If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Osteoarthritis Research Program, Division of Orthopedic Surgery, Schroeder Arthritis Institute, University Health Network, CanadaKrembil Research Institute, University Health Network, CanadaInstitute of Biomedical Engineering, University of Toronto, Canada
Osteoarthritis Research Program, Division of Orthopedic Surgery, Schroeder Arthritis Institute, University Health Network, CanadaKrembil Research Institute, University Health Network, Canada
Osteoarthritis Research Program, Division of Orthopedic Surgery, Schroeder Arthritis Institute, University Health Network, CanadaKrembil Research Institute, University Health Network, Canada
Osteoarthritis Research Program, Division of Orthopedic Surgery, Schroeder Arthritis Institute, University Health Network, CanadaKrembil Research Institute, University Health Network, Canada
Address correspondence and reprint requests to: S. Viswanathan, 60 Leonard Avenue, Office 5KD416-K1, Toronto, ON, M5T 0S8, Canada. Tel: 1(416)-634-7240.
Osteoarthritis Research Program, Division of Orthopedic Surgery, Schroeder Arthritis Institute, University Health Network, CanadaKrembil Research Institute, University Health Network, CanadaInstitute of Biomedical Engineering, University of Toronto, CanadaDivision of Hematology, Department of Medicine, University of Toronto, Canada
There is a need to incorporate multiple tissues into in vitro OA models to evaluate novel therapeutics. This approach is limited by inherent donor variability. We present an optimized research tool: a human OA cartilage-synovium explant co-culture model (OA-EXM) that employs donor-matched lower and upper limit response controls combined with statistical approaches to address variability. Multiple rapid read-outs allow for evaluation of therapeutics while cataloguing cartilage-synovium interactions.
Design
48-h human explant cultures were sourced from OA knee arthroplasties. An OA-like cartilage-synovium co-culture baseline was established relative to donor-matched upper limit supraphysiological pro-inflammatory cytokine and lower limit OA cartilage or synovium alone controls. 100 nM dexamethasone treatment validated possible “rescue effects” within the OA-EXM dual tissue environment. Gene expression, proteoglycan loss, MMP activity, and soluble protein concentrations were analyzed using blocking and clustering methods.
Results
The OA-EXM demonstrates the value of the co-culture approach as the addition of OA synovium increases OA cartilage proteoglycan loss and expression of MMP1, MMP3, MMP13, CXCL8, CCL2, IL6, and PTGS2, but not to the extent of supraphysiological stimulation. Conversely, OA cartilage does not affect gene expression or MMP activity of OA synovium. Dexamethasone shows dual treatment effects on synovium (pro-resolving macrophage upregulation, protease downregulation) and cartilage (pro-inflammatory, catabolic, and anabolic downregulation), and decreases soluble CCL2 levels in co-culture, thereby validating OA-EXM utility.
Conclusions
The OA-EXM is representative of late-stage OA pathology, captures dual interactions between cartilage and synovium, and combined with statistical strategies provides a rapid, sensitive research tool for evaluating OA therapeutics.
Crucial role of macrophages in matrix metalloproteinase-mediated cartilage destruction during experimental osteoarthritis: involvement of matrix metalloproteinase 3.
. There is a need for ex vivo assessment tools of OA therapeutics that can dually account for human cartilage-synovium tissue interactions.
Single chondrocyte or synoviocyte cultures have been mainstays of in vitro OA research; cells are typically expanded from primary tissue or a progenitor cell source. Alternatively, cartilage or synovium are kept intact as tissue explants to retain endogenous matrix topology as a trade-off for increased heterogeneity. Single tissue cultures are often treated with pro-inflammatory cytokine stimulation or mechanical stressors to mimic disease that are in far excess of physiological levels
. Although single tissue investigations allow greater control over experimental conditions, they silo outcomes and do not reflect key crosstalk in OA pathology
, given its predominance of synovial hyperplasia and inflammation. Recently, cartilage-synovium co-culture explant models of healthy animal tissue treated with compression injury or exogenous cytokines have been used to induce an OA-like state
Interleukin-1 receptor antagonist (IL-1Ra) is more effective in suppressing cytokine-induced catabolism in cartilage-synovium co-culture than in cartilage monoculture.
Osteoarthritic synovial tissue inhibition of proteoglycan production in human osteoarthritic knee cartilage: establishment and characterization of a long-term cartilage-synovium coculture.
that demonstrated proteoglycan synthesis inhibition by OA synovium without additional pro-inflammatory stimuli; this effect could be dampened by anti-inflammatory corticosteroid rescue. This study importantly suggests that a human cartilage-synovium explant co-culture is capable of capturing known pathophysiological effects of synovium on cartilage and OA clinical outcomes
. However, this study used an extended timeline and did not address human donor heterogeneity, which is a limiting factor in widespread use of co-culture explant models as a research tool.
Building on existing contributions, we present a “human OA cartilage-synovium explant co-culture research model” (referred to as OA-EXM). The OA-EXM encompasses an entire suite of culture conditions, methods, and statistical strategies to establish a useful late-stage OA cartilage-synovium baseline to evaluate therapeutics, with significant features including: (1) optimized readouts at a rapid 48-h timepoint sensitive to significant changes in multiple variables; (2) statistical analyses and clustering strategies to account for donor-derived variability; (3) an expanded toolkit of simultaneous, quantitative readouts of cartilage and synovium gene expression, glycosaminoglycan (GAG) loss, soluble matrix metalloproteinase (MMP) activity and soluble protein concentrations. The OA-EXM provides a 48-h “snapshot” of multivariate outcomes in an intact, human disease setting accounting for dual cartilage-synovium tissue interactions and is thus a useful research tool for evaluating OA therapeutics.
Method
Total knee arthroplasty processing and explant tissue culture
Cartilage and synovium from late-stage OA (Kellgren–Lawrence grade 3–4) total knee arthroplasties (Table I) were acquired with patient consent and institutional ethics approval (UHN REB 14-7483-AE). Full-depth cartilage is cut from subchondral bone and sectioned with a biopsy punch (1.5 mm diameter) and pooled before random distribution per 20 or 40 mg wet weight into 24-well plates (800 μL). Synovium is minced with scissors after removing fat tissue, then distributed per 40 mg wet weight into 0.4 μm transwell inserts (100 μL).
Table ITissue donor demographics for total knee replacement explants
Explant medium is a 1:1 ratio of Roswell Park Memorial Institute (RPMI) 1640 and Dulbecco's modified Eagle medium (DMEM), low glucose, supplemented with 1% fetal bovine serum (FBS), 1% gentamicin, 1 mM sodium pyruvate, 50 μg/mL l-proline, 50 μg/mL L-ascorbic acid 2-phosphate, and 1X insulin-transferrin-selenium (ITS-G). Explant tissue is incubated separately for a 2-day acclimatization period
Prostaglandin PGE2 at very low concentrations suppresses collagen cleavage in cultured human osteoarthritic articular cartilage: this involves a decrease in expression of proinflammatory genes, collagenases and COL10A1, a gene linked to chondrocyte hypertrophy.
Explant tissue and conditioned medium are harvested at days 2 and 7 of culture by snap freezing and stored at −80°C for analysis.
Explant tissue co-culture experimental design for statistical analyses
The OA-EXM [Fig. 1(A)] is constructed with experimentally controlled fixed factors of multiple levels for compatibility with mixed linear modeling. Recommended options for analysis are detailed in Statistics.
Fig. 1Human OA joint explant model (OA-EXM) consisting of cartilage-synovium co-culture is viable for 2–7 days. (A) Schematic of model. Cartilage and synovium from late-stage OA knee replacements are processed for a 48-h acclimatization period. Cartilage controls are established alone (baseline; SOLO-BASE) or with OSM and IL1β as a positive pro-inflammatory control (SOLO-POS). Synovium controls are alone (SOLO-BASE) or pro-inflammatory IFNγ-treated control (SOLO-POS). Cartilage and synovium in co-culture (CCUL-BASE) is representative of an OA baseline state. Treatment within the cartilage-synovium co-culture is validated in this iteration using 100 nM dexamethasone (CCUL-DEXA). (B) Representative (N = 1) DAPI stain on cartilage and synovium show distribution of cells in day 7 explant co-cultures. Bars represents 400 μm. (C) Representative (N = 1) hematoxylin & eosin staining of synovium show structure before and after day 7 explant co-culture. Bars represent 100 μm. (D) Representative (N = 1) cell populations within synovium digest on day of knee replacement. A CD14+ population is identified within a CD45+/CD90- lymphocyte gate. (E) Representative (N = 3) Safranin-O/Fast Green staining of cartilage show proteoglycan content in day 7 explant culture between cartilage SOLO-BASE and pro-inflammatory SOLO-POS.
Two fixed factors are used: “Culture” and “Treatment”. “Culture” is a 2-level factor: (1) single tissue alone (SOLO; cartilage-alone or synovium-alone) or (2) cartilage-synovium co-culture (CCUL). “Treatment” is a 3-level factor: (1) baseline (BASE; medium alone); (2) positive cytokine control (POS; 5 ng/mL oncostatin M (OSM) and 5 ng/mL interleukin 1β (IL1β) for cartilage, or 100 ng/mL interferon γ (IFNγ) for synovium; PeproTech); or (3) corticosteroid (DEXA; 100 nM dexamethasone
Dexamethasone treatment alters the response of human cartilage explants to inflammatory cytokines and mechanical injury as revealed by discovery proteomics.
; BioReagent). “Treatment” is nested under “Culture” for a total of four experimental “Culture-Treatment” groups per tissue: SOLO-BASE, SOLO-POS, CCUL-BASE, CCUL-BASE clusters, dexamethasone treatment (CCUL-DEXA). SOLO-BASE and supraphysiological pro-inflammatory positive control (SOLO-POS) represent the lower and upper limit responses specific to each donor. SOLO-POS uses supraphysiological cytokines to elicit known pro-inflammatory and pro-catabolic effects tailored specifically for cartilage
. The OA-EXM therefore comprises of the OA co-culture environment testing baseline (CCUL-BASE) with established upper (SOLO-POS) and lower (SOLO-BASE) limits. The OA-EXM also includes test therapeutic “Treatment” groups on top of its co-culture platform; in this study, we validate the current version of the OA-EXM with known therapeutic, dexamethasone, generating an additional CCUL-DEXA subset.
Gene expression by quantitative polymerase chain reaction (qPCR)
Frozen explant tissue is fragmented using a liquid nitrogen pre-chilled BioPulverizer (BioSpec) for RNA isolation using a Plant Total RNA Mini Kit (Geneaid) for cartilage or RNeasy Plus Universal Mini Kit (QIAGEN) for synovium. cDNA is generated with SuperScript™ IV VILO™ Master Mix (Invitrogen). qPCR is run on cartilage- and synovium-specific gene panels (Supplementary Table 1) using custom primers (Supplementary Table 2, Invitrogen) and FastStart Universal SYBR Green Master Mix (Roche) on a QuantStudio™ 5 (ThermoFisher) or 7900HT (Applied Biosystems). Results are normalized (ΔΔCT), represented as log2(fold-change) against reference genes B2M, TBP, and RPL13A (geometric mean) and to the donor-matched single tissue baseline (SOLO-BASE) median. Undetected CT values of samples are imputed as 40 for fold-change calculations.
Proteoglycan loss, soluble proteins, and soluble protease activity
Proteoglycan loss represented by sulfated GAG concentration in medium is detected by dimethylmethylene blue (DMMB) assay
against a chondroitin sulfate (sodium salt from shark cartilage, Sigma–Aldrich) standard A525 curve. Soluble protein concentrations in medium are detected using LEGENDplex™ immunoassay (BioLegend) on a FACSCanto™ II (BD Biosciences). Values below or above the limit of detection were imputed as half or twice the lower and upper limits if ≥ 67% of samples were within range. Soluble total MMP activity is measured using the MMP Activity Assay kit (abcam). All kits are used according to manufacturer's protocol.
Statistics
JMP Pro 14 (SAS) is used. Each randomized well is assigned into an experimental “Culture-Treatment” group with 3 technical replicates per group from a single arthroplasty. “Donor” is used as a random effect, blocking factor, or normalization factor in statistical analyses to account for donor variability; each donor is considered a biological replicate for each experimental group. The number of replicates and “Culture-Treatment” groups differ between donors due to yield variability in surgical tissue acquisition. Table II details statistical methods used to account for donor variability. Supplementary Fig. 1 illustrates the study datasets, sample sizes, and selected analyses as a practical guideline to adopt the OA-EXM.
Table IIStatistical methods can account for donor variability depending on experimental data and groups
Readout
MethodA
GroupingB
Donor as FactorC
Soluble GAG loss
Mixed linear regression
Yes
Yes, as Random variance factor
Block-centered t-test or ANOVA
Yes
Yes, as Block factor (equal n replicates within blocks)
Gene expression fold-change profile
Hierarchical clustering
No
No, unbiased multivariate clustering with donor-matched ΔΔCT normalization
Discriminant analysis
Yes
No, multivariate clustering based on known experimental group with donor-matched ΔΔCT normalization
Secreted protein concentration
Hierarchical clustering
No
Yes, unbiased multivariate clustering with donor as post hoc variable
Block-centered t-test or ANOVA
Yes
Yes, as Block factor (equal n replicates within blocks)
MMP activity
Block-centered t-test or ANOVA
Yes
Yes, as Block factor (equal n replicates within blocks)
A = Statistical method used in JMP 14 software. B= If the method uses “Culture-Treatment” experimental grouping or if the analysis is unbiased. C=How the OA tissue donor is accounted for in the statistical method.
ANOVA, analysis of variance; GAG, glycosaminoglycan; MMP, matrix metalloproteinase.
Linear regression (Fit Model, Mixed Model) was used for GAG loss. Regression assumptions are fulfilled by test of normal residuals (Supplementary Fig. 2). Effect size of each factor is represented by mixed model parameter estimates.
Unbiased hierarchical clustering (Ward method) is used for multivariate gene and soluble protein profiles. “Donor” labels are appended post hoc. Discriminant analysis (Linear, common covariance method; grouped by “Culture-Treatment”) is used for dimensionality reduction. The percent misclassified and entropy R2 value indicate the fit of “Culture-Treatment” groups into distinct clusters. Gene expression fold-change outcomes are summarized by mean and 95% confidence interval (CI) using un-pooled estimates of standard error. Statistical testing of individual genes was omitted in favour of cluster-based classification representative of overall effects to expression profile.
Block-centered two-sided t-test or analysis of variance (ANOVA; Fit Y by X) with “Donor” as the blocking factor is used when “Culture-Treatment” groups are matched in number. Normality and independence were assumed. Outcomes are summarized by mean and 95% CI using pooled estimates of standard error. P-values are reported with corrections for multiple comparisons (Tukey–Kramer HSD or Dunnett's Test).
Results
Co-culture environment of the OA-EXM is conducive to structural maintenance and viability
Osteoarthritic synovial tissue inhibition of proteoglycan production in human osteoarthritic knee cartilage: establishment and characterization of a long-term cartilage-synovium coculture.
Pre-treatment of human mesenchymal stem cells with inflammatory factors or hypoxia does not influence migration to osteoarthritic cartilage and synovium.
. Media supplementation ensures explant viability for up to 7 days (Supplementary Fig. 3), with mass per well of cartilage and synovium based on feasible harvest per donor and minimum-volume requirements for downstream multivariate analysis. Low serum-supplemented conditions allow investigation of multiple tissues without serum growth factor interference and is conducive to future investigations of cellular therapies.
Cartilage and synovium morphology are maintained for 7 days as seen by 4′,6-diamidino-2-phenylindole (DAPI) nuclear [Fig. 1(B)] and hematoxylin & eosin [Fig. 1(C)] staining. Both cluster of differentiation (CD) 90+ fibroblasts and CD45+/CD14+ monocytes/macrophages are present within explant synovium [Fig. 1(D)], corroborating with reports of elevated percentages of macrophage populations in the OA joint
Enumeration and localization of mesenchymal progenitor cells and macrophages in synovium from normal individuals and patients with pre-osteoarthritis or clinically diagnosed osteoarthritis.
. Proteoglycan retention in cartilage explants at baseline contrasts with visible loss when treated with supraphysiological of pro-inflammatory cytokines [Fig. 1(E)], demonstrating the upper limit response of cartilage degradation within the OA-EXM.
Cartilage-synovium co-culture conditions in the OA-EXM set an OA-like baseline demonstrating significant effects of synovium on cartilage matrix loss, anabolic and inflammatory gene expression
OA cartilage alone (SOLO-BASE) has a lower limit of steady GAG loss (μg/mL) at baseline at 48-h while supraphysiological cytokine stimulation (SOLO-POS) defines the upper limit [Fig. 2(A)]. Addition of OA synovium in co-culture (CCUL-BASE) significantly increases GAG loss (P = 0.0479) compared to the lower limit, but not as much as the upper limit (P < 0.09001). Using donor-specific limits, least squares regression (Fig. 2(B), Table III) accounts for “Donor” contribution to total variance and reattributes effect estimates of “Culture” and “Treatment”. This mixed linear analysis predicts GAG loss of SOLO-BASE, CCUL-BASE, and SOLO-POS as: 0.838 (0.595, 1.07), 1.14 (0.908, 1.38), and 3.03 (2.79, 3.27). The OA-EXM thus captures additional GAG loss in cartilage-synovium co-culture compared to cartilage alone, with further capacity to statistically account for donor variability using upper and lower limits.
Fig. 2OA-EXM cartilage-synovium co-culture baseline increases cartilage proteoglycan loss when accounting for tissue donor heterogeneity as a factor. (A) GAG loss (μg/mL) in day 2 conditioned medium can be represented by scatterplot of raw concentrations by “Culture-Treatment” group. Diamonds represent mean and 95% confidence interval, N = 19 donors ∗ 3 replicates. (B) GAG loss (μg/mL) represented by mixed linear regression of fixed effects and random effect of tissue donor on GAG loss. Magenta line represents 100% linear fit. Parameter estimates and P values are listed in Table III. Nested hierarchy of “Culture-Treatment” experimental groups, where “Culture” and “Treatment” are used as fixed effects, and “Donor” as random effect for regression analyses. “Treatment” level of dexamethasone (DEXA, greyed out) is analyzed as a separate subset [Fig. 5(B)].
The OA-EXM captures 48-h cartilage expression changes within a curated OA-relevant 28-gene panel representing extracellular matrix (ECM) metabolism, chemotaxis and adhesion, and inflammation (Supplementary Table 1). The OA-EXM detects the effects of synovium on cartilage in the CCUL-BASE group (Fig. 3(A), Supplementary Table 3): the top three downregulated genes are anabolic (ACAN, COL2A1, SPP1), while the top three upregulated genes are pro-inflammatory and chemotactic (CXCL8, IL6, NOS2). Matrix proteases (MMP1, MMP3, MMP13, ADAMTS4, ADAMTS5) and other pro-inflammatory (PTGS2, IL1B) genes are also upregulated. Addition of synovium affects 48-h cartilage gene expression in the same direction as SOLO-POS, albeit at a diminished magnitude. Individual donor effects are accounted for as fold-change in gene expression is relative to each donor's cartilage-alone expression (SOLO-BASE).
Fig. 3OA-EXM cartilage-synovium co-culture baseline downregulates anabolic matrix and upregulates pro-inflammatory cartilage gene expression, which correspond to pro-inflammatory changes to the soluble factor profile. (A) Day 2 cartilage gene expression as log2(fold change) relative to reference genes and cartilage baseline (SOLO-BASE). Diamonds represent mean and 95% confidence interval. All experimental groups represented are significantly different from SOLO-BASE by post-hoc Dunnett comparison (p < 0.05), N = 21 donors ∗ 3 replicates. (B) Soluble factor profile represented by hierarchical clustering and (C) select block-centered scatterplots comparing cartilage baseline (SOLO-BASE, cartilage), synovium baseline (SOLO-BASE, synovium) and cartilage-synovium baseline (CCUL-BASE). Heatmap represents high-to-low protein concentration with maximum–minimum relative to each secreted protein (where minimum ≥0 pg/mL). Diamonds represent mean and 95% confidence interval, block factor = donor, ∗ = p < 0.05 by post-hoc Tukey–Kramer comparison, N = 4 donors ∗ 3 replicates.
The OA-EXM concurrently captures 48-h synovium gene expression changes within a curated 9-gene panel focused on macrophage polarization and fibrosis (Supplementary Table 4). Supraphysiological IFNγ (SOLO-POS) upregulates fold-change of NOS2 and pro-inflammatory macrophage surface marker HLADRA. In contrast to cartilage gene expression, the OA-EXM shows non-significant change in synovium gene expression with the addition of OA cartilage (CCUL-BASE) within the 48-h co-culture period. Again, donor-matched normalization to synovium-alone (SOLO-BASE) accounts for donor variability.
An alternative configuration of the OA-EXM explored increasing the cartilage: synovium ratio from 1:2 (20:40 mg) to 1:1 (40:40 mg). The effects of synovium on cartilage are maintained despite reduced synovium per cartilage (Supplementary Fig. 4). Meanwhile, increasing the cartilage: synovium ratio does not change synovium gene expression or produce significant effects of cartilage on synovium (Supplementary Fig. 5). Given the lack of changes to synovium even in the presence of additional OA cartilage, the suggested cartilage: synovium ratio for the OA-EXM is more practically limited to the lower 1:2 ratio due to limited availability of cartilage from end-stage OA arthroplasty.
The OA-EXM captures donor-specific soluble protein profiles
The OA-EXM allows for analysis of soluble factors from both cartilage and synovium. Soluble OA-relevant
cytokines (interleukin (IL) IL6, IFNγ, IL10, IL1β, tumor necrosis factor (TNF) α), chemokines (chemokine (C–C motif) ligand (CCL) 2, chemokine (C-X-C motif) ligand (CXCL) 8, CXCL10), and adipokines (adiponectin, resistin, leptin, adipsin) were measured in a donor subset (N = 4). . All proteins, except IL10, IL1β, and TNFα, were present in ≥67% of samples (Supplementary Table 5).
Both OA cartilage and synovium alone produce detectable levels of the remaining 9 proteins. IL6, IFNγ, CCL2, CXCL8, adiponectin, resistin, leptin, and adipsin were found in higher concentrations in the OA synovium-alone, whereas CXCL10 was primarily secreted by OA cartilage. Further, the OA-EXM captures a differential co-culture baseline soluble factor profile compared to OA cartilage or synovium alone. All soluble factors except CXCL10 are higher in co-culture than cartilage-alone [Fig. 3(B)]. Comparing co-culture with synovium-alone with donor blocking (P < 0.005 for all proteins), CCL2 and IL6 have the greatest differences [Fig. 3(C)]. Both CCL2 and IL6 levels are higher in cartilage-synovium co-culture compared to synovium alone (P = 0.097; P = 0.0081, respectively). All means and comparisons are found in Supplementary Tables 6 and 7 The cartilage-synovium baseline in the OA-EXM is a distinctive soluble factor environment of OA-relevant factors that is comparable to OA synovial fluid, albeit with higher levels of chemokines CCL2 and CXLC8 and lower levels of adipokines
Hierarchical clustering [Fig. 3(B)] suggests that cartilage-alone profiles are donor independent, whereas synovium-alone profiles individually form donor-dependent clusters. The co-culture baseline profile, although distinct from both cartilage- or synovium-alone, is influenced by both the cartilage and synovium, and thus also clusters in a donor-dependent manner.
The OA-EXM is validated by testing with dexamethasone
To validate the OA-EXM as a research tool for evaluating therapeutics, we tested a known anti-inflammatory corticosteroid, dexamethasone
, adding 100 nM of dexamethasone to the cartilage-synovium co-culture. Cartilage and synovium gene expression, co-culture soluble factors, GAG loss, and MMP activity 48-h readouts were tested in this validation dataset. The OA-EXM captures multiple changes in OA cartilage and synovium mediated by dexamethasone.
With dexamethasone treatment in co-culture (CCUL-DEXA), catabolic and pro-inflammatory genes MMP1, MMP13, CCL2 are downregulated when previously they were upregulated by the effect of synovium on cartilage (CCUL-BASE) relative to cartilage-alone (SOLO-BASE; Supplementary Table 3, Supplementary Fig. 6). MMP3, CXCL8, and IL6 remain upregulated relative to SOLO-BASE, but decreased in magnitude compared to CCUL-BASE. Dexamethasone treatment also worsens the anti-anabolic effect of synovium observed in cartilage-synovium baseline. We further break down the mixed effects of dexamethasone in an additional subset that shows similar concurrent catabolic MMP and anabolic collagen downregulation when OA cartilage-alone is treated with dexamethasone (Supplementary Fig. 7).
These observations are complemented by direct discriminant analysis [Fig. 4(A)]; dexamethasone treatment (CCUL-DEXA) produces a cartilage gene expression profile that is significantly different from OA co-culture baseline (CCUL-BASE) and is furthest distanced from the supraphysiological pro-inflammatory upper limit (SOLO-POS). Fig. 4(A) represents the log2(fold-change) values from the cartilage gene expression panel of 28 outputs reduced to three canonical dimensions. This differentiates the four “Culture-Treatment” groups of SOLO-BASE (cartilage), SOLO-POS (cartilage), CCUL-BASE, and CCUL-DEXA with only 5% sample misclassification (Supplementary Table 8). Adding dexamethasone (CCUL-DEXA) not only “rescues” the profile from CCUL-BASE to SOLO-BASE (downwards on Canonical3), but further shifts on both the Canonical1 and 2 axes. Unbiased two-way hierarchical clustering [Fig. 4(B)] similarly shows the four “Culture-Treatment” groups as separate clusters.
Fig. 4OA-EXM detects promotion of anti-inflammatory gene expression profiles in cartilage and synovium induced by dexamethasone. (A) Discriminant canonical plot of log2(fold change) relative to SOLO-BASE gene expression profiles of cartilage (top, N = 21 donors ∗ 3 replicates) and synovium (bottom, N = 6 donors ∗ 3 replicates). 3D ellipsoids represent 95% confidence interval. (B–C) Hierarchical clustering of log2(fold change) relative to SOLO-BASE cartilage (B) and synovium (C) gene expression. Suggested clusters are annotated by labels below heatmap. Heatmap indicates upregulation-downregulation of expression with maximum–minimum scale relative to each gene.
In parallel, dexamethasone alters the synovium gene expression profile away from cartilage-synovium co-culture baseline (CCUL-BASE; Fig. 4(A)). Dexamethasone added in co-culture downregulates expression of MMP1 and MMP3 combined with upregulation of pro-resolving macrophage scavenger receptors
CD163 and MRC1 normalized to synovium-alone (SOLO-BASE; Supplementary Table 4, Supplementary Fig 8). An additional subset demonstrates similar MMP1, MMP3 downregulation and CD163, MRC1 upregulation in synovium-alone treated with dexamethasone (Supplementary Fig. 9).
Discriminant analysis (Fig. 4(A), Supplementary Fig. 10) and unbiased hierarchical clustering [Fig. 4(C)] confirms an overlap between the synovium SOLO-BASE and CCUL-BASE groups, which is the source of a 19.6% misclassification for synovium gene expression (Supplementary Table 8). Addition of pro-inflammatory IFNγ to synovium (SOLO-POS) shifts the cluster away from both SOLO-BASE and CCUL-BASE synovium profiles. Regardless of the overlap between SOLO-BASE and CCUL-DEXA produces a separate expression profile shifted from these two groups (downwards on Canonical1).
Treatment with dexamethasone (CCUL-DEXA) significantly reduces MMP activity [Fig. 5(A)], but not GAG loss [Fig. 5(B)] compared to co-culture baseline (CCUL-BASE). Soluble MMP activity was significantly reduced with CCUL-DEXA compared to CCUL-BASE (P = 0.0397) when accounting for donor blocking (donor effect P < 0.0001). In contrast, there is no difference (P = 0.181) in GAG loss between CCUL-BASE and CCUL-DEXA, even after accounting for significant donor effect (P = 0.0177). This is consistent with downregulation of MMP1 and MMP3 expression in the CCUL-DEXA group relative to SOLO-BASE and CCUL-BASE for both cartilage (Fig. 4(B), Supplementary Fig. 6) and synovium (Fig. 4(C), Supplementary Fig. 8).
Fig. 5OA-EXM detects mixed effects across soluble MMP activity, proteoglycan (GAG) loss, and chemokine levels that are donor dependent after dexamethasone treatment. (A) Block-centered scatterplot of soluble MMP activity (log10(RFU/min)). SOLO-BASE represents synovium baseline. Diamonds represent mean and 95% confidence interval, block factor = donor, ∗ = p < 0.05 by post-hoc Dunnett comparison against CCUL-BASE, N = 3 donors ∗ 3 replicates. (B) Block-centered scatterplot of GAG loss (μg/mL). SOLO-BASE represent cartilage baseline. Diamonds represent mean and 95% confidence interval, block factor = donor, ∗ = p < 0.05 by post-hoc Dunnett comparison against CCUL-BASE, N = 3 donors ∗ 3 replicates. (C) Hierarchical clustering of secreted factor profile comparing cartilage-synovium baseline (CCUL-BASE) and dexamethasone rescue (CCUL-DEXA). Heatmap represents high-low protein concentration with maximum–minimum relative to each secreted protein (where minimum ≥0 pg/mL). (D) Block-centered scatterplot of secreted factor levels of CCL2 (left) and CXCL8 (right). Diamonds represent mean and 95% confidence interval, block factor = donor, ∗ = p < 0.05 by pooled t-test, N = 3 donors ∗ 3 replicates.
The OA-EXM detects changes to soluble CCL2 and CXCL8 with dexamethasone treatment while others were not significantly altered (Fig. 5(C), Supplementary Table 9). Addition of dexamethasone decreases CCL2 compared to co-culture baseline (P = 0.0072, Fig. 5(D)). CXCL8 concentrations unexpectedly increased (P = 0.0116) with dexamethasone treatment. Unbiased hierarchical clustering [Fig. 5(C)] of soluble protein concentrations shows no distinct clustering of dexamethasone (CCUL-DEXA) and co-culture baseline (CCUL-BASE) groups. Only one of three donors shows separation between CCUL-DEXA and CCUL-BASE soluble protein profiles within its donor cluster.
Taken together, the OA-EXM is a sensitive tool that captures the nuanced effects of dexamethasone in terms of reduced pro-inflammatory and catabolic cartilage and synovium gene expression, reduced MMP activity, and reduced soluble CCL2 levels without concomitant rescues in GAG loss or soluble CXCL8levels at a 48-h “snapshot”.
Discussion
We present a novel human OA joint explant cartilage-synovium co-culture model (OA-EXM) that includes a full suite of explant culture protocols, experimental designs to account for an OA co-culture baseline with donor-specific upper and lower limit controls, treatment groups, as well as appropriate statistical methods. The OA-EXM rapidly captures dual cartilage and synovium gene expression, soluble protein concentrations, GAG loss, and MMP activity at a rapid 48-h timepoint “snapshot”. Human late-stage OA explants allow for non-culture expanded cells with baseline human disease phenotypes. The OA-EXM queries an end-stage OA-like baseline using cartilage-synovium co-culture; supraphysiological factors are only used to capture donor-matched upper limits of multiple readouts. Tissue donor heterogeneity, often a deterrent to the use of human explant tissue is directly addressed by use of single tissue and supraphysiological cytokine controls to capture donor-specific response ranges combined with various statistical tools such as clustering and incorporation of donor as a block factor in sample sizes ranging from N = 3 to N = 21. The OA-EXM incorporates OA synovium effects and confirms increased GAG loss, pro-inflammatory and catabolic cartilage gene expression relative to OA cartilage-alone, and reduced secretion of CXCL10. Conversely, the OA-EXM shows minimal effects of OA cartilage on OA synovium regardless of cartilage: synovium co-culture ratios in terms of synovium gene expression at 48-h.
Dexamethasone was used to validate the utility of the OA-EXM. Clinically, dexamethasone has shown acute paint reduction in OA
Effects of short-term glucocorticoid treatment on changes in cartilage matrix degradation and chondrocyte gene expression induced by mechanical injury and inflammatory cytokines.
. The OA-EXM captured dual effects of dexamethasone on cartilage and synovium, inhibiting cartilage catabolic, chemotactic, and pro-inflammatory gene expression as previously reported
Dexamethasone treatment alters the response of human cartilage explants to inflammatory cytokines and mechanical injury as revealed by discovery proteomics.
Effects of short-term glucocorticoid treatment on changes in cartilage matrix degradation and chondrocyte gene expression induced by mechanical injury and inflammatory cytokines.
while upregulating pro-resolving macrophage marker expression and downregulating both synovium MMP gene expression in synovium and concomitant MMP activity. Dexamethasone has been observed to have strong pro-resolving macrophage polarization effect
Unexpectedly, the OA-EXM captured non-significant effects of dexamethasone treatment on GAG loss; the model captures downregulated cartilage expression of anabolic matrix genes to magnitudes greater than anabolic downregulation in the OA cartilage-synovium co-culture baseline. This aligns with controversial clinical reports of greater cartilage loss and radiographic OA progression
with repeat corticosteroid treatment. Despite a consensus on dexamethasone's effects in MMP downregulation, mixed effects on apoptosis and GAG loss have been reported in cell and cartilage explant cultures
. Inhibition of MMP expression and activity concurrently with anabolic matrix gene downregulation has not been previously reported in a human arthritic co-culture setting. The OA-EXM simultaneously captures mixed effects of dexamethasone and can thus be a useful research tool for investigating dosage, duration, and novel formulation or delivery strategies for treatments including dexamethasone conjugates
that may recapitulate beneficial effects without activating anti-anabolic pathways.
The OA-EXM has a number of important limitations. Our data shows that it can be vaunted for its rapid 48-h “snapshot” readout utility and 7-day viability, but it is not optimized for prolonged culture durations. Longer culture durations could capture more extensive matrix remodeling or long-term treatment effects with additional histological and matrix protein quantification. Long-term explant cultures have been previously used
Osteoarthritic synovial tissue inhibition of proteoglycan production in human osteoarthritic knee cartilage: establishment and characterization of a long-term cartilage-synovium coculture.
and are valuable in modeling extended changes in OA. Further, the OA-EXM does not use healthy or early-stage joint tissue and thus is not currently configured for capturing longitudinal disease evolution.
The OA-EXM configuration is optimized with considerations to ease-of-use and to primarily establish a controlled human OA-like cartilage-synovium baseline environment using donor-matched tissues and control conditions. The use of donor-matched single tissue controls allows the capture of a range of response to test therapeutics, but only in relevant to late-stage OA. Incorporation of additional joint components including fat pad, synovial fluid, infiltrating monocytes/macrophages, and subchondral bone would be considered in future iterations of the OA-EXM research tool. Additionally, the use of a transwell system recapitulates soluble factor interactions between cartilage and synovium that are more reflective of physiological conditions
but does not account for possible direct tissue interactions.
The OA-EXM did not capture significant effects of cartilage on synovium. The synovium gene panel was curated toward macrophage-mediated response and was thus less comprehensive than the cartilage panel. Further, the synovium is a heterogeneous tissue with additional variability arising from its mixed cellular composition that could obfuscate cell population-specific changes. Despite this, incorporation of OA synovium is essential to the OA-EXM as a contributor of active soluble proteins and its multiple effects on OA cartilage. Single cell RNA sequencing, cell sorting, or flow cytometry could be integrated into a future OA-EXM toolset to parse meaningful synovial changes in face of population heterogeneity.
Although the OA-EXM accounts for donor variability through upper and lower limit control groups and statistical strategies, this utility is limited to determining the significance of treatment or culture effects as a whole. In its current iteration, the OA-EXM effectively screens for novel therapies with significant effects but cannot predict which patients could be responders to test treatments. Within our sample datasets, the effects of OA synovium on cartilage were of greater magnitude in a subset of donors, suggesting the existence of donor subtypes that could be segregated within the OA-EXM. Annotation of these donors with existing clinical data yielded no correlations (data not shown). More sensitive clinical measurements and potential biomarkers
could be used in future correlations to patient heterogeneity.
The OA-EXM allows for rapid 48-h multivariate readouts that provide holistic simultaneous “snapshots” of the OA joint cartilage and synovium tissue's inflammatory, anabolic, and catabolic status. The model is sufficiently sensitive to capture potentially favourable and less favourable gene, protein, enzyme activity and proteoglycan losses upon challenge with a well-known anti-inflammatory treatment, and thus represents a useful tool to investigate novel OA therapeutics. Importantly, the OA-EXM as presented incorporates statistical tools and controls to capture donor heterogeneity characteristic to OA. The OA-EXM represents a significant advance in in vitro tools available to OA researchers to evaluate therapies and their mechanisms.
Author contributions
All authors have made substantial contributions to all three requirements for authorship. MWYC, AGA, SV contributed to the design and conception of the study. MWYC and AGA contributed to conducting experiments. MWYC, AGA, NM, and RG contributed to acquiring reagents, patient tissue, and data. MWYC, AGA, SV contributed to analyzing and interpreting data. All authors contributed to the writing and final review of the manuscript.
All authors declare no conflicts of interest within the scope of this manuscript. MWYC and RG declare no conflicts of interest exist outside of this scope. AGA receives income from BlueRock Therapeutics and holds co-inventorship on the patent WO2007128115A1 on immune privileged and modulatory progenitor cells. NM declares equity in biotechnology company Arthritis Innovation Corp, and equity and funds from Arthur Health as its Chief Executive Officer. SV declares 60% ownership in Regulatory Cell Therapy Consultants, Inc., a private regulatory consulting company.
Role of the funding source
This research is funded by the Arthritis Society (grants #YIO-15-321 to SV and #TGP-18-0195 to MWYC). The funding source was not involved in the study's design, interpretation of data, and writing or submission of the manuscript for publication.
Acknowledgments
We thank Kim Perry, Amanda Weston, Tamara Wagner, and the Orthopedic Surgery team for assistance with acquisition of donor samples. This work is in part supported by the Schroeder Arthritis Institute via the Toronto General and Western Hospital Foundation (University Health Network). Flow cytometry was performed in the Toronto Western KDT-UHN Flow Cytometry Facility, with funding from the Canada Foundation for Innovation, and Toronto General and Western Hospital Foundation. We thank Dr. A. Robin Poole for his expert advice on working with joint explant tissues and matrix degradation assays.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
Crucial role of macrophages in matrix metalloproteinase-mediated cartilage destruction during experimental osteoarthritis: involvement of matrix metalloproteinase 3.
Interleukin-1 receptor antagonist (IL-1Ra) is more effective in suppressing cytokine-induced catabolism in cartilage-synovium co-culture than in cartilage monoculture.
Osteoarthritic synovial tissue inhibition of proteoglycan production in human osteoarthritic knee cartilage: establishment and characterization of a long-term cartilage-synovium coculture.
Prostaglandin PGE2 at very low concentrations suppresses collagen cleavage in cultured human osteoarthritic articular cartilage: this involves a decrease in expression of proinflammatory genes, collagenases and COL10A1, a gene linked to chondrocyte hypertrophy.
Dexamethasone treatment alters the response of human cartilage explants to inflammatory cytokines and mechanical injury as revealed by discovery proteomics.
Pre-treatment of human mesenchymal stem cells with inflammatory factors or hypoxia does not influence migration to osteoarthritic cartilage and synovium.
Enumeration and localization of mesenchymal progenitor cells and macrophages in synovium from normal individuals and patients with pre-osteoarthritis or clinically diagnosed osteoarthritis.
Effects of short-term glucocorticoid treatment on changes in cartilage matrix degradation and chondrocyte gene expression induced by mechanical injury and inflammatory cytokines.
00910-9&id=gr1.jpg){kind=link}
00910-9&id=gr2.jpg){kind=link}
00910-9&id=gr3.jpg){kind=link}
00910-9&id=gr4.jpg){kind=link}
00910-9&id=gr5.jpg){kind=link}