Sequential Alterations in Catabolic and Anabolic Gene Expression Parallel Pathological Changes during Progression of Monoiodoacetate-Induced Arthritis

Chronic inflammation is one of the major causes of cartilage destruction in osteoarthritis. Here, we systematically analyzed the changes in gene expression associated with the progression of cartilage destruction in monoiodoacetate-induced arthritis (MIA) of the rat knee. Sprague Dawley female rats were given intra-articular injection of monoiodoacetate in the knee. The progression of MIA was monitored macroscopically, microscopically and by micro-computed tomography. Grade 1 damage was observed by day 5 post-monoiodoacetate injection, progressively increasing to Grade 2 by day 9, and to Grade 3–3.5 by day 21. Affymetrix GeneChip was utilized to analyze the transcriptome-wide changes in gene expression, and the expression of salient genes was confirmed by real-time-PCR. Functional networks generated by Ingenuity Pathways Analysis (IPA) from the microarray data correlated the macroscopic/histologic findings with molecular interactions of genes/gene products. Temporal changes in gene expression during the progression of MIA were categorized into five major gene clusters. IPA revealed that Grade 1 damage was associated with upregulation of acute/innate inflammatory responsive genes (Cluster I) and suppression of genes associated with musculoskeletal development and function (Cluster IV). Grade 2 damage was associated with upregulation of chronic inflammatory and immune trafficking genes (Cluster II) and downregulation of genes associated with musculoskeletal disorders (Cluster IV). The Grade 3 to 3.5 cartilage damage was associated with chronic inflammatory and immune adaptation genes (Cluster III). These findings suggest that temporal regulation of discrete gene clusters involving inflammatory mediators, receptors, and proteases may control the progression of cartilage destruction. In this process, IL-1β, TNF-α, IL-15, IL-12, chemokines, and NF-κB act as central nodes of the inflammatory networks, regulating catabolic processes. Simultaneously, upregulation of asporin, and downregulation of TGF-β complex, SOX-9, IGF and CTGF may be central to suppress matrix synthesis and chondrocytic anabolic activities, collectively contributing to the progression of cartilage destruction in MIA.


Introduction
Osteoarthritis (OA) is a debilitating joint disease, causing severe pain and physical disabilities to millions of people worldwide [1,2,3]. The etiopathology of OA is multifactorial. Chronic inflammation, degeneration of the extracellular matrix and abnormal remodeling of the underlying bone, all take part in cartilage destruction [4,5,6,7]. Cartilage matrix is mainly composed of collagens and proteoglycans synthesized by chondrocytes residing in the matrix. Collagens, mainly Type II, type IX and type XI, provide the tensile strength, whereas proteoglycans rich in water act as shock absorbents during cartilage loading.
At the onset of the disease, an etiologic agent or an insult/ trauma of the joint causes focal edema and minor surface erosion in the cartilage. Progression of cartilage destruction is marked by further fibrillation accompanied by loss of matrix and chondrocytes in the superficial layers of the cartilage. Proliferating chondrocytes become apparent as cell clusters in the middle zone of the cartilage. Further progression of fibrillation and cartilage erosion leads to increased loss of cartilage matrix resulting in bone denudation, the hallmark of osteoarthritic lesions. The rate of disease progression and amount of joint damage is not predictable, and varies among patients. However, the progression of joint damage appears to follow a similar pattern and therefore can be categorized into various grades according to the extent of damage in the cartilage and bone [8,9].
Chondrocytes take part in cartilage damage by synthesizing catabolic cytokines and enzymes that breakdown the matrix as well as impairing their ability to repair the matrix. Studies focusing on the molecular events in human OA inception and progression by genomic or proteomic profiling of intra-articular lesions have revealed distinct gene profiles in OA specimens as compared to visually unaffected cartilage [10,11,12,13,14,15]. Similarly, gene association studies in large populations have identified a number of genes that might confer susceptibility to OA [16,17,18,19]. However, the knowledge of the discrete molecular events that support the time-dependent progression of OA remains incomplete.
In this study, we aimed to conduct a systematic longitudinal examination of molecules and pathways associated with the progression of cartilage damage. A widely used model of monoiodoacetate-induced arthritis (MIA) of the rat knee was utilized [20,21]. The progression of MIA was analyzed by macroscopic, microscopic and micro-tomographic (mCT) analyses and categorized into various stages of cartilage damage using the grading system of Pritzker et al. [9]. A transcriptome-wide analysis was conducted on the cartilage of temporally well-defined stages of MIA and compared to those of sham control cartilage. Ingenuity Pathways Analysis (IPA) was employed to obtain key insights into molecular relationships and networks/mechanisms during the progression of cartilage destruction. This analysis linked the microarray data to relevant, manually curated information from periodically updated knowledge databases in order to interpret the global impact of differentially regulated molecules during MIA progression. We believe that this study is the first to systematically elucidate the longitudinal time-dependent gene regulation and molecular networks/mechanisms throughout the course of MIA progression and cartilage destruction.

Results
Macroscopic and microscopic changes in cartilage and subchondral bone during the progression of MIA The progression of MIA was monitored by overall macroscopic and microscopic changes at the distal ends of femurs ( Figure 1). The articular surface of Cont femurs exhibited normal cartilage morphology, histology and bone imaging by mCT, typical of Grade 0/healthy cartilage ( Figure 1 a-d, Movie S1). The progression of MIA followed the similar pathologies as described by Guzman et al. [22]. Typically, femurs from MIA afflicted knees exhibited greater extent of cartilage damage around the patellar groove than on femoral condyles and intercondylar fossa (Figure 1 e, f, i, j, m, n). The examination of time-dependent progression of knee cartilage damage showed that, on day 5 post MIA induction (MIA5), femurs showed cartilage damage typical of Grade 1, i.e., superficial fibrillation, chondrocyte proliferation, clustering and disorientation, and some loss of tidal ridge demarcation (Figure 1eg) [9,22]. Bone damage was not apparent microscopically or by mCT imaging at both patellar and condylar surfaces (Figure 1e-h, Movie S2).
Analysis of MIA9 cartilage revealed marked lesions at the apexes of condyles and ridges of the patellar groove (Figure 1i-k). The loss of the tidal layer and deeper lesions in some areas were observed. Chondrocytes appeared larger, some with multiple nuclei and disarrayed. Subchondral bone marrow extensions towards cartilage and deposition of fibrous tissue in the lesions typical of Grade 2 cartilage degeneration were apparent. The mCT images revealed scattered subchondral bone lesions on the femoral condyles and patellar groove (Figure 1l, Movie S3).
On day 21 post-monoiodoacetate injection (MIA21), increased cartilage and bone damage in the patellar groove and ridges, fulldepth lesions and pits on the femoral condyles were observed (Figure 1m-o). Histology revealed fissuring with matrix loss, fibrocartilage formation within the denuded cartilage and abnormal subchondral bone marrow intrusion typical of Grade 3 to 3.5 damage. Micro-CT imaging showed pitted areas of bone loss on the femoral condyles and patellar groove (Figure 1p, Movie S4).
Transcriptome-wide regulation of gene expression during the progression of MIA We next determined the changes in transcriptome-wide gene expression profiles during the progression of MIA in the distal end of femoral cartilages in Cont, MIA5, MIA9 and MIA21 rats exhibiting Grade 0, Grade 1, Grade 2 and 3-3.5 cartilage damage, respectively. Principal components analysis (PCA) revealed relatively uniform distribution of overall gene expression among the samples in each group (n = 3) except in MIA9 group, where the overall gene expression was distributed between MIA5 and MIA21 ( Figure 2A). Significant differences in gene expression over the course of MIA progression were observed, as evidenced by the average F ratio (signal to noise ratio) of 18.8.
Of the 27,342 transcripts detectable by Affymetrix GeneChips array, 2,034 (7.44%) transcripts were significantly (p,0.05) and differentially up-or downregulated at one or more time points by more than two-fold change. In the hierarchical clustering analysis of the differentially regulated genes (p,0.05, over 62-fold change), distinct sets of genes were regulated at each stage of MIA progression ( Figure 2B). The most interesting information derived from the hierarchical clustering was that: (i) as compared to Cont, the maximal changes in gene expression occurred in MIA5, judging by its farthest distance from Cont ( Figure 2B), followed by MIA21 and MIA9; and (ii) distinct individual sets of genes were temporally either upregulated or suppressed during the progression of MIA.
Cluster analysis of major functional genes during the progression of MIA Among the 2,034 transcripts that were significantly up-or downregulated during the progression of MIA, 1,971 were unique genes annotated by Ensembl. These genes were then analyzed by Davies-Bouldin index [23] to render optimal number of clusters for partition clustering and were assigned to one of the five trends of temporal gene regulation ( Figure 3). The graphs represent 10 most regulated genes in each cluster, and were groups of genes that exhibited: peak-upregulation at day 5 after MIA induction, followed by decrease in gene expression (Cluster I ); peak-upregulation at day 9 after MIA induction (Cluster II ); gradual increase in gene expression that peaked at day 21 after MIA injection (Cluster III ); peak-downregulation at day 5 after MIA injection, followed by relative increase in gene expression (Cluster IV ); and peak-downregulation at day 9 after MIA induction (Cluster V ). Validation of at least two genes in each cluster by rt-PCR exhibited similar trends in the differences in gene expression as in microarray analysis ( Figure 4). However, rt-PCR technique being more sensitive contributed to greater fold changes in gene expression as compared to the microarray analysis.
Among the five distinct biologically functional gene clusters, IPA identified three clusters mainly associated with inflammation and immunological disorders (Clusters I, II and III ), and the remaining two clusters associated with musculoskeletal function and disorders (Clusters IV and V ) ( Figure 3, Table 1). To delineate the overall functional relevance, the genes were further categorized into 7 functional sets: (i) Inflammation (cytokines, chemokines, and their receptors); (ii) Inflammation regulators (mediators, transcription factors, and signaling molecules that regulate inflammation); (iii) Cell division/proliferation; (iv) ECM (molecules of the matrix); (v) ECM regulators (molecules that regulate matrix synthesis and degradation); (vi) Growth factors (growth factors and their receptors); (vii) Growth factor regulators (signaling molecules and transcription factors that regulate growth factors) ( Figure 5, Tables 2, 3, 4, 5 and 6). Genes including molecules involved in cell metabolism, transporters and ion channels, and those with unknown functions were not included in the present analysis. The genes in these Tables reflect: genes with known function, the degree of gene regulation, and are in proportion to the group of genes regulated in a particular cluster shown in Figure 5.
Cartilage with Grade 1 damage (MIA5) exhibits gene expression associated with innate immunity and cell proliferation.
Interestingly, the genes associated with cell cycle/division/ differentiation such as Diap3, Anln, Prc1, Emb, Kif4, Kif23, Dusp6, Vav1, Ccnb1, Ccna2, Ccnb2, Ccne1, Ccnf, and Cdk6 were also highly upregulated (Table 2, Figure 5A, Table S1). The expression of Figure 1. Progression of MIA at the distal femoral ends by macroscopic, microscopic, and mCT analyses. Right knees of rats were given an intra-articular injection of MIA on day 0, and distal ends of right femurs examined on post-injection days 5 (Grade 1 damage, MIA5), 9 (Grade 2 damage, MIA9) and 21 (Grade 3-3.5 damage, MIA21) and compared to saline-injected sham control (Cont). Macroscopic view of condyles, patellar grooves of cartilage, histology, and subchondral bone imaging by mCT of: (a, b) Cont femur showing smooth surface, (c) normal histology and no bone lesions on the femoral condyles and patellar grove and (d) lack of lesions in the subchondral bone (Movie S1); (e, f) MIA5 cartilage showing superficial abrasions on the condyles (black arrows) and patellar groove (white arrows), (g) superficial fibrillation (black arrow), chondrocyte clustering and disorientation (blue arrow), and (h) no bone lesions in mCT images (Movie S2); (i, j) MIA9 cartilage exhibiting lesions at the apexes of condyles (black arrow) and ridges of the patellar groove (white arrow), (k) thinning of cartilage, matrix and cell loss above the tidal layer with large disarrayed chondrocytes (black arrow), and some multinucleated chondrocytes (blue arrow), subchondral bone marrow/fibrous tissue extension in the cartilage typical of Grade 2 damage (white arrow), and (l) scattered subchondral bone lesions on the femoral condyles and patellar groove in mCT images (Movie S3); (m, n) MIA21 cartilage exhibiting increased lesions and damage on the condyles (black arrows) and patellar groove and ridges (white arrow), (o) delamination of surface, full depth cartilage lesions and denuded cartilage layer at some places (black arrow), and (p) increased subchondral bone lesions on the femoral condyles and patellar groove in mCT images (Movie S4). Each figure shows representative right femur from separate rats from each group (n = 10). Arrows indicate cartilage damages. The distal ends of femurs showing 360u mCT projection can be found in Movie files S1 to S4. doi:10.1371/journal.pone.0024320.g001 these genes paralleled the chondrocyte proliferation characteristically observed as disoriented clusters of chondrocyte distributed in the cartilage (Figure 1g).
Despite the presence of cytokines like IL-1b and IL-33, genes for several ECM proteins involved in cell-matrix attachment were significantly upregulated in Grade 1 cartilage damage. These genes included Vcan, Fbln2, and Spon1. Additionally, proteinases with broad specificity involved in protein/matrix breakdown were upregulated such as Hpse, Ctsc, Ctss, Arsb, and Plau (Table 2).
Strikingly, asporin, a suppressor of TGF-b/receptor interactions was more than 9 fold upregulated in Cluster I [25]. Additionally, genes for growth factors involved in cell division or immune response such as, Fgf7, Csfrb, the regulators of Wnt signaling Sfrp1 and Sfrp2, were dynamically upregulated in cartilage with Grade 1 damage.
Cartilage with Grade 1 damage (MIA5) exhibits suppression of genes associated with matrix synthesis (Cluster IV) In parallel to marked upregulation of genes in cartilage with Grade 1 damage (MIA5, Cluster I ), several genes were significantly downregulated and were assigned to Cluster IV. These genes were associated with genetic disorders (163 genes, p-value 1.37E-06 -2.08E-02) and musculoskeletal development and function (95 genes, p-value 2.10E-07 -1.73E-02), and consisted of relatively higher proportion of the genes for extracellular matrix and their regulators ( Figures 3D & 5D, Table 3, Table S2). Interestingly, along with genes that induce cell division (Cluster I ), genes associated with suppression of cell growth and apoptosis were downregulated such as Scrg1 and Cidea in this cluster. Among cytokines, Cytl1 [26], IL23r, and the inhibitor of osteoclastogenesis Tnfrsf11b (osteoprotegerin), were major molecules suppressed, along with several proinflammatory mediators Sod3, Alox12, and Ptgds.
The scrutiny of global gene expression in cartilage with Grade 1 damage, also showed that several growth factors required for cartilage growth/homeostasis were dramatically downregulated, such as Gdf10, Ig f2, Ig fbp7, Bmp6, Fg frl1, Spock1, and Veg fa. Among growth factor regulatory proteins the most suppressed genes were Crim1, Sox9, Ltbp4, and htra, which may cumulatively retard cartilage repair.
Major genes upregulated in cartilage with Grade 2 damage were associated with chronic inflammation The Grade 2 cartilage damage showed upregulation of genes in Cluster II, belonging to family of genes prevalent in genetic disorders (116 genes, pvalue 3.21E-10 -6.79E-04) and inflammatory response and immune trafficking (96 genes, p-value 4.81E-12 -1.15E-03) ( Table 1). Overall gene expression profiles of articular cartilage from 3 separate rats in each experimental group as compared to Cont. Hierarchical clustering representing the transcripts that were significantly (p,0.05) and differentially up-or downregulated at one or more time points by more than twofold change. Note the maximal changes in overall gene expression occurred in MIA5, followed by MIA 21 and MIA9 as compared to gene expression in cont cartilage. doi:10.1371/journal.pone.0024320.g002 As compared to Grade 1, significantly fewer genes associated with cell cycle/division were upregulated in Grade 2 cartilage damage (Figures 3B & 5B, Table 4, Table S3). In fact, a number of genes involved in the inhibition of cell division, Dapk1 and Ccng1, were upregulated. The majority of genes significantly upregulated in Grade 2 (Cluster II) were associated with chronic inflammation such as chemokines and their receptor Ccl2, Ccl7, Ccl9, Ccr1, Ccr5, Cx3cr1, and Cxcl16 as well as cytokines involved in amplification of immune response Lif, Il7, Il18, and Ifngr2. More notably, cytokines that induce bone resorption such as Tnfsf11 (RANKL), Tnfrsf11a, Tnfrsf1B were significantly upregulated explaining the initiation of bone damage observed in m-CT images (Figure 1l). In parallel, genes involved in the regulation of inflammation were upregulated such as those associated with clotting cascade, Tfpi2, adherence, Itgam, Itgax, Itga2, and NF-kB signaling cascades Tank, Ripk2, NFkB2, NFkbie, Map2k3, and enzymes necessary for the regulation of inflammation Pik3cb, Dusp4, Ptpre, and Ptpn22.
Interestingly, the expression of genes for two matrix proteins, Col5a3 and Sdc1 was significantly upregulated. Besides these, the expression of genes associated with cartilage matrix degradation was prevalent in the cartilage with Grade 2 damage (MIA9). These genes were matrix metallopeptidase (MMP)-9, Mmp12, Mmp19, Adamts4, Adamts7, Adamts12, Hyal1, Hyal3, Arsb, and Adam8. Simultaneously, genes for inhibitors of proteases such as Timp1 and Serpine2 were also upregulated.
Major genes suppressed in the cartilage with Grade 2 damage were ECM and growth factor associated genes During the progression of cartilage damage, we also observed that a significant number of genes were downregulated in the cartilage with Grade 2 damage (MIA9, Cluster V). Genes in Cluster V, in parallel to Cluster II, were mainly matrix associated and demonstrated maximal suppression on day 9, and associated with genetic disorders (235 genes, pvalue 1.87E-12 -2.88E-02) and skeletal and muscular disorders (134 genes, p-value 1.31E-10 -2.88E-02) (Figures 3E &5E, Table 1).
There were several proinflammatory genes suppressed including IL-7, IL-16 and IL-17b involved in amplification of immune response, and Nrk in NF-kB signaling cascade (Table 5). Nevertheless, the most dramatically suppressed gene was matrilin 3, a major component of ECM, involved in the formation of filamentous networks [27]. The expression of several collagens integral to cartilage matrix such as collagens type -IXa1, -IIa1, -IXa2, -IXa3, -XIa1, -XXIVa1, and -Va3 were significantly downregulated. The expression of other cartilage matrix components involved in cell-matrix and . Partition clustering of significantly regulated genes. Partition clustering analysis of the genes that showed two fold or greater changes in their expression at one or more time points (p,0.05). The graphs represent 10 most regulated genes in each cluster. Identification of five gene clusters that exhibited maximal upregulation on day 5 (Grade 1 damage) followed by their downregulation (Cluster I); upregulation on day 9 (Grade 2 damage) followed by their downregulation (Cluster II); upregulation in a sustained manner showing maximal expression on day 21 (Grade 3-3.5 damage, Cluster III); downregulation of genes on day 5 followed by their upregulation (Cluster IV); and downregulation of genes on day 9 followed by their upregulation (Cluster V). Detailed description of these genes is given in Tables 1, 2, 3, 4, 5, and 6, and in Tables S1, S2, S3, S4, and S5. doi:10.1371/journal.pone.0024320.g003 matrix-matrix adhesion were suppressed such as Chad, Scin, Hapln1, Vit, Mgp, and Fbln5 (Table 5, Table S4) [24]. Additionally, gene expression of several molecules involved in collagen, chondroitin and hyaluran synthesis were suppressed (Adamts3, Adamts6, Chsy3, Has2) as well as those involved in mineralization (Alpl ).
Among the inflammatory genes, those involved in the regulation of T and B cell functions Cxcl13, IL15, and suppression of inflammation such as Il10rb, Lilbr4, Nfkbia, Socs3, and Pld2 were simultaneously upregulated. Additionally, genes involved in LPS responses and acute inflammation such as Tlr4, Lbp, F3, Alox5, Lpl, and Ptges were upregulated.
Interestingly, we also observed that several of the ECM associated proteins that were suppressed in Grade 2 cartilage damage, were relatively upregulated in cartilage with Grade 3-3.5 damage. For example, Col2a1, 50.3-fold downregulated in cartilage with Grade 2 damage, was only 9.06-fold suppressed in Grade 3-3.5 damage. Similar upregulation of ECM associated genes in Grade 3-3.5 damage relative to lesser damaged cartilage included Matn3, Col10a1, Col9a1, Col9a3, Col11a2, Col11a3, Col5a3, Major molecular networks involved in cartilage damage during the progression of MIA We next subjected genes in individual clusters to IPA to generate major functional molecular networks (Figures 6, 7 and 8). The significance and specificity of IPA-generated networks were based on the score of each network. The high score numbers signified that gene networks are extremely specific to each cluster. For example, a score of 43 of the molecular network in Cluster I ( Figure 6A) indicates that there is only a 1 in 10 43 chance of getting a network containing the same member of Network Eligible molecules, when same numbers of molecules are randomly picked from the IPA knowledge base.
The molecular network maximally upregulated in the specimens with Grade 1 cartilage damage (MIA5) were (i) acute inflammation and (ii) cell cycle/cell division-related genes in Cluster I. Genes that typically regulate innate immunity directly or via activation of other mediators formed this network. For example, IL-1b, which auto-regulates its own expression, may also upregulate expression of Ccr2, Trem2 (stimulates production of cytokines and chemokines in macrophages), IL10ra (receptor of IL-10), Ptgfr, Cyba and Cybb (phagocytic oxidases that generate superoxide), and NCF1 and -2 (oxidases that produce superoxides) ( Figure 6A). Strikingly, the genes associated with cell cycle including Vav1, Emb, Prc1, Kif4A, Kif23, Kif20A, and Dock2 were also prevalent in this network despite the presence of inflammation ( Figure S1).
Interestingly, in parallel to upregulation of genes associated with innate immunity and cell cycle in Cluster I, other pathways were simultaneously suppressed as observed in the major molecular network for the Cluster IV (score 32, Figure 6B). For example, asporin, an inhibitor of TGF-b [25] and a member of Cluster I, was considerably upregulated at this stage of cartilage damage, and may be responsible for preventing activation of TGF-b complex, consequently downregulating matrix proteins and growth factors such as Sox9, alkaline phosphatase, aggrecan, Cilp, Cilp2, and other proteoglycans/collegens, directly or via activating intermediary molecules in Cluster IV ( Figure 6B).
The IPA of genes upregulated in cartilage with Grade 2 damage, revealed a molecular network (score 34) involved in chronic inflammation, immune cell trafficking and perpetuation of inflammatory response (Cluster II, Figure 7A). This network appeared to be activated by TNF receptor and may invoke the activities of the NF-kB signaling cascade, RIPK2, a potent activator of NF-kB and inducer of apoptosis and chemokines. The activation of NF-kB complex in turn may play a central role in upregulating the expression of MMPs that cleave matrix proteins, chemokines that attract immune cells, and Cd44 that mediates cell adhesion/migration via hyaluronate/matrix attachment. Similarly, based on the existing role of chemokines, their upregulation may further augment activity/gene expression of chemokines and their receptors, such as Ccl7, Ccl9, Ccl13, Ccr1, Ccr5 and Pf4 (Cxcl4) that are important for amplification of immune response and recruitment of immune cells to the site of inflammation.
Simultaneous with persistent inflammation in the cartilage with Grade 2 damage, the suppression of genes involving matrix synthesis in Cluster V was observed (score 39, Figure 7B). IPA network analysis suggested that the major foci of the molecular network suppressed were TGF-b complex, Ig fbp, Ctg f and Eg f. Suppression of these genes may have downregulated matrix proteins such as collagens (-type II alpha-1, -type X alpha1, -type XI alpha-1 and -2), and molecules involved in matrix synthesis such as Adamts3 and Hapln1 (stabilizes cartilage matrix). More importantly, a significant suppression of TGF-b complex in this network may have also downregulated many genes associated with bone formation such as Bglap, Dlx5, Alpl, and Bmpr1. The downregulation of these genes during chronic inflammation may result in the failure of matrix repair, thus accelerating the damage.
In the major molecular network in Cluster III (score 29, Figure 8A), related to pathologies observed in Grade 3-3.5 cartilage damage, many of the genes were associated with immune suppression and adaptation such as Socs3, Osmr, Gas7 and Il10rb [28]. Interestingly, at this stage, except for IL-15, the upregulation of other inflammation-associated genes such as NF-kB complex, IL-1 complex, IFN alpha and IFN beta complex, MHC complex, and IL-12, was not evident. However, several genes that are associated with B cell, T cell and macrophage proliferation, differentiation, and migration, such as complement cascade (innate immunity and macrophage activation), IL-15 (stimulates T-lymphocyte proliferation), and interferon-induced transmembrane protein 3 (Ifitm3, mediates cellular immunity) were upregulated.

Discussion
To the best of our knowledge, this study documents the first evidence of temporally controlled global gene regulation and identifies the major determining molecular networks that likely control the progression of cartilage damage in a well-established rat model of MIA. We examined changes in the gene expression profiles by transcriptome-wide microarray analysis in relation to the progression of MIA determined by macroscopic, microscopic, and mCT imaging to assess bone involvement [22,29,30,31]. This model of experimental OA was considered useful due to its similarities to the pathogenesis of OA, reproducibility, reasonable duration of the test period, and ability to induce cartilage damage without confounding effects of surgical wounding on the joint tissues [21,22,29]. In this experimental model, the first 3 weeks of MIA progression showed major changes in the cartilage destruction and Grade 6 damage is achieved over a period of 8 weeks (56 days) [22]. After 3 weeks of MIA progression, the cartilage loss is slowly replaced by fibrocartilage and bone. Therefore, we have focused on the initial period of 3 weeks (21 days) where the cartilage damage advanced to Grade 3-3.5. Although the progression of MIA in this model was much faster, it exhibited a sequential progression of cartilage damage observed over a longer period of time in other models of OA. Furthermore, as described earlier, less than 2% cell death was observed due to the monoiodoacetate-induced injury on day 1 after monoiodoacetate injection [32]. Nevertheless, rodent models cannot depict arthritis exactly to humans, as the joint mechanics differ in small quadrupeds [33].
The foremost findings from the transcriptome-wide gene expression profiles are that the MIA afflicted cartilage showed stage specific reproducible changes in gene expression, as demonstrated by the hierarchical and partition clustering analyses. Strikingly, MIA progression involves up-or downregulation of approximately 7.44% of the transcripts by more than two-fold, at one or more time points (p,0.05). Furthermore, discrete sets of genes at each stage of cartilage damage appear to maximally regulate set of genes associated with inflammation and ECM degradation.
The overall gene expression profiles and the IPA derived from these profiles suggest that Grade 1 cartilage damage is likely associated with upregulation of genes required for: (i) acute inflammation/innate immunity such as superoxides, complement components, integrins, IL-1/IL-1r, chemokines and their receptors, and monocyte activating factors; (ii) chondrocyte-matrix interactions, i.e., versican, fibulin and microfibril; (iii) cleavage of matrix and cell associated proteins such as broad specificity proteases cathepsins and heparanase [34]; and (iv) cell proliferation such as cell cycle/proliferation and mitogenic growth factors (FGF7 and CSF receptor b). The presence of proliferating cells may support increased cell division in the cartilage with Grade 1 damage (Figure 1g). In parallel, suppression of genes essential for: (i) proteoglycan synthesis/assembly, i.e., Cilp and Cilp2, Acan, Bgn, Eln, Sdc2 and cell-matrix adhesion such as Col XVI 1a, and ColXVII 1a; and (ii) inhibitors of peptidases (Timp3, Serpina3, Pi15), may further support increased proteolytic breakdown of cartilage matrix. More importantly, upregulation of asporin that mediates downreg- ulation of TGF-b activity, and consequently suppression of Sox9 may be responsible for the dramatic suppression of the above proteoglycan-associated genes. Interestingly, genes such as asporin, IL-1b, IL-1 receptor-like 1, cathepsin S, PGE receptor (EP4) and integrins are also upregulated in OA in humans and experimental animals, suggesting their possible role in the early stages of the disease progression [35,36,37,38,39,40,41,42].
In summary, the present study provides evidences that the progression of cartilage damage is driven by complex but precise regulation of gene clusters that are induced or suppressed during a specific stage of cartilage damage (Figure 9). Cartilage with close to Grade 1 damage exhibited upregulation of genes associated with acute inflammation and innate immunity, broad specificity proteases, and cell cycle/division and suppression of genes for proteoglycan synthesis. Gene expression in cartilage with Grade 2 damage was associated with dynamic upregulation of genes driven by NF-kB such as inflammatory mediators/cytokines, metallopeptidases, and immune trafficking. Chronic inflammation was paralleled by suppression of growth factors and collagens. Cartilage with Grade 3-3.5 damage exhibited an adaptive response evidenced by upregulation of anti-inflammatory genes. Simultaneously, there is a significant reduction in the suppression of matrix-associated proteins and growth factors as compared to cartilage with Grade 1 or Grade 2 damage. Collectively, the precise modulation of sequential up and down regulation of these genes may support the cartilage damage observed during the progression of MIA. Further elucidation of the key molecules that regulate the expression of catabolic as well as anabolic genes is critical in understanding the mechanisms of cartilage damage in experimental and human OA.

Monoiodoacetate-induced arthritis
The work was performed under the protocol number 2009A0138 approved by the Institutional Animal Care and Use Committee, The Ohio State University. Female Sprague-Dawley rats, 12-14 weeks old (Harlan Labs, IN) were randomly assigned to 4 groups (15 rats/group). The right knees of rats were given intra-articular injection of 50 ml saline in sham controls (Cont, n = 15), or monoiodoacetate (2 mg/50 ml saline) in experimental animals to induce MIA (n = 45). Following administration of monoiodoacetate, the cartilage exhibited Grade 1, Grade 2, or Grade 3-3.5 on days 5, 9, and 21, respectively. Therefore, progression of cartilage damage and changes in gene expression profiles were carried out on day 5 (MIA5; n = 15), day 9 (MIA9; n = 15), or day 21 (MIA21; n = 15) post-monoiodoacetate injection. Among them, 5 femurs from each group were snap-frozen in liquid nitrogen for microarray and real time-Polymerase Chain Reaction (rt-PCR) analyses (n = 5), and the remaining 10 femurs were immediately examined macroscopically using a stereomicroscope and then fixed in 10% buffered formalin for microscopic examination of the cartilage and bone, or mCT imaging to assess the overall subchondral bone loss.

Macroscopic and microscopic examination
Gross morphologies of femurs were recorded photographically under a stereomicroscope. The microscopic examination was performed in paraffin embedded and Hematoxylin-Eosin (H&E) stained femurs. The cartilage damage was graded according to Pritzker et al. [9].

MicroCT analysis
To assess the involvement of subchondral bone in MIA, the femurs were scanned at approximately 19.4 mm resolution on an Inveon microCT from Siemens Preclinical (Knoxville, TN). The scans were run as 220 degree half scans with a theta of 0.5 degrees, with 500 ms exposure, and 700 projections/360 degrees. The source for the acquisition was run at 80 kV and 500 mA with Please see Table 2 for group description. A full list of these genes is given in Table S3. doi:10.1371/journal.pone.0024320.t004

RNA extraction and microarray analysis
The cartilage from the distal end of individual femur (10-15 mg/femur) was examined under a stereomicroscope (Zeiss, Germany). Superficial articular cartilage on the patellar and condylar surfaces of the distal ends of femur was chipped off in a frozen state, avoiding the areas immediately around lesions. The cartilage chips from each knee were collected separately, and pulverized into 1 mm fragments in a Mikrodismembrator S (Sartorious, France) at 2500 rpm for 30 seconds [32]. RNA was extracted with Trizol reagent (Invitrogen, CA), and each sample of RNA was analyzed in a 2100 Bioanalyzer (Agilent, CA) to ensure optimal quality of RNA [64].
A total of 300 ng of RNA was used for cDNA synthesis and labeling using Whole Transcript (WT) cDNA Synthesis and Amplification Kit, and WT Terminal Labeling Kit (Affymetrix, CA). The labeled samples were hybridized on Affymetrix GeneChip Rat Gene 1.0 ST Array and scanned at the Microarray Shared Resource Facility at the OSU Comprehensive Cancer Center.
The intensity scans from three biologically independent arrays per treatment were subjected to gene expression analysis using Partek Genomic Suite version 6.4 (Partek Inc., MO). The significance of differences among the conditions was calculated by the analysis of variance (ANOVA) and only significantly regulated transcripts (p,0.05) were considered for further analyses. Variations among the samples in each condition were examined by principal components analysis (PCA), and subjected to both hierarchical and partition clustering by Partek Genomic Suite. All data is MIAME compliant and the raw data has been deposited in a MIAME compliant database GEO (accession number GSE28958).

Functional gene network analysis
The gene expression data derived from microarray analysis was used to generate functional and molecular networks through the use of IPA (Ingenuity Systems, CA). A fold-change cutoff of 2.0 was set to identify and assign the molecules to the Ingenuity's Knowledge Base. In these analyses, gene expression changes were considered in the context of physical, transcriptional or enzymatic interactions of the gene/gene products, and then grouped according to interacting gene networks at a particular point. The score assigned to any given gene network took into account the total number of molecules in the data set, the size of the network and the number of assigned network eligible genes/molecules in the data set at a given time point. The significance value and network score were based on the hypergeometric distribution and calculated with the right-tailed Fisher's exact test. The network score was the negative log of the p value.

Validation of salient genes differentially expressed in cluster analysis
Expression of selected genes from clustering analysis was confirmed by rt-PCR as previously described [65]. Briefly, extracted RNA was subjected to first strand cDNA synthesis using the Superscript III Reverse Transcriptase Kit (Invitrogen, CA).

Statistical Analysis
All time dependent analyses were performed on 15 animals per group. Microarray analyses were performed on cartilage extracted from three separate animals. The significance among the conditions in the microarray data was tested by Partek Genomic suite by ANOVA to render significantly regulated genes (p,0.05) during the progression of MIA at each time point. ANOVA with Tukey's HSD post hoc test by SPSS v 17 was used to determine the significance levels of rt-PCR data that include two additional independent samples per group to microarray-examined specimens (n = 5). p,0.05 was regarded as significant. Please see Table 2 for group description. A full list of these genes is given in Table S4. doi:10.1371/journal.pone.0024320.t005 Table 5. Cont. Please see Table 2 for group description. A full list of these genes is given in Table S5. doi:10.1371/journal.pone.0024320.t006  Figure 8B. Red, green, and white colors represent upregulation, downregulation and no regulation as compared to cont cartilage, respectively. The shading of each color represents fold change in gene expression; dark, higher changes and light lower changes. doi:10.1371/journal.pone.0024320.g006  Figure 8B. Red, green, and white colors represent upregulation, downregulation and no regulation as compared to cont cartilage, respectively. The shading of each color represents fold change in gene expression; dark, higher changes and light lower changes. doi:10.1371/journal.pone.0024320.g007        Table S1 for group description. (DOC) Movie S1 3606 mCT projection of the knee of Cont. Author Contributions