Increased 15-PGDH expression leads to dysregulated resolution responses in stromal cells from patients with chronic tendinopathy

The mechanisms underpinning the failure of inflammation to resolve in diseased musculoskeletal soft tissues are unknown. Herein, we studied bioactive lipid mediator (LM) profiles of tendon-derived stromal cells isolated from healthy donors and patients with chronic tendinopathy. Interleukin(IL)-1β treatment markedly induced prostaglandin biosynthesis in diseased compared to healthy tendon cells, and up regulated the formation of several pro-resolving mediators including 15-epi-LXA4 and MaR1. Incubation of IL-1β stimulated healthy tendon cells with 15-epi-LXA4 or MaR1 down-regulated PGE2 and PGD2 production. When these mediators were incubated with diseased cells, we only found a modest down regulation in prostanoid concentrations, whereas it led to significant decreases in IL-6 and Podoplanin expression. In diseased tendon cells, we also found increased 15-Prostaglandin Dehydrogenase (15-PGDH) expression as well as increased concentrations of both 15-epi-LXA4 and MaR1 further metabolites, 15-oxo-LXA4 and 14-oxo-MaR1. Inhibition of 15-PGDH using either indomethacin or SW033291 significantly reduced the further conversion of 15-epi-LXA4 and MaR1 and regulated expression of IL-6, PDPN and STAT-1. Taken together these results suggest that chronic inflammation in musculoskeletal soft tissues may result from dysregulated LM-SPM production, and that inhibition of 15-PGDH activity together with promoting resolution using SPM represents a novel therapeutic strategy to resolve chronic tendon inflammation.

of stromal fibroblast activation including Podoplanin (PDPN), VCAM-1 (CD106) and Endosialin (CD248) compared to healthy tendon tissues and cells 11 . Stromal fibroblast activation is a feature of Rheumatoid Arthritis (RA) in which resident stromal cells fail to switch off their inflammatory programme. These phenotypic alterations in RA synovial fibroblasts play an important role in the switch from resolving inflammation to persistent disease 12,13 . Collectively, these studies support the concept that resident stromal fibroblasts are implicated in the persistence of chronic inflammation, although the mechanisms underpinning the failure of inflammation to resolve are not understood.
Inflammation resolution is an active and highly coordinated process whereby a repertoire of pro-resolving lipid mediators and proteins promote the timely resolution of inflammation after injury and/or infection [14][15][16] . Perturbed resolution is thought to contribute to the development of many systemic chronic inflammatory diseases 17,18 . Proresolving lipid mediators are well studied in experimental mouse models of systemic inflammation 19,20 as well as in humans 21,22 . Evidence for their protective roles in chronic inflammatory diseases is growing, including periodontal disease 23 , inflammatory arthritis 24 and pulmonary fibrosis 25 . Receptors implicated in mediating the effects of proresolving lipid mediators including the lipoxin A 4 receptor ALX/FPR2 and the Resolvin E1 receptor ERV1/ChemR23 have been identified in diseased human tendons 9 , suggesting a role for these mediators in disease etiopathology. Of note, to date the presence of these pro-resolving mediators and their regulation in diseased human tendon cells remains of interest.
The present study focused on identification of mechanisms underpinning the development of chronic inflammation in diseased human tendon tissues, which are currently poorly understood. We utilised an omics approach to perform a comprehensive analysis of pro-inflammatory and pro-resolving lipids in cultures of stromal fibroblasts derived from healthy and diseased human tendons. Using lipid mediator profiling, we identified differences in bioactive lipid mediator profiles between healthy and diseased tendon-derived stromal cells after treatment with IL-1β. We also investigated the biological actions of proresolving lipid mediators 15-epi-LXA 4 and MaR1 on counter-regulating dysregulated resolution processes in diseased tendon cells. The findings from this study provide improved understanding of the biological roles of SPM in diseased musculoskeletal soft tissues. We identify a mechanism underpinning dysregulated resolution responses in stromal cells from patients with tendinopathy, and propose a novel therapeutic strategy to promote resolution of chronic tendon inflammation.

Results
Diseased tendon-derived stromal cells display dysregulated resolution responses. Lipid mediator (LM) profiling of healthy hamstring and diseased supraspinatus tendon-derived stromal cell cultures identified specialized pro-resolving lipid mediators (SPM) including D-series Resolvins (RvD1, RvD2, RvD3, RvD4, RvD5, RvD6, 17R-RvD1 and 17R-RvD3), Protectins (PD1, 17R-PD1), Maresins (MaR1), E-series Resolvins (RvE1, RvE2, RvE3), arachidonic acid-derived Lipoxins (LXA 4 , LXB 4 , 15-epi-LXA 4 and 15-epi-LXB4) and n-3 DPA-derived Resolvins RvD1 n-3 DPA , RvD2 n-3 DPA and RvD5 n-3 DPA ), Protectins (10 S,17S-diHDPA) and Maresins (MaR1 n-3 DPA ). These mediators were identified in accordance with published criteria that include matching retention times and at least 6 ions in the tandem mass spectrum 26 (Fig. 1A,B, Supplementary Figure 1). Multivariate analysis uncovered differences in bioactive LM profiles between healthy and diseased tendon cells following incubation with IL-1β (10ngml −1 ) for 24 hours as demonstrated by the distinct clustering of the LM profiles ( Fig. 1A and B). Assessment of individual LM concentrations demonstrated significant increases in several SPM including Maresin (MaR) 1, n-3 DPA derived D-series resolvin (RvD1 n-3 DPA ), LXA 4 and 15-epi-LXA 4 . In these incubations we also found significant increases in several inflammation initiating eicosanoids including PGE 2 ( Fig. 1C and Table 1). This increase in both SPM and eicosanoids in tendon derived stromal cells from patients with tendinopathy was coupled with a significant increase in the expression of several of their biosynthetic enzymes including ALOX12, ALOX15 and PTGS2 ( Fig. 1D and E). These findings suggest that although SPM are up regulated in stromal cells from patients with tendinopathy, their concentrations are not sufficient to counter regulate the ongoing inflammatory processes, reminiscent of a dysregulated resolution response characteristic of chronic inflammatory conditions 27-29 . 15-epi-LXA 4 and MaR1 up regulate SPM production and reduce inflammatory responses in both diseased and healthy tendon-derived stromal cells. Having found altered LM-SPM profiles following addition of IL-1β, we next tested whether addition of 15-epi-LXA 4 and MaR1, two of the mediators up regulated in diseased tendon stromal cell incubations (Fig. 1), modulated responses in stromal cells isolated from healthy hamstring and diseased supraspinatus tendons. Incubation of tendon-derived stromal cells with either 0.1 nM or 10 nM 15-epi-LXA 4 dose-dependently up regulated SPM production, including the DHA derived RvD, PD and MaR as well as the EPA derived E-series resolvins in healthy volunteer cell incubations ( Fig. 2A, Supplemental Table 1). In addition, in these cell incubations we also found dose dependent decreases in the concentrations of inflammation initiating prostaglandins (PG), primarily PGE 2 , a mediator that carries both pro-inflammatory and nociceptive actions 30 (Fig. 2B, Supplemental Fig. 2 and Supplemental Table 1). Of note, when 15-epi-LXA 4 was incubated with cells derived from patients with tendinopathy, these cells displayed blunted actions in up regulating SPM production (Fig. 2, Supplemental Fig. 2, Supplemental Table 2). We also found a dose dependent decrease in PG levels, however these decreases were less pronounced than those observed with cells from healthy volunteers. Moreover, 15-epi-LXA 4 induced expression of ALOX15 mRNA relative to IL-1β-stimulated vehicle controls (Fig. 2C) and induced ALOX15 protein in diseased tendon stromal cells (n = 3 donors) (Fig. 2D).
We next assessed whether 15-epi-LXA 4 also regulated other markers of tendon inflammation in patient-derived stromal cells. Incubation of IL-1β-stimulated diseased tendon cells with 10 nM 15-epi-LXA 4 reduced PDPN mRNA (p = 0.03) and protein, moderated STAT-1 mRNA and protein and reduced IL-6 mRNA and protein levels in tissue culture media (p = 0.002) (Fig. 2E-G). Distinct SPM profiles in IL-1β stimulated healthy and diseased tendon stromal cells. Tendon stromal cells (60,000 cells per well) were derived from healthy hamstring (n = 8 donors) or diseased supraspinatus tendons (n = 8 donors). Cells were cultured in DMEM F12 phenol red free medium containing 1% heat inactivated human serum to 80% confluence and incubated with IL-1β for 24 h. Media and cells were harvested and placed in ice-cold methanol containing deuterium labeled internal standards. LM were then extracted and profiled. (A) 2-dimensional score plot and (B) corresponding 2-dimensional loading plot of plasma LM-SPM from human tendon derived-stromal cell incubations isolated from healthy volunteers or patients with tendinopathy after stimulation with IL-1β (10ngml -1 ) for 24 h. Grey ellipse in the score plot denotes 95% confidence regions. (C) Concentrations for mediators found to be differentially regulated between healthy (grey bars) and diseased (black bars) tendon stromal cell incubations. Results are shown as means and SEM and representative of n = 8 donors per group. (D) mRNA expression of lipid mediator biosynthetic enzymes determines using real time qPCR. Gene expression is normalized to β-actin, bars show median values. (E) Representative immunofluorescence images showing staining for ALOX15 (green), PTGS2 (red) and nuclei (cyan) in IL-1β-stimulated healthy and diseased tendon cells. Scale bar, 20μm. Results are representative of n = 3 donors.
Incubation of tendon-derived stromal cells with 10 nM MaR1 also significantly up regulated SPM concentrations in incubations with healthy cells. Here we found increases in RvD and LX as well as statistically significant decreases in inflammation initiating eicosanoids including PGE 2 ( Fig. 3A and B, Supplemental Fig. 3 and Supplemental Table 1). Incubation of diseased tendon cells with MaR1 also lead to a decrease in the levels of PG and Tx, although as observed for 15-epi-LXA 4 the reduction in these pro-inflammatory eicosanoids was less than that observed with cells from healthy volunteers (Fig. 3, Supplemental Fig. 3 and Supplemental Table 2). In addition to regulating eicosanoid production, incubation of MaR1 with diseased tendon cells also lead to down-regulation of mRNA and protein of PDPN, STAT-1 and IL-6 ( Fig. 3C-E). Together these findings demonstrate that MaR1 and 15-epi-LXA 4 counter regulate IL-1β initiated inflammation in tendon-derived stromal cells. They also point to a dysregulated resolution response in cells derived from patients with tendinopathy, given the lower potency of these mediators at regulating the production of both pro-resolving and pro-inflammatory mediators.

Diseased tendon cells and tissues show increased PGDH expression.
Since tendon stromal cells from diseased donors had an enhanced ability to convert 15-epi-LXA 4 and MaR1 into metabolites that possess reduced biological actions, we investigated the expression of enzymes implicated in SPM metabolism in healthy and diseased tendon tissues and cells. NAD + dependent 15-Prostaglandin Dehydrogenase (15-PGDH) is a short-chain dehydrogenase/reductase (classified as SDR36C1) 32,33 , that converts PGE 2 and LXA 4 to 15-keto-PGE 2 and 15-oxo-LXA 4 respectively 34 . We therefore investigated expression of 15-PGDH mRNA and protein in cells and tissues derived from healthy and diseased human tendons. 15-PGDH mRNA and protein were highly expressed in cells isolated from diseased supraspinatus compared to healthy hamstring tendons after IL-1β treatment for 24 h (p = 0.008) ( Fig. 4C and D). Tissues derived from patients with supraspinatus tendinopathy (n = 14) showed increased 15-PGDH mRNA compared to healthy (subscapularis) shoulder tendons (n = 4) (p = 0.018) (Fig. 4E). Expression of 15-PGDH protein was also determined in these healthy and diseased human shoulder tendon tissues. Immunostaining confirmed increased 15-PGDH protein in diseased compared to healthy shoulder tendons (Fig. 4F). Together these results suggest that the increase in 15-PGDH leads to rapid further metabolism and inactivation of the SPM in cells from diseased patients, thereby blunting their biological actions.

15-PGDH inhibition prevents 15-epi-LXA 4 and MaR1 further conversion.
Having identified increased 15-PGDH and enhanced SPM further metabolism in cells from patients with tendinopathy, we investigated whether 15-PGDH was indeed responsible for inactivation of these mediators in cells derived from these patients. For this purpose, we incubated cells with either the 15-PGDH inhibitor SW033291 35 or indomethacin which in addition to its effects on COX also inhibits 15-PGDH activity 36 , and is used clinically to moderate inflammation 37 . In incubations of IL-1β stimulated diseased tendon stromal cells with either indomethacin or SW033291, we found significantly lower concentrations of 15-oxo-LXA 4 and 14-oxo-MaR1 levels, and a corresponding increase in the levels of 15-epi-LXA 4 and MaR1 (Fig. 5A,B Tables 2,3). These results indicated that the elevated 15-PGDH expression was responsible for the blunted actions of these mediators in regulating lipid mediator profiles in stromal cells from patients with tendinopathy. We therefore next questioned whether incubating cells with both an SPM and indomethacin or SW033291 would have additive actions on regulating markers of inflammation in tendon stromal cells. Assessment of IL-6 production in these cells demonstrated that whereas there were no additive actions with indomethacin, incubation of SW033291 together with 15-epi-LXA 4 displayed additive actions in down-regulating the concentrations of this inflammatory cytokine (Fig. 5C). We also found that co-incubation of 15-epi-LXA 4 with indomethacin displayed additive actions in regulating PDPN and STAT-1 expression (Fig. 5D) compared to incubation in 15-epi-LXA 4 alone ( Fig. 2E and G).

Discussion
Specialized proresolving mediators (SPM) including lipoxins, resolvins, protectins and maresins initiate the highly active and coordinated process of resolution 38 , regulating the duration and magnitude of inflammation and promoting restoration of tissue homeostasis after infection and/or injury 14,16,39 . Whilst SPMs are implicated in resolving acute inflammation via cells of the innate immune system, these bioactive mediators are also associated with chronic inflammatory diseases 17,18 . Tissue-resident stromal cells such as fibroblasts are emerging as an important cell type implicated in mediating the resolution of inflammation in wound healing, periodontal disease, pulmonary inflammation and Rheumatoid Arthritis 24,25,40,41 . Stromal fibroblasts actively participate in inflammatory responses and are implicated in governing the persistence of inflammatory disease through a variety of mechanisms including stromal fibroblast activation, recruitment and retention of immune cells, inhibition of leucocyte apoptosis and immunological memory 13 .
Chronic inflammation is a common feature of musculoskeletal soft tissue diseases including tendinopathy 9 . Tendons possess a low rate of tissue turnover 42 , therefore damage accumulated may be long lasting as diseased tissue heals by fibrosis and not regeneration. Current therapeutic strategies focus on ameliorating the pain associated with disease but do not address the underlying biological mechanisms underpinning the development of to β-actin, bars show median values. (F) ELISA assay of IL-6 protein secretion from IL-1β stimulated diseased tendon cells incubated in the presence and absence of 10 nM 15-epi-LXA 4 . Data are shown as means and SEM, n = 4 separate donors. (G) Representative immunofluorescence images showing staining for STAT-1 (green), IL-6 (red), PDPN (green), and nuclei (cyan) in IL-1β stimulated diseased tendon stromal cells incubated in 10 nM 15-epi-LXA 4 . All images are representative of n = 3 donors. Scale bar, 20 μm. chronic inflammation. Resolution of inflammation has not been well studied in the context of diseased human musculoskeletal soft tissues.
In the present study we found that incubation of tendon-derived stromal cells with IL-1β up regulated the production of pro-inflammatory eicosanoids and proresolving SPM. Furthermore, the production of select  pro-resolving mediators was significantly higher in tendon stromal cells from patients with tendinopathy compared with those from healthy volunteers. These results are in line with an increased SPM biosynthetic enzyme observed in these cells, pointing to a status of dysregulated resolution, where in response to an inflammatory stimulus, in this case IL-1β, the up regulation of tissue protective mediators is not sufficient to counter regulate the inflammatory profile. Of note, addition of either 15-epi-LXA 4 or MaR1 to the cell incubations also led to a feed forward production in endogenous SPM production by both healthy volunteer and diseased tendon stromal cells. These findings are also in line with an increased expression of the SPM biosynthetic enzyme ALOX15 observed in these cells as well as with published findings in other experimental systems including peritonitis 43 and asthma 44 , where the administration of one SPM triggers the formation of different SPM that contribute to the resolution of inflammation. These findings suggest that diseased tendon cells display a pro-inflammatory and dysregulated resolution profile as summarized in Figure S4. Our findings are consistent with earlier reports that demonstrate pro-inflammatory mediators including IL-1β induce the production of PGE 2 in cultures of tendon cells 45,46 .
Having observed that a select group of pro-resolving mediators was up regulated in patient-derived compared to healthy tendon stromal cells, we queried whether these autacoids carried biological actions in regulating molecular aspects of tendon inflammation. Indeed 15-epi-LXA 4 and MaR1 regulated lipid mediator production in both healthy and patient tendon-derived stromal cells. In addition, incubation of patient derived tendon cells with these SPM also led to a reduction of IL-6, STAT-1 and PDPN, that we have previously found to be associated with disease severity 9,11 . Of note, we found that the biological actions of both 15-epi-LXA 4 and MaR1 were blunted in patient-derived tendon cells, this decreased effectiveness was associated with an increased further metabolism of these mediators to their inactive metabolites. This process of inactivation was at least in part reliant on 15-PGDH since this enzyme was found to be up regulated in patient cells compared with tendon stromal cells from healthy volunteers. Furthermore, inhibition of this enzyme using either indomethacin or SW033291 led to increased recovery of both 15-epi-LXA 4 and MaR1 and a reduction in the further metabolism of these mediators in diseased tendon stromal cell incubations. 15-PGDH has been recently shown to negatively regulate tissue repair and regeneration in murine models of bone marrow, colon and liver injury 35 . Zhang et al., illustrated that 15-PGDH blockade potentiated repair in multiple murine tissues without apparent adverse effects. Having found that the 15-PGDH inhibitors reduced both SPM further metabolism and prostaglandin production, we also queried whether co-incubation of either 15-epi-LXA 4 or MaR1 with these inhibitors moderated the pro-inflammatory phenotype of diseased tendon stromal cells. In these incubations we found a further reduction in the expression of IL-6, STAT-1 and PDPN as well as prostaglandins, although this was not statistically significant. Future experiments will need to investigate the potential of this approach and the possibility of obtaining additive or even synergistic actions with a dual pronged approach in controlling soft tissue inflammation, in line with actions observed in bacterial infections where SPMs lower the required doses of antibiotics required to clear infections 47 .
We acknowledge there are potential limitations with the use of hamstring tendon as a comparator to diseased tendons including tendon type and donor age differences. However, hamstring tendon was taken from live healthy donors without history of tendinopathy. We believe this is a more suitable comparator than cadaveric shoulder tendon tissues where little is known about whether the tendons were healthy or diseased and tendons were not affected by post mortem changes.
The findings from this study suggest that tendinopathy is characterized by a status of dysregulated resolution that results from an up regulation of 15-PGDH leading to the rapid inactivation of SPM. Incubation of tendon-derived stromal cells from both healthy volunteers and patients with tendinopathy in SPM including 15-epi-LXA 4 or MaR1 induced further release of proresolving mediators and counter regulated the expression of pro-inflammatory molecules including PGE 2 , IL-6, STAT-1 and PDPN. In addition, our findings suggest that a dual pronged approach, using pro-resolving mediators together with inhibitors to 15-PGDH may represent a novel therapeutic strategy to reduce local inflammation and promote tissue repair and regeneration.

Materials and Methods
Study Approval. Healthy and diseased tendon tissues were collected under research ethics from the Oxford Musculoskeletal Biobank (09/H0606/11). Full informed consent according to the Declaration of Helsinki was obtained from all patients. Experimental protocols were performed at the University of Oxford in accordance with research ethics from the Oxford Musculoskeletal Biobank (09/H0606/11).
Collection of human tendon tissues. All patients were recruited from orthopedic referral clinics where the structural integrity of the supraspinatus tendon was determined ultrasonographically. Patients presenting to the referral shoulder clinic had failed non-operative treatment, including a course of physical therapy, and had experienced pain for a minimum of 3 months. Patients completed the Oxford Shoulder Score (OSS), a validated and widely used clinical outcome measure scoring from 0 (severe pathology) to 48 (normal function). Diseased (D) Representative immunofluorescence images of tendon stromal cells isolated from healthy hamstring donors (n = 3) and patients (n = 3) with supraspinatus tendinopathy showing staining for 15-PGDH (red) and nuclei (cyan) after stimulation with IL-1β (10ngml -1 ) for 24 h. (E) 15-PGDH mRNA expression in healthy subscapularis (n = 4 donors) and diseased supraspinatus (n = 14 donors) shoulder tendon tissues. Gene expression is normalized to β-actin, bars show median values. (F) Representative immunofluorescence images of sections of diseased and healthy shoulder tendons stained for 15-PGDH (red). Cyan represents POPO-1 nuclear counterstain. Scale bars, 20μm. tendon tissues (supraspinatus tendon tears) were collected at the time of surgical debridement of the edges of the torn tendons from 28 male and female patients aged between 44 and 75 (mean 55 ± 18.3 years). All patients were symptomatic and had small to medium tears (≤1 cm to ≤3 cm in anterior-posterior length). Exclusion criteria for all patients in this study included previous shoulder surgery, other shoulder pathology, rheumatoid arthritis and systemic inflammatory disease. Samples of healthy supraspinatus (n = 5) and subscapularis tendons (n = 4) were collected intra-operatively from male and female patients between 25-65 years of age that were undergoing   Results are expressed as pg/incubation. Mean ± SEM of n = 3 per incubation. *P < 0.05, **P < 0.01 vs Disease + IL1β incubations. $ P < 0.05, $$ P < 0.01 vs Disease + IL1β + Indo; # P < 0.05, ## P < 0.01 vs Disease + IL1β + SW. The detection limit was ~0.1 pg. -, Below levels found in media alone.
shoulder surgery for post-traumatic instability. Healthy hamstring tendons were collected from 15 male and female patients undergoing surgical reconstruction of their anterior cruciate ligament. All patients were aged between 18 and 48 (mean 25.2 ± 11 years).   Treatment of tendon-derived stromal cells with IL-1β for bioactive lipid mediator profiling. As IL-1β induces NF-κB target genes known to be highly expressed in early stage tendinopathy 9 , we investigated bioactive pro-resolving lipid mediator profiles in cells derived from healthy and diseased human tendons in the presence of IL-1β. Tendon-derived stromal cells from healthy hamstring (n = 8) and diseased supraspinatus tendons (n = 8) were seeded at a density of 60,000 cells per well in a 6 well plate. Tendon cells were allowed to reach 80% confluence prior to stimulation with IL-1β (Merck, 10ngml -1 ) in DMEM F12 phenol red free medium (Gibco) containing 1% heat inactivated human serum (Sigma) and 1% penicillin-streptomycin. Non-treated cells (vehicle only, containing 0.1% endotoxin free BSA, Sigma) served as controls for each experiment. After treatment, cells were then incubated at 37 °C and 5% CO 2 until harvest of the media and lysate for bioactive lipid mediator profiling after 24 h. Bioactive lipid mediator profiling of IL-1β stimulated healthy and diseased tendon stromal cells. Media

15-PGDH mRNA expression in healthy and diseased tendon tissues and cells. Tendon-derived
stromal cells from healthy hamstring (n = 5) and diseased supraspinatus tendons (n = 5) were seeded at a density of 15,000 cells per well in a 24 well plate. Tendon cells were allowed to reach 80% confluence prior to stimulation with IL-1β (Merck, 10ngml -1 ) in DMEM F12 phenol red free medium (Gibco) containing 1% heat inactivated human serum (Sigma) and 1% penicillin-streptomycin. Non-treated cells (vehicle only, containing 0.1% endotoxin free BSA, Sigma) served as controls for each experiment. After cytokine treatment, cells were then incubated at 37 °C and 5% CO 2 until harvest of the cell lysate for mRNA after 24 h. For tissues, samples of healthy subscapularis (n = 4) and diseased supraspinatus tendons (n = 14) were snap frozen and stored at −80 °C. RNA isolation, cDNA synthesis and quantitative PCR were performed using previously published protocols 9 . Pre-validated Qiagen primer assays (15-PGDH, STAT-1, IL6, PDPN, β-actin and GAPDH) were used for qPCR. Results were calculated using the ddCt method using reference genes for human β-actin and GAPDH. Results were consistent using these reference genes and data are shown normalized to β-actin.
Immunofluorescence for 15-PGDH in healthy and diseased tendons. Tendon samples were immersed in 10% buffered formalin, processed using a Leica ASP300S tissue processor and embedded in paraffin wax. Tissues were sectioned to 4μm onto adhesive glass slides. For antigen retrieval, slides were baked at 60 °C for 60 minutes and tissue sections were taken through deparaffinisation and target retrieval steps (high pH heat mediated antigen retrieval) using an automated PT Link (Dako). Immunofluorescence staining protocols and image acquisition are adapted from a previously published protocol 9 . Sections were incubated with the primary antibody anti-15-PGDH (Abcam, ab118185). For negative controls the primary antibody was substituted for universal isotype control antibodies: cocktail of mouse IgG 1 , IgG 2a , IgG 2 b, IgG 3 and IgM (Dako) ( Figure S5).

Immunocytochemistry for healthy and diseased tendon stromal cells. Tendon stromal cells were
grown in chamber slides and treated as above. Cells were fixed in ice cold methanol for 5 mins and washed twice in PBS. Immunofluorescence staining protocols and image acquisition are adapted from a previously published protocol 9 . Tendon stromal cells isolated from 3 healthy hamstring donors and 3 patients with supraspinatus tendinopathy were incubated with the following primary antibodies: anti-ALOX15 (Abcam ab119774), anti-15-PGDH (Abcam, ab118185), anti-STAT-1 (Abcam ab29045), anti-COX2 (Abcam ab15191), anti-Podoplanin (Abcam ab10288) and anti-IL-6 (Abcam 9324) in PBS containing 5% goat serum in Saponin for 3 hrs at room temperature. For negative controls the primary antibody was substituted for universal isotype control antibodies: cocktail of mouse IgG 1 , IgG 2a , IgG 2 b, IgG 3 and IgM (Dako) and rabbit immunoglobulin fraction of serum from non-immunized rabbits, solid phase absorbed. Isotype control staining is shown in Figure S5.
Immunofluorescence imaging. Immunofluorescence images were acquired on a Zeiss LSM 710 confocal microscope using a × 40 oil immersion objective (NA = 1.3). The fluorophores of POPO-1, Alexa Fluor 488, Alexa Fluor 568, and Alexa Fluor 633 were excited using the 405 nm, 488 nm, 561 nm, and 633 nm laser lines respectively. To minimize bleed-through, all channels were acquired sequentially. Averaging was set to 2 and the pinhole was set to approximately 1 Airy unit. Two-dimensional image reconstructions were created using ZEN 2009 (Zeiss).

Treatment of diseased tendon stromal cells with inhibitors of 15-PGDH. Tendon stromal cells
were derived from diseased supraspinatus tendons. Cells were seeded at a density of 60,000 cells per well in 6 well plates for SPM profiles, (6 donors). Once cells were at 80% confluence, they were pre-incubated in 10μM indomethacin (Sigma) or 25μM 15-PGDH inhibitor (SW033291, Tocris) in DMEM F12 phenol red free medium containing 1% heat inactivated human serum and 1% penicillin-streptomycin. After 2 h, 10 nM 15-epi Lipoxin A 4 (Cayman Chemical) or 10 nM Maresin 1 (Cayman Chemical) were added and cells further incubated for 24 h. Healthy and diseased cells were then stimulated with IL-1β (10ngml -1 ), non-treated (vehicle only) IL-1β stimulated cells served as controls for each experiment. Cells were shielded from light and incubated at 37 °C and 5% CO 2 until harvest of the media and lysate for bioactive lipid mediator profiling 24 h after stimulation with IL-1β.
Quantification of Interleukin-6 in tissue culture media from diseased tendon stromal cells. IL-6 is an important cytokine implicated in tissue inflammation. IL-6 in tissue culture supernatants was measured using enzyme-linked immunosorbent assay (ELISA) reagents (BD Biosciences). Minimum detectable IL-6 concentration for this assay was 2.2 pgml -1 . Optical density was read on a spectrophotometric ELISA plate reader (FLUOstar Omega, BMG Labtech) and analysed using MARS data analysis software.
Statistics. Statistical analyses were performed using GraphPad Prism 7 (GraphPad Software). Normality was tested using the Shapiro-Wilk normality test. Pairwise Mann Whitney U tests were used to test for differences between mRNA expression of ALOX12, ALOX15, PTGS2, IL6, STAT-1, PDPN and 15-PGDH in IL-1β treated tendon stromal cells in the presence or absence of SPM and inhibitors of 15-PGDH. Pairwise Mann Whitney U tests were used to test for differences between IL-6 protein secretion in IL-1β treated tendon stromal cells in the presence or absence of SPM and inhibitors of 15-PGDH. Pairwise Mann Whitney U tests were used to test for differences between 15-PGDH mRNA expression in healthy and diseased tendon tissues and cells. Analysis of bioactive lipid mediator profiles from healthy and diseased tendon cells was performed using multivariate statistical analysis, orthogonal-partial least squares-discriminant analysis (o-PLS-DA) was performed using SIMCA 14.1 software (Umetrics, Umea, Sweden) following unit variance scaling of LM amounts. PLS-DA is based on a linear multivariate model that identifies variables that contribute to class separation of observations (cell incubations) on the basis of their variables (LM levels). During LM classification, observations were projected onto their respective class model. The score plot illustrates the systematic clusters among the observations (closer plots presenting higher similarity in the data matrix). Loading plot interpretation identified the variables with the best discriminatory power (Variable Importance in Projection greater then 1) that were associated with tight clusters for lipid mediator profiles obtained from incubations with cells from healthy volunteers or patients with tendinopathy. Data are shown as mean and SEM, where n is the biological replicate (human donor of cells derived from healthy or diseased tendons). P < 0.05 was considered statistically significant.
Data Availability. All data generated from this study are included in this published article and its Supplementary Information files.