Targeting the peripheral neural-tumour microenvironment for cancer therapy

Abstract

As the field of cancer neuroscience expands, the strategic targeting of interactions between neurons, cancer cells and other elements in the tumour microenvironment represents a potential paradigm shift in cancer treatment, comparable to the advent of our current understanding of tumour immunology. Cancer cells actively release growth factors that stimulate tumour neo-neurogenesis, and accumulating evidence indicates that tumour neo-innervation propels tumour progression, inhibits tumour-related pro-inflammatory cytokines, promotes neovascularization, facilitates metastasis and regulates immune exhaustion and evasion. In this Review, we give an up-to-date overview of the dynamics of the tumour microenvironment with an emphasis on tumour innervation by the peripheral nervous system, as well as current preclinical and clinical evidence of the benefits of targeting the nervous system in cancer, laying a scientific foundation for further clinical trials. Combining empirical data with a biomarker-driven approach to identify and hone neuronal targets implicated in cancer and its spread can pave the way for swift clinical integration.

Introduction

Apart from their roles as messengers of the central nervous system (CNS), neurons guide tissue growth, upkeep, functionality and differentiation through a mixture of electrical and chemical signals. This fundamental biological communication regulates numerous physiological processes during development as well as processes involved in neurodegeneration and other disease states, including cancer. More than a century ago, initial studies depicted the presence of peripheral nerves in malignant tumours¹. Subsequent research connected tumorigenesis to nerve development and regeneration. This connection is often attributed to the secretion of neurotrophic growth factors within the tumour microenvironment (TME), prompting nerve sprouting into the tumour stroma^2,3. Concomitantly, these nerves can discharge growth factors, morphogens and neurotransmitters in the TME, and these dynamics spur the growth of cancer cells and the activity of tumour-supporting cells, fuelling cancer growth^4-6.

Modern breakthroughs in spatial biology such as spatial transcriptomics and high-plex protein labelling, coupled with the use of denervated tumour models and in vivo imaging, have equipped researchers with robust tools to explore nerve–cancer cell interplay and to help identify the roles of distinct nerve phenotypes within different types of cancer and different tumour compartments. Highly sensitive experimental techniques can pinpoint molecules in the axoplasm even at low levels of expression that are instrumental to cancer development, progression or response to therapy and can be targeted using existing neuromodulatory drugs in clinical oncology^7-9.

In this Review, we present the foundational biology of peripheral tumour innervation and cancer–neuron interaction and demonstrate how recent empirical data can augment biomarker-driven strategies to repurpose nervous system-focused drugs to treat cancer. Given that more than 95% of tumours originate outside the CNS, we focus on the interplay between communicating elements in the peripheral nervous system (PNS) and tumours of various origins. Finally, we examine the actionable targets nestled within the cancer–immune system–neuron interface that may show clinical efficacy.

The PNS in health and disease

The neurons of the PNS can be broadly categorized into four primary systems: cranial and spinal nerves, and the somatosensory and autonomic nervous systems. The autonomic nervous system further splits into the sympathetic and parasympathetic branches, with noradrenaline (also known as norepinephrine) and acetylcholine (ACh) as their main neurotransmitters, respectively. On a structural level, peripheral nerves comprise multiple axon bundles or fascicles. The size, number and configuration of these fascicles (such as myelination status and conduction velocity) can vary between and throughout these nerves^10,11.

The PNS transmits signals between the CNS and the rest of the body. It has numerous functions, such as controlling movement, regulating blood pressure and sensing touch and pain. However, the PNS also has an unappreciated crucial role in tissue development, regeneration and pathological processes^12-14.

The peripheral nerves develop in the early stages of embryogenesis, and there is evidence for their role in coordinating the development and morphology of several organs^12,13, which is attributed to PNS capacity to modulate the fate and stemness of multipotent cells¹⁵. Later in life, nerve-derived soluble factors coordinate the regenerative process as part of a physiological wound-healing response by modifying differentiation and replicative state of non-neuronal cells^12,14,16,17. Soluble factors derived from non-neuronal cells can also stimulate neurite outgrowth from peripheral nerves to the regenerative site, supporting the existence of coordinated molecular crosstalk between neuronal and non-neuronal cells¹². Specifically, during wound healing, neuronal signals stimulate the recruitment and differentiation of cells to the injured site, induce cell proliferation and control the assembly of tissue architecture^14,16,17. In addition to these effects, the PNS exerts an important control over the inflammatory cells¹⁶. The peripheral nerves sense immune signals and damage-associated molecular patterns (DAMPs) and respond to these stimuli by releasing immunoregulatory molecules^14,18. As observed in regenerating tissues, non-neuronal cells mobilize neurons to spurt dendrite projections towards the damaged tissue by releasing neurotrophic factors. This is shown by the increased nerve density in the skin during wound healing¹⁴.

All these processes, including regulation of differentiation, migration and proliferation of non-neuronal cells, extracellular matrix remodelling and immune response modulation, contribute to cancer progression¹⁹. Although the roles of nerves are well established in tissue homeostasis in health and in response to various pathological processes outside of cancer — either exogenous (such as infection and trauma) or endogenous (such as autoimmune disease) — we start to realize the crosstalk between neurons and cancer cells and how it impacts the phenotype of different cancers. Data suggest that cancer cells hijack the PNS and promote the growth of nerves in cancer in a process similar to angiogenesis. These interactions poise neural regulation of cancer as a potential actionable hallmark of cancer²⁰.

Neuron signalling in cancer

The exchange of molecules through synapses is the best-known form of neuronal communication; however, release of soluble factors in a paracrine or endocrine fashion represents the main communication channel between neurons and non-neuronal cells¹⁹. Neuronal signals are widely conserved, and most cells, including cancer cells, are sensitive and responsive to these signals^21,22. Most of these signals are mediated by G protein-coupled receptors (GPCRs, such as muscarinic acetylcholine receptors (mAChRs)), but others include enzyme-linked receptors (such as the neuropeptide nerve growth factor (NGF) receptors, NTRK1 and p75^NTR) and ion channel-linked receptors (such as nicotinic acetylcholine receptors (nAChRs)).

These receptors can be found in non-neuronal cells, including malignant cells^23-28, and are upstream in pathways vital to cancer biology. Many of the pathways downstream of the neuronal signals have been studied clinically as drug targets^16,24,29. For example, the activation of NTRK1 triggers survival and differentiation mechanisms in neurons, along with neurite outgrowth, and these activities are carried by activating downstream signalling pathways such as PI3K and Ras–Raf–MAPK³⁰. In cancer, PI3K and Ras–Raf–MAPK signalling are also responsible for the maintenance and progression of malignant cells and have been extensively explored as targets for cancer therapies³¹. Furthermore, NTRK1 downstream events are also known to activate transcription factors such as CREB, which is one of the most important transcription factors in the nervous system, controlling the expression of genes involved in neuronal development and survival, neuronal plasticity and neurotransmission³². CREB activation is also an important mechanism of cancer cells³³, and its overexpression has been linked to transcriptional changes associated with uncontrollable cell growth, apoptosis resistance, angiogenesis, immune evasion, epithelial–mesenchymal transition and poor clinical outcomes^34-36. Targeting CREB, NGF or NTRK1 may block various conserved downstream signalling pathways involved in the progression of multiple cancers, thus presenting a potential therapeutic target. Elevated levels and increased activity of CREB have been noted in cancerous tissues from patients with prostate cancer, breast cancer, non-small-cell lung cancer (NSCLC) and acute leukaemia³⁴. Conversely, reducing CREB levels in various cancer cell lines led to decreased cell growth and the initiation of cell death³⁴. However, it is important to consider that inhibiting a transcription factor such as CREB can be challenging owing to the inherent non-specificity of such targets, which may lead to off-target effects and toxicity^37-40.

Another interesting target is the cholinergic signals. The neurotransmitter ACh is expressed by human and mouse keratinocytes, along with the enzyme choline O-acetyltransferase (ChAT), which is involved in the biosynthesis of ACh. Interestingly, the expression of both ACh and ChAT has been found to be directly linked to cellular growth rate⁴¹. Mechanistically, the role of ACh in the growth of non-neuronal cells includes modulation of cytoskeleton molecules. ACh and ChAT expression promotes activation of WNT signalling and was directly associated with the development of EGFR tyrosine kinase inhibitor resistance in NSCLC⁴². Post-therapy residual cancer cells showed accumulation of ACh mediated by increased expression of ChAT induced by EGFR tyrosine kinase inhibitor administration. In silico analysis revealed that increased expression of ChAT was associated with worse clinical outcomes⁴². The similarities in the mechanisms that underlie neuronal communication and cancer signalling could potentially pave the way to develop novel therapeutic strategies for cancer treatment.

PNS tumours

Much of our initial understanding of cancer neuroscience comes from studies of tumours originating in the PNS, a heterogeneous group of mostly benign tumours that are rare in the general population. A fundamental understanding of the pathophysiology of these tumours, as well as of cancers that recapitulate characteristics of these tumours, could support targeting of tumour-innervating neurons in other solid tumours.

Neurofibromas.

These are benign PNS tumours relatively common in patients with neurofibromatosis type 1 (NF1), carriers of germline monoallelic neurofibromin 1 (NF1) gene mutations. Neurofibroma development relies on a complex signalling interplay between the neoplastic Schwann cells and other components such as nerves, mast cells and fibroblasts⁴³. Schwann cells with NF1 loss of heterozygosity (LOH) promote the recruitment and proliferation of mast cells by secreting the stem cell factor (SCF) that binds to the KIT receptor of mast cells. In turn, transforming growth factor-β (TGFβ) released by mast cells promotes an aberrant extracellular deposition by fibroblasts in neurofibromas⁴⁴. Owing to the significant extracellular matrix remodelling and high collagen content of neurofibromas⁴⁵, strategies to reduce collagen, possibly by targeting Schwann and mast cells, could benefit NF1-mutation carriers⁴⁶. The MEK inhibitor selumetinib was the first FDA-approved drug for neurofibroma treatment, and phase II clinical trials suggest effective neurofibroma growth inhibition by counteracting the activation of the RAS–MAPK pathway mediated by NF1 LOH^4,48. Drugs that target the recruitment and activation of mast cells have also been tested, as hindering the recruitment of mast cells to nerves seems to interfere with neurofibroma growth⁴⁹. The tyrosine kinase inhibitor imatinib inhibits SCF-mediated KIT signalling, and clinical studies indicate a potential clinical benefit in reducing neurofibroma volume⁵⁰. Another tyrosine kinase inhibitor, cabozantinib, which has KIT inhibition activity, has also shown clinical benefit in patients with neurofibroma⁵¹.

Schwannomas.

Schwannomas originate from Schwann cells, but other TME components — such as axons, macrophages, T cells and endothelial cells — also play an essential part in their development⁵². The pathogenesis of schwannomas is tied to a flawed nerve repair process, in which Schwann cells are unable to undergo the necessary phenotypic transitions to facilitate nerve regeneration. Following a nerve injury, Schwann cells lose their myelinating properties, adopting pro-inflammatory and phagocytic characteristics to aid in myelin clearance caused by axonal degeneration. This myelin phagocytosis triggers a transformation in Schwann cells, leading to increased proliferation and migration, forming cellular tracks to guide axonal regrowth. The process culminates when Schwann cells regain their myelinating differentiation state, facilitated by intrinsic mechanisms and extrinsic signals from regenerated axons and macrophages. The inability to regain the myelinating phenotype while sustaining a proliferative state is suggested to be pivotal for Schwannoma development. Both sporadic and syndromic schwannomas, associated with neurofibromatosis type II (NF2), are characterized by the loss of MERLIN protein activity, primarily caused by mutations in the NF2 gene⁵³. This loss hinders the ability of Schwann cells to redifferentiate into their final myelinating state, leading to the uncontrolled activation of pathways such as Ras–MAPK, mTOR and HER2 signalling, perpetuating a continuous proliferative state in these cells⁵². In vitro studies have shown that Nf2 mutations are also associated with an accumulation of macrophages at the site of nerve injury⁵². Although macrophages are essential during the initial phases of nerve repair by promoting most myelin clearance, they impair the repair resolution and can promote tumour growth⁵⁴. Such mechanisms may also be significant in the context of cancer. Nerve injury caused by tumour compression or PNI can also trigger similar nerve repair responses by Schwann cells and macrophages, reshaping the TME and fostering a pro-tumorigenic inflammatory environment and facilitating cancer cell dissemination^55,56.

Malignant peripheral nerve sheath tumours.

Malignant peripheral nerve sheath tumours (MPNSTs) are aggressive cancers that represent up to 10% of all soft tissue sarcomas⁵⁷. Approximately half of these tumours develop when neurofibromas transform into MPNSTs in patients with NF1. As a result, these tumours are linked to the loss of neurofibromin due to NF1 mutations. Interestingly, although an NF1-heterozygous TME is required for benign neurofibroma development, in vivo studies showed that it impairs malignant transformation⁵⁸. The remaining cases emerge sporadically⁵⁹, or more rarely, are associated with radiation therapy⁶⁰, suggesting that radiation-treated primary cancers might evolve into MPNSTs owing to changes in TME cellular components. Mouse models have highlighted the significance of the TME in MPNSTs⁵⁹, as cells in the TME respond differently to certain compounds from their counterparts found in healthy tissue⁶¹. This difference underscores the differential impact that these interventions (such as radiation or targeted therapy) have on the neuronal tissue in cancer and the potential for selective targeting of the TME while sparing healthy surrounding nerves.

Neuronal-like tumour characteristics.

Some cancer cells mimic neuronal behaviour by altering transcriptional programmes, specifically by inhibiting the RE1 silencing transcription factor (REST)⁶². This factor keeps many neuronal genes silent in differentiated non-neuronal cells. Notably, decreased REST levels have been documented in various cancers such as those of the breast⁶³, lung⁶⁴, prostate⁶² and colon⁶⁵. As a result, these cells express neuroendocrine genes, such as chromogranin A (CHGA) and synaptophysin (SYP), typically repressed by REST⁶⁶. In prostate cancer, diminished REST activity correlates with the emergence of a neuroendocrine phenotype, which is linked to aggressive, androgen-independent disease progression. The loss of REST is linked to a neuroendocrine phenotype that triggers various mechanisms that promote tumour growth, such as sustaining proliferative signals, avoiding cell death, promoting blood vessel formation and facilitating cell movement. These mechanisms are influenced by well-known cancer signalling pathways. For example, inhibition of REST with short hairpin (shRNA) significantly increased the PI3K signalling pathway after EGF stimulation⁶⁵. Although the regulation of REST expression is not fully understood and no compounds currently target REST⁶⁶, new technologies to modulate the neuroendocrine phenotype are under development. For instance, an aberrant splicing isoform of REST is overexpressed in lung and prostate tumours with a neuroendocrine phenotype. An antisense oligonucleotide designed to block the binding of a splicing factor (splice-switching oligonucleotide) to REST was able to prevent the formation of the aberrant REST splicing form and restore the expression levels of REST-controlled genes. Moreover, it effectively suppressed tumorigenesis in vitro and in vivo⁶⁷.

Tumours within the CNS and tumour cells that metastasize to the brain can exhibit neuronal characteristics, such as the development of neurite-like projections capable of forming functional synapses with neurons^68,69. The synapses between cancer cells and neurons offer specific advantages for cancer cells. In the case of glioma, these synapses, which are controlled by glutamate receptors, have a significant impact on prompting aggressive behaviours in cancer cells. This stimulation leads to increased cell proliferation and invasion, furthering the progression and severity of the disease. Currently, there is limited evidence to suggest the existence of synapses between cancer cells and neurons in the PNS^27,70,71.

These hybrid cells create microcircuits within the TME and can trigger other cancer cells to activate similar neuronal gene programmes, which leads to cell proliferation^{27 ,72}. From a clinical perspective, high mRNA levels of neurogenesis-related genes are associated with poor prognosis in prostate, ovarian and colorectal cancer^73-75. Moreover, breast cancer cells expressing γ-aminobutyric acid (GABA)⁷⁶ and amyloid β-secreting melanoma⁷⁷ cells demonstrate neuronal behaviour and interact with peripheral nerves in the TME. This engagement may be influenced by cancer stem cells, resembling embryonic neural progenitors^78,79. In certain tumour models, such as those of the stomach and colon, these stem cells have even differentiated into functional neurons, hinting at a unique neurogenesis and synaptogenesis pathway and the creation of neuron–tumour circuits⁸⁰. This transcriptional reprogramming might propose novel therapeutic targets in cancer treatment.

Nerve co-option in the TME

The traditional view of the TME consists of malignant cells, fibroblasts, blood vessels, lymphatics, stromal cells and immune cells. Introducing peripheral nerves into this concept brings a new layer of complexity in terms of cellular components (neurons and the peripheral glial cells, such as Schwann cells) and signalling mechanisms (such as efferent-like sensory nerves, and afferent-like motor and autonomic nerves). These components engage in continual crosstalk, creating functional microcircuits between nerves, cancer cells and the other TME elements (Fig. 1). These physiological mechanisms, including PNS regulation of the fate of multipotent cells (such as those in the basal layer of the skin), fibroblast activity and immune cell infiltration into tissues, are fundamentally important in TME homeostasis and can be hijacked by cancer⁸¹. However, the first step in this process would be the growth of nerves into the TME and the formation of the neural niche. It is now evident that cancer cells can manipulate and recruit neurons to the TME, and in the following paragraphs, we review the relevant mechanisms involved in these processes.

Fig. 1 ∣. Structure of the neural niche.

The neural niche comprises three integrated components: neural, immune and cancer cells. Notable cells within the nervous system include neurons and Schwann cells. Predominant immune cells in the tumour microenvironment include dendritic cells, T cells, natural killer cells, macrophages, neutrophils and myeloid-derived suppressor cells. These immune cells have receptors for neurotransmitters and neuropeptides on their extracellular membranes, enabling them to detect and respond to neuronal cues. Peripheral nerves typically comprise multiple fascicles, or bundles, of axons. Each fascicle is ensheathed (enclosed) by the endoneurium. Groups of axons are further enveloped by another connective tissue layer known as the perineurium, which has an epithelium-like structure. The entire nerve is encapsulated by an external sheath termed the epineurium. Nerves may be either unmyelinated or myelinated. Certain Schwann cells have the capability to both envelop and myelinate nerve fibres. These fibres rely on macrophages and specific proteins to execute various cellular functions such as regeneration and debris clearance. Another group of Schwann cells remain prepared to undertake diverse repair functions. Interestingly, these repair-ready Schwann cells can be exploited by cancer cells, which use them to establish pathways, facilitating progression and expansion of the cancer along peripheral nerves. The blood–nerve barrier surrounds these nerve layers, delineating a functional space for the axons, Schwann cells and other peripheral nerve cells. Structurally, the blood–nerve barrier consists of the endoneurial microvessels located within the fascicle and the surrounding perineurium, which are interconnected by specialized tight junctions that restrict capillary permeability. Positioned next to these capillaries, the perineural sheath features abundant tight junctions among its cells, ensuring that each fascicle remains separated from the surrounding interfascicular and epineural spaces. Pericytes, which are interspersed along capillary walls, play a pivotal part in blood vessel formation and in sustaining the integrity of the blood–nerve barrier²¹¹. To access a peripheral nerve, blood-borne molecules must traverse the endoneurial vascular endothelium to enter the endoneurium, and they need to pass through the perineurium to reach the nerve fascicles^212,213. A compromised blood–nerve barrier, often indicated by disrupted tight junctions²¹⁴, results in permissive signalling between cancer cells and nerve filaments. Such intrusions permit immune cells to enter the endoneurial space, triggering an inflammatory response that can further harm the neuroglia. A permeable blood–nerve barrier has also been associated with the growth of peripheral nervous system tumours, including neurofibromatosis and schwannomas. From a therapeutic perspective, nerves represent a relatively privileged environment, and the integrity of the blood–nerve barrier might hinder the efficacy of cancer therapeutics.

Tumour innervation

Similar to neo-angiogenesis, in which cancer cells emit factors that promote blood vessel growth into the tumour, tumours can modulate their interaction with nerves by supporting nerve outgrowth. Unlike blood vessels, neurons comprise mature, highly specialized, differentiated cells. Growth signals must overcome the functional barrier of these mature cells to induce tumour innervation, which frequently requires long-distance travel as the nerve cell body (soma) is often located in a ganglion distant from the tumour. The process varies by location, as nerve composition differs between organs^82-84. The key soluble factors shown to induce tumour-associated axonogenesis and increased nerve density within the TME include NGF, brain-derived neurotrophic factor (BDNF), glial cell line-derived neurotrophic factor (GDNF), semaphorins, netrins, axon-guidance molecules (such as ROBO and Slit proteins) and other regulators of cell differentiation such as microRNAs^{5,12,78,80,85-87}. Physiologically, these factors support neuron development and response to injury^85,88,89, but they are also pivotal for cancer-associated neuroplasticity within the TME. This neuroplasticity entails the growth of nerve filaments (such as neurites, which include both axons and dendrites) into the TME in a process termed neuritogenesis, as well as the acquisition of de novo neuronal features in pre-existing neurons (such as adrenergic signalling in afferent sensory neurons). The expression of such factors has been associated with the clinicopathological features of cancer⁸⁷. For instance, the expression of NGF by prostate cancer cells was linked to tumour axonogenesis and advanced tumour grade^90,91. In gastric cancer, tumour NGF expression was stimulated by cholinergic signalling from nerves and specialized epithelial chemosensory cells, triggering tumour axonogenesis; ACh blockade by vagotomy caused downstream Hippo signalling pathway inhibition and hindered tumour progression⁹². In head and neck squamous cell carcinoma (HNSCC) microRNAs carried by exosomes secreted by TP53-deficient cancer cells were shown to control the fate of intratumoural nerves, triggering the nerve reprogramming to an adrenergic phenotype. Specifically, miR-34a, miR-21 and miR-324 have pivotal roles in regulating nerve growth and establishing neuronal identity. miR-34a acts as an inhibitory regulator, suppressing neurite outgrowth, whereas miR-21 and miR-324 promote neuritogenesis and neuronal differentiation. Together, these microRNAs create a balanced regulatory network that activates transcriptional programmes essential for defining specific neuronal subtypes and ensuring proper nervous system function. Importantly, the adrenergic signalling blockade resulted in tumour growth inhibition⁹³.

Neurogenesis in adults is limited to the brain (to the subgranular zone of the dentate gyrus in the hippocampus and subventricular zone of the lateral ventricles), yet it has been described in cancer; neural progenitors from the CNS expressing doublecortin (a protein expressed by neuronal precursor cells that initiates cell division, used as a neurogenesis and neuronal progenitor marker) were found to migrate through the bloodstream and infiltrate primary and metastatic prostate tumours⁹⁴. These changes promote the formation of morphologically and functionally distinct nerves, referred to as tumour-associated neurons. It is currently uncertain whether such an event occurs in other peripheral tumours.

Perineural invasion

Cancer cells can invade nerves in the TME in a process called perineural invasion (PNI). Similar to haematogenous, lymphatic and local invasion routes, PNI is a pathway for tumour spread. Clinically, its presence signifies higher-grade disease, leading to worse outcomes and the need for intensified treatments (Box 1). The process of PNI involves intricate signalling between nerves and cancer, immune and Schwann cells. Schwann cells are essential for the maintenance of the axonal myelin sheath, releasing factors that ensure neuronal homeostasis and response to injuries^95,96. Unlike the invasion of blood and lymphatic vessels, as tumours invade the lumen-lacking nerves, the resulting nerve damage activates regeneration and repair programmes in both Schwann cells and the nerve itself^97-99. Activated Schwann cells adopt a repair phenotype that creates pathways that ease cancer cell movement along and within the lumen-lacking nerve fibres⁵⁵. Myelin degradation (also known as demyelination) leads to nerve damage response involving the release of cytokines, including IL-1, IL-6, IL-10 and TGFβ^{95,96,100-103}. Downstream immune response includes the recruitment of growth factor-releasing macrophages, like those producing GDNF, that enhance cancer cell migration along the nerve and remodel the extracellular matrix to be more conducive to PNI^56,104,105. Subsequently, this distinct state of invaded tumour-associated nerves results in a unique immune response in the neural niche that negatively influences the global tumour immune microenvironment, promoting regulatory and immunosuppressive responses¹⁰⁶.

Box 1 ∣. Nerves as cancer biomarkers.

With the realization that cancer alters the normal neuronal landscape, both qualitatively and quantitatively, the role of nerves as potential biomarkers for cancer has recently been studied. Currently, the only neuronal marker used in the clinic is the presence of perineural invasion (PNI). PNI is an ominous feature in multiple tumours and is incorporated as a dichotomous variable (present or absent) in the staging system of various carcinomas (such as cutaneous and mucosal head and neck squamous cell carcinoma (HNSCC), as well as penile carcinomas)²⁵¹. Yet, the predictive value of PNI that remains localized to the excised tumour and of the PNI burden is unclear^104,252. A recent study showed that the presence of PNI was a strong prognostic factor in oral cavity cancer²⁵³, and spatial molecular mapping of neuroinvasive cancer cells showed that the distance between nerves and tumours was also a significant prognostic factor, even among patients without PNI; indeed, shorter tumour-to-nerve distance was associated with worse disease-specific and overall survival²⁵³.

Beyond PNI, tumour nerve density in the tumour microenvironment (TME), like levels of immune cells in the TME, is linked to tumour characteristics and outcomes. This marker was associated with increased extracapsular extension, poor recurrence-free survival and increased tumour proliferation in prostate cancer^5,254,255; TP53 mutation status⁹³ and worse disease-specific survival²⁵⁶ in HNSCC; lymph node metastasis in breast cancer⁹¹; deeper tumour invasion in gastric cancer³; and poor disease-specific survival in colorectal cancer²⁵⁷. Increased nerve density using metrics normalized for different anatomical sites was also associated with PNI detection and worse disease-specific survival in patients with oral cancer. Increased normalized nerve density was associated with worse outcomes, even among patients with no PNI²⁵⁶. These data support the idea of considering the density of nerve fibres in the TME as a prognostic factor.

In addition, the density of specific nerve phenotypes was also investigated as a potential biomarker. Increased density of nerve fibres from the autonomic system (TH⁺) is a clinical predictor of prostate biochemical recurrence²⁵⁸ and is associated with overall and recurrence-free survival in HNSCC⁹³. Interestingly, the density of regenerating neurons (such as those expressing growth-associated protein 43 (GAP43)) is associated with worse outcomes in patients with pancreatic cancer, which might indicate a prognostic role of nerve response to injury^259-261. In breast cancer, increased sympathetic nerve density is associated with worse recurrence-free survival, whereas increased parasympathetic nerve density is associated with improved recurrence-free survival¹¹⁵.

Tumour addiction to PNS signalling

Distinct types of nerves innervate different tumours. Nerves in the TME might be part of the pre-existing normal innervation patterns of the organ or induced to grow to the TME by the cancer cells. These nerves carry divergent functions that distinctly affect cancer cells and cells in the TME by modulating known cancer-related pathways⁴². However, the pharmacological blockade of specific downstream signalling poses challenges owing to the adverse effects of available inhibitors¹⁰⁷. In contrast to directly targeting these intracellular signalling processes, targeting upstream neuronal signals may offer a safer way to modulate multiple key cancer pathways at once⁴². This type of upstream inhibition might not only be safer but can also aid in overcoming resistance to targeted therapy¹⁰⁸. This section delves into the peripheral neuronal subtypes that innervate peripheral tumours and spotlights components of neuron–immune cell–cancer crosstalk that could be pharmacologically targeted (Fig. 2).

Fig. 2 ∣. Interactions within the tumour niche.

The interaction between the nervous system and the immune system has a pivotal role in tumour progression, immune evasion and therapy resistance. Neurons emit neurotransmitters that influence immune cell activity and promote cancer cell growth, whereas cytokines released by immune cells enhance neuronal activity and stimulate cancer cell aggressiveness. Below are examples of how neuronal signals modulate interactions in the tumour microenvironment and potential therapeutic compounds. Preclinical studies have indicated antitumour activity associated with β-adrenergic receptor inhibition using selective or non-selective β-blockers^109,133. Clinical evidence in support of oncological benefits of β-adrenergic blockers is primarily derived from retrospective studies limited to specific tumours^116-120. Manipulation of adrenergic signalling may augment immunotherapy response, with α₂-adrenergic signalling potentiating antitumour immune response and β-adrenergic signalling hindering it. Thus, α2 adrenergic receptor agonists (for example, clonidine, guanfacine)¹³⁹ and β-adrenergic antagonists^138,141,143 could potentially enhance the anti-tumour efficacy of immune checkpoint blockade (ICB) therapy. The γ-aminobutyric acid (GABA)ergic signalling pathway can have pro-tumorigenic or anti-tumorigenic effects depending on the tumour type^176,215. GABA_B receptors have an antitumour role in certain breast, colon, lung and liver tumours^176,215. Baclofen, a GABA_B receptor agonist, exhibits potential clinical benefits, inhibiting cancer cell migration and tumour growth in colorectal and liver cancers¹⁷⁴. Acetylcholine (ACh) promotes cholinergic stimulation in cancer cells expressing nicotinic or muscarinic ACh receptors. Agonists such as nicotine stimulate nicotinic acetylcholine receptors (nAChRs) and induce tumour-promoting signalling^162,216,217, whereas antagonists such as atropine and pirenzepine inhibit this mechanism^165,218-220. Tumour cells can synthesize and release ACh to support their own growth through an autocrine/paracrine loop mediated by the vesicular acetylcholine transporter (VAChT). Vesamicol, an inhibitor of VAChT, hinders ACh production and secretion in lung cancer cells, inhibiting proliferation^216,218. Dopamine signalling has been observed in various tumours, and experimental manipulation of this pathway influences tumour biology²²¹. Compounds that target dopamine receptors, such as thioridazine, ONC201 and trifluoperazine, have shown promising anticancer effects in preclinical and clinical trials^221-223. The neurokinin-1 receptor (NK-1R) is the receptor for Substance P (SP). The NK-1R/SP axis has been extensively associated with pro-tumour signaling^224,225. The NK1R antagonist, aprepitant, has demonstrated significant antitumour activity in various tumour types^199,226-228. Neurokinin receptors play a vital part in regulating the immune system. However, the impact of NK-1R/SP antagonism varies across different tumour types. In tumours with high inflammatory components, NK-1R/SP inhibition was associated with anti-tumour activity, whereas in tumours with low inflammatory components, NK-1R/SP inhibition led to increased metastatic activity¹⁹⁹. Calcitonin gene-related peptide (CGRP), derived from tumour-associated nociceptors, interacts with the RAMP1/CALCRL receptor. As these receptors are present in immune cells from various tumor types, CGRP can modulate immune response^181,229. For instance, CGRP-RAMP1 signalling leads to CD8⁺ T cell exhaustion in melanoma models¹⁸¹. Inhibiting CGRP–RAMP1 signalling with antagonists such as Rimegepant, reduces immune exhaustion and tumour growth, and enhances anti-tumour activity of other oncological drugs¹⁹². The vasoactive intestinal peptide (VIP) and its receptor (VIPR) are implicated in pro-tumorigenic signalling^230,231, and inhibition of VIP–VIPR signalling with experimental VIPR antagonists ANT008 or ANT308 improved antitumour immunity in pancreatic cancer models²³². Transient receptor potential (TRP) channels, particularly TRPV1, play a part in modulating neurogenic inflammation. Agonists like capsaicin, stimulate the neuronal secretion of the inflammatory neuropeptides CGRP and SP²³³. Although TRPV1 upregulation has been reported in some tumours, there is debate about whether these receptors are active in cancer cells²³⁴. Pharmacological approaches targeting TRPV1 have been explored as a potential co-adjuvant to promote an antitumour immune response, but preclinical results are contradictory^234,235. AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; BDNF, brain-derived neurotrophic factor; BK, large conductance calcium-activated potassium channels; β₂-AR, β₂-adrenergic receptor; CALCRL, calcitonin gene-related peptide type 1 receptor; CCK, cholecystokinin; eCB, endocannabinoid; IGF3, insulin-like growth factor 3; NA, noradrenaline; NGF, nerve growth factor; NMDA, N-methyl-D-aspartate; NPY, neuropeptide Y; NT, neurotensin; P2Y, purine; PTEN, phosphatase and tensin homologue; R, receptor; 5-HTR, serotonin receptor; SP, substance P.

Adrenergic neurons

Adrenergic neurons mediate adrenergic signalling by releasing noradrenaline, which binds to α- or β-adrenergic receptors. The interaction between noradrenaline and the adrenergic receptors leads to stimulatory downstream signals. The α₂-adrenergic receptor is the only exception. The adrenergic receptors are GPCRs abundantly expressed within the TME, including in malignant cells¹⁰⁹ (Fig. 3). The α₂-adrenergic receptor couples with the inhibitory G protein, acting as a negative regulator to conclude the response to adrenergic signalling. Many preclinical studies indicate that the expression of adrenergic receptors in cancer cells has significant implications for cancer treatment. In vivo, adrenergic agonists promote pro-tumorigenic signalling in cancer cells, whereas antagonists abrogate apoptosis evasion¹¹⁰, cell proliferation¹¹¹, survival¹¹² and invasion¹¹³, notably through β₂-adrenergic receptor signalling¹⁰⁹. Adrenergic signalling also influences other cellular components in the TME, such as endothelial and immune cells, impacting mechanisms of angiogenesis¹¹⁴, metastasis⁹¹ and immune evasion¹¹⁵.

Fig. 3 ∣. Neuronal signatures in cancer.

a, An analysis of the transcriptome data from the Cancer Genome Atlas (TCGA) database reveals that various human tumours show an enrichment of neuronal-related signatures and demonstrates differential neuronal signalling between different tumour types. Enrichment scores for specific Gene Ontology database terms tied to neuronal signalling pathways were determined via single-sample Gene Set Enrichment Analyses (ssGSEA). The enrichment score (ES) gauges the representation of a gene set based on its expression ranking among all sequenced genes. Positive enrichment score values indicate upregulation, and negative values indicate downregulation. In the provided plots, the x-axis showcases the ES from the ssGSEA distribution for samples in each tumour cohort (bladder urothelial carcinoma (BLCA)^236,237, breast invasive carcinoma (BRCA)^238,239, colon adenocarcinoma (COAD)²⁴⁰, glioblastoma (GBM)^241,242, head and neck squamous cell carcinoma (HNSCC)²⁴³, lung squamous cell carcinoma (LUSC)^244,245, pancreatic adenocarcinoma (PAAD)²⁴⁶, prostate adenocarcinoma (PRAD)²⁴⁷, rectum adenocarcinoma (READ)²⁴⁰, skin cutaneous melanoma (SKCM)²⁴⁸) split by a dashed line representing an enrichment score of ‘0’. Peaks to the right of this line signify gene set upregulation and those on the left indicate downregulation. b, Relevant genes involved in cancer–neuron signalling crosstalk have a variable range of expression between tumour types and within the same tumour. For instance, semaphorin 4F (SEMA4F) expression is consistently high among tumour types, whereas γ-aminobutyric acid type B receptor subunit β3 (GABRB3) expression is highly heterogeneous within samples from the same group and between different tumour types. The normalized RNA sequencing data underwent a log₂ transformation for better visualization. A higher y-axis value signifies more transcripts for a particular gene, whereas values near 0 denote scarce transcript detection. The colour code displays the percentage of samples in each tumour type with a specified gene expression level. ACh, acetylcholine; ACHE, acetylcholinesterase; ADRA1B, adrenoceptor α1B; ARTN, artemin; BDNF, brain-derived neurotrophic factor; CALCA, calcitonin-related polypeptide α; CHAT, choline O-acetyltransferase; CHRM1, cholinergic receptor muscarinic 1; CHRNA5, cholinergic receptor nicotinic α5 subunit; DCC, DCC netrin 1 receptor; GABA, γ-aminobutyric acid; GABRA3, γ-aminobutyric acid type A receptor subunit α3; GAP43, growth-associated protein 43; GDNF, glial cell line-derived neurotrophic factor; GFRA1, GDNF family receptor α1; GRIA1, glutamate ionotropic receptor AMPA type subunit 1; GRIN2A, glutamate ionotropic receptor NMDA type subunit 2A; GRM4, glutamate metabotropic receptor 4; NGF, nerve growth factor; NGFR, nerve growth factor receptor; NRP1, neuropilin 1; NTN1, netrin 1; NTRK1, neurotrophic receptor tyrosine kinase 1; SEMA3A, semaphorin 3A; TAC1, tachykinin precursor 1; TACR1, tachykinin receptor 1; VIP, vasoactive intestinal peptide; VIPR1, vasoactive intestinal peptide receptor 1.

Partial evidence exists for the therapeutic benefits of targeting adrenergic signalling using β-blockers — which are β-adrenergic receptor antagonists — in cancer patients. Retrospective epidemiological studies in patients with melanoma¹¹⁶, breast^{117 ,118} or prostate cancer^119,120 found that those patients taking non-selective β-blockers for other conditions experienced better survival rates¹²⁰. Despite the benefits observed in patients who were taking these agents before the cancer diagnosis, use of β-blockers after diagnosis did not impact survival in patients with colorectal¹²¹, melanoma^122,123, lung¹²⁴, breast^125-127, ovarian¹²⁸ or prostate cancer^125,129. Although single studies^130,131 have reported a protective survival effect with β-blockers, a meta-analysis that synthesized most of these studies did not find a definitive survival benefit associated with β-blocker use before or after diagnosis¹³². The retrospective design of these studies and methodological limitations, including inadequate adjustment for confounding factors and the influence of immortal time bias, have contributed to the inconsistent findings across these investigations¹³³.

Despite significant research, inconsistent data on the relationship between β-blocker use and clinical outcomes across various cancers may explain why these relatively safe and cost-effective agents have not fully transitioned into clinical practice. Considering the diverse impacts of adrenergic signalling on various cancer types and various TME cellular components, the definitive answer about the benefit of β-blockers in cancer therapy should come from the multiple clinical trials testing the β-blockers in combination with other oncological treatments that are currently under way (NCT03384836, NCT06145074, NCT04848519, among many others). It is noteworthy that phase III clinical trials, which typically involve a larger and more genetically diverse patient population, have the potential to reveal unforeseen hereditary factors that may not have been apparent in the earlier phases of the study. These different effects might be attributed to the upstream roles of adrenergic receptor signalling (such as stress response pathway and diverse pro- and anti-tumorigenic responses to adrenergic signalling by the different cellular components in the TME) but also to differences in drug metabolism and signal transduction downstream. Thus, genetic polymorphisms in adrenergic receptor genes found in patients of European descent were associated with an increased risk of developing hepatocellular carcinoma (HCC) in patients using β-blockers. These genetic polymorphisms were absent in an Asian population, in which β-blockers were not associated with increased HCC risk¹³⁴.

A crucial confounder that drives the differential response to adrenergic blockade is the diverse influence of adrenergic signalling on immune cells, as the α- and β-adrenergic receptors are expressed in immune cells ranging from neutrophils¹³⁵ and monocytes¹³⁶ to natural killer cells¹³⁷, and targeting the adrenergic system can alter the immune cell state, leading to immune activation or exhaustion^138,139. As monotherapies, α₂-adrenergic receptor agonists demonstrated robust antitumour activity in multiple immunocompetent mouse models, including in models of immune checkpoint blockade (ICB)-resistant cancer and in mice transplanted with human cancer cell xenografts and reconstituted with human lymphocytes¹³⁹. In these models, the α₂-adrenergic receptor agonist clonidine was found to enhance the ability of macrophages to activate T cells, thereby boosting the antitumour immune response^139. Noradrenaline can also directly modify the T cell phenotype through β-adrenergic receptor¹³⁸. Exhausted CD8⁺ T cells showed upregulation of the ADRB1 gene and were often distributed near sympathetic nerves. The exhaustion state was found to be induced when ADRB1-expressing CD8⁺ T cells were exposed to catecholamines. Impairment of β₁-adrenergic signalling with β-adrenergic receptor antagonists limited the progression to an exhausted state and improved CD8⁺ T cell effector functions when combined with ICB in melanoma treatment. Furthermore, in a pancreatic cancer mouse model resistant to ICB, combining β-blockers with ICB enhanced CD8⁺ T cell responses and promoted the development of tissue-resident memory-like T cells¹³⁸.

Much evidence points to a pro-tumour interaction between the adrenergic system and immune response, particularly regarding β-blockers. Clinical evidence supports this by linking adrenergic tumour innervation to poorer survival rates, regional lymph node spread and a pro-metastatic phenotype in HNSCC⁹³, breast¹¹⁵ and prostate cancers^94,114,140. Furthermore, in vivo stress models that use subthermal housing conditions to induce stress have demonstrated that β-adrenergic signalling undermined the effector phenotype in CD8⁺ T cells, compromising the effectiveness of checkpoint inhibitor therapy¹⁴¹. Clinical studies have indicated that β-blockers when administered as a perioperative stress reduction strategy sensitized melanoma and lung tumours to immunotherapy^142,143. Although the favourable outcomes observed in these clinical trials suggest that this intervention may hold promise as a potential adjunct therapy, this effect was not observed in other studies^132,144-146, indicating that further research is needed to understand the underlying conditions in which β-blockers can be effective. Nonetheless, the findings highlight a crucial link between adrenergic signalling and the antitumour immune response; these nuanced interactions between the adrenergic system and various immune cells, are cancer specific. Understanding these cancer-specific effects can guide us on how to harness these signalling pathways to augment the efficacy of cancer immunotherapy or overcome resistance to therapy.

Cholinergic neurons

Cholinergic (parasympathetic) neurons release ACh as the primary neurotransmitter. ACh binds to two classes of receptor, the nAChRs and the mAChRs¹⁴⁷. Both classes are extensively expressed in various tumours (Figs. 3 and 4), having roles in neo-angiogenesis, cell proliferation, migration and cell survival¹⁴⁸. A fine-tuned approach to manipulation of cholinergic levels may derive from activation of these specific classes of AChR localized in cancer cells^95,99. Muscarinic signalling in cancer cells via M1, M3 and M5 mAChRs (also known as CHRM1, CHRM3 and CHRM5) resulted in increased cancer cell viability¹⁴⁹ and motility and, downstream of that, increased metastasis rates in HNSCC¹⁵⁰, lung¹⁵¹, gastric¹⁵², brain¹⁵³ and prostate¹⁵⁴ cancers. These effects were reversed by administering atropine, a known muscarinic antagonist¹⁵⁵. In high-risk prostate adenocarcinomas, CHRM1 expression levels are increased, and blocking this receptor through pharmacological means or genetic methods, hindered tumour invasion and metastasis induced by cholinergic-signalling in vivo¹⁴⁰. The nAChRs α7 and α9 have been found in lung, colon, breast and bladder cancer cells^156,157. Direct activation of these nAChRs induces a calcium-triggered autocrine release of growth factors, which fosters cancer cell proliferation and forestalls apoptosis¹⁵⁸. Activation of α7 nAChR by nicotine stimulates Ras–ERK–MAPK and JAK2 pathways, driving cancer cell proliferation and migration^158,159. This upstream regulation by α7 nAChR of these canonical pathways, which are known to drive and promote the progression of various cancers, provides a strong rationale for a combinatorial approach to target these pathways together with the neuronal signalling cascades that activate them, especially in patients whose cancer does not respond to MAPK or JAK2 inhibitors¹⁶⁰.

Fig. 4 ∣. Expression of neuronal-related genes across different cell types in the tumour microenvironment (TME).

The expression variability of neuronal-related genes between tumours can be determined by their anatomical distribution, but also by the cellular composition of the TME. Single-cell RNA sequencing data obtained from samples of colorectal cancer (part a) and head and neck cancer (part b)^{249 ,250} demonstrate how neuronal-related genes are expressed not only in tumour cells but also in other constituents of the TME. ACHE, acetylcholinesterase; ADRA2A, adrenoceptor α2A; ADRB1, adrenoceptor β1; BDNF, brain-derived neurotrophic factor; CHRNA6, cholinergic receptor nicotinic α6 subunit; DDC, dopa decarboxylase; NTRK2, neurotrophic receptor tyrosine kinase 2; GABRE, γ-aminobutyric acid type A receptor subunit ε; NGF, nerve growth factor; SLC29A4, solute carrier family 29 member 4.

Several studies investigated parasympathetic nerve density as a surrogate biomarker indicating cholinergic levels in the TME (Box 1). Analysis of breast cancer biopsy samples indicated a direct relationship between decreased parasympathetic nerve density and worse clinical outcomes¹¹⁵. In prostate cancer, however, poor survival rates were associated with heightened parasympathetic nerve fibre densities within the TME¹⁶¹. These tumour-specific effects raise questions regarding the source of ACh in these tumours. Interestingly, it was shown that lung¹⁶², gastric¹⁶³ and liver¹⁶⁴ cancer cells can produce ACh^162-164. These ACh-secreting cancer cells have been found to respond to ACh, creating an autocrine loop that promotes a continuous proliferative stimulus¹⁶⁵. Of note, immune cells, specifically T cells, also produce and secrete ACh¹⁶⁶, suggesting that cholinergic signalling may impact various cell types in the TME. For instance, the density of parasympathetic nerves was associated with higher expression of immune checkpoint receptors. In vivo, genetic ablation of parasympathetic innervation led to decreased PD1 and PDL1 expression on tumour-infiltrating lymphocytes, a finding that mirrors the unfavourable prognosis seen in patients with reduced parasympathetic nerve density¹¹⁵.

Cholinergic signalling was also studied in cancer prevention settings. In vitro findings indicated that anaesthetics with anticholinergic properties can inhibit breast cancer tumorigenesis¹⁶⁷, and reduce NETosis and matrix metalloproteinase 3 (MMP3), both of which drive cancer recurrence¹⁶⁸. These findings suggest significant cholinergic signalling in the cancer cells, highlighting its potential as a therapeutic target. Yet, the data indicate a differential, cancer-specific effect; this should be taken into consideration in clinical trial design.

GABAergic signalling

GABA is the main inhibitory neurotransmitter in the CNS, operating through ionotropic GABA_A receptor (GABRA1), GABA_A-ρ receptor (previously known as GABA_C receptor) and the metabotropic GABA_B receptor (GABBR)¹⁶⁹. GABA influence extends beyond the CNS and significantly regulates cancer cell growth, metastasis and antitumour immune responses in extracranial tumours^169-171. Several types of cancer cells, including prostate, breast and pancreatic cancer cells, exhibit upregulated GABA_A receptors. Tumours positive for GABA_A are generally more aggressive, and activation of this receptor stimulates cancer cell proliferation^172-174. By contrast, GABA_B receptors are downregulated in liver and pancreatic cancer cells^174,175. Activation of GABA_B receptors can impede cancer cell growth¹⁷⁶. Baclofen, a GABA_B receptor agonist often prescribed for muscle spasms, can diminish HCC proliferation¹⁷⁴. Two key enzymes involved in GABA metabolism are GAD67, associated with GABA synthesis, and GABA transaminase, linked to GABA degradation. Elevated expression of GAD67 is associated with advanced tumour status and poor outcomes in various solid cancers¹⁷⁷. By contrast, GABA transaminase levels are reduced in breast, liver and kidney cancer cells^76,178,179. Like other neuro-immune interactions discussed above, recent studies illuminate the role of GABA_A receptor in regulating the tumour immune microenvironment¹⁸⁰. In a colon cancer mouse model, B cell-derived GABA stimulated monocytes to differentiate into anti-inflammatory macrophages that released IL-10 and restrained CD8⁺ T cell killer functions¹⁸⁰.

Somatosensory signalling

Pain, a cardinal feature of several types of cancer, is attributed to sensory neurons, which populate many solid tumours and have been linked to a worse clinical prognosis^{6,27 ,181}. Pain is a response to stimuli in the peripheral sensory nociceptive nerves that can be formed by unmyelinated (C) or myelinated (Aδ) fibres. Nociceptors primarily use glutamate as their neurotransmitter, but neuropeptides such as substance P and calcitonin gene-related peptide (CGRP) also have a crucial role in communicating with the CNS and surrounding peripheral cells. Nociceptive neurons react to tissue injuries and interact with their environment^15,21,145. Cancer- or immune-derived signals, such as growth factors¹⁸², cytokines (IL-1β, IL-6, IL-10, MIP, TNF)¹⁸³ and immunoglobulins¹⁸⁴, also sensitize nociceptor neurons^90,185. This cancer-induced activation is characterized by a hyperexcitable state as well as the release of various neuropeptides — such as substance P and CGRP — that: support cancer cell proliferation^27,185,186; suppress tumour-associated pro-inflammatory cytokines^106,181; induce neovascularization^{187 ,188}; promote metastasis²⁷; modulate antigen flow in the lymphatics^166,189; or drive immune escape and exhaustion^181,185.

The complex interplay between sensory neurotransmitters and cancer progression is increasingly being recognized. In melanoma, sensory neurons in the TME can bolster tumour growth while inhibiting effective antitumour immune responses¹⁸¹. Ablation of these sensory neurons increased the formation of tertiary lymph nodes, thereby enhancing antitumour immunity¹⁹⁰. Studies using syngeneic mouse models of malignant melanoma demonstrated that skin cancer cells engage nociceptors, enhancing neurite outgrowth, responsiveness to harmful ligands and neuropeptide release¹⁹¹. Consequently, CGRP released by tumour-associated neurons led to augmented exhaustion of cytotoxic T cells, restricting their ability to eradicate melanoma cells. Pharmacological intervention, either by blockade of neuropeptide release from tumour-associated nociceptors or by blockade of the CGRP receptor RAMP1 using antagonists, mitigated the exhaustion of tumour-infiltrating CD8⁺ T cells. These strategies collectively reduced tumour growth and nearly tripled mouse survival rates¹⁸¹. Importantly, an analysis of human melanoma single-cell RNA sequencing datasets revealed that RAMP1-expressing CD8⁺ T cells exhibited greater exhaustion, and the heightened expression of RAMP1 in these CD8⁺ T cells was linked to a diminished responsiveness to ICB²⁷.

In addition, increased CGRP production by nerves is evident in low-glucose environments typical of oral mucosa carcinomas and melanoma. In these tumours, cancer cells secrete NGF in response to nutrient starvation, which stimulates CGRP release by nociceptive neurons. In turn, the increased levels of CGRP lead to the stimulation of cytoprotective autophagy in cancer cells, a survival pathway. This pathway can be disrupted by blocking neurogenic CGRP with a FDA-approved antimigraine medication, rimegepant¹⁹². Notably, in germline Calca (encoding CGRP) knockout mice, HNSCC tumours were significantly smaller, and there was an enhanced presence of tumour-infiltrating immune cells, including CD4⁺ T cells, cytotoxic CD8⁺ T cells and natural killer (NK) cells, indicating a crucial connection between sensory neurons and cancer therapies¹⁸⁵. In summary, CGRP seems to influence cancer cell metabolism in nutrientdeficient environments, shape immune cell recruitment to tumours and contribute to the exhaustion of cytotoxic CD8⁺ T cells. Several FDA-approved drugs that target the RAMP1–CGRP axis, including those currently used to treat migraines¹⁹³, might be effective in counteracting immunomodulatory effects of CGRP on cytotoxic CD8⁺ T cells, offering a promising therapeutic strategy to enhance immunosurveillance.

Similarly, substance P can drive T cell exhaustion¹⁹⁴ and directly regulate cell growth and apoptosis via MAPK activation. Elevated substance P levels have been correlated with the initiation and progression of melanoma¹⁹⁵, breast cancer¹⁹⁶ and glioma¹⁹⁷. The primary receptor for substance P, NK1R, is expressed by various human cancer cell lines. Aprepitant, a selective NK1R anti-nausea antagonist, can overcome the electrical substance P–neurons–cancer functional circuits that are formed in lung cancer¹⁹⁸ and HNSCC to induce tumour growth inhibition²⁷. Interestingly, the efficacy of NK1R inhibition as a tumoursuppressive agent is contingent upon the characteristics of the TME. In preclinical models of metastatic breast cancer, NK1R inhibition demonstrated a significant reduction in metastatic spread in highly inflammatory models characterized by elevated leukocyte-secreted levels of IL-6 and TNF. Conversely, in animals with cold, less-inflammatory, tumours, NK1R inhibition increases the rate of metastasis. These findings suggest that the effectiveness of NK1R inhibition depends on the tumour capacity to induce an inflammatory response and underlines the crucial role of the TME in regulating the effectiveness of oncological treatments¹⁹⁹.

On a functional level, the vagus nerve, which is the primary parasympathetic nerve, also possesses sensory fibres. Surgical ligation of these neurons reduces gastric tumour recurrence by blocking Wnt signalling, which directly stops cancer stem cells from proliferating³. Chemical ablation of TRPV1⁺ nociceptor neurons, a sensory subset of the vagal neurons, protected mice from pancreatic ductal adenocarcinoma recurrence⁶. The neuropeptide vasoactive intestinal peptide (VIP), typically produced by vagal nociceptors, was observed at elevated plasma levels in patients with pancreatic ductal adenocarcinoma. Simultaneously, its receptor was more prevalent in activated T cells. Inhibition of VIP receptor (VIPR) signalling with compounds with affinity for both of its isoforms VPAC1 and VPAC2, boosted antitumour immunity, especially when combined with anti-PD1 anti-body treatment, hinting at potential new combination therapies²⁰⁰. These findings suggest a key role for nociceptors in shaping the tumour metabolome and immune microenvironment. These afferent neuro-immune interactions can be intercepted in conjugation with existing immunomodulatory approaches in clinical oncology to restore immune cell function, enhance immunotherapy efficacy and overcome resistance.

Neuromodulatory strategies

The different branches of the PNS contain varying subtypes of neuronal signals and receptors that are expressed at different levels in different tumour types and within specific cancer-associated cell types (Fig. 3). These differences could be due to unique neurogenic cues, TME cellular composition (for example, immunogenic and non-immunogenic tumours) or location-specific neuronal landscape (the different neuronal phenotypes observed in a specific tissue). However, these differences dictate the sensitivity of the tumour and TME to neuronal cues. Characterization of the neuronal signals involved in each cancer type is still lacking but clinical oncology databases such as The Cancer Genome Atlas (TCGA) can help to identify putative actionable neural targets for a specific disease site that are associated with biological and clinical phenotypes (such as tumour growth inhibition or therapeutic resistance), as well as the compounds to target these pathways. Although foundational signatures such as ‘catecholamine uptake’ and ‘axonal growth cone’ are globally prevalent in most tumours, signatures such as the ‘GABAergic regulation of synaptic transmission’ are more variable between tumour types and even within the same tumour group. Because innervation patterns vary by tumour type, TCGA gene expression data show variable enrichment of neuronal-related gene signatures (Fig. 3a) without indicating the specific cellular source of these transcripts. Additional cell types in the TME also show diverse gene expression patterns in different tumours. For instance, T cells showed high NTRK2 expression in colorectal cancer, but the levels of adrenergic (ADRB1) and cholinergic (CHRNA6) receptors are elevated in HNSCC (Fig. 4). These data suggest different neuronal signal dependencies in the TMEs of different tumour locations.

Drugs that can modify physiological processes at the synaptic level and influence the production, release and metabolism of neurotransmitters and their receptors are of great interest^5,93. Beyond these, neurotrophic factors that have roles in axonogenesis and neuronal reprogramming represent another class of compelling drug targets. A case in point are the ongoing clinical trials using drugs such as entrectinib, larotrectinib and cabozantinib, which target the BDNF–tropomyosin receptor kinase B pathway, as these drugs influence synaptic plasticity and axon regrowth²⁰¹. The latter is pivotal for BDNF-promoted neurogenesis, cancer cell survival and chemotherapy resistance²⁰². Furthermore, neuromodulatory strategies and the removal of specific neuronal inputs, for instance, adrenergic nerve ablation, have profound effects on the TME and represent another avenue for treatment. These effects include metabolic changes that aid immune escape^203,204, modulation of checkpoint signals²⁰⁵, altered gene expression in peripheral lymph nodes¹⁸⁹ and increased angiogenesis¹⁰⁹.

Importantly, measuring neural composition and activity in different cancers to assess the relative contribution of various neuronal signals (either electrical or chemical) to cancer cell and TME evolution and treatment response is now feasible thanks to advances in selective loss-of-function (pharmacological or genetic denervation) or gain-of-function (through designer receptor exclusively activated by designer drugs (DREADD), optogenetics) approaches targeting specific neuron subtypes⁶. Genetic tools such as CRISPR–Cas9 and cell-specific conditional gene knock-in/knockout (such as cre/lox), coupled with in vivo imaging, electrical recording and spatial quantitative techniques (such as spatial transcriptomics and high-plex protein labelling) have expanded our ability to manipulate (either ablate or activate, for example, through optogenetics) and phenotype neuron–cancer interactions and better understand the role of specific neuronal populations in tumorigenesis, cancer progression and response to therapy.

Conclusion

Our understanding of the intricate interplay between the nervous and immune systems and cancer is expanding. As we delve deeper, we recognize the strong capacity of cancer cells to hijack neuronal signals and suppress the immune system. It is now pressing to further unveil the specific interactions between various tumour types, each with unique genetic composition and abnormalities, and the nervous and immune systems. Different branches of the nervous system have been shown to fulfil a significant role in the cancer biology and clinical progression of various tumour types. However, targeting the molecules involved in the cancer–nerve crosstalk is not a clinical reality yet. As we move forwards, expanding experimental research and gaining insights into the neuronal characteristics and neurotransmitter dependencies of specific cancers are crucial. This will enable us to uncover novel pathways to develop cancer treatments that could transform clinical oncology practice. To turn our understanding of cancer–nerve interactions into real patient benefits, clinical trials focused on targeting the cancer–nerve crosstalk for cancer therapy are essential. Despite the compelling scientific evidence indicating the potential benefits of this approach for cancer care, the current number of active clinical trials investigating this avenue is insufficient. Therefore, increasing the number of such trials is urgent to expedite advancements in cancer therapeutics. It is worth noting that many molecules involved in cancer–neuron signalling have been well characterized pharmacologically^206-209. This means that many existing compounds, including clinically approved drugs, can already be used, offering an established mechanism of action and known safety and toxicology profiles. A strategy focused on drug-repurposing trials also shortens the drug development timeline and substantially reduces the costs associated with novel drug development²¹⁰. Finally, this approach could pave the way for more accessible treatments for cancer patients globally, addressing health-care disparities. We believe that the fields of oncology and neuroscience are well positioned to take advantage of this extensive pre-existing therapeutic arsenal to fight cancer.

Glossary

Axoplasm

The fluid interior of the axon of a neuron, which contains cellular components and molecules essential for axon function.

High-plex protein labelling

An advanced technique that allows researchers to simultaneously tag and measure more than 100 different proteins in a single sample.

Immortal time bias

A potential source of error in cohort studies in which the analysis incorrectly includes a period during which the outcome of interest cannot occur for the exposed group, leading to biased results.

Myelin sheath

A fatty insulating layer that surrounds nerve fibres that enhances the speed of electrical impulse transmission along neurons.

NETosis

A type of cell death that is characterized by the release of decondensed chromatin and granular contents to the extracellular space, also known as neutrophil extracellular traps (NETs).

Neurotrophic

Describes a substance that promotes the survival, growth and function of neurons in the nervous system.

Optogenetics

An innovative neuroscience technique that uses light-sensitive proteins to control and study the activity of specific neurons in living tissue.

ROBO

A family of proteins that have a crucial role in guiding the growth and direction of axons during nervous system development.

Slit proteins

Signalling molecules that interact with ROBO receptors to regulate axon guidance and cell migration in the nervous system.

Somatosensory

Relating to the sensory system that detects and processes information about touch, temperature, body position and pain.