Nociceptor neurons affect cancer immunosurveillance

Solid tumours are innervated by nerve fibres that arise from the autonomic and sensory peripheral nervous systems1–5. Whether the neo-innervation of tumours by pain-initiating sensory neurons affects cancer immunosurveillance remains unclear. Here we show that melanoma cells interact with nociceptor neurons, leading to increases in their neurite outgrowth, responsiveness to noxious ligands and neuropeptide release. Calcitonin gene-related peptide (CGRP)—one such nociceptor-produced neuropeptide—directly increases the exhaustion of cytotoxic CD8+ T cells, which limits their capacity to eliminate melanoma. Genetic ablation of the TRPV1 lineage, local pharmacological silencing of nociceptors and antagonism of the CGRP receptor RAMP1 all reduced the exhaustion of tumour-infiltrating leukocytes and decreased the growth of tumours, nearly tripling the survival rate of mice that were inoculated with B16F10 melanoma cells. Conversely, CD8+ T cell exhaustion was rescued in sensory-neuron-depleted mice that were treated with local recombinant CGRP. As compared with wild-type CD8+ T cells, Ramp1−/− CD8+ T cells were protected against exhaustion when co-transplanted into tumour-bearing Rag1-deficient mice. Single-cell RNA sequencing of biopsies from patients with melanoma revealed that intratumoral RAMP1-expressing CD8+ T cells were more exhausted than their RAMP1-negative counterparts, whereas overexpression of RAMP1 correlated with a poorer clinical prognosis. Overall, our results suggest that reducing the release of CGRP from tumour-innervating nociceptors could be a strategy to improve anti-tumour immunity by eliminating the immunomodulatory effects of CGRP on cytotoxic CD8+ T cells.


Melanomas are innervated
Although the expression of genes of neuronal origin is not detected by RNA-sequencing approaches in human malignant cells or immune cells (Extended Data Fig. 1a-c), we observed a significant increase in their expression in biopsies from patients with melanoma [19][20][21][22] (Extended Data Fig. 1d).As these clinical data suggested increased innervation of melanomas, we tested for the presence of nociceptor neurons by assessing TRPV1 + neurons in biopsies from patients with melanoma.TRPV1 immunolabelling was increased by around twofold in the tumour compared to adjacent healthy tissue in each of the ten biopsies examined.The numbers of tumour-infiltrating lymphocytes (TILs) correlated (R 2 = 0.63) with increased TRPV1 immunolabelling (Extended Data Fig. 2).These data indicate that melanomas are innervated by sensory neurons and that these neurons may affect the intratumoral numbers of immune cells.
To investigate this in more detail, we inoculated a GFP-expressing melanoma (B16F10-eGFP) cell line into Nav1.8cre ::tdTomato fl/WT mice (Nav1.8 is also known as Scn10a).Twenty-two days after implantation, we found abundant Na V 1.8 + nociceptor neurons around and within the tumour (Fig. 1a).RNA sequencing of samples from B16F10-bearing mice revealed that malignant and melanoma-infiltrating immune cells had no detectable levels of neuronal markers (Nav1.8 or Trpv1), indicating that the Na V 1.8 signal could be ascribed to tumour-infiltrating nerves (Extended Data Fig. 3).We next used an in vitro co-culture approach to assess whether malignant cells modulate the function of nociceptor neurons.When co-cultured, TRPV1 + nociceptors directly extended neurites towards the B16F10-eGFP melanoma cells, and the average length of neurites increased, whereas the overall neuronal arborization or branching decreased (Extended Data Fig. 4a-c).Together, these data indicate that nociceptor outgrowth is enhanced when in proximity to melanoma cells and that skin sensory neuron collaterals sprout directly into the tumour bed.Such tumour neo-innervation may be akin to cancer's neoangiogenesis.

Melanoma cells sensitize nociceptors
Given that melanoma promotes axonogenesis, leading to tumour innervation (Fig. 1a and Extended Data Fig. 2), we examined whether this physical proximity allows melanomas to modulate the sensitivity of the nociceptor.As nociceptor neurons detect signals from the local environment, we measured changes in calcium flux in response to sub-threshold concentrations of various noxious ligands.When nociceptors were cultured without melanoma cells, few responded to the ligands at the concentrations selected.However, the number of responsive neurons increased when they were co-cultured with B16F10 cells (Fig. 1b).Similarly, the amplitude of calcium flux responses to the ligands was greater in lumbar DRG neurons (L3-L5) that were collected ipsilateral to a 14-day tumour inoculation in mice, as compared to those collected from mice that were injected with non-tumorigenic keratinocytes (Extended Data Fig. 4d).Signals released from melanoma, therefore, heighten nociceptor sensitivity.
We next tested whether this neuronal hypersensitivity would lead to an increased release of immunomodulatory neuropeptides.In contrast to B16F10 cells alone, DRG neurons co-cultured with B16F10 cells (5 × 10 4 cells, 96 h) actively release CGRP in the medium (Fig. 1c).These data prompted us to test whether exposure to melanoma alters the transcriptome of nociceptor neurons.To do so, we cultured naive DRG neurons (Trpv1 cre ::-CheRiff-eGFP fl/WT ) alone or in combination with B16F10-mCherry-OVA cells.After 48 h, TRPV1 + nociceptors were purified by fluorescence-activated cell sorting (FACS) and RNA sequenced.Differentially expressed genes (DEGs) were calculated, and Calca-the gene that encodes CGRP-and the NGF receptor Trka (also known as Ntrk1)were found to be overexpressed in nociceptors that were exposed to cancer (Fig. 1d-e and Extended Data Fig. 4e).Overexpression of Trka may help to drive melanoma-induced hypersensitivity to pain, whereas CGRP, when released from activated nociceptors, may immunomodulate TILs.
To identify the mechanism through which melanoma sensitizes nociceptor neurons, we used a co-culture system designed to mimic the interactions that take place in the melanoma microenvironment.Type 1 (T c1 )-stimulated (ex-vivo-activated by CD3 and CD28, IL-12 and anti-IL-4 for 48 h) OVA-specific cytotoxic CD8 + T cells (OT-I mice), naive DRG neurons (Trpv1 cre ::CheRiff-eGFP fl/WT ) and B16F10-mCherry-OVA melanoma cancer cells were cultured alone or in combination.After 48 h, the cells were collected, purified by FACS and RNA sequenced, and DEGs were calculated.Among others, we found that Slpi (secretory leukocyte protease inhibitor) was overexpressed in the melanoma cancer cells when co-cultured with either DRG neurons (around 3.6-fold) or OVA-specific cytotoxic CD8 + T cells (around 270-fold), and when exposed to both populations (around 150-fold) (Fig. 2a,b and Extended Data Fig. 5a-e).We also found that B16F10-mCherry-OVA cells, when co-cultured with naive DRG neurons and OVA-specific cytotoxic CD8 + T cells, increased the secretion of SLPI into the culture medium, with this effect being maximal after 48 h (around 200-fold; Fig. 2c).
In addition to protecting epithelial cells from the activity of serine proteases, SLPI enhances the regeneration of transected retinal ganglion cell axons 23 and the proliferation of neural stem cells 24 .Although these data provide evidence of the effect of SLPI on neurons, its role in nociception is unclear.To address this, we measured whether SLPI directly activates cultured DRG neurons using calcium microscopy.We found that SLPI (0.01-10 ng ml −1 ) activates around 20% of DRG neurons and that-consistent with these neurons being nociceptors-SLPI-sensitive neurons were mostly small (with a mean area of 151 μm 2 ) capsaicin-responsive (around 90%) neurons (Fig. 2d,e and Extended Data Fig. 5f-i).Given that SLPI triggered calcium influx, we investigated whether this is the means by which B16F10 cells drive the release of CGRP from neurons (Fig. 1c).SLPI, when used at a concentration similar to that secreted by melanoma cells (Fig. 2c), induced the release of CGRP from cultured naive DRG neurons (Fig. 2f).Finally, we sought to test whether SLPI can drive pain hypersensitivity in vivo.When administered into the right hindpaw of naive mice, SLPI generated transient thermal hypersensitivity (Extended Data Fig. 5j).
Melanoma-secreted SLPI acts on nociceptors to trigger calcium influx, neuropeptide release and thermal hypersensitivity, which indicates that these sensory neurons detect and react to the presence of cancer cells.Whether this gives the malignant cells a functional advantage over the host cells remains unknown.To assess this, we implanted B16F10-mCherry-OVA cells (intradermally (i.d.), 2 × 10 5 cells) into the hindpaw of eight-week-old male and female mice.We found that mice with larger tumours had a higher proportion of intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells and greater hypersensitivity to thermal pain (not shown).Notably, heightened sensitivity to thermal pain positively correlated (n = 60; R 2 = 0.55, P < 0.0001) with increased frequency in intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells (Fig. 3a; measured on day 13 after implantation).

Melanoma-innervating nociceptors control tumour growth
The expression of adrenergic and cholinergic axon markers in tumours correlates with poor clinical outcome 2 .Gastric tumour denervation limits growth and patients who have undergone vagotomy have lower rates of mortality from intestinal cancer 16,25,26 .To investigate the nature of the three-way interaction between cancer, nociceptors and CD8 + T cells, we next used a syngeneic mouse model of triple-negative melanoma, which is an established model of immunosurveillance 9 .B16F10-mCherry-OVA cells were inoculated (i.d., 5 × 10 5 cells) into eight-week-old male and female nociceptor-ablated (Trpv1 cre ::DTA fl/WT ) or intact (littermate control; Trpv1 WT ::DTA fl/WT ) mice.In nociceptor-ablated male and female mice, the median length of survival increased by 2.5-fold (evaluated until day 22; Fig. 3b).In another set of mice that were analysed 16 days after tumour inoculation, we found that genetic ablation of nociceptors reduced tumour growth (Fig 3c).In addition, nociceptor-ablated mice showed an increase in the total number and relative frequency of cytotoxic (IFNγ + , TNF + or IL-2 + ) tumour-infiltrating CD8 + T cells, but a reduced proportion of PD-1 + LAG3 + TIM3 + CD8 + T cells (Fig. 3d,e and Extended Data Fig. 6a,b).
Up to this point, our data suggest that nociceptor neurons are an upstream driver of intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells.To assess whether this is indeed the case, we mapped out the kinetics of thermal pain hypersensitivity, increased frequency in intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells and tumour growth.When compared to their baseline threshold and to that of sensory-neuron-ablated mice (Trpv1 cre ::DTA fl/WT ; n = 19), eight-week-old littermate control mice (Trpv1 WT ::DTA fl/WT ; n = 96) that were inoculated with B16F10-mCherry-OVA (left hindpaw, i.d., 2 × 10 5 cells) showed significant thermal hypersensitivity on day 7, an effect that peaked on day 21 (Extended Data Fig. 6c).In these mice, the intratumoral frequency of PD-1 + LAG3 + TIM3 + (Extended Data Fig. 6d) or IFNγ + (Extended Data Fig. 6e) CD8 + T cells was significantly increased 12 days after tumour inoculation and peaked on day 19.Finally, B16F10-mCherry-OVA tumour volume peaked on day 22 (Extended Data Fig. 6f).Altogether, these data show that thermal hypersensitivity precedes any significant exhaustion of intratumoral CD8 + T cells by around five days and that pain hypersensitivity develops before the tumour is measurable using a digital caliper (Extended Data Fig. 6g).
Blocking the activity of immune checkpoint proteins releases a cancer-cell-induced 'brake' on the immune system, thereby increasing its ability to eliminate tumours 6,[8][9][10] .Immune checkpoint inhibitors (ICIs), including those that target PD-L1, improve clinical outcomes in patients with metastatic melanoma 8 ; however, the efficacy of ICIs varies considerably among patients, half of whom will not benefit 27 .We set out to assess whether the presence (Trpv1 WT ::DTA fl/WT ) or absence (Trpv1 cre ::DTA fl/WT ) of tumour-innervating nociceptor neurons would affect responsiveness to treatment with anti-PD-L1.Anti-PD-L1 (intraperitoneally (i.p.), days 7, 10, 13 and 16) was given either to mice whose tumour cells (B16F10-mCherry-OVA, i.d., 5 × 10 5 cells) were inoculated on the same day, or to mice with established tumours (around 85 mm 3 ; achieved by inoculating Trpv1 cre ::DTA fl/WT around 3 days before).In both scenarios, ablation of nociceptors increased the anti-PD-L1-mediated reduction in tumours and the infiltration of tumour-specific CD8 + T cells (Extended Data Fig. 6h-k).
To test whether the reduction in tumour growth that was observed in the absence of nociceptor neurons depends on their action on Article immune cells, we compared the respective effects of nociceptors on the growth of an immunogenic and a non-immunogenic isogenic melanoma model.YUMMER1.7 is a highly immunogenic derivative of the Braf V600E Cdkn2a −/− Pten −/− cell line modified by ultraviolet (UV) exposure, and provides a clinically relevant model of melanoma 28 .As in the case of B16F10-OVA, ablation of nociceptors decreased the growth of tumours (Extended Data Fig. 6l) and reduced their frequency in intratumoral PD-1 + LAG3 + TIM3 + CD8 T cells.while increasing their number and cytotoxic potential (IFNγ + or TNF + ; not shown).By contrast, YUMM1.7 (the parental and non-immunogenic 29 counterpart of YUMMER1.7)showed similar tumour growth (Extended Data Fig. 6m) and a similar frequency of intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells in both the presence and the absence of nociceptors (not shown).
Next, we assessed whether these differences were due to nociceptor neurons directly modulating intratumoral T cells.We observed no major changes in tumour growth between nociceptor-intact and nociceptor-ablated mice after systemic depletion of CD8 + (Fig. 3f) or CD3 + (Extended Data Fig. 6n) T cells.Although chemoablation of nociceptor neurons with resiniferatoxin (RTX) reduced tumour growth in B16F10-inoculated wild-type mice (Extended Data Fig. 6o), we found that naive OT-I CD8 + T cells enhanced tumour shrinkage when transplanted in RTX-exposed Rag1 −/− mice (Fig. 3g).In doing so, the chemoablation of nociceptor neurons shielded the naive OT-I CD8 + T cells from undergoing exhaustion (Fig. 3h).These data imply that the slower tumour growth found in Trpv1 cre ::DTA fl/WT and RTX-exposed mice depends on the modulation of CD8 + T cells by nociceptors.
Optogenetic activation of skin nociceptor neurons triggers the antidromic release of neuropeptides that mediate anticipatory immunity against microorganisms 30 and potentiate skin immunity 31 .We used transdermal illumination to stimulate tumour-innervating Na V 1.8 + channelrhodopsin-expressing neurons (Nav1.8cre ::ChR2 fl/WT ).Daily stimulation with blue light enhanced the growth of B16F10 when exposure began in mice bearing visible (around 20 mm 3 ) or well-established (around 200 mm 3 ) tumours (Extended Data Fig. 6p).This increase in tumour volume was also linked to an increase in the intratumoral levels of CGRP, confirming the engagement of pain-transmitting neurons (Extended Data Fig. 6q).Laser exposure had no effect on tumour growth in light-insensitive mice (Nav1.8WT ::ChR2 fl/WT ; not shown).
The neonatal or embryonic ablation of neuronal subsets may lead to compensatory changes.To circumvent this possibility, we silenced neurons using botulinum neurotoxin A (BoNT/A), a neurotoxic protein produced by Clostridium botulinum, which acts by cleaving SNAP25 (ref. 32).BoNT/A causes a long-lasting (20 days) abolition of neurotransmitter release from skin-innervating neurons 33 .BoNT/A reduces tumour growth in prostate cancer 2 and blocks nociceptor-mediated modulation of neutrophils during skin infection 33 .BoNT/A does not affect the function of cultured B16F10 or CD8 + T cells in vitro (Extended Data Fig. 7a-f).When BoNT/A (25 pg μl −1 , 50 μl, five i.d.sites) was administered one and three days before the B16F10-OVA cell inoculation, it reduced subsequent tumour growth and preserved the cytotoxic potential of intratumoral CD8 + T cells (Extended Data Fig. 7g-n; as measured 18 days after inoculation).Pre-treatment with BoNT/A also reduced the growth of YUMMER1.7 tumours and enhanced anti-PD-L1-mediated tumour regression (Extended Data Fig. 7o,p).When administered to mice with established tumours (around 200 mm 3 ), BoNT/A had limited efficacy (Extended Data Fig. 7g-n).BoNT/A also did not affect tumour growth when given to mice in which TRPV1 + nociceptor neurons were genetically ablated (Extended Data Fig. 7o), which suggests that its anti-tumour effectiveness depends on the presence of tumour-innervating nociceptor neurons.
We next tested the anti-tumour efficacy of a proven nociceptor-selective silencing strategy 34 .This protocol uses large-pore ion channels (TRPV1) as cell-specific drug-entry ports to deliver QX-314-a charged and membrane-impermeable form of lidocaine-to block voltage-gated sodium (Na V ) channels.During inflammation, similar to what we observed in tumour microenvironments, these large-pore ion channels open, which allows QX-314 to permeate the neurons and results in a long-lasting electrical blockade 17 .Although QX-314 did not affect cultured B16F10-mCherry-OVA cells or CD8 + T cell function in vitro (Extended Data Fig. 8a-f), we confirmed that it silences tumour-innervating nociceptors in vivo, as shown by reduced B16F10-induced release of CGRP and pain hypersensitivity (Extended Data Fig. 8g-i).We found that vehicle-exposed B16F10-mCherry-OVA-bearing mice succumbed at a 2.7-fold higher rate (P ≤ 0.02) than QX-314-exposed mice (Extended Data Fig. 8j; measured until day 19).As observed 17 days after tumour inoculation, QX-314-mediated silencing of sensory neurons (0.3%; daily i.d., surrounding the tumour) reduced melanoma growth and limited the exhaustion of intratumoral CD8 + T cells (Extended Data Fig. 8k-n).Nociceptor silencing also increased the intratumoral numbers of CD8 + T cells and preserved their cytotoxic potential (IFNγ + or TNF + ) as well as their proliferative capacity (IL-2 + ; Extended Data Fig. 8o-r).Similar to what was observed in nociceptor-ablated mice (Extended Data Fig. 6j-j), silencing tumour-innervating neurons with QX-314 enhanced anti-PD-L1-mediated tumour regression (Extended Data Fig. 8s,t).When administered to mice with an established (around 200 mm 3 ) B16F10-mCherry-OVA tumour, QX-314 still reduced tumour growth and preserved the anti-tumour capacity of CD8 + T cells (Extended Data Fig. 8k-r), suggesting that it could be used as a therapeutic agent in cancer.

CGRP attenuates the activity of RAMP1 + CD8 + T cells
In breast cancer, tumour-specific sympathetic denervation downregulates the expression of PD-L1, PD-1 and FOXP3 in TILs 15 .Human and

Article
mouse cytotoxic CD8 + T cells express multiple neuropeptide receptors (10 or more), including the CGRP receptor RAMP1 (Extended Data Figs.1b and 3b).Given that nociceptors readily interact with CD8 + T cells in culture and that the neuropeptides they release block anti-bacterial immunity 33,[35][36][37] , we aimed to test whether these mediators drive the expression of immune checkpoint receptors in CD8 + T cells.First, splenocyte-isolated CD8 + T cells were cultured under type 1 (T c1 ) CD8 + T cell-stimulating conditions for two days and then co-cultured with DRG neurons for an additional four days.We found that nociceptor stimulation with capsaicin increased the proportion of PD-1 + LAG3 + TIM3 + -expressing CD8 + T cells but decreased the levels of IFNγ + , TNF + and IL-2 + .Capsaicin had no measurable effect on CD8 + T cells in the absence of DRG neurons (Extended Data Fig. 9a,b).When T c1 -activated CD8 + T cells were exposed to fresh conditioned medium (1:2 dilution) collected from KCl (50 mM)-stimulated DRG neurons, this treatment increased the proportion of PD-1 + LAG3 + TIM3 + cytotoxic CD8 + T cells and reduced that of IFNγ + cells (Extended Data Fig. 9c,d; measured after four days of co-culture).These effects were prevented when the cytotoxic CD8 + T cells were challenged (1:2 dilution) with fresh KCl-induced conditioned medium from BoNT/A-silenced neurons (50 pg per 200 μl) or when they were co-exposed to the RAMP1 blocker CGRP 8-37 (2 μg ml −1 ; Extended Data Fig. 9c,d).To confirm that nociceptor-released neuropeptides drive T cell exhaustion, we exposed T c1 -activated CD8 + T cells to CGRP.CGRP-treated cells expressing wild-type RAMP1 showed increased exhaustion and limited cytotoxic potential.These effects were absent in CGRP-exposed CD8 + T cells that were collected from CGRP-receptor-knockout (Ramp1 −/− ) mice (Fig. 4a and Extended Data Fig. 9e,f).
We then assessed whether neuropeptides released by nociceptor neurons blunt the anti-tumour responses of cytotoxic CD8 + T cells through exhaustion.OT-I cytotoxic T cells induced robust apoptosis of cultured B16F10-mCherry-OVA cells (AnnexinV + 7AAD + B16F10-mCherry-OVA; Extended Data Fig. 9g-i).However, this apoptosis of B16F10-mCherry-OVA cells was decreased when the T cells were exposed to capsaicin-or KCl-stimulated neuron-derived conditioned medium, or when the cells were stimulated with CGRP (Fig. 4b and Extended Data Fig. 9g-i).OT-I cytotoxic T cells did not eliminate cultured B16F10-mCherry-OVA when co-exposed to KCl-induced neuron-conditioned medium supplemented with the RAMP1 blocker CGRP 8-37 (2 μg ml −1 ; Extended Data Fig. 9h).When taken together with previous evidence that CGRP limits the activity of CD8 + T cells 12,38 , our data suggest that, through the CGRP-RAMP1 axis, nociceptors lead to the functional exhaustion of CD8 + T cells, as defined by a simultaneous loss of expression of cytotoxic molecules (that is, IFNγ and TNF) and proliferative capacity (IL-2), increased co-expression of several exhaustion markers (PD-1 + LAG3 + TIM3 + ) and a reduced capacity to eliminate malignant cells.
We then set out to rescue CGRP levels (by daily intratumoral injection) in sensory-neuron-ablated mice and measured the effect on tumour growth and TIL exhaustion.At 11 days after inoculation, CGRP-treated sensory-neuron-ablated mice (Trpv1 cre ::DTA fl/WT ) showed similar tumour growth and CD8 + T cell exhaustion to that of nociceptor-intact mice (Fig. 4c and Extended Data Fig. 10f).Next, we treated tumour-bearing mice with the selective RAMP1 antagonist BIBN4096 (5 mg kg −1 , i.p., once every two days).The latter was previously found to block neuroimmune interactions during microorganism infections and rescues host anti-bacterial activity 35 .BIBN4096-exposed mice succumb at a rate 2.6-fold lower (P ≤ 0.02) than that of vehicle-exposed B16F10-bearing mice (Extended Data Fig. 10g; measured until day 19).As measured In the presence of IL-2 (10 ng ml −1 ), the cells were stimulated with CGRP (100 nM; challenged once every two days) for 96 h.Wild-type cytotoxic CD8 + T cells showed an increased proportion of PD-1 + LAG3 + TIM3 + cells; this effect was absent when treating cytotoxic CD8 + T cells that were collected from Ramp1 −/− mice (a).In co-culture (48 h), CGRP (100 nM; once daily) also reduced the ability of OT-I cytotoxic CD8 + T cells (4 × 10 5 cells) to eliminate B16F10-mCherry-OVA cancer cells (b).c, Orthotopic B16F10-mCherry-OVA cells (5 × 10 5 cells, i.d.) were inoculated into eight-week-old female mice with sensory neurons intact or ablated.In nociceptor-ablated mice, peritumoral recombinant CGRP injection (100 nM, i.d., once daily) rescues B16F10 growth (assessed until day 12).d,e, Orthotopic B16F10-mCherry-OVA cells (5 × 10 5 cells, i.d.) were inoculated into eight-week-old male and female mice.Starting one day after inoculation (defined as prophylactic), the RAMP1 antagonist BIBN4096 (5 mg kg −1 ) was administered systemically (i.p.) once every two days.In another group of mice, BIBN4096 (5 mg kg −1 , i.p., every two days) injections were started once the tumour reached a volume of around 200 mm 3 (defined as therapeutic).Prophylactic or therapeutic BIBN4096 treatments decreased tumour growth (d) and reduced the proportion of intratumoral PD-1 + LAG3 + TIM3 + CD8 + T cells (e; assessed until day 13).Data are shown as box-and-whisker plots (as defined in Fig. 1b, c on day 13, BIBN4096 (5 mg kg −1 , i.p., every other day) reduced B16F10 growth, tumour weight and frequency of PD-1 + LAG3 + TIM3 + CD8 + T cells (Fig. 4d-e and Extended Data Fig. 10h-m).As BIBN4096 showed no effect when administered to nociceptor-ablated mice and did not affect cultured B16F10 cells or CD8 + T cell function in vitro (Extended Data Fig. 10n-t), we conclude that the anti-tumour property of BIBN4096 relies on the presence of active nociceptor neurons.
When compared with benign nevi, patient melanomas showed increased expression of Calca (Extended Data Fig. 1d).Along with other markers of nociceptor neurons, overexpression of RAMP1 in these biopsies 40 correlates (P ≤ 0.05) with reduced patient survival (Fig. 5d and Extended Data Fig. 11a-l).Whether RAMP1 does this by affecting intratumoral CD8 + T cell exhaustion is unknown.
To answer this, we analysed two independent unbiased single-cell RNA-sequencing datasets of human melanomas 41,42 , and found that around 1.5% of tumour-infiltrating CD8 + T cells expressed RAMP1.The melanoma-infiltrating RAMP1 + CD8 + T cells of the patients overexpressed the immune checkpoint receptors PD-1 (also known as PDCD1), TIM3 (HAVCR2), LAG3, CTLA4 and CD27 (Fig. 5e and Extended Data Fig. 11m).This analysis also revealed that tumour-infiltrating CD8 + cells collected from patients who were resistant to ICIs markedly overexpressed RAMP1 (Extended Data Fig. 11n-p).Such an expression profile resembles the functional exhaustion of effector CD8 + T cells and suggests that the CGRP receptor RAMP1 influences CD8 + T cell exhaustion and the clinical response to ICI in patients with melanoma.

Article
modelling in mice, human data imply that RAMP1-expressing CD8 + T cells are more prone to exhaustion and are associated with lower responsiveness to ICIs.
Tumour-innervating nociceptors dampen the immune response to melanoma by upregulating multiple immune checkpoint receptors on cytotoxic CD8 + T cells.Blocking the CGRP-RAMP1 axis attenuates this immunomodulatory action of the nervous system on CD8 + T cells, thereby safeguarding the anti-tumour immunity of the host (Extended Data Fig. 12) and providing potential therapeutic opportunities by interrupting pro-cancerous neuro-immune links.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-05374-w.

Secondary use of biopsies as research specimens
The ten melanoma samples used in this study were collected by Sanford Health and classified by a board-certified pathologist.Their secondary use as research specimens (fully de-identified formalin-fixed, paraffin-embedded (FFPE) blocks) was approved under Sanford Health IRB protocol 640 (titled 'Understanding and improving cancer treatment of solid tumours').As part of this Institutional Review Board (IRB)-approved retrospective tissue analysis, and in accordance with the US Department of Health and Human Services (HHS) secretary's advisory committee on human research protections, no patient consent was necessary as these secondary use specimens were free of linkers or identifiers and posed no more than minimal risk to the human individuals.

Immunohistochemistry and scoring
In compliance with all the relevant ethical regulations and as approved by Sanford Health IRB protocol 640, ten fully de-identified FFPE melanoma blocks were randomly selected for secondary use as research specimens.The notes of a board-certified pathologist on these specimens are provided in Supplementary Table 1.The specimens were stained using a BenchMark XT slide staining system (Ventana Medical Systems).The Ventana iView DAB detection kit was used as the chromogen, and slides were counterstained with haematoxylin and anti-TRPV1 (Alomone Labs, ACC-030; 1:100).Haematoxylin and eosin (H&E) staining followed standard procedures.TRPV1 immunohistochemistry-stained specimens were analysed on an Olympus BX51 bright-field microscope.Sections were viewed under 20× magnification.Five random fields per sample for both tumour and adjacent normal tissue were analysed and scored on a scale from 0 to 3. Scores were averaged.A score of 0 indicates no appreciated nerve fibres in the evaluated field; +1 indicates sparse nerve fibres; +2 indicates 5-20 nerve fibres; +3 indicates more than 20 nerve fibres.

IACUC approval
The Institutional Animal Care and Use Committee (IACUC) of Boston Children's Hospital and of the Université de Montréal (Comité de Déontologie de l'Expérimentation sur les Animaux; #21046; 21047) approved all animal procedures.

Housing of mice
Mice were housed in standard environmental conditions (12-h lightdark cycle; 23 °C; food and water ad libitum) at facilities accredited by the Canadian Council of Animal Care (Université de Montréal) or the Association for Assessment and Accreditation of Laboratory Animal Care (Boston Children's Hospital).

IACUC end-points
As per our IACUC-approved protocol, the following end-points were used in all of the experiments and were not exceeded.Along with excessive body weight loss (maximum of 10%), the end-points include excessive tumour volume (10% of the mouse's body weight; around 17 mm × 17 mm), skin ulceration, necrosis, bleeding, infection and self-inflicted injury, prostration, lethargy, unresponsiveness to stimulation and/or lack of grooming.
We used the Cre-lox toolbox to engineer the various mice lines used (Trpv1 cre ::DTA fl/WT , Trpv1 cre ::CheRiff-eGFP fl/WT , Trpv1 cre ::tdTomato fl/WT , Nav1.8 cre ::DTA fl/WT , Nav1.8 cre ::tdTomato fl/WT ; Nav1.8 cre ::ChR2 fl/WT and littermate control) by crossing heterozygote Cre mice with homozygous loxP mice.Mice of both sexes were used for these crosses.All Cre driver lines used were viable and fertile, and abnormal phenotypes were not detected.Offspring were tail-clipped and tissue was used to assess the presence of the transgene by standard PCR, as described by The Jackson Laboratory or the donating investigators.Offspring of both sexes were used at 6-20 weeks of age.
All the cell lines tested negative for mycoplasma, and none are listed by the International Cell Line Authentication Committee registry (v.11).Non-commercial cell lines (B16F10-OVA, B16F10-OVA-mCherry and B16F10-eGFP) were authenticated using antibodies (against OVA, eGFP and mCherry) and/or imaging as well as morphology and growth properties.Commercial cell lines were not further authenticated.

BoNT/A
BoNT/A 35 (List Biological Labs, 130B) was injected (25 pg μl −1 , i.d., five neighbouring sites injected with 20 μl) into the skin three days and one day before tumour inoculation (defined as prophylactic).BoNT/A (25 pg μl −1 ; i.d., five neighbouring sites injected with 20 μl) was injected around the tumour one day and three days after the tumour reached a volume of around 200 mm 3 (defined as therapeutic) in other groups of C57BL/6J mice.
In another group of mice, BIBN4096 (5 mg kg −1 ) was administered systemically (i.p., 50 μl) on alternate days once the tumour reached a volume of around 200 mm 3 (defined as therapeutic).RTX RTX (ref. 33; Alomone Labs, R-400) was injected (subcutaneously; s.c.) in three dosages (30, 70 and 100 μg kg −1 ) into the right flank of Rag1 −/− and C57BL/6J mice of around three weeks of age.Denervation was confirmed 28 days after RTX by an absence of pain withdrawal reflex (paw flinching) when exposed to heat (see 'Thermal hypersensitivity' for details on the test).

Survival
In specific groups of mice, orthotopic B16F10-mCherry-OVA (5 × 10 5 cells, i.d.) cells were administered to intact and nociceptor-ablated mice and survival was measured until day 22 and determined by the tumour reaching a volume of 1,000 mm 3 or greater, or according to the ethical end-points described above.In B16F10-mCherry-OVA-inoculated mice treated with QX-314 or BIBN4096, survival was measured until day 19 and determined by the tumour reaching a volume of 800 mm 3 or greater, or according to the ethical end-points described above.As the survival analysis of vehicle-injected, QX-314-treated and BIBN4096-treated mice was performed simultaneously, the same group of vehicle-injected mice is shown in the respective panels for QX-314 and BIBN4096.

iDISCO imaging
Whole-mount immunohistochemistry of tumours was performed using an iDISCO protocol 54,55 with methanol pre-treatment optimized for tumours.In brief, adult mice (eight weeks old) were perfused with 25 ml of PBS (HyClone) and 25 ml of 4% paraformaldehyde (PFA; Sigma) sequentially at room temperature.Tumours were post-fixed with 4% PFA for 6 h at 4 °C.For methanol pre-treatment, fixed tumours were washed sequentially in 50% methanol (in PBS) for 1 h and 100% methanol for 1 h, and then bleached in 5% H 2 O 2 in 20% DMSO and methanol overnight at 4 °C.Tumours were subsequently rehydrated in 100% methanol for 1 h twice, 20% DMSO and methanol for 1 h twice, 50% methanol in PBS for 1 h, PBS for 1 h twice and PBS and 0.2% Triton X-100 for 1 h twice at room temperature.Tumours were then left in PBS, 0.2% Triton X-100, 20% DMSO and 0.3 M glycine (Sigma) overnight at room temperature and blocked in PBS, 0.2% Triton X-100, 10% DMSO, 6% donkey serum ( Jackson ImmunoResearch) and anti-CD16/CD32 (Fc block; Bio X Cell) overnight at room temperature.Tumours were subsequently washed in PBS, 0.2% Tween-20 and 10 mg ml −1 heparin (PTwH; Sigma) for 1 h twice at room temperature before incubation with antibody mix (GFP (Aves Labs) at 1:500, mCherry (OriGene) at 1:500, in PTwH, 5% DMSO, 3% donkey serum and Fc block 1:100 for four days at room temperature).Tumours were extensively washed in PTwH at least six times over the course of one day at room temperature.Tumours were further incubated with a secondary panel of species-specific anti-IgG (H+L) Alexa Fluor 488 or 546-conjugated antibodies (Invitrogen or Jackson ImmunoResearch), all at 1:500, in PTwH, 5% DMSO, 3% donkey serum and Fc block 1:100 for three more days at room temperature.Tumours were washed in the same way as after primary antibody incubation for one day.Immunolabelled tumours were then processed for clearing, which included sequential incubation with 50% methanol for 1 h at room temperature, 100% methanol for 1 h three times at room temperature, and a mixture of one part benzyl alcohol (Sigma):two parts benzyl benzoates (Sigma) overnight at 4 °C.For tdTomato and GFP immunolabelling, mCherry and GFP antibodies were preabsorbed against tumours from tdTomato − mice overnight at room temperature before use.Cleared whole-mount tissues were imaged in BABB between two cover glasses using Olympus FV3000 confocal imaging s ys te m.

In vivo CGRP rescue experiment
Trpv1-ablated mice were injected (i.d.) once daily with recombinant CGRP (100 nM) at five points around the tumour (treatment began once the tumour was visible), and tumour growth was measured daily by a handheld digital caliper.Mice were euthanized, and tumour-infiltrating CD8 + cell exhaustion was immunophenotyped by flow cytometry using an LSRFortessa or a FACSCanto II (Becton Dickinson).
Anti-PD-L1 treatment in mice with similar tumour sizes Orthotopic B16F10-mCherry-OVA cells (5 × 10 5 cells; i.d.) were injected into a cohort of nociceptor neuron-ablated mice three days before nociceptor-intact mice were injected.Mice from each group with a similar tumour size (around 85 mm 3 ) were selected and exposed to anti-PD-L1 14,57 (Bio X Cell, BE0101, 6 mg kg −1 , i.p., 50 μl) or isotype control once every three days for a total of nine days.The effect of anti-PD-L1 treatment on tumour growth was analysed until day 18.
One and three days before tumour inoculation, the skin of eight-week-old male and female mice was injected with BoNT/A (25 pg μl −1 , i.d., five neighbouring sites injected with 20 μl) or vehicle.One day after the last injection, orthotopic B16F10-mCherry-OVA cells (5 × 10 5 cells, i.d.) were inoculated into the area pre-exposed to BoNT/A.On days 7, 10, 13 and 16 after tumour inoculation, the mice were exposed to anti-PD-L1 (6 mg kg −1 , i.p.) or isotype control.Eighteen days after tumour inoculation, we found that neuron silencing using BoNT/A potentiated anti-PD-L1-mediated tumour reduction.
Orthotopic B16F10-mCherry-OVA (5 × 10 5 cells, i.d.) were injected into mice treated with QX-314 (0.3%, i.d.) two to three days before being given to vehicle-exposed mice.Mice from each group with a similar tumour size (around 100 mm 3 ) were selected and exposed to anti-PD-L1 14 (Bio X Cell, BE0101, 6 mg kg −1 , i.p.) or isotype control once every three days for a total of nine days.Eighteen days after tumour inoculation, the effect of anti-PD-L1 on tumour growth was analysed, and TIL exhaustion was immunophenotyped using an LSRFortessa or a FACSCanto II (Becton Dickinson).

RNA sequencing of adoptive transferred Ramp1 WT or Ramp1 −/− CD8 T cells
For FACS-purified cells, Ramp1 −/− and Ramp1 WT CD8 + T cell RNAsequencing libraries were constructed using KAPA Hyperprep RNA (1 × 75 bp) following the manufacturer's instructions.Nextseq500 (0.5 Flowcell High Output; 200 M defragments; 75 cycles single-end read) sequencing was performed on site at the Institute for Research in Immunology and Cancer (IRIC) genomic centre.Sequences were trimmed for sequencing adapters and low-quality 3′ bases using Trimmomatic v.0.35 and aligned to the reference mouse genome version GRCm38 (gene annotation from Gencode v.M23, based on Ensembl 98) using STAR v.2.5.1b (ref. 58).Gene expression levels were obtained both as a read count directly from STAR and computed using RSEM to obtain normalized gene and transcript level expression, in TPM values, for these stranded RNA libraries.DESeq2 v.1.18.1 (ref. 59) was then used to normalize gene read counts.Individual cell data are shown as a log 10 of (TPM × 1,000).These data have been deposited in the National Center for Biotechnology Information (NCBI)'s Gene Expression Omnibus (GEO) 60

Mechanical hypersensitivity
B16F10-mCherry-OVA (2 × 10 5 cells, i.d.) or non-cancerous keratinocytes (MPEK-BL6; (2 × 10 5 cells, i.d.) were inoculated intradermally in the left hindpaw of the mice.On alternate days, mechanical sensitivity was evaluated using von Frey filaments (Ugo Basile, 52-37450-275).To do so, the mice were placed in a test cage with a wire mesh floor and allowed to acclimatize (three consecutive days: 1 h per session).Von Frey filaments of increasing size (0.008-2 g) were applied to the plantar surface and the response rate was evaluated using the up-down test paradigm 61 .

Thermal hypersensitivity
To measure thermal sensitivity, the mice were placed on a glass plate of a Hargreaves's apparatus (Ugo Basile) 62 and stimulated using radiant heat (infrared beam).The infrared beam intensity was set at 44 and calibrated to result in a withdrawal time of around 12 seconds in acclimatized wild-type mice.An automatic cut-off was set to 25 s to avoid tissue damage.The radiant heat source was applied to the dorsal surface of the hindpaw and latency was measured as the time for the mouse to lift, lick or withdraw the paw 62 .
Before any treatment, the mice were allowed to acclimatize in the apparatus (minimum of three consecutive days: 1 h per session) and three baseline measurements were taken on the following day.In some instances, B16F10-mCherry-OVA (2 × 10 5 cells; i.d.) or non-cancerous keratinocytes (MPEK-BL6; (2 × 10 5 cells; i.d.) were inoculated intradermally to the mouse's left hindpaw and thermal pain hypersensitivity was measured on alternate days (10:00).In other instances, SLPI (1 μg per 20 μl) or saline (20 μl) were injected in the left and right hindpaw, respectively, and thermal hypersensitivity was measured in both hindpaws at 1, 3 and 6 h after treatment.

Kinetics of pain and intratumoral CD8 T cell exhaustion
We implanted B16F10-mCherry-OVA (2 × 10 5 cells, i.d.) in several groups of littermate control (Trpv1 WT ::DTA fl/WT ; n = 96) and nociceptor-ablated (Trpv1 cre ::DTA fl/WT ; n = 18) mice.We then evaluated the level of thermal hypersensitivity (daily), tumour size (handheld digital caliper), and intratumoral CD8 + T cell exhaustion (flow cytometry) at the time of euthanasia (days 1, 4, 7, 8, 12, 13, 14, 19 and 22).We processed these data by determining the percentage change of each data point to the maximal value obtained in the pain, CD8 + T cell exhaustion and tumour size datasets, and then presented these data as percentages of the maximum (100%).

Calcium imaging
L3-L5 DRG neurons were collected and co-cultured with B16F10, B16F0 or MPEK-BL6 for 24-48 h.The cells were then loaded with 5 mM Fura-2 AM (BioVision, 2243) in complete Neurobasal-A medium for 30 min at 37 °C, washed in Standard Extracellular Solution (SES, 145 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose and 10 mM HEPES, pH 7.5), and the response to noxious ligands (100 nM capsaicin, 100 μM AITC or 1 μM ATP) was analysed at room temperature.Ligands were flowed (15 s) directly onto neurons using perfusion barrels followed by buffer washout (105-s minimum).Cells were illuminated by a UV light source (Xenon lamp, 75 watts, Nikon), 340-nm and 380-nm excitation alternated by an LEP MAC 5000 filter wheel (Spectra services), and fluorescence emission was captured by a Cool SNAP ES camera (Princeton Instruments).The 340/380 ratiometric images were processed, background-corrected and analysed (IPLab software, Scientific Analytics), and Microsoft Excel was used for post-hoc analyses.Responsiveness to a particular ligand was determined by an increase in fluorescence (F 340 /F 380 ) of at least 5-10% above baseline recording (SES).To test neuronal sensitivity in mice inoculated with B16F10 or non-tumorigenic keratinocytes, the mice were euthanized two weeks after inoculation (left hindpaw, i.d.), and L3-L5 DRG neurons were collected and cultured (3 h).Calcium flux to noxious ligands (1 μM capsaicin or 10 μM ATP) was subsequently tested.For SLPI, the DRG neurons were cultured for 24 h, loaded with 5 mM Fura-2 AM in complete Neurobasal-A medium for 45 min at 37 °C and washed into SES, and the responses to noxious ligands (0-10 ng ml −1 of mouse recombinant SLPI (LifeSpan BioSciences, LS-G13637-10), 1 μM capsaicin or 50 mM KCl) were analysed at room temperature.

In vitro stimulation of cytotoxic CD8 + T cells with neuronconditioned medium
Naive or ablated DRG neurons were cultured (48 h) in Neurobasal-A medium supplemented with 0.05 ng μl −1 NGF (Life Technologies, 13257-019) and 0.002 ng μl −1 GDNF (PeproTech, 450-51-10).After 48 h, the neurobasal medium was removed, neurons were washed with PBS and 200 μl per well of T cell medium supplemented with 1 μl ml −1 peptidase inhibitor (Sigma, P1860) and, in certain cases, capsaicin (1 μM) or KCl (50 mM) was added to DRG neurons.The conditioned medium or vehicle were collected after 30 min and added to T c1 CD8 + T cells for another 96 h.The expression of exhaustion markers (PD-1, LAG3 and TIM3) and cytokines (IFNγ, TNF and IL-2) by CD8 + T cells was analysed by flow cytometry using an LSRFortessa or a FACSCanto II (Becton Dickinson).Cytokine expression levels were analysed after in vitro stimulation (PMA-ionomycin; see 'Intracellular cytokine staining').

In vitro stimulation of cytotoxic CD8 + T cells with CGRP
CD8 + T cells were isolated and stimulated under T c1 conditions in a 96-well plate.After 48 h, cells were treated with either CGRP (0.1 μM) or PBS in the presence of peptidase inhibitor (1 μM) for another 96 h.The expression of PD-1, LAG3 and TIM3, as well as IFNγ, TNF and IL-2, was immunophenotyped by flow cytometry using an LSRFortessa or a FACSCanto II (Becton Dickinson).Cytokine expression levels were analysed after in vitro stimulation (PMA-ionomycin; see 'Intracellular cytokine staining').

In vitro silencing of DRG neurons with BoNT/A
Naive DRG neurons (2 × 10 4 ) were seeded in a 96-well plate with neurobasal medium supplemented with NGF and GDNF.Neurons were pre-treated with 50 pg ml −1 of BoNT/A for 24 h.After 24 h, the culture medium was removed, neurons were washed with PBS and 200 μl per well of T cell medium supplemented with 1 μl ml −1 peptidase inhibitor, and KCl (50 mM) was added to DRG neurons.The conditioned medium or vehicle were collected after 30 min and added to T c1 CD8 + T cells for another 96 h.
In vitro RAMP1 blockade CD8 + T cells were treated with CGRP 8-37 (Tocris, 1169) 6 h before being exposed to the neuron-conditioned medium.In other instances, the neuron-conditioned medium was incubated for 1 h with 2 μg ml −1 of CGRP 8-37 before being added to the CD8 + T cells.

ELISA on co-cultures of B16F10 cells, CD8 + T cells and DRG neurons
Levels of SLPI (R&D Systems, DY1735-05) were measured in the cells' supernatant using a commercial ELISA.

Survival of B16F10 cells
A total of 1 × 10 5 B16F10 cells were cultured in six-well plates and challenged with BoNT/A (0-50 pg μl −1 ) for 24 h, QX-314 (0-1%) for 72 h, BIBN4096 (1-8 μM) for 24 h or their vehicle.The survival of B16F10 cells was assessed using anti-Annexin V staining and measured by flow cytometry using an LSRFortessa or a FACSCanto II (Becton Dickinson), or counted using a haemocytometer.

In silico analysis of neuronal expression profiles using RNA-sequencing and microarray datasets
Publicly available RNA gene expression data from seven datasets were downloaded from the NCBI GEO portal 66 .RNA gene expression values of genes of interest were extracted.Expression values from single-cell sequencing were averaged for all cells.To be able to compare expression from datasets that were generated using different techniques (single-cell RNA sequencing, bulk RNA sequencing and microarrays) and normalization methods (TPM, RPKM (reads per kilobase per million mapped reads), RMA (robust multiarray analysis) and UMI (unique molecular identifiers)), all genes of interest were ratioed over TRPV1 expression, then multiplied by 100, and the log 10 values of these values were plotted as a heat map 66 .Kupari et al. 67 used single-cell RNA sequencing of JNC neurons, whereas Usoskin et al. 68 .and Li et al. 69 .used single-cell RNA sequencing of lumbar neurons.Chiu et al. 70 measured gene expression by microarrays of whole and FACS-sorted Na V 1.8 + lumbar neurons.Goswami et al. 71 performed RNA sequencing of TRPV1 + lumbar neurons, whereas Ray et al. 72 performed RNA sequencing of human lumbar neurons.

In silico analysis of tumour expression profiles of patients with melanoma using single-cell RNA sequencing
Using the publicly available Broad Institute single-cell bioportal, we performed an in silico analysis of single-cell RNA sequencing of human melanoma biopsies.We assessed the gene profile of RAMP1-expressing and RAMP1-negative T cells in the tumours of patients with metastatic melanoma 42 .Similarly, we assessed the genetic program of RAMP1-expressing and RAMP1-negative CD8 + T cells in patients with melanoma 41 .The latter dataset was also used to analyse the genetic profile of CD8 + T cells in patients who were responsive to immune checkpoint blockers or unresponsive to such treatment, as well as the genetic profile of malignant melanoma cells (defined as CD90 − CD45 − ) from the biopsies of ten different patients 41 .Individual cell data are shown as a log 2 -transformed 1 + (TPM/10).Experimental details and cell clustering have been defined in previous studies 41,42 .

In silico analysis of the expression profiles of human immune cells
Publicly available RNA gene expression data from a previous study 73 were downloaded from the NCBI GEO portal.Read counts normalized to transcripts per million protein-coding genes (pTPM) values for genes of interest were extracted.Expression values from single-cell sequencing were averaged for all cells.Experimental details and cell clustering have been defined before 73 .

In silico analysis of the expression profiles of cultured B16F10 cells
Publicly available RNA gene expression data from a previous study 74 were downloaded from the NCBI GEO portal.Read counts normalized to TPM for genes of interest were extracted.Experimental details and cell clustering have been defined before 74 .

In silico analysis of the expression profiles of mouse immune cells using the ImmGen database
Using the publicly available ImmGen database, we performed an in silico analysis of RNA-sequencing data (DESeq2 data) of various mouse immune cells.As per ImmGen protocol, RNA-sequencing reads were aligned to the mouse genome GENCODE GRCm38/mm10 primary assembly (GenBank assembly accession GCA_000001635.2) and gene annotations vM16 with STAR 2.5.4a.The ribosomal RNA gene annotations were removed from the general transfer format file.The gene-level quantification was calculated by featureCounts.Raw read count tables were normalized by the median of ratios method with the DESeq2 package from Bioconductor and then converted to GCT and CLS format.Samples with fewer than 1 million uniquely mapped reads were automatically excluded from normalization.Experimental details can be found at https://www.immgen.org/Protocols/ImmGenULI_RNAseq_methods.pdf.

Oncomine
In silico analysis of the expression profiles of biopsies from patients with melanoma using bulk microarray sequencing.As described previously 19 , samples from 45 cutaneous melanomas and 18 benign melanocytic skin nevus biopsies (around 5-20 μm) were collected and amplified, and their transcriptomes were profiled using Affymetrix U133A microarrays.Data were downloaded from the Oncomine database (https://www.oncomine.com/)as log 2 -transformed (median centred intensity) and genes of interest were shown as heat maps.Experimental details and cell clustering have been defined before 19 .

Survival analysis of patients with melanoma
OncoLnc (http://www.oncolnc.org/)contains survival data for 8,647 patients from 21 cancer studies performed by TCGA 40 .Using OncoLnc, we assessed the transcript expression of a user-defined list of 333 neuronal-enriched genes (neuronal membrane proteins, neural stem cell markers, transcription factors, ion channel receptors and neuropeptides) in 459 skin cancer (SKCM) tumour biopsies from the TCGA database.Of these genes, 206 were expressed, and 108 were selected on the basis of their negative Cox coefficient value, indicating a link between lower gene expression and improved patient survival.Kaplan-Meier curves show the survival of the patients after segregation into two groups defined by their low or high expression of a gene of interest.Details of patients can be found in TCGA 40 and computational analyses can be found at https://doi.org/10.7717/peerj-cs.67.

Sample size
Statistical methods were not used to predetermine sample size.The size of the cohort, based on similar studies in the field, was validated by pilot studies.All sample sizes are indicated in the figures and/or figure legends.All n values are indicated within the figure legends.In the only case in which a range is used (Fig. 2a,b), exact n values are provided in the source data files.For in vivo experiments, we used n > 8 mice.For in vitro experiments in which replicate samples were used, we repeated the experiments at least three independent times to confirm the findings.For other mouse experiments a minimum of five mice were used to ensure that proper statistics could be used.We determined this to be sufficient as per our pilot data, use of internal controls and/or the observed variability between within experimental groups.

Replication
The number of replicates is indicated in the figures, figure legends and/or methods.On the graphs, individual dots represent individual samples or mice used.For each experiment, all attempts at replication were successful and our findings showed comparable results.

Randomization
Breeding pairs and their offspring (nociceptor-intact and -ablated mice) were co-housed and, in respect with the ARRIVE guidelines 75 , were randomly allocated into each experimental group.For in vitro experiments, randomization was used for treatment selection.In some calcium microscopy experiments, the investigators performing the data collection were tasked to select all ligand-responsive cells for downstream analysis.In these rare cases.randomization was not used for cell selection.

Blinding
Double blinding was used for all in vivo treatments.In calcium microscopy experiments involving co-culture (for example, nociceptors and cancer cells), the differences in cell morphology are obvious and, therefore, the investigator performing the experiment was not blind.However, this investigator was always blinded to the treatment being applied to the cells and a second blinded investigator performed the downstream data analysis.
Extended Data Fig. 7 | BoNT/A silencing of B16F10-innervating neurons decreases tumour growth.(a-e) Splenocytes-isolated CD8 + T cells from naive C57BL6J mice were cultured under T c1 -stimulating conditions (ex vivo activated by CD3 and CD28, IL-12, and anti-IL4) for 48h.The cells were then exposed to BoNT/A (10-50 pg/μL) for 24h; effects on apoptosis, exhaustion, and activation were measured by flow cytometry.When compared to vehicle-exposed cells, BoNT/A did not affect the survival (a) of cultured cytotoxic CD8 + T cells, nor their relative expression of IFNγ + (b), TNF + (c), IL-2 + (d) and PD-1 + LAG3 + TIM3 + (e).(f) B16F10 (1x10 5 cells) were cultured for 24h and subsequently exposed to BoNT/A (1.6-50 pg/μL) or its vehicle for an additional 24h.BoNT/A did not trigger B16F10 cells apoptosis, as measured by the mean fluorescence intensity of Annexin V. (g-n) One and three days prior to tumour inoculation (defined as prophylactic), the skin of 8-week-old male and female mice was injected with BoNT/A (25 pg/μL; i.d.) or its vehicle.One day after the last injection, orthotopic B16F10-mCherry-OVA (5x10 5 cells; i.d.) were inoculated into the area pre-exposed to BoNT/A.In another group of mice, BoNT/A was administered (25 pg/μL; i.d.) one and three days after the tumour reached a volume of ~200mm3 (defined as therapeutic).The effect of neuron silencing on tumour size and tumour-infiltrating CD8 + T cell exhaustion was measured.Nineteen days post tumour inoculation, we found that the tumour volume (g,h) and weight (i) were reduced in mice treated with BoNT/A (Prophylactic group).In parallel, we found that silencing tumour-innervating neurons increased the proportion of IFNγ + (k), TNF + (l), and IL-2 + (m) CD8 + T cells.BoNT/A had no effect on the total number of intratumoral CD8 T cells (j) or the relative proportion of PD-1 + LAG3 + TIM3 + (n) CD8 + T cells.(o) One and three days prior to tumour inoculation, the skin of 8-week-old male and female sensory neuron-intact or ablated mice was injected with BoNT/A (25 pg/μL; i.d.) or its vehicle.One day following the last injection, orthotopic YUMMER1.7 cells (5×10 5 cells; i.d.) were inoculated into the area pre-exposed to BoNT/A.The effects of nociceptor neuron ablation on tumour size and volume were measured.Thirteen days post tumour inoculation, we found that the tumour growth was lower in mice treated with BoNT/A or in sensory neuron-ablated mice.BoNT/A had no additive effects when administered to sensory neuronablated mice.(p) One and three days prior to tumour inoculation, the skin of 8-week-old male and female mice was injected with BoNT/A (25 pg/μL; i.d.) or its vehicle.One day following the last injection, orthotopic B16F10-mCherry-OVA cells (5×10 5 cells; i.d.) were inoculated into the area pre-exposed to BoNT/A.On days 7, 10, 13 and 16 post tumour inoculation, the mice were exposed to αPD-L1 (6 mg/kg, i.p.) or its isotype control.Eighteen days post tumour inoculation, we found that neuron silencing using BoNT/A potentiated αPD-L1-mediated tumour reduction.Data are shown as box-and-whisker plots Extended Data Fig. 8 | QX-314 silencing of B16F10-innervating neurons reduces tumour growth.(a-e) Splenocytes-isolated CD8 + T cells from naive C57BL6J mice were cultured under T c1 -stimulating conditions (ex vivo activated by CD3 and CD28, IL-12, and anti-IL4) for 48h.The cells were then exposed to QX-314 (50-150 μM) for 24h, effects on apoptosis, exhaustion and activation were measured by flow cytometry.When compared to vehicle-exposed cells, QX-314 did not affect the survival of cultured cytotoxic CD8 + T cells (a), nor their relative expression of PD-1 + LAG3 + TIM3 + (b), IFNγ + (c), TNF + (d) and IL-2 + (e).(f) B16F10 (1x10 5 cells) were cultured for 24h.The cells were then exposed or not to QX-314 (0.1-1%) for an additional 24-72h, and cell count was analysed by bright-field microscopy.QX-314 did not affect B16F10 cells' survival, as measured by relative cell count changes (at each time point) in comparison to vehicle-exposed cells.(g-i) One and three days prior to tumour inoculation, 8-week-old male and female wild-type mice's right hindpaws or flanks were injected with BoNT/A (25 pg/μL; i.d.) or its vehicle.On the following day, orthotopic B16F10 cells (g: 5x10 5 cells; i.d.; h-i: 2x10 5 cells; i.d.) were inoculated into the area pre-exposed to BoNT/A.Starting one day post inoculation, QX-314 (0.3%) or its vehicle was administered (i.d.) once daily in another group of mice.The effects of sensory neuron silencing were tested on neuropeptide release (g), as well as mechanical (h) and thermal pain hypersensitivity (i).First, CGRP levels were increased in B16F10 tumour surrounding skin explant (assessed on day 15) in comparison to control skin; an effect further enhanced by capsaicin (1 μM; 3h) but was absent in skin pre-treated with BoNT/A (25 pg/μL) or QX-314 (0.3%; g).We also found that B16F10 injection induced mechanical (h) and thermal pain hypersensitivities (i) fourteen days post tumour inoculation.These effects were stopped by sensory neuron silencing with QX-314 or BoNT/A (h-i).( j) Orthotopic B16F10-mCherry-OVA cells (5x10 5 cells; i.d.) were inoculated into 8-week-old male and female mice.Starting one day post inoculation, QX-314 (0.3%; i.d.; 5 sites) was injected once daily around the tumour.The effect of nociceptor neuron silencing on tumour size and tumourinfiltrating CD8 + T cell exhaustion was measured.We found that silencing tumour-innervating neurons increased the mice's median length of survival (~270% Mantel-Haenszel hazard ratio; measured on day 19).(k-r) Orthotopic B16F10-mCherry-OVA cells (5x10 5 cells; i.d.) were inoculated into 8-week-old male and female mice.Starting one day post inoculation (defined as prophylactic), In other groups of mice, QX-314 daily injection started once the tumour reached a volume of ~200mm 3 (defined as therapeutic).As measured seventeen days post tumour inoculation, silencing tumour innervation also decreased tumour volume (k,l) and weight (m), as well as the relative proportion of PD-1 + LAG3 + TIM3 + (n) CD8 + T cells.QX-314 treatment also increased the total number of intratumoral CD8 + T cells (o), as well as relative proportion of IFNγ + (p), TNF + (q), and IL-2 + (r) CD8 + T cells.(s-t) Orthotropic B16F10-mCherry-OVA cells (5x10 5 cells, i.d.) were injected into mice treated with QX-314 (0.3%; i.d.) 2-3 days prior to being injected into vehicle-exposed mice.Mice from each group with similar tumour size (~100mm 3 ) were selected and exposed to αPD-L1 (6 mg/kg, i.p.) once every 3 days for a total of 9 days.Eighteen days post tumour inoculation, we found that αPD-L1-reduced tumour growth was higher (~61%) in nociceptor silenced mice than was observed in isotype vehicle-exposed mice (~49%; s-t).Data are shown as box-and-whisker plots (runs from minimal to maximal values; the box extends from 25 th to 75 th percentile and the middle line indicates the median), for which individual data points are given (a-e, g, l-r), as mean ± S.E.M (f,h,i,k,s,t), or as Mantel-Cox regression analysis ( j).Extended Data Fig. 9 | Nociceptor-released CGRP increases cytotoxic CD8 + T cell exhaustion.(a-b) Splenocytes-isolated CD8 + T cells were cultured under T c1 -stimulating condition (ex vivo activated by CD3 and CD28, IL-12, and anti-IL4) for 48h.The cells were then cultured or not with wild-type DRG neurons and exposed to capsaicin (1 μM, challenged once every two days) or its vehicle.As measured after 4 days stimulation, capsaicin-stimulated intact neuron increased the proportion of PD-1 + LAG3 + TIM3 + (a) cytotoxic CD8 + T cells, while it decreased the one of IFNγ + (b).(c-d) Splenocytes-isolated CD8 + T cells were cultured under T c1 -stimulating conditions (ex vivo activated by CD3 and CD28, IL-12, and anti-IL4) for 48h.In the presence of peptidase inhibitors (1 μL/mL), naive DRG neurons were cultured in the presence of BoNT/A (50 pg/mL) or its vehicle for 24h.The cells were then washed, stimulated (30 min) with KCl (50mM), and the conditioned medium collected.On alternate days for 4 days, the cytotoxic CD8 + T cells were exposed or not to a RAMP1 blocker (CGRP 8-37 ; 2 μg/mL) and challenge (1:2 dilution) with fresh KCl-induced conditioned medium from naive, or BoNT/A-silenced neurons.As measured after 4 days stimulation, KCl-stimulated neuron-conditioned medium increased the proportion of PD-1 + LAG3 + TIM3 + (c) cytotoxic CD8 + T cells, while it decreased the one of IFNγ + (d).Such effect was absent when cytotoxic CD8 + T cells were co-exposed to the RAMP1 blocker CGRP 8-37 or challenged with the neuron conditioned medium collected from BoNT/A-silenced neurons (c-d).
and Extended Data Fig. 5f-j and did the in silico analysis for Fig. 5d,e and Extended Data Figs.1a-d, 3 and 11.A.A. performed the experiments in Fig. 3g,h and Extended Data Figs.6o and 9. A.M. performed the experiments in Fig. 3f.Karine Roversi performed the experiments in Fig. 3a and Extended Data Fig. 6c-g,p.Katiane Roversi performed the experiments in Fig. 3a.C.T.L. performed the experiments in Extended Data Fig. 2. A.C.R. performed the experiments in Extended Data Fig. 2. S.H. performed the experiments in Fig. 1a.L.J. did the bioinformatic analysis of the experiments in Figs.1d,e and 2a,b and Extended Data Figs.4e and 5a-e.D.W.V. performed the experiments in Extended Data Fig. 2. S.T. performed the experiments in Fig. 1b,c and Extended Data Figs.4d and 8f,h,i.Competing interests S.T. and C.J.W. have an equity stake in Nocion Therapeutics.S.T. and C.J.W. have deposited a provisional patent (WO 2021/173916) on the use of charged sodium channel blockers to silence nociceptor neurons as a means to safeguard host anti-tumour immunity.Extended Data Fig. 1 | TRPV1, NAV1.8, SNAP25 or RAMP1 transcripts are expressed in patient melanoma biopsies but are not detected in human immune cells or malignant cells.(a-b) In silico analysis of single-cell RNA sequencing of human melanoma-infiltrating cells revealed that Trpv1, Nav1.8 (Scn10a), Snap25 (the molecular target of BoNT/A), Calca (gene encoding for CGRP) transcripts are not detected in malignant melanoma cells (defined as CD90 -CD45 -) from ten different patients' biopsies (a) nor in cancer-associated fibroblasts, macrophage, endothelial, natural killer, T, and B cells (b).Individual cell data are shown as a log 2 of 1 + (transcript per million / 10).Experimental details and cell clustering were defined in Jerby-Arnon et al 41 .N are defined in the figures.(c) In silico analysis of human immune cells revealed their basal expression of Cd45.Using RNA sequencing approaches, Calca, Snap25, Trpv1 or NaV1.8 are not detected in these cells.Heat maps show the read counts normalized to transcripts per million protein-coding genes (pTPM) for each of the single-cell clusters.Experimental details and cell clustering were defined in Monaco et al 73 .(d) Forty-five cutaneous melanomas and 18 benign melanocytic skin nevus biopsies transcriptomes were profiled using Affymetrix U133A microarrays 19 .In silico analysis of this dataset revealed that cutaneous melanoma heightened expression levels of Calca (1.4-fold), Pouf4f1 (2-fold), Eno2 (1.4-fold), and Tubb3 (1.1-fold), as well as other neuronal-enriched genes.Heat map data are shown as log 2 (median centred intensity); two-sided unpaired Student's t-test; p-values and n are shown in the figure.Experimental details were defined in Haqq et al 19 .Extended Data Fig. 2 | TRPV1 + neurons innervate patient melanomas.Patients' melanoma sections were stained with hematoxylin eosin (a-f), and the presence of TRPV1 (d-f; brown) neurons was analysed by immunohistochemistry. Increased levels of TRPV1 + neurons (g) were found in the tumour (delimited by red square; a-b, d-e) compared to adjacent healthy skin (delimited by blue square; a,c,d,f).Increased TRPV1 immunolabelling in tumour sections primarily correlated with enhanced levels of tumour-infiltrating leukocytes (h) as scored from a retrospective correlation analysis performed on the patients' pathology reports.Data are shown as representative immunohistochemistry images (a-f), box-and-whisker plots (runs from minimal to maximal values; the box extends from 25 th to 75 th percentile and the middle line indicates the median), for which individual data points are given (g) or as a heat map (h) displaying Pearson's correlation (R 2 ).N are as follows: a-f: n=10, g: intact (n = 8), tumour (n = 10), h: n = 10.Slides were scored blindly by two experienced medical pathologists.P-values are shown in the figure and determined by two-sided unpaired Student's t-test (g).Scale = 100 μm (a,d), 50 μm (b,c,e,f).Extended Data Fig. 4 | B16F10 cells interact with nociceptor neurons.(a-c) When co-cultured with B16F10-eGFP cells (green), TRPV1 + nociceptor (Trpv1 cre ::tdTomato fl/WT ; orange) neurons form neuro-neoplasic contacts (a), show longer neurites (b), and exhibit reduced arborization (c) than when cultured alone or with non-tumorigenic keratinocytes (b-c).(d) L3-L5 DRG neurons were collected from mice 2-weeks after they were inoculated (left hindpaw; i.d.) with B16F10-or non-tumorigenic keratinocytes, cultured and calcium flux to ligands tested (ATP (10 μM), and capsaicin (1 μM)).Compared to neurons from keratinocytes-injected mice, the one from tumour-bearing mice showed increased sensitivity to capsaicin.(e) Naive DRG neurons (Trpv1 cre ::CheRiff-eGFP fl/WT ) were cultured alone or in combination with B16F10-mCherry-OVA.After 48h, the cells were collected, FACS purified, and RNA sequenced.Hierarchical clustering of sorted neuron DEG show distinct groups of transcripts enriched in TRPV1 + neuron vs cancer-exposed TRPV1 + neuron populations.Pairwise comparison of naive TRPV1 + neuron vs cancer-exposed TRPV1 + neuron populations showing differentially expressed transcripts as a volcano plot (p<0.05).Among others, Calca (gene encoding for CGRP) was overexpressed in TRPV1 + (FACS-purified eGFP-expressing cells) neurons when co-cultured with B16F10-mCherry-OVA.Data are shown as representative image (a), mean ± S.E.M (b-d), or volcano plot (e).N are as follows: a: n = 4, b: neuron (n=8), neuron + keratinocytes (n = 7), neuron + B16F0 (n = 7), neuron + B16F10 (n = 7), c: n = 15/groups, d: keratinocytes inj.+ ATP (n=5), B16F10 inj.+ ATP (n=36), Keratinocytes inj.+ caps (n = 6), B16F10 inj.+ caps (n = 44), e: n = 4/ groups.Experiments were independently repeated two (d) or three (a-c) times with similar results.Sequencing experiment was not repeated (e).P-values are shown in the figure and determined by one-way ANOVA post-hoc Bonferroni (b-c) or two-sided unpaired Student's t-test (d).Scale bar = 100 μm (a).Extended Data Fig. 5 | B16F10-secreted SLPI activates nociceptor neurons.