Unique inflammatory signature in haemophilic arthropathy: miRNA changes due to interaction between blood and fibroblast‐like synoviocytes

Abstract In haemophilia, the recurrence of hemarthrosis leads to irreversible arthropathy termed haemophilic arthropathy (HA). However, HA is a unique form of arthropathy in which resident cells, such as fibroblast‐like synoviocytes (FLS), come into direct contact with blood. Therefore, we hypothesized that FLS in HA could have a unique inflammatory signature as a consequence of their contact with blood. We demonstrated with ELISA and ELISPOT analyses that HA‐FLS expressed a unique profile of cytokine secretion, which differed from that of non‐HA‐FLS, mainly consisting of cytokines involved in innate immunity. We showed that unstable cytokine mRNAs were involved in this process, especially through miRNA complexes as confirmed by DICER silencing. A miRNOME analysis revealed that 30 miRNAs were expressed differently between HA and non‐HA‐FLS, with most miRNAs involved in inflammatory control pathways or described in certain inflammatory diseases, such as rheumatoid arthritis or lupus. Analysis of transcriptomic networks, impacted by these miRNAs, revealed that protein processes and inflammatory pathways were particularly targeted in LPS‐induced FLS, and in particular vascularization and osteoarticular modulation pathways in steady‐state FLS. Our study demonstrates that the presence of blood in contact with FLS may induce durable miRNA changes that likely participate in HA pathophysiology.

musculoskeletal disability and poor quality of life. Although prophylactic treatment with a clotting factor versus episodic infusion has demonstrated superiority in preventing joint disease, 3 HA remains a significant comorbidity of the haemophilic patient.
Indeed, the pathophysiology of blood-induced joint disease remains unclear. Some data suggest that the pathobiology of HA may be similar to the pathogenicity of rheumatoid arthritis (RA), given that chronic proliferative synovitis and cartilage destruction are present in both. 4 Moreover, metalloprotease enzymes, inflammatory cells and inflammatory cytokines, such as interleukin-1 (IL-1) or tumour necrosis factor alpha (TNF-α), have been found in the synovium of HA as would be expected in RA. 5 One of the main difference between the two diseases is the origin of inflammation: autoimmunity in RA and blood injury in HA. Synovial iron deposition as a result of blood in the joint is considered the main component in triggering and sustaining the inflammatory response and cell proliferation within the synovial membrane. 6 Thus, synovial macrophages from blood injury promote the inflammatory pathway of hemarthrosis and HA. However, these elements alone cannot explain the entire pathophysiology of HA. Indeed, in hemochromatosis, another arthropathy characterized by synovial iron deposition, there is no evidence of significant synovial hyperplasia, although inflammatory cells and articular destruction have been observed. 7 If we focus on the main difference between these two forms of arthritis, the only element found purely in HA is the presence of blood in the joint in direct contact with the synovial membrane, and, particularly, with the unique resident cells in joints, the fibroblast-like synoviocytes (FLS).
The exact role of FLS in HA pathophysiology is still unknown, but assumed to be central. As widely demonstrated in RA, FLS straddles the boundary between innate and adaptive immunity, given that they synthesize inflammatory cytokines while cooperating with lymphocytes and macrophages. 8,9 Owing to their high proliferation rate in HA, these cells are directly involved in synovial proliferation and perpetuate inflammation, thereby contributing to cartilage damage. Due to this proliferative activity, the apoptotic pathway is disrupted to the advantage of anti-apoptotic activity.
C-MYC and Mdm2, two of the main pro-mitotic factors, are overexpressed in the FLS of HA. 10,11 The use of an agonistic anti-human Fas monoclonal antibody can induce FLS apoptosis in HA and reduce inflammatory cytokine production, thereby abolishing synovial hypertrophy. 12 We hypothesized that, in haemophilia, chronic interaction between blood and FLS in the joint leads to a unique cytokine signature due to acquired and lasted changes, explaining its inflammatory and proliferative characters. First, from the synovium of patients who underwent orthopaedic surgery, we isolated pure HA-FLS versus non-HA-FLS. Following stimulation, the expression pattern of cytokines from HA-FLS completely differed from that from non-HA-FLS, mainly reflected by messenger RNA (mRNA) instability and micro-RNA (miRNA) interaction. A miRNA microarray assay showed HA as possessing a unique signature closer to the miRNA of patients with chronic inflammatory diseases, such as RA, than other forms of arthritis.
Our data demonstrated that blood could induce durable epigenetic changes in resident cells due to pathological interactions.
After stimulation, supernatants were harvested and assayed for cytokine content using commercially available ELISA tests.

| Protein profiling from supernatant
Protein profiling was performed based on the supernatant of cell activation. Some 300 µL of supernatant were loaded on proteome profiler antibody array membranes (Human XL Cytokine Array Kit, R&D Systems), as suggested by the supplier. These membranes were washed and incubated with biotinylated detection antibody cocktail, streptavidin/horseradish peroxidase and chemiluminescent detection reagents, as suggested by the supplier. Membranes were exposed to a ChemiDoc and analysed using ImageJ Software.
Image analysis was performed with background subtraction and normalization to membrane reference points. Differences in the expression levels of various proteins were calculated pairwise as fold changes, with comparisons made between HA and non-HA-FLS.

| Stimulation of cells for total extraction
Total RNA was extracted from human FLS or THP-1 cells incubated for 6 hours with either medium alone or medium containing LPS used the RNeasy Plus Mini Kit according to the manufacturer's instructions. Total RNA isolated from the FLS and THP-1 cells was reverse transcribed using the iScript Reverse Transcription Supermix for RT-qPCR, according to the manufacturer's instructions (Bio-Rad), and amplified.

| mRNA decay measurement
The stability of mRNA was assessed by adding 5 µg/mL of actinomycin D to the cell medium following activation with foetal bovine serum to inhibit mRNA transcription. At the indicated time points, it was possible to correlate the relative amount of specific mRNA remaining in each sample with mRNA degradation. Total RNA was extracted 2 hours after treatment with actinomycin D, and endogenous mRNA levels were analysed by RT-qPCR. Because the mRNA levels for GAPDH and β actin were unchanged after actinomycin D treatment, the GAPDH and β actin gene were employed as a reference, and the ratio of IL-1α/IL-6 and GAPDH/β actin in each sample was calculated.

| Real-time quantitative polymerase chain reaction
RT-qPCR was performed on a total of 20 µL using the SsoFast EvaGreen Supermix (Bio-Rad) and gene-specific primers for (i)TNF-α After initial denaturing at 95°C for 3 minutes, the temperatures used were 95°C for 5 seconds, 58°C for 15 seconds and 72°C for 15 seconds. Some 40 cycles were performed using the C1000 Touch Instrument (Bio-Rad).
RT-qPCR analyses for miRNAs were conducted using the miScript System, and the primers (Qiagen) and RNA concentrations were determined with a NanoDrop instrument (Thermo Fisher). A 1000 ng RNA sample was employed for the assays. Reverse transcriptase reactions and RT-qPCR were carried out according to the manufacturer's protocols. An endogenous control was employed for normalization. All reactions were run in triplicate on a C1000 Touch Instrument (Bio-Rad).

| Single interfering RNA transfection
The DICER siRNA used in our study was designed to effectively inhibit DICER activity, being supplied by Qiagen.
Cells were transfected with siRNA using the HiPerFect Transfection Reagent kit (Qiagen). Cells were plated in 24-well plates (1 × 10 3 cells/well). All assays were performed 24 hours after transfection.

| Data were available on web Array Express, with the number: E-MATB-7442
Model of onset haemophilic arthropathy after repeated contact between plasma and normal HR-FLS.
Regular Human Fibroblast-Like Synoviocytes (HR-FLS) were contacted with plasma repeatedly for 3 weeks. Briefly, HR-FLS were inoculated in flask, and the next day, a ratio of 50% plasma/50% medium is used for cell culture. Every 3 days, the exposure with fresh plasma with the same ratio is renewed. After 3 weeks, HR-FLS (5.10 5 cells) were stimulated with 1 mL of either medium alone or medium containing LPS for 6 hours. Total RNA was extracted following the instructions below. miRNA-specific primers were obtained from Qiagen.

| HA-FLS and non-HA-FLS are solely composed of a pure homogenous fibroblastic population
As only little data described the use of human FLS from haemophilic patients are available, we first verified that cell cultures of HA and non-HA-FLS were composed of pure fibroblastic cell populations. A study of cell surface expression, using flow cytometry, revealed that HA-FLS ( Figure S1A), non-HA-FLS ( Figure S1B), RA-FLS ( Figure S1C) and HR-FLS ( Figure S1D) exhibited similar cell surface markers (CD45, CD55, CD68, CD90 and CD106), completely differing from the THP-1 cell line ( Figure S1E). 8

| An unique release of inflammatory cytokines in HA-FLS
To examine the FLS cytokines expression profile, we first per- iron metabolism (TfR) ( Table 1).
Given that innate immunity cytokines appeared particularly dysregulated, in an effort to confirm the inflammatory signature of HA-FLS, we focused on the cytokine expression of three proinflammatory proteins (IL-1α, IL-6 and TNF-α) after 6 hours of activation with LPS.
These three cytokines were selected because they are widely in-  Figure 2D) and THP-1 cell lines ( Figure 2E) did release significant detectable amounts of IL-1α. As previously described, no TNF-α expression was found after LPS stimulation in any of the fibroblast subtypes, as compared to the THP-1 cell line ( Figure 2E). 15

| IL-1α inflammatory cytokine synthesis is repressed post-transcriptionally by a miRNAdependent pathway in LPS-activated HA-FLS
To determine the mechanism underlying the regulation of inflam- According to the miRNA microarray results from the FLS, we found that more than a thousand miRNAs were differentially expressed in all HA-FLS and non-HA-FLS (P50, p0.05, f1.2) ( Figure S2).
We then focused on differences in miRNA expression levels between HA-FLS and non-HA-FLS. As a result, 14 miRNAs in the steady-state condition with medium control (Table 2) were revealed to display significantly different expression levels, as did 16 different miRNAs after LPS stimulation ( Table 3). Most of these 30 miRNAs could potentially target mRNAs involved in inflammatory processes directly or indirectly, such as proinflammatory cytokines (IL-1α and L-6 with miR-146 or miR-224, TNF-α with miR-146a-5p or miR185a-5p or miR125b-2-3p, etc) or specific signalling pathways (PI3K and micro-RNA-204-3p, SOCS3/STAT3 and miR-196b-5p, etc).

| Correlation between acquired miRNA profile in HA-FLS and repeated plasma exposure
To demonstrate that epigenetic changes are due to a chronic contact of blood with FLS, we propose to use an in vitro model of onset haemophilic arthropathy after repeated contact between plasma and normal HR-FLS. We also compared expression of these 11 miRNA after 3 weeks of repeated exposure of normal HR-FLS by plasma or medium. We observed the same variations between plasma exposed versus non-exposed HR-FLS that HA versus non-HA-FLS, after medium activation ( Figure 4A) or LPS ( Figure 4B).

| Transcriptomic network potentially impacted by miRNA expression in HA-FLS: From inflammatory to ubiquitous changes
To further investigate the potential impact of miRNA epigenetic changes in FLS following chronic interaction with blood, we performed a computer analysis pertaining to the targets of previous  Note: The global expression profile of miRNAs is analysed using a microarray-based approach after exposing HA and non-HA-FLS to LPS. FLS were stimulated with medium (Table 2) and LPS (Table 3) for 6 hours, and the extracted RNAs were compared using a DNA microarray containing 2578 probes complementary to miRNAs of human origin. A HA-FLS/non-HA-FLS miRNA ratio was used to determinate expression levels of each miRNA. After used of a significant cut-off ratio, respectively, at < 0.75 and > 1.5, 14 miRNAs had significantly different expression levels after medium stimulation ( Table 2) and 16 miRNAs had significantly different expression levels after LPS stimulation (Table 3).

TA B L E 3
Microarray analysis of miRNA profile of HA-FLS shows a single miRNA signature close to that of inflammatory arthritis and diseases Note: The global expression profile of miRNAs is analysed using a microarray-based approach after exposing HA and non-HA-FLS to LPS. FLS were stimulated with medium (Table 2) and LPS (Table 3) for 6 hours, and the extracted RNAs were compared using a DNA microarray containing 2578 probes complementary to miRNAs of human origin. A HA-FLS/non-HA-FLS miRNA ratio was used to determinate expression levels of each miRNA. After used of a significant cut-off ratio, respectively, at < 0.75 and >1.5, 14 miRNAs had significantly different expression levels after medium stimulation ( Table 2) and 16 miRNAs had significantly different expression levels after LPS stimulation (Table 3) Upon LPS activation, we detected that 55 mRNAs could be targeted by three or more selected miRNAs ( Figure 5B). As in steadystate condition, proteins encoded by these mRNAs are known to be particularly involved in protein trafficking, processing and destruction, as well as inflammatory pathways. These proteins eoncoded by mRNAs are highly involved in controlling gene transcription and post-transcriptional regulation, while playing a specific role in angiogenesis and osteocartilaginous changes.

| D ISCUSS I ON
In this study, we demonstrate that episodic blood interaction is likely to induce durable epigenetic changes in specific resident cells that do not normally interact directly with blood. In haemophilia, hemarthrosis is the most common complication affecting haemophilic patients, and repeated intra-articular bleeds lead to progressive joint destruction. 34 The resident joint cells, namely the FLS, are repeatedly exposed to blood during flare-ups of hemarthrosis. HA is a unique form of arthritis characterized by a repeated interaction between blood and FLS. Currently, there are two theories that are not exclusive, as for the nature and type of this arthritis. On the one hand, HA may result from a mechanical process much like osteoarthritis with little involvement of inflammation. On the other hand, HA may be envisioned as an inflammatory process underlying HA similar to that of RA. [35][36][37] Our results provide evidence for the second hypothesis.
Given that so far little data on the use of primary FLS from haemophilic patients are available, we have first characterized FLS and shown that they were entirely composed of synoviocytes. The results of flow cytometry confirmed that the primary FLS exhibited exactly the same profile than the FLS cell line, with no marker of cells of other origin found, in particular no macrophages ( Figure S1).
Therefore, the cytokines, mRNAs and miRNA epigenetic changes observed solely involved FLS.

Cytokine expression profile using ELISPOT analysis between
HA-FLS and non-HA-FLS, either activated or not by LPS, seems to be in line with the inflammatory theory ( Figure 1 and Table 1).
Cytokines involved in innate immunity were particularly impacted by the expression difference between non-HA and HA-FLS. We also noticed that HA-FLS were likely to participate in the bone remodelling process and iron recycling, which confirms the cen- IL-1β's role in the pathogenesis of HA has been clearly demonstrated. 38 Indeed, when activated by erythrocyte phagocytosis, monocytes and macrophages were reported to produce IL-1β and TNF-α in the joint after hemarthrosis, which could explain the control of FLS expression by a negative feedback pathway of IL-1α but F I G U R E 4 Chronic plasma exposure in normal HR-FLS lead to the same miRNA profile than HA-FLS. qRT-PCR was performed of 11 selected miRNAs from microarray discovery in HA and non-HA-FLS and also in plasma exposed normal HR-FLS (+ plasma) and non-exposed HR-FLS (-/-), on the condition medium (A) (has-miR-181b-5p, hsa-miR-25-3p, has-miR-1271-5p, hsa-miR-146b-5p and hsa-miR-196b-5p) and on the condition LPS (B) (hsa-miR146a-5p, hsa-miR-6511b-5p, hsa-miR1246, hsa-miR-1271-5p, hsa-miR-585-5p and hsa-miR34a-3p). miRNA expression was verified using miScript SYBER Green Kit and miScript Primer Assays. The 2-ΔΔCt method was used to calculate relative changes in miRNA expression. Results were normalized to SNORD and RNU. Data are expressed as the mean of triplicate samples ± SD and are representative of three independent experiments (***P < .0001, **P < .001, *P < .01) F I G U R E 5 Transcriptomic network potentially impacted by miRNA expression in HA-FLS: from inflammatory to ubiquitous changes. The extracted RNAs were compared using a DNA microarray containing 2578 probes complementary to miRNAs of human origin. Chord diagram of the regulated genes (A: Medium; B: LPS) shared between the miRs. Each chord represent a number genes regulated by two miRs. The total number of regulated genes by each miR is mentioned between brackets. of FLS in RA, non-expression of TNF-α has been reported due to epigenetic regulation. 17 Of note is that TNF-α is likewise key to inflammation in HA, as recently published with regard to the role of the iRhom2/ADAM17/TNF-α pathway 39 and also as a crucial mediator of proliferative synovitis in haemophilia A. 39 In contrast to TNF-α and IL-1α, we did observe high IL-6 expression levels after LPS treatment in HA-FLS, in line with previous data that demonstrated the utility of targeting IL-6 in HA. 41 Interestingly, the cytokine profile of LPS-induced HA-FLS mostly correlated with the potential post-transcriptomic control of the 30 miRNAs selected ( Figure 1 and Table 3). After LPS activation, we observed that HA-FLS up-regulated the expression of cytokines involved in macrophage polarization (also control by miR-1246) and modulated the TH17 pathway (miR-10b-5p) involved in ankylosing spondylitis.
Moreover, most of the other miRNAs may influence inflammatory cytokine expression directly or indirectly, for instance through the NFkB, PI3K or JAK/STAT pathways (miR-146 and miR-196b-5p, the miRNAs with the highest expression ratios).
We confirm that the FLS cytokine profile was influenced by mRNA instability, especially in HA-FLS after actinomycin D treatment ( Figure 3). However, there is more than one process involved in mRNA stability, particularly the 5' and 3' UTR mRNA regulation region with RNA-binding proteins or miRNA complexes. To demonstrate that miRNAs were primarily involved in mRNA instability in HA-FLS, we decided to silence DICER, an essential enzymatic step for miRNA maturation. Following siRNA DICER transfection, LPSinduced HA-FLS did not exhibit proinflammatory mRNA instability after actinomycin D treatment ( Figure 3A). The transcriptomic network of the selected miRNAs in LPS-induced HA-FLS revealed the inflammatory process and protein synthesis to be severely impacted ( Figure 5B and Table 3). Moreover, on analysing the transcriptomic network of the selected miRNAs in steady-state HA-FLS without LPS activation, we observed that pathways involved in angiogenesis, osteocartilaginous changes and inflammatory control process were impacted without any stimulation (Table 2 and Figure 5A).
To confirm that repeated hemarthrosis could induced those miRNA epigenetic changes into articular resident cells, we proposed to create an in vitro model of haemophilic arthropathy with normal HR-FLS repeatedly exposed to plasma every 3 days for 3 weeks. We observed that selected miRNAs acquired the same expression patterns as the HA-FLS, in medium and LPS conditions ( Figure 4A et   4B). It would also be interesting to determine the 'plasma free time' needed for the plasma exposed HR-FLS to restore the same miRNA profile than non-exposed. However, FLS culture conditions and passages induce major limiting factors to perform this experiment.
Indeed, in this model, we used normal plasma samples with the same volume and duration of exposure. In a clinical point of view, when a haemophilic patient present intra-articular bleeding, blood in contact with the synovial membrane could have a variable volume and duration of exposure, with the presence of blood cells and different concentration of proteins especially coagulation factors. These elements could also modulate FLS signature after blood exposure, and this could be an explanation of variability of joint damage and degree of haemophilic arthropathy between each patient.
These findings confirm the clinical theory that just a few repeated episodes of hemarthrosis are sufficient to cause irreversible changes in the joint, while triggering the process of progressive articular degradation. Targeting the goal of 'very few or no bleeds' in haemophilic patients is therefore justified, primarily designed to avoid these irreversible epigenetic changes in synoviocytes.
Recently, some data demonstrated that other triggers than classical FVIII or FIX concentrates could protect haemophilic patients from joint bleeding, such as thrombin-activatable fibrinolysis inhibitors or TAFI, which underscores the relevance of clot protection in addition to clot formation. 42 Different prohemostatic strategies must thus be considered in an effort to avoid definite haemophilia joint bleeding.
It could be interesting to extend this concept of interaction between blood and cells to other cell types, which, under normal conditions, do not interact with blood, such as neuronal and microglial cells following repeated intracranial bleeds in certain sports (boxing, American football, etc).
In conclusion, these findings demonstrate that resident cells in the joints of haemophilic patients develop a unique proinflammatory pattern, close to rheumatoid arthritis, due to durable miRNA changes owing to repeated episodes of hemarthrosis.

This study was supported by the Investigator-Initiated Research
Grant from Baxalta US Inc, now part of Shire, Bannockburn, IL, USA.

CO N FLI C T O F I NTE R E S T
None of the authors have any conflicts of interest to disclose, relating to this study.