Myostatin and CXCL11 promote nervous tissue macrophages to maintain osteoarthritis pain

.


Introduction
Osteoarthritis (OA) is a highly prevalent chronic joint disease with increasing prevalence due to the aging and obese world population (Allen, Thoma, & Golightly, 2022;Bijlsma, Berenbaum, & Lafeber, 2011;Safiri et al., 2020).Pain is the primary symptom of OA causing disability and a substantial socioeconomic burden (Neogi, 2013).Current treatments against OA pain are not effective, and some, such as opioids, are associated with long-term adverse effects (Suzuki et al., 1987).Clinical findings show that pain in OA does not correlate well with radiographic damage in the affected joint (Ostojic et al., 2019;Trouvin & Perrot, 2018).Moreover, in patients with a likely neuropathic component, as determined with the painDETECT questionnaire, local and overall joint damage appears to not be the main cause of pain and disability (van Helvoort et al., 2021).Importantly, pain persists in 44 % of OA patients after total knee replacement (Wylde et al., 2011), suggesting that there must be mechanisms other than the local knee joint injury that contribute to OA pain.
At the early stages of OA, pain is thought to be initiated by the activation and sensitization of nociceptive fibers because of pathological changes in the injured joint (Fingleton et al., 2015;Steen Pettersen et al., 2019).However, at later stages central mechanisms also contribute to OA pain (Gwilym et al., 2009;Lluch et al., 2018;Pinto, Lima, & Tavares, 2007).Recently, it has become clear that macrophages have important functions in the regulation of pain.Macrophages with a tissue-healing phenotype (M2-like) attenuate or even actively resolve pain in different rodent models of pain (Raoof, Willemen, & Eijkelkamp, 2018;van der Vlist et al., 2022).Conversely, in animal models of chronic pain, including OA, macrophages with an inflammatory (M1-like) phenotype accumulate in the DRG and contribute to persistent pain (Miller et al., 2012;Raoof et al., 2021;Simeoli et al., 2017).As example, reducing F4/ 80 + iNOS + (M1-like) macrophages in the DRG, or injecting F4/ 80 + CD206 + (M2-like) macrophages intrathecally, reduced painassociated behaviors in a rodent model of OA (Raoof et al., 2021).
Overall, these data suggest that the phenotype of macrophages affects their regulatory functions within the nervous system.However, until now it remains unknown how macrophages are engaged in the DRG during OA to maintain pain.
Therefore, we assessed the neuro-immune mechanisms in OA that govern DRG macrophage accumulation and their programming into proalgesic macrophages.

Animals
All experiments comply with the ARRIVE guidelines, were performed in accordance EU Directive 2010/63/EU for animal experiments and were approved by the local experimental animal welfare body and the national Central Authority for Scientific Procedures on Animals (CCD, license numbers AVD115002015323 and AVD11500202010805).
Adult (age 8-15 weeks) male and female C57BL/6J or Lysm cre/+ x Csf1r DTR/+ (Cat# 024046, Jackson laboratories) (MM dtr ) in a C57Bl/6 background were used and maintained in the animal facility of the University of Utrecht.
Mice were housed in groups under a 12:12 h light/dark cycle, with food and water available ad libitum.The cages contained environmental enrichments including tissue papers and shelter.To minimize bias, animals were randomly assigned before the start of experiment using RandoMice.We used baseline mechanical sensitivity, sex and age as factor to equally distribute between groups.Before each experiment the number of mice required was calculated using GPower with an effect size of 1.5-2, power of 80-90 %, and an alpha of 5 % (corrected for multiple testing when needed).

Pain-associated behavioral tests
1-2 weeks prior to the start of experiments, mice were first acclimatized to the testing environment by placing them in the test environment during 1 h.At least 3 baseline measures were performed at different days during one week before starting the experiment, with the last one at the day of the start of the experiment.Experiments were performed in the same room and same test set-up over the duration of the experiment.Involved experimenters were well trained to perform the assays and were blinded for the experimental groups and genotypes.
To avoid variation, always the same experimenter performed pain behavioral tests in each experiment.Mice were put into the testing environment (von Frey) at least 30 min before performing measurements with the experimenter present in the same room.
Mechanical thresholds were assessed in both hind paws using the von Frey test (Stoelting, Wood Dale, IL, US) with the up-and-down method to determine the 50 % threshold.In short, von Frey filaments were applied for 5 s to the plantar surface of the paw.After applying the first filament (0.4 g), in case of a non-response the next filament with a higher force was used.In case of a response, the next lower force filament was used.A minimum of 30 s between the application of filaments was taken.Four readings were obtained after the first change of direction.Von Frey filaments were applied perpendicularly, smoothly, without moving the filaments horizontally during application.If the response was ambiguous, the measurement was repeated at least one minute later.
Changes in weight bearing were evaluated using the dynamic weight bearing (DWB) apparatus (Bioseb, Vitrolles, France) and the following parameters were used for the analysis: 1) low weight threshold of 0.5 g, 2) high weight threshold of 1 g, 3) surface threshold of 2 cells and 4) minimum 5 images (0.5 s) for stable segment detections.The device consists of a small Plexiglas chamber (11.0 x 19.7 x 11.0 cm) with a sensor mat containing pressure transducers.The system records the average weight that each limb exerts on the floor.Mice were placed in the chamber at least 1 min prior to starting the measurements and were allowed to move freely within the chamber.Each mouse was recorder for period of 5 min.A camera at the top of the enclosure was used to record all movements to ensure accurate validation of the position of the mouse by the experimenter.The relative weight bearing of the affected paw (ipsilateral paw) was expressed as ratio between body weight bore on ipsilateral paw divided by body weight bore on the unaffected paw (contralateral paw).

Depletion of monocytes and macrophages
To deplete monocytes and macrophages in vivo, MM dtr mice received a first intraperitoneal injection of 20 ng/g body weight diphtheria toxin (DT; Sigma-Aldrich) followed by daily intraperitoneal injections of 4 ng/ g body weight DT at all subsequent days as described previously (Schreiber et al., 2013).
To obtain L929 cell-conditioned medium, 10x10 6 L929 (ATCC CCL-1) cells were seeded in a 75 cm 2 flask with cell culture-medium supplemented with 1 % non-essential amino acids (Sigma-Aldrich) for one week.L929 cells were passaged to a 162 cm 2 flask with 50 ml medium and after a week the supernatants were collected and filtered through a 0.2-µm filter and stored at − 20 • C (L929-derived M− CSF).

Flow cytometry analysis
DRG (L3-L5) or spinal cord were collected to analyze infiltrating immune cells.In brief, tissues were gently minced and dissociated at 37 • C for 30 min with an enzyme cocktail (1 mg collagenase type I with 0.5 mg trypsin in 1 ml DMEM; Sigma-Aldrich).Cells were stained with various combinations of fluorochrome-labeled antibodies (Table 1).Absolute number of cells was calculating using counting beads (Count-Bright™ Absolute Counting Beads, for flow cytometry, Cat# C36950, ThermoFisher).A defined set of counting beads was added to the DRG cell suspension.Based on the number of measured beads during flow cytometry the total number of cells in the DRG cell suspension was calculated.

Immunofluorescent stainings
Mice were killed by cervical dislocation and DRG were collected.For some experiments, the retrograde marker Wheat Germ Agglutinin (WGA) was injected intra-articular 48 h prior to killing the mice (WGA-594, Cat.# W11262, ThermoFisher) to identify sensory neurons innervating the knee joint.
Cryosections (10 µm) of lumbar DRG (L3-L5) were cut on a cryostat (Leica CM3050 S Cryostat), post-fixed in 4 % PFA during 5 min and stained with primary antibodies overnight at 4 • C, followed by 2 h incubation with fluorescent-tagged secondary antibodies (Table 2).Nuclei were counterstained with or without 4,6-diamidino-2-phenylindole (DAPI).Immunostaining images were captured with an Olympus IX83 microscope using identical exposure times for all slides within one experiment.

PCR array
Mature RNA was isolated as described above.RNA quality was determined using a spectrophotometer and was reverse transcribed using a cDNA conversion kit.cDNA from 5 mice belonging to the same group were pooled.The cDNA was used on the real-time RT 2 Profiler PCR Array (Cat# PAMM-150Z, QIAGEN) in combination with RT 2 SYBR® Green qPCR Mastermix (Cat# 330529,QIAGEN).Fold change/ regulation were calculated using delta delta CT method, in which delta CT is calculated between gene of interest and an average of reference genes, followed by delta-delta CT calculations (delta CT (Test Group)delta CT (Control Group)).Fold Change is then calculated using 2^ (-delta delta CT) formula.Samples passed PCR Array Reproducibility, RT Efficiency and Genomic DNA Contamination quality checks.For normalization, CT values of β-actin, β2-microglobulin, GAPDH, β-glucuronidase and HSP 90β expression were used to calculate the average of reference genes.

Real-time RT-PCR
Quantitative real-time PCR reactions were performed using a QuantStudio 3 (ThermoFisher) following the manufacturer's instructions.We used 1-5 ng cDNA input per qPCR reaction and used the mean of 18S rRNA, β2-microglobulin and TBP expression as reference.Primers used are included in Table 3.

Statistical analysis
All data are presented as mean ± SD and were analyzed with GraphPad Prism version 9.3 using unpaired two-tailed t-tests, one-way or two-way ANOVA, or as appropriate two-way repeated-measures ANOVA, followed by post hoc analysis.The used post hoc analyses are indicated in each figure legend.Data are represented as mean ± SD.A p  value less than 0.05 was considered statistically significant.

Data availability
All data are available in the article or in Supplementary materials.Raw data and materials are available on reasonable request.Some materials used in this manuscript are subject to a material transfer agreement.

F4/80 + iNOS + macrophages accumulate in the DRG during early OA and maintain pain
Knee joint damage was induced by an unilateral intra-articular injection of MIA (10 % w/v).At day 3 after MIA-induced OA, mice showed signs of pain-like behaviors such as mechanical hypersensitivity as assessed with von Frey test, and weight bearing deficits assessed with dynamic weight bearing (Fig. 1A/B).Earlier findings show that macrophages are at least present from one week up to 4 weeks after induction of OA with MIA (Raoof et al., 2021).To investigate whether DRG macrophages accumulate earlier during initiation of OA pain, the L3-L5 DRG that contain the somas of sensory neurons innervating the knee joint, were analyzed using flow cytometry (Fig. S1).Macrophages were gated based on CD45 + Ly6G − CD11b + F4/80 + , further referred to as F4/80 + DRG macrophages.The number of F4/80 + DRG macrophages was increased at day 3 and day 7 after intra-articular MIA injection (Fig. 1C).Ly6G + neutrophils, gated based on CD45 + Ly6G + , and other immune cell numbers, gated based on CD45 + Ly6G − CD11b − F4/80 − , were also significantly increased in the DRG at these time points (Fig. 1D).
Next, we selectively depleted monocytes and macrophages by daily intraperitoneal injections of diphtheria toxin (DT) in Lysm cre/+ x Csf1r DTR/+ mice (Schreiber et al., 2013), referred to as ''MM dtr ''.Daily intraperitoneal DT injection reduced the number of F4/80 + macrophages in the DRG and spinal cord compared to saline injected MM dtr mice at day 7 after injection of MIA (Fig. 1E/F).In macrophage-depleted mice, MIA-induced mechanical hypersensitivity and weight bearing deficits were indistinguishable from non-depleted mice until day 7.However, at day 7 after MIA injection, macrophages-depleted mice had progressively resolved from mechanical hypersensitivity and deficits in weight bearing, and stayed resolved until day 13 after MIA injection, compared to non-depleted mice (Fig. 1G/H).
To further support the role of DRG macrophages to maintain OA pain, 30.000 (van der Vlist et al., 2022), DT resistant, wild type bone marrow-derived macrophages were injected intrathecally into macrophages-depleted MM dtr mice 7 days after MIA injection.Proinflammatory, M1-like macrophages were gated based on CD45 + Ly6G − CD11b + F4/80 + iNOS + , further referred to as F4/80 + iNOS + macrophages.Intrathecal injection of bone marrow-derived macrophages resulted in an increase in the number of total F4/80 + macrophages and F4/80 + iNOS + macrophages in the DRG (Fig I /J).Intriguingly, these macrophages were sufficient to reinstate mechanical hypersensitivity in the ipsilateral paw in macrophage-depleted MM dtr mice, but not in the contralateral paw (Fig. 1K).Similarly, intrathecal macrophages reinstated weight bearing deficits in macrophage-depleted MM dtr mice (Fig. 1L).

CXCL11 and myostatin expression in the DRG during early OA
To identify which factors may drive the accumulation and programming of macrophages in the DRG during OA we used a PCR array for 96 mouse cytokines and chemokines (Table S1).Analysis of DRG transcript at 3 days after intra-articular MIA injection identified CXCL11, IL-6, IL-18 and myostatin as top 4 of most highly differentially expressed genes.As CXCL11 is known as chemoattractant for monocytes/macrophages (Zeng et al., 2016), and myostatin can inhibit macrophages pro-inflammatory programming in adipose and muscle tissue (Dong et al., 2016), we focused further on these 2 novel mediators as potential targets of interest.
To validate the increase in CXCL11 and myostatin expression, we determined CXCL11 and myostatin mRNA levels in the DRG using qPCR and using a larger experiment.At day 3 after MIA injection, we confirmed that expression of CXCL11 was ~ 1.5 fold higher than controls and myostatin mRNA was increased with ~ 4-fold.At 7 and 21 days after MIA injection, CXCL11 expression did not differ from DRG from control mice.Myostatin mRNA expression was doubled compared to controls at days 7 and 21, but not significant (Fig. 2A/B).
To confirm the increase of CXCL11 and myostatin protein expression in the DRG, we measured CXCL11 and myostatin after immunofluorescent staining of the DRG from control mice and mice at day 3 and 21 after MIA injection.CXCL11 immunofluorescence was increased in the ipsilateral lumbar DRG (L3-L5) compared to the DRG of mice without MIA injection at days 3 and 21 after MIA injection (Fig. 2C).CXCL11 expression appeared to colocalize in part with FABP7 and GFAP (satellite glial cell markers) (Fig. 2D), suggesting satellite glial cells may be a source of CXCL11.Staining with isotype IgG did not give any visible staining (Fig. S2).
Myostatin expression, in contrast, colocalized with β3-tubulin positive cells in the DRG (Fig. 2E), suggesting myostatin is produced within sensory neurons.Anti-myostatin IgG blocked recombinant myostatin did not give any staining (Fig. S3).To quantify neuronal myostatin expression, we binned sensory neurons according to their cross-sectional area.Myostatin expression was increased at day 3 and day 21 in small DRG neurons (0-300 µm 2 ).In medium (300-600 µm 2 ) and large (600-2000 µm 2 ) DRG neurons myostatin expression was increased at day 21, but not day 3 (Fig. 2F).To assess whether the observed changes in myostatin expression occur in neurons innervating the affected knee joint, we injected the retrograde tracer, wheat germ agglutin (WGA) 2 days priors to isolation of DRG.Myostatin fluorescent intensity was Fig. 1.DRG macrophages maintain pain but are not involved in the initiation of OA pain.Course of (A) mechanical hypersensitivity and (B) weight bearing of ipsilateral hind paw after an unilateral intra-articular injection of MIA.As control, mice received saline in the contralateral knee.Two-way ANOVA with Sidak's post hoc.Flow cytometry of a cell suspension of ipsilateral lumbar DRG (L3-L5) of (C) CD45 + CD11b + Ly6G − F4/80 + macrophages, (D) CD45 + Ly6G + neutrophils and other immune cells defined as CD45 + Ly6G − CD11b − F4/80 − comparing day 0 with day 3 and 7 days after MIA injection.One-way ANOVA with Dunett's post hoc.To deplete monocytes and macrophages, mice received daily intraperitoneal DT injections starting one day prior intra-articular injection of MIA.Flow cytometry analysis of a cell suspension showing total CD45 + CD11b + Ly6G − F4/80 + macrophages in (E) ipsilateral lumbar DRG (L3-L5) and (F) lumbar section from spinal cord of MM dtr mice injected intraperitoneally with DT or saline at day 7 after injection with MIA.Gating strategy is displayed in Fig. S1.Unpaired two-tailed t-test.Course of (G) mechanical hypersensitivity and (H) weight bearing on ipsilateral hind paw in MM dtr mice injected daily with DT or vehicle.Two-way ANOVA with Dunnet's post hoc comparing vehicle and DT treatments at each time point.MM dtr mice injected with DT starting one day prior intra-articular injection of MIA in the ipsilateral knee and saline in the contralateral knee, received intrathecal exogenous macrophages or vehicle at day 7. Analysis of the number of (I) total CD45 + CD11b + Ly6G − F4/80 + macrophages and (J) CD45 + CD11b + Ly6G − F4/80 + iNOS + macrophages in the ipsilateral lumbar DRG (L3-L5) at 24 h after this intrathecal injection.Unpaired two-tailed t-test.Course of (K) mechanical hypersensitivity and (L) weight bearing on ipsilateral hind paw in MM dtr mice injected with DT starting one day prior unilateral intra-articular injection of MIA.At day 7, mice received intrathecal exogenous macrophages or vehicle.Two-way ANOVA with Sidak's post hoc.Open dots (○) represent males and closed dots represent females (•) in column graphs from panels B, C, E, F, H, I, J and L. A p-value less than 0.05 was considered statistically significant and each significance is indicated with *p < 0.05, **p < 0.01, ***p < 0.001.

Role of CXCL11 in OA pain
To address whether CXCL11 in DRG is required for the persistence of OA pain, we neutralized endogenous CXCL11 by administration of daily intrathecal injections of anti-CXCL11 for five days starting the day of MIA injection.Intrathecal injection of anti-CXCL11 IgG prevented the development of persistent mechanical hypersensitivity and weight bearing deficits compared to mice injected with a control IgG antibody.In contrast, the initiation of OA pain until day 3 was not affected by the inhibition of CXCL11 compared to the control IgG antibody treatment (Fig. 3A/B).Neutralization of CXCL11 reduced the number of F4/80 + macrophages in the DRG (Fig. 3C).In addition, the number of F4/ 80 + iNOS + macrophages in the DRG was reduced with 33 % compared to control IgG-treated mice (Fig. 3D).Ly6G + neutrophils and other immune cells in the DRG remained unaffected after CXCL11 inhibition (Fig. 3E).
Next, we tested whether CXCL11 intrathecal injection is sufficient to induce pain-like behaviors and DRG macrophage accumulation.Intrathecal injection of 100 and 500 ng CXCL11 (Piotrowska et al., 2018) dose dependently increased the number of F4/80 + macrophages in lumbar DRG (Fig. 3F).In contrast, these doses of CXCL11 did not induce mechanical hypersensitivity (Fig. 3G), nor increased the number of F4/ 80 + iNOS + macrophages or other immune cells in the DRG (Fig. 3H/I).These data suggest that CXCL11 alone is not sufficient to induce painassociated behaviors.

Myostatin maintains OA pain
Next, we assessed the role of myostatin in OA pain by daily intrathecal injections of neutralizing anti-myostatin IgG, starting just prior to intra-articular MIA injection for five days.Anti-myostatin IgG resolved MIA-induced mechanical hypersensitivity starting from day 3, whilst control IgG had no effect.Seven days after MIA injection mechanical  hypersensitivity fully resolved in anti-myostatin IgG treated mice (Fig. 4A).Similarly, at day 7 after MIA, weight bearing deficits were normalized in mice that received anti-myostatin IgG (Fig. 4B).Inhibition of myostatin did reduce the number of F4/80 + iNOS + macrophages in the DRG with 56 % compared to mice treated with a control IgG (Fig. 4C), without affecting the total number F4/80 + macrophages, Ly6G + neutrophils or other immune cells in the DRG (Fig. 4D/E).
Intrathecal injection of either 100 ng or 500 ng myostatin induced mechanical hypersensitivity within 3 h that persisted for at least 24 h (Fig. 4F).Injection of 500 ng myostatin significantly increased the total F4/80 + macrophages, F4/80 + iNOS + macrophages, Ly6G + neutrophils and other immune cells (CD45 + but negative for macrophage and neutrophil markers) number in the DRG compared to the vehicle (Fig. 4G/H/I).
To determine whether myostatin-induced pain requires macrophages, we injected myostatin in monocyte/macrophages-depleted mice.MM dtr mice received 4 daily intraperitoneal injections of DT prior to intrathecal injection of recombinant myostatin.Depletion of macrophages prior intrathecal injection of myostatin completely prevented myostatin-induced increase in F4/80 + iNOS + DRG macrophages, while Ly6G + neutrophils and other immune cells remained unaffected (Fig. 4J/K).Importantly, macrophages depletion completely prevented myostatin-induced mechanical hypersensitivity (Fig. 4L).
Because myostatin neutralization reduced mechanical sensitivity earlier in time than CXCL11 neutralization (Fig. 3A vs 4A), we hypothesized that myostatin may be upstream of CXCL11.To test this hypothesis, we intrathecally injected myostatin whilst neutralizing CXCL11 with anti-CXCL11.Administration of anti-CXCL11 together with myostatin significantly reduced the development of myostatininduced pain compared to mice injected intrathecally with myostatin and control IgG antibody (Fig. 4M).

Inhibition of myostatin resolves persistent OA pain
To assess if targeting myostatin is a therapeutic approach to treat persistent OA pain, mice received daily intrathecal injections during five days with neutralizing anti-myostatin IgG starting at 21 days after induction of OA.Anti-myostatin IgG resolved MIA-induced mechanical hypersensitivity from 3 days onwards after starting the intrathecal treatment.At day 27, mechanical hypersensitivity had fully resolved and weight bearing deficits were normalized in mice that had received anti-myostatin IgG (Fig. 5A/B).At day 32, the number of total F4/80 + macrophages, F4/80 + iNOS + macrophages, Ly6G + neutrophils and other immune cells in the DRG had fully normalized in anti-myostatin IgG treated mice (Fig. 5C/D/E).

Discussion
In this study we focused on the neuro-immune mechanisms that contribute to the accumulation and programming of macrophages that maintain pain-associated behaviors in a mouse model of OA.We found that macrophages with an inflammatory phenotype accumulate in the lumbar DRG, innervating the damaged knee, early during OA development.These macrophages are not required for the initiation, but rather for the maintenance of OA pain, independent of the knee joint damage.Intriguingly, we identified myostatin is produced by sensory neurons, whilst CXCL11 appeared to be expressed by some satellite cells, and together contribute to the accumulation of F4/80 + iNOS + macrophages in the DRG of OA mice and regulate the maintenance of OA pain.Only intrathecal administration of myostatin, in naive mice was sufficient to increase F4/80 + iNOS + macrophages in the DRG and provoke painassociated behaviors in a CXCL11-dependent manner.In contrast, CXCL11 contributes to the accumulation of total macrophages in the DRG, but was not sufficient to induce pain-associated behaviors.Therefore, we conclude that likely a range of factors contribute to neuron-macrophages communication in the maintenance of OA pain, where CXCL11 contributes the accumulation of DRG macrophages and together with myostatin plays and essential role in programming these macrophages into a pain maintaining pro-inflammatory phenotype.Finally, we identified that inhibition of myostatin may represent a novel unexpected therapeutic target for the treatment of chronic OA pain.
Different studies have pointed to DRG macrophages in the regulation of various pain conditions.DRG macrophages resolve inflammatory pain by transferring mitochondria to sensory neurons (van der Vlist et al., 2022).In contrast to a role in inflammatory pain resolution, DRG macrophages initiate and maintain neuropathic pain (Yu et al., 2020).More recently, OA was added to this list.The number of macrophages in the DRG is increased in different rodent models of OA (Miller et al., 2012;Raoof et al., 2021;Raoof, Willemen, & Eijkelkamp, 2018).Here, we identified that these macrophages are key for the maintenance of OA pain, whilst they are not needed for the initial development of pain induced by joint damage.This specific role of macrophages in only the maintenance of OA pain differs from that found in neuropathic pain, where DRG macrophages are required for the initiation and maintenance of neuropathic pain (Yu et al., 2020).This indicates that neuroimmune interactions may be pain condition-specific in the initiation phase, but the pain maintaining role of macrophages may be shared across pain conditions.
Macrophages depletion reduced pain-associated behaviors from day 3 onwards, whilst the number of macrophages, myostatin and CXCL11 are already increased at day 3 in the DRG.This questions what else than macrophages is driving pain at day 3 or earlier.Likely, knee tissue damage and the resulting DAMPS and inflammatory molecules cause pain-associated behaviors during the early phase of OA.As example, OA chondrocytes express cytokines, including IL-1β, TNF and TGFβ, that either directly or indirectly cause pain by activating nociceptors (Blaney Davidson et al., 2015;Manni et al., 2003;Melchiorri et al., 1998;Takano et al., 2019).In line with an inflammatory component driving pain, is that paracetamol and NSAIDs reduce pain at the first day of MIAinduced OA while at the later phase they do not (de Sousa Valente, 2019).Thus, inflammatory and tissue damage mediators in the joint likely cause pain at the early stage of OA (Vincent, 2020) by direct activation and sensitization of joint nociceptors independent of macrophages.Whether these early inflammatory triggers also engage the myostatin-CXCL11-macrophages axis in the DRG to maintain pain remains to be determined.
CXCL11 neutralization prevented the development of persistent OA pain, but intrathecal administration of CXCL11 was not sufficient to induce pain.CXCL11 is a well-known chemoattractant that activates CXCR3, CXCR4 and CXCR7, expressed on monocytes, neurons and glial cells (Bhangoo et al., 2007;Biber et al., 2002;Groom & Luster, 2011;Odemis et al., 2010;Oh et al., 2001;Shimizu et al., 2011;Singh et al., 2013).CXCR3 and CXCR4 are increased in the DRG in different rat models of neuropathic pain, and receptor antagonists reduced mechanical hypersensitivity in these models (Bhangoo et al., 2007;Chen et al., 2019).Although we did not verify receptor levels in our model, other studies in OA models (destabilization of the medial meniscus; mechanical joint loading) did not identify changes in CXCR3 and CXCR4 expression (Bangash et al., 2018;Miller et al., 2020).Different studies showed that spinal cord astrocytes express CXCL11 following spinal nerve injury (Guo et al., 2017;Piotrowska et al., 2018;Wu et al., 2018).Although we observed an increase in CXCL11 immunofluorescence in the DRG of mice with experimental OA that colocalized with some satellite glial cells markers, future work will need to assess whether satellite glial cells actually release CXCL11 or other cells may also express and release CXCL11, as not all CXCL11 immunofluorescence colocalized with satellite glial cells markers, and whether myostatin is a possible inducer of CXCL11.In line with our observations, CXCL11 is also upregulated relatively early, 2 days, after nerve damage, and blocking CXCR3 reduced allodynia in chronic constriction injury-induced rat model of neuropathic pain (Piotrowska et al., 2018).A study found that intrathecal injection of 400 ng of CXCL11 induced pain-associated behavior at 1.5 h after intrathecal injection (Piotrowska et al., 2018).We did not observe that a similar dose of intrathecal CXCL11 produced pain-associated behaviors at either 1 or 3 h after injection.Possibly CXCL11 only induces a very short episode of pain through direct actions on sensory neurons.The same study, however, did not observe CXCL11induced allodynia 3 h after injection.Despite absence of pain behaviors at 24 h after injection, CXCL11 was sufficient to increase the number of macrophages in the DRG.Yet, macrophages did not acquire a pro-algesic F4/80 + iNOS + phenotype, possibly explaining why CXCL11 did not induce pain.In support of the requirement of F4/80 + iNOS + macrophages to induce pain, intrathecal injection of in vitro stimulated proinflammatory F4/80 + iNOS + macrophages, but not of unstimulated macrophages, into WT mice induces mechanical hypersensitivity (Raoof et al., 2021).
IL-6 and IL-18 were also in the top 4 list of increased factors in mice DRG during OA pain establishment.Although we did not focus on the two other proteins to test for their contribution, IL-6 is a classical cytokine expressed by pro-inflammatory macrophages (Qin et al., 2012).Moreover, there are several studies linking IL-6 protein production and mRNA expression in the DRG with pain (De Jongh et al., 2003).IL-18 has also been linked to inflammation, but requires cleavage from its precursor form (pro-IL-18) to its active form, which is initiated by NLRP3 inflammasome activation (van de Veerdonk et al., 2011).We recently identified a potential role of IL-18 and inflammasome activation in OA pain specifically in male mice (Ribeiro et al., 2023).Whether myostatin, CXCL11 and IL-18 are all required in concert remains to be determined.
In this study we identified a hitherto novel role of myostatin in promoting pro-algesic F4/80 + iNOS + macrophages in the DRG and pain maintenance in OA.Myostatin is an extracellular cytokine member within the transforming growth factor-β (TGF-β) family and mostly expressed in skeletal muscles.It plays a crucial role in the negative regulation of muscle mass.Upon the binding to activin receptor type-2b (Acvr2b) (Lee & McPherron, 2001), myostatin initiates several different signaling cascades resulting in the upregulation of genes that promote iatrogenesis of muscles, whilst downregulating important myogenesis genes.Inhibition of myostatin, genetically or pharmacologically, with anti-myostatin antibody or propeptide, increases skeletal muscle mass (Camporez et al., 2016;Collins-Hooper et al., 2014;McPherron, Lawler, & Lee, 1997).Although its role in the sensory system has not been described, single cell RNAseq indicates that sets of sensory neurons express myostatin (Usoskin et al., 2015).Here we also confirmed that, during OA, sensory neurons innervating the affected joint but also other neurons express myostatin, with the largest increase in expression predominantly in small diameter neurons.But how does myostatin induce pain?Mice deficient in myostatin have decreased osteoclast numbers, increased cortical thickness, cortical tissue mineral density in the tibia, and increased vertebral bone mineral density (Omosule & Phillips, 2021).Additionally, there is some evidence that links myostatin with inflammation and macrophages infiltration and programming.Inhibition of myostatin reduced expression of TNF, IL-6 and F4/80 in muscles from mice with chronic kidney disease (Zhang et al., 2011) and obesityinduced insulin resistance (Dong et al., 2016;Wilkes, Lloyd, & Gekakis, 2009).Myostatin inhibition also suppresses macrophages infiltration in adipose tissue, while in vitro stimulation of macrophages with myostatin promoted IL-6 production (Dong et al., 2016).In line with these results, we observed that neutralization of myostatin during developing or established OA prevented accumulation of pro-inflammatory F4/ 80 + iNOS + macrophages in the DRG.Furthermore, intrathecal myostatin administration increased the number of total and F4/80 + iNOS + macrophages in the DRG.Therefore, myostatin's role regarding DRG macrophages appears not to be restricted to pro-inflammatory programming, but also to their accumulation.In line with these findings, we show that myostatin-induced pain-associated behaviors required CXCL11, possibly explaining as to how myostatin promotes macrophage accumulation.Overall, these findings extent the role of the cytokine myostatin beyond an immune regulator in adipose tissue, to one that involves neuro-immune communication in nervous tissues.Despite the hitherto unknown role of myostatin in OA pain, we do not know whether myostatin directly programs macrophages into a proalgesic F4/80 + iNOS + macrophages, or whether this involves other cells present in the DRG.We show that macrophages stimulated with myostatin do not express the M1-like marker iNOS (Fig. S5), whilst classically stimulated macrophages with LPS and IFNγ do (Raoof et al., 2021).This would be in line with recent sequencing data of nervous tissue macrophages that shows that these cells hardly express activin receptor type-2a (Acvr2a) or Acvr2b.Possibly, myostatin-driven programming of DRG macrophages to pro-inflammatory phenotype requires an intermediate cell/factor.As examples, subsets of DRG neurons (e.g., neuro-filament+ neurons, peptidergic and non-peptidergic neurons) and satellite glial cells, express Acvr2a, one of the receptors of myostatin (Jager et al., 2020;Usoskin et al., 2015).Possibly, myostatin only promotes an F4/80 + iNOS + phenotype when expressed together with other inflammatory molecules such as CXCL11.The fact that neurons also express myostatin receptor Acvr2a may also explain why neutralization of myostatin has faster effects on pain-associated behaviors than macrophages depletion.Future research has to elucidate whether sensoryneuron derived myostatin acts on satellite glial cells to facilitate programming of macrophages into a pro-algesic state during OA.
A limitation of this study is that we did not assess whether the various treatments affected knee swelling and/or histology.However, most of the treatments were applied with intrathecal injection and were sufficient to reduce pain-associated behaviors.In addition, intrathecal injection of DT in MM dtr mice only depleted macrophages in the DRG, not systemically, but was sufficient to reduce OA-induced mechanical hypersensitivity (Raoof et al., 2021).Therefore, we postulate that it is rather unlikely that our treatments targeting the myostain-CXCL11 axis have affected knee joint pathology.
Intrathecal treatments in this study target the DRG and spinal cord area, so it cannot be completely excluded that these treatments only affect the DRG and not the spinal cord.However, the increase of myostatin is observed in sensory neurons and CXCL11 in some extent in satellite glial cells in the and neutralizing antibodies targeting these proteins reduced MIA-induced mechanical hypersensitivity.These observations suggest that treatment effects are peripherally driven.
Targeting myostatin has been a therapeutic strategy for some diseases associated with debilitating loss of musculoskeletal tissues.This has yielded in the development of myostatin inhibitors, ranging from natural antagonists to antibodies and soluble receptors (Suh & Lee, 2020).Some of these inhibitors, such as BYM-338, an anti-Acvr2b antibody, have been tested in Phase III trials to treat sarcopenia (Rooks et al., 2017) and chronic obstructive pulmonary disease (Polkey et al., 2019).Peptibody AMG-745, developed by Amgen, has been used in trials for treating patients receiving androgen deprivation therapy for prostate cancer (Padhi et al., 2014).Despite studies showed safety, clinical efficacy in various cases have been somewhat disappointing (Padhi et al., 2014;Polkey et al., 2019;Rooks et al., 2017;Rybalka et al., 2020).In this study we show that myostatin neutralizing antibodies resolve persistent OA pain, opening the possibility for a new use of these inhibitors.A limitation, however, is that we have only tested it in the mouse MIA-model of OA, which is a model with fast developing, and vigorous signs of cartilage damage and synovial inflammation.Given that OA is a slowly progressing disease, future research is needed to assess whether myostatin role in pain is consistent among other OA pain models.

Conclusion
We identified that CXCL11 and myostatin are ligands in neuron-tomacrophages communication in a preclinical model of OA (Fig. 6).CXCL11 drives the accumulation of DRG macrophages whilst myostatin predominantly programs DRG macrophages into pro-algesic inflammatory macrophages that maintain OA pain.These findings may explain clinical findings pointing that the severity of OA pain does not correlate well with the tissue changes in the injured joint.In addition, these findings point to completely novel potential therapeutic approaches to treat pain inOA, focused on preventing macrophage accumulation in nervous tissue or preventing their programming into pro-algesic phenotypes.

Fig. 2 .
Fig. 2. Increased CXCL11 and myostatin expression in the DRG of mice with OA. (A) CXCL11 and (B) myostatin mRNA expression in ipsilateral lumbar DRG (L3-L5) comparing control mice (day 0) with mice at days 3, 7 and 21 after MIA injection.One-way ANOVA with Dunnett's post hoc.(C) Quantification of CXCL11 signal in the lumbar DRG (L3-L5) from naive mice (day 0) and ipsilateral DRG from mice at days 3 and 21 after MIA injection.One-way ANOVA with Dunnett's post hoc.(D) Example images showing CXCL11, FABP7 and GFAP fluorescence in the DRG from naive mice (day 0) and ipsilateral lumbar DRG (L3-L5) from mice at days 3 and 21 after MIA injection.Control isotype IgG staining is shown in Fig. S2.(E) Example images showing myostatin and β3-Tubulin signal after immunofluorescent staining of ipsilateral lumbar DRG (L3-L5) of naive mice (day 0) and mice at days 3 and 21 after MIA injection.Control anti-myostatin IgG staining where the antibody was blocked with recombinant myostatin (Cat# 120-00, PeproTech) is displayed in Fig. S3.(F) Quantification of myostatin fluorescence in small (0-300 µm 2 ), medium (300-600 µm 2 ) and large (0-2000 µm 2 ) cross-sectional area neurons from lumbar DRG (L3-L5) of naive mice (day 0) and ipsilateral DRG from mice at days 3 and 21 after MIA injection.Two-way ANOVA with Dunnett's post hoc.(G) Mice were injected intra-articular with WGA to identify the somas of sensory neurons innervating the knee joint.Example images of myostatin, WGA and β3-Tubulin fluorescence in the ipsilateral lumbar DRG (L3-L5) of naive mice (day 0) and mice at 3 days after MIA injection.Isotype IgG was used as control and is displayed in Fig. S4.(H) Quantification of myostatin fluorescent signal in WGA-and WGA + sensory neurons in the ipsilateral lumbar DRG (L3-L5) of naive mice (day 0) and mice at day 3 after MIA injection.One-way ANOVA with Tukey's post hoc.Open dots (○) represent males and closed dots represent females (•) in column graphs from panels A, B, C, F and H.A p-value less than 0.05 was considered statistically significant and each significance is indicated with *p < 0.05, **p < 0.01, ***p < 0.001.White arrows in panel D point to the areas where CXCL11 and FABP7 signal colocalize.White scale bars on the bottom right corner of panels D, E and G represents 100 µm.

Fig. 5 .
Fig. 5. Myostatin neutralization resolves persistent OA pain.Course of (A) mechanical hypersensitivity and (B) weight bearing of ipsilateral hind paw after an unilateral intra-articular MIA injection.Saline was injected into the contralateral knee joint.Five daily intrathecal injections of anti-myostatin IgG (2.5 µg/injection) or control IgG (2.5 µg/injection) were administered starting at day 21 after MIA injection.Two-way ANOVA with Sidak's post hoc.Analysis of the number of (C) total CD45 + CD11b + Ly6G − F4/80 + macrophages, (D) CD45 + CD11b + Ly6G − F4/80 + iNOS + macrophages, (E) CD45 + Ly6G + neutrophils and other immune cells defined as CD45 + Ly6G − CD11b − F4/80 − in the ipsilateral and contralateral lumbar DRG (L3-L5) at day 32 after MIA injection.One-way ANOVA with Dunnett's post hoc.Open dots (○) represent males and closed dots represent females (•) in column graphs from panels B, C and D. A p-value less than 0.05 was considered statistically significant and each significance is indicated with, **p < 0.01, ***p < 0.001.In panel E, asterisks (*) indicate statistically significant differences in number of Ly6G + neutrophils between MIA compared to saline.No significant differences were found in the number of other immune cells among groups.

C
. Martin Gil et al.

Fig. 6 .
Fig. 6.Summary diagram.During OA, DRG sensory neurons produce myostatin, and CXCL11 is sequentially produced possibly by satellite glial cells.These factors contribute to the attraction and programming of DRG macrophages into a pro-algesic phenotype that maintains pain in OA.

Table 1
Antibodies used for flow cytometry.

Table 2
Antibodies used for immunofluorescence.

Table 3
Primers used for Real-Time RT-PCR.
(caption on next page)C.Martin Gil et al.