ERK1/2 drives IL-1β-induced expression of TGF-β1 and BMP-2 in torn tendons

Diseased and injured tendons develop fibrosis, driven by factors including TGF-β, BMPs and CTGF. IL-1β and its signal transducer Erk1/2 are known to regulate TGF-β expression in animal tendons. We utilised tissues and cells isolated from patients with shoulder tendon tears and tendons of healthy volunteers to advance understanding of how inflammation induces fibrosis in diseased human tendons. ERK1/2 expression was reduced in torn (diseased) compared to healthy patient tendon tissues. We next investigated the fibrotic responses of tendon-derived cells isolated from healthy and diseased human tendon tissues in an inflammatory milieu. IL-1β treatment induced profound ERK1/2 signalling, TGFB1 and BMP2 mRNA expression in diseased compared to healthy tendon-derived cells. In the diseased cells, the ERK1/2 inhibitor (PD98059) completely blocked the IL-1β-induced TGFB1 and partially reduced BMP2 mRNA expression. Conversely, the same treatment of healthy cells did not modulate IL-1β-induced TGFB1 or BMP2 mRNA expression. ERK1/2 inhibition did not attenuate IL-1β-induced CTGF mRNA expression in healthy or diseased tendon cells. These findings highlight differences between ERK1/2 signalling pathway activation and expression of TGF-β1 and BMP-2 between healthy and diseased tendon tissues and cells, advancing understanding of inflammation induced fibrosis during the development of human tendon disease and subsequent repair.

Musculoskeletal diseases contribute a significant global burden 1,2 by causing pain and disability impacting upon activities of daily living and quality of life 3,4 . Tendon diseases account for one third of all musculoskeletal complaints to a general practitioner 5 with the supraspinatus tendon of the rotator cuff being frequently affected 6,7 . Diseased tendons are characterised by fibrosis, the inappropriate and often excessive production of extracellular matrix (ECM) proteins 8 . Fibrotic tendons are biologically and mechanically inferior compared to healthy tendons, predisposing the diseased tendon to tear 9,10 . Transforming growth factor beta (TGF-β) and its closely related growth factors including connective tissue growth factor (CTGF) and bone morphogenetic proteins (BMPs) have important roles in cell proliferation, differentiation and ECM metabolism 11,12 . These factors are implicated as mediators of fibrosis by studies of fibrotic diseases in other organs. TGF-β is regarded as the key fibrotic mediator, and its increased expression has been shown in animal studies of tendon overuse and injury healing 13 . CTGF has gained interest as a downstream regulator of TGF-β 13 . Dysregulated expression of BMPs such as BMP-2, BMP-7 and BMP-12 has also been reported in relation to tendon injury and healing 13 , implicating their involvement in tendon fibrosis.
Fibrosis develops as a result of dysregulated and persistent inflammation 14 , and an increasing number of studies support the regulatory roles of inflammatory cytokines in the development and progression of a tendon disease 3,15,16 . The effects of inflammatory cytokines on the expression of ECM-related genes and proteins have been investigated in both human and animal tendon-derived cells 17 . Tendon-derived cells of injured rat patellar tendons show an altered response to IL-1β treatment compared to those of un-operated tendons in vitro 18 . IL-1β is a pro-inflammatory cytokine known to take part in early stage human rotator cuff tendon disease 19 . Previous works from our group have also identified that tendon-derived cells from diseased human rotator cuff tendons show increased responsiveness to pro-inflammatory cytokines 19,20 and increased expression of IL-1 receptor in diseased human rotator cuff tendon tissues 21 . IL-1β activates mitogen-activated protein kinases (MAPKs) such as the extracellular signal-regulated kinase (ERK) 1/2 22 . Activation of the Erk1/2 signalling pathway has been implicated in a rat supraspinatus tendinopathy model 23 and in regulating Tgfb1 mRNA expression in rat Achilles tendon-derived cells 24 . However, to date no studies have investigated how inflammatory cytokines regulate fibrotic mediators including TGF-β, CTGF and BMP during the development of a human tendon disease and subsequent repair.
The aim of this study was to identify the mechanism by which inflammatory cytokines regulate fibrotic mediators including TGF-β, CTGF and BMP in tendon-derived cells of healthy and diseased patient tissues. We focused on large to massive tears for the diseased cohort as they represent end-stage disease and established pathology where the tendon has undergone fibrotic changes. We hypothesised that the ERK1/2 signalling pathway regulates IL-1β-induced expression of these fibrotic mediators in tendon-derived cells isolated from patients with end-stage torn rotator cuff tendons.

Results
Diseased tendon tissues show reduced levels of active ERK1/2. We investigated the differences in the expression of phosphorylated ERK1/2 between supraspinatus tendon tissues collected from healthy volunteers and patients with diseased tendons (large to massive tears) by immunohistochemistry (Fig. 1). Semiquantitative analysis of immunopositive staining showed a 3.6-fold decreased expression of phosphorylated ERK1/2 in diseased compared to healthy tendons (P = 0.0047).
Diseased tendon-derived cells show increased ERK1/2 signalling pathway activation by IL-1β treatment. Having shown diminished tissue expression of active ERK1/2 in diseased tendon tissues, we next investigated whether IL-1β treatment induces ERK1/2 signalling pathway activation in tendon-derived cells of healthy (healthy cells) and torn (large to massive tear) supraspinatus tendons (diseased cells) in an in vitro cell culture model. IL-1β treatment (5 ng/ml) for 30 minutes induced expression of phosphorylated ERK1/2 in both healthy and diseased cells determined by Western blotting. Representative images of the blots for healthy and diseased cells are shown in Fig. 2A. Semi-quantitative analysis of the blots indicated increased ERK1/2 signalling pathway activation in the diseased compared to healthy cells ( Supplementary Fig. 1).

ERK1/2 drives TGFB1 and BMP2 expression in IL-1β stimulated diseased tendon-derived cells.
Having shown activation of ERK1/2 signalling by IL-1β in tendon-derived cells, we sought to establish firstly whether IL-1β induces expression of the fibrotic mediators including TGF-β1, CTGF and BMPs (BMP-2, BMP-7 and BMP-12), and to determine whether induction of these fibrotic mediators was mediated via ERK1/2 signalling. IL-1β treatment (5 ng/ml) for 24 hours induced TGFB1 mRNA expression compared to vehicle control in diseased cells only (P < 0.0001) (Fig. 2B). This treatment also induced TGFBR1 and BMP2 mRNA expression compared to vehicle control in both healthy (P < 0.0001 and P < 0.0001, respectively) and diseased cells (P = 0.0011 and P < 0.0001, respectively) (Fig. 2C,D). IL-1β treatment suppressed CTGF mRNA expression compared to vehicle control in healthy and diseased cells (P = 0.0002 and P < 0.0001, respectively) ( Fig. 2E), but did not modulate TGFBR2 mRNA expression in both cells (Fig. 2F). mRNA expression of BMP7 and BMP12 were low-level regardless of treatment in both healthy and diseased cells. Having identified the regulatory role of IL-1β in the mRNA expression of TGFB1, TGFBR1, BMP2 and CTGF, we next investigated the consequences of ERK1/2 pathway inhibition on the IL-1β-induced expression of these fibrotic mediators using the ERK1/2 inhibitor PD98059 (8 μg/ml, pre-treatment for 1 hour) 25 . ERK1/2 pathway inhibition diminished IL-1β-induced TGFB1 mRNA expression in diseased cells (P = 0.0021) (Fig. 2B), but did not attenuate TGFBR1 or CTGF mRNA expression in healthy or diseased cells (Fig. 2C,E). ERK1/2 pathway inhibition partially suppressed the IL-1β-induced BMP2 mRNA expression in diseased but not healthy cells (P = 0.0019) (Fig. 2D). BMP-2 is a target gene of the NF-κB pathway [26][27][28][29] and similar effects of ERK1/2 pathway inhibition were observed in the IL-1β-induced mRNA expression of additional NF-κB target genes PTGES and COX in diseased cells (P = 0.0005 and P = 0.0232, respectively) but not healthy cells (Fig. 2G). Taken together these results demonstrate that the ERK1/2 signalling pathway drives the abnormal expression of TGF-β1 and BMP-2 in response to IL-1β treatment in diseased tendon-derived cells. (G) ERK1/2 inhibition partially suppressed IL-1β-induced mRNA expression of NF-κB signalling pathway target genes PTGES and COX2 in diseased cells only. Bars shown represent median. *Indicates significant difference to the respective vehicle control. # Indicates significant difference between with and without ERK1/2 inhibition. */# P < 0.05, **/## P < 0.01, ***P < 0.001, ****P < 0.0001. diseased tendon-derived cells, we next investigated the involvement of the ERK1/2 signalling pathway in the IL-1β-induced TGF-β1 and BMP-2 cell signalling activities by measuring their canonical SMAD signalling pathway target gene SERPINE1 30 and ID1 31 expression, respectively. Serpin E1 (serpin family E member 1) also known as plasminogen activator inhibitor 1 represses the plasmin-dependent protease activities and negatively regulates the breakdown of ECM collagens and fibrins resulting in fibrosis 32 . ID-1 (inhibitor of DNA-binding 1) is a pleiotropic transcription factor taking part in cell growth, differentiation and angiogenic activities 33 , and also counteracts TGF-β1, subsequent collagen production and hence fibrosis 34 . IL-1β treatment induced SERPINE1 and ID1 mRNA expression compared to vehicle control in healthy cells only (P = 0.0330 and 0.0063, respectively). However, when ERK1/2 signalling was inhibited, IL-1β elicited a significant induction in SERPINE1 mRNA expression compared to IL-1β treatment alone in both healthy and diseased cells (P = 0.0048 and 0.0005, respectively). In contrast, ERK1/2 pathway inhibition down-regulated IL-1β-induced ID1 mRNA expression in both healthy and diseased cells (P = 0.0004 and P < 0.0001, respectively) (Fig. 3). These results identify that IL-1β stimulated diseased cells show diminished TGF-β1 and BMP-2 SMAD cell signalling activities. The ERK1/2 pathway may have differential roles in IL-1β-induced TGF-β1 and BMP-2 SMAD cell signalling activities.

Discussion
Dysregulated repair and inflammation drive fibrosis in a diseased tendon 3,14,15 . Numerous animal studies of tendon overuse and injury healing have focused on TGF-β and its closely related CTGF and BMPs as mediators of fibrosis 13 . This study shows for the first time that diseased human tendon-derived cells exhibit dysregulated gene expression of fibrotic mediators in response to IL-1β treatment compared to healthy tendon-derived cells.
We investigated if ERK1/2 signalling pathway contributes to the IL-1β-induced TGFB1 and BMP2 expression in diseased cells. IL-1β treatment induced TGFB1 in diseased cells. Increased Erk1/2 signalling pathway activation in relation to tendon overuse and Erk1/2-driven Tgfb1 mRNA expression have been reported in studies using rat tendons 23,24 . In the current study, ERK1/2 signalling pathway inhibition negated the IL-1β-induced TGFB1 mRNA expression, and also partially suppressed the IL-1β-induced BMP2 mRNA expression in diseased cells. These results indicate the involvement of the ERK1/2 signalling pathway in end-stage tendon disease. ERK1/2 signalling pathway has been reported to suppress the expression of NF-κB pathway target genes in a monocytic cell line (THP-1) 35 . The cellular effects of IL-1β treatment in vitro may differ between cells isolated from clinical samples and cell lines. Furthermore, immunostaining in this study showed that the level of active ERK1/2 is significantly reduced in diseased compared to healthy tendon tissues. Reduced expression of fibrotic proteins in tissue samples from patients with chronic supraspinatus tendon tears has also been reported 36 . Increased responsiveness to inflammatory stimuli in diseased compared to healthy human tendon-derived cells 19,20 and increased tissue expression of their receptors in diseased tendons have been shown 21 . Whilst diseased tendon tissues showed reduced expression of TGF-β1 and BMP-2 proteins, increased IL-1β-induced expression of TGFB1 and BMP2 by diseased cells may be attributable to this capacity for increased responsiveness to inflammatory stimuli. Increased responsiveness to inflammatory stimuli may subsequently influence fibrotic healing during tendon repair. The nature of the IL-1β-induced expression of fibrotic mediators driven by the ERK1/2 signalling pathway in torn www.nature.com/scientificreports www.nature.com/scientificreports/ human tendons is likely to differ from the pathobiology of increased Erk1/2 signalling pathway activation noted in a rat model of supraspinatus tendinopathy 23 .
CTGF is a downstream regulator of TGF-β 49 , which has been implicated in other fibrotic diseases [50][51][52] . CTGF and IL-1β are both expressed in tendon overuse animal models 53 , but our results showed that IL-1β treatment suppresses CTGF mRNA expression in human tendon-derived cells in vitro. Similar cell culture experiments treating rat mesangial cells, human dermal fibroblasts, human healthy and diseased (dihydropyridine-induced gingival overgrowth) gingival fibroblasts with IL-1β all show inconsistent results in CTGF mRNA expression between studies [54][55][56] . The effects of IL-1β on CTGF mRNA expression is likely cell and context specific, and the interpretation of the results from this study should not be expanded beyond the comparison of human tendon-derived cells of healthy and torn tendons.
We acknowledge that there are limitations in this study. The cells utilised for the experiments may include a variety of cells such as tendon fibroblasts (tenocytes), tendon stem or progenitor cells owing to the cell isolation and culture method 57 . Hence, we collectively termed these cells as 'tendon-derived cells' 58,59 . The donors of diseased tendons were significantly older than those of healthy tendons, and other studies have indicated changes in the activity of ERK1/2 and NF-κB signalling pathways in relation to aging and senescence 60,61 . Increased responsiveness of diseased human tendon-derived cells to inflammatory stimuli 20 have also been related to aging 62,63 . There is a possibility that our results are confounded by age-related changes. Future studies addressing whether the diminished tissue expression of active ERK1/2 and the increased IL-1β-induced activation is related to aging or represents end-stage tendon disease is warranted. IL-1β activates other signalling pathways such as p38 and SAPK/JNK, and how they interact will also require further investigation. Moreover, IL-1β and the subsequent expression of fibrotic mediators may act on cell morphology, cytoskeleton, proliferation, vascularity and the ECM in the development of fibrosis, which were not addressed by this study. Healthy cells for the in vitro experiments were derived from an anatomically different hamstring tendon. There are limitations in using functionally distinct tendons as a comparator, but it is a positional tendon alike the rotator cuff that has been used as an ethically feasible source to obtain 'healthy' tendon samples 36,64 , and this would be preferable to using cadaveric tissue where the health status of the donor tissue is not known 21 . Furthermore, our study sampled diseased tendons tissues from patients with large to massive supraspinatus tears. Histological and molecular characteristics of torn tendon tissues may differ according to the chronicity of the tear 65 , and hence the findings of this study should only be applicable to the understanding of end-stage disease of large to massive supraspinatus tendon tear.
The findings of this study suggest that the profound ERK1/2 signalling pathway activation in response to IL-1β treatment regulates the increased TGFB1 and BMP2 mRNA expression in the diseased cells. Contrarily, canonical TGF-β1 and BMP-2 cell signalling activities measured by the mRNA expression of their target genes were diminished in IL-1β-induced diseased cells (Fig. 4). Tissue expression of active ERK1/2 is also reduced in diseased compared to healthy tendons. These findings highlight differences between ERK1/2 signalling pathway activation and expression of TGF-β1 and BMP-2 between healthy and diseased human tendon tissues and cells, advancing the understanding of fibrosis in diseased human tendons. Improved understanding of inflammation-induced fibrotic pathways should inform future strategies to therapeutically target tendon fibrosis.

Materials and Methods
Study approval. Healthy  Collection of human tendon tissues. All patients were recruited in shoulder referral clinics. The structural integrity of the supraspinatus tendon was confirmed by high-definition ultrasound examination 66 . The size of a torn supraspinatus tendon was measured intraoperatively in the longest diameter, and defined as small, medium, large or massive 67 . The exclusion criteria were prior history of diseases of the affected shoulder; other shoulder diseases that do not relate to the rotator cuff such as osteoarthritis, idiopathic frozen shoulder and fracture; bilateral shoulder problems; systemic inflammatory diseases; and symptoms related to the cervical spine or neurological disorders. Clinical diagnoses were not confirmed by pathology.
Diseased supraspinatus tendon tissues were obtained from the edge of the torn tendon as part of the debridement procedure of a repair surgery in patients operated for large to massive supraspinatus tears. Tissues from 7 male and 3 female patients aged 62.7 ± 7.8 years were used for explant culture. Tissues from 4 male and 2 female patients aged 64.8 ± 5.5 years were used for tissue protein expression analysis.
Healthy tendon tissues for explant culture were obtained from surplus hamstring tendons of 6 male and 4 female patients aged 24.6 ± 7.7 years going under autograft anterior cruciate ligament reconstruction surgery. (2019) 9:19005 | https://doi.org/10.1038/s41598-019-55387-1 www.nature.com/scientificreports www.nature.com/scientificreports/ Healthy supraspinatus tendon tissues for tissue protein expression analysis were obtained from 6 male and 1 female patients aged 34.7 ± 16.6 years by a validated biopsy technique using a long core biopsy instrument (Bard) under ultrasound guidance and local anaesthesia 68 . There were no adverse patient effects reported for this procedure.

Immunohistochemistry for phosphorylated ERK1/2 in healthy and diseased tendon tissues.
Fresh tendon tissue samples were immediately fixed in 10% neutral-buffered formalin and left for the fixative to infiltrate sufficiently. Fixed tissues were processed using ASP300 S tissue processor (Leica) and embedded in paraffin wax. Sections 5 μm thick were sliced using RM2125 RTS manual rotary microtome (Leica), collected onto glass slides, dried at 60 °C for 30 minutes and then 37 °C for 60 minutes. Tissue sections were baked at 60 °C for 60 minutes and then placed in an automated pre-treatment system (PT Link, Dako). Autostainer Link 48 (Dako) was used for antibody staining and EnVision FLEX DAB+ Chromogen System (Dako) for visualisation of its binding to the antigen using 3,3′-diaminobenzidne (DAB). Haematoxylin was used for counterstain. The procedures were performed according to the manufacturer's instructions. A rabbit polyclonal anti-phospho-ERK1/2 antibody (Abcam) was used at a dilution of 1:1000 after validation using sections of human tendon tissues. After staining, slides were taken into graded industrial methylated spirit (twice, 3 minutes each) for dehydration, then xylene (twice, 5 minutes each) for clearance and mounted (DPX Mounting medium, Fisher Scientific).
Ten to 20 images of each stained slides were captured using LSM 700 inverted brightfield microscope (Carl Zeiss MicroImaging) and Axiovision Rel. 4.8 imaging software (Carl Zeiss) at 100x magnification with oil immersion. Attention was paid so that the images did not overlap. Histological fibrotic changes (e.g. cell morphology, cellularity, disorganised ECM and vascularity) were confirmed in the diseased tissues, but were not quantified as this was not the objective of this study. The number of DAB and haematoxylin stained cells in each image was counted using ImageJ using a previously validated algorithm, with a manually set threshold for each dataset [69][70][71] . Semi-quantitative analysis of immunopositive cells by immunohistochemistry was carried out by calculating the percentage of DAB positive cells relative to the total number of cell nuclei imaged by haematoxylin. The details of the macros used are shown in Supplementary Method 1.
Isolation and culture of tendon-derived cells. Cells were isolated from tendon tissues and cultured based on protocols described previously 19,20 . In detail, fresh samples of tendon tissues were immediately incised into pieces 1 to 2 mm large 3 and cultured in media containing 50% foetal bovine serum (FBS; Labtech), Dulbecco's Modified Eagle Medium/Nutrient Mixture (DMEM F-12; Fisher Scientific) and 1% penicillin-streptomycin (PS; Lonza) at 37 °C and 5% CO 2 . After confirming by light microscopy that the cells had migrated out from the tissue after 7 to 10 days, culture media were changed to 10% FBS, DMEM F-12 and 1% PS until the cells were sub-confluent (approximately 80 to 90% confluent) for further passage or frozen storage. Cells were suspended in 90% FBS and 10% dimethyl sulfoxide (Sigma Aldrich) and stored at −80 °C for short-term (up to 3 months) and liquid nitrogen (approximately −200 °C) for long-term storage 72 . Our group has previously shown that these cells are negative for the leukocyte marker CD45 and vascular markers CD31 and CD34 by fluorescence-activated cell sorting and immunohistochemistry 20,73 . Culture media were changed twice a week. Cells cultured up to passage 2 or 3 were used for experiments 74 . Culture and treatment of tendon-derived cells with IL-1β. Healthy (N = 10) and diseased (N = 10) tendon-derived cells were seeded in 24-well plates at a density of 30,000 cell/ml (0.5 ml) and cultured to sub-confluence for the mRNA expression study. For analysis of cell signalling pathway activation by Western www.nature.com/scientificreports www.nature.com/scientificreports/ blotting, healthy (N = 3) and diseased (N = 3) cells were seeded in 6-well plates at a density of 60,000 cells/ml (2 ml) and cultured to sub-confluence. Following serum starving overnight, the cells were pre-treated with a small molecule inhibitor of the ERK1/2 pathway (PD98059; 8 μg/ml (30 μM)) for 1 hour prior to addition of human recombinant IL-1β (5 ng/ml; Merck Millipore) in serum-free medium or vehicle control (serum-free medium). The treatment dosage was determined by a validation experiment (Supplementary Method 3). Cells from both cohorts were treated for 24 hours and 30 minutes for the mRNA expression and cell signalling pathway activation analysis experiments, respectively. Vehicle only treatments served as experimental controls. mRNA expression analysis of cell lysates. mRNA was isolated from tendon-derived cells using RNeasy Minikit (Qiagen) with treatment by DNase I (Thermo Scientific). NanoDrop 1000 spectrophotometer (Thermo Scientific) was used to confirm the quality of the processed samples to be within 1.8 and 2.2 by the ratio of absorbance at 260 nm and 280 nm. mRNA was reverse transcribed into cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and a ribonuclease inhibitor (RNasin Plus RNase Inhibitor; Promega). The generated cDNA was further diluted with nuclease-free water resulting in 0.5 to 2.5 ng μl −1 19 .
Real-time quantitative polymerase chain reaction (RT-qPCR) was performed using a ViiA 7 Real-Time PCR System (Applied Biosystems) using SYBR Green Master Mix (Applied Biosystems) and validated primers from QuantiTect Primer Assays (Qiagen) or Primerdesign (Southampton, UK). The details of the primers used and their sequences are shown in Supplementary Table 1. Relative expression was calculated by the comparative CT method with a set threshold was of 0.15 for all genes 75 . All of the experiments and measurements were performed in technical duplicates, and the average CT values were used for analysis with a coefficient of variation between duplicates less than 0.05. β-actin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used as endogenous housekeepers, which the geometric mean of their CT values was used for analysis 76 . Cell signalling pathway activity analysis. Tendon-derived cells were treated with 1 mM sodium orthovanadate (Sigma-Aldrich) for phosphatase inhibition and then scraped in a cell lysis buffer that contains 0.5 mM sodium orthovanadate, 32 mM Tris base (Fisher Scientific), 50 mM (5%) glycerol (Sigma-Aldrich), 34.7 mM (1%) sodium dodecyl sulphate (SDS; Fisher Scientific) and 0.5% Protease Inhibitor Cocktail Set III (Merck Millipore). Protein separation was performed by electrophoresis by loading approximately 2.5 μg of protein per well to 10% polyacrylamide gel electrophoresis gels (BioRad). The amounts of the proteins loaded were measured by a bicinchoninic acid assay. Proteins were transferred from gels on to a 0.2 μm polyvinylidene difluoride membrane (BioRad) using a Western blot transfer system (BioRad). Molecular weight was indicated by Precision Plus Protein Dual Color Standards (BioRad). The details of the antibodies used and their working dilutions are shown in Supplementary Table 2. GAPDH was used as the endogenous housekeeper. Anti-rabbit IgG antibody was used as the secondary antibody. Proteins were detected by enhanced chemiluminescence (GE Healthcare Life Sciences) and visualised using Alliance 6.7 system and NineAlliance 4.7 17.00 (uvitec). Statistical analysis. Statistical analyses were performed using Prism 7 for Mac OS X Version 7.0c (GraphPad Software Inc.). D' Agostino and Pearson normality test was carried out to determine normal distribution for each data. Differences in patient demographics, mRNA and protein expression between healthy and diseased cells or tissues, or by treatments were compared by unpaired t test when the data were normally distributed. Mann Whitney test was used when the data were not normally distributed or the numbers were small (less than 8). Statistical significance was set at p < 0.05.

Data availability
The data supporting the findings of this study are available upon request from the corresponding author.