Inhibition of ERK 1/2 kinases prevents tendon matrix breakdown

Tendon extracellular matrix (ECM) mechanical unloading results in tissue degradation and breakdown, with niche-dependent cellular stress directing proteolytic degradation of tendon. Here, we show that the extracellular-signal regulated kinase (ERK) pathway is central in tendon degradation of load-deprived tissue explants. We show that ERK 1/2 are highly phosphorylated in mechanically unloaded tendon fascicles in a vascular niche-dependent manner. Pharmacological inhibition of ERK 1/2 abolishes the induction of ECM catabolic gene expression (MMPs) and fully prevents loss of mechanical properties. Moreover, ERK 1/2 inhibition in unloaded tendon fascicles suppresses features of pathological tissue remodeling such as collagen type 3 matrix switch and the induction of the pro-fibrotic cytokine interleukin 11. This work demonstrates ERK signaling as a central checkpoint to trigger tendon matrix degradation and remodeling using load-deprived tissue explants.

www.nature.com/scientificreports/ upstream of the differentially expressed genes (DEG). E2K algorithm predicted ~ 100 upstream protein kinases that are likely responsible for regulating DEG. Mapping the identified kinases into Kinome phylogenetic tree showed that the majority of these kinase hits clustered around the CMGC kinases, which is a large kinase group including the mitogen-activated protein kinase (MAPK) family (Fig. 1B). The most significantly enriched kinases are depicted in red color and the MAPK members ERK 1/2 were among the 5 most significantly enriched kinases ( Fig. 1B,C, Supplementary table 1). Taking this into consideration, we reasoned that ERK 1/2 may be centrally involved in the niche-dependent degradation of tendon fascicles. Next, we experimentally validated ERK 1/2 predictions using Western Blot analysis and indeed found that the ERK kinases 1/2 are phosphorylated under standard culture conditions but not under reduced culture conditions nor in native, uncultured fascicles (fresh tissue) (Fig. 1D).
To further delineate the role of ERK 1/2, we pharmacologically inhibited their downstream activity using a small molecule inhibitor (SCH772984). First, we tested whether ERK inhibition affects viability in tendon explants ( Fig. 2A,B). We observed a high cell viability (> 80%) in all conditions except for the methanol devitalized control tissue ( Fig. 2A,B). These results show that the ERK inhibitor has no cytotoxic effects, and that the ex vivo culture retains the viability of the tendon fascicles. In contrast to fascicles cultured under standard conditions, ERK inhibition suppressed macroscopic contraction of fascicles after 12 days of culture similar to fascicles cultured under reduced, non-degrading conditions (Fig. 2C). Next, we assessed the functional properties of tendon explants (Fig. 2D). Mechanical testing revealed a previously observed, drastic drop in elastic moduli (EMod) of tissues maintained in standard culture (mean: 70 MPa) when compared to fresh tissue (mean: 1300 MPa) and reduced culture conditions (mean: 1260 MPa) 12 . Surprisingly, ERK inhibition fully rescues the mechanical properties in standard conditions (mean: 1090 MPa). Additionally, we tested whether mechanical unloading alone was responsible for ERK 1/2 phosphorylation and the loss of functional properties in standard conditions. Hence, we cultured tendon fascicles under static load conditions using a custom-made loading device. While static mechanical load fully prevents fascicles from losing their mechanical properties, ERK 1/2 were phosphorylated under this loaded ex vivo condition (Supplementary figure S1A, S1B).
To examine how ERK inhibition suppresses tissue breakdown in tendon fascicles, we measured the gene expression of ECM degrading enzymes (matrix metalloproteinases Mmp3, 9,10,13), which we had identified previously by RNA-sequencing and mass spectrometry-based proteomics 12 (Fig. 3A). The expression of all measured MMPs was strongly increased under standard conditions compared to the fresh, uncultured control tissue (Mmp3: 26-fold, Mmp9: 50-fold, Mmp10: 230-fold, Mmp13: 269-fold). Non-degrading, reduced culture showed generally lower up-regulation of MMPs than standard culture. Strikingly, when ERK 1/2 activity was inhibited in standard culture conditions the expression of Mmp3, Mmp9, and Mmp10 was not induced and fully remained at baseline level of fresh control tissue. For Mmp13 we found that its expression was increased even in the ERK inhibition condition (16-fold) but was much lower than in the ex vivo standard condition (269-fold). To verify that these effects occur in other tendon types and in human tissue, we cultured human tendon tissues ex vivo and measured MMP expression. Indeed, we could confirm the MMP induction and inhibition in human tendons without and with ERK-inhibition, respectively (Fig. 3B). Collectively, these data show that ERK inhibition prevents the induction of ECM degrading enzymes (MMPs) and the loss of mechanical properties in tendon explants.
To investigate whether ERK influences further components of degenerative ECM remodeling we assessed the gene expression of collagen type 1 (Col1a1), collagen type 3 (Col3a1) and interleukin 11 (Il11) (Fig. 3C). Expression of collagen type 1, the most abundant component of the healthy tendon ECM, was reduced in all ex vivo conditions including when ERK 1/2 was inhibited. However, ERK 1/2 inhibition partially rescues the expression of Col1a1. In contrast, collagen type 3 expression, which is a known early phenotypic marker in tendinopathic ECM remodeling [19][20][21] , was increased in tendon fascicles. Interestingly, blocking ERK 1/2 inhibited the increased expression of Col3a1. We have previously observed that interleukin 11 (Il11), a stromal-derived cytokine and strong fibrosis driver in many tissues [22][23][24] , might be an important cytokine for tendon deterioration as it was heavily up-regulated in mechanically degraded tissues compared to non-degrading conditions (Supplementary figure S2A). Here, we show that Il11 was increased in degrading fascicles (Fig. 3C), and required ERK 1/2 activity to be expressed and released into the supernatant (Fig. 3D). When further confirming that Il11 is only expressed in degraded tendon tissue, we observed that mechanical loading of the tissue fully blocks the protein expression of IL11 (Supplementary figure S2B).

Discussion
The molecular events leading to tendon degeneration are poorly understood. Mechanical unloading of tendon sub-structures due to tissue rupture or microdamage is thought to be an ignition switch of the tissue breakdown cascade 16,20,25,26 . Here, we show that the ERK pathway drives tendon matrix degradation in mechanically unloaded tissue in a culture niche-dependent manner. We found that ERK 1/2 are highly phosphorylated in degrading tendon fascicles, and that ERK 1/2 activity is required for tendon deterioration. However, ERK 1/2 phosphorylation alone is not sufficient to induce loss of tendon mechanics as mechanically loaded explants show ERK 1/2 phosphorylation but retain their mechanical properties. These findings nicely show that biomechanical and biochemical cues act together in the maintenance of tendon mechanics and functional properties. In addition, the strong increase of MMP gene expression under the degrading conditions of a vascular-like niche depends on ERK 1/2 activity. We have previously shown that under standard culture conditions reactive oxygen species (ROS) are implicated in the MMP-driven tendon breakdown in load-deprived explants 12 . Since ERK is known to mediate ROS stress 27 , it seems plausible that a ROS-ERK-MMP axis pathway is a central mechanism of tendon tissue breakdown. Our data, establishing a link between ERK, MMPs and tendon matrix degradation, are in line   www.nature.com/scientificreports/ Our finding showing that ERK 1/2 inhibition protects tendon tissue from degradation is in agreement with previous studies reporting the involvement of ERK 1/2 in early tendinopathy. Using cultured human tenocytes, ERK 1/2 have been recognized to be key signal transducers between pathological triggers and tendinopathic phenotypes. For instance, Morita et al. showed that ERK 1/2 mediates the IL1β induced fibrotic phenotype of diseased tendon cells 32 . Furthermore, IL17A, IL33 and hypoxia were all shown to induce the expression of inflammatory cytokines and collagen type 3 in human tenocytes in an ERK-dependent manner 19,21,33 . Accordingly, we found that ERK 1/2 inhibition abrogates collagen type 3 expression, which is induced in load-deprived tendon explants. Similarly, we found that the stromal cytokine IL11, which is a potent fibrosis driver and therapeutic target in many tissues [22][23][24] , is highly upregulated in degrading tendon tissue but not when ERK 1/2 was blocked. This finding mirrors previous reports by Nishina et al. which demonstrated that the ERK pathway mediates ROS-induced production of IL11 during early events of acute damage in hepatic stroma 34 .
In this study we focused on gene expression, functional read-outs and ERK 1/2 protein levels using the fascicle ex vivo explant system as it allows for the controlled parametric investigation of the load-bearing tendon stromal compartment. Although the relatively simple structure of tail tendon fascicles offers advantages for the study of isolated cellular mechanisms, we acknowledge that there are limitations with using this model. In tendon fascicles the ratio of ECM to cells is very high and measuring ECM turnover, in particular cellular collagen, at the protein level by bulk methods is difficult due to these high ECM baseline amounts. Although we cannot rule out that the observed decrease in collagen type I gene expression in standard culture is responsible for the measured loss of tissue stiffness, we know from our previous proteomic quantification studies 12 that the overall content of collagen type 1 protein does not significantly differ between standard and reduced culture despite clear differences in ERK 1/2 phosphorylation state. Further, recent literature has revealed that tendon is a complex tissue with heterogeneous cell www.nature.com/scientificreports/ populations, including various stromal, endothelial and immune cells 35,36 . Crosstalk between the cells in tendon core and the surrounding synovial-like sheath is vital for tendon physiology and healing mechanisms 20,37 . The finite synovial tissue layer is often damaged during fascicle extraction ex vivo which compromises the cross-compartmental signaling 12,37 . Furthermore, murine tendons differ from their human counterparts in terms of function, composition and cross-linking patterns, with the mechanical and structural properties of tail tendons significantly differing from those of energy-storing tendons 38,39 . Nonetheless, there are compelling similarities between the murine tail tendon and human load bearing tendon, both showing an ERK-dependent MMP regulation that strongly resonates with previous studies on diseased human tendons. In this sense, the present work underscores the utility of the ex vivo fascicle model for dissecting the early molecular events that may precede the clinical presentation of chronic tendon pathologies. Future ongoing work is envisioned to investigate the role of ERK 1/2 pathway in degenerated human tendons using patient-derived materials from different anatomical locations.

Experimental procedures
Kinase enrichment analysis. Upstream protein kinases associated with the DEG from dataset (Array-Express: E-MTAB-7832) 12 were identified using Expression2Kinases suite (MacOS, Version 1.6.1207, http:// www. maaya nlab. net/ X2K/) 40 . Briefly, the top 3000 DEG were fed to Expression2Kinases pipeline which applies enrichment analysis to infer and rank potential transcription factors regulating the quired genes. Next, it identifies binding partners and constructs protein-protein transcriptional regulatory subnetwork, which are loaded to the Kinase Enrichment Analysis (KEA) module 41 . The identified kinases were mapped to kinome tree dendrogram using Coral tool and phylogenetic information derived from Manning et al. 42,43 (http:// phans tiel-lab. med. unc. edu/ CORAL/. Source code is available at https:// github. com/ dphan sti/ CORAL).
Murine tissue explantation and load-deprived culture. Twelve 11-weeks-old wild-type C57BL/6 male and female mice were obtained from the ETH Phenomics Center (EPIC) and euthanized with CO 2 previous to tissue harvest. Tendon fascicles were cautiously extracted from the murine tails using a surgical clamp.
For fresh tissue controls we used untreated fascicles directly after extraction. For ex vivo culture and treatments, fascicles were immediately hydrated in PBS after extraction and randomly distributed among the test groups. Subsequently, fascicles were maintained deprived from load in 6-well plates filled with 2 ml culture medium for 12 days in a humidified atmosphere at 5 kPa/5% CO 2  Tissue culture under mechanical load. Fascicles from 5 different mice were clamped at a length of 20 mm and their mechanical properties were measured before (day 0, fresh tissue) and after (day 12, static loading) cultivation under static mechanical load using a custom-designed uniaxial stretcher (see section "Biomechanical testing"). Fresh, clamped fascicles were placed into silicon chambers (filled with 2 ml medium) and cultured in a custom-designed bioreactor 45 at crimp disappearance length (static load) and under standard culture conditions. Crimp disappearance length was visually set for each fascicle using a stereomicroscope (Nikon SMZ 745T). As control, load-deprived fascicles (n = 5) were cultured in standard conditions as described above and mechanical tests were performed after 12 days of culture.
Human tissue explantation and culture. Hamstring tendon pieces (Gracilis, Semitendinosus) were collected from male and female patients undergoing anterior cruciate ligament reconstruction surgeries (mean age: 28 years, sd: 10.7). Immediately after dissection, tendons were immersed in standard culture medium and cut into smaller pieces (3-5 mm × 5-10 mm). Small pieces were either snap frozen as fresh control or subsequently cultivated in a 6-well plate with 5 ml culture medium for 3 days in standard culture or reduced culture (see section "Murine tissue explantation and load-deprived culture" www.nature.com/scientificreports/ cross-sectional area and the pre-load length at 0.015 N corresponded to 0% strain. Tangential elastic modulus was calculated after 5 preconditioning cycles in the linear part of the stress-strain curve (0.5-1% strain) by fitting a linear slope (Matlab R2016a, Version 9.0.0.341360). Any spurious negative values of elastic modulus resulting from severe tissue failure during mechanical characterization were set to zero to accord with the total functional impairment of the tissue.

Cell viability analysis.
To measure cell viability, dead cells were stained with Ethidium Homodimer-1 (EthD-1) solution (2 mM, AS 83208, Anaspec) in PBS for 10 min. Cells in negative controls were devitalized by treatment with 100% methanol (MeOH) for 10 min. Fascicles were thoroughly washed in PBS after EthD-1 and MeOH treatment and subsequently fixed in 10% formalin for 20 min, previous to nuclear staining with NucBlue reagent (1 drop/ml, R37606, Thermo Fisher). Stained fascicles were embedded in a 50% Glycerol:PBS solution and Z-stacks of 40 µm height (9 stacks by 5 µm) were acquired on a laser scanning confocal microscope (A1R Nikon) in triplicates per fascicle by using a 20 × objective (Nikon Ti2) (https:// www. micro scope. healt hcare. nikon. com/ en_ EU/ produ cts/ softw are/ nis-eleme nts). Quantification of the viability was performed using ImageJ (Fiji version 2.0.0-rc-59/1.51n, https:// imagej. net/). We separated the channels and maximum projected all images of a stack into a single image before subtracting background using the rolling ball method with a ball radius of 20 pixels. Next, we adjusted image contrast/brightness and converted the image into 8-bit grayscale to then binarize it. To avoid counting cell clusters instead of single cells, we watershed the image. Finally, we used the analyze particles tool to count the imaged dead cells (stained with EthD- graph pad. com/ scien tific-softw are/ prism/). Between group differences of normally distributed datapoints were evaluated by repeated measure ANOVA tests with the Tukey's post hoc multiple comparison method. Differences in IL11 secretion between culture conditions were evaluated using the non-parametric Friedman test and Dunn's post hoc multiple comparison test. In all cases, significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data were represented either as box and whisker plots with individual data points, showing 25th and 75th percentiles (box) and 5th and 95th percentiles (whiskers) or bar charts (showing mean + SD). Negative values of viability and elastic moduli resulting from limitations in the quantification analysis were set to 0.