Drs. Olivotto, Borzí, and Marcu contributed equally to this work.
Differential requirements for IKKα and IKKβ in the differentiation of primary human osteoarthritic chondrocytes
Article first published online: 28 DEC 2007
Copyright © 2008 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 58, Issue 1, pages 227–239, January 2008
How to Cite
Olivotto, E., Borzi, R. M., Vitellozzi, R., Pagani, S., Facchini, A., Battistelli, M., Penzo, M., Li, X., Flamigni, F., Li, J., Falcieri, E., Facchini, A. and Marcu, K. B. (2008), Differential requirements for IKKα and IKKβ in the differentiation of primary human osteoarthritic chondrocytes. Arthritis & Rheumatism, 58: 227–239. doi: 10.1002/art.23211
- Issue published online: 28 DEC 2007
- Article first published online: 28 DEC 2007
- Manuscript Accepted: 7 SEP 2007
- Manuscript Received: 7 JUN 2007
- Rizzoli Institute
- Carisbo Foundation
- European Union FP6 MAIN, Network of Excellence grant
- Carisbo Foundation
- European Union FP6 MAIN, Network of Excellence grant
- MIUR 40% grant
- Progetto Strategico, Università degli Studi di Bologna
- NIH. Grant Number: GM-066882
- European Union FP6 MAIN, Network of Excellence grant
- Senior Scholar of the Institute of Advanced Studies of the University of Bologna
Osteoarthritic (OA) chondrocytes behave in an intrinsically deregulated manner, characterized by chronic loss of healthy cartilage and inappropriate differentiation to a hypertrophic-like state. IKKα and IKKβ are essential kinases that activate NF-κB transcription factors, which in turn regulate cell differentiation and inflammation. This study was undertaken to investigate the differential roles of each IKK in chondrocyte differentiation and hypertrophy.
Expression of IKKα or IKKβ was ablated in primary human chondrocytes by retro-transduction of specific short-hairpin RNAs. Micromass cultures designed to reproduce chondrogenesis with progression to the terminal hypertrophic stage were established, and anabolism and remodeling of the extracellular matrix (ECM) were investigated in the micromasses using biochemical, immunohistochemical, and ultrastructural techniques. Cellular parameters of hypertrophy (i.e., proliferation, viability, and size) were also analyzed.
The processes of ECM remodeling and mineralization, both characteristic of terminally differentiated hypertrophic cells, were defective following the loss of IKKα or IKKβ. Silencing of IKKβ markedly enhanced accumulation of glycosaminoglycan in conjunction with increased SOX9 expression. Ablation of IKKα dramatically enhanced type II collagen deposition independent of SOX9 protein levels but in association with suppressed levels of runt-related transcription factor 2. Moreover, IKKα-deficient cells retained the phenotype of cells in a pre–hypertrophic-like state, as evidenced by the smaller size and faster proliferation of these cells prior to micromass seeding, along with the enhanced viability of their differentiated micromasses.
IKKα and IKKβ exert differential roles in ECM remodeling and endochondral ossification, which are events characteristic of hypertrophic chondrocytes and also complicating factors often found in OA. Because the effects of IKKα were more profound and pleotrophic in nature, our observations suggest that exacerbated IKKα activity may be responsible, at least in part, for the characteristic abnormal phenotypes of OA chondrocytes.
Osteoarthritis (OA), the rheumatic disease with the highest prevalence and economic impact, is a degenerative malady driven by inflammatory factors but lacking the classic profile of inflammation (1). These inflammatory factors activate chondrocytes, the unique cell component in cartilage, which consequently affects their ability to maintain tissue homeostasis. Chondrocytes undergo a number of cell reaction patterns, including hypertrophy and terminal differentiation, similar to what occurs in the terminal hypertrophic zone of the growth plate (2). Endochondral ossification is the final outcome of the chondrogenic differentiation program, which begins with chondroprogenitor differentiation from mesenchymal stem cells and ends in cartilage matrix calcification (3).
RNA and protein expression patterns suggest that normal chondrocytes are kept in a state of maturational arrest (4). In contrast to this phenotype, OA chondrocytes exhibit increased expression of hallmarks of hypertrophy (i.e., type X collagen [Col10], alkaline phosphatase, and matrix metalloproteinase 13 [MMP-13]), indicating that OA-derived chondrocytes have differentiated into a mature phenotype (2, 3). Because OA chondrocytes appear to be inappropriately pushed toward a hypertrophic-like state, endochondral ossification has been proposed as a developmental model to understand the pathogenesis of OA (5). Under appropriate culture conditions, chondrocytes exhibit a level of phenotypic plasticity that is comparable with that of mesenchymal stem cells undergoing chondrogenesis by giving rise to adipose tissue, cartilage, and bone, with the latter function underlying a change in lineage commitment from hypertrophic chondrocytes to osteoblast-like cells (6, 7).
A variety of signaling pathways and transcription factors are known to play specific roles at each stage of chondrogenesis (for review, see ref.3). Tumor necrosis factor α (TNFα)–mediated activation of canonical NF-κB transcription factors was shown to inhibit the differentiation of mesenchymal chondrocytes through posttranscriptional down-modulation of SOX9, a master chondrocytic transcription factor (8). The process of chondrogenesis is orchestrated by interactions between SOX9 and runt-related transcription factor 2 (RUNX-2), the effects of which determine the fate of the differentiated chondrocyte in terms of either continuing to remain within the cartilage or undergoing hypertrophic maturation prior to ossification (3, 9). SOX9 has opposite effects at the early prehypertrophic phase as compared with the terminal hypertrophic phase of chondrogenesis (3, 9). Down-regulation of SOX9 (8) by NF-κBs could reflect the ability of the NF-κBs to promote the progression of chondrogenesis (3, 10). The NF-κBs are known to orchestrate most proinflammatory processes, and thus represent a potential therapeutic target in OA (11). NF-κBs provide functional connections between inflammation-like responses, normal as compared with abnormal cellular growth, and developmental programming (12, 13). In addition, NF-κBs act as homodimers and heterodimers of 5 different subunits that are kept in a cytoplasmic inactive state in complexes with inhibitory IκB proteins (14).
IKKα and IKKβ initiate the release of active NF-κBs from IκBs (12–14). In response to a host of proinflammatory stimuli, IKKβ is the dominant acting IκBα kinase in vivo. Activation of IKKβ is essential for the nuclear translocation of canonical NF-κB heterodimers (including p65/RelA:p50 and c-Rel:p50). In contrast, IKKα acts only occasionally as an IκBα kinase (15), and instead has been reported to activate canonical NF-κB targets in established cells (16) by acting as a nucleosomal kinase (17, 18).
Unlike IKKβ, IKKα plays a unique role in vivo by inducing activation of the noncanonical, or alternate, NF-κB pathway (12). This pathway involves the targeted phosphorylation and processing of the p100 precursor of the NF-κB p52 subunit, which liberates p65/p52 and RelB/p52 heterodimers to activate other NF-κB targets (12, 19). Recently, IKKα was shown to activate a subset of cytoplasmic p50/p65 complexes tethered to p100 (20). Although the IKKα-dependent noncanonical NF-κB pathway is not essential for normal mouse development (21, 22), IKKα has a kinase-independent function that is essential for keratinocyte differentiation and can also act as a serine/threonine kinase outside of the NF-κB pathway (for review, see ref.23). In addition to displaying a lack of epidermal differentiation, IKKα-knockout mice also show selective skeletal abnormalities, which initially were thought to be suggestive of incomplete and/or asymmetric ossification; however, this latter defect was subsequently attributed to the collateral effects of systemically high levels of fibroblast growth factor activity produced by the undifferentiated epidermis (24).
In the present study we investigated the contributions of IKKα and IKKβ in the physiology and differentiation of OA chondrocytes. To this end, expression of IKKα and of IKKβ was individually ablated in primary human chondrocytes by retroviral transduction of specific short-hairpin RNAs (shRNA). We then evaluated the effects of the IKK knockdowns on the ultrastructural organization of major extracellular matrix (ECM) components (proteoglycans and type II collagen [Col2]) and also assessed other factors indicative of chondrocyte hypertrophy (including cell size and expression of Col10) and terminal differentiation (cell viability and calcium deposition) in proliferating chondrocytes and in differentiated micromasses of the chondrocytes.
MATERIALS AND METHODS
Isolation and preparation of primary chondrocytes from patients with OA.
Primary chondrocytes were obtained from 15 patients with OA undergoing knee arthroplasty. Experiments were performed under University of Bologna ethics committee guidelines and with all patient identifiers removed. The chondrocytes were isolated from minced tissue by sequential enzymatic digestion as previously described (25), and then expanded in vitro at high density for up to 1 week prior to retroviral transduction. After validation of the IKK knockdowns, stably transduced chondrocytes were seeded into cultures of differentiating micromasses, as previously described (25). In some cases, the micromasses were embedded in OCT (Tissue-Tek; Sakura, Torrance, CA), snap-frozen in liquid nitrogen, and stored at −80°C.
IKKα and IKKβ shRNA in retroviral vectors.
Knockdowns of IKKα and IKKβ were achieved by transduction of early-passage primary chondrocytes with retroviral vectors containing IKKα- or IKKβ-specific shRNA. IKKα- and IKKβ-specific oligonucleotides for each shRNA (shOligos) were subcloned into the pSuper.retro (Puro) moloney retroviral vector (26) according to the manufacturer's instructions (OligoEngine, Seattle, WA). To avoid potential off-target effects, up to 3 shOligos were designed for IKKα (accession no. AF012890) and IKKβ (accession no. AF080158), with each containing 19–22 nucleotides (nt) complementary to sequences in the different exons. These were IKKα3 (19 mer, starting at nt 1288), IKKα4 (22 mer, starting at nt 1474), IKKβ1 (19 mer, starting at nt 457), IKKβ3 (21 mer, starting at nt 2937), and IKKβ4 (23 mer, starting at nt 2029). The phenotypes of the chondrocytes stably transduced with IKKα- or IKKβ-specific shRNA were compared with that of a negative control (GL2), comprising cells obtained from the same patient and infected by a retroviral vector harboring a firefly luciferase–specific shRNA (27).
Second-passage primary OA chondrocytes were transduced by spinoculation with amphotyped retroviruses prepared from Phoenix A packaging cells (provided by Dr. Gary Nolan at Stanford University). Briefly, viral supernatants were applied to the cells by centrifugation at 1,100g at 32°C for 45 minutes, with continued incubation for 5 hours at 32°C in 5% CO2 followed by replacement with regular growth medium. Seventy-two hours later, shRNA-expressing cells were selected for puromycin resistance (1.5 μg/ml), with 3 changes of medium over 6 days. Canonical NF-κB activity was inhibited by stable transduction with a derivative of a puromycin retroviral vector (BIP) that coexpresses an IκBα superrepressor (IκBαSR) (IBIP) (16, 28).
Toluidine blue and calcein staining.
Quantitative GAG assays.
The GAG content of the chondrocyte micromasses was quantified by dimethylmethylene blue (DMMB) assay. This was carried out essentially as described previously (31).
MMP-13 release assay.
Release of MMP-13 from the micromass cultures under basal conditions or following stimulation with 100 units/ml interleukin-1β (IL-1β) for 72 hours was quantified by an enzyme-linked immunosorbent assay (ELISA). The specific ELISA used detects the MMP-13 proenzyme and its cleaved active form (25).
Transmission electron microscopy (TEM).
Analysis by TEM provided direct visual information on ECM organization, cellular viability, mitochondrial morphology, and the presence of matrix vesicles and areas of electron-dense mineralization (32, 33). Cells were scored as viable when both the nuclear and cytoplasmic membranes appeared integral. Cells were scored as nonviable when either the membranes or nucleus displayed fragmentation, indicative of necrosis or apoptosis, respectively.
Efficiencies of the IKK knockdowns and levels of SOX9 protein were determined by immunoblotting with rabbit anti-human IKKα and IKKβ antibodies (Cell Signaling Technology, Beverly, MA) and affinity-purified rabbit anti-SOX9 antibody (Chemicon, Temecula, CA), respectively. Immunoblotting was performed on 20 μg of total cellular protein, with bands visualized by chemiluminescence. Results were quantified by densitometry with Kodak 1D Image Analysis software (Kodak, New Haven, CT). For a normalization control, Western blots were reprobed with an antitubulin antibody (Sigma, St. Louis, MO).
Immunohistochemical staining for Col2 was performed on 3-week micromass cultures using 5-μm tissue sections fixed with 4% paraformaldehyde. After antigen unmasking (with 1 mg/ml pepsin in Tris HCl, pH 2.0, for 15 minutes at 37°C), sections were incubated with anti-Col2 mouse monoclonal antibodies (2 μg/ml, MAB8887; Chemicon) and signals were developed with a biotin–streptavidin–amplified, alkaline phosphatase–based detection system with fuchsin as substrate. Images were captured and quantified with a Nikon Eclipse 90i microscope equipped with Nikon Imaging Software elements (Nikon, Melville, NY).
ECM neoepitopes, which appear after the catalytic activity of MMPs on both major ECM components (DIPEN on aggrecan and Col2-3/4C on Col2) or of aggrecanases on aggrecan (NITEGE), were revealed by immunohistochemistry (34, 35) after antigen unmasking with 0.02 units/ml chondroitinase ABC for 20 minutes at room temperature. Col2-3/4C neoepitopes were detected with a C1,2C polyclonal rabbit antibody (diluted 1:100; IBEX Pharmaceuticals, Montreal, Quebec, Canada). DIPEN and NITEGE neoepitopes were detected with rabbit anti-sera (kindly provided by Dr. John Mort, Shriners Hospital for Children, Montreal) diluted to 0.33 μg/ml (35). Primary antibody signals were developed with the SuperSensitive Link Label Immunohistochemistry Detection System using Multilink goat antiserum (Biogenex, San Ramon, CA) and fast red as substrate.
Immunohistochemical staining for Col10 was performed on sections of 1-, 2-, and 3-week micromass cultures, after antigen unmasking with 2 mg/ml hyaluronidase for 30 minutes at 37°C. Col10 was detected with anti-Col10 monoclonal antibodies (diluted 1:1,000; IBEX Pharmaceuticals) followed by signal detection with the SuperSensitive MultiLink Label Immunohistochemistry Detection System with fast red as substrate.
Immunohistochemical staining for RUNX-2 was performed with a goat anti-human RUNX-2/CBFA1 polyclonal antibody (0.5 μg/ml, AF2006; R&D Systems, Minneapolis, MN), developed with a Goat Link biotinylated rabbit anti-goat secondary antibody (Biogenex, San Ramon, CA), and detected as described above for ECM neoepitopes and Col10. In all cases, counterstaining with hematoxylin was performed to reveal cell nuclei.
Flow cytometry of chondrocytes.
The sizes of the retrovirally transduced cells were evaluated in samples from 7 patients. Cell size was assessed according to the forward side scatter profiles, determined using a Vantage Flow Cytometer (Becton Dickinson, Mountain View, CA).
Cell proliferation analysis.
Growth rates of the IKKα3-knockdown, IKKβ1-knockdown, and GL2 control chondrocytes were compared by pico green quantification of DNA from the proliferating cells, which were initially seeded at 1,000 cells per well, in quintuplicate, in 96-well plates (28). To correct for differences in cell counts, values were calculated as the percentage increase over the starting (day 0) values.
Cell cycle analysis.
Cell cycle profiles of the IKKα3-knockdown, IKKβ1-knockdown, and GL2 primary chondrocytes were analyzed in cells (at ∼80% confluence) from up to 5 patients. The cell cycle profiles were determined by intracellular 4′,6-diamidino-2-phenylindole staining (5 μg/ml for 30 minutes at 37°C) of 4% paraformaldehyde–fixed cells, carried out essentially as described previously (36). Prior to DNA staining, cells were resuspended with 50 μl RNase ONE buffer (Promega, Madison, WI), heated to 65°C for 10 minutes, cooled on ice, treated with 22 units of RNase for 30 minutes at 37°C, centrifuged, and washed with phosphate buffered saline.
RNA expression analysis.
Total cellular RNA was prepared from 1-week micromass cultures of IKKα3-knockdown, IKKβ1-knockdown, and GL2 primary chondrocytes, as previously described (25). Col2a1 and RUNX-2 messenger RNA (mRNA) were quantified by real-time polymerase chain reaction (PCR) (using SYBR Green), with values normalized to the expression of GAPDH mRNA. Annealing temperatures were 56°C for GAPDH primers (NM_002046, forward 579–598 and reverse 701–683) and RUNX-2 primers (transcript variant 3 [NM_004348], forward 864–883 and reverse 968–949; transcript variant 2 [NM_001015051] and transcript variant 1 [NM_001024630], forward 716–735 and reverse 820–801), and 58°C for Col2a1 primers (transcript variant 1 [NM_001844], forward 4247–4264 and reverse 4499–4485; transcript variant 2 [NM_033150], forward 4040–4057 and reverse 4292–4278).
Nonparametric statistical methods were used, due to the limited size of the primary patient data sets. The mean values of the groups were compared by Wilcoxon's matched pairs test. Data were analyzed using CSS statistical software (StatSoft, Tulsa, OK). All results are expressed as the mean ± SEM.
Differential augmentation of the ECM-generating capacity of differentiated OA chondrocytes between IKKα- and IKKβ-knockdown cultures.
To investigate the individual contributions of IKKβ and IKKα in the differentiation of primary human OA chondrocytes, we ablated their expression by transduction with amphotyped retroviruses expressing shRNA specifically targeting each kinase. To avoid the possibility of off-target effects, up to 3 different shOligos targeted to different exons of IKKα and IKKβ were applied to cells from multiple OA patients; all results were compared with chondrocytes obtained from the same patients that were stably infected with an irrelevant luciferase-specific shRNA retrovirus (GL2) as a negative control. In all cases, >80% of the P0-passage chondrocytes were converted to puromycin resistance within 6 days postinfection, and penetrant knockdowns of either IKKα or IKKβ were produced, as assessed by Western blotting of stably infected cells from different patients (i.e., for IKKα3, ∼92% knockdown in 10 infections; for IKKα4, ∼91% knockdown in 4 infections; for IKKβ1, ∼92% knockdown in 10 infections; for IKKβ3, >75% knockdown in 2 infections; for IKKβ4, ∼90% knockdown in 5 infections) (see representative immunoblots in Figure 1A).
We first assessed the effects of knockdown of either IKKβ or IKKα on the accumulation of modified GAGs in the ECM of 3-week differentiated micromass cultures (32, 37, 38). Toluidine blue staining revealed elevated GAG deposition in the IKKβ- and IKKα-knockdown micromasses as compared with the GL2 controls, with the highest GAG levels accumulating in the absence of IKKβ (i.e., IKKβ > IKKα > GL2) (Figure 1B). Importantly, this qualitative result was confirmed by quantitative DMMB assays (31) of IKKα- or IKKβ-knockdown micromasses derived from 6 patients (P = 0.0044 and P = 0.0033 for GAG release in IKKα- and IKKβ-knockdown micromasses, respectively, versus controls) (Figure 1C).
Enforced expression of IκBαSR by retroviral transduction with the IBIP vector also resulted in increased GAG levels, both in undifferentiated monolayer cultures and in differentiated micromass cultures (P = 0.043 for GAG release in monolayers and micromasses combined versus controls), with a stronger effect on micromass cultures (Figure 1D). To confirm that the activities of canonical NF-κBs were inhibited by IBIP infection, cells were stimulated with IL-1β, and total cellular RNAs were subjected to quantitative real-time PCR analysis. As expected, the expression of canonical NF-κB target genes, including RANTES, IL8, MCP1, IP10, and MIP1α, was inhibited in cells stably transduced by IBIP in comparison with control cells cultured with a BIP empty vector (results not shown).
We next examined the consequences of IKKα and IKKβ knockdowns for micromass ECM organization, using TEM. Results of TEM analyses of cells stably infected with the GL2 negative control retrovirus or with IKK-knockdown retroviruses are presented for a representative patient in Figure 2A (panels a, b, and c showing results for GL2, IKKβ knockdown, and IKKα knockdown, respectively). Transfection with alternate IKKβ and IKKα shRNA retroviruses harboring shRNA targeted to different exons yielded similar results (detailed data available online from the corresponding authors' Web site at http://www.sunysb.edu/biochem/marcu/olivottoA&R_07.html), thus ruling out the possibility of off-target effects.
Assessment of the accumulation of collagen fibrils and fibers in all micromasses indicated that the relative degree of accumulation was highest in the IKKα-knockdown micromasses (IKKα knockdown > IKKβ knockdown >> GL2), and that the ECM of IKKα-knockdown micromasses was replete with highly organized collagen fibers. Careful visual inspection of more than 20 TEM fields for the presence of organized ECM (including collagen fibers, fibrils, and GAG) in samples from each of 6 patients revealed that ablation of IKKα had a more potent enhancing influence on the overall ECM formation (P = 0.028 with IKKβ1 knockdown versus controls; P = 0.018 with IKKα3 knockdown versus controls) (see graph in Figure 2A).
Mediation of enhanced Col2 accumulation at the posttranslational level by IKKα or IKKβ knockdown.
Immunohistochemical analysis of frozen sections of chondrocyte micromasses revealed that, in comparison with matched GL2 negative controls, there was a strong enhancement of Col2 deposition in the absence of either IKKβ or IKKα (Figure 2B, upper left panels). To quantify the relative levels of Col2 (expressed as the percentage of signals per μm2 unit area), immunohistochemical signals were captured with a Nikon Eclipse 90i microscope and analyzed with Nikon Imaging Software. The results again revealed a strong enhancement of relative expression of Col2 in knockdown micromassess (Figure 2B, upper right panels; see also detailed data online). In addition, Western blotting of collagen proteins, extracted from micromass ECM by pepsin digestion, also revealed higher levels of Col2 in the absence of IKKα or IKKβ, with a rank order of IKKα knockdown > IKKβ knockdown >> GL2 micromasses (results not shown).
To begin to address how IKKβ and IKKα knockdowns alter the accumulation of ECM components in OA chondrocytes, we examined the levels of Col2a1 mRNA by quantitative real-time PCR, and also assessed the presence of Col2 neoepitopes produced by MMPs, as a relative measure of ECM remodeling activity. Quantitative real-time PCR assays of total cellular RNAs extracted from 1-week micromass cultures of multiple infected cell samples revealed similar increases, of up to ∼2.5 fold, of Col2a1 mRNA in cells deficient in either IKKα or IKKβ expression (graph in Figure 2B). Similar results (not shown) were obtained with 5× and 3× independent knockdowns of IKKα and IKKβ with multiple micromasses prepared from different patients. (P = 0.007 and P = 0.027, respectively, versus controls).
To examine the effects on Col2 turnover, we performed immunohistochemical analyses of GL2, IKKα-knockdown, and IKKβ-knockdown micromasses for the presence of Col2-3/4C neoepitopes, which are produced by the action of specific MMPs (MMPs 1, 8, and 13). As shown in Figure 2C, Col2-3/4C neoepitopes were clearly present in the ECM of GL2 micromasses, but were largely absent from both IKKα- and IKKβ-knockdown micromasses. Similar results (available online) were obtained for the DIPEN and NITEGE neoepitopes of aggrecan, which appear after the action of MMPs and aggrecanases, respectively (35). Taken together, these findings suggest that ECM increases of Col2 in IKKα- and IKKβ-knockdown micromasses are only partially mediated by changes in Col2a1 mRNA levels, and that Col2 remodeling is blocked in both IKKα- and IKKβ-knockdown micromasses, suggesting that each IKK strongly influences the accumulation of Col2 by posttranslational mechanisms and that this is presumably associated with the control of MMP activities.
Since MMP-13 is the major MMP responsible for Col2 remodeling at the hypertrophic phase of chondrogenesis (3, 5, 34, 39), we next explored the effects of each IKK knockdown on MMP-13 regulation. Basal and IL-1β–induced levels of MMP-13 secreted by micromass cultures were evaluated by quantitative ELISAs of duplicate samples from 3 independent patients. Neither IKKβ knockdown nor IKKα knockdown significantly affected the basal MMP-13 levels, but IKKβ knockdown abolished the IL-1β–induced MMP-13 secretion (P = 0.0277) (graph in Figure 2C). An inhibitory effect of IKKβ knockdown on IL-1β–induced levels of MMP-13 would also be consistent with MMP-13 being a target of the canonical NF-κB pathway (40). Even though there appears to be little, if any, effect on the basal MMP-13 levels in IKKβ-knockdown micromasses, the lack of proinflammatory factor–induced MMP-13 expression in IKKβ-knockdown micromasses might explain, in part, some of the increased accumulation of Col2. However, we found no apparent association between MMP-13 levels and the absence of Col2-3/4C neoepitopes in IKKα-knockdown micromasses.
Differential effects of IKKα and IKKβ loss on chondrogenic progression toward hypertrophy.
To investigate whether the effects of ablation of IKKα or IKKβ on ECM accumulation could be explained, in part, by differential blockade of the terminal differentiation process, we evaluated other signs of alterations in cell physiologic parameters associated with chondrogenesis, including 1) the size and proliferation rate of monolayer cells prior to being seeded in micromasses, 2) the viability of differentiated micromass cultures, and 3) the presence of calcium and mineralization deposits in long-term (3-week) micromasses. Primary micromass cultures are particularly well suited for this type of analysis because, akin to cartilaginous rudiments, they recapitulate all steps of chondrogenesis from the early prehypertrophic phase to the terminal hypertrophic phase, culminating in mineralization and subsequent ossification (37, 41, 42).
As shown in Figure 3A, flow cytometric measurement of forward and side scatter profiles revealed that IKKα-knockdown chondrocytes were significantly smaller in size than their GL2 counterparts (P = 0.027 for IKKα knockdowns [derived from 6 patients] versus controls). These same differences in cell size were also apparent by light microscopy after propidium iodide staining of cell nuclei (Figure 3B). In contrast to these clear effects of IKKα knockdown, IKKβ loss had either no effect or a very modest, and statistically nonsignificant, effect (Figure 3A and results not shown). In addition, IKKα-knockdown chondrocytes also proliferated at a significantly faster rate than their GL2 or IKKβ-knockdown counterparts (P = 0.043 versus both, at days 3–7 of the growth curve) (Figure 3C), which also correlated with a cell count of ∼2 times as many IKKα-knockdown cells as GL2 cells in the S–G2 phases of the cell cycle (Figure 3D).
Furthermore, IKKα-knockdown micromasses exhibited enhanced cellular viability in comparison with their matched GL2 negative controls, as shown by TEM analysis of 3-week infected micromasses from 7 patients (P = 0.0028 versus controls) (Figures 4A and B). In contrast, the viabilities of IKKβ-knockdown micromasses (in infected samples from 6 patients) were not significantly different from those of their matched GL2 samples (Figure 4A and results not shown). Cells were considered viable when both the nuclear and cytoplasmic membranes appeared integral and euchromatin was present, and were scored as nonviable when plasma membranes or nuclei appeared fragmented, which is indicative of necrosis or apoptosis, respectively (see representative TEM images of micromass cells in Figure 4B).
However, IKKα and IKKβ knockdowns both had a similar physiologically beneficial effect on mitochondrial morphologic features in comparison with their GL2 controls. The GL2 chondrocytes exhibited spheric, swollen, and emptied mitochondria, which are characteristic of terminally differentiated hypertrophic chondrocytes, whereas the mitochondria in IKKα3- and IKKβ1-knockdown micromasses were elongated, larger, darker, and also filled with septa, which is indicative of a healthier physiologic status (Figure 4C; see also additional results in cells from another patient, available online).
In addition, 2-week micromasses were examined for the presence of Col10, a marker of the transition to chondrocyte hypertrophy (2–4). Immunohistochemical analysis revealed clear Col10 staining in micromasses of GL2 control chondrocytes but no Col10 staining in IKKα-knockdown micromasses and only mild to weak staining in IKKβ-knockdown micromasses (see representative examples of staining results in cells from 1 patient in Figure 4D; see also similar results in cells from another patient, available online).
We then examined long-term micromasses for the accumulation of calcium deposits and other evidence of mineralization areas. Calcein staining revealed strong calcium deposition in control GL2 micromasses (30) but not in the absence of either IKKβ or IKKα (Figure 5A). The results in control GL2 micromasses were consistent with the presence of mineralization vesicles and electron-dense mineral areas (37), as revealed by TEM (Figure 5A, right panel). Analogous to these results, enforced expression of IκBαSR in the micromasses also inhibited calcium deposition and the appearance of mineralization areas and vesicles as compared with the effects of a BIP empty retroviral vector control (Figure 5B, right panel).
Differential effects of IKKα and IKKβ knockdowns on SOX9 and RUNX-2 expression.
We next determined the effects of each IKK on SOX9 (the major chondrogenic initiating transcription factor) and RUNX-2 (one of the essential transcription factors orchestrating terminal chondrogenic hypertrophy). Both knockdown of IKKβ and enforced expression of IκBαSR (using the IBIP retroviral vector) resulted in significantly enhanced levels of SOX9 in primary proliferating chondrocytes, whereas loss of IKKα caused a reduction in SOX9 expression (Figures 6A and B). Six IKKα3-knockdown patient samples showed decreased SOX9 expression (P = 0.043), while the IKKβ1 and IKKβ4 infections of 3 and 2 patient samples, respectively, increased SOX9 expression (P = 0.043). Moreover, because the effects of enforced IκBαSR expression and IKKβ ablation were similar, this strongly implies that IKKβ has intrinsic activity in OA chondrocytes in the absence of an overt extracellular inflammatory/stress-like stimulus. Indeed, since SOX9 is known to have opposing effects early and late in chondrogenesis (9), elevated levels of SOX9 in the IKKβ-knockdown micromasses could enhance ECM formation while simultaneously inhibiting endochondral ossification. However, the pronounced inhibitory effect of IKKα knockdown on differentiation toward hypertrophy would appear to be independent of SOX9.
Finally, we evaluated RUNX-2 expression by quantitative real-time PCR and immunohistochemistry (Figures 6C and D). Whereas IKKβ knockdown had no significant effect on RUNX-2 levels, IKKα knockdown repressed RUNX-2 mRNA levels in 1-week micromasses (P = 0.043) (Figure 6C). Similarly, RUNX-2 protein levels were repressed by ablation of IKKα in long-term (3-week) micromasses (Figure 6D; see also similar results from another patient, available online).
Evidence continues to accumulate to indicate that OA chondrocytes, in comparison with their normal articular counterparts, exhibit abnormal transcriptional programming (4, 43), such as the enhanced expression of a variety of inflammation mediators including IL-1β, IL-8, monocyte chemoattractant protein 1, nitric oxide, and TNFα (which are all direct targets of NF-κB signaling) (1). Moreover, the abnormal in vivo phenotypes and physiology of OA cartilage can be recapitulated in differentiating micromass cultures (44), and as shown in the present and previous studies, these cultures can also be used to parallel the stepwise process of chondrogenesis (2, 5).
Indeed, during the maturation process, micromasses established from OA chondrocytes show, in progression, the following in vitro expression patterns, which reflect those displayed by their in vivo counterparts: 1) deposition of GAG and Col2 (as in healthy cartilage and in growth plate chondroprogenitor proliferation and differentiation), 2) initiation of ECM remodeling (as in OA cartilage, with enhanced degradation and hypertrophy in the growth plate cartilage), 3) appearance of Col10 (as in hypertrophy in OA and growth plate cartilage), 4) continued ECM remodeling in conjunction with a loss of cell viability (as in advanced OA, in which cell death is a dominant event, and in growth plate terminal differentiation), and 5) calcium deposition and other evidence of mineralization (as in OA osteophytes and growth plate cartilage matrix calcification). This stepwise process was differentially impaired by ablation of IKKα and IKKβ in the present study, thus suggesting that IKKβ and IKKα are important positive effectors of ECM remodeling and terminal differentiation toward a hypertrophic-like state, with IKKα, surprisingly, having a more pronounced role than IKKβ. Any in vitro culture system has its own inherent limitations with respect to direct comparisons with the in vivo scenario; nevertheless, in vitro models serve as a crucial link between the in vivo observations in patients and the validation of disease hypotheses in in vivo animal models.
Inhibition of canonical NF-κB with an IκBαSR or knockdown of IKKβ expression enhanced formation of the ECM in conjunction with a block in hypertrophy and endochondral ossification of primary OA chondrocytes in differentiating micromass cultures. Our results showed that the enhanced GAG and Col2 deposition in the absence of IKKβ is the result of a combination of factors: 1) increased anabolism, caused, at least in part, by elevated SOX9 levels, and 2) reduced catabolism, exemplified by a suppression in ECM remodeling in conjunction with a differentiation block that prevented terminal calcification. Some of these observations are consistent with, and extend, the findings of earlier work in which inhibition of canonical NF-κB activation was found to enhance GAG deposition in the early phase of chondrogenesis in murine chondrocytic cells, which was correlated with the posttranscriptional enhancement of SOX9 mRNA levels (8, 45). However, our results indicate that anabolic effects are not the major cause of enhanced Col2 accumulation by IKKβ-knockdown micromasses and, particularly, by IKKα-knockdown micromasses; instead, we found that the effects on Col2 were mostly due to a posttranslational phenomenon linked to suppression of ECM remodeling.
Moreover, our findings suggest that the canonical NF-κB pathway via IKKβ signaling takes on an intrinsic role in the context of differentiating OA chondrocytes. Indeed, our results revealed that the major consequence of IKKβ ablation (mirrored by the action of an IκBαSR) on chondrocyte physiology is blockade of terminal differentiation. This is in keeping with our previous observations showing that pharmacologic or IκBαSR-mediated NF-κB inhibition in immortalized chondrocyte cell lines hampers their proinflammatory factor–induced expression of known promoters (IL-8) or markers (MMP-13) of chondrocyte hypertrophy (28). Similarly, findings from other studies have shown that activators as well as downstream effectors of canonical NF-κB signaling regulate the expression of collagenase 3 (MMP-13) (46–48).
IKKα ablation had a broader range of effects on OA chondrocyte physiology. IKKα knockdown dramatically enhanced ECM formation, as evidenced by the accumulation of highly organized Col2 fibers, a phenomenon that was independent of SOX9 levels but associated with suppression of RUNX-2 levels. Increased Col2 deposition in IKKα-knockdown micromasses was largely due to strong inhibition of collagen remodeling, as evidenced by the general absence of MMP-specific Col2 neoepitopes, which was also observed following ablation of IKKβ. Interestingly, this pronounced blockade of ECM remodeling appeared to be independent of any effects on basal MMP-13 levels, but possibly dependent on the subsequent suppression of MMP activity in IKK-knockdown micromasses, an interesting possibility that is under further investigation. Moreover, IKKα ablation uniquely resulted in an increase in the proliferative capacity and reduction in size of the undifferentiated chondrocytes and enhanced survival of the differentiated micromass cultures. The enhanced viability of IKKα-knockdown cells is not likely to be attributable to stronger, ECM-dependent survival signals, because the ECM was well preserved in IKKβ-knockdown micromasses, which displayed even higher GAG content than that in IKKα-knockdown cells. It is more likely that the selective blockade of the canonical activation pathway by IKKβ knockdown abrogated the antiapoptotic activity of NF-κB. Taken together, our findings suggest that the loss or inhibition of IKKα expression could potentially ameliorate the degenerative aspects of OA that are characterized by exacerbated endochondral ossification and excessive ECM remodeling coupled with increased cell death, each of which is a frequent complication of OA. Moreover, because IKKα knockdown increased the replicative potential and survival of OA chondrocytes, this could also provide an additional means of attenuating the stress-induced senescence of OA chondrocytes.
OA pathogenesis is a complex phenomenon triggered by abnormal joint biomechanics, and the disease can also occur in aged cartilage, in which cellular senescence has been documented at the molecular level (49). Unlike OA chondrocytes, proliferating chondrocytes in normal articular cartilage progress through their differentiation steps without being “inappropriately” pushed to exit their prehypertrophic time window of ECM formation to prematurely undergo terminal ECM remodeling and endochondral ossification (3, 4, 43). Maturational arrest of normal chondrocytes appears to be a dynamic state maintained by molecular constraints, whose failure results in inappropriate maturation of the chondrocytes toward a hypertrophic phenotype (4).
In terms of a propensity to hypertrophy, OA or aged chondrocytes are poles apart compared with normal chondrocytes, but even the latter under particular “mineralizing” culture conditions can be induced to progress toward hypertrophy and terminal differentiation (50). Conceivably, IKKα activity could drive either OA or aged normal chondrocytes into a mature, terminal phenotype. Even if, in the future, studies show that IKKα has similar properties in normal (i.e., non-OA) chondrocytes, which would be obtained from age-matched donors under in vitro conditions, we propose that IKKα can still be considered a potential target for ameliorating the more damaging effects of OA and also for maintaining articular chondrocytes in a state of maturational arrest. Moreover, defining the effects of IKKα ablation on cartilage as either a template for bone formation or as a target of ectopic endochondral ossification (as occurs in OA) now represents an important goal of future in vivo experiments in the context of an appropriate animal model. In this regard, we suggest that, except in articular cartilage, the loss of IKKα in vivo could conceivably be of less clinical relevance in older patients with debilitating OA, since the final outcome of terminal chondrocyte differentiation is normal bone formation.
Thus, our observations surprisingly reveal that exacerbated IKKα activity may be responsible for at least part of the idiosyncratic behavior of OA chondrocytes, including their accentuated collagen remodeling and ectopic ossification in vivo, which is often associated with the appearance of osteophytes. Future studies will be devoted to defining the mechanisms of action and downstream targets of IKKα involved in the chondrogenic programming of OA cells and their normal counterparts.
Dr. Marcu had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Olivotto, Borzi, Andrea Facchini, Marcu.
Acquisition of data. Olivotto, Vitellozzi, Pagani, Annalisa Facchini, Battistelli, Penzo, Xiang Li, Jun Li, Marcu.
Analysis and interpretation of data. Olivotto, Borzi, Battistelli, Flamigni, Falcieri, Marcu.
Manuscript preparation. Olivotto, Borzi, Marcu.
Statistical analysis. Borzi.
- 25Chondrocyte hypertrophy and apoptosis induced by GROα require three-dimensional interaction with the extracellular matrix and a co-receptor role of chondroitin sulfate and are associated with the mitochondrial splicing variant of cathepsin B. J Cell Physiol 2007; 210: 417–27., , , , , , et al.
- 46Interleukin-1 induction of collagenase 3 (matrix metalloproteinase 13) gene expression in chondrocytes requires p38, c-Jun N-terminal kinase, and nuclear factor κB: differential regulation of collagenase 1 and collagenase 3. Arthritis Rheum 2000; 43: 801–11., , , , .
- 50Expression of type X collagen and matrix calcification in three-dimensional cultures of immortalized temperature-sensitive chondrocytes derived from adult human articular cartilage. J Bone Miner Res 1998; 13: 432–42., , , , , , et al.