Article Text

Extended report
Blocking p38 signalling inhibits chondrogenesis in vitro but not ankylosis in a model of ankylosing spondylitis in vivo
  1. Kirsten Braem,
  2. Frank P Luyten,
  3. Rik J U Lories
  1. Laboratory for Skeletal Development and Joint Disorders, Division of Rheumatology, KULeuven, Leuven, Belgium
  1. Correspondence to Rik Lories, Laboratory for Skeletal Development and Joint Disorders, Division of Rheumatology, UZ Gasthuisberg – Rheumatology, Herestraat 49, 3000 Leuven, Belgium; rik.lories{at}


Objectives To investigate p38 mitogen activated protein kinase (MAPK) signalling in an in vitro model of bone morphogenetic protein (BMP) and transforming growth factor β (TGFβ)-induced chondrogenesis and in vivo, with specific attention to its potential role in ankylosing enthesitis.

Methods Human periosteum-derived cells (hPDCs) were cultured in pellets and stimulated with BMP2 or TGFβ1 in the presence or absence of a p38 inhibitor SB203580 or proinflammatory cytokines. Chondrogenic differentiation was evaluated using quantitative PCR. Male DBA/1 mice from different litters were caged together at the age of 8 weeks and treated with SB203580 in both a preventive and therapeutic strategy. The mice were evaluated for prospective signs of arthritis and the toe joints were analysed histologically to assess disease severity.

Results p38 inhibition by SB203580 and proinflammatory cytokines downregulated chondrogenic markers in pellet cultures stimulated by BMP2 or TGFβ1. In contrast, the in vivo experiments resulted in an increased clinical incidence of arthritis and pathology severity score, reflecting progression towards ankylosis in mice given SB203580.

Conclusion Inhibition of p38 inhibited chondrogenic differentiation of progenitor cells, showing that not only the SMAD signalling pathways and also alternative activation of MAPKs including p38 contribute to chondrogenesis. Such an inhibitory effect is not found in an in vivo model of joint ankylosis and spondyloarthritis. Increased incidence and severity of disease in preventive experiments and shifts in disease stages in a therapeutic experimental set-up suggest that specific inhibition of p38 may have deleterious rather than beneficial effects.

Statistics from


Ankylosing spondylitis (AS) is the best known entity of the spondyloarthritis (SpA) group of diseases. These common inflammatory and chronic arthritic diseases are characterised by enthesitis, synovitis and osteitis, and also by new cartilage and bone formation which can result in spine or joint ankylosis.1 Both inflammation and structural changes, especially in the spine and sacroiliac joints, contribute to symptoms and can lead to loss of function and permanent disability.2 Recent advances in SpA research and treatment have identified the proinflammatory cytokine tumour necrosis factor α (TNFα) as a key player. Targeted therapies against this cytokine have signficant benefits by limiting the symptoms caused by inflammation. Nevertheless, structural changes as determined by radiographic progression of disease do not appear to be influenced by anti-TNF strategies, suggesting that the processes of inflammation and ankylosis are at least partially independent.3,,6 These findings also highlight the need for alternative or complementary drugs.

The process of ankylosis originating from the enthesis at least partially recapitulates developmental endochondral bone formation in which bone morphogenetic proteins (BMPs), growth factors that are members of the transforming growth factor β (TGFβ) superfamily, play an important role.7 TGFβ family members can act by at least two intracellular signalling pathways. The canonical pathway induces ligand-dependent type I and type II receptor heterodimerisation resulting in phosphorylation of SMAD signalling molecules that bind common SMAD4. The SMAD complex translocates to the nucleus and interacts with multiple activators and repressors to regulate gene expression of target genes.8 The alternative cascade involves activation of different mitogen activated protein kinases (MAPKs).9 10 The induction of SMAD or MAPK pathways depends on the type I and type II receptor-complex assembly.11

MAPKs convert various extracellular stimuli into different cellular responses and mediate distinct effects on a wide array of biological processes. In these pathways, key effector enzymes such as p38, extracellular signal regulated-kinase (ERK) and c-Jun N-terminal kinase 1-3 (JNK) are part of a multistep cascade which is tightly regulated by phosphorylation and dephosphorylation.12 Among the various MAPK subfamilies, p38 kinase has attracted much attention in the last years.13 Besides its role in tissue differentiation, p38 is also involved in the catabolic actions of the proinflammatory cytokines interleukin 1 (IL-1) and TNFα.14 15 Studer et al demonstrated that blocking p38 signalling could inhibit the negative effects of IL-1 on chondrocyte proliferation, suggesting that p38 blockade helps to restore anabolic/catabolic imbalances which could be important in inflammatory bone diseases such as SpA.15

We previously identified the BMP pathway as a therapeutic target to inhibit radiographic disease progression in SpA based on experiments in the spontaneous arthritis model in male DBA/1 mice, which is characterised by ankylosing enthesitis. In these experiments, noggin, an extracellular BMP antagonist, was used, resulting in inhibition of both onset and progression of ankylosing enthesitis.7 Here we study the role of p38 activation in in vitro and in vivo cartilage and bone formation, since both SMAD and p38 activation is associated with chondrogenesis and osteogenesis and inflammatory cascades.16


Animal experiments

Male DBA/1 mice (Janvier) from different litters were mixed and caged together at the age of 8 weeks. In a first set of experiments, mice were injected daily at age 10–17 weeks with 1 mg SB203580 (LC Laboratories, Woburn, Massachusetts, USA) or 20 µl dimethyl sulfoxide (DMSO) vehicle control by intraperitoneal injection (n=18 and 15 animals per group, respectively). In a second set of experiments, mice were injected daily with SB203580 or DMSO after the first symptoms appeared and the mice were killed 3 weeks later (n=10 animals per group). Both hind paws were scored twice a week for clinical signs of arthritis and histologically evaluated as described previously.6 7 Healthy and arthritic paws of 21-week-old DBA/1 mice were used for gene expression analysis.

Cell culture

Mouse musculoskeletal progenitor C2C12 cells were cultured in complete culture medium consisting of high-glucose Dulbecco's Modified Eagle Medium/Glutamax (Invitrogen, Merelbeke, Belgium) supplemented with 10% fetal bovine serum (Gibco, Merelbeke, Belgium), antibiotic-antimycotic solution (100 units/ml penicillin, 100 µg/ml streptomycin and 0.25 µg/ml amphotericin B; Invitrogen) and sodium pyruvate (Gibco). After 3 h of serum starvation, cells were stimulated with 10 ng/ml TGFβ1 or 100 ng/ml BMP2 (R&D Systems Europe, Abingdon, UK) with or without prior treatment for 30 min with 10 µM SB203580 or DMSO vehicle control.

Human periosteum-derived cells (hPDCs) were isolated as described elsewhere17 and expanded in complete culture medium. Cell density was adjusted to 1.25×106 cells/ml in serum-free chemically-defined chondrogenic medium supplemented with sodium pyruvate and antibiotics.18 200 µl cell suspension was dispensed into the wells of a sterile 96-well V-bottom polypropylene microplate. The plate was centrifuged for 5 min at 500g. The pellets were stimulated with 10 ng/ml TGFβ1 or 100 ng/ml BMP2 and treated with 10 µM SB203580, 10 ng/ml human recombinant IL-1β (Thermo Scientific, Rockford, Illinois, USA) or TNFα (BioSource, Merelbeke, Belgium) from day 2 onwards. The medium was changed every 2–3 days. Duplicate pellets were harvested after 7 days for real-time PCR and western blot analysis.

Western blot analysis, immunohistochemistry and RNA isolation

Details are given in the online supplement.


For survival analysis, the Gehan–Wilcoxon test was used. For comparison between groups in the in vivo experiments, the non-parametric Mann–Whitney test was used. The results were considered statistically significant with two-sided p values <0.05.


In vitro stimulation of BMP and TGFβ signalling cascades and activity of SB203580

We determined the effect of p38 inhibition with chemical inhibitor SB203580 in vitro after BMP2 or TGFβ1 stimulation in BMP-responsive C2C12 cells. BMP2 addition caused SMAD1/5/8 phosphorylation and TGFβ1 stimulation resulted in activation of canonical SMAD3 (figure 1A,B). Both ligands also induced phosphorylation of p38 MAPK which was reduced to control levels by SB203580 pretreatment. Effective inhibition of downstream activation of MAPK-activated protein kinase 2 (MAPKAP2) was detected in SB203580-treated C2C12 cells. No BMP2 or TGFβ ligand-dependent ERK1/2 activation could be demonstrated. However, pretreatment with the p38 inhibitor increased ERK MAPK activation.

Figure 1

SB203580 inhibits transforming growth factor β (TGFβ)- and bone morphogenetic protein 2 (BMP2)-induced activation of p38. Serum-starved C2C12 cells were treated with 10 µM SB203580 or dimethyl sulfoxide (DMSO) prior to ligand stimulation. Samples were incubated with (A) 100 ng/ml BMP2 or (B) 10 ng/ml TGFβ for 30 min. After cell lysis, samples were immunoblotted against regulatory p-SMAD1/5/8 or p-SMAD3, p-p38, p-MAPKAP2 and p-ERK1/2 as well as glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as loading control.

Inhibition of chondrocyte-specific gene expression by p38 inhibition in periosteal pellet culture model

To study the effect of p38 modulation with SB203580 on endochondral bone formation in vitro, hPDCs were used in pellet cultures. The results shown in figure 2 are representative of four different experiments. After 7 days of culture, consistent upregulation of COL2 expression (median fold increase 68.2 (range 42.8–173.6)) and COL10 expression (median fold increase 30.6 (range 10.9–149.1)) was observed in TGFβ1-stimulated pellets (figure 2A). Also, consistent increases in RUNX2 (median 2.9 (range 1.6–4.4)) and SOX9 (median 2.1 (range 1.4–2.4)) were detected, confirming chondrogenic differentiation in this in vitro assay. ACAN (median 8.0 (range 1.0–24.8)) and ALP (median 2.5 (range 0.8–4.2) were also induced, but with variation in response between the different experiments. Treatment of the pellet cultures with SB203580 from the onset of the culture consistently reduced gene expression of COL2 (median decrease compared with vehicle-treated controls 93% (range 87–97%)), COL10 (58% (range 55–82%)) and ALP (87%). SOX9 (43% (range 0–67%)), RUNX2 (22% (range 0–52%)) and ACAN (8% (range 0–57%)) were also reduced compared with vehicle-treated controls, but not consistently. When hPDCs were cultured in the presence of BMP2, COL2 expression was consistently increased by day 7 (median fold increase 463.1 (range 53.1–873.1)). Expression of COL10 was still low at day 7 and even appeared to be downregulated, but steadily increased throughout the culture period (days 11 and 14) (data not shown). Compared with TGFβ1-stimulated pellets, BMP2 treatment resulted in a strong induction of SOX9 (8.1 (range 5.2–10.9)), RUNX2 (9.4 (range 7.8–11.0)), ACAN (212.0 (range 59.4–364.6)) and ALP (452.2 (range 247.3–657.1)). Inhibition of p38 with SB203580 resulted in a consistent reduction in BMP2-induced expression of COL2 (median decrease compared with vehicle-treated controls 92% (range 90–94%)), COL10 (63% (range 53–74%)) and ALP (59% (range 42–75%)). The effects on BMP2-induced RUNX2 (22% (range 0–46%)) and ACAN (17% (range 0–70%)) expression upon SB203580 treatment were not consistent. SOX9 expression was consistently increased in SB203580- and BMP-treated pellet cultures compared with vehicle ((18% (range 14–21%)).

Figure 2

Effect of p38 mitogen activated protein kinase (MAPK) inhibition on chondrogenic differentiation of human periosteum-derived cells (hPDCs) induced by transforming growth factor β (TGFβ) or bone morphogenetic protein 2 (BMP2). hPDCs were cultured in chemically-defined chondrogenic medium with TGFβ or BMP2 in the absence or presence of 10 µM SB203580. Real-time PCR analysis of chondrogenic markers SOX9, RUNX2, COL2, COL10, ALP and ACAN (A) and BMP2 (C) was performed on day 7. Values are expressed as fold change relative to control condition of day 2. (B) Protein lysates were subjected to western blot analysis and samples were immunobloted against p-SMAD3, p-SMAD1/5/8, p-p38, p-MAPKAP2 and p-ERK1/2 as well as for GAPDH as loading control. One of four representative experiments is shown.

TGFβ1-induced phosphorylation of SMAD3 and BMP2 stimulation resulted in SMAD1/5/8 phosphorylation in this pellet culture model (figure 2B). TGFβ1 stimulation of hPDC pellets also resulted in a small induction of p-SMAD1/5/8. This is supported by the fact that TGFβ1-stimulated pellet cultures showed increased BMP2 expression, suggesting an indirect effect on SMAD1/5/8 activation (figure 2C). Phosphorylated p38 was increased in TGFβ1-stimulated pellets compared with control conditions at day 7 (figure 2B). In contrast, BMP2 did not increase p38 activation. The amount of phosphorylated p38 appeared to be increased after SB203580 treatment in both the TGFβ1- and BMP2-stimulated pellet cultures. P-MAPKAP2, a p38 downstream target, was inhibited by SB203580 in both pellet culture systems, demonstrating the effect of the inhibitor. Again, SB203580 increased ERK activation in the BMP2-driven pellet system.

Effect of inflammation on chondrogenic differentiation in vitro

We performed an 84 gene PCR screening assay for cytokines and growth factors on arthritic hind paws compared with hind paws that appeared clinically normal. All of the 84 genes were detectable (see table S1 in online supplement) and 24 were changed more than twofold (18 upregulated and six downregulated) compared with control paws. These included upregulation of Bmp2, growth and differentiation factor-10 (Gdf10) and Bmp6, but also IL-1 and TNF family members (Il1f9, Tnfsf13 and Tnfsf15) and downregulation of IL-1 antagonists (Il1f10 and Il1rn) (figure 3A).

Figure 3

Effect of proinflammatory cytokines on chondrogenic differentiation of human periosteum-derived cells (hPDCs) induced by bone morphogenetic protein 2 (BMP2). (A) Inflammatory interleukin 1 (IL-1) and BMP signature in arthritic paws of 21-week-old male DBA/1 mice. PCR array analysis was performed on control and arthritic hind paws from 21-week-old male DBA/1 mice. Genes that exhibit a twofold or greater change in gene expression between healthy and arthritic paws are shown (n=2 and 3 mice per group, respectively). (B) hPDCs were cultured in chemically-defined chondrogenic medium with BMP2 in the absence or presence of 10 ng/ml tumour necrosis factor α (TNFα) or IL-1β. Real-time PCR analysis of chondrogenic markers RUNX2, COL2, COL10, ALP and ACAN was performed on day 7. Values are expressed as fold change relative to control condition on day 2.

Activation of MAPK signalling including p38 is a downstream event for BMP and TGFβ signalling and also for IL-1 and TNF families. To understand the impact of proinflammatory cytokines on chondrogenic differentiation of hPDCs induced by BMP2, pellets were treated with 10 ng/ml IL-1 or TNFα. Both cytokines resulted in an early increase in BMP ligand expression (BMP2, BMP4 and BMP6) (data not shown). However, the general effect on chondrogenic markers was a reduction of RUNX2, COL10, ACAN and ALP expression at day 7 (figure 3B). COL2 could not be detected in pellet cultures stimulated with BMP2 and treated with TNF or IL-1. p38 inhibition with SB203580 was not sufficient to rescue the effect of the proinflammatory cytokines (data not shown).

Figure 4

SB203580 administration before disease onset results in increased incidence of arthritis. (A) Immunohistochemistry of phosphorylated p38 (p–p38) in ankylosing enthesitis. Immunoreactivity for p–p38 is detected in entheseal fibroblast-like cells (arrowheads), prehypertrophic chondrocyte-like cells (arrows) and bone. Negative control staining using normal rabbit IgG. Magnification 10× (1) and 20× (2–3). (B) Cumulative incidence and (C) clinical severity score of dimethyl sulfoxide (DMSO)- (white triangles) and SB203580- (black triangles) treated mice (n=14 and 19 animals/group). Data are shown as mean ±SEM. (D) Daily SB203580 administration significantly increased pathological disease severity compared with DMSO-treated controls. *p<0.05, Mann–Whitney U test. Data are presented as individual cumulative score/mouse of all interphalangeal joints from the hind feet ◆ + median score — (n=14 and 19 animals per group).

Exacerbation of spontaneous arthritis in vivo by p38 inhibition in male DBA/1 mice

We further studied p38 MAPK signalling in vivo in the spontaneous arthritis model in DBA/1 mice characterised by ankylosing enthesitis. Immunohistochemistry was performed to identify cell populations that expressed phoshorylated p38 in spontaneous arthritis. Positive cells were detected in proliferating entheseal fibroblast-like cells, prehypertrophic chondrocyte like cells and bone (figure 4A). In contrast, negative control staining using normal rabbit IgG did not show any positive signal.

To examine the in vivo effect of p38 inhibition on the clinical incidence and severity of spontaneous arthritis, ageing male DBA/1 mice were injected daily with SB203580 or vehicle control from week 10 onwards. The mice were monitored twice a week for clinical signs of spontaneous arthritis until the age of 17 weeks. This early time point in the disease course was chosen to detect differences in incidence between the two groups. Animals treated with SB203580 showed an increased cumulative incidence compared with vehicle controls (figure 4B; p<0.05, Gehan–Wilcoxon test). However, the mean clinical severity score was not affected (figure 4C). Histomorphological analysis demonstrated an increased severity score for the p38 inhibitor-treated group compared with vehicle controls (figure 4D; p<0.05, Mann–Whitney test). When detailed histological analysis of pathological bone formation was performed, 10 out of 19 SB203580-treated animals showed cell proliferation at the enthesis (table 1).

Table 1

Effect of SB203580 treatment before and after disease onset on different features of ankylosing enthesitis

Progression towards cartilage and bone formation was seen in seven and four animals, respectively. In contrast, eight of the 14 control mice appeared normal. Proliferation was seen in four animals. Only one animal showed signs of cartilage differentiation and bone formation was documented in two animals. No ankylosis could be detected in any animals in either group, probably due to the short duration of the experiment (weeks 10–17).

The effect of SB203580 as a therapeutic strategy was evaluated by daily injections starting immediately after disease onset. Mice were evaluated further for 3 weeks. SB203580 treatment did not result in a statistically increased histomorphological severity score compared with control mice (data not shown). However, detailed analysis of the disease stages showed that cell proliferation was seen in nine of 10 mice treated with SB203580 but in only six of 10 mice in the control group (table 1). Progression to cartilage and bone formation in the p38 inhibitor-treated group was demonstrated in nine and eight animals, respectively, compared with six and five control animals. Three of 10 SB203580-treated mice showed ankylosis compared with only one control animal. These observations suggest that p38 inhibition may accelerate progression of pathological endochondral bone formation after initiation of the disease process.


Our earlier work showing the effects of noggin overexpression and haploinsufficiency in the spontaneous mouse model of ankylosing enthesitis in DBA/1 mice provided strong arguments for an important role of BMP signalling in the initiation and progression of ankylosing enthesitis.7 19 We have used SB203580 as p38 inhibitor to modulate the non-canonical BMP signalling pathway and demonstrated that both the SMAD signalling pathways and p38 contributed to chondrogenesis. Although inhibition of p38 activity affected both BMP and TGFβ-driven differentiation in the hPDC-derived in vitro model, such an inhibitory effect was not found in our in vivo model of joint ankylosis and SpA. Moreover, increased incidence and severity of disease in preventive experiments and shifts in disease stages in a therapeutic experimental set-up suggested that specific inhibition of p38 may have deleterious rather than beneficial effects.

The outcome of p38 inhibition in models and patients could be specifically determined by the pharmacokinetics of the compound used. p38 inhibitors have shown efficacy in animal models of arthritis by reducing joint inflammation. Both mice and rats treated with synthetic p38 inhibitors were protected from collagen-induced arthritis.20,,22 In contrast, clinical trials showed that highly selective p38 inhibitors had only modest clinical efficacy.23 SB203580 is a pyridinylimidazole highly selective for p38 α and β isoforms. By competing with ATP for binding to p38, SB203580 renders p38 inactive but does not necessarily interfere with its phosphorylation.24 In our experimental set-up with C2C12 cells, SB203580 was able to block the activation of p38 and MAPKAP2, a downstream effecter of p38, induced by BMP2 or TGFβ1. In the pellet cultures, exposure to the antagonist was necessary over a longer time period. Despite repeated renewal of the compound in the pellet culture, chronic SB203580 treatment resulted in increased phosphorylated p38 protein. Nevertheless, MAPKAP2 was clearly inhibited, demonstrating the successful inhibition of p38 activity. However, these in vitro data suggest that a relatively short half-life of compounds such as SB203580 may lead to compensatory upstream events which, in vivo, can result in net increased activity. The multipathway regulation of p38 activation is another factor that complicates the analysis and outcome of the inhibitory effects. Although anti-TNF therapy reduced disease severity in the human TNF transgenic murine model of rheumatoid arthritis, p38 activation was not reduced to the level of healthy mice upon treatment.25 Similar results are emerging in the clinic as a recent study has shown that, in psoriatic arthritis, p38 was expressed and activated in the synovium.26 Anti-TNF treatment did not change p38 phosphorylation status.

Off target effects can also provide an explanation for the apparent discrepancy between the in vitro and in vivo results. Although selectivity of SB203580 for p38 α and β isoforms has been shown, recent studies have demonstrated that the compound can also inhibit other protein kinases with similar or even greater potency.27 Nevertheless, no off target effects of SB203580 on WNT signalling, another pathway involved in ankylosis, were detected as demonstrated by western blot analysis (results not shown). Diarra et al showed that TNF induced the expression of DKK1, a secreted negative regulator of the WNT pathway, through p38 MAPK.28 We have no evidence that DKK1 plays a role in the spontaneous model as serum levels in both healthy and arthritic mice are below the level of detection. Local expression levels are low and do not differ between healthy and arthritic paws (data not shown). Of note, p38 inhibition in the human TNF transgenic mouse model reduced inflammation and bone loss but had no effect on osteoproliferation.29

The different steps in chondrogenesis are associated with the expression of stage-specific markers. In our in vitro system both BMP2 and TGFβ treatment enhanced the expression of chondrocyte differentiation markers including COL2, COL10, ACAN and ALP, and was associated with the activation of SMAD and p38 pathways. SOX9 and RUNX2 are considered as master regulators of chondrogenic and osteogenic differentiation. The lack of effect of SB203580 on SOX9 expression induced by BMP2 and the relatively low inhibitory effect on RUNX2 expression induced by TGFβ1 and BMP2 suggested that these factors are induced preferentially by the canonical SMAD rather than the alternative MAPK pathway.

We have previously shown that inflammatory changes in the spontaneous arthritis model are short-lived.30 PCR array analysis showed a proinflammatory profile. Undetectable cytokine levels in serum samples of mice with arthritric paws (data not shown) suggested that the pathological events resulting in entheseal inflammation and new tissue formation are local rather than systemic phenomena. The IL-1 expression signature is of particular interest since there is evidence of an IL-1 cluster being genetically associated with AS, including IL-1 receptor antagonist, IL1f10 and IL1f9.31 Tnfsf15 was differentially regulated as well, although the expression levels were rather low. However, this could be of interest since Zinovieva et al recently showed that this gene is associated with AS.32

In summary, our study demonstrates the importance of non-SMAD driven BMP and TGFβ signalling in chondrogenic differentiation but failure to detect a protective effect in vivo. Our finding that p38 inhibition aggravated disease also highlights the difficulties in defining an appropriate strategy to inhibit progression of ankylosis in patients with AS and related spondyloarthritides.


The authors thank Mrs Inge Derese, Ann Hens, Lies Storms and Jenny Peeters for technical assistance.


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:

    • Web Only 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.
    • Web Only 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.


  • Funding This work was supported by a GOA grant from KULeuven and FWO grant G.0213.07. KB is the recipient of a PhD fellowship from the Institute for Science and Technology (IWT Vlaanderen).

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

  • Competing interests None.

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

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.