CGRP Signals from Endosomes of Schwann Cells to Elicit Migraine Pain


 Efficacy of monoclonal antibodies against calcitonin gene-related peptide (CGRP) or its receptor (calcitonin receptor-like receptor/receptor activity modifying protein-1, CLR/RAMP1) implicates peripherally-released CGRP in migraine pain. However, the site and mechanism of CGRP-evoked migraine pain remain unknown. By cell-selective RAMP1 gene deletion, we reveal that CGRP released from mouse cutaneous trigeminal fibers targets CLR/RAMP1 on surrounding Schwann cells to evoke periorbital mechanical allodynia. CLR/RAMP1 activation in human and mouse Schwann cells generates long-lasting signals from endosomes that evoke cAMP-dependent formation of NO. NO, by gating Schwann cell transient receptor potential ankyrin 1 (TRPA1), releases ROS, which in a feed-forward manner sustain allodynia via nociceptor TRPA1. When encapsulated into nanoparticles that release cargo in acidified endosomes, a CLR/RAMP1 antagonist provides superior inhibition of CGRP signaling and allodynia in mice. The CGRP-mediated neuronal/Schwann cell pathway is critical to mediate allodynia associated with neurogenic inflammation, thus contributing to the pro-migraine action of CGRP.


Introduction
For almost a century it has been known that cutaneous tissue injury elicits a local vascular response, referred to as neurogenic in ammation, that is associated to a wider area of increased sensitivity to mechanical stimuli 1 . A subset of C-ber primary afferents, which mediate neurogenic in ammation, is the main source of the neuropeptides substance P (SP) and calcitonin gene related peptide (CGRP) 2, 3 . In rodents, noxious stimuli such as capsaicin, a pungent agonist of the transient receptor potential vanilloid (IMS32) recapitulated features of primary MSCs, including expression of CLR and RAMP1 mRNA and immunoreactivity ( Supplementary Fig. 1a,b) and responsiveness to allyl isothiocyanate, which evoked a TRPA1-dependent increase in Ca 2+ response ( Supplementary Fig. 1c). Immunoreactive CLR and RAMP1 were also detected in S100 + ve Schwann cells in nerve bundles in biopsies of human abdominal and mouse periorbital skin (Fig. 1b).
Schwann cell CLR/RAMP1 mediates the CGRP-dependent PMA evoked by GTN Systemic GTN administration provokes migraine-like pain in humans 19 . In mice intraperitoneal GTN elicits PMA 18 , that is in part mediated by CGRP release from periorbital trigeminal terminals 18 . Here, we show that systemic (intraperitoneal) GTN elicited PMA (Fig. 3a) and paw allodynia (Fig. 3b) that were similar in Control and Adv-Cre + ;Ramp1 / mice. Olcegepant transiently and partially inhibited PMA (Fig. 3a), while did not affect paw allodynia (Fig. 3b) evoked by intraperitoneal GTN in both mouse strains. Deletion of RAMP1 by 4-OHT in Plp-Cre ERT+ ;Ramp1 / mice partially inhibited PMA (Fig. 3c), but not paw allodynia (Fig. 3d). Importantly, treatment with olcegepant reduced GTN-evoked PMA in Control mice, but failed to further inhibit the response in Plp-Cre ERT+ ;Ramp1 / mice (Fig. 3c). Paw allodynia was unchanged by olcegepant (Fig. 3d). Together, the ndings suggest that Schwann cell CLR/RAMP1 mediates the CGRP-dependent component of PMA in an established mouse model of migraine pain.

Clathrin-and dynamin-mediated endocytosis of CLR/RAMP1 in Schwann cells mediates PMA
We investigated the CLR/RAMP1 signaling pathway in Schwann cells that mediates CGRP-evoked PMA. CGRP-stimulated cAMP formation was measured in HSCs using a virally-encoded cAMP cADDis reporter. CGRP stimulated a prompt concentration-dependent increase in cAMP formation in HSCs that was sustained for > 300 s (Fig. 4a,b; Video S1). The CLR/RAMP1 antagonist olcegepant caused a concentration-dependent inhibition of CGRP-stimulated cAMP formation ( Fig. 4c-e).
CLR activates Gα s , Gα q and Gα i and recruits βARR2 to the plasma membrane and endosomes GPCRs, including CLR/RAMP1, can signal from endosomes by Gα s , Gα q and βARR-mediated mechanisms 31,32,33 . We used enhanced bystander bioluminescence resonance energy transfer (EbBRET) to study the activation of Gα and recruitment of βARR to the plasma membrane and early endosomes of HEK293T cells expressing human (h) CLR and RAMP1 (HEK-hCLR/RAMP1). CGRP-dependent activation of Gα s , Gα sq and Gα si was assessed using an EbBRET assay that detects recruitment of mini (m) Gα coupled to Renilla (R)luc8 to the plasma membrane marker CAAX coupled to RGFP or the early endosome marker Rab5a coupled to tandem (td)RGFP. mGα proteins are N-terminally truncated Gα proteins that freely diffuse throughout the cytoplasm and bind to active conformations of GPCRs. Their translocation to GPCRs re ects Gα activation. mGα sq and mGα si were developed by mutating mGα s residues to equivalent Gα q and Gα i residues. Recruitment of βARR was assessed by measuring EbBRET between Rluc2-βARR2 and RGFP-CAAX or tdRGFP-Rab5a. CGRP induced a rapid increase in EbBRET between Rluc8-mGα s , Rluc8-mGα sq , Rluc8-mGα si and Rluc2-βARR2 with RGFP-CAAX, which was maximal at ~ 300 s and declined over 1000 s (Fig. 5a,b). CGRP increased EbBRET between Rluc8-mGα s , Rluc8-mGα sq , Rluc8-mGα si and Rluc2-βARR2 with tdRGFP-Rab5a that was fully sustained for 1300 s (Fig. 5c,d).
To examine the contribution of endosomal CLR/RAMP1 signaling to CGRP-induced cAMP formation, we preincubated HSCs expressing the cADDis cAMP reporter with sucrose or vehicle. In vehicle-treated cells, CGRP stimulated a rapid (1 min) increase in cAMP formation that was sustained for 30 min (Fig. 5n,o). Sucrose reduced but did not abolish the initial response, yet strongly inhibited the sustained phase of CGRP-stimulated cAMP formation (Fig. 5n,o). Thus, CGRP initially activates Gα and βARR at the plasma membrane, which is followed by sustained activation of Gα and βARR in early endosomes. Endocytosis is necessary for the recruitment of Gα and βARR to endosomes. Gα s continues to signal in endosomes, leading to sustained cAMP formation.

CLR/RAMP1 activation in Schwann cells releases NO, which initiates but does not sustain PMA
We investigated the mechanisms that sustain PMA following CLR/RAMP1 activation and endocytosis in Schwann cells. Pretreatment with CLR/RAMP1 antagonists, olcegepant or CGRP8-37, attenuated PMA evoked by capsaicin and in accordance with previous studies 17,18 PMA evoked by CGRP ( Supplementary  Fig. 4a,b). In contrast, antagonists had no effect when administered 60 min after CGRP or capsaicin (Supplementary Fig. 4c-f). Similarly, inhibitors of clathrin-and dynamin-mediated endocytosis had no effect when administered 60 min after CGRP or capsaicin ( Supplementary Fig. 4g-j). Thus, once induced by CGRP, CLR/RAMP1 antagonists or inhibitors of CRL/RAMP1 internalization are unable to attenuate PMA. Pre-but not post-treatment with the protein kinase A (PKA) inhibitor, H89, reduced PMA by CGRP and capsaicin (Supplementary Fig. 4k-n). NO has been implicated in CGRP-mediated vascular responses 2 . Although NO can release CGRP with proalgesic functions, the contribution of NO to CGRPevoked allodynia is uncertain. Pretreatment with an NO synthase (NOS) inhibitor (L-NAME) or an NO scavenger (cPTIO) (Fig. 6a), abrogated CGRP-evoked PMA (Fig. 6b,c). L-NAME and cPTIO pretreatment also attenuated capsaicin-evoked PMA (Fig. 6d,e). However, L-NAME and cPTIO did not affect PMA when administered 60 min after CGRP ( Supplementary Fig. 4o,p) or capsaicin ( Supplementary Fig. 4q,r). Thus, PKA-dependent NO release 37 evoked by CGRP is necessary to initiate, but is not su cient to sustain, allodynia.
In vitro ndings recapitulated in vivo results. HSCs, MSCs and IMS32 cells predominantly expressed NOS3 (eNOS) mRNA, with little or no expression of NOS1 and NOS2 (nNOS and iNOS, respectively) mRNA ( Fig. 6f,g; Supplementary Fig. 4s). In both HSCs and IMS32 cells, CGRP elicited a transient increase in NOS3 phosphorylation (i.e., activation), consistent with NO generation, which peaked at 5-10 min and declined within 30-60 min (Fig. 6h). In HSCs and IMS32 cells, the cAMP increase elicited by CGRP was prevented by olcegepant, CGRP8-37 and an adenylyl cyclase inhibitor (SQ22536), but not by L-NAME (Fig. 6i). The increase in cAMP evoked by CGRP but not that elicited by forskolin was reduced in cultured MSCs obtained from Plp-Cre ERT+ ;Ramp1 / mice as compared to Control mice treated with intraperitoneal 4-OHT (Fig. 6j). In contrast, the CGRP-evoked increase in NO was attenuated by all these interventions, including NOS inhibition (Fig. 6k). The cAMP increase evoked by forskolin was unaffected by CLR/RAMP1 antagonism and NOS inhibition, and olcegepant failed to inhibit NO release by the NO donor NONOate ( Supplementary Fig. 4t,u), indicating selectivity. PS2 and Dy4, but not their inactive analogs, inhibited CGRP-mediated NO release but did not affect NONOate-stimulated NO release ( Supplementary  Fig. 4v), further supporting selectivity. These results suggest that clathrin-and dynamin-dependent endocytosis and endosomal CLR/RAMP1 signaling evoke NOS activation and NO generation in Schwann cells.
Schwann cell TRPA1 mediates CGRP-evoked PMA NO belongs to a series of reactive oxygen species (ROS) that target TRPA1 38 . TRPA1 is coexpressed with TRPV1 and CGRP in a subpopulation of primary sensory neurons 39 . TRPA1 is expressed in Schwann cells of nerve bundles of human skin and mouse sciatic nerve, where it mediates mechanical allodynia in rodent models of pain 28, 40 . Immunoreactive TRPA1 was coexpressed with RAMP1 in S100+ve Schwann cells in human abdominal and mouse periorbital cutaneous nerves bundles (Fig. 7a). Thus, CLR/RAMP1 might engage signaling pathways that activate TRPA1 in trigeminal Schwann cells to initiate allodynia (Fig. 7b). The hypothesis that CLR/RAMP1 activates TRPA1 to induce allodynia was supported by the observation that both CGRP-and capsaicin-evoked PMA were reduced in Trpa1 -/mice and in mice with sensory neuron-speci c deletion of TRPA1 (Adv-Cre + ;Trpa1 / ) (Fig. 7c,d).
We next investigated the signaling pathway by which the CLR/RAMP1 activates TRPA1. In HSCs and IMS32 cells, CGRP stimulated a slowly developing yet sustained increase in Ca 2+ response and increased  Supplementary Fig. 5a). These results support the hypothesis that CGRP liberates NO, which activates Schwann cell TRPA1; activated TRPA1 promotes a Ca 2+ -dependent H 2 O 2 generation that sustains a feed-forward mechanism comprising TRPA1 channel engagement and ROS release.
In vivo results support the existence of a CGRP-evoked proalgesic pathway that is initiated by an NOdependent mechanism and sustained by ROS-mediated TRPA1 engagement. Whereas CLR/RAMP1 antagonists or NO inhibitors attenuated PMA only if given before CGRP or capsaicin, both pre-and posttreatment with a TRPA1 antagonist, a ROS scavenger and a NOX1 inhibitor reduced PMA ( Targeting endosomal CGRP signaling provides superior relief of CGRP-and capsaicin-evoked PMA The nding that persistent GPCR signaling from endosomes mediates pain transmission suggests that GPCRs in endosomes rather than at the plasma membrane are a valid and perhaps superior target for the treatment of pain 31, 32, 33 . Nanoparticles have been used to deliver chemotherapeutics to tumor, where endocytosis and endosomal escape are necessary for drug delivery to cytosolic and nuclear targets 41 . The realization that GPCRs within endosomes are a therapeutic target, raises the possibility of exploiting the acid microenvironment of endosomes as a stimulus for nanoparticle disassembly and release of antagonist cargo 32 . To target CLR in endosomes, we generated self-assembling soft polymer nanoparticles containing a CLR antagonist. Diblock copolymers were synthesized with a hydrophilic shell of P(PEGMA-co-DMAEMA) and a hydrophobic core of P(DIPMA-co-DEGMA) (Fig. 8a). Gel permeation chromatography and 1 H-nuclear magnetic resonance ( 1 H-NMR) con rmed the molecular weight and composition of nanoparticles (Supplementary Fig. 6a-d). Nanoparticles were self-assembled with MK-3207, a potent hydrophobic antagonist of human CLR/RAMP1, forming DIPMA-MK-3207 (Fig. 8a). Empty nanoparticles (DIPMA-Ø) were used as a control. Nanoparticles were uniformly spherical, with similar diameter (30-35 nm) and ζpotential (-0.4-1.3 mV) (Fig. 8b,c). DIPMA nanoparticles demonstrate a pH-dependent cargo release at pH<~6.5, consistent with the protonation of the DIPMA tertiary amine (pK a 6.1), charge repulsion and disassembly 32 . DIPMA nanoparticles enter cells by clathrin-and dynamin-mediated endocytosis and disassemble in acidic early endosomes 32 . To determine whether DIPMA nanoparticles target endosomes containing CLR/RAMP1, HSCs expressing early endosomal antigen-1-GFP (EEA1-GFP) were incubated with DIPMA-Cy5 for 30 min to allow accumulation in EEA1-GFP+ve endosomes (Fig. 7d, Video S6). Cells were then incubated with TAMRA-CGRP, which was detected in endosomes containing Cy5-DIPMA within 5-10 min (Fig. 8d, Video S6). Thus, DIPMA nanoparticles accumulate with CLR/RAMP1 in early endosomes of Schwann cells.
To determine whether DIPMA-MK-3207 can antagonize CLR in endosomes, we measured CGRPstimulated cAMP formation using the CAMYEL cAMP BRET sensor, which detects total cellular cAMP.
HEK293 cells expressing rat CLR/RAMP1 (HEK-rCRL/RAMP1) were preincubated with graded concentrations of DIPMA-MK-3207 or free MK-3207, DIPMA-Ø or vehicle (control) for 30 min. Beginning at 0 min, baseline BRET was measured for 5 min, and cells were then challenged with CGRP. At 10 min, cells were washed to remove extracellular CGRP, and BRET was measured up to 35 min. In vehicle-treated cells, CGRP stimulated a prompt increase in cAMP formation (1 st phase, 6-10 min) that gradually declined after agonist removal from the extracellular uid (2 nd phase, 11-35 min) (Fig. 8e). DIPMA-Ø did not affect this response. Free MK-3207 and DIPMA-MK-3207 (100, 316 nM) both inhibited CGRP-evoked cAMP in the 1 st phase to a similar extent (Fig. 8f). During the 2 nd phase, free MK-3207 was inactive at all concentrations whereas DIPMA-MK3207 (31.6, 100, 316 nM) strongly inhibited responses (Fig. 8g). The results suggest that DIPMA-MK3207 can antagonize the sustained phase of CGRP-stimulated formation of cAMP, which is attributable to endosomal CLR/RAMP1 signaling.
To assess antagonism of the pain signaling pathway in HSCs, we measured CGRP-evoked changes in Ca 2+ response, which depend on endosomal CGRP signaling and activation of TRPA1. HSCs were preincubated with graded concentrations of DIPMA-MK-3207 or MK-3207 for 20 min to allow accumulation in endosomes, and washed to remove extracellular compounds. At 10 min after washing, cells were challenged with CGRP and Ca 2+ response was measured as an index of TRPA1 activity.

Discussion
The major ndings of the present study, supported by genetic and pharmacological evidence, are that CGRP causes migraine pain by activating CLR/RAMP1 of Schwann cells, CLR/RAMP1 signal from endosomes of Schwann cells to activate pain pathways, and endosomal CLR/RAMP1 can be targeted using nanoparticles and endocytosis inhibitors to relieve CGRP-evoked migraine pain. CLR/RAMP1 activation and tra cking to endosomes results in a persistent cAMP-dependent activation of NOS and generation of NO, a mediator of migraine pain 19 . The role of NO in PMA is crucial, yet transient, as it is temporally limited to the engagement of TRPA1/NOX1, which releases ROS with a dual function. On one hand, ROS target TRPA1/NOX1 of Schwann cells to maintain ROS generation by a feed-forward mechanism. On the other hand, as suggested by experiments with selective TRPA1 deletion in primary sensory neurons, ROS target TRPA1 on nociceptors to signal allodynia to the CNS.
Periorbital capsaicin injection elicited acute nociception mediated by TRPV1 excitation and ensuing afferent discharge, which signals pain to the CNS. In a larger cutaneous area, capsaicin evoked delayed and prolonged PMA. While the acute pain response is most likely dependent on ion in ux associated with TRPV1 activation, the mechanism underlying mechanical hypersensitivity 8, 42 has remained elusive. Our ndings support the existence of a paracrine mechanism that underlies PMA associated with neurogenic in ammation. We suggest that capsaicin locally activates TRPV1 + ve nerve bers to generate action potentials that propagate antidromically into collateral bers which release CGRP in a broader area, thus eliciting widespread PMA. PMA depends on the interaction between peptidergic nerve bers, surrounding Schwann cells and nociceptive neurons that convey allodynic signals to the CNS. CGRP liberated from the varicosities of trigeminal TRPV1 + ve nerve bers binds to CLR/RAMP1 of adjacent Schwann cells.
CGRP released from peptidergic C-bers has been hypothesized to target CLR/RAMP1 on adjacent nonpeptidergic Aδ-bers 43 within the node of Ranvier 26 to mediate migraine pain. Present results in mice with selective RAMP1 deletion in primary sensory neurons rmly support the view that nociceptors are not the target of CGRP to evoke PMA, which, instead, implicates the role of Schwann cells that wrap their terminals. A limitation of the present study is that we cannot distinguish between TRPV1-expressing nerve bers that release CGRP and TRPA1-expressing nerve bers that are targeted by Schwann cell ROS and convey allodynic signals centrally. Indeed, TRPV1 and TRPA1 may coexist in the same population of CGRP-expressing Aδ-or C-ber primary sensory neurons 2,39 . Most Schwann cells in Remak bundles contain multiple unmyelinated axons belonging to morphologically and functionally different neurons, including CGRP + ve and isolectin B4 + ve bers 44 . Moreover, the bulk of CGRP is expressed in C-ber nociceptors 2 . Thus, we cannot discriminate between the three possibilities that ROS from Schwann cells target TRPA1 on the same Aδ-or C-ber that releases CGRP, on a different C-ber of the same Remak bundle, or on a different adjacent Aδ-ber. The observation that both C-ber and Aδ-ber nociceptors contribute to capsaicin-evoked hypersensitivity in humans 45  Limitations of the present study include uncertainty about the nature of the CLR/RAMP1 signaling complex in endosomes of Schwann cells, which warrants further investigation by proteomics approaches. Although some of the pharmacological inhibitors used to dissect the signaling pathway can have non-speci c actions, we bolstered con dence in selectivity by using inhibitors of the same pathway and by genetic deletion of GPCRs and TRP channels. Our ndings reveal a prominent role for CLR/RAMP1 in Schwann cells for CGRP-evoked periorbital pain. Future studies will investigate the role of this pathway in preclinical models of migraine pain.
Monoclonal antibodies to CGRP, although bene cial, are not effective in all patients 11 . While non-CGRPdependent mechanisms might explain this failure 49 , CGRP or CLR/RAMP1 antibodies likely do not inhibit CGRP signaling in endosomes. The small molecule CLR/RAMP1 antagonist, rimegepant, was found to resolve migraine attacks in patients treated with the anti-CLR/RAMP1 monoclonal antibody, erenumab 50 . This unexpected result was interpreted by the inherent membrane permeability of the lipophilic antagonist rimegepant 51 that might favor inhibition of CGRP signaling in endosomes 50 , while neither receptor-targeted nor ligand-targeted monoclonal antibodies internalized with CLR/RAMP1 activated by CGRP 52 . Our results showing a superior inhibition of CGRP signaling in Schwann cells and of PMA by DIPMA-MK-3207, which selectively targets receptor activity in endosomes, reveals a better approach to control migraine pain. By inhibiting CLR/RAMP1 translocation to endosomes, inhibitors of clathrin-and dynamin-mediated endocytosis might offer an alternate therapeutic approach for migraine pain.
In 1936, Sir Thomas Lewis postulated 1 that in human skin action potentials are carried antidromically from the injured nerve terminal to collateral branches from where a chemical substance is released that produces the are and increases the sensitivity of other bers responsible for pain. CGRP has been previously identi ed as the mediator responsible for neurogenic vasodilatation in rodents 2 , and in humans 9 . Herein, we propose that CGRP is the 'chemical substance' that, via the essential role of endosomal CLR/RAMP1, TRPA1/NOX1 and oxidative stress of surrounding Schwann cells, sustains the enhanced sensitivity of primary sensory neurons associated with neurogenic in ammation (Fig. 9).
Results obtained in the two mouse models of migraine evoked by systemic CGRP or GTN suggest that the proalgesic pathway activated by CGRP in Schwann cells is implicated in migraine pain and is the target of peripherally acting anti-CGRP medicines.
The group size of n = 8 animals for behavioral experiments was determined by sample size estimation using G*Power (v3.1) 57 to detect size effect in a post-hoc test with type 1 and 2 error rates of 5 and 20%, respectively. Mice were allocated to vehicle or treatment groups using a randomization procedure (http://www.randomizer.org/). Investigators were blinded to the identities (genetic background) and treatments, which were revealed only after data collection. No animals were excluded from experiments. Mechanical stimuli were applied homolaterally outside the periorbital area at a distance of 6-8 mm from the site where stimuli were injected. The response was considered positive by the following criteria: mouse vigorously stroked its face with the forepaw, head withdrawal from the stimulus, or head shaking.
Mechanical stimulation started with the 0.16 g lament. Absence of response after 5 s led to the use of a lament with increased force, whereas a positive response led to the use of a weaker (i.e. lighter) lament. Six measurements were collected for each mouse or until four consecutive positive or negative responses occurred. The 50% mechanical withdrawal threshold (expressed in g) was then calculated from these scores by using a δ value of 0.205, previously determined.
Paw mechanical allodynia. Paw mechanical allodynia was evaluated by measuring the paw withdrawal threshold by using the up-down paradigm 63, 64 . Mice were acclimatized (1 hr) in individual clear plexiglass boxes on an elevated wire mesh platform, to allow for access to the plantar surfaces of the hind paws. von Frey laments of increasing stiffness (0.07, 0.16, 0.4, 0.6 and 1.0, 1.4 and 2 g) were applied to the hind paw plantar surfaces of mice with enough pressure to bend the lament. The absence of a paw being lifted after 5 s led to the use of the next lament with an increased force, whereas a lifted paw indicated a positive response, leading to the use of a subsequently weaker lament. Six measurements were collected for each mouse or until four consecutive positive or negative responses occurred. The 50% mechanical withdrawal threshold (expressed in g) was then calculated.
qRT-PCR. Detailed method and the sets of primers for human and mouse cells are reported in Supplementary information and Table S3. EbBRET assays of G protein and βARR recruitment to the plasma membrane and endosomes cDNAs.
Detailed method is reported in supplementary information.
Synthesis and Characterization of Nanoparticles. Detailed method is reported in supplementary information.
Statistical analysis. Results are expressed as mean ± standard error of the mean (SEM). For multiple comparisons, a one-way analysis of variance (ANOVA) followed by the post-hoc Bonferroni's test or Dunnett's test was used. Two groups were compared using Student's t-test. For behavioral experiments with repeated measures, the two-way mixed model ANOVA followed by the post-hoc Bonferroni's test was used. Statistical analyses were performed on raw data using Graph Pad Prism 8 (GraphPad Software Inc.). IC 50 values and con dence intervals were determined from non-linear regression models using