Introduction

Melanoma represents one of the leading causes of cancer-related deaths, being the most aggressive skin cancer type worldwide [124]. It emerges from molecularly altered melanocytes, which are the producers of melanin in the skin [48]. These cancer cells are embedded within the cutaneous microenvironment where they reside and interact dynamically with its constituents during disease progression [15, 54]. Understanding the interplay between the different components within the tumor microenvironment is crucial for the success of therapeutic applications, since each component can be influenced by the others, resulting in impacts on the cancer cells [9, 10, 45, 52, 88, 106]. The presence of individual nerve fibers within the tumor microenvironment was ignored for many years as they are difficult to detect in classical histology. For a long time, only large nerve trunks were detected within tumors, and they were always associated with perineural invasion of cancer cells, a process in which these cells grow and migrate along native passive tissue nerves [84]. Recently, a different phenomenon was described, by which the tumor itself is infiltrated pro-actively by newly developed peripheral nerve projections [32, 36, 71, 89, 108, 115, 116, 121, 154, 158].

To understand how peripheral innervations behave within the tumors, functional studies, in which intra-tumoral nerves were eliminated, have relied on the surgical or pharmacological manipulation of nerves. Each such strategy, however, has its disadvantages. Peripheral nerves contain mixtures of different nerve fiber types [41, 82], and therefore, surgical denervation of a peripheral nerve leads to the disruption of all the nerve fibers present within that specific nerve [101]. Consequently, the role of particular nerve projections in the tumor cannot be isolated, as other nerve fibers are also affected. On the other hand, pharmacological drugs cause systemic reactions in several organs and indirect effects on unexpected targets. Thus, achieving the neuronal type-specificity that is needed to understand the role that specific nerve fibers perform in the tumor microenvironment is difficult with these methods, and the observed outcomes could be due to the unwanted effects on other innervations in addition to the targeted neurons. Wherefore, conclusions drawn from studies based on surgical or pharmacological denervation may be imprecise. These are some of the reasons, in addition to tumor tissue specificity, for some of the ambiguity about the roles of specific nerve fibers in cancer behavior. Accordingly, contradictory reports have been published: while some studies have claimed that certain neuronal types promote cancer progression [57, 158], others concluded that they suppress tumorigenesis [32, 116].

Therefore, to study the role of specific innervations, these should be directly manipulated in a nerve-fiber-type-specific manner. Recently, this approach became possible with the advent of powerful genetically-based tools, that precisely allow the targeting and elimination of specific peripheral nerve fibers for studying their functions in vivo [13]. Our group showed that specific genetic depletion of sensory neurons promotes melanoma growth [108]. Nevertheless, genetic ablation of these innervations may result in the generation of a pro-inflammatory microenvironment, secondary to cell death in the site where the neurons were ablated (Männ et al. 2016; Christiaansen, Boggiatto, and Varga 2014; Bennett et al. 2005), which itself is strongly tumor growth promoting [49, 67], and can affect cancer cells’ behavior [50]. Thus, it remains unclear which facets of the sensory neuron-ablated tumor phenotype are due to the loss of sensory innervations, rather than indirect effects due to the local inflammation caused by the death of these neurons. To circumvent this issue, in the present study, we used chemogenetics, an experimental strategy that has empowered neuroscience studies [131, 147], to determine the precise role of sensory neurons in the regulation of melanoma progression. Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) enable the silencing or overactivation of genetically defined neuronal populations upon binding to small-molecule designer drugs [119]. This approach allowed for highly selective and non-invasive modulation of sensory neurons’ activity in the tumor. Here, we revealed that silencing of sensory neurons’ activity, without ablating them, is sufficient to trigger increase in melanoma growth and in intra-tumoral new blood vessel formation. In contrast, chemogenetic stimulation of sensory neurons counteracted melanoma progression, by regulating tumoral growth, angiogenesis and immunosurveillance. Our results provide unequivocal evidence of the influence of sensory neurons in cancer progression.

Materials and methods

Animals

Generation of Nav1.8-Cre mice, in which Nav1.8 + sensory neurons express Cre recombinase, have been previously described. These animals were obtained from Infrafrontier (EMMA ID: 04 582). R26-LSL-hM4Di-DREADD (hM4Di) and CAG-LSL-hM3Dq-DREADD (hM3Dq) mice were purchased from the Jackson Laboratory (Jax) (Bar Harbor, ME).

To silence neuronal activity in sensory innervations in vivo, Nav1.8-Cre mice were crossed with R26-LSL-hM4Di-DREADD (hM4Di), a mouse line conditionally expressing a Gi-coupled engineered human muscarinic 4 receptor (hM4Di) [159]. hM4Di is a mutant G protein-coupled receptor which induces the canonical Gi pathway following binding to the pharmacologically inert drug clozapine-N-oxide (CNO). In Nav1.8-Cre + /hM4Di + mice, upon removal of the loxP-stop-loxP cassette by Cre recombination, the Gi-coupled hM4Di is expressed only in Nav1.8 + sensory neurons. Thus, sensory neuronal activity can be silenced by the administration of CNO. Nav1.8-Cre-/hM4Di + mice were used as controls.

To promote sensory neuron overactivation in vivo, Nav1.8-Cre mice were crossed with CAG-LSL-hM3Dq-DREADD (hM3Dq) animals, a mouse line conditionally expressing an evolved Gq protein-coupled receptor (hM3Dq), to generate Nav1.8-Cre + /hM3Dq + mice. In Nav1.8-Cre + /hM3Dq + animals, upon removal of loxP-stop-loxP cassette by Cre recombination, the Gq-coupled hM3Dq is expressed specifically in Nav1.8-sensory neurons. hM3Dq is a mutant G protein-coupled receptor which induces the canonical Gq pathway following the binding to CNO. Thus, sensory neuron firing can be chemically induced by administration of CNO. Nav1.8-Cre-/ hM3Dq + animals were used as controls.

All animal care and experimental procedures were approved by the Ethics Animal Care and Use Committee (CEUA), in accordance with the Guide for the Care and Use of Laboratory Animals from the Federal University of Minas Gerais. All colonies were housed in a pathogen-free animal facility of the Department of Pathology, UFMG, under controlled light cycle (12:12-h light/dark cycle) and fed ad libitum. Age-matched 8- to 12-week-old mice were used for all experiments. All experiments used mice heterozygous for both NaV1.8-Cre and DREADD receptors.

Cell culture

Murine B16-F10 melanoma is a common cell line that naturally originated in melanin-producing epithelia of C57BL6 mice. These cells were originally obtained from ATCC (USA), and were used to study melanoma development in vivo. The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) fetal calf serum/2 mM L-glutamine/100 U/ml penicillin/100 μg/ml streptomycin. The cells have been tested and found negative for mycoplasma. Cells were cultured in a humidified atmosphere of 95% air and 5% (v/v) CO2 at 37 °C.

Melanoma tumor implantation

B16-F10 cells were suspended in PBS and examined for viability using trypan blue staining. B16-F10 cells were used for transplantation only when more than 90% of cells were viable. For subcutaneous injection, the skin of all mice at an age of 8–12 weeks was shaved at the place of application. 1 × 105 cells in 100 μL were injected subcutaneously into the right flank of each mouse and the growth of the tumors was monitored until sacrifice. Growth of the tumors was assessed over time with a caliper as previously reported [12]. For determination of tumor volume, tumor-bearing animals were anesthetized with isoflurane in O2 by manually restraining the mice and placing their heads in in-house-built nose cones. Tumors were removed 16 days after transplantation and weighted. Length (L) and width (W) were calculated to measure tumor volume (V) using the formula V = 0.5 × (L × W2) [40]. Tumor area was determined using calibrated photographs of each tumor using Fiji software®, version 1.53 (National Institute of Health, Bethesda, MD).

CNO treatment

The DREADD ligand clozapine-N-oxide (CNO) (1 mg/kg in saline) (Sigma-Aldrich, St Louis, MO, USA) [7] was administered intra-peritoneally using a 25-gauge needle daily to test the effect of neuronal inhibition or activation on melanoma progression in Nav1.8-Cre + /hM4Di + and Nav1.8-Cre + /hM3Dq + animals, respectively. Control Nav1.8-Cre-/hM4Di + and Nav1.8-Cre-/hM3Dq + mice were similarly injected with CNO.

Capsaicin-induced spontaneous behavior

To confirm sensory neurons inhibition efficiency, following acclimation, Nav1.8-Cre-/hM4Di + and Nav1.8-Cre + /hM4Di + mice were injected with an intra-plantar subcutaneous dose of 10 μl of capsaicin (1 μg/10 μl; Sigma-Aldrich). A video recording was taken for 5 min post-capsaicin injection. The time that the animals spent performing spontaneous behaviors of licking, lifting, and flicking the paw were measured for 5 min after injection of capsaicin from these videos.

Immunohistochemistry and microscopy

Adult mice were deeply anesthetized with isoflurane and transcardially perfused with saline followed by 4% buffered paraformaldehyde (PFA, pH = 7.4). After dissection, B16F10 tumors were fixed overnight at 4 °C in 4% buffered paraformaldehyde, incubated overnight at 4 °C with 30% sucrose diluted in PBS, embedded and frozen in optimal cutting temperature compound (OCT, Tissue‐Tek). Embedded tumors were stored at − 80 °C. 20 μm cryosections were cut and blocked for 2 h in 3% BSA in PBS + 0.5% Triton and immunostained with the following antibodies: CD31‐PE (dilution 1:100) (BioLegend), Ki67 (dilution 1:100) (BD Biosciences), and anti‐Guinea pig‐AlexaFluor‐647 (1:1000) (Life Technologies) [10, 23]. After this, the sections were washed with PBS containing 4',6-diamidino-2-phenylindole (DAPI, 5 μg/ml, Invitrogen) and mounted using Dako fluorescence mounting medium (Dako, Santa Clara, CA). Stained tumor sections were imaged and analyzed by confocal microscopy using an inverted Zeiss LSM 880 confocal microscope (Oberkochen, Germany). CD31 area and the number of Ki67+ cells were quantified using Fiji software®, version 1.53 (National Institute of Health). Multiple random fields of each section were used for quantification.

Tumor-infiltrating leukocytes immunophenotyping and intracellular cytokine measurement

Tumors, their draining lymph nodes and non-tumor draining lymph nodes were harvested. Tissues were macerated and filtrated trough cell strainers of 40 um (Falcon) to isolate the cells used for immunophenotyping. Cells were washed in phosphate-buffered saline (PBS), incubated with Live/Dead solution (Invitrogen), for dead cell exclusion, and with monoclonal antibodies, washed, fixed, and permeabilized (FoxP3 staining buffer set, eBioscience) according to manufacture's instructions. Antibodies are listed in Table 1. Acquisition was realized on a LSR-FORTESSA. For analyses, to exclude debris, combinations of fluorochromes was done, to remove doublets a forward scatter area (FSC-A) versus forward scatter height (FSC-H) gate was used, and then cells were gated in function of time versus FSC-A to avoid a possible interference of flux interruptions. Only live leukocytes were used using a Live/Dead gate versus CD45. We assessed different immune cell subpopulations based on molecular markers of each cell subset: CD4 + T cells (CD4+/CD3+), CD8 + T cells (CD8+/CD3+), γδ T cells (CD3+/CD4−/CD8−/TCRγδ+), NKT cells (CD3+/NK+), regulatory T cells (Foxp3+/CD4+/CD3+), NK cells (CD3−/NK+), neutrophils (CD11b+/CD11c−/Ly6C−/LyG6+), PMN/MDSCs (CD11b+/Ly6C−/LyG6+) and dendritic cells (CD11b−/CD11c+/Ly6C−/LyG6−). In each T-cell subset, frequencies of cells expressing checkpoint inhibitors CTLA-4 and PD1 were evaluated. Cytokine analyses in lymphocytes from the tumor microenvironment and lymph nodes were done using intracellular staining. Briefly, cells were isolated from tumor samples and lymph nodes and cultivated for 4 h at 37 °C in 10% FBS RPMI supplemented with 2 mM L-glutamine, 50 units/mL penicillin, and 50 μg/mL streptomycin, in the presence of Brefeldin A (ThermoFisher) and Monensin (ThermoFisher). Following this, cells were washed in FACS buffer and stained for cell surface markers. Cells were then fixed for 35 min at 4 °C with eBiosciences Cytofix/Cytoperm buffer and, subsequently, washed once in eBioscience Perm/Wash buffer. Then, cells were stained for 45 min at 4 °C with anti-IFN-γ and anti-IL-17 (Table 1) antibodies diluted in eBioscience Perm/Wash [78]. Cells were washed twice and the data was acquired. Ki67 is a nuclear factor transcript in the late G1, S, G2, and M of cell cycle, therefore marks proliferating cells [44, 128]. Thus, we evaluated proliferation in viable CD45 negative cells, suggesting tumoral proliferation. GraphPad Prism V7.0 (GraphPad software) and FlowJo V10.4.11 (TreeStar) were used for data analysis and graphic presentation.

Table 1 Antibodies used in flow cytometry
Full size table

Quantification of CGRP within tumors

Tumor samples from Nav1.8-Cre + /hM4Di + and Nav1.8-Cre + /hM3Dq + animals, as well as from their respective controls (Nav1.8-Cre-/hM4Di + and Nav1.8-Cre-/hM3Dq +) were analyzed to measure the amount of CGRP using commercially available Sandwich-CGRP ELISA kit purchased from Elabscience (Catalog # E-EL-M2744). Briefly, tumor pieces were weighed and then homogenized in PBS (0.01 M, pH = 7.4) with a glass homogenizer on ice. The homogenates were centrifuged for 5 min at 5000 × g at 4 ℃ to get the supernatant. ELISA of CGRP were performed according to manufacturer’s instructions. After ELISA, Optical Density (OD) was measured using Varioskan Flash (Thermo) set at 450 nm.

In silico analysis

RNA sequencing count data of 103 Skin Cutaneous Melanoma (SKCM) patients was downloaded from The Cancer Genome Atlas (TCGA) repository (https://portal.gdc.cancer.gov/). Differential gene expression analyses were performed between samples of alive and dead patients (considering a 5-year interval) using DESeq2 [83]. We stratified patients in these two groups, alive or dead, based on their vital status in a 5-year interval of their tumor diagnosis (clinical data available at TCGA and curated by Liu et al. (2018) [85]. Genes with absolute log2(Fold-change) ≥ 1 and False Discovery Rate (FDR) adjusted P-value < 0.05 were considered differentially expressed. To identify biological processes associated with genes differentially expressed, we performed a Gene Ontology (GO) enrichment analysis using ShinyGO [42]. We used the STRING database [132] (parameters: full STRING network, considering only text-mining, databases and experiments interactions with score > 0.400, and only genes with 3 or more interactions) and Cytoscape (https://cytoscape.org/) to construct protein–protein interactions (PPIs) among our manually curated list of 34 gene related to sensory neurons selected based on the literature [29, 37, 51, 114, 141, 144]. The set of 18 genes showing at least two PPI interactions are shown. For the remaining analyses, RNA sequencing counts were first Transcripts Per Million (TPM)-normalized using a local R script. To identify a gene signature associated with SKCM cancer patient survival we used Reboot [31] with parameters "-B 100 -G 5 -P 0.3 -V 0.01". Briefly, Reboot finds genes associated with cancer patient prognosis using multivariate penalized Cox regression combined with a bootstrap approach. In the first step of Reboot, it produces regression coefficients (numerical values) that determine the contribution of each submitted gene to patients' survival. These coefficients may be positive or negative values indicating that high expression of a particular gene potentially contributes to worse or better prognosis, respectively. Once these coefficients are produced, Reboot then calculates the score of each patient (sample) as the sum of each gene coefficient multiplied by the corresponding gene expression level in that patient. Finally, when all patients' scores are calculated, we then stratify them into groups with high/low scores based on the median score of all patients to create the survival curve (Kaplan–Meier). For further information, see [31]. SCN10A box plots and survival curves were created using R (https://www.r-project.org/) scripts.

Statistical analysis

Graphs were plotted using GraphPad Prism 7 (San Diego, CA). Shapiro‐Wilk normality test was performed, and unpaired t test was used to determine statistical significance.

Results

Chemogenetic inhibition of Nav1.8 + neurons accelerates melanoma progression

We have previously demonstrated that melanomas are infiltrated by Nav1.8 + sensory innervations, and that those tumors grow slower when these neurons are pharmacologically or genetically ablated [108]. However, these investigations were performed using Nav1.8-Cre + /DTA + mice, in which a diphtheria toxin fragment A is constitutively activated in Nav1.8 + sensory neurons, resulting in the toxin induced-death of these cells. Therefore, this technique lacks temporal control of neuronal ablation, and enables compensatory effects during the development of these animals. Importantly, the approach by which specific neurons are ablated from the tissue microenvironment is also limited because of the secondary consequences, such as inflammation, caused in the tissue where sensory neurons are eliminated. Thus, it remains unclear whether these damage-induced changes in the tissue may influence the observed cancer outcomes. Here, we applied a chemogenetic approach to specifically inhibit the activity of Nav1.8-expressing sensory neurons without killing these cells. We used DREADDs to specifically control sensory neuron activity. DREADDs are derived from different types of mutated muscarinic receptors that have been engineered to lose affinity to their endogenous ligand acetylcholine [5], but to gain responsiveness to a synthetic ligand, clozapine-N-oxide (CNO). Inhibitory DREADDs (hM4Di) elicit an intracellular cascade that results in the silencing of neuronal activity [113], without changing the number of innervations as previously reported [68]. This method allows for the selective silencing of specific types of neurons in vivo without physical manipulation or destruction in the tissue. DREADDs were expressed specifically in sensory neurons, using a transgenic murine approach: mice carrying the construct encoding for Cre-dependent expression of hM4Di were crossed to Nav1.8-Cre animals. In the resulting mice, Nav1.8-Cre + /hM4Di + , only Nav1.8 + sensory neurons expressed inhibitory DREADDs. As controls in this study, we used littermate mice carrying Cre-dependent hM4Di, but lacking the Cre gene (Nav1.8-Cre-/hM4Di +) (Fig. 1A). This allowed us to control for any potential side effects from CNO administration.

Fig. 1
figure 1

Chemogenetic inhibition of neuronal activity in sensory Nav1.8 + nerve fibers triggers melanoma growth. A Schematic diagram of the Nav1.8-Cre + /hM4Di + experimental mouse model. Cre recombinase directs the expression of hM4Di specifically to sensory neurons in those mice. After the administration of CNO to those mice, neuronal activity in sensory neurons is inhibited. B Tumor-infiltrating sensory neurons are targeted in Nav1.8-Cre mice. 1 × 105 B16F10 melanoma cells were subcutaneously injected into Nav1.8-Cre/TdTomato mice, and tumor tissues were surgically removed 16 days later. Representative image of a Nav1.8-Cre/TdTomato mouse tumoral section with sensory nerve fibers infiltrating the tumor labelled with TdTomato fluorescence (red) and nuclei with DAPI (blue). C Capsaicin-induced spontaneous behavior test corroborates chemogenetic inhibition of sensory Nav1.8 + nerve fibers in Nav1.8‐Cre+/hM4Di+ mice after CNO treatment. Column charts show the licking time after capsaicin application of Nav1.8‐Cre−/hM4Di+ (n = 5) and Nav1.8‐Cre+/hM4Di+ (n = 5) animals. D Representation of the protocol for subcutaneous allograft melanoma growth. 1 × 105 B16F10 melanoma cells were subcutaneously injected into Nav1.8‐Cre−/hM4Di + (n = 5) and Nav1.8‐Cre + /hM4Di + (n = 5) mice, followed by tumors removal for analysis after 16 days. CNO was daily intra-peritoneal injected at 1 mg/kg. E Development curve of tumor growth from Nav1.8‐Cre−/hM3Dq+ and Nav1.8‐Cre+/hM3Dq+. Tumor volumes were assessed over time with a caliper. F Representative macroscopic image of B16F10 melanoma after dissection, left panel (Nav1.8‐Cre−/hM4Di+) and right panel (Nav1.8‐Cre+/hM4Di+). G Tumor weight. (Nav1.8‐Cre−/hM4Di+: 0.50 ± 0.04 g; Nav1.8‐Cre+/hM4Di+: 0.98 ± 0.23 g). Data are shown as mean ± SEM. Unpaired t test (ns P > 0.05; *P < 0.05; **P < 0.01)

Full size image

In order to ascertain that the expression of DREADD receptors was driven to intra-tumoral Nav1.8-expressing neurons in Nav1.8-Cre + /hM3Di + mice, we used the solid tumor model B16F10. We assessed tumor sections from melanoma grown in Nav1.8-Cre + /TdTomato + mice and detected Nav1.8 + neurons expressing TdTomato present within the tumor microenvironment (Fig. 1B). To validate sensory neuronal inhibition following daily CNO injection, we used a behavioral test to evaluate the sensitivity to capsaicin, confirming the silencing of sensory neurons, as previously described [2]. Indeed, Nav1.8-Cre + /hM4Di + animals spent less time (25.53 ± 2.27 s) licking their paws after intra‐plantar injection of capsaicin, compared to control animals (56.53 ± 3.92 s) (Fig. 1C). To analyze the effect of sensory neurons silencing on tumor growth, we subcutaneously transplanted B16F10 cells to the lower right flank of both inhibitory DREADD-expressing mice (Nav1.8-Cre + /hM4Di +) and their controls (Nav1.8-Cre-/hM4Di +). Following cancer cell injection, we treated the animals daily with CNO to induce sensory neuronal activity inhibition (controls were also treated with CNO) (Fig. 1D). After 14 days of continuous sensory inhibition, tumor volume was significantly enhanced in the sensory neuron-silenced mice when compared to the controls (tumor volume was increased from 82.1 ± 29.6 to 319.6 ± 72.8 mm3; Fig. 1E). After 16 days of repeated sensory inhibition, tumor weight was also significantly enhanced in the sensory neuron-silenced mice when compared to the controls (tumor weight was increased from 0.50 ± 0.04 to 0.98 ± 0.23 g; Fig. 1F, G). Animal weights were not affected by sensory inhibition (data not shown).

Increase in neoangiogenesis within melanoma tumors is correlated with worse outcomes in these patients [111]. We detected, in melanoma-bearing animals with silenced sensory neurons, an enhancement in the intra‐tumoral blood vessels’ area (from 0.02 ± 0.00 to 0.03 ± 0.01 µm2) (Fig. 2A, B). Expression of Ki67 is used to determine the proliferation rate of malignant cancer cells [139], which is also associated with melanoma aggressiveness [76]. We found that genetic silencing of sensory innervations led to an increase in the proliferation rate within the melanoma (from 2074 ± 55.32 to 2454 ± 168.4 Ki67 + cells per μm2) (Fig. 2C, D). We also observed after inhibition of sensory neurons firing a decrease in tumor-infiltrating CD4 + T cells (from 4.47 × 107 ± 1.15 × 107 to 1.73 × 107 ± 7.92 × 106 cells per mg of tumor) (Fig. 2E), in special, in IL-17-producing CD4 + T cells (from 1.63 × 107 ± 1.30 × 106 to 3.77 × 106 ± 3.27 × 106 cells per mg of tumor) (Fig. 2F), and a decrease in melanoma-infiltrating CD8 + T cells (from 3.27 × 106 ± 5.22 × 105 to 7.62 × 105 ± 6.78 × 105 cells per mg of tumor) (Fig. 2G). Our results indicate that inhibition of neuronal activity in sensory neurons promotes melanoma tumor advancement.

Fig. 2
figure 2

Chemogenetic inhibition of neuronal activity in sensory Nav1.8 + innervations increases intra-tumoral proliferation and angiogenesis, and blocks anti-tumoral immune response. 1 × 105 B16F10 melanoma cells were subcutaneously injected into Nav1.8‐Cre−/hM4Di + (n = 5) and Nav1.8‐Cre + /hM4Di + (n = 5) mice, followed by tumors removal for analysis after 16 days. A Representative immunofluorescence images of tumors labelled for endothelial cells (CD31; red) to identify blood vessels and nuclei (DAPI; blue). B Quantification of angiogenesis in melanomas by blood vessel area. C Representative immunofluorescence images of tumors labelled for Ki67 (Ki67; green) to identify cell proliferation and nuclei (DAPI; blue). D Quantification of proliferation in melanomas by the counting of Ki67 + cells per μm2. Absolute number of CD4 + E and CD8 + G T cells from the melanomas of B16F10–inoculated mice. F Graph shows absolute numbers of CD4 + T cells producers of IL-17. IL-17 levels were measured in cells isolated from tumors of B16F10–inoculated Nav1.8-Cre−/hM4Di+ and Nav1.8-Cre+/hM4Di+ animals. Data are shown as mean ± SEM. Unpaired t test (ns P > 0.05; *P < 0.05)

Full size image

Chemogenetic activation of hM3Dq excitatory DREADD receptors in Nav1.8 + neurons promotes melanoma regression

As we found that inhibition of sensory neuron activity promotes melanoma advancement, we hypothesized that increasing sensory excitability would result in the reverse: blockage of melanoma progression. To test this hypothesis, we used again chemogenetics, by which we induced the expression of excitatory hM3Dq DREADDs [127] only in Nav1.8 + sensory neurons. We crossed Nav1.8-Cre mice to a mouse line with a Cre-dependent evolved Gq protein-coupled receptor (hM3Dq) expression. In the resulting Nav1.8-Cre + /hM3Dq + mice, upon removal of loxP-stop-loxP cassette by Cre recombination, the Gq-coupled hM3Dq is expressed specifically in Nav1.8-sensory nerve fibers. Sensory neurons in those mice can thus be overactivated by the administration of CNO. It has been shown previously that Gq-DREADD activation by CNO increases neuronal activity in the targeted neurons, including sensory neurons [68, 90], without changing the number of neurons [110]. To evaluate the role of sensory stimulation on tumor growth, we transplanted subcutaneously B16F10 melanoma cells to the lower right flank of both stimulatory DREADD-expressing mice (Nav1.8-Cre + /hM3Dq +) and their controls (Nav1.8-Cre-/hM3Dq +). Following the cancer cell implantation, we treated mice daily with CNO to induce Nav1.8 + sensory neuron activation (controls were also treated with CNO) (Fig. 3A, B). After repeated sensory neuron activation, melanoma development was decreased in the sensory neuron-overactivated mice when compared to the controls (at day 14, tumor volume per body weight was reduced from 3.51 ± 0.89 to 0.71 ± 0.20 mm3; at day 16, tumor weight was reduced from 0.38 ± 0.07 to 0.17 ± 0.03 g; Fig. 3C–F). Animal weights were not affected by genetic stimulation of sensory neurons in melanoma‐bearing mice (data not shown). Moreover, genetic overactivation of sensory neurons led to a decrease in proliferating cells within the tumor (from 3050 ± 203 to 1292 ± 367 Ki67 + cells per μm2, analyzed by immunohistochemistry) (Fig. 3G, H), corroborated by flow cytometry analysis of CD45- cells for Ki67 expression (the was a decrease from 8.13 ± 1.00 to 5.07 ± 0.70% of CD45-/Ki67 + cells within the population of CD45- cells) (Fig. 3I). Additionally, there was a decrease in the intra‐tumoral blood vessels’ area (from 0.010 ± 0.001 to 0.006 ± 0.001 µm2 of CD31 + area / µm2 of tumor area) (Fig. 3J, K). Our data suggest that increase in neuronal activity in sensory neurons counteracts melanoma development.

Fig. 3
figure 3

Overstimulation of sensory Nav1.8 + nerve fibers decreases melanoma growth. A Schematic diagram of the Nav1.8-Cre + /hM3Dq + experimental mouse model. Cre recombinase directs the expression of hM3Dq specifically to sensory neurons in those mice. After the administration of CNO to those mice, neuronal activity in sensory neurons is overactivated. B Representation of the protocol for subcutaneous allograft melanoma growth. 1 × 105 B16F10 melanoma cells were subcutaneously injected into Nav1.8‐Cre−/hM3Dq + (n = 14) and Nav1.8‐Cre + /hM3Dq + (n = 13) mice, and tumors were removed for analysis after 16 days. CNO was injected daily intra-peritoneally at 1 mg/kg. C Development curve of tumor growth from Nav1.8‐Cre−/hM3Dq+ and Nav1.8‐Cre+/hM3Dq+. Tumor volumes were assessed over time with a caliper. D Representative macroscopic images of B16F10 melanoma tumors after dissection, left panel (Nav1.8‐Cre−/hM3Dq+) and right panel (Nav1.8‐Cre+/hM3Dq+). E Tumor weight. (Nav1.8‐Cre−/hM3Dq+: 0.38 ± 0.07; Nav1.8‐Cre+/hM3Dq+: 0.17 ± 0.03). F Tumor weight corrected by animal body weight. G Representative immunofluorescence images of tumors labelled for Ki67 (Ki67; green) to identify cell proliferation and nuclei (DAPI; blue). H Quantification of proliferation in melanomas from Nav1.8‐Cre−/hM3Dq+ and Nav1.8‐Cre+/hM3Dq+ animals. I Quantification of proliferation (Ki67 +) by flow cytometry in CD45- cells from tumors of Nav1.8‐Cre−/hM3Dq+ and Nav1.8‐Cre+/hM3Dq+ mice. J Representative immunofluorescence images of tumor sections labelled for endothelial cells (CD31; red) to identify blood vessels and nuclei (DAPI; blue). K Quantification of angiogenesis in melanomas by blood vessel area. Data are shown as mean ± SEM. Unpaired t test (ns P > 0.05; *P < 0.05; **P < 0.01)

Full size image

Increase in sensory neuron activty affects melanoma immunosurveillance

Functional studies in combination with histological analysis have demonstrated that tumor-infiltrating immune cells modulate melanoma cells’ behavior, altering cancer outcomes [38, 72, 79, 83, 112, 130, 133, 134, 152, 153]. Given that sensory neurons may influence immune responses in non-cancer contexts, we sought to probe whether sensory neurons stimulation alters immune surveillance within the tumor.

Accumulating evidence has demonstrated that tumor-infiltrating neutrophils and PMN-MDSCs promote tumor development and progression [21, 39, 65, 107, 136, 138, 140, 150]. Thus, we evaluated whether these cells are affected by sensory neurons’ overactivation. We found that the number of melanoma-infiltrating neutrophils and PMN-MDSCs was significantly decreased in the sensory neuron-overactivated mice (Nav1.8-Cre + /hM3Dq +) when compared to the controls (Nav1.8-Cre-/hM3Dq +) (from 12.02 × 107 ± 3.45 × 107 to 4.69 × 107 ± 7.10 × 106 PMN-MDSCs per mg of tumor; and from 10.45 × 107 ± 3.70 × 107 to 2.78 × 107 ± 5.65 × 106 neutrophils per mg of tumor) (Fig. 4A, B). On the other hand, we found that the number of tumor-infiltrating dendritic cells, which counteract the proliferation of melanoma cells [137], was significantly increased (from 5.53 × 107 ± 8.80 × 106 to 1.07 × 108 ± 2.27 × 107 dendritic cells per mg of tumor) (Fig. 4C).

Fig. 4
figure 4

Sensory neurons overactivation improves anti-tumor immunity by decreasing tumor-infiltrating immunosuppressive cells, increasing dendritic cells and by promoting CD4 + T, CD8 + T, γδT, NKT, and NK-cell infiltration. Immune cells from B16F10–inoculated mice were analyzed ex vivo in Nav1.8-Cre−/hM3Dq+ (n = 14) and Nav1.8-Cre+/hM3Dq+ (n = 13) mice. Column charts show the proportion of PMN/MDSC (A) Neutrophils (B) and Dendritic cells (C) quantified in the tumor microenvironment. (D-I) TIL from B16F10–inoculated Nav1.8-Cre-/hM3Dq + (n = 14) and Nav1.8-Cre + /hM3Dq + (n = 13) mice were analyzed ex vivo. Absolute number of CD4 + T cells (D), CD8 + T cells (E), γδ T cells (F), NKT cells (G), NK cells (H), and Treg cells (I) from the melanomas of B16F10–inoculated mice. Data are shown as mean ± SEM, Unpaired t test, *.01 < P < .05; **.001 < P < .01

Full size image

Recent breakthroughs in cancer immunotherapy have revealed the remarkable ability of the immune system to fight different types of cancers, including melanoma. The phenotypes and numbers of prevalent tumor-infiltrating lymphocytes are predictive of response to immunotherapy and key modulators of disease progression. Thus, we examined how tumor-infiltrating lymphocytes are affected by sensory neurons’ overstimulation. We detected an increase in tumor-infiltrating CD4 + T cells (from 2.91 × 106 ± 1.04 × 106 to 1.09 × 107 ± 2.92 × 106 cells per mg of tumor), CD8 + T cells (from 8.94 × 106 ± 1.60 × 106 to 1.96 × 107 ± 5.20 × 106 cells per mg of tumor), γδ T cells (from 31.76 × 107 ± 7.32 × 107 to 74.14 × 107 ± 20.40 × 107 cells per mg of tumor), NKT cells (from 16.34 × 107 ± 4.6 × 107 to 34.92 × 107 ± 10.42 × 107 cells per mg of tumor) and NK cells (from 1.72 × 107 ± 2.90 × 106 to 3.47 × 107 ± 4.40 × 106 cells per mg of tumor) (Fig. 4D–H), while regulatory T cells, which mediate immunosuppression in the tumor microenvironment [66], were not altered (Fig. 4I). Immune checkpoint molecules, such as cytotoxic T lymphocyte antigen 4 (CTLA-4) and programmed cell death 1 (PD-1), act fine-tuning the intense immune responses that might kill healthy cells [27, 55, 122]. Their expression in cytotoxic T cells may lead to dysfunction of these cells, affecting their effector function [11, 146]. We found that increase in the firing of sensory neurons prevented the increase of immune checkpoint markers of tumor infiltrating CD8 + T cells and CD4 + T cells (Fig. 5 and Additional file 1: Figure 1). The percentage of CTLA-4-expressing CD4 + tumor-infiltrating lymphocytes decreased from 29.43 ± 4.04% in Nav1.8-Cre−/hM3Dq+ to 19.08 ± 2.80% in Nav1.8-Cre+/hM3Dq+ animals (Fig. 5A); similarly, the percentage of PD-1-expressing CD4 + tumor-infiltrating lymphocytes decreased from 15.02 ± 2.62% in Nav1.8-Cre−/hM3Dq+ to 7.85 ± 1.43% in Nav1.8-Cre+/hM3Dq+ mice (Fig. 5B, C). The percentage of PD-1-expressing CD8 + tumor-infiltrating cytotoxic lymphocytes also decreased from 22,03 ± 2,66% in Nav1.8-Cre−/hM3Dq+ to 12.99 ± 3.85% in Nav1.8-Cre+/hM3Dq+ animals (Fig. 5E), while the expression of CTLA-4 did not vary in these cells (Fig. 5D, F). In addition, no differences were found in CTLA-4 and PD-1 expression on γδ T cells (Fig. 5G, H and I), NKT cells (Fig. 5J, K and L), NK cells (Fig. 5M, N and O) and Treg cells (Fig. 5P, Q and R). Overall, our data suggest that sensory neurons overactivation induces improvement of T cells effector functions within the tumor microenvironment.

Fig. 5
figure 5

Sensory neurons overstimulation prevent the increase of immune checkpoint markers in tumor infiltrating CD8 + T cells and CD4 + T cells. Immune cells from tumors of B16F10–inoculated mice were analyzed ex vivo in Nav1.8-Cre−/hM3Dq+ (n = 14) and Nav1.8-Cre+/hM3Dq+ (n = 13) mice. Column charts show proportion of CTLA-4 (A, D, G, J, M, P), PD-1 (B, E, H, K, N, Q) and CTLA-4/PD-1 co-expressing (C, F, I, L, O, R) CD4 + T cells (A, B, C), CD8 + T cells (D, E, F), γδ T cells (G, H, I), NKT cells (J, K, L), NK cells (M, N, O), and Treg cells (P, Q, R) from tumors of B16F10–inoculated mice. Data are shown as mean ± SEM, Unpaired t test, *.01 < P < .05; **.001 < P < .01

Full size image

It has been reported that CD4 + and CD8 + lymphocytes secreting IL-17 promote melanoma regression [91, 99]. Here, we detected in response to sensory neuron firing an increase in melanoma-infiltrating IL-17-producing CD4 + T cells (from 2.45 × 107 ± 6.05 × 106 to 30.78 × 107 ± 9.20 × 107 cells per mg of tumor) as well as in melanoma-infiltrating IL-17-producing CD8 + T cells (from 5.02 × 107 ± 0.90 × 107 to 20.08 × 107 ± 5.92 × 107 cells per mg of tumor) (Fig. 6A). In parallel, we did not detect significant changes in the number of other tumor-infiltrating lymphocytes producing IL-17 or in IFN-γ-producing lymphocytes after sensory neuron overactivation (Fig. 6B). Altogether, our results suggest that sensory neurons induce a Th17-immune response in the melanoma microenvironment.

Fig. 6
figure 6

Sensory neurons overactivation promote an increase in tumor-infiltrating IL-17-producing CD4 + and CD8 + T cells. TIL from melanomas of B16F10–inoculated Nav1.8-Cre-/hM3Dq + (n = 14) and Nav1.8-Cre + /hM3Dq + (n = 13) mice were analyzed. TIL from B16F10–inoculated mice were analyzed after 4 h of culture. A Column charts show absolute numbers of CD4 + and CD8 + T cells producers of IFN-γ and IL-17. B Column charts show absolute number of γδ T cells, NKT cells and NK cells producing IFN-γ and IL-17. Cytokines levels were measured in cells isolated from tumors of B16F10–inoculated Nav1.8-Cre−/hM3Dq+ and Nav1.8-Cre+/hM3Dq+ mice. Data are shown as mean ± SEM, Unpaired t test, *.01 < P < .05; **.001 < P < .01

Full size image

Lymph nodes are an integral part of the adaptive immune system in our organism and are essential for the effective immune responses. Melanoma draining lymph nodes are influenced by the primary tumor, but may also prime the immune suppressive microenvironment, playing critical roles in promoting cancer immune escape [33, 86, 87, 97]. It is completely unknown whether sensory neuron overactivation may affect the immune cells also within the tumor draining lymph nodes. Herein, we analyzed immune cells from tumor draining and non-tumor-draining lymph nodes from CNO-treated stimulatory DREADD-expressing animals (Nav1.8-Cre + /hM3Dq +) and their controls (Nav1.8-Cre-/hM3Dq +). Tumor draining and non-tumor-draining lymph nodes were isolated from the ipsilateral and contralateral side, respectively, of the implanted melanoma (Fig. 7A, B). We found that the effect of sensory neurons stimulation in tumor draining lymph nodes mimics the immune response within the primary tumor, but not in non-tumor-draining lymph nodes. In the tumor-draining lymph nodes, we found an increase in the number of CD8 + cytotoxic T cells after sensory neurons’ overstimulation (from 13.68 × 107 ± 3.50 × 107 to 22.41 × 107 ± 2.17 × 107 cells per mg of tumor) (Fig. 7C); while we did not detect any differences in the numbers of T cells in the tumor non-draining lymph nodes (Fig. 7D). These data indicate a possible priming effect of tumor on the adjacent draining lymph nodes.

Fig. 7
figure 7

Tumor-draining lymph nodes present an increase in effector CD8 + T-cells after overstimulation of sensory neurons, while the number of lymphocytes in tumor non-draining lymph nodes doesnt change. A Schematic representation of the collected tumor draining lymph nodes. TIL from tumor-draining lymph nodes of B16F10–inoculated Nav1.8-Cre-/hM3Dq + (n = 14) and Nav1.8-Cre + /hM3Dq + (n = 13) mice were analyzed. B Absolute number of CD4 + , CD8 + , γδ, NKT and NK cells from tumor-draining lymph nodes of B16F10–inoculated mice. C IFN-γ and IL-17 were quantified in CD4, CD8, γδ, NKT and NK cells. D Schematic representation of the collected tumor non-draining lymph nodes. TIL from tumor non-draining lymph nodes of B16F10–inoculated Nav1.8-Cre-/hM3Dq + (n = 14) and Nav1.8-Cre + /hM3Dq + (n = 13) mice were analyzed. E Absolute number of CD4 + , CD8 + , γδ, NKT and NK cells from tumor non-draining lymph nodes of B16F10–inoculated mice. F IFN-γ and IL-17 were quantified in CD4, CD8, γδ, NKT and NK cells. Cytokines levels were measured in cells from tumor-draining and tumor non-draining lymph nodes of B16F10–inoculated Nav1.8-Cre-/hM3Dq + and Nav1.8-Cre + /hM3Dq + mice. Data are shown as mean ± SEM. Unpaired t test, *.01 < P < .05; **.001 < P < .01; ****P < 0.0001

Full size image

We also observed, in the tumor-draining lymph nodes, increases in IFN-γ-producing CD4 + T cells (from 8.31 × 106 ± 2.65 × 106 to 1.94 × 107 ± 3.36 × 106 cells per lymph node), in IFN-γ-producing NK cells (from 4.72 × 104 ± 2.46 × 104 to 8.44 × 105 ± 2.22 × 105 cells per lymph node), in IL-17-producing CD8 + T cells (from 2.38 × 106 ± 4.27 × 105 to 1.28 × 107 ± 1.64 × 106 cells per lymph node), and in IL-17-producing NKT cells (from 2.35 × 106 ± 5.19 × 105 to 4.08 × 106 ± 5.00 × 105 cells per lymph node) after sensory neurons CNO-stimulation (Fig. 7E). In tumor non-draining lymph nodes, we detected increases in IFN-γ-producing CD4 + T cells (from 5.73 × 106 ± 2.17 × 106 to 1.92 × 107 ± 4.23 × 106 cells per lymph node), in IFN-γ-producing CD8 + T cells (from 1.36 × 107 ± 5.57 × 106 to 4.77 × 107 ± 1.19 × 107 cells per lymph node), in IFN-γ-producing γδ T cells (from 2.27 × 106 ± 8.55 × 105 to 1.03 × 107 ± 3.22 × 106 cells per lymph node), and in IFN-γ-producing NK cells (from 1.80 × 105 ± 4.70 × 104 to 6.82 × 105 ± 1.53 × 105 cells per lymph node) (Fig. 7F). Our data suggest that lymphocytes at the lymph nodes may be contributing to the response against the melanoma observed within the tumor microenvironment after sensory neurons’ overactivation as both, IFN-γ and IL-17, may contribute to enhance the anti-tumoral response in the context of melanoma [92, 109, 151]. Altogether, our data suggest that sensory neurons stimulation alters immune surveillance that impacts melanoma development.

High expression of genes related to sensory neurons correlates with best prognosis in human melanoma patients

In order to investigate our findings also in human tumors, we analyzed The Cancer Genome Atlas (TCGA) samples from 103 Skin Cutaneous Melanoma (SKCM) patients. First, we stratified SKCM patients in two groups, alive or dead based on a 5-year interval. Next, we searched for differentially expressed genes between these groups. We found 112 up-regulated and 195 down-regulated genes (|log2(Fold change)|≥ 1 and adjusted P-value < 0.05; Table 2). Next, we performed a Gene Ontology (GO) analysis of Biological Processes (BP) in which these genes are involved. The up-regulated genes set are enriched for three biological processes, while the down-regulated genes are enriched to a wide variety of processes (Fig. 8A; Table 3). Curiously, for the former (up-regulated genes), two out of three Biological Processes represent nervous system development (Fig. 8A), indicating the importance of neuronal networks in melanoma outcomes. Next, we investigated the interactions (and putative regulation) of 34 sensory neurons-related genes which were selected from the literature [29, 37, 51, 114, 141, 144] (Table 4). Figure 8B shows a strong connection among 18 of these genes, suggesting that they work on the same cellular pathways or cell types. Next, using the expression levels of these 34 gene markers for sensory neurons we investigated their potential to be "a signature" associated with SKCM cancer patient survival. Figure 8C shows that high expression of these genes (lower patient scores) are associated with a better overall survival of SKCM patients. Next, we investigated the expression of these genes in the two sample sets (alive and dead patients). We found that SCN10A, which codifies Nav1.8, a key gene based on which our mouse models target sensory neurons, is more expressed in alive than in dead patients (Fig. 8D). Finally, we investigated the impact of SCN10A expression on SKCM patients' survival. Figure 8E shows that high expression of SCN10A trends to be associated with a better overall survival of SKCM patients, even without statistical support (P-value = 0.26). Taken together, these results confirm that genes related to nervous system development are enriched in samples from live Skin Cutaneous Melanoma patients. Focusing on gene markers for sensory neurons, we confirmed that these genes are strongly connected, suggesting a synergistic activity, and that the higher expression of some of these genes are associated with a higher overall survival. Strikingly, the high expression of SCN10A is potentially associated with better SKCM patient survival, indicating that the presence of sensory neurons within melanoma counteracts cancer progression. We also found that TCGA samples from tumors with a worst prognosis (dead patients) have an enrichment of genes promoting angiogenesis (Tables 3 and 5; 15 genes related to angiogenesis). We focused on this set of 15 genes related to angiogenesis and we confirmed that they are strongly connected (Additional file 2: Fig. 2), indicating a synergic function in promoting angiogenesis. Our results indicate that there is an increase in genes related with angiogenesis in tumors with worst prognosis (from dead patients) and a decrease in their expression in tumors from alive patients which show an increased expression of SCN10a, a sensory neuron marker used in this study (Fig. 8C, D, E). By using the CIBERSORT tool [102], we investigated immune infiltrated cells in the same TCGA cohort of alive vs. dead patients (Additional file 3: Fig. 3). CIBERSORT uses gene expression (RNA sequencing data) and support vector regression combined with prior knowledge of expression profiles from purified leukocyte subsets (gene signatures) to produce an estimation of the abundances of immune infiltrated cells subpopulations in a sample. In line with our data presented in this manuscript, we have checked the enrichment of immune infiltrated cells in the tumors of patients alive vs. dead (Additional file 4: Fig. 4). We found an increase of CD4 + T cells, CD8 + T cells, NK cells and dendritic cells in patients with better prognosis (alive). Thus, tumors showing a better prognosis (alive) have an increased infiltration of some key immune cells. Additionally, microarray data evidenced a down-regulation of genes related to the Th17 immune response in melanoma patients (Additional file 5: Fig. 5). These analyses are consistent with the data generated in our mouse models: that the overactivation of sensory innervations in the tumor microenvironment was associated with suppressed melanoma progression. Albeit gene expression in tumor biopsies from human cancer patients is used as a tool to define novel biomarkers and  to contribute to prognosis, the obtained data should be also validated by the quantification of sensory neuron-related proteins in human melanoma biopsies and correlation with clinical outcomes in future research.

Table 2 Analyzes of genes from The Cancer Genome Atlas (TCGA) samples from 103 Skin Cutaneous Melanoma (SKCM) patients
Full size table

 

Fig. 8
figure 8

available at TCGA and curated by Liu et al. (2018) [85]. We stratified patients in two groups, alive or dead, based on a 5-year interval of their tumor diagnosis. B Interactions among genes related to sensory neurons. C Gene signature using sensory neurons-related genes. High expression of these genes (lower patient scores) is associated with a better overall survival of SKCM patients (patients were stratified based on their median Reboot score). Overall survival of patients showing expression of sensory neurons-related genes. More negative scores are associated with higher gene expression. D Expression of SNC10A (Nav1.8) in SKCM samples. Only patients presenting tumors expressing SNC10A were used. E We evaluated the survival probability of patients with melanoma based on their tumor transcriptome. Patients were stratified based on lower/upper quartiles of SCN10A expression values. Overall survival of SKCM patients based on the level of expression of SCN10A. High expression of SCN10A (Nav1.8) correlates with best outcomes in patients with melanoma

Overexpression of genes related to sensory nerves is associated with Skin Cutaneous Melanomas (SKCM) patients improved survival. A Biological Processes of genes overexpressed in Skin Cutaneous Melanomas (SKCM) samples from alive patients versus dead patients. Patients were stratified (alive or dead) based on their vital status in a 5-year interval of their tumor diagnosis (clinical data

Full size image
Table 3 Gene Ontology (GO) and analysis of Biological Processes (BP) in which specific genes are involved
Full size table
Table 4 Sensory neuron markers
Full size table
Table 5 Angiogenesis-related genes down-regulated in alive x dead
Full size table

Discussion

In the present study, we examined how melanoma progression is affected by sensory neurons activity. Our chemogenetic approach revealed that inhibition of sensory activity promotes tumor growth and intra-tumoral angiogenesis. In contrast, excitation of sensory neurons induces melanoma regression with decrease in tumor growth and in new blood vessel formation, as well as a boost in the anti-tumor immune surveillance (Fig. 9). This work indicates that induction of hyperactivation in Nav1.8-expressing sensory neurons represents a potential new therapeutic path in the battle against melanoma.

Fig. 9
figure 9

Schematic illustration summarizing the results of sensory neurons’ activity inhibition and overactivation in the melanoma microenvironment

Full size image

Just as the role of particular genes in a specific biological process can be examined by evaluating examining the assessable consequences that result from their removal (e.g. using knockout animals), the role of neurons in the tumor microenvironment was previously assessed by eliminating them [32, 36, 71, 89, 108, 115, 116, 121, 154, 158]. Nevertheless, although biomedical research has gained some insights into the function of intra-tumoral neurons using loss-of-function studies with surgical or pharmacologic denervation, these strategies are mostly not specific to a given neuronal type. Importantly, a disadvantage of all these studies is that neuronal killing may generate secondary undesirable indirect side effects caused by the inflammatory tissular response which may influence the observed phenotypes (Männ et al. 2016; Christiaansen, Boggiatto, and Varga 2014; Bennett et al. 2005). To circumvent these limitations, novel powerful technologies have been created in the field of modern neuroscience which allow to manipulate the firing of specific neurons without killing them: optogenetics and chemogenetics. These methods use genetic strategies to deliver the expression of light-sensitive proteins or designer receptors exclusively activated by designer drugs, respectively, to the membrane of defined neuronal populations. Therefore, by using these techniques, manipulation of neurons by exposure to light or to designer drugs, without killing them, became feasible. As melanoma is a chronic disease, the use of optogenetics for longer periods, may not be the best approach, as the unavoidable surgical preparation with the chronic implantation of hardware for stimulation and prolonged exposure to highly energetic laser light will eventually culminate in confounding regional inflammation and tissue degradation. Therefore, we chose to use chemogenetics to examine the participation of sensory neurons in melanoma development, as it is more suitable to evaluate the long-term effects of sensory stimulations with less side-effects. Future studies may use similar approaches to explore the role of sensory neurons and other innervations in other cancers.

Our findings suggest that sensory neurons’ overactivation affects the immune response to melanoma. Melanoma progression is influenced by the complex interplay between cancer cells and different components of the immune system [8]. Melanoma cells may cause disruption of the organism’s immunity to overrun and escape the immune system control [96, 104]. The role of sensory innervations in these interactions remains completely unknown. Lymphocytes are the dominant immune elements found infiltrating the melanoma microenvironment. Their composition correlates with patients’ survival [38]. While regulatory T cells play pro-tumorigenic roles; CD8 + T cells, CD4 + T cells, γδ T cells, and NK cells have been shown to act against the transformed cells, [38, 43, 46, 47, 58, 64, 73, 118]. Conversely neutrophils and myeloid-derived suppressor cells have been associated with poor prognosis and are largely pro-tumorigenic [22, 28, 69, 126, 129, 145, 155]. Our data shows that sensory overactivation induce an increase in the number of tumor-infiltrating anti-cancer lymphocytes (CD8 + T cells, CD4 + T cells, γδ T cells, and NK cells), while we did not detect changes in the number of tumor-infiltrating regulatory T cells. We also found that there is a decrease in the number of neutrophils and myeloid-derived suppressor cells within the tumor. We found that these changes were tumor-specific, as we did not detect any alterations in the number of lymphocytes in the non-draining lymph nodes. Tumor-draining lymph nodes presented an increase in some of the anti-tumor lymphocytes probably because of the tumor-priming effect previously reported [135]. Signals transmitted to T cells via PD-1 or CTLA-4 (considered markers for T cells “exhaustion”) promote T cell dysfunction, thereby turning off the immune response [59, 98, 149]. We found that the tumor-infiltrating lymphocytes decrease their expression of PD1 and CTLA-4, possibly indicating that these cells are “less exhausted” within the melanoma microenvironment after sensory hyperactivation. The more active phenotypes of lymphocytes have been associated to the increase in the production of cytokines. A variety of lymphocytes are capable of producing IL-17 [14, 20, 24, 53, 56, 77, 94, 103] which has presented anti-tumorigenic effects in melanoma [3, 63, 74, 75, 91, 99, 100]. We found an increase in tumor-infiltrating lymphocytes producing IL-17 after sensory stimulation. Thus, in light of our overall findings, we suggest that induced increase in firing of sensory innervations contributes to boosting of the immune response against melanoma. Future studies will need to explore the exact molecular mechanisms involved in the interactions of sensory neurons and immune cells in the melanoma microenvironment.

A variety of cellular and molecular mechanisms may be involved in the effect of sensory neurons’ modulation on melanoma behavior. For instance, it has been documented that the same drug that is used to denervate sensory neurons, resiniferatoxin (RTX), an analogue of capsaicin, also induces stress by causing hyperactivation of the sympathetic nervous system [16, 62, 157]. These studies also revealed that sensory nerves may tune down sympathetic nerve activity [16, 62, 105]. Sympathetic neurons release norepinephrine [70, 123], which has been shown to strongly induce tumorigenesis [1, 60, 154]. It remains open the important question whether the effect of sensory innervations in the tumor microenvironment depend also on the modulation of the sympathetic tone.

Future perspectives

The present study reveals the short-term impact of chemogenetic modulation of sensory neurons on melanoma behavior. It remains to be examined what are the long-term effects of this manipulations. In the current study, the sensory neurons’ activity is being continuously inhibited or overactivated. Are changes in sensory neurons’ activity at specific time points sufficient to influence cancer outcomes? Also, it remains to be determined what are the changes within the tumor microenvironment at different stages of cancer progression. Are some stages more sensible to changes in the activity of sensory neurons? Moreover, this study focuses on melanoma tumors. Future studies should explore what is the role of sensory neurons in the development of other solid tumors.

A variety of factors secreted from sensory neurons may be implicated in the regulation of the melanoma microenvironment described here [18]. The overactivation of sensory nerve fibers may induce the release of neuropeptides, such as substance P, CGRP, VIP, GRP, neurokinin A, neurokinin B, neuropeptide Y (NPY) and adrenomedullin, which have been shown to interact directly with several components within the tumor microenvironment, including cancer cells, immune cells, endothelial cells and others [4, 6, 17, 25, 26, 30, 35, 61, 80, 81, 93, 95, 117, 120, 142, 143, 148]. We detected an increase in intra-tumoral CGRP after overactivation of sensory neurons (Fig. 10). However, it remains unknown whether the decrease in the tumor size after sensory neurons overactivation is due to this increase in the intra-tumoral concentration of CGRP. Future studies will need to genetically eliminate specific receptors for this and other neuropeptides (such as receptor activity-modifying protein 1 (RAMP1) for CGRP) from different cellular components of the tumor microenvironment to reveal whether those communications are relevant for melanoma outcomes.

Fig. 10
figure 10

Increase in sensory neurons activity leads to increase in intra-tumoral calcitonin gene-related peptide (CGRP). CGRP concentration was measured in tumor samples from Nav1.8-Cre + /hM4Di + and Nav1.8-Cre-/hM4Di + animals (A) and from Nav1.8-Cre + /hM3Dq + and Nav1.8-Cre-/hM3Dq + mice (B). (C) Schematic representation of the association between intra-tumoral concentration of CGRP and tumor size. (n = 5). Data are shown as mean ± SEM

Full size image

Conclusion

In conclusion, this work identifies sensory neurons overactivation as a potential strategy for blocking melanoma progression and improving patient outcomes. We also anticipate that drugs that reduce sensory neurons hyperexcitability used for analgesic treatment may cause undesired effects in cancer patients and need to be carefully evaluated in pre-clinical models of cancer before using them in these patients [34, 125]. Moreover, the relationship between sensory hyperexcitability and cancer outcomes is likely to inform studies of other cancers that are also infiltrated by sensory neurons.