Article pubs.acs.org/molecularpharmaceutics
Suitability of Porcine Chondrocyte Micromass Culture To Model Osteoarthritis in Vitro Niels Schlichting,†,‡ Tilo Dehne,*,†,‡ Karsten Mans,§ Michaela Endres,∥ Bruno Stuhlmüller,§ Michael Sittinger,† Christian Kaps,∥ and Jochen Ringe† †
Tissue Engineering Laboratory & Berlin-Brandenburg Center for Regenerative Therapies (BCRT), Department of Rheumatology and Clinical Immunology, CharitéUniversitätsmedizin Berlin, 10117 Berlin, Germany § Department of Rheumatology and Clinical Immunology, CharitéUniversitätsmedizin Berlin, 10117 Berlin, Germany ∥ TransTissue Technologies GmbH, 10117 Berlin, Germany S Supporting Information *
ABSTRACT: In vitro tissue models are useful tools for the development of novel therapy strategies in cartilage repair and care. The limited availability of human primary tissue and high costs of animal models hamper preclinical tests of innovative substances and techniques. In this study we tested the potential of porcine chondrocyte micromass cultures to mimic human articular cartilage and essential aspects of osteoarthritis (OA) in vitro. Primary chondrocytes were enzymatically isolated from porcine femoral condyles and were maintained in 96-multiwell format to establish micromass cultures in a high-throughput scale. Recombinant porcine tumor necrosis factor alpha (TNF-α) was used to induce OA-like changes documented on histological (Safranin O, collagen type II staining), biochemical (hydroxyproline assay, dimethylmethylene blue method), and gene expression level (Affymetrix porcine microarray, real time PCR) and were compared with published data from human articular cartilage and human micromass cultures. After 14 days in micromass culture, porcine primary chondrocytes produced ECM rich in proteoglycans and collagens. On gene expression level, significant correlations of detected genes with porcine cartilage (r = 0.90), human cartilage (r = 0.71), and human micromass culture (r = 0.75) were observed including 34 cartilage markers such as COL2A1, COMP, and aggrecan. TNF-α stimulation led to significant proteoglycan (−75%) and collagen depletion (−50%). Comparative expression pattern analysis revealed the involvement of catabolic enzymes (MMP1, -2, -13, ADAM10), chemokines (IL8, CCL2, CXCL2, CXCL12, CCXL14), and genes associated with cell death (TNFSF10, PMAIPI, AHR) and skeletal development (GPNMB, FRZB) including transcription factors (WIF1, DLX5, TWIST1) and growth factors (IGFBP1, -3, TGFB1) consistent with published data from human OA cartilage. Expression of genes related to cartilage ECM formation (COL2A1, COL9A1, COMP, aggrecan) as well as hypertrophic bone formation (COL1A1, COL10A1) was predominantly found decreased. These findings indicating significant parallels between human articular cartilage and the presented porcine micromass model and vice versa confirm the applicability of known cartilage marker and their characteristics in the porcine micromass model. TNF-α treatment enabled the initiation of typical OA reaction patterns in terms of extensive ECM loss, cell death, formation of an inflammatory environment through the induction of genes coding for chemokines and enzymes, and the continued... Special Issue: Engineered Biomimetic Tissue Platforms for in Vitro Drug Evaluation Received: Revised: Accepted: Published: © 2014 American Chemical Society
2092
January 19, 2014 March 13, 2014 March 17, 2014 March 17, 2014 dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
modulation of genes involved in skeletal development such as growth factors, transcription factors, and cartilage ECM-forming genes. In conclusion, the porcine micromass model represents an alternative tissue platform for the evaluation of innovative substances and techniques for the treatment of OA. porcine chondrocyte high-density micromass culture, comparative gene expression analysis, in vitro 3D model, KEYWORDS: tumor necrosis factor alpha
■
INTRODUCTION Osteoarthritis (OA) as the worldwide leading form of arthritis in the western world has particular medical and socioeconomic importance and is characterized by gradual destruction of the articular cartilage and subchondral bone.1 The etiology of this disease is multifactorial in which also mechanical factors play a central role.2 Although OA is considered as a noninflammatory disease, it is widely accepted that synovial inflammation is a common feature of OA, but its role in pathogenesis is not clarified yet.3,4 On the cellular level, chondrocytes can undergo phenotypic changes during OA progress summarized as categories of reaction pattern: (i) proliferation and cell death (apoptosis), (ii) changes in anabolic activity, (iii) degradation, (iv) phenotypic modulation of the articular chondrocytes, and (v) osteophyte formation.5 Finally, all these implications are leading to a loss of joint function, for which to date no cure is available.6 In vitro models are important tools for the development of novel therapeutic strategies in cartilage repair offering a costeffective analysis of potential active substances in highthroughput approaches under standardized conditions. To date, only few human OA in vitro models based on cell lines,7 explant cultures,8 monolayer cultures,9 and cultures providing a three-dimensional (3D) cell surrounding are available.10,11 In the future, induced pluripotent stem cells (iPS) and their chondrogenic-differentiated descendants may overcome the limited cell availability and open the perspective of personalized testing. So far, the generation and maintenance of iPS is costly and often results in mixed populations of terminal differentiated cells. Furthermore, the variability between cells derived from patients of different age and gender is wide as intensively reviewed for cardiac repair strategies.10−12 Hence, the limited availability of human cells hampers the implementation of extensive standardized testing. Instead, many studies with animal-derived cells have been conducted preferentially using small animal tissue.13,14 Although animal models are thought to be limited in terms of utility due to differences with the human system, such models can overcome the limited availability of primary tissue. Especially, large animalderived in vitro models such as pig, cow, and sheep have the potential to provide large batches of cells as already shown by many groups using these chondrocytes in in vitro cartilage models.15−17 Since the in vivo OA conditions of chondrocytes can only be maintained for a short time under standardized in vitro cell culture conditions, long-term evaluations require the addition of OA-inducing agents such as synovial fluid from OA patients,18 or at least OA mediators such as tumor necrosis factor alpha (TNF-α).19,20 In order to find an in vitro OA model for preclinical tests suitable for screening in high-throughput approaches, we evaluated the potential of porcine chondrocyte micromass culture to mimic essential aspects of OA. Therefore, suspension of porcine primary chondrocytes was maintained for 14 days in medium containing 10% serum without any inducing agents to
develop cartilage-like expression pattern and ECM rich in proteoglycans and collagens. To induce OA-like changes on established in vitro cartilage, TNF-α was supplemented for up to 14 days. The reaction pattern was documented on histological (Safranin O staining), biochemical (collagens, hydroxyproline assay; proteoglycans, DMMB method), and gene expression level (Affymetrix porcine microarray, RT-PCR), and results were compared with published data from human OA cartilage in order to assess the suitability of porcine chondrocyte micromass culture to model human OA in vitro.
■
METHODS Chondrocyte Isolation. Articular cartilage was harvested from the medial and lateral femoral condyle of domestic pigs (n = 6; 6−12 month old). An animal approval was not necessary because the samples were obtained from a slaughterhouse. Chondrocytes were isolated according to a protocol previously published.15 Briefly, cartilage slices were incubated for 19 h in spinner flasks containing RPMI 1640 medium, supplemented with 10% fetal bovine serum (FBS, v/v), 100 U/mL penicillin and 100 μg/mL streptomycin, 333.3 U/mL collagenase II (all Biochrom, Berlin, Germany), 1 U/mL collagenase P (Roche Diagnostics, Mannheim, Germany), and 33.3 U/mL hyaluronidase (Sigma-Aldrich, Steinheim, Germany). Subsequently, incubated cell suspensions were strained through a nylon mesh with 100 μm pore diameter (Becton Dickinson, Heidelberg, Germany) washed in Hanks solution (Biochrom), and resuspended in maintenance medium (RPMI 1640, 10% FBS, penicillin/streptomycin as above) supplemented with 170 μM L-ascorbic acid (Sigma-Aldrich). Preparation of High-Density Micromass Cultures. To generate a high-density micromass culture, a volume of 200 μL containing 6 × 105 freshly isolated chondrocytes in maintenance medium was transferred to a well of uncoated flat bottom 96-well plates (Becton Dickinson). Subsequently, culture plates were incubated for 24 h (37 °C, 5% CO2) to ensure sedimentation of the cells. Medium was changed every day. In order to induce OA-like changes in micromass cultures, after 14 days of micromass establishment 10 ng/mL recombinant porcine TNF-α (R&D Systems, Wiesbaden-Nordenstadt, Germany) was added to the maintenance medium for up to 14 days. Histological and Immunohistochemical Analyses. To document ECM formation or loss, histological and immunohistochemical stainings were performed on 8 μm cryosections obtained from micromasses embedded in optimal cutting temperature (OCT) compound (Sakura Finetek, Alphen aan den Rijn, The Netherlands). Cartilage-typical sulfated glycosaminoglycans (GAG) were stained with 0.7% Safranin O in 66% ethanolic solution, and cell nuclei were counterstained with 0.2% Fast Green FCF (Sigma-Aldrich) in 0.3% acetic acid. Additionally, cartilage-specific type II collagen was detected immunohistochemically with polyclonal rabbit anti-porcine type II collagen antibodies (Acris Antibodies, Herford, Germany). Rabbit IgG (DAKO, Hamburg, Germany) served as control. EnVision horseradish peroxidase (HRP) rabbit Kit (DAKO) was used for 2093
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
(each group n = 3 donors). Gene expression profiling was performed with a Porcine Genome Array (Affymetrix, Freiburg, Germany) according to the manufacturer’s recommendations. Raw gene expression data were normalized and analyzed with GeneChip Operating Software 1.4 (GCOS, Affymetrix). Comparative analysis of all three groups was performed on signal intensities, detection and change calls, and signal log ratios (SLR) all provided by GCOS software. Comparison between native cartilage and nonstimulated micromass was performed on the basis of a limited set of cartilage markers complemented by published data from human normal donor and OA cartilage tissue as well as from human micromass culture published previously.23,24 For analysis of gene expression differences between TNF-α stimulated and nonstimulated micromasses genes were considered that (i) revealed for each possible comparison between samples of both groups (n = 3 control vs n = 3 TNF-α) a significant signal change detected by GCOS, (ii) showed a mean fold change (FC) ≥4 or ≤−4, and (iii) showed a p-value less than 0.05 (p < 0.05) applying Student’s t test. Functional classification of all selected genes was manually performed with annotations from the Gene Ontology Annotation Database25 and Entrez Gene database26 based on the gene symbols and was further annotated according to OA literature reports (Supplementary Table 3 in the Supporting Information). To conduct comparison between human and porcine Affymetrix data, the complex comparison sheet (porcine vs human genome U133 Plus Array, Affymetrix) was used to find cross-species relationships between probe set IDs and their conjugated gene information. Real-Time PCR Analysis. RNA from native porcine cartilage micromasses used for microarray gene expression analysis (n = 3) and from micromasses of additional donors (n = 3) was reverse transcribed using the iScript cDNA synthesis kit (BioRad, München, Germany). Real-time PCR was performed in triplicates in 96-well plates (Becton Dickinson) on a Mastercycler ep gradient realplex (Eppendorf, Hamburg, Germany) using expression assays for TaqMan probes and primer sets (order no. in parentheses): collagen type II alpha 1 (COL2A1, Ss03373344_g1), collagen type I alpha 1 (COL1A1, Ss003373341_g1), collagen type X alpha 1 (COL10A1, Ss033911766_m1), aggrecan (ACAN, SS03373387_S1), matrix metallopeptidase 2 (MMP2, SS03394318_m1), matrix metallopeptidase 13 (MMP13, SS03373279_m1), ADAM metallopeptidase domain 10 (ADAM10, SS03373280_m1), chemokine (C-C motif) ligand 2 (CCL2, SS03394377_m1), and interleukin 8 (IL8, SS03392435). In order to normalize the samples, the expression of GAPDH (glyceraldehyde-3-phopshate dehydrogenase, SS03375435_u1) was used. Marker gene expression is given as a percentage related to GAPDH expression or as fold change compared to control samples applying the efficiency corrected ΔΔ-Ct method.27 Statistical Analysis. The significance level of log2tranformed data was determined with the independent two sample t test statistics of the Excel 2007 software package (Microsoft, Redmond, WA, USA). Normality distribution was checked applying the Anderson−Darling test,28 and equal variance of compared sample groups was tested applying the f test.29 In all groups signals were normally distributed. If the equal variance test was not passed, Welch’s t test was applied.30 The dependence (correlation) was determined by calculating the Pearson correlation coefficient (r) with Excel 2007 giving a value between +1 and −1 inclusive, where 1 is total positive correlation, 0 is no correlation, and −1 is total negative correlation. The distribution
antibody detection, and nuclei were counterstained with hematoxylin (DAKO). Stainings were photodocumented using a light microscope (CX 41, Olympus, Hamburg, Germany) with a ColorView III camera and AnalySIS docu 5.1 software (both Olympus Soft Imaging Solutions, Münster, Germany). A histomorphometric analysis was performed to quantitatively determine the intensity of the Safranin O stain. For each image the proportion of red stained areas (pixel) was determined in redgreen-blue (RGB) color mode using a programming software (Xcode, Apple Inc.). A pixel was counted as red if the 1.8-fold of the red value was greater than the sum of green and blue (R-value × 1.8 > G-value + B-value). Intensity of each counted pixel was considered by correcting the 1.8-fold of the red value by the green and blue values (intensity = R × 1.8 − G − B) giving a scale from 0 to 459. Mean intensity (sum of intensities/area of interest) is calculated from each image, and for inspection of histomorphometric analysis an image was created visualizing intensities in different colors (intense, reddish; moderate, orange to yellow; weak, green; not considered as red, blue). Biochemical Analysis. For the assessment of total GAG and collagen content, two micromasses were digested with 1 mL of papain solution (Sigma-Aldrich, 0.125 mg/mL in 0.1 M disodium hydrogen phosphate, 0.01 M EDTA, and 0.01 M cysteine hydrochloride at 60 °C for 6 h). GAG quantity was measured spectrophotometrically using dimethylmethylene blue with chondroitin sulfate (Sigma-Aldrich) as a standard at 525 nm.21 The hydroxyproline assay allows the direct measurement of collagen content.22 Therefore, papain-digested samples were hydrolyzed with hydrochloric acid (3 M, 24 h at 110 °C), oxidized with chloramine T reagent (0.056 M, 20 min at RT), and detected with Ehrlich’s reagent (1 M p-dimethylaminobenzaldehyde in isopropanol/perchloric acid (2:1 v/v), 15 min, 60 °C). Collagen quantity was measured spectrophotometrically at 595 nm with hydroxyproline (Sigma-Aldrich) as a standard. Collagen and GAG content were normalized to DNA amounts, measured spectrofluorometrically using bisbenzimide (Life Technologies, Darmstadt, Germany; 0.7 μg/mL in 2 M sodium chloride, 0.05 M disodium hydrogen phosphate) and calf thymus DNA (Life Technologies) as a standard. GAG and collagen contents were reported as μg/μg DNA. RNA Preparation. For total RNA from native cartilage or micromass cultures, 300 mg of freshly prepared porcine cartilage biopsies or three micromasses were snap-frozen and stored at −80 °C. Frozen samples were transferred to TriReagent (Sigma-Aldrich) and mechanically homogenized. Subsequently, 1-bromo-3-chloro-propane (Sigma-Aldrich) was admixed followed by centrifugation for 45 min at 13000g. The aqueous phase was collected and nucleic acids were precipitated by the addition of an equal volume of ice-cold isopropanol. After 30 min of incubation, precipitated nucleic acids were collected and resolved in RNA isolation buffer (RLT, Qiagen, Hilden, Germany). Further purification was performed according to a protocol for animal tissues of the RNeasy Mini Kit (Qiagen). Integrity and purity of RNA were analyzed using Agilent Bioanalyzer 2100 (Agilent, Palo Alto, USA) and a NanoDrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA) (Supplementary Table 4 in the Supporting Information). Microarray Analysis. Gene expression profiling was performed on three groups: (1) native cartilage, (2) micromasses cultured for 14 days, and (3) micromasses cultured for 14 days and stimulated for 14 days with 10 ng/mL TNF-α 2094
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
Figure 1. Porcine chondrocytes in micromass culture develop a compact hydrogel-like structure rich in cartilage-typical proteoglycans and collagens as well as a cartilage-specific gene expression pattern, which do not reach the high levels observed in native cartilage. (A) Cells were seeded on the bottom of 96-well plates and (B) formed a tissue-like structure clearly visible after 14 days. (C) Safranin O staining documented the increase of thickness and proteoglycan content of micromass cultures from day 2 to day 28 (i−vii) and the typical state of native porcine cartilage (viii). The left third of the image shows the visualization of the histomorphometric analysis that was performed with these stainings. (D) Histomorphometric analysis of Safranin O stainings of porcine native cartilage (nat) and micromass culture from day 2 to 28 considering the proportion and intensity of the stained area as mean intensity. An increase of intensity was observed up to day 14. (E, F) Immunohistochemistry demonstrated the presence of cartilage-specific collagen type II as a component of the formed matrix in micromass culture after 14 and 28 days. (G) Biochemical analysis of glycosaminoglycan (GAG) content using the DMMB method showed a slight increase from day 14 to day 28. (H) Total collagen content determined with hydroxyproline (HP) assay revealed a stable content even after 28 days. (I−K) Quantitative gene expression analysis of cartilage-specific collagen type II (COL2A1) and aggrecan (ACAN) detected constant levels up to day 28. Collagen type I (COL1A1) was found at lower levels at day 14 compared to day 28. Bar = 200 μm, * p < 0.05.
of the sample correlation coefficient (significance) was tested using Student’s t distribution.31 A p-value lower than 0.05 was considered as significant.
Figure 1C-viii, 1D). No further considerable increase could be detected in prolonged culture. Immunostainings of day 14 and 28 micromass cultures demonstrated the clear presence of cartilagespecific collagen type II in the ECM (Figure 1E,F). Biochemical analysis of glycosaminoglycans (GAG) revealed a content between 8 and 12 μg/μg DNA after 14 and 28 days of culture, which is not more than 33% of the content observed in native porcine cartilage (37 μg/μg DNA, Figure 1G), but a typical value of tissue engineered cartilage.32 Also total collagen content of day 14 (1.8 μg/μg DNA) and day 28 cultures (1.9 μg/μg DNA) was found more than 6-fold decreased when compared with porcine cartilage (12.8 μg/μg DNA, Figure 1H). Gene expression of the cartilage markers collagen type II (COL2A1, Figure 1I) and aggrecan (Figure 1J) was detected at constant levels up to day 28. In comparison to native cartilage these levels are significantly decreased, which is typical for in vitro cultures without additional growth factors.24 Collagen type I that is related to dedifferentiation was found at low levels in day 14 cultures (1.3% of GAPDH expression) and higher levels at day 28 (4%). As expected,
■
RESULTS Cartilage Qualities of Chondrocyte Micromass Culture. Freshly isolated (primary) porcine chondrocytes were seeded as a suspension in 96-multiwell plates. After 1−2 days, cells sedimented and grew in high-density multilayer micromass culture, and within 14 days, a compact hydrogel-like structure formed (Figure 1A) showing the circular geometry of the culture well with a thickness up to 1 mm (Figure 1B). During culture, chondrocyte micromasses developed an increasing amount of ECM rich in proteoglycan histologically detected by Safranin O stainings (Figure 1C). Histomorphometric analysis verified the continuous increase of staining intensities up to day 14 (mean intensity 201), which is in total 60% of the intensity detected in native porcine cartilage (mean intensity 333; 2095
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
expression in native cartilage was close to the detection limit (0.01%, Figure 1K). The results show that porcine micromass cultures exhibited substantial characteristics known from human tissue-engineered and native cartilage. Already after 14 days most properties of micromass cultures were established that did not change significantly while prolonging the culture. For deeper analysis comparative gene expression profiling with published data from human native cartilage and human chondrocyte micromass culture was performed. Comparative Gene Expression Profiling of Cartilage Markers. To identify the gene expression pattern of chondrocytes in micromass cultures, microarray analyses with porcine genome wide GeneChips from Affymetrix were performed. The expression profiles of porcine micromass cultures at day 14 were compared with those of native porcine cartilage. Furthermore published data from human native cartilage and human chondrocyte micromass culture were integrated into the analysis (for all probe set signals see Supplementary Table 1 in the Supporting Information).23,24 Since probe sets of human and porcine microarrays differ, comparisons were performed on the basis of the gene symbol diminishing the number of comparable data sets and overlapping genes (Figure 2A). Global analysis of expression signals revealed a high correlation (r = 0.90) between porcine native cartilage and micromass culture at day 14 (Figure 2B). Out of 10,409 genes comparable between porcine arrays, 9,504 genes (91%) were found “detected” both in cartilage and in micromass culture (71%), or “absent” in all replicates tested (20%, Table 1). Human micromass cultures demonstrated a similar expression in about 61% of all comparable genes (Table 1). The correlation coefficient of detected genes was found at 0.75 (Figure 2C). Human native cartilage expression resembled the porcine micromass in 53% of all considered genes whereof the expression values of detected genes still correlated with r = 0.71. Apart from these global statistics the expression of known cartilage- and bone-related genes was considered on the basis of a review of Martel-Pelletier et al.33 The selection comprised mainly ECM-building genes, growth factors and receptor, enzymes, and transcription factors (Table 2). The majority (31 from 41) of this cartilage-describing selection showed a comparable expression both in human and porcine micromass cultures and in native cartilage including collagens (type II α1 (COL2A1), type IX α1 (COL9A1)), proteoglycans (aggrecan (ACAN), biglycan (BGN), chondroitin sulfate proteoglycan 4 (CSPG4), decorin (DCN), heparan sulfate proteoglycan 2 (HSPG2), versican (VCN)), ECM connectors (C-endopeptidase enhancer 2 (PCOLCE2), carbohydrate-binding protein 35 (LGALS3), f ibromodulin (FMOD), hyaluronan and proteoglycan link protein 1 (HAPLN1), prolargin (PRELP)) and other players of structural integrity of cartilage (cartilage oligomeric protein (COMP), cartilage intermediate layer protein (CILP), matrix gla protein (MGP), Chitinase 3-Like 1 (CHI3L1), f ibronectin 1 (FN1), f ibrillin 1 (FBN1)). Apart from ECM components transcription factors (SRY (Sex Determining Region Y)-Box 6 and 9 (SOX6, SOX9), osterix (SP7)), enzymes (matrix metallopeptidase 2 (MMP2), serpin peptidase inhibitor, clade A (SERPINA1)), growth factors (f ibroblast growth factors 2 (FGF2), transforming growth factor beta 1 and 2 (TGFB1, TGFB2)), and receptors (f ibroblast growth factor receptor 1−3 (FGFR1−3)) were found expressed in all compared groups. Interestingly, the suspected cartilage-related gene33 FGF9 was detected in neither cartilage nor micromass culture. FGF18 was detected only in human cartilage, but not in porcine cartilage or any micromass. Among genes that were not detected or only
Figure 2. Comparative microarray analysis of porcine and human micromass culture and cartilage demonstrated a high correlation of the expression patterns of detected genes. (A) In addition to the expression data obtained in this study from porcine micromass culture and cartilage with Affymetrix porcine GeneChips, published data from human micromass culture available in Affymetrix human genome (HG)-U133A GeneChip format24 and human cartilage available in Affymetrix HGU133+ GenChip format23 was included in the analysis to examine the suitability of porcine micromass culture as a model for human tissue. The different structure of the detecting probe sets (white number) in porcine and human micorarrays allowed only a comparison based on the gene symbol (black number). Since all chip formats partly cover different genes, an overlap of 5,406 genes was found between porcine and HG-U133A microarray, and 6,420 genes were overlapping in porcine and HG-U133+ arrays. (B−D) To assess the similarities of the gene patterns a correlation analysis of the signal value of genes detected in both groups was performed. For more details see Table 1. (B) The mean expression value of 7,447 genes of porcine and native cartilage correlated with a coefficient r = 0.90. (C) Expression values of human and porcine micromass cultures were found lower but clearly correlating at r = 0.75 considering 2,393 genes. (D) Human cartilage expression values still correlated at r = 0.71. Gray data dots represent all genes included in the correlation analysis, and black dots are genes related to bone and cartilage described in Results and presented in Table 2. * Significance of correlation, p < 0.05. 2096
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
Table 1. Comparison of Expression Pattern between Porcine Micromass Culture with Porcine Cartilage, Human Micromass Culture, and Human Cartilage Based on the Detection of Corresponding Genesa expression pattern
equal varying
indefinite
detection of genes overlap with porcine MM present both absent both absent pMM, present other present pMM, absent other heterogeneous
porcine cartilage 10409
human MMb 5406
Table 2. Overview of Gene Expression Values of Selected Cartilage and Bone Markers in Porcine and Human Articular Cartilage (AC) and Micromass (MM) Culturea human
human cartilagec
gene symbol
6420
7447 (71%) 2393 (44%) 2716 (42%) 2057 (20%) 926 (17%) 692 (11%) 85 (0.1%) 572 (11%) 247 (4%) 34 (0.1.%)
634 (12%)
636 (10%)
786 (7%)
881 (16%)
2129 (33%)
a
MM = micromass, complete gene list available in Supplementary Table 2 in the Supporting Information. bDehne et al.24 cKarlsson et al.23
partly detected, genes related to bone such as alkaline phosphatase (ALPL) and osterix (SP7) were found. The genes coding for the bone ECM components collagen type X α1 (COL10A1) and integrin-binding sialoprotein (IBSP) were detected at low levels only in human native cartilage specimen, but not in porcine cartilage or micromass culture. COL1A1 and osteocalcin (BGLAP) were detected only in human micromass culture. A species-specific difference was observed for the genes insulin growth factor 1 (IGF1), transforming growth factor 3 (TGFB3), and thyroid hormone receptor (THRA), which were detected in porcine specimens but not in human, and for SERPINA3, which was detected only in humans. Bone morphogenetic protein 2 (BMP2) was not detectable in porcine micromass cultures and vice versa BMP4 and BMP7 were not detected in human chondrocytic cells. Insulin growth factor binding protein 3 (IGFBP3) was the only cartilage-related marker that was not detected in cartilage, but in micromass culture independent of human or porcine origin. These comprehensive considerations identified extensive similarities of expression with minor differences that were mainly limited to bone related genes. ECM Alterations in Micromass Culture As Response to TNF-α Treatment. In order to examine the ability of the porcine micromass model to feature osteoarthritis, 10 ng/mL TNF-α was added to the culture. Already after 3 days of treatment, a decline of Safranin O staining intensity was observed culminating in a weak appearance of proteoglycans in the micromass model after 14 days of induction (Figure 3A). Histomorphometric analysis specified a decline of more than 76% of the mean intensity compared to the starting condition (day 0, mean intensity 201) or respective control (day 14, mean intensity 184) after 14 days of induction (mean intensity 44, Figure 3B). Collagen type II immunohistochemistry demonstrated an intense staining (Figure 3C) comparable with detections made at day 0 and day 14 controls (Figure 1E,F). Biochemical quantitative GAG analysis showed a significant loss of 69% or 81% in TNF-α stimulated micromass cultures compared to controls at day 0 or day 14, respectively, whereas total collagen decreased only about 50% at the same time (Figure 3D). Summarized, TNF-α treatment led to a significant reduction of proteoglycan in ECM of micromass culture. Total collagen loss was not that pronounced and collagen type II was still detectable even after 14 days of treatment. Gene Expression Profiling of Significantly Dysregulated Genes. To identify the patterns of osteoarthritis in TNF-α
AC
b
ACAN BGN BMP2 BMP4 BMP7 CHI3L1 CILP COL1A2 COL2A1 COL9A2 COMP CSPG4 DCN ECM1 FBN1 FGF18 FGF2 FGF9 FGFR1 FGFR2 FGFR3 FMOD FN1 HAPLN1 HSPG2 IGF1 IGFBP3 LGALS3 MGP MMP3 PCOLCE2 PRELP SERPINA1 SERPINA3 SOX6 SOX9 TGFB1 TGFB2 TGFB3 THRA VCAN
13918 5689 526 27 27 15684 16893 1536 5022 1217 17344 982 19282 209 763 277 650 29 790 2562 898 12093 16290 7161 538 37 181 7441 16879 12424 11043 16628 15589 17825 255 3959 640 302 82 163 270
ALPL BGLAP COL10A1 COL1A1 IBSP SP7 SPP1
28 89 436 29 2034 8 670
porcine MM
c
AC
Cartilage 4115 6312 2156 5116 1322 38 6 69 8 44 6237 472 331 6493 4481 206 1873 1726 361 4646 3619 7098 189 709 5900 7749 330 203 933 271 3 5 346 482 4 8 655 983 734 493 234 2437 2573 9582 5521 9470 2037 5197 152 441 4 138 5579 10 3085 964 6460 7491 5188 41 2150 2054 3407 4557 254 2481 6141 2 ne 173 838 390 327 194 58 166 90 357 50 125 482 141 Bone 3 73 51 28 30 149 385 11 33 490 ne 3 801 7682
FC porcine
MM
AC vs MM
7899 3781 19 69 12 3104 100 186 733 3926 1524 1195 4207 45 256 5 268 7 800 644 2221 8874 6598 5463 395 1162 80 848 8850 63 2621 3113 145 1 229 93 34 340 382 237 2831
−1.3 1.3 2.3 1.0 3.3 −6.5 74.1 1.2 2.4 1.1 4.6 −1.3 1.8 3.3 −1.0 1.1 1.8 1.3 1.2 −1.5 1.1 1.1 1.5 −1.1 1.1 −8.6 −5.4 1.1 −1.2 −1.8 −1.3 1.5 15.4 1.0 −1.4 3.6 5.8 −1.9 −1.0 −1.8 −24.4
10 10 14 12 14 6 2811
6.1 1.9 4.5 1.0 14.6 −3.9 2.8
a
Bold = detected in all biological replicates, underlined = not detected in all biological replicates, ne = gene not existent on microarray type, FC = fold change. bKarlsson et al.23 cDehne et al.24
stimulated micromass cultures, a comparative expression analysis of nonstimulated and stimulated porcine micromass cultures was performed using oligonucleotide arrays. 645 probe sets were 2097
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
Figure 3. TNF-α induced micromass cultures demonstrated a decrease of ECM content. (A) Safranin O staining documented the decrease of thickness and content of micromass cultures by TNF-α treatment starting from established cultures (i) up to 14 days induction (ii−v). The left part of each picture shows the visualization of the histomorphometric analysis that was performed with these stainings. (B) Histomorphometric analysis of Safranin O stainings of porcine TNF-α induced micromass cultures starting from established micromass cultures (day 0) up to 14 days of treatment considering the proportion and intensity of the stained area as mean intensity. The mean intensity decreased down to 25% of the corresponding control cultures and starting condition, respectively. (C) Immunohistochemistry demonstrated the presence of cartilagespecific collagen type II as a component of the remaining ECM after 14 days of induction. (D) Biochemical analysis showed a significant decrease of GAG content (−70%) after 14 days TNF-α addition and moderate decline (−50%) of total collagen content compared to day 14 or 28. Bar = 200 μm, * p < 0.05.
Figure 4. Distribution of all relevant and significantly dysregulated genes from microarray analysis in functional clusters comparing TNF-α induced micromass and control cultures. Bright bars: Percentage of genes with lower expression in TNF-α induced micromass cultures (decreased). Dark bars: Percentage of genes with higher expression in micromass cultures induced with TNF-α (increased).
decreased in OA and in TNF-α stimulated micromass culture, respectively. Out of the 349 genes significantly changed 92 genes were found already described in the context of OA shown in Table 3. Consistent expression alterations achieved through TNF-α treatment were predominantly found in the groups “cytokines”, “cell death”, “enzymes”, “transcription”, “skeletal development”, and “growth factors”, whereas genes involved in the formation of ECM components deviated from known OA reaction pattern (Table 3). The remaining groups “inflammation”, “signaling”, and “proliferation or differentiation” do not clearly match with the expression pattern observed in human OA or groups were too small to derive a trend. To confirm the findings of global expression analysis PCR analysis of genes involved in the best matching and most deviating groups was performed (Figure 5). Consistently, the expression of cartilage-specific ECM components COL2A1 (183-fold) and aggrecan (15-fold) was found significantly decreased in TNF-α stimulated micromass culture compared to corresponding control cultures at days 14 and 28. Furthermore, the ECM degrading matrix metallopeptidases 2 (MMP2, 4-fold), -13 (MMP13, 64-fold) and a disintegrin and
found significantly changed with a more than 4-fold alteration coding for 349 known genes (Supplementary Table 3 in the Supporting Information). Genes were functionally annotated on the basis of the Gene Ontology database25 and Entrez Gene database,26 leading to 10 main functional groups (181 genes) and miscellaneous genes not assigned to any main group (168 genes, Figure 4). A clear trend was found for the genes related to cytokines and cell death, which showed an increase of more than 90% of the genes of this group. Furthermore, a majority of the genes associated with inflammation (73%), enzymes (69%), growth factors (67%), and proliferation/differentiation (62%) were found increased in TNF-α stimulated micromass culture. In contrast, genes of the groups signaling (36%), transcription (26%), and ECM components (19%) were mainly decreased (Figure 4). These global expression statistics confirm the ECM decline observed on histological and biochemical levels and additionally indicate mechanisms associated with cell death and inflammation including cytokines. A more detailed view was set on genes that had already been described in literature in the context of OA, and furthermore it was also examined whether these genes were increased or 2098
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
Table 3. Genes Differentially Regulated (>4-fold) in TNF-α Induced Micromass Cultures Compared to Noninduced Cultures and Selected According to the Association to Osteoarthritisa gene symbol
FCb
ref
Inflammation PTGS2 TAC1 VCAM1 PTX3 LY96 VLDLR CEBPD CEBPB
14.0 −8.2 7.9 −6.9 5.0 4.8 4.4 4.1
47,46 23 23,48 23 23 23 23
Cytokines CCL2 IL8 CXCL12 CXCL2 CXCL14
144.8 36.8 35.9 11.9 6.0
23,40 23,41 42,43 23,45 23
Cell Death TNFSF10 PMAIP1 AHR HMOX1 ADM
9.2 7.5 6.9 5.2 4.2
23,49 23 23,50 23,51 23
Proliferation/Differentiation PTN −7.6 ID4 −7.1 NUAK1 6.1 Skeletal Development GPNMB 28.7 FRZB −25.0 CLEC3B −15.5 RUNX1 6.0 ITGB8 6.0 CDH11 4.9 HOXA3 −4.6 MAMDC2 4.3
23
23,54 23 23
23 23,52 23,55 23,53 23 23 23 23
gene symbol
FC
Transcription WIF1 DLX5 TWIST1 MAFB ELL2
−26.0 −6.4 6.1 5.4 −4.4
ref
gene symbol
23 23 23 23 23
ECM Components COL9A2 COL2A1 CILP THBS3 HAPLN1 COMP LUM NID2
−19.7 −18.1 17.5 −16.9 −8.1 −7.8 5.8 −5.5
55 51,68 55,69 23 55,60 55,69 55,70 23
Signaling RGS5 GPR125 MARCKS
−24.1 −4.7 4.1
23 23 23
Growth Factors IGFBP3 IGFBP6 ZMAT3 TGFBI
15.0 −7.0 5.2 −5.0
23,54 23 23 59,71
Enzymes MMP1 CTSC ANPEP CTSS PDK4 HTRA1 PLK2 MMP3 FAM108C1
22.5 15.6 15.4 7.7 −7.6 5.2 5.0 4.8 4.4
23,56 23,56 23,56 23 23 23,56 23 56,55 23
FC
ref
Enzymes (continued) ADAM10 4.0 MMP13 4.0 MMP2 4.0
23 23 23
Miscellaneous C1orf54 SOD2 RCAN2 C6orf115 GGTA1 MYLK SNX10 MAP3K8 THY1 S100A4 SAMD9 B4GALT5 DUSP6 BASP1 ALDH1L1 APOD SLC25A37 PHLDB2 GLIPR1 GUCY1B3 NCAM1 UGCG MSN WWP2 NAP1L2 LIFR SM22A RAI14 MARCH3 PDLIM1 SLA1
23 23 23 23 23 23 23 23 23 23,71 23 23 23 23 23 23,71 23 23 23 23 23 23 23 23 23 23 23 23 23 23 23
36.8 19.7 14.8 13.6 11.2 10.8 10.7 10.6 10.1 10.0 9.9 7.4 7.3 6.4 6.3 −6.3 6.0 5.9 5.4 5.4 −5.4 5.3 5.2 −4.7 −4.5 4.5 4.2 4.2 4.2 4.1 4.0
a
For gene titles and signal values refer to Supplementary Table 3 in the Supporting Information; bold = consistence of increase or decrease with OA literature, underlined = controversial in literature ref = reference. bFC = fold change.
metalloprotease domain 10 (ADAM10, 7-fold) were detected highly induced as well as genes coding for the chemokines monocyte chemotactic protein 1 (CCL2, 245-fold) and interleukin 8 (IL8, 7-fold). In addition to significant findings from microarray analysis the expression of collagen type 1 and 10 (COL1A1, COL10A1), both increased in OA-associated hypertrophic bone formation, was examined. Consistent with comparative expression analysis, COL1A1 was detected not significantly altered. On the contrary, COL10A1, not detected in any condition in microarray analysis, was found significantly decreased. The OA pattern-focused analysis of TNF-α induced porcine micromasses demonstrated marked parallels in terms of cell death, formation of an inflammatory environment through secretion of cytokines and enzymes, and modulation of genes involved in skeletal development such as growth factors and transcription factors. Contrarily, the OA-typical induction of ECM-forming genes as response to ECM loss was not observed
as well as the consistent regulation of some players of the inflammatory machinery.
■
DISCUSSION
In vitro models offer cost-effective screenings for the development of novel therapeutic strategies. Especially, the chondrocyte micromass culture model built of primary cells is an easy to manage 3D model, which has already been shown to mimic essential aspects of human chondrocyte biology, pathophysiology, and differentiation.24 The downside of the model is the extensive need of primary chondrocytes impeding highthroughput approaches including human cells. The use of cells from animal tissue sources such as porcine byproducts like cartilage used in this study can overcome these limitations, but the suitability of such porcine cell-based in vitro cultures to model human cartilage and its pathophysiology in OA was not investigated comprehensively so far. 2099
dx.doi.org/10.1021/mp5000554 | Mol. Pharmaceutics 2014, 11, 2092−2105
Molecular Pharmaceutics
Article
the applicability of known human cartilage markers and characteristics in the porcine micromass model. In the second stage of this study, the ability of porcine micromass cultures to model aspects of osteoarthritis was investigated, since naturally or spontaneously developing OA in vivo is only known from mini guinea pigs but not from domestic pigs.37 To induce such changes we used the inflammatory mediator TNF-α. Inflammation is occurring globally in synovium during OA and affects chondrocytes and thus cartilage homeostasis. Among factors such as interleukin (IL) 1β, IL-12, and IL-15 and various associated chemokines, TNF-α is highly upregulated in OA and known to mediate the catabolic process.20 In this study, the daily addition of TNF-α led to a continuous depletion of proteoglycans in porcine micromass cultures documented by histomorphometric analysis of Safranin O stainings and biochemical detection of GAG. Within 14 days of TNF-α treatment GAG mass was diminished by about 70%. GAG depletion plays a central role in the histopathologic assessment of OA grade and stage38 and is a detectable feature in this model. Total collagen loss was not that pronounced and reproducible as observed in GAG measurements also seen in the moderate collagen type II immunostaining. Genome-wide gene expression profiling of TNF-α treated cultures identified groups of genes with expression patterns predominantly consistent with or deviating from human OA expression patterns resulting in the consideration of 92 genes (Table 3). Human OA data was obtained from a study of Karlsson et al., who analyzed the expression pattern of human articular cartilage biopsies collected from donors with OA23 serving in combination with additional literature reports (see Table 3) as reference for determining OA relevance and consistency of expression. TNF-α addition to micromass cultures led to the induction of several genes coding for cytokines such as the chemokines monocyte chemotactic protein 1 (MCP1/CCL2), interleukin 8 (IL8/CXCL8), stromal cell-derived factor 1 (SDF1/CXCL12), macrophage inf lammatory protein 2 (MIP2/CXCL2), and bolekine (CXCL14) all involved in immunoregulatory and inflammatory processes. Karlsson et al. have shown elevated levels of all these chemokine genes in human cartilage from patients with endstage OA.23 Pulsatelli et al. already described the upregulation of chemokines in chondrocytes such as MCP1 and IL8 in response to TNF-α treatment,39 and Eisinger et al. proposed the role of MCP1 in innate immune system-associated joint inflammation.40 Kaneko et al. detected higher IL8 levels in serum and synovial fluid of patients with OA compared to normal patients.41 Kanbe et al. observed 4-fold higher levels of CXCL12 in synovial fluid, and Xu et al. correlated synovial fluid levels with radiographic severity in knee osteoarthritis and suggested CXCL12 as an alternative biomarker for OA progress.42,43 Apart from human OA cartilage CXCL2 induction was also shown in chondrocytes exposed to IL-1β, a highly upregulated cytokine during OA,44 and Scaife et al. demonstrated that CXCL12 is selectively overexpressed in OA fibroblasts rather than in rheumatoid fibroblasts.45 Besides cytokines, prostaglandin-endoperoxide synthase 2 (PTGS2), vascular cell adhesion molecule 1 (VCAM1), and lymphocyte antigen 96 (LY96) all related to inflammation displayed an OA-like expression response. Lopez-Armada et al. demonstrated the induction of PTGS2 expression in chondrocytes by addition of TNF-α, and Amin et al. noticed a 50-fold higher release of PTGS2 in OA cartilage specimen compared to normal cartilage.46,47 Grogan et al. observed an increased expression of the stem cell marker
Figure 5. Gene expression analysis by real-time PCR of selected genes obtained from microarray analysis. Genes of different functional categories obtained from gene expression profiling of TNF-α induced micromass and corresponding control cultures were verified. Except COL1A1 all genes were detected significantly differentially expressed. The expression of genes related to cartilage (COL2A1, ACAN) and bone (COL10A1) was decreased, and genes related to ECM degradation (MMP2, MMP13, ADAM10) and inflammation (CCL2, IL8) were found induced correlating with results obtained from microarray analysis. The expression of TNF-α induced micromass cultures is given as ratio to control cultures at day 14 and day 28. * p < 0.05.
In the first stage of this study the ability of the porcine chondrocyte micromass model to establish a cartilage-like tissue was investigated. The time-consuming formation of ECM including proteoglycans and collagens allowscontrary to single cell or monolayer culturethe maintenance of chondrocyte phenotype34 as well as the interaction with ECM components and its degradation products, which can in addition to OA mediators further drive the OA progress.35 In this study, micromass cultures were maintained for at least 14 days to establish a cartilage-like ECM documented on histological (collagen type II and Safranin O staining), biochemical (GAG and collagen content), and gene expression level (COL2A1, aggrecan, COL1A1). The high levels of GAG and cartilage marker expression observed in native cartilage were not achieved, which is a typical feature of in vitro culture not utilizing additional growth or differentiation factors such as bone morphogenetic proteins (BMPs) or transforming growth factors (TGFs).24 Comparative genome-wide expression analysis of porcine micromass cultures revealed high correlations with data from porcine cartilage (r = 0.90), human micromass culture (r = 0.75), and human cartilage (r = 0.71) considering all detected genes. Detailed analysis of 48 cartilage and bone markers demonstrated considerable consistency (37 genes) of porcine micromass culture with porcine and human cartilage including important cartilage markers such as COL2A1, COL9A1, aggrecan, COMP, HAPLN1, and many more comprehensively reviewed by MartelPelletier et al.33 Among genes with strongly deviating expression pattern, COL10A1 and IBSP both associated with hypertrophy and calcification were found induced in native human cartilage and were not detected in porcine micromass culture or cartilage. Since the human reference did not show any signs of OA (Mankin score
