Dorsal Root Ganglia CX3CR1 expressing monocytes/macrophages contribute to arthritis pain

Pain is a persistent symptom of Rheumatoid Arthritis, and the K/BxN serum transfer model recapitulates both association and dissociation between pain and joint inflammation in RA. Furthermore, this model features monocyte/macrophage infiltration in joints and lumbar dorsal root ganglia (DRG), where these immune cells are close to nociceptive neurons. We focussed on CX 3 CR 1 -monocyte/macrophage trafficking and show that at peak paw swelling associated with nociception, CX 3 CR 1 infiltration/phenotype in paws. However, acute nociception and DRG non-classical monocyte numbers were reduced in CX 3 CR 1GFP/GFP (KO) compared to CX 3 CR 1+/GFP (WT). Nociception that persisted despite swelling had resolved was attenuated in KO and correlated with DRG macrophages displaying M2-like phenotype. Still in the DRG, neurons up-regulated neuropeptide CGRP and olcegepant treatment reduced acute swelling, nociception, and leukocyte infiltration in paws and DRG. We delineate in-vitro a signalling pathway showing that CGRP liberates the CX 3 CR 1 ligand fractalkine (FKN) from endothelium, and in bone marrow-derived macrophages, FKN promotes activation of intracellular kinases, polarisation towards M1-like phenotype and release of pro-nociceptive IL-6. These data implicate nonclassical CX 3 CR 1 -expressing monocyte and macrophage recruitment into the DRG in initiation and maintenance of arthritis pain. These in-vitro data indicate that in macrophages FKN results in activation of intracellular kinases, polarisation towards an M1-like phenotype and release of the pro-nociceptive cytokine IL-6.


Introduction
Pain is a persistent feature of rheumatoid arthritis (RA). It is associated with the flares of RA and can persist even when joint inflammation is controlled by current anti-arthritic drugs. To advance our understanding of the mechanisms underlying RA pain and identify novel analgesic strategies, we employ models of RA that mimic both association and dissociation of pain with inflammation as they constitute an ideal platform to investigate mechanisms of acute and persistent nociception. Specifically, in the K/BxN model of inflammatory arthritis, a few days after passive immunization, mice exhibit hind paw nociceptive hypersensitivity (allodynia) in concomitance to clinical scores (paw swelling) (Christianson et al., 2010). However, allodynia persists after swelling has spontaneously resolved at around 4 weeks from immunization (Allen et al., 2020;Christianson et al., 2010;Woller et al., 2019). In this model, joint pathology is driven by neutrophils and macrophages and the infiltration of non-classical monocytes into the joints is crucial for the initiation, progression, and resolution of inflammation (Misharin et al., 2014). Monocytes are recruited to the synovium during the early phase of the model, where they display plasticity and differentiate into macrophages in situ, with changes from macrophage M1-like (pro-inflammatory) to M2-like (anti-inflammatory) phenotype over time. Notably, resident synovial macrophages possess a protective M2-like phenotype throughout the time course of the model (Culemann et al., 2019), highlighting the importance of recruited cells for initiation and progression of arthritis. We have recently observed that monocytes/macrophages also populate the dorsal root ganglia (DRG) and can be found in vicinity of the cell bodies of sensory neurons innervating the joint (Allen et al., 2020). Within DRG, macrophage interaction with sensory neurons contributes significantly to peripheral mechanisms of nociception via release of pro-inflammatory mediators such as cytokines and prostaglandins (Allen et al., 2020;Massier et al., 2015;Segond von Banchet et al., 2009).
Here we investigated the modality governing monocyte/macrophage invasion and modulation of nociceptive signalling in the DRG under inflammatory arthritis conditions.
Since monocytes/macrophages express CX 3 CR 1 receptors, which mediate monocyte adhesion and transmigration via binding to endothelial chemokine CX 3 CL 1 (Fractalkine, FKN) (Geissmann et al., 2003;Montague et al., 2018;Old et al., 2014), we hypothesised that the FKN/CX 3 CR 1 pair plays a functional role in mediating macrophage communication with nociceptive neurons in DRG under inflammatory arthritis conditions. Although CX 3 CR 1 is expressed by dendritic cells which play functional roles in trans-endothelial mechanisms (Vollmann et al., 2021), dendritic cells are not critical for K/BxN arthritis progression (Wu et al., 2007) and were not examined in this study. Instead we examined neutrophils which do not express CX 3 CR 1 (Jung et al., 2000), but play a critical role in K/BxN paw inflammation (Norling et al., 2016).

Animals
8-12-week-old male and female CX 3 CR 1 +/GFP (used as WT) and CX 3 CR 1 GFP/GFP (used as KO) littermate mice (B6.129P2(Cg)-Cx3cr1 tm1Litt /J) were used for experiments according to the United Kingdom Animals (Scientific Procedures) Act 1986 and following the guidelines of the Committee for Research and Ethical Issues of the International Association for the Study of Pain. All experimental study groups were randomized and blinded. Mice were housed in groups of up to 5 per standard cage. All animals were kept at room temperature with a 12-h light/dark cycle. Animals received food and water ad libitum.

Induction of K/BxN serum transfer inflammatory arthritis
Induction of K/BxN serum-transfer model of inflammatory arthritis was induced as previously described (Allen et al., 2020;Norling et al., 2016). Mice were injected with 50 μl of arthritogenic serum on days 0 and 2. Control mice received equivalent volume injections containing pooled sera from KRN/C57 mice (non-arthritic). Clinical signs of arthritis were evaluated using a 12point scoring system. Each limb was scored separately; 0 to 3 points per limb with the following criteria: 0, no sign of redness/swelling; 1, redness/swelling observed in either ankle/wrist, pad, or any of the digits; 2, redness/swelling in 2 regions; and 3, redness/swelling seen in all limb sections. Scores for all 4 limbs were combined to give a maximum score of 12 points per animal. Ankle thickness was measured using a digital caliper. Ankle joint measurements were acquired before the first injection of serum and repeated daily after injection.

Drug
Olcegepant (Tocris Biosciences, Abingdon, UK) was dissolved in 2.5% dimethyl sulfoxide in saline and injected at 1 mg/kg intraperitoneally, daily for 6 days, starting one day prior injection of K/BxN serum. Animals were assessed by von Frey 3 hours after each injection (n = 8 per group).

Behavioural tests
Hind paw mechanical withdrawal thresholds were assessed by application of calibrated von Frey monofilaments (0.02-1.0 g) to the hind paw plantar surface. Testing started with the application of a 0.07 g filament and each paw was assessed alternately between application of increasing stimulus intensity until a withdrawal response was achieved or application of 1.0 g filament failed to induce a response, to avoid tissue damage. The 50% paw withdrawal threshold (PWT) was determined by increasing or decreasing stimulus intensity and evaluated using Dixon's "up-down" method. Experiments were performed blind.

Immunohistochemistry of dorsal root ganglia
For immunohistochemistry of perfuse-fixed tissue, animals were trans-cardially perfused with saline solution followed by 4% paraformaldehyde (PFA, VWR chemicals) in PBS before L3, L4 and L5 DRGs were excised. Transverse DRG sections (10µm) were then cut with a cryostat (Bright instruments) and mounted onto Superfrost Plus microscope slides (ThermoFisher Scientific). Sections were first permeabilised with PBS containing 0.1% Triton-X-100 (PBS-T; Sigma-Aldrich) for 15 min, then blocked with 3% Bovine Serum Albumin (BSA) (Sigma-Aldrich) for 1 h and finally incubated overnight with goat anti-rabbit CGRP (2 μg/ml, Cell Signalling), followed by anti-rabbit Alexa Flour 594 secondary antibody (1 μg/ml, Invitrogen) for 1 h. Slides are then washed 3 times for 5 min in 0.1% PBS-T and mounted in DAPI containing mounting solution (Invitrogen). All antibodies were prepared in 0.1 M PBS with 0.1% BSA and 0.1% Triton X-100 (Sigma). For negative controls, the primary antibody was omitted; this resulted in the absence of staining. Images for immunofluorescence analysis were captured using a Zeiss LSM400 fluorescence microscope and analysed using ImageJ software (1.50i, Wayne Rasband, National Institutes of Health, USA). CGRP + profiles, indicative of number of CGRP positive neuron cells, were quantified within fixed areas (4 x 10 4 µm) per section. At least four sections from three mice per group were analysed.

Primary culture of BMDMs, treatments and Flow Cytometry.
Hematopoietic stem cells were harvested from the femur and tibia bone marrow of mice. Bone marrow cells were differentiated into macrophages by culturing for 7 days at 37 • C and 5% CO 2 in 8ml high glucose Dulbecco's modified eagle media (DMEM) supplemented with 10% heatinactivated fetal bovine serum (FBS, Gibco), 1% penicillin/streptomycin (P/S) and 10% supernatant derived from L929 fibroblasts (L929-condition media) as a source of macrophage colony-stimulating factor (Englen et al., 1995) in 100 mm non-tissue culture treated Petri dishes (ThermoFisher Scientific). On day 5, 3ml of medium was removed and an additional 5ml of medium was added. Gentle scrapping was used to lift cells off dish surface. Cells were then counted and resuspended in DMEM at the concentration of 1x10 6 per well of a 6 well plate.
Stained cells were acquired on an LSR Fortessa cytometer. All gating strategies were generated using Fluorescent Minus One (FMO) controls (Supplementary Figure 1).

Statistical analysis
All statistical calculations were performed using GraphPad Prism 7 (GraphPad Software, Inc.).
Data are shown as mean ± S.D. Differences between groups were identified by one-way ANOVA followed by Kruskal-Wallis, one-way ANOVA with Tukey or Bonferroni multiplecomparison tests, and two-way ANOVA with Tukey or Bonferroni multiple-comparison test.
Since the CX 3 CR 1 transgenic mouse is a global KO and CX 3 CR 1 receptor is expressed by monocytes, macrophages, but also microglia, we examined microglia activation in the dorsal horn of the spinal cord of K/BxN WT and KO. At day 5 and day 25 after K/BxN serum transfer, we observed significant microgliosis quantified as increased number of CD45 low CD11b + GFP + cells (flow cytometry) and Iba1 + cells ( Altogether, these behavioural data suggest that whilst CX 3 CR 1 expression in monocytes/macrophages is not critical for development of paw swelling, confirming previous work (Jacobs et al., 2010;Wang et al., 2012), it is likely to contribute to the development and maintenance of allodynia in inflammatory arthritis. Therefore, in WT and KO mice, we quantified and assessed the phenotype of monocytes/macrophages in the hind paw, which is the area where nociceptive fibres innervate the joint, and in lumbar DRG which is the place for cell bodies of nociceptors.

Monocytes and macrophages in hind paws and lumbar DRG of CX 3 CR 1 KO in K/BxN inflammatory arthritis
To study dynamics of monocyte influxes, we monitored GFP + cells, together with lineage markers Ly6C and CCR 2 which identify classical monocytes ( Figure 2A). In WT hind paw cell suspension, we measured a significant increase of Ly6C high cells at day 5, but not day 25 after K/BxN serum transfer in comparison to control serum ( Figure 2B). Infiltrating monocytes were Ly6C high CCR 2 + but also Ly6C low CCR 2 - (Figure 2 B,C). Similarly in KO paw cell suspension, Ly6C high cells were found at day 5 but not day 25 after K/BxN serum transfer ( Figure 2B) and both Ly6C high CCR 2 + and Ly6C low CCR 2cells were significantly higher than in control serum cell suspension (Figure 2 B,C). At day 5 after K/BxN serum transfer Ly6C high CCR 2 + cells were lower in KO than in WT ( Figure 2B). At day 5 and 25 after K/BxN serum transfer female and male hind paws showed similar infiltration of monocytes with no difference between the sexes. Representative scatter plots of gating strategy for monocytes. Cells were first gated based on FSC-A and SCC-A, doublets were discriminated using FSC-A and FSC-H and viable cells using L/D dye and SSC-A. Non-classical monocytes were gated as CD45 + F4/80 -Ly6C Low CCR 2 -, while classical monocytes were CD45 + F4/80 -Ly6C High CCR 2 + (B, C) Bar charts present numbers of classical and non-classical monocytes. Data are mean ± S.D.; n=8-13 animals per group. * p<0.05, **p < 0.01, ***p<0.05, same group, day comparisons, # p < 0.05, ## p<0.01, ### p<0.001 same day, group comparisons, two-way ANOVA with Tukey multiplecomparison test.
Since monocyte influx in the hind paw was noticeable in males and females at day 5 after K/BxN serum transfer and given that this model is associated with neutrophil infiltration (Norling et al., 2016), we investigated further the presence of both cell types in WT and KO in male mice. Monocytes were first identified as CD45 + CD11b + CD115 + and then further subdivided in classical (Ly6C high CD43 -CCR2 + ), non-classical (Ly6C low CD43 + CCR2 -), while neutrophils were identified as CD45 + CD11b + CD115 -Ly6G + events ( Figure 3A). This gating strategy confirmed presence of both classical and non-classical monocytes in WT and KO compared to control paws (Figure 3 B,C) with lower number of classical monocyte in KO compared to WT ( Figure 3B). Neutrophil numbers were higher in both WT and KO after K/BxN serum transfer compared to control serum ( Figure 3C), as expected since neutrophils do not express CX 3 CR 1 (Jung et al., 2000).
Altogether, these analyses in hind paws, support existing evidence (Brunet et al., 2016;Misharin et al., 2014) that at peak of paw swelling associated with allodynia, both classical and non-classical monocytes infiltrate the hind paw and macrophages display a proinflammatory phenotype (Supplementary Figure 3). Moreover, results in KO mice indicate that CX 3 CR 1 expression in monocytes is associated with reduced infiltration of classical monocyte which correlates with the less severe allodynia in KO at 5 K/BxN serum transfer and may contribute to sensitization of nociceptive fibre terminals in arthritis joints. Nevertheless, CX 3 CR 1 expression in macrophages is unlikely to impact on number and phenotype of these cells in the hind paw. Overall, our data point to a pro-nociceptive role for CX 3 CR 1 -expressing monocytes but not macrophages in arthritic joints. However, to step away from the site of primary inflammation, we moved our attention to the lumbar DRG which contain the cell bodies of nociceptive neurons that innervate the joint and performed flow cytometry analysis of monocytes and macrophages at day 5 and 25 after K/BxN serum transfer. In WT DRG, we observed that Ly6C cell (monocyte) numbers were higher at day 5 but not day 25 after K/BxN serum compared to control serum ( Figure 5 A-C). These cells were predominantly Ly6C low CCR 2with small percentage (≤10%) Ly6C high CCR 2 + (Figure 5 B,C). In KO DRG, Ly6C high CCR 2 + cell numbers were higher in K/BxN serum compared to controls at day 5 ( Figure 5B). However, Ly6C low CCR 2cell numbers in K/BxN serum transfer DRG were lower than in WT ( Figure 5C). Overall, these analyses of DRG data suggest that CX 3 CR 1 expressing monocytes (Ly6C low CCR 2 -) infiltrate the DRG at day 5 after K/BxN serum transfer, and absence of this chemokine receptor impairs such an infiltration (Supplementary Figure 4A). Classical monocytes were CD45 + CD11b + F4/80 -Ly6C High CCR 2 + cells, while non-classical were identified as CD45 + CD11b + F4/80 -Ly6C Low CCR 2events. (B-C) Bar charts for classical and nonclassical monocytes in WT and KO DRG. Data are mean ± S.D.; n=7-13 animals per group. Data are mean ± S.D.; * p<0.05, **p < 0.01, ***p<0.001, same group, day comparisons, # p < 0.05, ## p<0.01, ### p<0.001 same day, group comparisons, two-way ANOVA with Tukey multiple-comparison test.
Next, we assessed numbers and phenotypes of macrophages in DRG. In WT DRG, we observed an increase of F4/80 + cells at day 5 which reached statistical significance at day 25 after K/BxN serum transfer compared to control serum (Figure 6 A,B). Instead in KO DRG, at day 5 after K/BxN serum transfer the numbers of F4/80 + cells were comparable to both control serum and WT K/BxN, while at day 25 after K/BxN serum transfer, macrophage numbers were lower than in WT ( Figure 6B), indicating that in KO DRG lower infiltration of monocytes resulted in no increase of macrophages.
As observed in hind paws, monocyte and macrophage numbers were similar in male and female DRG not only under control condition, but also at day 5 and day 25 after K/BxN serum transfer (Figures 5,6).
Considering that M2-like macrophages are antinociceptive when injected intrathecally (Raoof et al., 2021;van der Vlist et al., 2022), their presence in the DRG correlates with the development of less severe allodynia after K/BxN serum transfer in CX 3 CR 1 KO. Moreover, a further consideration based on WT data is that CX 3 CR 1 -expressing monocytes/macrophages in DRG contribute to nociceptive mechanisms in K/BxN inflammatory arthritis. Existing evidence, including ours, indicates that CX 3 CR 1 -expressing monocytes can bind to FKN, a transmembrane chemokine expressed by endothelial cells, to facilitate monocyte adhesion.
Following cleavage of FKN chemokine domain by ADAM17, soluble FKN binding to CX 3 CR 1 promotes monocytes transmigration through the endothelium (Landsman et al., 2009;Old et al., 2014;White and Greaves, 2012). Therefore, our data suggest the possibility that in response to DRG sensory neuron activity triggered by peripheral inflammation, sFKN is liberated from the endothelium and promotes CX 3 CR 1 -monocyte adhesion and infiltration in the DRG parenchyma. Thus, we next investigated the mechanism which underlies sensory neuron-endothelium communication, and focussed on the neuropeptide CGRP, which is known to promote plasma extravasation in neurogenic inflammation (Chiu et al., 2012).

Sensory neuron CGRP in K/BxN inflammatory arthritis
Sensory neuron peptide content undergoes significant changes under peripheral inflammation conditions (Nieto et al., 2015;Staton et al., 2007). We found that CGRP expression was upregulated in DRG small and medium size neurons at day 5 but not 25 after K/BxN serum transfer compared to controls (Figure 7 A,B). In both WT and KO DRG at day 5 after K/BxN transfer, CGRP expression was increased by ~40%, suggesting that CX 3 CR 1 expression in monocytes/macrophages is unlikely to influence neuronal CGRP expression which is increased in sensory neurons under inflammatory arthritis. Since CGRP is released with activity by nociceptive neurons (Ebbinghaus et al., 2015) and promotes hind paw nociception and joint swelling, we tested the effect of systemic administration of the CGRP receptor antagonist, olcegepant, on swelling, allodynia and leukocyte accumulation in paws and DRG. We selected olcegepant, a small molecule antagonist, as this was recently shown to attenuate mechanical hypersensitivity in female mice in a model of neuropathic pain (Paige et al., 2022) and not to exert any effect on baseline pain responses in both male and female rodents (Christensen et al., 2019;Kopruszinski et al., 2021). Olcegepant administration (1 mg/kg daily for 5 days) to female mice ( Figure 8A), reduced clinical scores and ankle swelling at 4 th injection on day 3 after K/BxN serum transfer, which is when swelling started to show, and this effect persisted for up to day 5 (Figure 8 B,C).

Olcegepant attenuated mechanical hypersensitivity from 3 hours after the first injection of
KBxN serum at day 0 for up to day 4 ( Figure 8D) and this anti-nociceptive effect was still detected at 24 h after the last administration at day 5 ( Figure 8E).
Altogether these behavioural data suggest that CGRP contributes to swelling and allodynia in inflammatory arthritis. Next, we quantified the phenotype of monocyte/macrophages and neutrophils in hind paw cell suspensions and in lumbar DRG using the same markers as above (Figures 3-6) In the paws, at day 5 after K/BxN serum transfer, olcegepant treatment resulted in fewer classical and non-classical monocyte, neutrophil and M-1 like macrophage numbers compared to vehicle group ( Figure 9A-F), indicating that CGRP drives generation of the inflammatory infiltrate within the arthritic paw.
When DRG were analysed for immune cells numbers and phenotype, we then observed that olcegepant treatment reduced Ly6C cells (both classical and non-classical monocytes) at day 5 after K/BxN serum transfer ( Figure 10A,B). In the macrophage analyses, olcegepant reduced MHCII + (M1-like) macrophage recruitment, ( Figure 10D) without affecting either the M1/M2 or the M2 like macrophages (Figure 10 E,F).
Overall, these data indicate that sensory neuron derived CGRP promotes infiltration of classical monocytes, non-classical monocytes, neutrophils, and macrophages in the hind paws as well as of classical monocytes, non-classical monocytes and M1 like macrophages in DRG.
Since these effects of CGRP would be concomitant to endothelial activation (Crossman et al., 1990), and CX 3 CR 1 expressing monocytes infiltrate the DRG at day 5 after K/BxN serum transfer, we postulated that CGRP may be critical for the liberation of sFKN from the endothelium which would then attract monocytes.

Liberation of endothelial fractalkine by sensory neuron CGRP
To test this hypothesis, we evaluated the effect of CGRP incubation with HUVEC cells on the expression of FKN and selected adhesion molecules using flow cytometry. Incubation of HUVECs with CGRP (1 µM overnight) resulted in a significant decrease of membrane-bound FKN expression both when expressed as % of single cells and MFI units (Figure 11 A,B and Supplementary Figure 5) and this effect was reversed by the CGRP receptor antagonist CGRP 8-37 (Figure 11 A,B, Supplementary Figure 5). The effect of CGRP was specific as it did not alter expressions of ICAM-1, VCAM-1 and CD62E (Figure 11 C-E, Supplementary Figure 5).
Since membrane-bound FKN is cleaved into sFKN by the protease ADAM-17 which is also expressed by endothelial cells (Gooz et al., 2009), we quantified ADAM-17 expression and release of sFKN after CGRP application to HUVECs. We observed that CGRP application resulted in up-regulation of ADAM-17, and this effect was blocked by CGRP 8-37 ( Figure 11F).
Furthermore, under the same conditions CGRP increased sFKN content in the incubation media and CGRP 8-37 blocked this effect ( Figure 11G). Finally, using immunohistochemistry we could visualise that CGRP incubation reduced expression of membrane-bound FKN expression in HUVEC (CD31 + cells) and this effect was inhibited by CGRP 8-37 (Supplementary Figure 5). These in-vitro data indicate that in endothelial cells CGRP upregulates ADAM-17 which liberates sFKN and suggest that in K/BxN DRG, CGRP released with activity by nociceptive neurons would result in liberation of endothelial sFKN that can activate CX 3 CR 1 receptor in monocytes and promotes monocyte infiltration in the DRG parenchyma. In our final set of experiments with bone marrow-derived macrophages, we identified the mechanism by which sFKN activation of CX 3 CR 1 promotes nociceptive signalling.

IL-6
Since monocyte/macrophage invasion in K/BxN DRG is associated with behavioural allodynia in WT and attenuation of allodynia in CX 3 CR 1 KO mice, we evaluated the effect of FKN on pro-nociceptive cytokine release in bone marrow derived macrophages (BMDM). Our preliminary data in differentiated THP-1 cells indicated that IL-6 and IL-8 were up-regulated by FKN in macrophages more than IL-1 and TNF. Thus, we have only considered IL-6 in this study. We observed that FKN (200 ng/ml) induced significant release of IL-6 in WT, but not in CX 3 CR 1 KO BMDM (Figure 12 A,

Discussion
In this study we show that in the K/BxN serum transfer model of inflammatory arthritis, acute arthritic scores (swelling) and allodynia (mechanical hypersensitivity) are associated with invasion of monocytes and presence of M1-like macrophages in the joints and DRG. When K/BxN inflammatory arthritis resolve, macrophages return to basal levels and display M2-like phenotype in joints. However, allodynia persists and correlates with the presence of numerous M1-like macrophages in DRG. Furthermore, absence of CX 3 CR 1 receptor, expressed by monocytes/macrophages, results in significant attenuation of allodynia, monocyte recruitment in joints and DRG and M1-like macrophage infiltrate in DRG. These findings expand on the assumption that monocytes/macrophages in DRG are relevant to nociceptive mechanisms and suggest that CX 3 CR 1 receptor plays a significant role in the initiation and persistence of allodynia in inflammatory arthritis. Microglial CX3CR1 receptors appear not to contribute to nociception in the KBxN model of RA. This is inconsistent with our data showing FKNneutralising antibody reversal of nociception in the rat collagen-induced arthritis model of RA, which was associated with attenuation of microglial activation in the dorsal horn (Nieto et al., 2016). We speculate that active immunization in CIA versus passive immunization in K/BxN ST may be associated with distinct microglial pathways as for instance, microglial TLR-4 receptor plays a critical role in KBxN persistent allodynia (Christianson et al., 2011).
Under joint arthritis conditions, large numbers of monocytes infiltrate the joint and play critical roles in initiation, progression, and resolution of inflammation (Brunet et al., 2016;Misharin et al., 2014). Specifically, in K/BxN inflammatory arthritis, there is evidence that classical monocytes are critical for the progression of joint inflammation (Brunet et al., 2016). However, there is also proof that non-classical monocytes are crucial for the initiation of arthritis (Misharin et al., 2014). Nevertheless, severity of clinical scores is altered in neither CCR 2 deficient nor CX 3 CR 1 deficient mice (Jacobs et al., 2010;Wang et al., 2012) suggesting that neither chemokine receptors are critical for the onset and progression of joint swelling. We obtained similar findings and observed that CX 3 CR 1 receptor deficiency affects neither severity of clinical scores, nor macrophage phenotype in arthritic paws though classical monocyte infiltration was reduced. Classical monocytes express CX 3 CR 1 and CCR 2 and we observed that deficiency of CX 3 CR 1 results in infiltration of more CCR 2 + monocytes in sciatic nerve in vincristine model of neuropathic pain (Montague et al., 2018). Nevertheless, a possible interaction in the K/BxN inflamed paw would have resulted in opposite results indicating that interactions of these chemokine receptors are tissue and model dependent.
Away from the joint in the DRG, we observed recruitment of non-classical and classical monocytes as well as presence of M1-like macrophages at peak inflammation at day 5 K/BxN arthritis. With the progression of arthritis, when joint inflammation resolves by day 25, DRG still present high numbers of M1-like macrophages. However, at this time point a good proportion (~60%) also expresses CD206 and as such are identified as dual M1/M2-like macrophages. Relevantly, deficiency of CX 3 CR 1 receptor, impairs non-classical monocyte infiltration in DRG at day 5 K/BxN, and alters the macrophage dynamics; using our markers, in CX 3 CR 1 KO DRG the phenotype of macrophages is mostly M1/M2-like at day 5 and M2 at day 25 K/BxN. Taken together, these observations lead to two assumptions. First assumption is that in inflammatory pain conditions, CX 3 CR 1 receptors mediate monocyte/macrophage communication with sensory neurons responsible for nociceptive signalling. Second assumption is that the DRG are a preferable location over the joint for further delineation of neuro-immune mechanisms in nociceptive pain. Undeniably, inflamed joints would be a plausible site for such studies as they are innervated by primary afferents that i) transmit noxious signalling to the CNS, ii) can be sensitised by inflammatory mediators and iii) contribute to inflammation by releasing neuropeptides (neurogenic inflammation). However, our focus to the DRG is motivated by their location away from the joint since nociception is concomitant to paw inflammation but persists even when overt paw inflammation subsides.
Thus, we gathered evidence that the neuropeptide CGRP expression increases in sensory neurons in concomitance to joint inflammation, an observation which is in line with CGRP upregulation in DRG in other models of arthritis and peripheral inflammation (Nieto et al., 2015;Seybold et al., 1995).
CGRP is a pro-nociceptive peptide that is released by the central terminals of primary afferent fibres in the dorsal horn of the spinal cord where it mediates mechanical hypersensitivity (Nieto et al., 2015). CGRP is also released at peripheral fibre terminals where it mediates neurogenic inflammation and regulates immune response to bacterial infection (Chiu et al., 2013;Chiu et al., 2012). We show that the small molecule CGRP antagonist, olcegepant prevents development of swelling and mechanical hypersensitivity in inflammatory arthritis and precludes monocyte and neutrophil infiltration in paws while reducing classical monocytes, non-classical monocytes and M1 like macrophage numbers in DRG. We suggest that sensory neuron-derived CGRP orchestrates inflammatory cell trafficking in inflammatory arthritis and corroborate this hypothesis with an in-vitro set of experiments. Herein, we show that CGRP induces the expression of endothelial ADAM-17 which liberates soluble FKN chemokine domain in HUVEC cells. Since FKN activation of CX 3 CR 1 receptor induces M1-like phenotype in bone marrow-derived macrophages and release of pro-nociceptive cytokine IL-6, we suggest that in K/BxN DRG, neuronal CGRP activation of endothelial cells promotes monocyte infiltration via activation of endothelial FKN/monocyte CX 3 CR 1 signalling which in turn sensitise neurons via downstream cytokine release (Supplementary Figure 6).
Considering that the cell bodies of sensory neurons in DRG contribute to nociceptive mechanisms, and CX 3 CR 1 KO develop less sever allodynia than WT, our data suggest that CX 3 CR 1 receptors expressed in monocytes/macrophages play a critical role in nociceptive mechanisms in inflammatory arthritis A peculiarity and translational feature of the K/BxN serum transfer model of inflammatory arthritis is that whilst joint swelling resolves at 3-4 weeks after passive immunization, allodynia persists. We have recently proposed that DRG macrophages play a significant role in the maintenance of mechanical hypersensitivity in inflammatory arthritis (Allen et al., 2020). Here we add more prominence to these immune cells as in CX 3 CR 1 KO DRG, macrophages display a predominant M2-like phenotype and allodynia is significantly attenuated in CX 3 CR 1 KO mice.
We suggest that the absence of CX 3 CR 1 in monocytes/macrophages prevents macrophages from acquiring M1-like phenotype during progression of inflammatory arthritis by reducing the impact of endothelial FKN activation of CX 3 CR 1 receptors in these cells. Since macrophage polarization between M1 and M2 states is a regulated phenomenon (Mounier et al., 2013;Sica and Mantovani, 2012), our data indicate that DRG macrophages are highly dynamic in-vivo.
Under normal conditions, DRG macrophages distribute among 3 equally represented populations, namely M1-, M2-and M1/M2-like phenotypes, which suggests that M1 marker expression doesn't rule out expression of M2 markers. At both day 5 and day 25 K/BxN, WT macrophages go through phenotypic transition to M1/M2-like macrophages, which provides evidence of polarization skewing. Importantly, polarization also occurs in CX 3 CR 1 KO DRG, however the direction of skewing indicates preferred transition towards M2-like phenotype, which suggests a mechanistic role of CX 3 CR 1 receptors in conferring M1-like phenotype in WT. This was partially recapitulated in our in-vitro studies where BMDMs treated with FKN acquired M1-like phenotype and at the same time supported by evidence that intraganglionic injection of FKN induces proinflammatory phenotype in macrophages (Kwon et al., 2015).
In conclusion, in inflammatory arthritis, sensory neuron activity in DRG promotes recruitment of monocytes via a CGRP-led mechanism that includes endothelial liberation of FKN and transmigration of CX 3 CR 1 -expressing monocytes. In DRG monocyte/macrophage CX 3 CR 1 expression is likely to confer M1-like phenotype to macrophages and promote nociceptive neuron activation via release of factors, including IL-6 (see scheme in Supplementary Figure   6). Thus, we suggest that peripheral CX 3 CR 1 receptor offers a potential target to relieve arthritis pain.

Author contributions
Silvia Oggero planned, designed and performed experiments, and analysed the data. Chiara Cecconello, Rita Silva, George Sideris-Lampretsas conducted experiments. Mauro Perretti supported data analyses and provided comments and suggestions to the manuscript. Lynda Zeboudj helped with data interpretation. Marzia Malcangio devised the project, the main conceptual idea and proof outline. Silvia Oggero and Marzia Malcangio wrote the manuscript.