Article Text

Extended report
The effect of forced exercise on knee joints in Dio2−/− mice: type II iodothyronine deiodinase-deficient mice are less prone to develop OA-like cartilage damage upon excessive mechanical stress
  1. Nils Bomer1,2,
  2. Frederique M F Cornelis3,
  3. Yolande FM Ramos1,
  4. Wouter den Hollander1,
  5. Lies Storms3,
  6. Ruud van der Breggen1,
  7. Nico Lakenberg1,
  8. P Eline Slagboom1,2,4,
  9. Ingrid Meulenbelt1,4,
  10. Rik JL Lories3,5
  1. 1Department of Molecular Epidemiology, LUMC, Leiden, The Netherlands
  2. 2Integrated Research of Developmental Determinants of Ageing and Longevity (IDEAL), Leiden, Netherlands
  3. 3Laboratory of Tissue Homeostasis and Disease, Skeletal Biology and Engineering Research Centre, KU Leuven, Leuven, Belgium
  4. 4The Netherlands Genomics Initiative, sponsored by the NCHA, Leiden-Rotterdam, The Netherlands
  5. 5Division of Rheumatology, University Hospitals Leuven, Leuven, Belgium
  1. Correspondence to Department of Medical Statistics and Bioinformatics, Section Molecular Epidemiology, Leiden University Medical Center, LUMC Post-zone S-05-P, P.O. Box 9600, Leiden 2300 RC, The Netherlands; n.bomer{at}


Objective To further explore deiodinase iodothyronine type 2 (DIO2) as a therapeutic target in osteoarthritis (OA) by studying the effects of forced mechanical loading on in vivo joint cartilage tissue homeostasis and the modulating effect herein of Dio2 deficiency.

Methods Wild-type and C57BL/6-Dio2−/− -mice were subjected to a forced running regime for 1 h per day for 3 weeks. Severity of OA was assessed by histological scoring for cartilage damage and synovitis. Genome-wide gene expression was determined in knee cartilage by microarray analysis (Illumina MouseWG-6 v2). STRING-db analyses were applied to determine enrichment for specific pathways and to visualise protein–protein interactions.

Results In total, 158 probes representing 147 unique genes showed significantly differential expression with a fold-change ≥1.5 upon forced exercise. Among these are genes known for their association with OA (eg, Mef2c, Egfr, Ctgf, Prg4 and Ctnnb1), supporting the use of forced running as an OA model in mice. Dio2-deficient mice showed significantly less cartilage damage and signs of synovitis. Gene expression response upon exercise between wild-type and knockout mice was significantly different for 29 genes.

Conclusions Mice subjected to a running regime have significant increased cartilage damage and synovitis scores. Lack of Dio2 protected against cartilage damage in this model and was reflected in a specific gene expression profile, and either mark a favourable effect in the Dio2 knockout (eg, Gnas) or an unfavourable effect in wild-type cartilage homeostasis (eg, Hmbg2 and Calr). These data further support DIO2 activity as a therapeutic target in OA.

  • Osteoarthritis
  • Knee Osteoarthritis
  • Disease Activity
  • Treatment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Osteoarthritis (OA) is a prevalent, complex, chronic and disabling disease of articular joints, characterised by progressive destruction of joint cartilage, remodelling of the subchondral bone, formation of osteophytes and synovitis.1 ,2 It causes pain and disability to an increasing proportion of the population and is associated with the obesity pandemic, ageing of the population and not in the least by improved survival of patients with cardiovascular or oncological health problems. Altogether this imposes a large and growing social and economic burden.3 ,2 The development of novel therapeutic approaches is therefore urgently needed and should be based on insights into the underlying disease mechanisms.4 ,5 ,6 Several genetic studies identified robust signals for OA susceptibility,7 ,8 ,9 ,10 ,11 ,12 which suggest specific genes, involved in cartilage development and growth, to play a key role in the OA disease process.13 ,14 Notable examples of such OA susceptibility alleles are the C-variant of the rs225014 single-nucleotide polymorphism (SNP) located in the coding region of the deiodinase iodothyronine type 2 (D2) gene (DIO2)7 ,15 and the rs945006 SNP in the deiodinase iodothyronine type 3 (D3) gene (DIO3), another deiodinase with a counter-regulatory function for DIO2.15 For together, D2 and D3 primarily regulate the bioavailability of intracellular thyroid hormone in specific tissues such as the growth plate, but not systemically. The deiodinase type 2 protein (D2) catalyses the conversion of intracellular inactive thyroxine (T4) to active thyroid hormone (T3). During skeletal development, this conversion plays a critical role in the process of endochondral bone formation by facilitating terminal maturation of hypertrophic chondrocytes, subsequently leading to breakdown of the cartilage matrix and replacement by bone. This process is essential in skeletal development and growth, but loss of the chondrocyte's maturational arrested characteristics is considered deleterious for postnatal articular cartilage.16 ,17 DIO2 mRNA and D2 protein levels are highly upregulated in human osteoarthritic cartilage compared with healthy cartilage,18 ,19 ,20 suggesting that in disease DIO2 contributes to the loss of the highly specialised maturational arrested state of articular chondrocytes.16 Cartilage-specific overexpression of human DIO2 in rats was associated with increased damage to the articular cartilage in a surgical OA model. However, this was without clear evidence that hypertrophy of chondrocytes plays an essential role and rather pointing towards increased tissue-destructive enzyme activity and enhanced expression of interleukin-1 target genes.21 Upregulation of DIO2 expression in a human in vitro model resulted in a marked reduction of the capacity of chondrocytes to deposit extracellular matrix (ECM) components, including type II and type X collagen, while inducing OA-specific markers of cartilage matrix degeneration and mineralisation.17 In contrast, pharmacological inhibition of DIO2 increased the expression of collagens and aggrecan without a clear effect on hypertrophy or tissue-destructive enzymes. These accumulating data suggest that D2 inhibition and/or modulation may become a therapeutic target, but the in vivo impact of D2 loss of function in joint biology and disease remains largely unknown in particular at the molecular level.

In this study, we set out to study the molecular network of Dio2 in the healthy and challenged joint. We performed genome-wide expression analyses in ageing wild-type and Dio2−/− mice, including groups exposed to a moderately strenuous running regime. Our results indicate that Dio2 is effectively involved in specific gene networks that can be associated with OA and provide further insights into the complex molecular interactions involved in healthy and disease articular cartilage.

Materials and methods

Animal experiments

Dio2−/− mice were a kind gift of Dr V. Galton (Dartmouth Medical School, New Hampshire, USA)22 and were backcrossed onto the C57Bl/6 background.

Four-month-old to six-month-old male Dio2−/− (n=22) and wild-type mice (n=30) ran for 3 weeks 1 h/day, 5 days/week, at a speed of 11 m/min and with an inclination of 5°. For additional details, see the online supplementary methods.

Histological assessment of OA

Right knees were fixed overnight at 4°C in 2% formaldehyde, decalcified for 3 weeks in 0.5 M EDTA pH 7.5 and embedded in paraffin. Severity of disease was determined by histological scores on haematoxylin/eosin or Safranin O-stained sections (5 µm) throughout the knee (five sections at 100 µm distance). For additional details on histological assessment and statistical analyses, see the online supplementary methods.

RNA isolation

Snap frozen cartilage of the left knees was powderised using a Retsch Mixer Mill 200 under cryogenic conditions. RNA was isolated and washed using the RNeasy mini kit (Qiagen, Venlo, the Netherlands) according to the manufacturer's protocol. RNA quality was assessed using a Biocore lab-on-a-chip and quantity was assessed using a Nanodrop spectrophotometer. For additional details, see the online supplementary methods.

Microarray analysis

Complementary DNA synthesis, amplification, biotin labelling and hybridisation onto the microarrays were performed using the Ambion TotalPrep-96 RNA amplification kit (Life Technologies, Bleiswijk, the Netherlands) according to manufacturer's protocol. After hybridisation on Illumina MouseWG-6 v2 BeadChip microarrays (Illumina, Eindhoven, the Netherlands), the slides were scanned with the Illumina Beadscanner 500GX. For additional details on microarray handling and data analyses, see the online supplementary methods.

Pathway analysis and protein–protein interaction networks

Gene enrichment among the genes with significant differential expression was performed with STRING (Search Tool for the Retrieval of Interacting Genes/Proteins 9.1).23 Pathways with a p value ≤0.05 after false discovery rate (FDR) correction were considered significant. Enrichment in protein–protein interactions was also analysed using the STRING database. For additional details, see the online supplementary methods.

Quantitative RT-PCR assays (validation)

Validation of the microarray results was performed by quantitative real-time (RT)-PCR. In total, 500 ng of total RNA was processed with the First Strand cDNA Synthesis Kit according to the manufacturer's protocol (Roche Applied Science, Almere, the Netherlands) upon which cDNA was diluted five times. RT-qPCR measurements were performed on the Roche Lightcycler 480 II using Fast Start Sybr Green Master reaction mix according to the manufacturer's protocol (Roche Applied Science). For additional details, see the online supplementary methods.


Reduced severity of OA after forced running in Dio2−/− mice

To study the role of Dio2 in joint homeostasis, Dio2−/− and wild-type mice were studied in a forced-exercise setup and compared with non-running mice (figure 1). No striking developmental skeletal phenotype appears present in the Dio2−/− mice,24 and our observations in our mouse colony are in agreement with these findings. Wild-type and Dio2−/− mice displayed a similar running behaviour. As shown in figure 1, overall group analysis indicated a significant difference between wild-type and knockout mice with respect to cartilage damage (p=0.0006) and synovial hyperplasia (p=0.0536). No specific effects within wild-type or knockout mice of exercise were found, nor interaction between genotype and exercise. Sidak's multiple comparison test indicated that the effect was determined by the difference in genotype in the exercise group (p<0.0001 for cartilage damage—95% CI of the difference between the means (0.1820 to 0.5820) and p=0.0184 for synovial hyperplasia—95% CI of the difference between the means (0.03661 to 0.4627)) (figure 1).

Figure 1

Histological scoring of osteoarthritis. Comparing wild-type and Dio2−/− mice as well as the effect of forced running. (A) Frontal haematoxylin-Safranin O-stained sections of C57Bl/6 wild-type and Dio2−/− knees (medial) of mice subjected to a running regime (run) and control mice (no run) (magnification 10×). (B) Cartilage damage was increased in wild-type mice compared with Dio2−/− mice (two-way analysis of variance (ANOVA) p=0.0006) and (C) a similar trend was observed for synovitis (two-way ANOVA p=0.0536). None of the other comparisons (eg, ‘WT no run’ vs ‘KO no run’) were found significantly different. Data are shown as individual values, mean and 95% CIs. KO, knockout; WT, wild type.

Microarrays: differential expression in knee joints upon forced exercise

To identify genes responsive to the forced running regime in mice, genome-wide gene expression in knee cartilage was studied by microarray. Gene expression was detected before and after the running regime in interaction with the genetic background (wild-type and Dio2−/− mice). Figure 2 shows a schematic overview of the study strategy.

Figure 2

Overview of the study strategy. Following microarray analysis, data were statistically analysed. Secondary analyses using gene enrichment analysis and interaction analysis will lead to a set of genes that will be technically validated. FD, fold difference; KO, knockout; WT, wild type.

After quality control and normalisation, 20 872 of the 45 281 probes of the BeadChip array were used for analyses. Microarrays on human articular cartilage samples showed corresponding numbers for specific expression.25 In order to detect all articular cartilage genes that are responsive to the applied running regime, we performed differential expression analyses in three strata: total (A), Dio2−/− (B) and wild-type (C) group (figure 2). For each strata, significance was adjusted for multiple testing according to the ‘Benjamini and Hochberg’ method. In the total and wild-type strata, we independently detected respectively 1862 and 892 probes representing 1699 and 830 genes (558 overlapping) that were significantly differentially expressed between the forced exercise and the control group. In contrast, in the Dio2−/−-stratum, this comparison did not result in any significantly differentially expressed probes after multiple testing. Among the differentially expressed probes in the total and wild-type groups, we observed respectively 102 and 97 probes with a fold-change of 1.5 and higher that together consisted of 158 significant unique probes, representing 147 unique transcripts that were responsive to the running regime (see figure 2 and online supplement table S1). Notably, of the 158 differentially expressed probes, 31 were upregulated (20%) and 127 showed downregulation (80%). Among the 147 unique transcripts, we observed genes known for a potentially role in cartilage homeostasis and disease, such as proteoglycan 4, also known as lubricin (Prg4, 1.95-fold down; p=4.44×10−2),26 myocyte enhancer factor 2C (Mef2c, 1.9-fold down; p=3.05×10−2)27 ,28 and connective tissue growth factor (Ctgf, also known as Ccn2, 1.79-fold down; p=1.32×10−2).29

Technical validation of our microarray results was carried out by RT-qPCR in the discovery cohort (20 samples previously included in the microarray analyses). Genes to be validated were selected based on p value and fold difference (differential reaction on mechanical stress). Here, 19 out of the 20 genes tested showed similar effect sizes and direction as the original data; only Pfn1 did not show similar effect of expression (see online supplementary table S2).

Pathway analyses: protein–protein interactions and gene enrichment analyses

To identify functional connections between the genes that were found as differentially expressed (see online supplementary table S1), we determined the protein–protein interaction prediction using STRING-db. Analysing the 147 differentially expressed genes, representing 124 unique proteins, showed enrichment of interaction in the subset of proteins (p=1.61×10−4; 63 interactions; see online supplementary figure S1), indicating that the gene products that were identified are closely interacting as a response to the forced running regime.

Furthermore, we assessed whether the differentially expressed genes occurred more frequently in a specific pathways in mice. Gene enrichment analyses revealed enrichment for biological processes (GOTERM_BP_FAT) concerning, among others, ‘regulation of metabolic process’ (p=0.0063; GO: 0019222; N=46, eg, Calr, Igfbp5, Sox4 and Tcf4), ‘skeletal system development’ (p=0.00395; GO:0001501; N=12, eg, Ctnnb1, Ctgf, Sox4 and Mef2c), ‘anatomical structure development’ (p=0.039; GO:0048856; N=36, eg, Sox4, Notch3, Egfr, Ctnnb1 and Mef2c) and ‘regulation of response to stress’ (p=0.0171; GO:0080134; N=14; Setd8, Ankrd1, Mef2c, Egfr and Ctgf) with application of a FDR algorithm.

To connect thyroid hormone signalling with the effects of forced mechanical stress, we added genes involved in intracellular thyroid hormone signalling (Dio2, Thra, Thrb and Rxra) into the protein–protein interaction prediction. The network that is formed (figure 3) shows thyroid signalling genes/proteins to be incorporated into the large network, as previously seen (see online supplementary figure S1). Thyroid signalling was found to interact directly with the differential expressed genes through Ctgf (A) and Egfr (B), via thyroid receptor alpha (Thra) and retinoid x receptor (Rxra), both known factors in the development of OA.29 ,30 ,31

Figure 3

Gene networks in search tool for the retrieval of interacting genes. Screenshot of the protein–protein interactions (STRING-db) for the 147 genes differing >1.5-fold between ‘runners’ and controls, including thyroid signalling members; Dio2, Thra, Thrb and Rxra (encircled by the dashed line). Mayor hubs in the network are depicted by (A) ctgf, (B) egfr and (C) ctnnb1. Disconnected proteins are hidden.

Genes differentially expressed in Dio2−/− mice compared with wild-type mice upon a forced running regime

Since we found differences in differential expression patterns between knockout and wild-type mice upon the forced running regime, we assessed which of the 147 differentially expressed genes (see figure 1 and online supplementary table S1) showed significant interaction between running regime and genetic background based on nominal p values. In total, 29 probes, representing 29 genes, were found to be significantly differentially expressed in knockout mice compared with wild-type mice when undergoing a forced running regime (table 1). The significant differences in effect (β-values), as a result of the running regime, between the knockout and the wild-type group can be divided into three subgroups. Each group shows a different effect size based on ‘genotype’ for the 29 genes found to show a significant interaction. We identified four genes that showed no differential expression upon running in wild-type mice but are differentially expressed (downregulated) upon running in Dio2−/− mice (group 1; eg, Gnas and Rhbdl2). In total, 16 genes only showed differential expression in wild-type mice and not in the Dio2−/− mice (group 2; eg, Hmgb2, Calr and Lbh), whereas 9 genes showed significant differential effect sizes between wild-type and knockout mice, but having the same direction of effect (group 3; eg, Sox4 and Socs2). Depending on the gene expression pattern, the genes in the different groups could mark a favourable effect in the Dio2 knockout mice (group 1) or an unfavourable effect in wild-type cartilage homeostasis (group 2).

Table 1

Genes that show significant differential expression upon the running regime between wild-type and knockout mice (interaction)


Dio2-deficient mice showed less cartilage damage and reduced severity of synovitis in a treadmill running model of OA. The absence of significant differential gene expression between the running and no-exercise group in Dio2−/− mice suggests that degenerative pathways are not activated in this knockout strain despite the biomechanical burden that is imposed. These data provide novel support for inhibition of DIO2 as a therapeutic strategy in OA, particularly since no striking developmental skeletal phenotype appears present in the Dio2−/− mice (results not shown). Upon forced mechanical loading, wild-type mice showed clear signs of OA and differential expression of genes associated with the disease, supporting the use of forced running as an OA model in mice. A subset of the genes was found directly interacting with thyroid signalling through Thra, Rxra and Dio2, depicted in figure 3. This indicates the importance of thyroid hormone signalling as a regulatory system in response to stress and, when suppressed, for the maintenance of cartilage homeostasis.

With the differential expressed genes in the wild-type stratum and the combined knockout/wild-type stratum, pathway enrichment was found for expected biological processes, such as ‘skeletal system development’, but also for processes involved in ‘regulation of response to stress’. Notably, genes overlapping between these enriched processes as well as being ‘nodes’ in the recognised protein interaction networks (Egfr(B) and Ctnnb1(C); figure 3) are well known for their association with OA.29 ,30 ,32 These could, therefore, point at important modulators affecting the propensity to develop OA upon mechanical stress and as such at potential druggable targets for novel therapeutic approaches.

In contrast, when assessing gene expression in the knockout group alone, we found no significant differential gene expression upon severe mechanical loading after multiple testing adjustment. Taken together with the pathology observations, these data indicate that the repression of Dio2 is beneficial against the development of cartilage damage upon mechanical stress. This effect could theoretically also be caused by a difference in power between the different strata (wild-type and knockout). However, since the comparison made in knockout animals (6 running vs 3 controls) is comparable to the wild-type animals (8 running vs 3 controls), this is unlikely to explain the complete absence of a significantly differentially expressed gene in the knockout stratum.

Looking at the effect of Dio2 deficiency on gene expression, we found that the gene expression response upon exercise between wild-type and knockout mice was significantly different for 29 genes with nominal significance, of the total set of 147 differentially expressed genes. Genes categorised in group 1 (table 1), showing no differential expression upon forced running in wild-type mice, but are differentially expressed with nominal significance in Dio2−/− mice (E130112E08Rik, Gnas, Hist1h2an and Rhbdl2), could represent the respective favourable effects. Gnas, for example, is involved in skeletal development (GO:0001501) and complex skeletal disorders such as Albright Hereditary Osteodystrophy that can lead to early OA.33 Gnas upregulation was found directly regulating hypertrophic differentiation of growth plate cartilage in vivo.34 The knockout-specific downregulation of Gnas upon forced running could be a protective mechanism within the Dio2 knockout mouse.

We hypothesise that the 16 genes within group 2, only showing differential expression in wild-type mice and not in the knockout group (eg, Hmgb2 and Calr), represent the unfavourable damaging effects of forced running in wild-type mice. Hmgb2 was shown to be expressed at higher levels in human mesenchymal stem cells (MSCs) compared with human articular chondrocytes and the expression declined during chondrogenic differentiation of MSCs.35 The upregulation of Hmgb2 seen in wild–type mice upon forced running could be explained as a marker that changes in the differentiation status of the chondrocytes are occurring. A second interesting member of this group is calreticulin (Calr). Calr interacts with the glucocorticoid receptor. It may also interact with other steroid receptors or thyroid receptors in a similar way.36 Furthermore, Calr was shown to be involved in cartilage thinning of mandibular cartilage in a rat model that studied the effects of compressive mechanical loading.37 In our forced running experiment, we see a similar induction of Calr expression upon stress, but not in the Dio2 knockout mice, possibly giving the knockout mouse an advantage against the formation of OA-like degradation or cartilage thinning.

Nine additional genes showed significant differential effect sizes between wild-type and knockout mice, although the same direction of effect (eg, Sox4 and Socs2). Notable within this third group (see online supplementary table S3) is Sox4, which was previously found to be expressed very early during chondrogenesis,38 much earlier than the well-defined Sox5 and Sox6.39 Furthermore, it was shown that expression of Sox4 could be stimulated by adding physiological concentrations of human parathyroid hormone (PTH), indicating involvement of the PTH/PTHrP receptor.40 Here we show that Sox4 expression is influenced by the absence of Dio2. In the knockout mice, the expression of Sox4 is two times less downregulated upon running compared with the control group. How this influences the structural integrity of articular cartilage remains unknown. Other genes in the list have no straightforward connection to OA based on current literature.

In conclusion, in the current paper we find that Dio2 deficiency has a protective effect on the homeostasis of articular cartilage in the knee joints of mice undergoing a forced running regime. This is consistent with our earlier findings, showing that pharmacological inhibition of deiodinases in a human in vitro chondrogenesis model has a beneficial effect on the early formation and maintenance of articular cartilage ECM.17 It is therefore hypothesised that control of thyroid hormone signalling, both during development and adult cartilage maintenance, is essential to ensure normal bone and cartilage homeostasis, and that it could act as the master-switch that forces maturational arrested chondrocytes to reactivate the endochondral ossification process, leading to articular cartilage destruction. Our results show that interfering with intracellular thyroid hormone levels could be a powerful way to oppose the pathological events that are occurring in OA.


We acknowledge support by Treat∼OA and IDEAL, which are funded by the European Union's Seventh Framework Program (FP7/2007–2011) under respective grant agreement nos. 200800 and 259679. Furthermore, this work was supported by a grant from FWO Vlaanderen (Flanders Research Foundation) G.0828.11N and by the Dutch Arthritis Association. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • NB and FMFC shared first authors.

    IM and RJLL shared last authors.

  • Contributors NB, FMFC, PES, IM and RJLL conceived and designed the experiments. NB, FMFC, LS, RvdB and NL performed the experiments. NB, FMFC, YFMR, WdH, IM and RJLL analysed the data. NB, FMFC, IM and RJLL wrote the manuscript. All authors critically reviewed the manuscript.

  • Competing interests The authors declare no competing financial, personal, or professional interests.

  • Ethics approval All experiments were approved by the Ethics Committee for Animal Research (KU Leuven, Belgium).

  • Provenance and peer review Not commissioned; externally peer reviewed.