https://orcid.org/0000-0001-8272-3058
Figures
Abstract
To determine the transcriptomic changes seen in early- to mid-stage posttraumatic osteoarthritis (PTOA) development, 72 Yucatan minipigs underwent transection of the anterior cruciate ligament. Subjects were randomized to no further intervention, ligament reconstruction, or ligament repair, followed by articular cartilage harvesting and RNA-sequencing at three different postoperative timepoints (1, 4, and 52 weeks). Six additional subjects received no ligament transection and provided cartilage tissue to serve as controls. Differential gene expression analysis between post-transection cartilage and healthy cartilage revealed an initial increase in transcriptomic differences at 1 and 4 weeks followed by a stark reduction in transcriptomic differences at 52 weeks. This analysis also showed how different treatments genetically modulate the course of PTOA following ligament disruption. Specific genes (e.g., MMP1, POSTN, IGF1, PTGFR, HK1) were identified as being upregulated in the cartilage of injured subjects across all timepoints regardless of treatment. At the 52-week timepoint, 4 genes (e.g., A4GALT, EFS, NPTXR, ABCA3) that—as far as we know—have yet to be associated with PTOA were identified as being concordantly differentially expressed across all treatment groups when compared to controls. Functional pathway analysis of injured subject cartilage compared to control cartilage revealed overarching patterns of cellular proliferation at 1 week, angiogenesis, ECM interaction, focal adhesion, and cellular migration at 4 weeks, and calcium signaling, immune system activation, GABA signaling, and HIF-1 signaling at 52 weeks.
Citation: Donnenfield JI, Karamchedu NP, Proffen BL, Molino J, Fleming BC, Murray MM (2023) Transcriptomic changes in porcine articular cartilage one year following disruption of the anterior cruciate ligament. PLoS ONE 18(5): e0284777. https://doi.org/10.1371/journal.pone.0284777
Editor: Andre van Wijnen, University of Vermont College of Medicine, UNITED STATES
Received: January 31, 2023; Accepted: April 7, 2023; Published: May 3, 2023
Copyright: © 2023 Donnenfield et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Raw fastq files for the 1- and 4-week samples can be found in the data repository: https://doi.org/10.7910/DVN/KYSZH9 Raw salmon count files for the 52-week samples can be found in the data repository: https://doi.org/10.7910/DVN/KYSZH9 Appendix S1, S2, and S5 have been uploaded as Supporting Information files. Appendix S3 and S4 can be found in the data repository: https://doi.org/10.7910/DVN/KYSZH9 All other relevant data (e.g., figures) have been uploaded as Supporting Information files.
Funding: We gratefully acknowledge the support from the National Institutes of Health [National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) R01-AR056834, R01-AR065462, National Institute of General Medical Sciences (NIGMS) P30-GM122732 (Bioengineering Core of the COBRE Centre for Skeletal Health and Repair)], and the Lucy Lippitt Endowment. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official view of the National Institutes of Health.
Competing interests: Dr. Murray is a founder and equity holder, Dr. Proffen is a paid consultant and equity holder, and Dr. Fleming is a founder of Miach Orthopaedics, Inc, which was formed to upscale production of a scaffold for ACL restoration and is related to one of the ACL procedures described herein. Drs. Murray and Proffen maintain a conflict-of-interest management plan approved by Boston Children’s Hospital and Harvard Medical School. Dr. Fleming maintains a conflict-of-interest management plan with Rhode Island Hospital. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Introduction
Recent studies estimate there may be as many as 250,000 ACL tears per year in the United States [1, 2]. In addition to the multibillion-dollar toll these injuries put on the healthcare system, they can devastate the psychological and physical well-being of young athletes who comprise the majority of those injured [1, 3, 4]. In addition, the gold standard treatment for this condition, ACL reconstruction, has shown little-to-no benefit in terms of staving off one of the most common complications of ACL injury: posttraumatic osteoarthritis (PTOA) [4].
PTOA, which accounts for roughly 12% of all symptomatic osteoarthritis cases, develops within 15 years of ACL injury in about half of individuals [4, 5]. This, combined with the skew towards younger incidence, means the duration of disease burden per individual is generally longer than with idiopathic osteoarthritis. Despite the demographically wide-reaching and enduring aspects of PTOA, there remains no FDA-approved, disease-modifying treatment for it or even the more general form of osteoarthritis. In part, this is due to an incomplete understanding of how osteoarthritis develops.
In the past two decades, scientific advances in RNA sequencing (RNA-Seq) have provided a tool to study osteoarthritis. Gene expression of cartilage from humans undergoing total knee arthroplasty has been characterized in end-stage osteoarthritis [6–9], and preclinical studies have described the early gene expression changes upon induction of PTOA in animals (e.g., mice [10–12], dogs [13, 14], horses [15, 16], pigs [17]). However, none of these studies sequenced the cartilage beyond 16 weeks. Thus, even with these new tools, we still need mid-stage disease characterization.
To address this gap, we analyzed the transcriptomes of 90 cartilage samples from 78 Yucatan minipigs with healthy cartilage and PTOA cartilage at three different timepoints (1, 4, and 52 weeks) in disease development. This porcine model, and these animal cohorts in particular, has previously been shown to develop significant macroscopic cartilage damage at six months, with worsening at 12 months, even when treated with different methods of ligament stabilization [18, 19]. The current study applied updated differential gene expression analysis algorithms to determine the changes seen in early- to mid-stage PTOA development to help address the current knowledge gap in our field.
Methods
Study design
Seventy-eight adolescent Yucatan minipigs (Sinclair BioResources, Columbia, MO) were included in this study and came from two cohorts that have previously been reported on [17, 19]. Approval from the Brown University Institutional Animal Care and Use Committee was obtained prior to these studies (Protocol number: 1511000175), which were designed following the ARRIVE guidelines [20]. In the first (short-term) cohort, 36 Yucatan minipigs were allocated to undergo unilateral ACL transection (N = 36) followed by no further treatment (ACLT) or surgical repair (REPAIR) or reconstruction (RECON) of the ligament (N = 6 per group at 1 and 4 weeks). Six additional minipigs were allocated not to receive surgery (controls, N = 6). In the second (long-term) cohort, 36 Yucatan minipigs were randomized to ACLT (N = 12) or surgical treatment with RECON or REPAIR (N = 12 per group), followed by euthanasia and outcome assessment at 52 weeks (N = 36) after surgery. To establish controls for the long-term cohort, computerized randomization stratified by sex, surgical group, and knee laterality selected 4 contralateral knees from each surgical group to establish a total of 12 control samples to which surgical joints at 52 weeks would be compared against. Justification for the Yucatan minipig model, details for the IACUC-approved surgical procedures, animal husbandry, pain management have been reported and can be found in S1 Appendix along with the IACUC approval letter [17, 19]. After surgical operation, investigators were blinded to animal group assignments for all outcome assessments in both short-term and long-term studies.
Macroscopic cartilage evaluation and extraction
Immediately after euthanasia, knee joints were assessed for macroscopic damage to the articular cartilage surface by Osteoarthritis Research Society International (OARSI) guidelines for sheep and goat [21]. Cartilage assessment was performed by two orthopedics-trained physicians who arrived at a verbal consensus for each subject. Thus, there was no quantitative inter-observer analysis, but the same Yucatan minipig PTOA model was previously used by the group conducting the current study, and it was determined in Fleming et al. 2015 that the macroscopic damage score was reliable between independent examiners (intraclass correlation coefficient = 0.96) [22]. Damage to six articular surfaces, the medial femoral condyle, medial tibial plateau, lateral femoral condyle, lateral tibial plateau, femoral trochlea, and patella, were scored from 0 (normal) to 4 (large erosions down to subchondral bone). These scores were summed to create a macroscopic score sum that could range from 0 to 24. After cartilage assessment, osteochondral samples were extracted from the medial femoral condyle—with RNA isolation samples coming from the surface posterior to the frontal plane in the center of the medial femoral condyle. 4–8 5mm osteochondral biopsies were taken from each animal, and these samples were subsequently rinsed with water and separated from the attached bone. They were then flash frozen in liquid nitrogen and placed in a -80°C freezer until homogenization and RNA isolation.
Cartilage RNA-seq
Sample processing methods, from tissue extraction to RNA sequencing, for the 1- and 4-week samples can be found in Sieker et al. 2018 [17], while the 52-week samples were extracted de novo for this study. In brief, all samples were homogenized in 2 ml tubes (MP Biomedical) containing 500 μl of frozen TRIzol (Life Technologies) using a sterile drill bit while tubes were submerged in liquid nitrogen [17]. After one round of drilling, tubes received an additional 500 μl of liquid TRIzol, were flash-frozen in liquid nitrogen, and again homogenized using the drill bit. Total RNA was extracted using phenol-chloroform and purified using PureLink RNA Mini Kit (Life Technologies). Samples were then checked for purity with NanoDrop (Thermo Scientific) and checked for integrity using a combination of Agilent Tapestation High Sensitivity RNA Screen Tape and Agilent Tapestation RNA Screen Tape. For the 52-week samples, mean 260/280 absorbance ratios were 1.71 and 1.81 for controls and surgical groups, respectively. Mean 260/230 ratios were 1.22 and 1.46, and mean RIN integrity numbers (RINs) were 6.1 and 5.6, respectively (Table 1). All 52-week samples were then library-prepped using KAPA mRNA HyperPrep with RiboErase (Roche) and subsequently pooled and sequenced on a NovaSeq 6000 S2 Flow Cell with 100-bp paired-end reads (Biopolymers Facility, Harvard Medical School, Boston, MA). Newly generated fastq files for the 52-week samples, along with the legacy, short-term fastq files, were piped into FastQC version 0.11.9 to create individual sample reports that were then compiled using MultiQC version 1.12 [23, 24]. Summary statistics and histograms of mean quality values across each base position of the read were generated (S2 Appendix). All samples were contained within a Phred score range of 31.80 to 39.36. Using Salmon version 1.8.0, reads were then quasi-mapped, and transcript quantification files were generated [25]. For 1-, 4-, and 52-week surgical samples, the mean number (and %) of uniquely mapped reads was 14.2 million (93.3%), 10.1 million (81.7%), and 30.8 million (78.2%), respectively. The mean number of uniquely mapped reads for 1- and 4-week controls was 11.6 million (81.8%); for 52-week controls, it was 29.7 million (78.2%). Mapping employed the Sscrofa11.1 porcine genome, which was assembled by The Swine Genome Sequencing Consortium (SGSC) and hosted by Ensembl (http://ftp.ensembl.org/pub/current_embl/sus_scrofa/) [26]. STAR version 2.7.3a and Qualimap version 2.2.1 were also used to align and document alignment statistics, respectively, and the summary statistics from these programs can be found in the data availability repository. Differential gene expression analysis and functional pathway analysis were performed as detailed below.
Statistical analysis
Demographic and RNA quality characteristics for within-study comparisons (i.e., comparisons between 1-week and 4-week surgical subjects or between 52-week surgical subjects and 52-week controls) were analyzed with independent two sample t-tests. Cartilage score comparisons between short-term controls and long-term controls were analyzed with independent two sample t-tests. Four-surface scores were used for these comparisons as the short-term controls did not undergo assessment of the patella and trochlea like the long-term contralateral controls. One-way ANOVA tests were used to assess demographic differences across all subject groups (i.e., comparisons among 1-week, 4-week, and 52-week surgical subjects and controls). Randomization tests with 200 permutations were performed on differential gene expression lists to determine if the number of significant genes produced by a comparison was robust to sample label randomization. P-values ≤ 0.05 were considered statistically significant. A Gaussian mixture model, similar to that used by Donnenfield et al. 2022, was employed to soft cluster 52-week surgical subjects into two subpopulations based on macroscopic cartilage damage score sums [27]. Subjects were labeled as having either “less” or “more” damaged cartilage based on which subpopulation they had a greater than 50% chance of belonging to. “Less” damaged subjects were compared to “more” damaged subjects for differential gene expression analysis.
Differential gene expression analysis
Differential gene expression analysis was performed in R version 4.2.1 using DESeq2 on transcript quantification files produced by Salmon mapping [28, 29]. Gene outliers were handled using the DESeq2 default criteria for zero counts, extreme counts, and low mean normalized counts. P-values were adjusted for multiple testing using the Benjamini-Hochberg false discovery rate (FDR) with a value of < 0.05 as the cutoff for significance.
Functional pathway analysis
Overrepresentation analysis employed hypergeometric testing on the significantly differentially expressed gene lists and tested for representation of Gene Ontology (GO) terms [30]. These terms included biological processes, molecular functions, and cellular components, and the priority of reporting these terms was assigned in that order (i.e., if the biological process and molecular function terms were present, biological process terms were documented in the main text with molecular function terms and all other GO terms in S3 Appendix in the data availability repository). Because overrepresentation analysis is dependent upon statistically significant differentially expressed genes, it was only performed on comparisons that produced gene lists that were robust to randomization testing. Clusterprofiler was used to create dotplots and category netplots of the GO terms, and Revigo treemaps grouped GO terms by parent terms for high-level visualization [31, 32]. Dotplots were used in cases where many GO terms were present and highlighting significance among them was prioritized. Category netplots were used in cases where fewer GO terms were present, and there was an emphasis on showcasing influential genes. Gene Set Enrichment Analysis (GSEA) was performed to assess Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment independent of significant gene lists produced by DESeq2 [33, 34]. Signaling Pathway Impact Analysis (SPIA) was also used to evaluate significant KEGG pathways in a way that was more statistically robust than GSEA alone (i.e., SPIA incorporated a combination of hypergeometric testing and permutation testing) [35]. Complete lists of GSEA and SPIA results for all groups comparisons can be found in S4 Appendix in the data availability repository. An adjusted p-value cutoff of < 0.05 was used for all pathway analysis methods to determine significance.
Results
There were no differences in baseline age, weight, or sex among the 1-week and 4-week surgical subjects or their controls [17]. There were no demographic differences between 52-week surgical subjects or their controls (Table 2). There was also no difference in the macroscopic cartilage scores in the control knees for the short-term and long-term control groups (Table 2).
Differential gene expression analysis by experimental group
Expression differences between surgical groups and controls were more significant than those among surgical groups, where only REPAIR and RECON were significantly different at 4 and 52 weeks (Table 3; exact p-values shown in Table 1 in S5 Appendix). The top differentially expressed genes between surgical groups and controls at each timepoint are shown in Table 4. Based on the gaussian mixture model, 52-week surgical subjects with 6-surface cartilage damage scores ≥ 9 had a greater than 50% probability of belonging to the “more” damage group (N = 23), and surgical subjects with scores < 9 had a greater than 50% probability of belonging to the “less” damaged group (N = 13) (Fig 1). There was no significant difference in gene expression between these groups (Table 3). This lack of expression heterogeneity among surgical groups justified pooling surgical groups for timepoint comparisons.
Green and red lines represent the different gaussian distributions for “less” damaged and “more” damaged cartilage groups, respectively. The optimal separation was between macroscopic sum scores of 8 and 9.
Differential gene expression analysis by timepoint
There were significant differences in gene expression between controls and pooled surgical subjects at all timepoints (Table 5; exact p-values shown in Table 2 in S5 Appendix). At 1 week following surgical intervention, pooled surgical subjects featured 4206 differentially expressed transcripts compared to the no-surgery control group (Fig 2A). At 4 weeks, pooled surgical groups featured 5539 differentially expressed transcripts compared to the non-surgical controls (Fig 2B). And at 52 weeks, pooled surgical groups featured 314 differentially expressed transcripts compared to intact controls (Fig 2C). Across all timepoint comparisons, five genes were consistently differentially expressed: MMP1, POSTN, IGF1, PTGFR, and HK1 (Table 6).
Heatmaps of significantly differentially expressed genes for pooled timepoint ((A) 1-week, (B) 4-weeks, (C) 52-week) comparisons to controls. The red-blue color spectrum indicates the z-score for each gene. Surgical groups are labeled above columns, and subjects IDs are listed below columns.
Functional pathway analyses 1 week after surgery
At 1 week following surgery, pathways related to cell proliferation were the most enriched in the surgical groups compared to controls. This finding was supported by overrepresentation and upregulation of genes associated with the biological process cell cycle when surgical groups were compared to controls (Fig 3A). There was no overrepresentation of downregulated biological process GO terms for individual surgical group comparisons to controls. A treemap of biological process GO terms for pooled 1-week vs. controls was generated (Fig 4A). Cell cycle and nuclear division were the most notable parent terms, followed by vasculature development and terms related to the intranuclear mitotic arrangement and cellular movement.
The top 10 upregulated biological process GO terms based on overrepresentation analysis for (A) 1-week and (B) 4-week ACLT, RECON, and REPAIR compared to controls arranged by gene ratio (i.e., the number of genes related to the GO term divided by the total number of significant genes in the comparison). The number of genes represented by a term is indicated by the size of the circle, and the color represents the adjusted p value.
Treemap plots of upregulated biological process GO terms for (A) pooled 1-week vs. controls and (B) pooled 4-week vs. controls. Terms are grouped and colored according to unifying parent terms, and the amount of space a term occupies is proportional to gene set size and hypergeometric testing of overrepresented genes.
GSEA demonstrated that the most significantly enriched pathway was cell cycle when all three surgical groups were individually compared to controls (Fig 5A). Complete lists of the KEGG terms with significant GSEA results for these comparisons and an SPIA plot and table can be found in S4 Appendix in the data availability repository. Most pathways were involved in cell cycle, ECM-receptor interaction, and focal adhesion.
GSEA enrichment plots for (A) 1-week ACLT, RECON, and REPAIR compared to controls and (B) 4-week ACLT, RECON, and REPAIR compared to controls. The upper portion of each plot has lines which represent the log2 fold changes for each gene considered in the KEGG pathway. The lower portion of each plot shows the running total of these log2 fold changes, with a positive deflection representing upregulation in the first group listed in the comparison relative to the second group listed (i.e., ACLT vs. controls—first vs. second).
Functional pathway analyses 4 weeks after surgery
At 4 weeks after surgery, angiogenesis and cell adhesion pathways were the most enriched in the surgical groups compared to controls. This finding was supported by the overrepresentation and upregulation of genes related to the biological processes of angiogenesis, cell migration, and cell adhesion when compared to controls (Fig 3B). A treemap of biological process GO terms for pooled 4-week vs. controls revealed vasculature development, immune response, cell migration, actin filament-based process, and cell adhesion were the most represented parent terms (Fig 4B).
GSEA at 4 weeks demonstrated all three surgical groups featured cell adhesion or focal adhesion among their top 10 most significant terms as determined by adjusted p-value (Fig 5B). There was also upregulation of pathways related to inflammation and immune activation (e.g., Staphylococcus aureus infection). When individually assessed with SPIA, all 4-week surgical groups vs. controls featured upregulation of the terms ECM-receptor interaction, focal adhesion, PI3K-Akt signaling, Calcium signaling pathway, and complement-containing pathways. Concordantly, SPIA for the pooled 4-week vs. controls comparison featured upregulation of these terms within the top 10 most significantly represented pathways (S4 Appendix in the data availability repository).
REPAIR vs. RECON at 4 weeks following surgery featured a significant number of differentially expressed genes (Table 3). The top 5 biological process GO terms were all related to more expression in RECON than REPAIR of SMAD phosphorylation/signaling. These terms were influenced by the RECON group’s upregulation of INHBA, GDF5, TGFB1, and BMP2. GSEA featured increased expression in RECON relative to REPAIR of TGFβ-related terms, including cytokine-cytokine receptor interaction, rheumatoid arthritis, and TGF-beta signaling pathway (S4 Appendix in the data availability repository). Similarly, downregulation of TGFβ-related genes was involved in every SPIA-featured pathway: cytokine-cytokine receptor interaction, Hippo signaling pathway, TGF-beta signaling pathway, and osteoclast differentiation (GSEA plots available in S4 Appendix in the data availability repository).
Pooled 4-week surgical subjects were directly compared to pooled 1-week surgical subjects. Similar to above, cell adhesion, positive regulation of immune system process, and vasculature development were among the top upregulated biological process GO terms (S3 Appendix in the data availability repository).
Functional pathway analyses 52 weeks after surgery
At 52 weeks following surgery, surgical groups varied in their overrepresentation of GO terms across all three surgical groups compared to controls. ACLT vs. controls featured an overrepresentation of molecular function GO terms heparin binding, glycosaminoglycan binding, and sulfur compound binding—driven by POSTN, SFRP1, and THBS2 without clear directionality of expression (Fig 6A). There was also overrepresentation and mild upregulation of genes related to GABA receptor activity in ACLT samples relative to controls (Fig 6A). RECON vs. controls (and ACLT vs. controls) featured an overrepresentation of downregulated genes related to endoplasmic reticulum membrane regulation (Fig 6B). REPAIR vs. controls featured upregulation of biological process GO terms hexose metabolic process, monosaccharide metabolic process, response to hypoxia, and response to decreased oxygen levels (Fig 6C). Pooled 52-week vs. controls featured molecular function GO terms with mixed expression (e.g., glycosaminoglycan binding), cellular component GO terms with downregulation (e.g., endoplasmic reticulum membrane), and biological process GO terms with increased expression (e.g., positive regulation of immune system process) (Fig 6D). REPAIR vs. RECON at 52 weeks featured upregulated molecular function GO terms protein kinase binding and kinase binding dependent upon the genes NR4A3, MAPK6, and SOX9.
Category netplots for (A) 52-week ACLT vs. controls (molecular function GO terms), (B) RECON vs. controls (downregulated cellular component GO terms), (C) REPAIR vs. controls (upregulated biological process GO terms), and (D) pooled 52-Week vs. controls (upregulated biological process GO terms) based on overrepresentation analysis. GO term size is proportional to how many genes contribute to it, and fold change of expression is color coded.
GSEA at 52 weeks revealed that all three surgical groups (when compared to controls) featured upregulation of calcium signaling pathway among their top 10 most significant terms as determined by adjusted p-value. This pathway depends upon growth factor signaling, and each comparison featured several significantly upregulated growth factor genes. ACLT vs. controls featured upregulated neuroactive ligand-receptor interaction and GABAergic synapse. RECON vs. controls featured upregulation of pathways in cancer and PI3K-Akt, which were partly driven by growth factor signaling and ECM components of focal adhesion (S4 Appendix in the data availability repository). Amoebiasis was also upregulated, with contributions from laminin- and collagen-related genes. REPAIR vs. controls featured upregulation of the HIF-1 signaling pathway. This finding was driven by upregulation of HIF-, HK-, PDK-, VEGF-, and growth factor-related genes. Like ACLT vs. controls, neuroactive ligand-receptor interaction, GABAergic synapse, and pathways in cancer were upregulated for this comparison. Concordantly, SPIA for the pooled 52-week vs. controls featured upregulation of calcium signaling pathway, focal adhesion, PI3K-Akt signaling pathway, and complement and coagulation cascades (S4 Appendix in the data availability repository).
Discussion
The current study involved 78 porcine subjects and produced 90 transcriptomes across timepoints that extended up to a year. Differential gene expression analysis showed how transcriptomic differences between PTOA cartilage and healthy cartilage evolve as the disease progresses. It also displayed on a genetic level how different treatments modulate the course of PTOA following ACL transection. Specific genes (e.g., MMP1, POSTN, IGF1, PTGFR, HK1) were identified as being upregulated in surgical subjects across all timepoints. Functional pathway analysis of surgical groups compared to healthy cartilage revealed overarching patterns of cellular proliferation at 1 week, angiogenesis, ECM interaction, focal adhesion, and migration at 4 weeks, and calcium signaling, immune system activation, GABA signaling, and HIF-1 signaling at 52 weeks.
Overall, the number of differentially expressed genes increased (i.e., from 1 week to 4 weeks) and then substantially decreased (i.e., from 1 and 4 weeks to 52 weeks) when surgical groups were compared to controls. This trend suggests that, following injury, chondrocytes have a pronounced, reactive transcription phase that builds over the first few weeks of injury and then quiesces over a year. MMP1, POSTN, IGF1, PTGFR, and HK1 were still shared and upregulated among all pooled timepoint comparisons to controls, which suggests an enduring, typical response that all joints had following trauma.
MMP1 has repeatedly been shown to be upregulated in articular cartilage following surgical induction of PTOA in animal models—with expression positively correlated to time since injury and degree of cartilage damage for up to 12 weeks after surgery [13, 36, 37]. The current study demonstrates this continues through 52 weeks after surgery. Although much of the documentation of MMP-1 in the osteoarthritis literature is related to its role in cleaving type II collagen in articular cartilage, this matrix-degrading protease also has pivotal roles in modulating inflammatory cell recruitment, processing inflammatory cytokines, and activating VEGF to dilate pre-existing capillaries [38, 39]. The continued expression of MMP1 in injured cartilage suggests that these molecular pathways may remain active long after the initial insult.
Several studies have implicated periostin (encoded by POSTN) in the biology of cartilage degeneration [40]. Murine PTOA models have shown peak POSTN expression at 2–6 weeks [41] and diminished microscopic cartilage damage in POSTN knockouts [42]. In the current study, ACL transection in the porcine model featured consistent upregulation of POSTN at 1, 4, and 52 weeks relative to uninjured cartilage. This, along with evidence that POSTN expression is elevated in end-stage disease in human cartilage [43], suggests that chondrocyte periostin production may rise significantly in the early stages of cartilage compromise and continue at a lower (but still elevated) level in the subsequent months to years.
IGF-1 is known to have chondroprotective effects, such as stimulating neocartilage formation, promoting cell survival, and enhancing the integrity of ECM [44, 45]. A recent posttraumatic murine model has shown that subpopulations of resident knee joint macrophages constitute a significant source of this IGF1 expression, and chondrocytes (expectedly) have high expression of IGF1R in the 3–7 days following compressive ACL rupture [46]. In the current porcine study, expression of IGF1 was elevated at 7 days, and it remained elevated at 4 and 52 weeks following ACL transection—suggesting that cartilage may work to maintain a chondroprotective effort even a year after injury. Ongoing endeavors are to create IGF-1 delivery systems that increase this promising growth factor’s duration and cartilage penetration [47, 48]. The present findings suggest that exogenous agents could longitudinally work in parallel with native cellular efforts.
There is less documentation of PTGFR expression as it relates to the development of posttraumatic osteoarthritis [17]. There is evidence that the PGF2α receptor (encoded by PTGFR) is upregulated during chondrocyte differentiation and that it promotes ECM production (e.g., collagen type II and aggrecan) [49, 50]. Activation of this receptor has also been noted to hypertrophy chondrocytes, which mineralize the surrounding matrix and release VEGF [49]. In vitro testing has shown that chondrocyte PGF2α expression can be induced by shear stress, and administration of a COX-2 inhibitor can prevent this response—a finding that is concordant with the relief that NSAIDs often provide osteoarthritis patients, who have been found to have elevated PGF2α in serum and synovial fluid [51, 52]. Aside from being a potential source of pain, PGF2α may also play a substantial role in profibrotic changes and joint stiffening as osteoarthritis progresses [53]. Considering this evidence, yearlong PTGFR upregulation, as seen in the current study, might indicate an inflammatory mechanism induced by shear stress that promotes cartilage regeneration but eventually leads to cartilage mineralization. Further evidence that this mechanism becomes unbalanced over time is the significant expression of PTGFRN at 1 and 4 weeks but not at 52 weeks (i.e., PTGFR expression may be balanced by PTGFRN expression at earlier timepoints). Given its dynamic, opposed, and then unopposed expression in posttraumatic osteoarthritis, PTGFR is an understudied and promising target for future PTOA research.
Hexokinase 2 (HK2) has been a focus of osteoarthritis research in recent years. Still, another hexokinase isoform, HK1, has been less studied. Both isoforms have been found in cultured chondrocytes from osteoarthritis patients [54]. HK2 expression has been identified in fibroblast-like synoviocytes from osteoarthritis knees. It can be upregulated by TNF and hypoxia, which can increase the production of inflammatory cytokines if HK2 is overexpressed [55]. Both HK1 and HK2 serve as the first rate-limiting enzyme in glycolysis, implicating their involvement in metabolic derangements in osteoarthritis chondrocytes. There is evidence that chondrocytes undergo a metabolic shift from a state of regulatory rest to a state of high metabolic activity to meet the homeostatic needs required for cell survival [56]. Usually, this occurs in conditions of low oxygen tension, but it has been noted to occur in normoxic contexts such as cancer (i.e., Warburg effect) and degenerative/inflammatory diseases where HIF-1α plays a mediating role [56]. In end-stage osteoarthritis, chondrocyte hypoxia has been linked to chondrocyte apoptosis, cell senescence, dysregulated ECM, and increased abundance of ROS [57]. The dysregulation of ECM likely occurs because glucose provides fuel for cartilage development and is a precursor to glycosaminoglycans which are vital to ECM construction [58]. This intertwining of glucose processing and its HIF-1α-related regulation suggests that HK1 may be a principal agent in mediating changes in chondrocyte metabolism and its impacts on ECM degeneration. The current study found that HK1 was consistently upregulated at 1, 4, and 52 weeks relative to controls—evidence that challenges inconsistent findings in the literature, which have found HK1 levels in end-stage cartilage to be higher or no different than in healthy cartilage [55, 59].
Aside from identifying genes that are concordantly significant at 1, 4, and 52 weeks (Table 6), the current study also independently characterized the transcriptome of the PTOA porcine model at 52 weeks (Table 4C and Table 3 in S5 Appendix). This is a mid-range timepoint, which to the best of our knowledge, has yet to be explored in the literature—mainly due to logistical, physical, and financial hurdles that must be overcome to maintain a large animal model for over a year. There were 11 differentially expressed genes that were concordantly regulated across the top 100 gene lists of all 52-week surgical groups compared to controls (Table 4C). Four of these genes (A4GALT, EFS, NPTXR, ABCA3) have virtually no association with osteoarthritis or cartilage in the literature. They, therefore, represent new targets that have only been identified one year into PTOA development. Alternatively, six of the top concordant genes (HTR7, TRAK1, LAMA4, HS6ST1, PDE4A, GPNMB) at 52 weeks have been documented in prior osteoarthritis literature. Still, their identification in the current study implicates them at a timepoint in disease pathogenesis that was previously unknown [60–69].
When 52-week cartilage samples were stratified into less and more damaged groups, only 3 transcripts were differentially abundant. There are several potential reasons for the minor transcriptional difference between damage groups. Surgical groups may be mediated by different genetic signals, and pooling these groups may mix/obscure these different signals. For instance, REPAIR subjects with more damage may develop that damage in a way that is distinct from how ACLT subjects developed the same damage. Grouping these subjects may muddle the discernment of their individual mechanisms. However, if this were the sole explanation, one would expect a significant expression difference between REPAIR and ACLT, which there was not. Another possible reason is that cartilage differences between surgical groups and controls are substantial enough to produce a detectable transcriptomic signature, but cartilage differences among surgical groups are not significant enough to produce such a detectable signature—at least not at this timepoint. And lastly, the poor transcriptomic distinction between more and less damaged cartilage may indicate that we are dealing with a disease that has less transcriptional (and more mechano-physical) influence at 52 weeks than was previously thought.
Functional pathway analysis at 1 week was dominated by evidence of cell cycle upregulation. This suggests that the earliest phases of the chondrocyte reaction to joint injury are related to compensatory cellular proliferation, which has been associated with reduced chondrocyte apoptosis and increased secretion of cartilage matrix [70]. The unequivocal strength of this early proliferative signal in the cartilage of the current study stands in stark contrast to the varied role of the cell cycle pathway in the osteoarthritis literature, which has depicted the cell cycle as either not being enriched at 1 week in a murine PTOA model or possibly downregulated in a 1- to 9-week equine model [10, 16]. However, the latter evaluated synovial fluid for microRNA expression and did not characterize the cartilage or its cellular contents. An additional murine POSTN knockout, which protected against PTOA, showed downregulation of cell cycle pathways at 8 weeks after injury [42]. Human studies in end-stage disease have also found cell cycle pathways to be enriched, and their constituent genes were downregulated or had mixed expression directionality [71, 72]. Single-cell sequencing of end-stage human osteoarthritis cartilage further revealed a subpopulation of cartilage progenitor cells (CPCs) that has an enriched upregulation of cell cycle, chromosomal organization, and DNA replication pathways, but most cell types did not feature this enrichment [73]. The current and previous studies suggest that cellular proliferation may be a very early (i.e., ~1 week) phenomenon following the induction of osteoarthritis. An overwhelming compensatory surge in cellular division may occur following an injury that quiesces and then downregulates in the following weeks. And as the end-stage human studies have shown, the cartilage cellular profile eventually settles into a state where a select few chondrocytes remain reproductively active, but the majority do not.
Vasculature development, focal adhesion, and ECM-receptor interaction were upregulated at 1 week, and, along with cellular migration and immune response, became the predominantly upregulated pathways at 4 weeks. This is concordant with previous animal models, which have shown mild cartilage vascular development 7 days after disease induction and more prominent vascularization by 14 and 28 days [74]. Mechanistically, this angiogenesis is enabled by osteoclasts, which pave channels from the subchondral bone through the tidemark into normally avascular non-calcified cartilage [75]. These channels result in subchondral bone marrow being supplanted by VEGF-expressing fibrovascular tissue [76]. This process has been documented in rodents from 2 to 6 weeks following destabilization of the medial meniscus and in various stages of human disease when assessed in non-standardized post-mortem patients [77, 78]. Thus, these processes are not restricted to early- or late-stage disease. Along with this vascularization of non-calcified cartilage come exclusively perivascular sensory and sympathetic nerve fibers, which are believed to be a source of pain as the disease progresses [79]. On the other hand, upregulation of focal adhesion and ECM interaction pathways have been found in early- and end-stage osteoarthritis cartilage and—in the latter—may indicate collagen fibrosis and conversion of chondrocytes to fibroblasts [80, 81]. Additionally, end-stage enrichment of these pathways may indicate monocyte adhesion and migration into the synovium or differentiation (e.g., monocytes into macrophages or macrophages into osteoclasts) [82]. Therefore, enrichment of angiogenesis, adhesion, ECM, and cellular migration pathways at 4 weeks suggests a transition from chondrocyte replication at 1 week. Instead, there is a focus on the creation of osteoclasts that drive neovascularization, which lays the foundational infrastructure for future nerve pain, changing the cellular profile to one of fibrosis creation, and (through upregulation of MAPK signaling) stimulating the release of MMPs [83, 84]. Additionally, the upregulation of adhesion pathways in the current study may indicate some infiltration of immune cells through vascularized non-calcified cartilage networks. However, this infiltration has only been previously documented in synovium [76].
At 4 weeks following surgery, there was distinct enrichment of TGFβ-related pathways in RECON relative to REPAIR. Several studies have shown increased TGFβ signaling in cartilage associated with early-stage synthetic activity and chondroprotective mechanisms [85, 86]. This TGFβ upregulation is believed to be a temporary anabolic impulse that eventually loses out to catabolic pathways once TGFβ expression declines [86]. The entire mechanism of TGFβ-related signaling remains unresolved, but it does involve the interplay of SMAD family expression and transcription factors downstream of MAPK signaling [86]. There is evidence that SMAD2/3, like the TGFβ family constituents they interact with, decline in chondrocytes as osteoarthritis progresses while SMAD1/5/8 increase [87, 88]. The former stymies chondrocyte differentiation while the latter promotes it [87–89]. Increased TGFβ signaling in RECON compared to REPAIR subjects might suggest increased anabolic activity and less chondrocyte differentiation in the RECON group—something that contradicts the cartilage damage differences seen between the groups at 1 year. Perhaps, up until 4 weeks, the REPAIR subjects have incurred less damage than the RECON subjects and thus have less upregulation of compensatory mechanisms. Additionally, BMP2 expression is tied to TGFβ signaling, but there is conflicting evidence regarding its association with osteoarthritis progression [90, 91]. Similarly, GDF5, another member of the TGFβ family, has conflicting evidence; several studies relate GDF5 upregulation to ECM production and collagen type II production following osteoarthritis induction while others show reduced or absent expression in late-stage disease or soon after DMM [92–94]. Reduced GDF5 expression in late-stage disease or severely damaged cartilage may suggest that expression increases in early-stage disease but attenuates as degeneration progresses. In the current study, BMP2 and GDF5 were upregulated at 4 weeks in RECON compared to REPAIR subjects—associating these genes with worse cartilage outcomes following injury.
At 52 weeks following surgery, all surgical groups featured upregulation of calcium signaling and upregulation of growth factor constituent genes. This activation of calcium pathways may represent mechanosignaling or voltage-gated calcium regulation. The former requires non-motile cilia that protrude from chondrocytes and increase in length and number as osteoarthritis worsens [95, 96]. Mechanical perturbation of these cilia, modulated by growth factor signaling (i.e., TGFβ and BMPs), induces calcium influx into chondrocytes and plays a central role in downstream MAPK and β-catenin signaling [97]. Voltage-gated calcium signaling, on the other hand, has been much less studied in osteoarthritis, but there is evidence that administering antagonists to these channels could stymie osteoarthritis symptoms and the development of cartilage lesions following injury [14, 98]. Regardless of the predominant mechanism, the current study has shown that expression of genes related to calcium signaling is the most conserved genetic signal across all surgically induced PTOA subjects at 52 weeks. Additionally, the upregulation of γ-aminobutyric acid (GABA) receptor expression was prominently featured in pathway analysis for ACLT and REPAIR subjects. This finding, along with recent evidence that blocking GABA breakdown prevents cartilage degeneration after injury, suggests that GABAergic pathways remain relevant to osteoarthritis even one year after injury [99]. GABA signaling may also impact pain transmission as perivascular innervation forms in the non-calcified cartilage following injury. Lastly, there was upregulation of hypoxia pathways for REPAIR subjects at 52 weeks. There is an established connection between chondrocyte hypoxia and deleterious cartilage outcomes, and HIF-1α has been shown to be upregulated as a compensatory force for chondrocyte differentiation and survival [56, 100]. Nevertheless, some analyses of end-stage human cartilage have shown conflicting relationships between the expression of HIF-1-related genes and the degree of cartilage damage [8, 100]. The current study helps resolve this conflict by associating upregulation of the HIF-1 pathway in the surgical group with the best cartilage outcomes—strengthening the notion that HIF-1 has chondroprotective effects. When all 52-week surgical subjects were pooled together and compared to controls, there was a notable upregulation of humoral- and complement-mediated immune pathways. Immunoglobulins and immune complexes against cartilage, along with complement activation, have previously been identified in the cartilage of end-stage osteoarthritis patients [101, 102]. Complement likely comes from the surrounding synovial fluid (rather than from leakage from blood) [103]. The current study reaffirms the presence of these immune system compounds and implicates them at 52 weeks following joint trauma.
The current study is not without its limitations. One limitation is that the control group for the 1- and 4-week subjects had no prior surgery on either knee while the 52-week control group was composed of contralateral knees from subjects that had undergone surgery. This was done due to the prohibitive cost of humanely keeping animals alive through the 52-week timepoint only as controls. This decision was also supported by evidence that subchondral bone and cartilage thickness do not differ between surgery-naïve controls and contralateral knees of humans who have suffered ACL tears [104, 105]. Similar evidence of healthy contralateral knees following ACL injury has also been seen in quadrupeds, even 54 months after ACL transection [106]. Consistent with that prior quadruped study, we did not see significant differences in control groups macroscopically (Table 2), suggesting that at least macroscopic damage at this timepoint after ACL surgery was not different in the two control groups. While the 1-, 4-, and 52-week surgical and control samples all underwent the same RNA-Seq processing (which has substantially changed since prior reporting in 2018 of the short-term samples with an updated Sus scrofa genome, mapping algorithms, and gene expression analysis software) and utilized the same animal model, interventional procedures, and type of data (e.g., fastq files for RNA-Seq analysis), we recognize there may be differences in the gene expression in the two control groups that may confound results when normalized to controls. However, we felt this risk was mitigated as qualitative—not quantitative—transcriptomic comparisons were made between 1- and 4-week findings and 52-week findings. That is, log fold changes were not calculated between these studies, and differential gene expression analysis was not performed between short-term and long-term tissues. Rather, patterns were noted in the ways genes/pathways were differentially regulated within respective timepoints. This respects the boundary established by the different controls as well as the fact that short- and long-term samples were sequenced separately. 52-week samples were sequenced separately from 1- and 4-week samples because specimens came from studies that were performed at different times. Ideally, all samples would be sequenced together to allow for direct fold change comparisons to be made across all timepoints. Nevertheless—as mentioned above—sequencing samples separately still allows for qualitative comparisons to be made. Lastly, OARSI cartilage scoring guidelines were designed for goats and sheep but have been extended to pigs for this study.
The overwhelming majority of osteoarthritis basic science research uses samples from the end-stage disease, which has left a vacuum in the understanding of the early- and mid-stage [6–9]. The current study characterized the transcriptional profile of PTOA at 1, 4, and 52 weeks following injury—thereby creating a novel time course for understanding the trajectories of differential gene expression and molecular dynamics in the year following ACL disruption. This study revealed that transcriptomic differences between PTOA cartilage and healthy cartilage increase early and decrease as the disease progresses. It also showed on a genetic level how different treatments modulate the course of PTOA following ACL transection. Specific genes were identified as being upregulated in surgical subjects across all timepoints—reinforcing or, in the case of PTGFR and HK1, establishing their links to PTOA pathogenesis. At the 52-week timepoint, 4 genes (e.g., A4GALT, EFS, NPTXR, ABCA3) that were previously unassociated with PTOA were identified as being concordantly differentially expressed across all treatment groups when compared to controls. Functional pathway analysis of surgical groups compared to healthy cartilage also revealed overarching patterns of cellular proliferation at 1 week, angiogenesis, ECM interaction, focal adhesion, and migration at 4 weeks, as well as calcium signaling, immune system activation, GABA signaling, and HIF-1 signaling at 52 weeks.
Supporting information
S4 Appendix. MultiQC report for short-term cartilage.
https://doi.org/10.1371/journal.pone.0284777.s004
(XLSX)
S5 Appendix. Randomization testing and top 20 pooled vs. control transcripts.
https://doi.org/10.1371/journal.pone.0284777.s005
(DOCX)
S6 Appendix. MultiQC report for long-term cartilage.
https://doi.org/10.1371/journal.pone.0284777.s006
(HTML)
Acknowledgments
We thank the Bioinformatics Working Group at Boston Children’s Hospital and the BPF Genomics Core Facility at Harvard Medical School for their expertise and instrument availability that supported this work.
We sincerely thank our team members, Scott McAllister, Kaitlyn Chin, Kimberly Waller, Jillian Beveridge, Meggin Costa, Emma Fleming, and Jakob Sieker for assisting with surgical procedures and post-operative care, as well as technical support in the RNA isolation and sequencing. We also appreciate the support of the Brown University Center for Animal Resources and Education (CARE) veterinary technicians, Veronica Bouvier, Roxanne Burrill, and Pamela Norberg, for coordinating and assisting with the animal procedures. We sincerely thank the CARE veterinarians, Drs. James Harper and Lara Helwig, for their leadership and veterinary oversight for these studies.
References
- 1. CDC—Injury—ICRCs—CE001495 [Internet]. Centers for Disease Control and Prevention. Centers for Disease Control and Prevention; 2010 [cited 2022Dec19]. Available from: https://www.cdc.gov/injury/erpo/icrc/2009/1-R49-CE001495-01.html#print
- 2. Sanders TL, Maradit Kremers H, Bryan AJ, Larson DR, Dahm DL, Levy BA, et al. Incidence of anterior cruciate ligament tears and reconstruction: a 21-year population-based study. Am J Sports Med. 2016 Jun;44(6):1502–7. pmid:26920430
- 3. Carbone A, Rodeo S. Review of current understanding of post‐traumatic osteoarthritis resulting from sports injuries. J Orthop Res. 2017 Mar;35(3):397–405. pmid:27306867
- 4. Cheung EC, DiLallo M, Feeley BT, Lansdown DA. Osteoarthritis and ACL reconstruction—myths and risks. Curr Rev Musculoskelet Med. 2020 Feb;13(1):115–22. pmid:31894466
- 5. Wang LJ, Zeng N, Yan ZP, Li JT, Ni GX. Post-traumatic osteoarthritis following ACL injury. Arthritis Res Ther. 2020 Dec;22(1):1–8.
- 6. Ramos YF, den Hollander W, Bovée JV, Bomer N, van der Breggen R, Lakenberg N, et al. Genes involved in the osteoarthritis process identified through genome wide expression analysis in articular cartilage; the RAAK study. PLoS One. 2014 Jul 23;9(7):e103056. pmid:25054223
- 7. Pemmari A, Leppänen T, Hämäläinen M, Moilanen T, Vuolteenaho K, Moilanen E. Widespread regulation of gene expression by glucocorticoids in chondrocytes from patients with osteoarthritis as determined by RNA-Seq. Arthritis Res Ther. 2020 Dec;22(1):1–3.
- 8. Fisch KM, Gamini R, Alvarez-Garcia O, Akagi R, Saito M, Muramatsu Y, et al. Identification of transcription factors responsible for dysregulated networks in human osteoarthritis cartilage by global gene expression analysis. Osteoarthritis Cartilage. 2018 Nov 1;26(11):1531–8. pmid:30081074
- 9. Soul J, Dunn SL, Anand S, Serracino-Inglott F, Schwartz JM, Boot-Handford RP, et al. Stratification of knee osteoarthritis: two major patient subgroups identified by genome-wide expression analysis of articular cartilage. Ann Rheum Dis. 2018 Mar 1;77(3):423. pmid:29273645
- 10. Sebastian A, McCool JL, Hum NR, Murugesh DK, Wilson SP, Christiansen BA, et al. Single-cell RNA-Seq reveals transcriptomic heterogeneity and post-traumatic osteoarthritis-associated early molecular changes in mouse articular chondrocytes. Cells. 2021 Jun;10(6):1462. pmid:34200880
- 11. Chang JC, Sebastian A, Murugesh DK, Hatsell S, Economides AN, Christiansen BA, et al. Global molecular changes in a tibial compression induced ACL rupture model of post‐traumatic osteoarthritis. J Orthop Res. 2017 Mar;35(3):474–85. pmid:27088242
- 12. Sun H, Wen X, Li H, Wu P, Gu M, Zhao X, et al. Single-cell RNA-seq analysis identifies meniscus progenitors and reveals the progression of meniscus degeneration. Ann Rheum Dis. 2020 Mar 1;79(3):408–17. pmid:31871141
- 13. Fernandes JC, Martel-Pelletier J, Lascau-Coman V, Moldovan F, Jovanovic D, Raynauld JP, et al. Collagenase-1 and collagenase-3 synthesis in normal and early experimental osteoarthritic canine cartilage: an immunohistochemical study. J Rheumatol. 1998 Aug 1;25(8):1585–94. pmid:9712105
- 14. Boileau C, Martel‐Pelletier J, Brunet J, Tardif G, Schrier D, Flory C, et al. Oral treatment with PD‐0200347, an α2δ ligand, reduces the development of experimental osteoarthritis by inhibiting metalloproteinases and inducible nitric oxide synthase gene expression and synthesis in cartilage chondrocytes. Arthritis Rheum. 2005 Feb;52(2):488–500.
- 15. McCoy AM, Kemper AM, Boyce MK, Brown MP, Trumble TN. Differential gene expression analysis reveals pathways important in early post-traumatic osteoarthritis in an equine model. BMC Genomics. 2020 Dec;21(1):1–2.
- 16. Anderson JR, Jacobsen S, Walters M, Bundgaard L, Diendorfer A, Hackl M, et al. Small non-coding RNA landscape of extracellular vesicles from a post-traumatic model of equine osteoarthritis. Front Vet Sci. 2022 Aug;9:901269. pmid:36003409
- 17. Sieker JT, Proffen BL, Waller KA, Chin KE, Karamchedu NP, Akelman MR, et al. Transcriptional profiling of articular cartilage in a porcine model of early post‐traumatic osteoarthritis. J Orthop Res. 2018 Jan;36(1):318–29. pmid:28671352
- 18. Murray MM, Fleming BC. Use of a bioactive scaffold to stimulate anterior cruciate ligament healing also minimizes posttraumatic osteoarthritis after surgery. Am J Sports Med. 2013 Aug;41(8):1762–70. pmid:23857883
- 19. Karamchedu NP, Murray MM, Sieker JT, Proffen BL, Portilla G, Costa MQ, et al. Bridge-enhanced anterior cruciate ligament repair leads to greater limb asymmetry and less cartilage damage than untreated ACL transection or ACL reconstruction in the porcine model. Am J Sports Med. 2021 Mar;49(3):667–74. pmid:33534613
- 20. Percie du Sert N, Hurst V, Ahluwalia A, Alam S, Avey MT, Baker M. The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research. PLoS Biol. 2020;18(7):e3000410. pmid:32663219
- 21. Little CB, Smith MM, Cake MA, Read RA, Murphy MJ, Barry FP. The OARSI histopathology initiative–recommendations for histological assessments of osteoarthritis in sheep and goats. Osteoarthritis Cartilage. 2010 Oct 1;18:S80–92. pmid:20864026
- 22. Fleming BC, Proffen BL, Vavken P, Shalvoy MR, Machan JT, Murray MM. Increased platelet concentration does not improve functional graft healing in bio-enhanced ACL reconstruction. Knee Surg Sports Traumatol Arthrosc. 2015 Apr;23(4):1161–70. pmid:24633008
- 23. Andrews S. (2010). FastQC: a quality control tool for high throughput sequence data. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc
- 24. Ewels P, Magnusson M, Lundin S, Käller M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics. 2016 Oct 1;32(19):3047–8. pmid:27312411
- 25. Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware quantification of transcript expression. Nature methods. 2017 Apr;14(4):417–9. pmid:28263959
- 26. Yates AD, Achuthan P, Akanni W, Allen J, Allen J, Alvarez-Jarreta J, et al. Ensembl 2020. Nucleic Acids Res. 2020 Jan 8;48(D1):D682–8. pmid:31691826
- 27. Donnenfield JI, Karamchedu NP, Proffen BL, Molino J, Murray MM, Fleming BC. Predicting severity of cartilage damage in a post-traumatic porcine model: Synovial fluid and gait in a support vector machine. PLoS One. 2022 Jun 8;17(6):e0268198. pmid:35675298
- 28.
Team RC. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2020.
- 29. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014 Dec;15(12):1–21. pmid:25516281
- 30. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool for the unification of biology. Nature genetics. 2000 May;25(1):25–9.
- 31. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. The innovation. 2021 Aug 28;2(3):100141. pmid:34557778
- 32. Sayols S. rrvgo: a Bioconductor package to reduce and visualize Gene Ontology terms. 2020.
- 33. Subramaniana A, Tamayoa P, Moothaa VK, Mukherjeed S, Eberta BL, Gillettea MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005 Oct 25;102(43):15545–50. pmid:16199517
- 34. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000 Jan 1;28(1):27–30. pmid:10592173
- 35. Tarca AL, Draghici S, Khatri P, Hassan SS, Mittal P, Kim JS, et al. A novel signaling pathway impact analysis. Bioinformatics. 2009 Jan 1;25(1):75–82. pmid:18990722
- 36. Bluteau G, Conrozier T, Mathieu P, Vignon E, Herbage D, Mallein-Gerin F. Matrix metalloproteinase-1,-3,-13 and aggrecanase-1 and-2 are differentially expressed in experimental osteoarthritis. Biochim Biophys Acta. 2001 May 3;1526(2):147–58. pmid:11325536
- 37. Wu H, Du J, Zheng Q. Expression of MMP-1 in cartilage and synovium of experimentally induced rabbit ACLT traumatic osteoarthritis: immunohistochemical study. Rheumatol Int. 2008 Nov;29(1):31–6. pmid:18597092
- 38. Nissinen L, Kähäri VM. Matrix metalloproteinases in inflammation. Biochim Biophys Acta. 2014 Aug 1;1840(8):2571–80. pmid:24631662
- 39. Lee S, Jilani SM, Nikolova GV, Carpizo D, Iruela-Arispe ML. Processing of VEGF-A by matrix metalloproteinases regulates bioavailability and vascular patterning in tumors. J Cell Biol. 2005 May 23;169(4):681–91. pmid:15911882
- 40. Zhu D, Zhou W, Wang Z, Wang Y, Liu M, Zhang G, et al. Periostin: an emerging molecule with a potential role in spinal degenerative diseases. Front Med. 2021;8. pmid:34513869
- 41. Attur M, Yang Q, Shimada K, Tachida Y, Nagase H, Mignatti P, et al. Elevated expression of periostin in human osteoarthritic cartilage and its potential role in matrix degradation via matrix metalloproteinase‐13. FASEB J. 2015 Oct;29(10):4107–21. pmid:26092928
- 42. Attur M, Duan X, Cai L, Han T, Zhang W, Tycksen ED, et al. Periostin loss-of-function protects mice from post-traumatic and age-related osteoarthritis. Arthritis Res Ther. 2021 Dec;23(1):104. pmid:33832532
- 43. Chinzei N, Brophy RH, Duan X, Cai L, Nunley RM, Sandell LJ, et al. Molecular influence of anterior cruciate ligament tear remnants on chondrocytes: a biologic connection between injury and osteoarthritis. Osteoarthritis Cartilage. 2018 Apr 1;26(4):588–99. pmid:29391276
- 44. Zhang Z, Li L, Yang W, Cao Y, Shi Y, Li X, et al. The effects of different doses of IGF-1 on cartilage and subchondral bone during the repair of full-thickness articular cartilage defects in rabbits. Osteoarthritis Cartilage. 2017 Feb 1;25(2):309–20. pmid:27662821
- 45. Hossain MA, Adithan A, Alam MJ, Kopalli SR, Kim B, Kang CW, et al. IGF-1 Facilitates Cartilage Reconstruction by Regulating PI3K/AKT, MAPK, and NF-kB Signaling in Rabbit Osteoarthritis. J Inflamm Res. 2021 Jul 24;14:3555–68. pmid:34335042
- 46. Sebastian A, Hum NR, McCool JL, Wilson SP, Murugesh DK, Martin KA, et al. Single-cell RNA-Seq reveals changes in immune landscape in post-traumatic osteoarthritis. Front Immunol. 2022;13. pmid:35967299
- 47. Vedadgavami A, He T, Hakim B, Bajpayee AG. Cationic Peptide Carriers enable Long Term Delivery of Insulin-Like Growth Factor-1 to Suppress Osteoarthritis Induced Matrix Degradation. Arthritis Res Ther. 2022 Jul 20;24(1):172. pmid:35858920
- 48. Geiger BC, Wang S, Padera RF Jr, Grodzinsky AJ, Hammond PT. Cartilage-penetrating nanocarriers improve delivery and efficacy of growth factor treatment of osteoarthritis. Sci Transl Med. 2018 Nov 28;10(469):eaat8800. pmid:30487252
- 49. Kim J, Shim M. Prostaglandin F2α receptor (FP) signaling regulates Bmp signaling and promotes chondrocyte differentiation. Biochim Biophys Acta Mol Cell Res. 2015 Feb 1;1853(2):500–12.
- 50. Jakob M, Demarteau O, Suetterlin R, Heberer M, Martin I. Chondrogenesis of expanded adult human articular chondrocytes is enhanced by specific prostaglandins. Rheumatology. 2004 Jul 1;43(7):852–7. pmid:15113997
- 51. Guan PP, Ding WY, Wang P. The roles of prostaglandin F2 in regulating the expression of matrix metalloproteinase-12 via an insulin growth factor-2-dependent mechanism in sheared chondrocytes. Signal Transduct Target Ther. 2018 Nov 23;3(1):1–3. pmid:30510777
- 52. Basu S, Whiteman M, Mattey DL, Halliwell B. Raised levels of F2-isoprostanes and prostaglandin F2α in different rheumatic diseases. Ann Rheum Dis. 2001 Jun 1;60(6):627–31.
- 53. Bastiaansen‐Jenniskens YM, Wei W, Feijt C, Waarsing JH, Verhaar JA, Zuurmond AM, et al. Stimulation of fibrotic processes by the infrapatellar fat pad in cultured synoviocytes from patients with osteoarthritis: a possible role for prostaglandin f2α. Arthritis Rheum. 2013 Aug;65(8):2070–80.
- 54. Wang C, Silverman RM, Shen J, O’Keefe RJ. Distinct metabolic programs induced by TGF-β1 and BMP2 in human articular chondrocytes with osteoarthritis. J Orthop Translat. 2018 Jan 1;12:66–73.
- 55. Bao C, Zhu S, Song K, He C. HK2: a potential regulator of osteoarthritis via glycolytic and non-glycolytic pathways. Cell Commun Signal. 2022 Dec;20(1):1–10.
- 56. Mobasheri A, Rayman MP, Gualillo O, Sellam J, Van Der Kraan P, Fearon U. The role of metabolism in the pathogenesis of osteoarthritis. Nat Rev Rheumatol. 2017 May;13(5):302–11. pmid:28381830
- 57. Hu S, Zhang C, Ni L, Huang C, Chen D, Shi K, et al. Stabilization of HIF-1α alleviates osteoarthritis via enhancing mitophagy. Cell Death Dis. 2020 Jun 25;11(6):1–16.
- 58. Mobasheri A. Glucose: an energy currency and structural precursor in articular cartilage and bone with emerging roles as an extracellular signaling molecule and metabolic regulator. Front Endocrinol. 2012 Dec 17;3:153. pmid:23251132
- 59. Vázquez-Mosquera ME, Rego-Pérez I, Soto-Hermida A, Fernández-Moreno M, Fernández-Tajes J, Cortés-Pereira E, et al. New insights into the molecular basis of the metabolic alterations in the osteoarthritis (OA) disease. Osteoarthritis Cartilage. 2013 Apr 1;21:S170.
- 60. Lockwood SM, Bannister K, Dickenson AH. An investigation into the noradrenergic and serotonergic contributions of diffuse noxious inhibitory controls in a monoiodoacetate model of osteoarthritis. J Neurophysiol. 2019 Jan 1;121(1):96–104. pmid:30461363
- 61. Moazedi‐Fuerst FC, Gruber G, Stradner MH, Guidolin D, Jones JC, Bodo K, et al. Effect of Laminin‐A4 inhibition on cluster formation of human osteoarthritic chondrocytes. J Orthop Res. 2016 Mar;34(3):419–26. pmid:26295200
- 62. Feng T, Wu QF. A review of non-coding RNA related to NF-κB signaling pathway in the pathogenesis of osteoarthritis. Int Immunopharmacol. 2022 May 1;106:108607.
- 63. Chanalaris A, Clarke H, Guimond SE, Vincent TL, Turnbull JE, Troeberg L. Heparan sulfate proteoglycan synthesis is dysregulated in human osteoarthritic cartilage. Am J Pathol. 2019 Mar 1;189(3):632–47. pmid:30553836
- 64. Lyon M, Rushton G, Gallagher JT. The interaction of the transforming growth factor-βs with heparin/heparan sulfate is isoform-specific. J Biol Chem. 1997 Jul 18;272(29):18000–6.
- 65. Ai X, Do AT, Lozynska O, Kusche-Gullberg M, Lindahl U, Emerson CP Jr. QSulf1 remodels the 6-O sulfation states of cell surface heparan sulfate proteoglycans to promote Wnt signaling. J Cell Biol. 2003 Jul 21;162(2):341–51. pmid:12860968
- 66. Viviano BL, Paine-Saunders S, Gasiunas N, Gallagher J, Saunders S. Domain-specific modification of heparan sulfate by Qsulf1 modulates the binding of the bone morphogenetic protein antagonist Noggin. J Biol Chem. 2004 Feb 13;279(7):5604–11. pmid:14645250
- 67. Hopwood B, Tsykin A, Findlay DM, Fazzalari NL. Microarray gene expression profiling of osteoarthritic bone suggests altered bone remodelling, WNT and transforming growth factor-β/bone morphogenic protein signalling. Arthritis Res Ther. 2007 Oct;9(5):1–21.
- 68. Tenor H, Hedbom E, Häuselmann HJ, Schudt C, Hatzelmann A. Phosphodiesterase isoenzyme families in human osteoarthritis chondrocytes–functional importance of phosphodiesterase 4. Br J Pharmacol. 2002 Feb;135(3):609–18. pmid:11834608
- 69. Ball HC, Al-Adlaan AA, Hussein NJ, Cook B, Safadi FF. A CD44-dependent anti-inflammatory role of gpnmb in posttraumatic osteoarthritis. Osteoarthritis Cartilage. 2021 Sep 1;29:S26–7.
- 70. Li Y, Tu Q, Xie D, Chen S, Gao K, Xu X, et al. Triamcinolone acetonide-loaded nanoparticles encapsulated by CD90+ MCSs-derived microvesicles drive anti-inflammatory properties and promote cartilage regeneration after osteoarthritis. J Nanobiotechnology. 2022 Dec;20(1):1–20.
- 71. Sun J, Yan B, Yin W, Zhang X. Identification of genes associated with osteoarthritis by microarray analysis. Mol Med Rep. 2015 Oct 1;12(4):5211–6. pmid:26151199
- 72. Dunn SL, Soul J, Anand S, Schwartz JM, Boot-Handford RP, Hardingham TE. Gene expression changes in damaged osteoarthritic cartilage identify a signature of non-chondrogenic and mechanical responses. Osteoarthritis Cartilage. 2016 Aug 1;24(8):1431–40. pmid:26973327
- 73. Ji Q, Zheng Y, Zhang G, Hu Y, Fan X, Hou Y, et al. Single-cell RNA-seq analysis reveals the progression of human osteoarthritis. Ann Rheum Dis. 2019 Jan 1;78(1):100–10. pmid:30026257
- 74. Mapp PI, Avery PS, McWilliams DF, Bowyer J, Day C, Moores S, et al. Angiogenesis in two animal models of osteoarthritis. Osteoarthritis Cartilage. 2008 Jan 1;16(1):61–9. pmid:17659886
- 75. Mapp PI, Walsh DA. Mechanisms and targets of angiogenesis and nerve growth in osteoarthritis. Nat Rev Rheumatol. 2012 Jul;8(7):390–8. pmid:22641138
- 76. Walsh DA, McWilliams DF, Turley MJ, Dixon MR, Fransès RE, Mapp PI, et al. Angiogenesis and nerve growth factor at the osteochondral junction in rheumatoid arthritis and osteoarthritis. Rheumatology. 2010 Oct 1;49(10):1852–61. pmid:20581375
- 77. Su W, Liu G, Liu X, Zhou Y, Sun Q, Zhen G, et al. Angiogenesis stimulated by elevated PDGF-BB in subchondral bone contributes to osteoarthritis development. JCI Insight. 2020 Apr 23;5(8). pmid:32208385
- 78. Suri S, Gill SE, de Camin SM, McWilliams DF, Wilson D, Walsh DA. Neurovascular invasion at the osteochondral junction and in osteophytes in osteoarthritis. Ann Rheum Dis. 2007 Nov 1;66(11):1423–8. pmid:17446239
- 79. Bonnet CS, Walsh DA. Osteoarthritis, angiogenesis and inflammation. Rheumatology. 2005 Jan 1;44(1):7–16. pmid:15292527
- 80. Gardiner MD, Vincent TL, Driscoll C, Burleigh A, Bou-Gharios G, Saklatvala J, et al. Transcriptional analysis of micro-dissected articular cartilage in post-traumatic murine osteoarthritis. Osteoarthritis Cartilage. 2015 Apr 1;23(4):616–28. pmid:25545425
- 81. Li C, Luo J, Xu X, Zhou Z, Ying S, Liao X, et al. Single cell sequencing revealed the underlying pathogenesis of the development of osteoarthritis. Gene. 2020 Oct 5;757:144939. pmid:32640306
- 82. Shahrara S, Castro-Rueda HP, Haines GK, Koch AE. Differential expression of the FAK family kinases in rheumatoid arthritis and osteoarthritis synovial tissues. Arthritis Res Ther. 2007 Oct;9(5):1–0. pmid:17963503
- 83. Xia P, Shen S, Lin Q, Cheng K, Ren S, Gao M, et al. Low‐intensity pulsed ultrasound treatment at an early osteoarthritis stage protects rabbit cartilage from damage via the integrin/focal adhesion kinase/mitogen‐activated protein kinase signaling pathway. J Ultrasound Med. 2015 Nov;34(11):1991–9. pmid:26396170
- 84. Ding L, Guo D, Homandberg GA. The cartilage chondrolytic mechanism of fibronectin fragments involves MAP kinases: comparison of three fragments and native fibronectin. Osteoarthritis Cartilage. 2008 Oct 1;16(10):1253–62. pmid:18396067
- 85. van der Kraan PM, Glansbeek HL, Vitters EL, van den Berg WB. Early elevation of transforming growth factor-beta, decorin, and biglycan mRNA levels during cartilage matrix restoration after mild proteoglycan depletion. J Rheumatol. 1997 Mar;24(3):543–9 pmid:9058663
- 86. Davidson EB, Van der Kraan PM, Van Den Berg WB. TGF-β and osteoarthritis. Osteoarthritis Cartilage. 2007 Jun 1;15(6):597–604.
- 87. Zhao W, Wang T, Luo Q, Chen Y, Leung VY, Wen C, et al. Cartilage degeneration and excessive subchondral bone formation in spontaneous osteoarthritis involves altered TGF‐β signaling. J Orthop Res. 2016 May;34(5):763–70.
- 88. Yang X, Chen L, Xu X, Li C, Huang C, Deng CX. TGF-β/Smad3 signals repress chondrocyte hypertrophic differentiation and are required for maintaining articular cartilage. J Cell Biol. 2001 Apr 2;153(1):35–46.
- 89. Shen J, Li S, Chen D. TGF-β signaling and the development of osteoarthritis. Bone Res. 2014 May 27;2(1):1–7.
- 90. Davidson EB, Vitters EL, Van Der Kraan PM, Van Den Berg WB. Expression of transforming growth factor-β (TGFβ) and the TGFβ signalling molecule SMAD-2P in spontaneous and instability-induced osteoarthritis: role in cartilage degradation, chondrogenesis and osteophyte formation. Ann Rheum Dis. 2006 Nov 1;65(11):1414–21.
- 91. Aref-Eshghi E, Liu M, Harper PE, Doré J, Martin G, Furey A, et al. Overexpression of MMP13 in human osteoarthritic cartilage is associated with the SMAD-independent TGF-β signalling pathway. Arthritis Res Ther. 2015 Dec;17(1):1–8.
- 92. Mang T, Kleinschmidt‐Dörr K, Ploeger F, Lindemann S, Gigout A. The GDF‐5 mutant M1673 exerts robust anabolic and anti‐catabolic effects in chondrocytes. J Cell Mol Med. 2020 Jul;24(13):7141–50. pmid:32497388
- 93. Kania K, Colella F, Riemen AH, Wang H, Howard KA, Aigner T, et al. Regulation of Gdf5 expression in joint remodelling, repair and osteoarthritis. Sci Rep. 2020 Jan 13;10(1):1–1.
- 94. Garciadiego-Cázares D, Aguirre-Sánchez HI, Abarca-Buis RF, Kouri JB, Velasquillo C, Ibarra C. Regulation of α5 and αV integrin expression by GDF-5 and BMP-7 in chondrocyte differentiation and osteoarthritis. PLoS One. 2015 May 26;10(5):e0127166.
- 95. McGlashan SR, Cluett EC, Jensen CG, Poole CA. Primary cilia in osteoarthritic chondrocytes: from chondrons to clusters. Dev Dyn. 2008 Aug;237(8):2013–20. pmid:18330928
- 96. Wann AK, Zuo N, Haycraft CJ, Jensen CG, Poole CA, McGlashan SR, et al. Primary cilia mediate mechanotransduction through control of ATP‐induced Ca2+ signaling in compressed chondrocytes. FASEB J. 2012 Apr;26(4):1663–71. pmid:22223751
- 97. Hodgkinson T, Kelly DC, Curtin CM, O’Brien FJ. Mechanosignalling in cartilage: an emerging target for the treatment of osteoarthritis. Nat Rev Rheumatol. 2022 Feb;18(2):67–84. pmid:34934171
- 98. Daniilidis K, Georges P, Tibesku CO, Prehm P. Positive side effects of Ca antagonists for osteoarthritic joints—results of an in vivo pilot study. J Orthop Surg Res. 2015 Dec;10(1):1–4. pmid:25573427
- 99. Shen J, Wang C, Ying J, Xu T, McAlinden A, O’Keefe RJ. Inhibition of 4-aminobutyrate aminotransferase protects against injury-induced osteoarthritis in mice. JCI Insight. 2019 Sep 9;4(18). pmid:31534049
- 100. Wu L, Huang X, Li L, Huang H, Xu R, Luyten W. Insights on biology and pathology of HIF-1α/-2α, TGFα/BMP, Wnt/β-catenin, and NF-κB pathways in osteoarthritis. Curr Pharm Des. 2012 Jul 1;18(22):3293–312.
- 101. Jasin HE. Immune mediated cartilage destruction. Scand J Rheumatol. 1988 Jan 1;17(sup76):111–6. pmid:2908010
- 102. Cooke TD, Bennett EL, Ohno O. The deposition of immunoglobulins and complement components in osteoarthritic cartilage. Int Orthop. 1980 Nov;4(3):211–7.
- 103. Haseeb A, Haqqi TM. Immunopathogenesis of osteoarthritis. Clin Immunol. 2013 Mar 1;146(3):185–96. pmid:23360836
- 104. Kroker A, Bhatla JL, Emery CA, Manske SL, Boyd SK. Subchondral bone microarchitecture in ACL reconstructed knees of young women: A comparison with contralateral and uninjured control knees. Bone. 2018 Jun;111:1–8. pmid:29526780
- 105. Bhatla JL, Kroker A, Manske SL, Emery CA, Boyd SK. Differences in subchondral bone plate and cartilage thickness between women with anterior cruciate ligament reconstructions and uninjured controls. Osteoarthritis Cartilage. 2018 Jul;26(7):929–939. pmid:29678623
- 106. Brandt KD, Myers SL, Burr D, Albrecht M. Osteoarthritic changes in canine articular cartilage, subchondral bone, and synovium fifty-four months after transection of the anterior cruciate ligament. Arthritis Rheum. 1991 Dec;34(12):1560–70. pmid:1747141