Abstract
The current paradigm states that inflammatory pain passively resolves following the cessation of the inflammatory insult. Yet, in a substantial proportion of patients with inflammatory diseases, such as rheumatoid arthritis and inflammatory bowel disease, spontaneous or treatment-induced resolution of inflammation is not sufficient to resolve pain, resulting in chronic pain1–5. Mechanistic insight as how inflammatory pain is resolved is lacking. Here we show that macrophages actively control resolution of inflammatory pain remotely from the site of inflammation by transferring mitochondria to sensory neurons. During resolution of inflammatory pain in mice, M2-like macrophages infiltrate the dorsal root ganglia that contain the somata of sensory neurons, concurrent with the recovery of oxidative phosphorylation in sensory neurons. To resolve pain, macrophages transfer mitochondria to sensory neurons. This transfer requires expression of CD200 Receptor (CD200R) on macrophages and the non-canonical CD200R-ligand iSec1 on sensory neurons. Our data reveal a novel mechanism for active resolution of inflammatory pain and suggests a new direction for treatment of chronic pain.
Main text
Pain and pain hypersensitivity (hyperalgesia) are functional features of inflammation that serve to protect the tissue from further damage. At the site of inflammation, immune cells and inflammatory mediators, such as IL-1β, TNF, and bradykinin, sensitize and activate sensory neurons, which cause pain and hyperalgesia6,7. While the initiation of inflammatory pain is relatively well understood8,9, the mechanisms of inflammatory pain resolution are less well characterized. We here set out to better understand the resolution of inflammatory pain and the role of immune cells in this process.
We injected carrageenan into the hind paw of mice (intraplantar; i.pl.) as a model for transient inflammatory pain (Supplemental data fig. 1A)10. Treated mice displayed signs of pain, such as allodynia/hyperalgesia as assessed by the von Frey and Hargreaves tests, and postural changes measured with dynamic weight bearing. Carrageenan-induced hyperalgesia resolved within ~3-4 days (Figs. 1A and supplemental data fig. 1B). We analysed the cellular composition of lumbar (L3-L5) dorsal root ganglia (DRG) which contain the somata of sensory neurons innervating the hind paw and observed an influx of macrophages. Macrophage numbers peaked at day 3 and returned to baseline levels after resolution of inflammatory hyperalgesia (Fig. 1B, supplemental data figs. 1C-1D, and supplemental data movies 1 and 2). Infiltration of macrophages was specific to the DRG that innervate the inflamed paw, and was not observed at the contralateral side (supplemental data fig. 1E). In contrast, during the entire course of inflammatory hyperalgesia, T cells, B cells or other CD45+ immune cell numbers in the DRG did not change significantly (Figs. 1B and supplemental data fig. 1C). To address the function of these macrophages in pain resolution, we selectively depleted monocytes and macrophages by intraperitoneal (i.p.) injection of diphtheria toxin in Lysmcre x Csf1r LsL-DTR mice11 (from here on referred to as ‘MMdtr’, supplemental data fig. 2). The induction and magnitude of carrageenan-induced hyperalgesia in these mice was normal. However, MMdtr mice failed to resolve inflammatory mechanical hyperalgesia (Fig. 1C), thermal hyperalgesia (Supplemental data fig. 3A) and postural changes related to inflammatory pain (Supplemental data fig. 3B) in both male and female mice (Supplemental data fig. 3C). Similarly, MMdtr mice failed to resolve Complete Freund’s Adjuvant (CFA)-induced transient inflammatory hyperalgesia (Supplemental data figs. 3D and 3E). To directly target monocytes to the DRG12, we intrathecally (i.t.) injected wildtype (WT) bone-marrow-derived CD115+ monocytes into MMdtr mice (Supplemental data fig. 4A), which reconstituted macrophages in the ipsilateral DRG (Supplemental data fig. 4B). Within hours, i.t. injection of WT monocytes sustainably rescued the defective resolution of hyperalgesia in MMdtr mice (Fig. 1D and supplemental data fig. 4C). The pain-resolving capacity of monocytes was independent of their origin (bone marrow or spleen; supplemental data fig. 4C-E) or Ly6C expression (‘classical’ Ly6C+ or ‘non-classical’ Ly6C−; supplemental data figs. 4F-H). These data show that monocytes are essential to resolve inflammatory pain and sufficient when specifically grafted into the DRG.
Macrophages that reside in peripheral nerve tissue are different from microglia and or non-nervous residing macrophages13,14 and can be programmed by sensory neurons 15. We determined whether monocytes/macrophages that infiltrate the DRG had an inflammatory (‘M1’) - or resolution (‘M2’)-like phenotype. At day 3, the number of CD206+ M2-like or tissue-repair macrophages16,17 was increased in the DRG, whereas the number of iNOS+ M1-like or inflammatory macrophages did not significantly change (Supplemental data fig. 4I). Consistent with the dominant presence of CD206+ macrophages, i.t. injection of bone-marrow derived macrophages (from here on referred to as ‘macrophages’) differentiated in vitro with IL-4 (‘M2’) rescued resolution of hyperalgesia in MMdtr mice. In contrast, inflammatory macrophages differentiated with LPS and IFNγ (‘M1’) induced a transient hyperalgesia in the saline treated paws and were incapable of resolving inflammatory hyperalgesia in the carrageenan treated paws (Figs. 1E and supplemental data fig. 4J). Macrophages resolved pain through a pathway independent of neuronal IL10 receptor (IL10R) signalling because Nav1.8CreIl10rflox mice, which are deficient for the IL10 receptor in pain-sensing sensory neurons that mediate inflammatory hyperalgesia18, recovered normally (Supplemental data fig. 4K).
Metabolically, M2 macrophages depend on oxidative phosphorylation (OxPhos), while M1 macrophages are glycolytic (Supplemental data fig. 5A)19. Neurons have a very high metabolic demand20. We previously demonstrated that a deficiency in mitochondrial function in sensory neurons prevents the resolution of inflammatory pain12. Moreover, in chronic pain neuronal mitochondrial functions, such as OxPhos and Ca2+ buffering, are impaired21,22. Indeed, oxygen consumption in DRG neurons was reduced during the peak of inflammatory pain and resolved at day 3 (Supplemental data fig. 5B). Therefore, we posited that to resolve inflammatory pain, sensory neurons have to re-establish OxPhos by restoring a functional mitochondrial pool. Given that after ischemic stroke neurons can take up mitochondria released by adjacent astrocytes23, we hypothesized that during pain resolution monocytes/macrophages aid neurons by supplying new mitochondria.
We stained mitochondria from macrophages with MitoTracker Deep Red (MTDR) that binds covalently to mitochondrial proteins24 and co-cultured live macrophages, or an equivalent volume of sonicated macrophages, with the neuronal cell line Neuro2a (N2A). After 2 hours, macrophage-derived MTDR+-mitochondria were detectable in N2A cells by flow cytometry and image stream (Supplemental data figs. 6A/B). Macrophages transduced with mitochondria-targeted DsRed (mitoDsRed) also transferred mitochondria to primary sensory neurons upon co-culture in vitro, excluding that the signal was due to MTDR leak from macrophages to neurons (Supplemental data fig. 6C). Transfer of mitochondria from macrophages to neurons also occurred in vivo. During the resolution of pain, we detected a significantly higher percentage of MitoDentra2 positive sensory neurons in the lumbar ipsilateral DRG of LysMCre-MitoDendra2flox mice compared to the contralateral DRG (Figs. 2A/B). In contrast, MitoDendra2flox mice or LysMCre-GFPflox mice did not have any MitoDendra2 or GFP positive neurons (Supplemental data fig. 6D), suggesting that monocytes/macrophages transfer mitochondrial content to neurons during resolution of inflammatory pain. In addition, i.t. injection of MTDR-labelled macrophages in MMdtr mice at day 6 after carrageenan injection increased the MTDR intensity in sensory neurons of mice with persisting inflammatory hyperalgesia (Fig. 2C and Supplemental data fig. 6E), but not in control treated mice or in WT mice that had resolved inflammatory hyperalgesia. Injection of sonicated MTDR-labelled macrophages did not result in accumulation of MTDR in sensory neurons (Fig. 2C), confirming mitochondrial transfer to sensory neurons. Using flow cytometry, we found that macrophages released CD45+ extracellular vesicles that stained positive for macrophage plasma membrane proteins, such as MHC class II, CD11b and CD200 Receptor 1 (CD200R), and the mitochondrial dye MTDR (Supplemental data fig. 7A/B). In line with the MTDR staining in vesicles, in the supernatant of MitoDendra2+ macrophages we detected CD45+MitoDentra2+ vesicles (Supplemental data fig. 7C). The vesicles had a broad range in size (Supplemental data fig. 7D).
We hypothesized that the mitochondria-containing vesicles released by macrophages were sufficient to resolve pain. Indeed, i.t. administration of mitochondria-containing extracellular vesicles isolated from macrophage supernatant rapidly but transiently resolved inflammatory hyperalgesia in MMdtr mice (Fig. 2D). However, injection of sonicated extracellular vesicles did not affect hyperalgesia. Taken together, this suggests that functional mitochondria, but not their individual components, are sufficient to resolve pain. Furthermore, monocytes that have distressed mitochondria and significantly reduced mitochondrial DNA (mtDNA) content due to a heterozygous deletion of the Transcription Factor A/Tfam25 failed to resolve inflammatory hyperalgesia in MMdtr mice (Supplemental data figs. 8A/B). Finally, we obtained artificial vesicles containing mitochondria from macrophage cell bodies (MitoAV). MitoAV stained positive for macrophage plasma membrane markers and MTDR and had active OxPhos (Supplemental data figs. 8C and 10B). I.t. injection of MitoAV rapidly but transiently resolved inflammatory hyperalgesia in MMdtr mice (Fig. 2E and supplemental data fig. 8D). In contrast, MitoAV in which oxidative phosphorylation was blocked by complex III inhibitor myxothiazol26 failed to resolve hyperalgesia (Fig. 2E and supplemental data fig. 8D). Thus, to resolve inflammatory pain, macrophages transfer vesicles containing mitochondria that are functional in their oxidative phosphorylation capacity.
For efficient transfer of mitochondria, we hypothesized that docking of extracellular vesicles to sensory neurons requires receptor-ligand interactions. Macrophages, predominantly those with an M2 phenotype27, and macrophages-derived extracellular vesicles expressed CD200R (Fig Supplemental data fig. 7A), while neurons are known to express its ligand CD200 (ref. 28). In line with this reasoning, Cd200r−/− mice completely failed to resolve inflammatory hyperalgesia, which persisted for at least 2 weeks (Fig. 3A and Supplemental data fig. 9A). Place-preference conditioning with the fast-working analgesic gabapentin29 confirmed ongoing spontaneous pain in Cd200r−/− mice up to 2 weeks after carrageenan injection (Supplemental data figs. 9B/C), further supporting the role of CD200R in the resolution of acute inflammatory pain and to prevent the development of chronic pain. In contrast, acute inflammation and the resolution of inflammation at the site of carrageenan injection in Cd200r−/− mice did not differ from that of WT mice (Supplemental data fig. 9D/E). I.t. injection of WT monocytes or macrophages resolved persisting inflammatory hyperalgesia in Cd200r−/− mice, whereas injection of additional Cd200r−/− monocytes or macrophages did not (Fig. 3B; supplemental data fig. 9F). Consistent with these data, Cd200r−/− monocytes did not resolve inflammatory hyperalgesia in MMdtr mice (Supplemental data figs. 9G/H). These data indicate an intrinsic defect in the pain-resolution capacity of Cd200r−/− monocytes and macrophages independent from effects of macrophages at the site of local inflammation.
We found no evidence for a defect in mitochondrial respiration or vesicle release in Cd200r−/− macrophages (Supplemental data figs. 10A-C) and Cd200r−/− macrophages were normal in their capacity to migrate into the DRG and had a similar phenotype to WT mice (Supplemental data figs. 10D-H). This suggested instead that there was a defect in mitochondrial transfer between Cd200r−/− macrophages and neurons. MTDR transferred normally from intrathecally injected MTDR-labelled macrophages to neurons from Cd200r−/− mice (Fig. 3C). In contrast, Cd200r−/− macrophages failed to transfer MTDR-labelled mitochondria to sensory neurons of Cd200r−/− mice (Fig. 3C) and extracellular vesicles isolated from Cd200r−/− macrophage culture supernatant did not resolve inflammatory hyperalgesia in MMdtr mice (Fig. 3D; supplemental data fig. 11). Thus, CD200R expression on monocytes/macrophages and their mitochondria containing extracellular vesicles is required for transfer of mitochondria to sensory neurons and for the resolution of inflammatory pain.
CD200 is the best-known ligand for CD200R and in inflammatory models, such as arthritis, Cd200−/− and Cd200r−/− mice have a similar phenotype30,31. However, in sharp contrast to Cd200r−/− mice, Cd200−/− mice completely resolved inflammatory pain with similar kinetics to WT mice (Fig. 4A; supplemental data fig. 12A). This suggests the involvement of an alternative CD200R ligand. In 2016, iSec1/Gm609 was described as a CD200R ligand expressed specifically in the gut32. We found that iSec1/Gm609 mRNA is expressed in DRG along with CD200 (Supplemental data figs. 12B/C). Repetitive i.t. injections of iSec1/Gm609 targeting antisense oligodeoxynucleotides (ASO)33 silenced mRNA expression in WT mice (Supplemental data figs. 12D/12E) and partially prevented resolution of inflammatory hyperalgesia (Fig. 4B; Supplemental data fig. 12F). In Cd200−/− mice, i.t. injections of iSec1/Gm609-ASO completely prevented the resolution of hyperalgesia (Fig. 4C; supplemental data fig. 12G). Next, we injected Herpes Simplex Virus (HSV) i.pl. to specifically target sensory neurons innervating that area12. Expression of iSec1/gm609 (HSV-iSec1) that was mutated to resist ASO treatment (Supplemental data fig. 12H) in sensory neurons completely rescued the ability of iSec1/Gm609-ASO treated CD200−/− mice to resolve pain, while an empty vector (HSV-e) did not (Fig. 4D; Supplemental data fig. 12I). We conclude that monocyte/macrophage expression of CD200R and sensory neuron expression of its ligand iSec1 are required for the transfer of macrophage-derived mitochondria to sensory neurons in vivo to resolve inflammatory pain.
Discussion
We identified a previously unappreciated role for macrophages which transfer mitochondria to somata of sensory neurons to resolve inflammatory pain. Previous studies showed that tissue-resident cells can transfer mitochondria23,34. We now show that non-tissue resident monocytes are recruited to the DRG, acquire a M2/tissue-repair like phenotype and transfer mitochondria to sensory neurons via a CD200R:iSec1 interaction in order to resolve pain. In contrast to M2 macrophages, inflammatory M1 macrophages induced pain. Thus, a DRG-milieu that skews local macrophages towards a M1 phenotype could contribute to the development of chronic pain35–38.
CD200 has long been thought of as the only ligand for CD200R. Although previous studies implicated CD200 as a checkpoint for microglia cell activation in neuropathic pain39,40, we show that Cd200−/− mice fully resolve inflammatory pain. Furthermore, we found that iSec1/gm609 is expressed in DRG neurons and we demonstrated that sensory neuron-iSec1 is required to resolve inflammatory pain. Of note, iSec1/gm609 knockdown did have a greater effect on pain resolution in Cd200−/− mice than in WT mice, suggesting that the function of these ligands is partially redundant.
Various chronic pain states, such as chemotherapy-induced pain and neuropathic pain caused by trauma or diabetes, are associated with mitochondrial defects41–44. We show here that oxidative phosphorylation is reduced during the peak of transient inflammatory pain but is restored when inflammatory hyperalgesia resolves. Hence, we postulate that resolution of inflammatory pain requires the restoration of mitochondrial homeostasis in sensory neurons and that DRG macrophages facilitate this process. Given that the injection of isolated extracellular vesicles only transiently resolves pain, a more durable resolution of pain requires a prolonged flux of mitochondria and/or additional signals from intact macrophages. These mitochondria could replace mitochondria in neurons that have incurred mitochondrial damage. Future work should assess how exactly neuronal mitochondrial homeostasis is restored by macrophage-derived mitochondria.
Why would sensory neurons require external help to restore the integrity of their mitochondrial network? Sensory neurons face unique challenges in maintaining a functional mitochondrial network because of their exceptional architecture and their intense demand for energy to support energetically expensive processes such as resting potentials, firing action potentials and calcium signalling20. Stressed neurons, e.g. during inflammatory pain, turn to anabolic metabolism45. In the face of this high energy demand during stress, an energy consuming process such as rebuilding the mitochondrial network is difficult to support. Moreover, maintaining an excess mitochondrial pool that is capable of handing the stress of inflammatory pain, would come at a high fitness cost because it would require more energy intake for the organism. Thus, we propose that dispensable monocytes/macrophages supply mitochondria to stressed indispensable neurons.
Together, our data show that pain is actively resolved by an interaction between the immune and neuronal systems that is separate from the cessation of inflammation within the peripheral tissue. Novel therapeutic strategies to resolve chronic pain may focus on the restoration of mitochondrial homeostasis in neurons or on enhancing the transfer of mitochondria from macrophages46.
Funding
This work has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 642720. MV, HW and LM are supported by the Netherlands Organization for Scientific Research (NWO), (ALW Grants 863.14.016, 821.02.025, 016.VENI.192.053 and NWO Vici 918.15.608). NE/JP were partly funded by the Life Sciences Seed grant of the University Utrecht.
Author contributions
Conceptualization (LM, MV, NE), Methodology (NE, LM, MV, HW, RR, JP, RL), Formal Analysis (MV, NE, HW, RR, JP), Investigation (LM, NE, MN, RR, HW, MVos, WKH, MV, SV, RL), Resources (RJP), Writing original draft (MV, NE, RR), Writing reviewing and editing (MV, NE, LM, RR, JP, HW, WKH), Visualization (MV, NE, RR), Supervision (MV, HW, NE, LM, RR), Project Administration (NE, LM), Funding Acquisition (HW, MV, NE, LM).
Competing interests
Authors declare no competing interests.
Data and materials availability
All data is available in the manuscript or supplementary materials. Raw data and materials will be made available upon reasonable request. Some materials used in this manuscript are subject to Material Transfer Agreement (MTA).
Supplementary Materials
Materials and Methods
Figures S1-S12
Movies S1-S2
Acknowledgments
We would like to thank B. Burgering (UMC Utrecht) and S. Kaech (Yale University) for access to the seahorse machine; R. Medzhitov (Yale University) and lab members for discussions; M. Pascha (UMC Utrecht) for help with the macrophages cultures and setting up flow-cytometry protocols; R.J. Soberman (Massachusetts General Hospital and Harvard Medical School) for sharing Cd200r−/− mice; W. Muller (University of Manchester) for providing us with the LysMcre x Il10rflox mice; Dr. H. Karasuyama (Tokyo Medical and Dental University) for the gm609/iSec1 expression plasmid; F. Baixauli and E. Pearce (both Max Planck Institute of Immunobiology and Epigenetics) for MitoDendra2 expressing bone marrow; Y. Adolfs (UMC Utrecht) for assisting with lighsheet microscopy; P. Vader (UMC Utrecht) for help with particle analysis, and L. van Vliet, M. Bruel, and J. Raemakers (all UMC Utrecht) for their help with the analysis of immunofluorescence pictures.