N‐Glycan profiling of chondrocytes and fibroblast‐like synoviocytes: Towards functional glycomics in osteoarthritis

Abstract Purpose N‐Glycan profiling provides an indicator of the cellular potential for functional pairing with tissue lectins. Following the discovery of galectin expression by chondrocytes as a factor in osteoarthritis pathobiology, mapping of N‐glycans upon their phenotypic dedifferentiation in culture and in fibroblast‐like synoviocytes is a step to better understand glycobiological contributions to disease progression. Experimental design The profiles of cellular N‐glycans of human osteoarthritic chondrocytes and fibroblast‐like synoviocytes were characterized by mass spectrometry. RT‐qPCR experiments determined mRNA levels of 16 glycosyltransferases. Responsiveness of cells to galectins was quantified by measuring the mRNA level for interleukin‐1β. Results The shift of chondrocytes to a fibroblastic phenotype (dedifferentiation) is associated with changes in N‐glycosylation. The N‐glycan profile of chondrocytes at passage 4 reflects characteristics of synoviocytes. Galectins‐1 and ‐3 enhance expression of interleukin‐1β mRNA in both cell types, most pronounced in primary culture. Presence of interleukin‐1β leads to changes in sialylation in synoviocytes that favor galectin binding. Conclusions and clinical relevance N‐Glycosylation reflects phenotypic changes of osteoarthritic cells in vitro. Like chondrocytes, fibroblast‐like synoviocytes express N‐glycans that are suited to bind galectins, and these proteins serve as inducers of pro‐inflammatory markers in these cells. Synoviocytes can thus contribute to disease progression in osteoarthritis in situ.

molecular messages, which are turned into respective bioactivity by pairing with tissue lectins, has added a new functional dimension to their presence [1,2]. When reprogramming of glycan synthesis is encountered in pathophysiological processes, this could indeed be part of an intimately orchestrated co-regulation with cognate lectins (e.g., in inflammation together with the three selectins) [3,4]. Intriguingly, distinct aspects of protein glycosylation can even become switches for regulating cellular gene expression. This has recently been shown in vitro for an association between sialylation and transcriptional activity of genes maintaining breast cancer pathogenicity such as the epidermal growth factor receptor, CD44 or nucleolin [5]. These lines of evidence for clinically relevant glycan functionality give incentive to approach the study of glycan profiles from a new perspective, especially in the context of a common disease such as osteoarthritis (OA).
OA is considered a degenerative disease of the entire joint, involving all joint constituents (cartilage, meniscus, subchondral bone, synovial membrane and infrapatellar fat pad), with unknown etiology. Based on the hypothesis of a functional glycan-receptor (lectin) interplay in OA, glycophenotyping of human chondrocytes in primary culture has guided us to identify a new class of pathogenic effectors in OA, that is, galectins [6,7]. Upregulation of expression and the extracellular availability of galectins-1, -3 and -8 (Gal-1, -3, -8) was then shown to induce a pro-degenerative and -inflammatory gene signature in chondrocytes, with the galectins acting together as a team [8][9][10]. Considering the potential of chondrocytes from culture for cartilage regeneration [11] together with the related problem of cellular transition to a fibroblastlike phenotype after passaging in vitro [12,13], it is now timely to define any alterations in N-glycans within this process. Equally important, it is also warranted to include another cell type involved in joint degeneration processes, that is, fibroblast-like synoviocytes (FLS). Of note, synovial fluid cells secrete a galectin (Gal-8) that associates with the glycoprotein CD44vRA and hereby affects the local inflammatory reaction in rheumatoid arthritis [14]. Starting with glycan mapping of FLS, we intended to learn more about the potential of cells of the synovial tissue to drive OA progression via glycan-dependent processes.
Thus, this study aimed to investigate whether (i) the phenotypic change of OA chondrocytes to dedifferentiated fibroblast-like cells during passaging to p4 is associated with alterations of N-glycosylation and thus responsiveness to galectins, (ii) there is a similarity in Nglycosylation between OA chondrocytes (at p0 and at p4) and OA FLS, (iii) a similarity exists between OA FLS and immortalized human synovial fibroblasts (cell line K4IM), (iv) N-glycosylation in OA FLS is affected by pro-inflammatory mediators, and (v) OA FLS are responsive to galectins (to a similar extent as chondrocytes at p4 and in primary cell isolates).

Cell culture
Clinical specimens of articular cartilage and synovial tissue were obtained from OA patients (13 female, 13 male; age range: 47-

Clinical Relevance
The emerging role of tissue lectins (galectins) in the pro-  [17,18] were maintained in DMEM supplemented with 10% fetal bovine serum and used at 100% confluency at p36-p38.
Human OA chondrocytes were isolated from femoral condyles and tibial plateaus of eight patients, from whom OA FLS were also isolated (see above), and cultured following established protocols to allow direct comparison between the two cell types [19]. Chondrocytes of five additional patients were independently taken to p4 to allow direct comparison of OA chondrocytes in p0 and p4.

RT-qPCR measurements
Total RNA isolation, cDNA synthesis and SYBR-green-based RT-qPCR experiments (including details on primer sequences and efficiencies) had previously been described [6]. The protocols followed the minimal guidelines for the design and documentation of qPCR experiments [20]. mRNA expression levels for 16 glycosyltransferases involved in N-glycan processing and maturation at different stages from the conversion to hybrid-and complex-type structures to α2,3/6-sialylation were calculated as relative copy numbers with respect to the geometric mean of the expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-actin (ACTB) and succinate dehydrogenase complex, subunit A (SDHA) arbitrarily set to 1000.
The effect of interleukin-1β (IL-1β) or galectins on mRNA levels was quantified as fold changes relative to untreated cultures, considering normalization to GAPDH.

Stimulation of cells with cytokines and galectins
At 90% confluency, cell cultures were serum-starved overnight and treated for 24 h (RT-qPCR) or 5 days (mass spectrometry) with human recombinant IL-1β (10 ng/mL) or tumor necrosis factor-α (TNF-α) (40 ng/mL) (both from Biolegend) to induce an aspect of pro-inflammatory conditions. In another set of experiments, serum-starved cells were treated with 10 or 50 μg/mL human Gal-1 or Gal-3 for 24 h, prior to RT-qPCR analysis. Recombinant Gal-1 and -3 were prepared, purified, checked for maintained activity and tested under conditions as described previously [8,9]. Control cultures of cells from the same patient were processed in parallel.

Sample preparation for mass spectrometry (MS)
Adherent cells were washed thoroughly with phosphate-buffered saline to remove any components of the culture medium. Cells were lysed with 100 mM ammonium bicarbonate solution (Sigma-Aldrich) containing 2% SDS (Bio-Rad). Dithiothreitol (Sigma-Aldrich) was added to the solution to a concentration of 30 mM, the sample was incubated for 5 min at 95 • C followed by another 30 min at 56 • C. Addition of iodoacetamide (Sigma-Aldrich) to a final concentration of 75 mM followed, and the mixture was incubated for 30 min in the dark at room temperature. Then, the samples were centrifugated at 10,000 rcf for 5 min to remove cell debris, the resulting supernatant was treated with chloroform/MeOH for protein precipitation [21].

Statistical evaluation
Statistical analyses of RT-qPCR data were performed using IBM SPSS

N-Glycosylation of OA chondrocytes at p0/p4
In a previous study, we presented a survey on 21 N-and 3 mucintype O-glycans of OA chondrocytes in primary culture, providing first data on the distribution of glycans among the different structural categories [6]. Of note, we found evidence for the presence of possible galectin-binding structures, defined as "core-substituted,

N-Glycosylation of OA FLS
Resulting from experiments performed under identical analytical conditions, the data on the N-glycan profile of OA FLS established the basis for comparisons to those of OA chondrocytes (at p0 and p4). Detailed information about the samples is found in Table S2, the corresponding raw data are listed in Tables S3 (area under the curve) and S4 (retention time). In total, 147 N-glycan structures, listed in Table S5, were detected and assigned to the different categories, that OA chondrocytes appear to be reconcilable with RT-qPCR data on the respective glycosyltransferases (Table S6). The respective heatmaps presented in Figure 2E and S2E give an overview on the level of different isobaric N-glycan structures. Taken together, these data suggest an inherent difference in the N-glycan profile between OA FLS and OA chondrocytes. This difference is diminished by passaging of chondrocytes that leads to morphological resemblance with OA FLS.
To allow comparison of data from patient-derived OA FLS with a standardized cell line, immortalized synovial fibroblasts K4IM were included into the analysis. Comparative analyses disclosed a number of differences such as an increased ratio between complex-and hybridtype structures or higher levels of triantennary structures in K4IM cells ( Figure S3). Next, we aimed to examine the extent of susceptibility of N-glycosylation in OA FLS to the presence of functional disease markers (IL-1β, TNF-α), in order to delineate the influence of a proinflammatory microenvironment in situ.

Cytokines as modulators of OA FLS N-glycosylation
Initial evidence for a modulation of N-glycosylation by a proinflammatory cytokine (IL-1β) was collected by RT-qPCR measurements (Table S7). These results suggested effects on i) sialylation by ST6GAL1 and ST3GAL4, ii) branching via MGAT4/5B and, most markedly, iii) N-glycan maturation by MAN1C1. Detailed MS-based characterization of N-glycans in paired samples of 14 patients revealed that IL-1β led to N-glycan remodeling (affecting a total of 27 cases of N-glycans after Benjamini-Hochberg correction), stronger so than TNF-α ( Figure 3A,B). Fittingly, the shift from α2,6to α2,3-sialyation, induced by IL-1β but not by TNF-α ( Figure 3C-F

Detection of elicitor activity of-galectins on FLS
The level of gene expression for IL1B was used as a sensor for respective galectin activity. Figure 4 shows that Gal-1 and -3 upregulated this parameter in OA FLS, and that the measured effect was comparable to that in OA chondrocytes at p4. Primary cultures showed an enhanced level of activity that declined during the following steps of passaging ( Figure S4).

DISCUSSION
N-Glycans are a highly versatile means to dynamically fine-tune the communication between cells and their environment [2,23,24]. Local density and the structures of the branch ends are factors that com-monly specify their recognition by lectins. Concerning human galectins, which are multifunctional effectors triggering a host of clinically relevant outside-in signaling [25][26][27][28], the number of antennae in the complex-type category, status and linkage type of sialylation and also LacdiNAc presence can modulate the interaction [29,30]. For example, positioned at an early stage in N-glycan maturation, the Golgi α1,2mannosidase I (MAN1C1) is the control point for conversion of highmannose-to hybrid-type N-glycans. An inhibition at this site can not only affect correct routing but also galectin-dependent lattice formation of glycoproteins, as exemplified for basigin (CD147), an inducer of matrix metalloproteinases [31,32], and Gal-3 [33,34]. In turn, this galectin is a receptor for LacdiNAc [35,36]. Also presented at terminal positions, N-acetyllactosamine (LacNAc) made accessible by reduction of extent of α2,6-sialylation is a growth-regulatory signal in activated T and carcinoma cells 'read' by Gal-1 [37,38], whereas α2,3-sialylated structures bind Gal-8, a potent pro-and anti-inflammatory mediator [39], with nM affinity [40]. Such examples for an interplay illustrate the potential for a functional meaning of shifts in the glycophenotype, here assessed by MS-based profiling.
Our study has first added N-glycosylation to the list of changes during dedifferentiation in the course of chondrocyte passaging, that is, their conversion to a fibroblast-like phenotype. So far, shifts in gene expression upon dedifferentiation of articular chondrocytes have predominantly been attributed to matrix proteins, most prominently to the switch from type II to type I collagen, proteinases and cytokines [12,13,[41][42][43]. involving also protein recognition in osteoblasts [45], advises to consider such possibilities, too. Successful blocking of galectin binding to cell surfaces by bioactive peptides from the carbohydrate recognition domains of galectins [46] suggests the feasibility of devising such a type of competitive inhibitor, using the galectin as source.
In summary, this work underscores the non-uniform nature of events that regulate N-glycan presentation and defines this feature for OA FLS. The data further suggest to examine the potential of lectins to selectively manipulate glycoprotein function in chronic were treated for 24 h with 10 μg/mL Gal-1 or Gal-3. Fold changes of IL1B mRNA levels (normalized to GAPDH) were evaluated using RT-qPCR with respect to untreated control cells set to 1. p-Values from the comparison to the untreated control given (paired, one-sided t-test) inflammatory disease with therapeutic intention, as recently outlined [47]. A certainly ambitious aim is to find ways to interfere with disease progression by, for example, blocking galectin-glycoprotein pairing and manipulating in situ glycosylation [48], as recently also suggested in the case of coronaviral infection by their galectin-like adhesins [49]. In this context, insights into glycome representation may offer inspiration for innovations to master the enormous challenge of finding new treatment modalities for OA [50].