatant at 1200× g for 20 min at 4 • C. Finally, the supernatant was ultracentrifuged at <110,000× g for 90 min at 4 • C, and the EV pellet which would be equivalent to the EV secretion by an average of 85,000 cells for 2 days was diluted in sterile-filtered PBS and analyzed with gas chromatography. The FA methyl ester structures were confirmed by using electron impact mass spectra recorded by the Shimadzu GCMS-QP2010 Ultra with the mass selective detector. The results represent the FA composition (mol-%) of total lipids in the media, cells, or EV fraction. The DBI and TACL were calculated as previously described [65].

qPCR of FLSs
Total cellular RNA was extracted using TRI Reagent combined with standard chloroform-isopropanol precipitation. RNA concentration was measured with the Nan-oDrop ONE (Thermo Fisher Scientific, Waltham, MA, USA), and 1000 ng was synthesized to cDNA using Verso cDNA Synthesis Kit (Thermo Fisher Scientific) in the Biometra Personal Cycler (Analytik Jena GmbH, Jena, Germany). Eight biological replicates were analyzed using LightCycler 480 SYBR Green I Master reagent and the LightCycler 480 polymerase chain reaction apparatus (Roche, Basel, Switzerland) under the following thermal conditions: 95 • C for 5 min, 45 cycles of 95 • C for 20 s, 60 • C for 20 s, and 72 • C for 20 s, followed by 72 • C for 5 min. Melting curve analysis was performed in the range of 40 to 95 • C, 0.5 • C for each 2-s interval. Ribosomal protein lateral stalk subunit P0 (RPLP0) was used to normalize the amounts of mRNA between samples to get the relative mRNA levels at each timepoint. Fold changes were calculated using the formula 2-∆∆Ct, where ∆∆Ct is the ∆Ct (differentiated) − ∆Ct (non-differentiated) , where ∆Ct is Ct (target gene) − Ct (RPLP0) , and Ct is the cycle at which the threshold is crossed. The relative gene expression was set to 1 for timepoint 1 samples. Primer sequences are reported in Table S2.
CLSM was performed with the Zeiss Axio Observer inverted microscope equipped with the Zeiss LSM 800 confocal module (Carl Zeiss MicroImaging GmbH, Jena). Image acquisition was carried out using the ZEN 2.3 blue edition software (Carl Zeiss MicroImaging GmbH) as previously outlined [67]. The area and intensity of the stainings, counts of EVs, HA-particles, and HA-EVs, and size distribution of EVs and HA-particles were determined with the ImageJ/Fiji v1.53 software (NIH, Bethesda, MA) with various open-source plug-ins. Co-localization of EV and HA fluorescences was determined with the ComDet analysis (v0.4.2).

NTA
The sterile-filtered FLS-conditioned media were first ultracentrifuged at 110,000× g for 2 h and again for 90 min to wash the pellet, followed by resuspension in 100 µL of PBS. These enriched EV samples were analyzed by NTA using the Nanosight model LM14 (Malvern Panalytical, Malvern, UK) equipped with a blue laser (404 nm, 70 mW) and sCMOS camera (Hamamatsu Photonics, Hamamatsu, Japan). The samples were diluted in 0.1 µm-filtered (Millex-VV; Merck) PBS to obtain 40-100 particles/view, and five 30 s videos were recorded using camera level 14. The data were analyzed using the NTA software v3.0 (Malvern Panalytical) with the detection threshold 5 to track as many particles as possible with minimal background.

HA Determination
Temporal changes in the HA concentrations of FLS-conditioned medium were determined with a sandwich-type enzyme-linked sorbent assay [69].

Statistical Analyses
All statistical analyses were conducted with the IBM SPSS v25 software (IBM, Armonk, NY, USA). Comparisons between the sample types or timepoints of 9 experiments were performed with the Kruskal-Wallis ANOVA. Correlations were calculated with the Spearman correlation coefficient (r s ). The p value < 0.05 was considered statistically significant. The results are presented as the mean ± SE. To analyze how clearly the different sample types and timepoints differed from one another and which variables separated them most clearly, we also performed the discriminant analysis for the FA data. , J.C., S.P.S. and P.R.; visualization, P.N. and K.R.; supervision, P.N., K.R. and R.K.; project administration, P.N.; funding acquisition, P.N. and K.R. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: All relevant data analyzed during this study are included in this published article and its supplementary information files.