Articular Joint Lubricants during Osteoarthritis and Rheumatoid Arthritis Display Altered Levels and Molecular Species

Background Hyaluronic acid (HA), lubricin, and phospholipid species (PLs) contribute independently or together to the boundary lubrication of articular joints that is provided by synovial fluid (SF). Our study is the first reporting quantitative data about the molecular weight (MW) forms of HA, lubricin, and PLs in SF from cohorts of healthy donors, patients with early (eOA)- or late (lOA)-stage osteoarthritis (OA), and patients with active rheumatoid arthritis (RA). Methods We used human SF from unaffected controls, eOA, lOA, and RA. HA and lubricin levels were measured by enzyme-linked immunosorbent assay. PLs was quantified by electrospray ionization tandem mass spectrometry. Fatty acids (FAs) were analyzed by gas chromatography, coupled with mass spectrometry. The MW distribution of HA was determined by agarose gel electrophoresis. Results Compared with control SF, the concentrations of HA and lubricin were lower in OA and RA SF, whereas those of PLs were higher in OA and RA SF. Moreover, the MW distribution of HA shifted toward the lower ranges in OA and RA SF. We noted distinct alterations between cohorts in the relative distribution of PLs and the degree of FA saturation and chain lengths of FAs. Conclusions The levels, composition, and MW distribution of all currently known lubricants in SF—HA, lubricin, PLs—vary with joint disease and stage of OA. Our study is the first delivering a comprehensive view about all joint lubricants during health and widespread joint diseases. Thus, we provide the framework to develop new optimal compounded lubricants to reduce joint destruction.

Introduction boundary lubricant, alone or synergistically with HA [1,26]. Lubricin is expressed in human SF as various isoforms [27], the form that provides better boundary lubrication is unknown. SF from human knee joints with advanced OA was reported to have markedly elevated lubricin concentrations compared with normal human SF [8,20] and a similar friction-lowering lubricating function [8,20]. However, compared to normal the levels of lubricin can also be low in some OA SF concomitant with a diminished cartilage boundary lubricating function [7]. Lubricin has been proposed to be chondroprotective, potentially protecting articular chondrocytes against apoptosis [28].
The surface of articular cartilage was described to be hydrophobic, most likely due to the hydrophobic hydrocarbon layer that is generated by surface-active phospholipids that have attached [29,30]. Most phospholipid species possess a zwitterionic head group, harboring a negative charge on the phosphate group and a positive charge on the amine group. Water-insoluble phospholipids was reported to contribute to a small fraction (~11%) of lubricin in SF [4]. FLSs are believed to be a biosynthetic source of phospholipids [31]; other phospholipids can diffuse from the plasma to SF or be released during cell necrosis. Our recently published lipidomic study demonstrated significantly elevated concentrations of most phospholipid species in OA and RA SF [5]. Changes in the relative distribution and levels of certain phospholipids might alter the lubricating ability of SF and modulate synovial joint inflammation [5]. However, the paucity of detailed data on the biological functions of lipid species in RA and OA underscore the necessity for further studies.
HA, phospholipids, and lubricin have been examined in vitro individually and in combination, and all of them mediate boundary lubrication in SF [1,2]. Nevertheless, the SF lubricant that contributes most to boundary lubrication in healthy, OA, and RA joints remains debated [32][33][34][35]. Individual differences in the level and composition of lubricants might cause patientspecific variability in joint lubrication and thus the speed of joint deterioration.
No data exist on the collective levels of HA, lubricin, and phospholipids in SF in patients with common joint diseases, such as OA and RA. Thus, the aim of our study was to quantify lubricin, the various molecular weight forms of HA, and all phospholipid species in parallel in SF from cohorts of healthy donors, patients with early-or late-stage OA, and patients with active RA.

Synovial fluid donors
SF samples were collected from knee joints of 16 postmortem donors, 20 patients with rheumatoid arthritis (RA), and 48 patients with OA. This study was approved by the ethical review committee of the Faculty of Medicine (Justus-Liebig-University of Giessen, Germany), and all patients provided written informed consent for their donor samples to be used for research. The ethical review committee waived the need for consent to be obtained from relatives of deceased donors, based on a judicial order that allowed an autopsy to be performed and to avoid additional emotional strain on relatives.
Detailed information about the SF donors is provided in Table 1. Based on the macroscopic appearance of the 6 cartilage surfaces-patella, trochlea, and femur and tibia from the medial and lateral sides-we used the Outerbridge classification (OU) [36] to subcategorize OA patients. To estimate OU scores for the entire knee joint, the 6 individual OU scores were added up and divided by 6 to generate an average score. OA joints were categorized as early (eOA; average OU 2; n = 26)-or late-stage OA (lOA; average OU > 2; n = 22). RA was diagnosed per the American College of Rheumatology [37]. Active arthritis in SF of RA patients was defined by the cell count of granulocytes (>1500/μl SF) in absence of bacteria and crystals (polarization microscopy) to exclude other forms of arthritis.

Synovial fluid samples
SF was collected from knee joints by aspiration during arthroscopy (patients with early OA cartilage damages), knee replacement surgery of late OA patients or arthrocentesis during medical exam for RA [5,38] All SF samples were first incubated for 15 min at 37°C, passed through a 1.2-μm filter to eliminated cells and cellular debris, and centrifuged at 16,100 x g for 45 min at RT [5,38]. To quench protease and phospholipase activity, 10% (v/v) inhibitor cocktail was added (PPI) [5]. The supernatants were then collected and stored at -86°C until analysis.

Level and molecular weight distribution of HA in synovial fluid
To examine HA levels in SF, samples were initially diluted 1:40,000 in 5% Tween20 in PBS (pH 7.2-7.4) and serially diluted 1:4 three times. HA concentration in human SF was measured in triplicate per dilution by sandwich ELISA (DuoSet ELISA Development kit) according to the manufacturer's instructions (R&D Systems, Minneapolis, USA). This kit contained recombinant human aggrecan to capture HA and biotinylated recombinant human aggrecan to detect bound HA using streptavidin-conjugated horseradish peroxidase. According to the information provided by the manufacturer, the HA ELISA detects low molecular weight (15-40 kDa), medium molecular weight (75-350 kDa), and high molecular weight (>950 kDa) forms of HA.
The molecular weight distribution of HA forms in human SF was determined in duplicate by horizontal 1% agarose gel electrophoresis, as described previously [7,14]. In brief, SF samples without PPI were first treated with proteinase K (Roche Applied Science, Mannheim, Germany) overnight. The Lo-Ladder (30-500 kDa), Hi-Ladder (0.5-1.5 MDa), and Mega-Ladder (1.5-6.1 MDa; Hyalose, Oklahoma City, USA) markers were loaded in separate lanes as references.
Electrophoresis was performed at 50 V for 3 h. The gels were stained overnight using 0.005% Stains-All (Sigma-Aldrich, Taufkirchen, Germany) in 50% ethanol and destained with 10% ethanol for a minimum of 24 h. To visualize the bands better, the staining and destaining were occasionally repeated. The molecular weight distribution of HA forms according to MW markers being commercially available (<0.5, 0.5-1.1, 1.1-3.1, 3.1-6.1, >6.1 MDa) was determined densitometrically, with automatic subtraction of background noise, using ImageJ (National Institutes of Health).

Analysis of lubricin in human synovial fluid
Lubricin (PRG4) concentration was measured in triplicate in SF that was stored with PPI using a custom sandwich ELISA method [7]. Briefly, a capture antibody against AA1356-1373 at the C-terminus of full-length PRG4 [39] was used to coat high-binding ELISA plates, followed by incubation with SF or PRG4 standards and detection with horseradish peroxidase-conjugated peanut agglutinin, which recognizes glycosylation in the mucin domain [40]. Purified PRG4 for the standard curve was prepared from culture medium that was conditioned with bovine cartilage explants, as described [1]; purified by Superose 6 size exclusion chromatography; verified with regard to purity by western blot; and quantified by BCA protein assay. SF was digested sequentially with S. hyaluronidase (1 U/mL, 3 h at 37°C) and sialidase A-66 (overnight at 37°C) prior to quantification. Purified PRG4 standards were also treated with sialidase. SF samples, diluted 1:4, and PRG4 standard (320 μg/mL) were loaded and serially diluted (2X). If SF was diluted during aspiration from the joint and could not be loaded at 1:4 dilution, a higher dilution-up to approximately 15X-was used.

Analysis of lipids in human synovial fluid
Lipids were extracted from SF samples in the presence of internal standards (Avanti Polar Lipids, Alabaster, AL, USA) [5]. Phospholipid and sphingolipid species were quantified by electrospray ionization mass spectrometry (ESI-MS/MS) as described [5]. Lipid species were annotated according to a recently proposed shorthand nomenclature [41].

Gas chromatography
Total fatty acids (FAs) were analyzed by gas chromatography, coupled with mass spectrometry (GC-MS), as described [42]. In brief, 50 μl of SF from lOA patients was derivatized to FA methyl esters (FAMEs) in the presence of internal standards. FAMEs were separated in a highly polar BPX70 column using a GC-2010 that was coupled to a GCMS-QP2010 detector (Shimadzu, Duisburg, Germany). Quantification was performed in the selected ion monitoring mode using calibration with authentic standards.

Statistical analysis
Analysis of variance (ANOVA) was used to assess statistically significant differences between cohorts. Subsequently, confidence intervals for pairwise differences between cohorts (control, eOA, lOA, RA) were adjusted by Tukey's honestly significant difference (HSD) procedure. The statistical analysis was performed with R, version 3.1.0 using linear models of the logarithm or the logits of the response. The data were expressed as medians with interquartile ranges for the box plots. The values in the text are the medians with interquartile ranges in brackets.

Results
Level and molecular weight distribution of HA in human SF By sandwich ELISA, as shown in Fig 1A,   Relative HA concentration was calculated as the percentage of total HA, defined as 100%. By horizontal agarose gel electrophoresis [7,14], compared with control SF, the MW of HA shifted toward the lower range in OA and RA SF (Fig 2A). The percentage of HA in the upper range of 3.1-6.    (Fig 1C).

Correlation between lubricants
The concentrations of all lubricants were plotted against each other to examine their correlations (Fig 3); the data were analyzed by linear regression, and the slopes, 95% confidence intervals, and Pearson correlation coefficients (r) were calculated. In eOA SF, the concentrations of HA and lubricin correlated significantly (r = 0.69, Fig 3A, (Fig 3C-3E) corresponded to lower levels of lubricin, whereas higher percentages of high-molecular-weight HA were linked to greater concentrations of lubricin (Fig 3F).
The predominant LPC species in SF was LPC 16:0, accounting for 42% to 49% of all LPCs (Fig 5). The percentages of most LPC species peaked in control SF compared with other cohorts, except for LPC 18:0, which was 1.5-fold higher in lOA SF [ (Fig 5).
The biophysical properties of lipid species depend on the fatty acid (FA) chain length and the degree of FA saturation. Thus, we calculated the relative contribution of PC and LPC species with various FA chain lengths and degrees of saturation to the total amount of PCs and LPCs (Fig 6). PC contains 2 FAs, but our mass spectrometry method only analyzed the sum of carbon atoms and double bonds in FAs.
In all cohorts, most PCs were unsaturated (Fig 6A)-7.2% of PC species in control SF, 5.1% in eOA SF, 4.1% in lOA SF, and 3.6% in RA SF were present in saturated form (p<0.001). In contrast, FAs were more saturated among LPCs (Fig 6B) Analysis of FA chain lengths revealed that more than 80% of PC species contained 36 Catoms or less and that nearly 20% of PC species more than 36 C-atoms (Fig 6C), but these differences were not significant between cohorts. In all cohorts, approximately 90% of all LPC species had FA chain lengths with 18 (Fig 6D).
In order to get an overview of the FAs in SF, total composition was determined by gas chromatography coupled to mass spectrometry (GS-MS) for 6 lOA SF samples (Fig 6E). The principal FAs were FA 16:0, FA 18:1, and FA 18:2, each of which accounted for 20-25% of total FAs. The polyunsaturated FAs constituted 13.5%. Medium long (<14 C-atoms) and very long (22-

Lubricants and age dependency
Our cohorts differ with respect of their age (Table 1). Thus, the levels of lubricants were plotted against the age of each donor within each cohort (S1 Fig). Subsequently, linear regression was performed, and Pearson correlation coefficients were calculated. We did not find any age dependency on the levels of all three lubricants (S1 Fig, [43]). In addition, the HA MW distributions were plotted against the age. Again, our analysis revealed no correlation between the HA MW distribution in SF of our four cohorts and the age of donors (S2 Fig). We already reported that no correlation was found between the levels of PLs and the age of patients [5].

Discussion
HA, lubricin, and phospholipids contribute to cartilage boundary lubrication that is provided by SF [1,3,4]. However, the most important SF component in maintaining efficient joint lubrication has long been debated. This study is the first to measure the levels of HA, lubricin, and phospholipids in SF from the same cohorts of healthy donors, patients with early-and latestage OA, and patients with active RA. Further, we have detailed the MW forms of HA, the relative distribution of phospholipid classes, the phospholipid species composition, and the degree of saturation and FA chain lengths.
OA and RA SF contained less HA and lubricin and are enriched with phospholipids. Also, the MW distribution of HA shifted toward the lower range in OA and RA SF. These results indicate that the catabolic activities in OA and RA SF are enhanced, leading to decreased levels of lubricin and high-MW HA.
The levels of lubricin vary in OA SF compared with control SF. Some studies have demonstrated lower levels of lubricin in OA and RA SF [1,[44][45][46], whereas others have reported opposite results [8,20,44]. In our study, the levels of lubricin are consistent with those of Ludwig et al., who observed that some OA SF is lubricin-deficient and has poor boundary-lubricating ability compared with healthy control SF [7]. Further, we confirmed our previous findings that phospholipid levels are elevated under pathological conditions, which might compensate for the decreased levels of HA and lubricin in SF to protect cartilage surfaces against wear [5,43].
Several studies have attempted to characterize the components of SF, but none has determined which species is crucial for effective cartilage boundary lubrication. HA provides viscosity in SF and mediates the retention of water [9]. Lubricin has been suggested to be the most important component for cartilage boundary lubrication and has chondroprotective properties [3]. Phospholipids are believed to cover the surface of articular cartilage, where they generate a microscopically thick biofilm, contributing to cartilage boundary lubrication [4,30]. However, the interaction between these components and their functions in cartilage boundary lubrication remain poorly understood.
Several models of cartilage boundary lubrication have been proposed. A model that was developed by Hills claims that a mono-to multilayered membrane-like structure adheres to the surface of articular cartilage, where HA and phospholipids interact and form stable complexes. HA molecules constitute a core filament that is surrounded by a bilayer of phospholipids, generating oligolamellar structures [30,47]. Another model opines that a network of multilamellar   Subsequently, PC and LPC species were grouped by degree of saturation and number of carbon atoms in the FA chains, respectively, and calculated as a percentage of total PC and LPC, defined 100%. Composition of total FA was determined by gas chromatography for 6 lOA SF samples as described in Methods. Data are presented as median and interquartile ranges (A-D) or mean ± SD (E). Pvalues less than 0.05 were considered statistically significant: *0.01< p0.05, **0.001<p0.01, vesicles exists-by transmission electron microscopy and biotribological techniques, there are vesicular structures that comprise 3-7 lipid bilayers that encapsulated HA with seric proteins, such as albumin and y-globulin. Here lubricin functions solely to anchor this lipid layer to the articular surfaces [2]. We noted correlations between the concentrations of HA and lubricin, between the levels of HA and phospholipids, and between lubricin levels and the HA MW distribution in SF, supporting a model of tight interactions between these molecules.
The friction-reducing properties of HA at the cartilage-cartilage interface depend on the concentration and molecular weight of HA [6,44]. High-MW HA has distinct cartilage lubricating abilities, and in vitro supplementation with high-MW HA is believed to partially restore the cartilage boundary lubricating function of SF [44]. We found that OA and RA SF contain significantly higher levels of small-MW HA, mitigating the impact of HA on lubrication. However, HA in the MW range of 0.5-1.1 x 10(6) Da was reported to restore SF rheological properties in animal models of OA (13).
The phospholipid organization on the surface of articular cartilage is thought to depend strictly on the MW of HA [47]. Throughout the enzymatic degradation of HA during joint inflammation and OA, a membrane-like sheet that is constructed from phospholipids is destroyed. Thus, altered MW distribution of HA can impair cartilage boundary lubrication and might also explain the elevated levels of phospholipids in OA and RA SF.
The lubricating properties of phospholipids remain poorly understood. However, the tribological properties of phospholipids depend on the molecular structure of the lipids-especially chain length and the number of double bounds. To better understand the molecular mechanisms of the lubricating abilities of phospholipids, the pattern of phospholipid species must be analyzed in details. Dipalmitoylphosphatidylcholine has been considered to be the chief phospholipid in SF, as it is in the lung [30]. However, we and other groups have shown that unsaturated forms of PC, such as dioleoylphosphatidylcholine, palmitoyloleoylphosphatidylcholine, and palmitoyllinoleoylphosphatidylcholine, are the predominant phospholipids in SF [5,48].
Further, LPC chain length is significantly shorter in OA and RA SF compared with control SF. One study on boundary lubrication found that shorter FA chains increase the friction coefficient, whereas more extensive unsaturation reduces the friction [49]. Further studies are needed to elaborate the overall lubricating ability of these altered PC and LPC species in SF.

Conclusions
In conclusion, we have provided novel comprehensive measurements of HA, lubricin, and phospholipids levels in SF from the same cohorts of healthy donors, patients with early-and late-stage OA, and patients with RA. Our study provide insight into the detailed composition of SF, correlating the concentrations of lubricants and the alterations that occur during widespread joint diseases, which can impair the joint lubricating ability of SF. These results extend our current knowledge on cartilage boundary-lubricating molecules. Thus our study provides the framework to develop new optimal compounded lubricants able to reduce joint destruction by improved lubrication. lOA SF (C), and 20 RA SF samples (D) as described in Methods. The molecular weight distribution of HA was calculated as the percentage of total HA (= 100%). Linear regression was performed, and Pearson correlation coefficients were calculated. The Pearson correlation coefficients range as follow: from -0.30 to 0.10 for control SF; from -0.10 to 0.07 for eOA SF; from -0.11 to 0.19 for lOA SF, and from -0.10 to 0.14 for RA SF. (TIF)