Therapeutic avenues for cancer neuroscience: translational frontiers and clinical opportunities

Abstract

With increasing attention on the essential roles of the tumour microenvironment in recent years, the nervous system has emerged as a novel and crucial facilitator of cancer growth. In this Review, we describe the foundational, translational, and clinical advances illustrating how nerves contribute to tumour proliferation, stress adaptation, immunomodulation, metastasis, electrical hyperactivity and seizures, and neuropathic pain. Collectively, this expanding knowledge base reveals multiple therapeutic avenues for cancer neuroscience that warrant further exploration in clinical studies. we discuss the available clinical data, including ongoing trials investigating novel agents targeting the tumour–nerve axis, and the therapeutic potential for repurposing existing neuroactive drugs as an anti-cancer approach, particularly in combination with established treatment regimens. Lastly, we discuss the clinical challenges of these treatment strategies and highlight unanswered questions and future directions in the burgeoning field of cancer neuroscience.

Introduction

Evidence of nerve involvement in tumours dates back to the 19th century, when Dr Hugh H Young first reported the presence of abundant nerve fibres throughout several peripheral tumours.¹ Dr H J Scherer described a similar perineuronal satellitosis of nerves within intracranial gliomas, noting that “the first manifestation of the neoplastic process is a collection of glioma cells about all or a great many of the nerve cells”.² These foundational observations provided preliminary evidence suggesting an integral role for nerves in malignancy, as opposed to their incidental capture within the developing tumour, and set the stage for what has become known as the field of cancer neuroscience. For more on the parallels between cancer neuroscience and the field of tumour angiogenesis, please see the appendix (p 1).

Perineural invasion, a histopathological finding introduced in 1985, that describes malignant cells surrounding, invading, or within nerves, is a prominent manifestation of tumour–nerve interactions.^3,4 Clinically, perineural invasion is a potential indication for adjuvant radiotherapy and is considered in staging and risk stratification of several malignancies, including penile cancer, head and neck cancers, oesophageal cancer, and thyroid cancer.⁵

Beyond irradiating cancer-infiltrated nerves, molecular targeting of the tumour–nerve axis is potentially poised for a major impact in clinical oncology. In contrast to previous reviews and perspectives that have extensively summarised mechanisms of tumour–nerve crosstalk learned from preclinical models,^6–13 we focus herein on providing an integrated clinical and translational framework by examining current and prospective therapeutic avenues within cancer neuroscience.

Therapeutic avenues for cancer neuroscience

A surge of preclinical and clinical evidence in the past two decades has begun to elucidate the role of the nervous system in promoting cancer growth. In light of the many mechanisms of tumour–nerve crosstalk and their accompanying clinicopathological ramifications, we have organised their translational implications into six broad avenues of intervention: inhibiting tumour growth, combating treatment resistance and tumour adaptation to stress, enhancing anti-tumour immunity, preventing dissemination and metastasis, and managing morbidity by dampening electrical hyperactivity and seizures as well as neuropathic pain (figure 1). We will highlight the potential therapeutic opportunities and challenges galvanised by preclinical and preliminary clinical data.

Figure 1: Six therapeutic avenues of cancer neuroscience.

TCA=tricyclic antidepressant. SSRI=selective serotonin reuptake inhibitor. GFRα=glial cell line-derived neurotrophic factor family receptor alpha. ADC=antibody–drug conjugate. BDNF=brain-derived neurotrophic factor. NGF=nerve growth factor.

Inhibiting tumour growth via modulation of the nervous system

Many cancers originate from glandular and ductal epithelial tissues. During normal development of these tissues, cells secrete neurotrophic factors that promote nerve recruitment, axon navigation, and neurite outgrowth.^14,15 As reviewed elsewhere,^6–13 a wealth of evidence currently suggests that expression of these neurodevelopmental and regenerative processes is aberrantly activated again during tumorigenesis, leading to a marked increase in intratumoural nerve density and subsequent tumour growth.

Multiple mechanisms have been described invoking a clear role of neuronal innervation in tumour growth across a variety of cancers, including gliomas and prostate, pancreas, and gastric cancers (table 1). These data highlight common signalling processes involving neurochemicals (eg, glutamate, norepinephrine, acetylcholine) and secreted factors (NGF, BDNF, GDNF). In addition, neurons can directly communicate with tumour cells by forming tumour–nerve synapses. These signalling mechanisms typically activate a classical oncogenic pathway that promotes tumour growth. For example, seminal work by Venkatesh and colleagues (table 1, row 12) showed that in high-grade gliomas and optic gliomas, the protein neuroligin-3 (NLGN3) is released from surrounding neurons and oligodendroglial precursor cells in response to neuronal activity, undergoes cleavage to its extracellular form by the protease ADAM10, and binds to receptors on the tumour cell surface to activate PI3K-mTOR signalling (figure 2). The net effect is tumour proliferation secondary to increased neural activity. An ADAM 10 inhibitor is now currently undergoing phase 1 testing in paediatric high-grade gliomas (table 2), and trials in adult patients with glioblastoma are planned. Besides the NLGN3–ADAM10 pathway, multiple other mechanisms have similarly delineated the role of neuronal input on promoting tumour growth across cancers (table 1). In addition to molecular inhibition, surgical denervation strategies have also been attempted (appendix p 1).

Table 1:

Preclinical and early clinical functional evidence of tumour-nerve interactions

	Tumour type	Therapeutic avenue	Mechanism and target*	Contribution to cancer
1	Prostate cancer	Tumour growth	Adrenergic signalling	Sympathetic neurogenesis and an increase in the density of β-adrenergic nerves promotes early stages of tumorigenesis (eg, low-to-high-grade prostatic intraepithelial neoplasia).16,17
2	Prostate cancer	Tumour growth	Adrenergic signalling (ADRB2)	Adrenergic nerves promote a metabolic shift in endothelial cells towards aerobic glycolysis that triggers angiogenesis and malignant growth.16
3	Pancreatic cancer	Tumour growth	Sensory neuron input	Dorsal root ganglion sensory neurons convey inflammatory signals from pancreatic cancer tumours to the CNS in a Kras-induced genetically engineered mouse model, contributing to tumour initiation and growth.18
4	Pancreatic cancer	Tumour growth	Cholinergic signalling	Cholinergic signalling suppresses pancreatic cancer growth through stimulation of muscarinic receptors.19
5	Pancreatic cancer	Tumour growth	Catecholamine signalling (ADRB2, NGF)	Catecholamines bind to ADRB2 and increase nerve density and neurotrophin secretion, creating a feedforward loop that promotes tumour development in a mouse model of pancreatic cancer.20
6	Basal cell carcinoma	Tumour growth	Sensory neuron input (Hedgehog pathway)	Cutaneous innervation activates Hedgehog signalling in stem-cell niches within the skin to promote tumorigenesis.21
7	Medulloblastoma	Tumour growth	Neurochemical signalling (mGluR4)	Activation of mGluR4 receptors prevents development of medulloblastoma in mouse models.22
8	Gastric cancer	Tumour growth	Cholinergic signalling (NGF, Wnt pathway)	Vagal innervation results in NGF secretion via acetylcholine signalling, thereby promoting gastric tumorigenesis through Wnt activation.23
9	Pancreatic cancer	Tumour growth	Protein signalling (ARTN)	The neurotrophic factor ARTN contributes to perineural invasion in pancreatic cancer.24
10	Glioma	Tumour growth	Protein signalling (BDNF)	BDNF is secreted in an activity-dependent manner and binds to the receptorTrkB, promoting glioma growth.25
11	Pancreatic cancer	Tumour growth	Protein signalling (GDNF/RET)	GDNF secretion and its interactions with its receptor RET promote perineural invasion in pancreatic cancer.26
12	Glioma	Tumour growth	Protein signalling (NLGN3, ADAM10)	The synaptic adhesion protein NLGN3 is released from neighbouring non-glioma neurons and cleaved by the protease ADAM10 in a neuron activity-dependent manner. The resultant secreted form of NLGN3 promotes growth through a PI3K-mTOR-dependent pathway.27–30
13	Multiple myeloma	Tumour growth, immune remodelling	Protein signalling (NNT-1)	NNT-1 is a neurotrophic factor that demonstrates B cell-stimulating capability and promotes growth and survival in myeloma.31
14	Multiple myeloma	Tumour growth	Protein signalling (BDNF)	BDNF promotes tumour survival in multiple myeloma through binding to the receptorTrkB, thereby activating MAPK and PI3K/Akt signalling cascades.32
15	Thyroid carcinoma (NRTC), NSCLC	Tumour growth	Neurotrophic-tropomyosin receptor kinase (NTRK) rearrangement	NTRK gene fusions can produce chimeric oncoproteins with constitutively activated kinase function (TrkA, TrkB, andTrkC), resulting in proliferation through the RAS–RAF–MAPK pathway.33,34
16	Glioma	Tumour growth, electrical hyperactivity and seizures	Direct synapse formation (AMPAR)	Nerves form direct glutamatergic synapses onto glioma cells and modulate tumour microtube-mediated invasion in gliomas.29,35
17	Glioma	Electrical hyperactivity and seizures	Glutamate signalling (AMPAR)	Gliomas release high amounts of glutamate to surrounding tissue and also exhibit glutamate-response growth via Ca2+-permeable AMPA receptors, establishing a positive feedback loop of nerve–tumour hyperactivity and tumour proliferation.29,35–37
18	Glioma	Tumour growth, electrical hyperactivity and seizures	Neurochemical signalling (GABA)	GABA receptors are present on low-grade astrocytomas and oligodendrogliomas, and GABA currents have also been evoked in glioblastoma.38,39
19	Glioma	Tumour growth, electrical hyperactivity and seizures	Neurochemical signalling (DRD4, DRD2)	Blockade of DRD4 has been shown to inhibit glioblastoma proliferation via disruption of autophagy, and blockade of DRD2 alone or in combination with EGFR blockade might have therapeutic potential.40,41
20	Glioma	Tumour growth, electrical hyperactivity and seizures	Neurochemical signalling (GABA)	Gliomas downregulate inhibitory GABA currents in the surrounding tumour microenvironment.42
21	Brain metastases	Metastasis	Neurochemical signalling (GABA)	Breast cancer brain metastases upregulate GABA receptors and GABA catabolism.43
22	Brain metastases	Metastasis, electrical hyperactivity and seizures	Direct synapse formation (NMDAR)	Breast cancer cells form astrocyte-like synapses (pseudo-tripartite synapses) to access glutamate secreted from neurons, which activates the NMDAR pathway and promotes brain colonisation.44
23	Metastatic tumours	Metastasis	Anoikis resistance (TrkB)	TrkB is a neurotrophic receptor that is shown to suppress anoikis, contributing to metastatic spread through supporting circulation of cancerous cells.45
24	Prostate cancer	Metastasis	Parasympathetic signalling (CHRM1)	Parasympathetic innervation of cholinergic fibres stimulates prostate cancer dissemination at later stages.17
25	Brain metastases	Metastasis, immune remodelling	Adrenergic signalling (ADRB2)	β2-adrenergic signalling facilitated breast cancer metastases by recruiting macrophages and inducing a pro-metastatic and immunosuppressive M2 state. Treatment with propranolol reversed this response in animals.46
26	Glioma	Tumour growth, electrical hyperactivity and seizures	Neuronal hyperexcitability (GPC3)	Subsets of glioblastoma driven by PIK3CA secrete GPC3, which drives hyperexcitability and glioma growth.47
27	Glioma	Electrical hyperactivity and seizures, stress adaptation	Microtube formation (Connexin 43 gap junctions)	Glioma cells form microtubes that facilitate direct communication with surrounding glioma and non-gliomatous cells via Connexin 43-mediated gap junctions, allowing for propagation of calcium waves that have been implicated as a cause of treatment resistance.48
28	Pancreatic cancer	Neuropathic pain	Neurotrophin signalling (NGF)	Cancer cells, fibroblasts, and immune cells can release NGF to recruit and activate sensory neurons by way of the receptor TrkA. Elevated expression of NGF–TrkA in pancreatic cancer tissues is thought to be associated with increased perineural invasion and pain.49
29	Pancreatic cancer	Neuropathic pain	Neurotrophin signalling (TRPV1)	NGF and other neurotrophins can activate the receptor TRPV1, a cation channel that facilitates nociception in sensory neurons. TRPV1 activation can trigger sensory neuronal membrane depolarisation and the release of substance P or CGRP to transmit pain signals.50
30	Glioma	Stress adaptation	Dopamine signalling	Combination treatment with the dopamine antagonist quetiapine, atorvastatin, and radiation prolonged survival in glioblastoma xenografts.51
31	Pancreatic cancer	Stress adaptation	Decreased apoptosis (MALT1, TRAF)	In-vitro co-cultures of mouse dorsal root ganglia and human pancreatic cancer cells demonstrated that pro-survival (MALT1, TRAF) genes are upregulated in neoplastic cells.52
32	Prostate cancer	Stress adaptation	Decreased apoptosis (NFκB)	Prostate cancer cells exhibiting perineural invasion demonstrate increased proliferation and decreased apoptosis through a NFκB-dependent mechanism.53
33	Ovarian cancer	Stress adaptation	Neurotrophin signalling (BDNF)	Activation of the BDNF–TrkB pathway inhibits apoptosis, and knockdown of TrkB enhances apoptosis.54
34	Pancreatic cancer	Stress adaptation, tumour growth	Axonal recruitment (NGF)	Starved pancreatic cancer cells in nutrient-poor, serine-deprived desmoplastic environments increase production of NGF, which promotes axonal recruitment as a means of L-serine acquisition. Treatment with theTrk–NGF inhibitor larotrectinib decreased pancreatic cancer tumour growth.55
35	Breast cancer	Stress adaptation, metastasis	β-adrenergic signalling	Chronically stressed mice exhibited 38-times higher frequency of distant metastases relative to unstressed mice. Propranolol blocked metastases in chronically stressed animals but had no effect on metastatic burden in unstressed mice.46
36	Melanoma	Immune remodelling	β-adrenergic signalling (ADRB2)	Concurrent β-blockade and immunotherapy is associated with improved survival in metastatic melanoma and might be mediated by the ADRB2 receptor.56,57
37	Colon cancer, melanoma, and breast cancer	Immune remodelling	β-adrenergic signalling	Addition of β2-adrenergic blockade improved local control and abscopal effects at distant metastatic sites following irradiation.58
38	HPV E6+ and E7+ tumours	Immune remodelling	β-adrenergic signalling	Propranolol improved efficacy of an anti-tumour vaccine via enhancement of CD8+ T lymphocyte tumour infiltration.59
39	Lung cancer	Immune remodelling	CD8+ T-cell infiltration	Depletion of tumour-associated macrophages in combination with anti-PD-1 immunotherapy restores cytotoxic T-cell migration and infiltration.60,61
40	Melanoma	Immune remodelling	Tumour innervation (Schwann cells)	Melanoma cells activate Schwann cells in the tumour microenvironment that promote tumour growth.62
41	Pancreatic cancer	Immune remodelling	Cholinergic signalling (acetylcholine)	Acetylcholine impairs the recruitment of CD8+ by pancreatic adenocarcinoma cells and favours T-helper-2 (Th2) cells. Subdiaphragmatic vagotomy in tumour-bearing mice was associated with increased CD8+ concentrations, elevated Th1:Th2 ratio, and improved survival.63

GABA=γ-aminobutyric-acid. HPV=human papillomavirus. NGF=nerve growth factor. NSCLC=non-small cell lung cancer. NRTC=NTRK-rearranged thyroid carcinoma. TRAF=TNF receptor associated factors.

^*Targets are provided in parentheses where available.

Figure 2: Example mechanisms of therapeutic avenues of cancer neuroscience.

NGF=nerve growth factor. DRG=dorsal root ganglia. PDAC=pancreatic ductal adenocarcinoma.

Table 2:

Select completed and ongoing clinical trials targeting the tumour–nerve axis

	Therapy	Target*	Phase	Actual or target accrual	Primary endpoint	Status	Outcome	Reference
H3K27M spinal cord glioma	ONC201	DRD2/DRD3 antagonist	Phase 2	76	Progression-free survival	Ongoing	8 of 12 patients alive at <12 months median follow-up	NCT02525692
H3K27M diffuse midline glioma	ONC201	DRD2/DRD3 antagonist	Phase 2	95	Objective response rate	Ongoing	No dose-limiting toxicity	NCT03295396
Thalamic H3K27M glioma	ONC201	DRD2/DRD3 antagonist	Phase 2	95	Objective response rate	Ongoing	26·3% 6-month progression-free survival	NCT03295396
Recurrent glioblastoma	ONC201	DRD2/DRD3 antagonist	Phase 2	17	Progression-free survival	Completed	No difference in 6-month progression-free survival	NCT02525692
Neuroendocrine tumours	ONC201	DRD2/DRD3 antagonist	Phase 2	25	Tumour response	Ongoing	56% progressed	NCT03034200
Paediatric diffuse intrinsic pontine glioma	ONC201	DRD2/DRD3 antagonist	Phase 1	130	Recommended phase 2 dose	Ongoing	Not reported	NCT03416530
Endometrial cancer	ONC201	DRD2/DRD3 antagonist	Phase 2	30	Progression-free survival	Ongoing	Not reported	NCT03485729
Acute myeloid leukaemia	ONC101	DRD2/DRD3 antagonist	Phase 1	20	Dose-limiting toxicity	Ongoing	Not reported	NCT03932643
Ovarian, fallopian tube, or primary peritoneal cancer	ONC201	DRD2/DRD3 antagonist	Phase 2	62	Dose-limiting toxicity, adverse events, objective response rate, progression-free survival	Ongoing	Not reported	NCT04055649
Recurrent and rare CNS tumours	ONC206	DRD2 antagonist	Phase 1	102	Dose-limiting toxicity	Ongoing	Not reported	NCT04541082
Paediatric high-grade glioma	INCB7839	ADAM10/17 inhibitor	Phase 1	28	Adverse events	Ongoing	Not reported	NCT04295759
Glioblastoma	Talampanel	AMPAR regulator	Phase 2	72	Overall survival	Completed	Improved overall survival compared with historical controls	NCT00943826, NCT00689221, NCT00813943, NCT00884741
Glioblastoma	Memantine, mefloquine, metformin	NMDAR regulator	Phase 1/2	85	Dose-limiting toxicity	Completed	Dose-limiting toxicity: dizziness, gastrointestinal symptoms, lymphopenia	NCT01430351
Glioblastoma	Memantine	NMDAR regulator	Phase 2	4	Overall survival	Terminated	Not reported	NCT01260467
Glioblastoma	Meclofenamic acid	NSAID64	Phase 1/2	72	Dose-limiting toxicity, progression-free survival	Ongoing	Not reported	EudraCT2021-000708-39
Glioma	Talampanel	AMPAR inhibitor	Phase 2	30	Progression-free survival	Terminated	Early termination due to futility	NCT00062504
Ovarian, primary peritoneal, or fallopian tube cancer	Propranolol, chemotherapy	β-adrenergic antagonist	Phase 1	32	Feasibility	Completed	Not reported	NCT01504126
Ovarian cancer	Propranolol	β-adrenergic antagonist	Phase 1	26	Feasibility	Completed	Improved quality of life	NCT01308944
Colorectal cancer	Propranolol, COX2 inhibitor	β-adrenergic antagonist	Phase 3	34	3-year recurrence	Completed	No difference in recurrence	NCT00888797
Breast cancer	Propranolol, COX2 inhibitor	β-adrenergic antagonist	Phase 2	38	Pro-metastatic cell markers	Completed	Decreased pro-metastatic and invasive markers	NCT00502684
Breast cancer	Propranolol, chemotherapy	β-adrenergic antagonist	Phase 2	10	Compliance	Ongoing	Not reported	NCT01847001
Breast cancer	Propranolol	β-adrenergic antagonist	Phase 2	60	Gene expression biomarkers	Completed	Reduced metastatic potential	ACTRN12615000889550
Pancreatic cancer	Propranolol etodolac	β-adrenergic antagonist/COX-2 Inhibitor	Phase 2	210	Recurrence	Ongoing	Not reported	NCT03838029
Pancreatic cancer	Bethanechol	Muscarinic agonist	Phase 1	18	Change in Ki-67	Ongoing	Not reported	NCT03572283
Cancer pain	Tanezumab	Antibody against NGF	Phase 3	156	Change in pain	Completed	Not reported	NCT00545129 NCT02609828
Cancer pain	Resiniferatoxin	Capsaicin analogue	Phase 1	45	Safety	Ongoing	Not reported	NCT00804154
Melanoma	Propranolol, pembrolizumab	β-adrenergic antagonist	Phase 1	9	Safety	Completed	No dose-limiting toxicity; objective response rate 78%	NCT03384836
Bladder cancer	Propranolol, pembrolizumab	β-adrenergic antagonist	Phase 2	25	Objective response rate	Ongoing	Not reported	NCT04848519

COX-2=cyclooxygenase 2. DRD2=dopamine receptor D2. DRD3=dopamine receptor D3. NGF=nerve growth factor. NSAID=non-steroidal anti-inflammatory drug.

^*Where multiple therapies are listed, this column relates only to drugs targeting the tumour–nerve axis.

Clinically, the mechanisms by which neurochemicals and proteins promote tumour growth represent attractive therapeutic targets. The potential for utilising existing neuroregulatory medications (eg, β-adrenergic, dopamine, and glutamate receptor modulators) as anti-cancer therapies is appealing and has been explored through a mixture of retrospective work and small prospective trials (table 2). For example, some retrospective studies have found an association between β-blocker usage (especially non-selective) and improved oncological outcomes,^65,66 whereas use of tricyclic antidepressants and selective serotonin reuptake inhibitors has been examined in preclinical and retrospective studies in gliomas^67,68 with mixed outcomes. Some of these strategies have been tested in small prospective clinical studies, although larger randomised controlled trials are needed to determine whether use of these drugs results in improved clinical outcomes (table 2).

Drugs modulating neuronal paracrine signalling, in contrast to neurotransmitters, are less commonly used, limiting the power of retrospective epidemiological studies. However, some cognate receptors (eg, GFRα1, NTRKs) and their downstream effectors (eg, RET, PI3K) have been known oncogenes for decades, with well-described genetic alterations observed across numerous cancer types.⁶⁹ Since much of this paracrine signalling converges on classic oncogenic pathways, targeting the upstream neural activators of these pathways might be effective. A 2016 study showed that a small molecule inhibitor of dopamine receptors 2/3 (DRD2/3), ONC201, counteracts the pro-survival Akt–ERK pathway.⁷⁰ ONC201 and a second generation imipridone targeting DRD2 (ONC206) are currently undergoing phase 1 and 2 testing in multiple tumour types, including H3K27M gliomas, endometrial cancer, and neuroendocrine cancers (table 2).

Although multiple prospective trials are testing novel neurochemical modulators (table 2), repurposing existing medications leverages known safety profiles and could offer an accelerated path towards adjunct use with conventional cancer therapies. This option might be especially attractive for treatment of brain tumours, because many of these neuroregulatory drugs permeate the blood–brain barrier (appendix p 2). However, one major clinical challenge to repurposing neuroregulatory drugs for anti-cancer benefit is the lack of tumour specificity and potential morbidity at effective doses. Furthermore, off-patent strategies are probably less appealing for industry-funded clinical trials. Nonetheless, the recent surge of novel data on neuronal regulation of cancer might reignite interest in this approach. Ongoing prospective trials might also shed light on tolerability and efficacy of neuroregulatory therapeutics (table 2).

Combating treatment resistance by disrupting stress-adaptive mechanisms

A pervasive feature of aggressive cancers is their ability to adapt and proliferate in stressful environments created by surgery, systemic therapy, or radiotherapy, supported by multiple components of the nervous system (table 1, rows 30–35). For example, recent work showed that pancreatic cancer cells deprived of the amino acid serine can selectively increase translation of nerve growth factor (NGF), which in turn recruits axons as an exogenous source of serine (figure 2 and table 1, row 34). Interestingly, it is estimated that approximately 40% of pancreatic tumours cannot synthesise their own serine, highlighting a subpopulation of patients that could benefit from combination of a low-serine diet and larotrectinib, a highly selective TRK inhibitor that has been approved by the US Food and Drug Administration. Such mechanisms are in line with previous predinical findings that dietary restriction of serine (and glycine) can reduce tumour burden in lymphoma and intestinal tumours,⁷¹ suggesting that this therapeutic strategy might have broad applicability to densely innervated cancer types.

Beyond serving as a source of essential nutrients, nerves can also directly induce changes in cancer and stromal cells that facilitate survival under microenvironmental stress. For example, adrenergic signalling in prostate cancer promotes a shift in cell metabolism towards aerobic glycolysis to promote angiogenesis (table 1, row 2) and the neurotrophin BDNF inihibits cancer cell apoptosis in ovarian cancer (table 1, row 33).

In the central nervous system (CNS), ultra-long membranous protrusions, known as microtubes, have been described as a mechanism of treatment resistance in gliomas. These microtubes connect cancer cells through gap junctions that protect against cell death from radiotherapy and are associated with high grade tumours and poor clinical outcomes (table 1, row 27). Furthermore, these microtubes are composed of myosin and microtubules, the latter of which might suggest a potential sensitivity to taxanes and vinca alkaloids.⁷² One of the challenges of taxane-based treatment of CNS malignancies, however, is their limited penetration ofthe blood–brain barrier.⁷³ To this end, it remains to be seen whether novel taxanes with better CNS penetration will become an effective option to combat microtube-mediated resistance.⁷⁴ There are now industry efforts (eg, Divide and Conquer) focused on small molecule inhibition of GAP43 (neuromodulin), a crucial protein for membrane tube outgrowth, as a therapeutic strategy for glioma.⁷⁵ As these efforts progress, it will be equally important to identify subsets of patients with glioma that will benefit most from treatment.

Therapeutic strategies targeting pro-survival mechanisms in the context of microenvironmental stress could hold the greatest promise for cancer types known to propagate in nutrient-scarce microenvironments (eg, pancreatic cancer, high-grade gliomas), and treatment of local relapses, where recurrent growth emerges in previously treated tissue with altered vascular and nutrient supply (table 2).

Remodelling the immune microenvironment in combination with immunotherapy

Immune checkpoint blockade has been integrated into standard-of-care management for numerous cancer types. Preclinical studies investigating the neuro–immune axis have shown multiple mechanisms through which the nervous system modulates the activity of tissue-resident as well as lymphatic immune cells.^76,77 Given the many neuroactive drugs already used for alternative indications, there might be opportunities to repurpose these drugs in combination with immune checkpoint blockade, cytotoxic therapies, or cancer vaccines to modulate the tumour immune microenvironment.

Among patients with melanoma, use of non-specific β-blockers has been associated with decreased recurrence and improved overall survival in those with metastatic disease who also received immunotherapy (table 1, row 36). As such, early phase clinical trials have begun to explore combination propranolol and pembrolizumab in locally advanced and metastatic melanoma, reporting an initial response rate of 78% (table 2). Similar studies are being conducted for recurrent or metastatic urothelial cancer (table 2).

Emerging evidence also suggests β-blockade might enhance efficacy of cytotoxic therapies through modulation of the immune response. In a preclinical study of multiple cancers, adrenergic signalling suppressed anti-tumour immunity, including the abscopal response following local irradiation (table 1, row 37). When β2-signalling was disrupted, CD8⁺ T cells exhibited enhanced cytotoxic effector function and migratory capacity. Concurrent irradiation and β-blockade also reduced the abundance of immunosuppressive M2-polarised macrophages and regulatory T cells, both of which have been shown to stymie migration of cytotoxic T cells and their physical interaction with malignant cells in lung cancer (table 1, row 39). These observations are consistent with pervasive adrenergic and sensory innervation in lymph nodes and control of T-cell egress from lymph nodes via β2-adrenergic receptor signalling (figure 2).^78,79 Given emerging evidence that a substantial fraction of the response to immune checkpoint blockade occurs outside of the tumour site (eg, draining lymphoid tissue),^80–82 modulation of the neural signalling that controls retention or release of T cells can have marked effects on the native as well as the immune checkpoint blockade-induced immune response.

Taken together, these findings could have important implications for immunologically cold tumours for which immune checkpoint inhibitors have been largely unsuccessful, in part due to the dearth of CD8⁺ T cells within the vicinity of malignant cells.⁸³ The combination of immune checkpoint blockade, radiotherapy, and β2-blockade might therefore warrant further investigation for malignancies of the pancreas, prostate, and other tumours unresponsive to checkpoint inhibition alone. Of note, prostate and pancreatic cancer are also among the most densely innervated tumours, with high rates of perineural invasion⁴ and profound sympathetic signalling (table 1, rows 1 and 5). We additionally discuss the role of cholinergic signalling, nociceptive sensory neurons, and growth factor signalling (eg, midkine) on the immune microenvironment in the appendix (p 2).

Targeting metastatic spread

Evidence on the role of nerves in promoting metastasis is only beginning to emerge. Nerves can serve not only as a physical conduit, but also as a mediator of vascular or lymphatic dissemination.⁸⁴ Multiple preclinical and clinical studies have now demonstrated associations of perineural invasion,⁸⁵ ANXA2–SEMA3D–PLXND1 signalling,⁸⁶ ADRB2-driven cell signalling (table 1, rows 1 and 5), and NMDA receptor (NMDAR) expression (table 1, row 22) with metastases. Cholinergic signalling has additionally been described as promoting invasion and metastasis in prostate cancer (table 1, row 24). Enhanced metastatic potential can result from cancer cells using nerves as physical channels for transit,⁸⁵ upregulating neural gene expression to maximise colonisation efficiency in distal neural tissue (eg, the brain; table 1, row 22), or adopting pro-invasive features secondary to dialogue with nerves.⁸⁷

The association between metastases, angiogenesis, and vascular permeability is well established⁸⁸ and is regulated by VEGF proteins.⁸⁹ However, as control of vasculature and nerves are often tightly coupled, neurotrophins (eg, NGF, BDNF) and their receptors are increasingly recognised as regulators of aberrant blood vessel formation as well.⁹⁰ In preclinical studies, treatment of a breast cancer mouse model with an anti-NGF antibody decreased liver metastases and microvascular density.⁹¹ Monoclonal antibodies against NGF (eg, fasinumab, tanezumab) are currently in development and have thus far been clinically tested in non-oncological populations.⁹²

Upon initial escape from the local microenvironment and intravasation into the systemic circulation, malignant cells must resist anoikis, a mechanism of programmed cell death due to detachment from neighbouring cells or extracellular matrix.⁹³ Neurotrophic receptors, and in particular TrkB, are crucial mediators of anoikis resistance, thereby enhancing metastatic spread (table 1, row 23). This might provide a therapeutic rationale for TrkB inhibition (eg, larotrectinib, entrectinib) in locally advanced malignancies as a way of inducing malignant cell death before distant organ colonisation.

The metastatic cascade ends with successful colonisation of a distant organ. Recent work has elucidated the role of the nervous system in breast cancer brain metastases (table 1, row 22). Breast-to-brain metastases were found to exhibit glutamate-dependent signalling through NMDAR by infiltrating synapses between neurons. Breast cancer cells co-opt normal physiological synapse formation, creating so-called pseudo-tripartite synapses to hijack synaptic glutamate from surrounding neurons and feed their own tumour growth (figure 2).⁴⁴ Glutamate then stimulates NMDAR signalling in breast cancer cells that promotes successful colonisation in the brain. Indeed, mice with knockdown of the GluN2B subunit of NMDAR exhibited significantly longer brain metastasis-free survival and lower brain metastasis burden than mice without this knockdown.⁴⁴ These findings were supported by correlative clinical data from patients with breast cancer in The Cancer Genome Atlas (table 1, row 22).

Taken together, the potential clinical translation of these findings prompts consideration of therapies targeting the tumour–nerve axis to preclude or slow metastatic spread. Preliminary clinical studies are repurposing neuroactive drugs as a strategy for slowing metastases (table 2). Consistent with preclinical evidence (table 1, rows 25 and 35), preoperative propranolol reduced biomarkers associated with metastatic potential in a triple-blind, placebo-controlled, phase 2 trial for primary resectable breast cancer (table 2). However, given that some of the discussed therapeutic rationale presented in this section is geared toward prevention of (as opposed to treatment of) metastatic spread, optimal timing of administration and combination with other therapies needs to be carefully appraised. In situations where metastatic spread has already occurred, therapeutic strategies that target mechanisms of metastatic growth (as opposed to initial seeding)—such as NMDAR regulators (eg, memantine) for patients with brain metastases from basal-like breast cancer (table 1, row 22)—might be viable options.

Disrupting electrical hyperactivity for dual control of tumour growth and seizures

Neuronal electrical activity is crucial for neurodevelopment,⁹⁴ although evidence of the importance of electrical neuronal input in facilitating cancer proliferation is also emerging, particularly among brain tumours. Analogous to development of the normal human nervous system, glioma cells are affected by electrical activity (table 1, rows 16–17 and 26–27). Gliomas, in turn, release growth stimuli, including glutamate, into the surrounding tissue, increasing neuronal activity in the microenvironment (table 1, row 17). Given the presence of calcium-permeable AM PA receptor-dependent synapses described in both paediatric and adult high-grade gliomas, glioma-derived glutamate contributes a bidirectional, positive feedback loop of tumour–nerve hyperactivity and tumour proliferation (figure 2 and table 1, row 17). However, glutamate represents only one of multiple sources of hyperactivity and glioma growth: recent work has also described the glioma-secreted factor glypican 3 found to drive synaptogenesis and hyperexcitability (table 1, row 26), as well as the ability of gliomas to downregulate inhibitory GABA currents in the surrounding electrical microenvironment (table 1, row 20).²⁹ The precise biological consequences of each source of electrical activity remain to be fully elucidated. For example, glioma–nerve electrical interactions have been described both as potassium currents (table 1, row 12)²⁹ and calcium waves (table 1, row 17). These findings are consistent with existing preclinical data suggesting that multiple mechanisms might be at play, including those involving gap junctions and microtubes (appendix p 2).

Whether electrical networking and complex neural circuits exist in cancer types outside of the CNS remains to be shown. To date, the evidence remains scant for peripheral tumours, although tumour microtubes facilitating intercellular communication have been discovered in pancreatic cancer,⁹⁵ raising the possibility of electrical processes in non-glial lineages.

The electrical integration of brain tumours with neuronal networks prompts reconsideration of some clinical manifestations of gliomas, including seizures. Although often considered a byproduct of mass effect from the tumour or surrounding oedema, seizures can directly result from glutamate-secreting tumours augmenting neural hyperactivity. Indeed, subgroups of gliomas with higher synaptogenic potential were associated with increased seizure frequency and tumour invasion,⁹⁶ and recurrent postoperative seizures have been shown to be associated with glioblastoma tumour recurrence.⁹⁷ These findings raise the question of whether anti-epileptics have oncological therapeutic potential. Although sometimes used in these patients for seizure prevention, anti-epileptics might also confer an additional anti-cancer benefit by dampening neuronal activity; however, the cumulative evidence remains indeterminate, particularly in classes of drugs that modulate postsynaptic signalling. Preliminary work has investigated use of the anti-epileptics perampanel,⁹⁸ talampanel,⁹⁹ valproic acid,¹⁰⁰ and levetiracetam100 in glioblastoma, although additional work is needed.

There is much that remains unknown regarding how electrical activity from the nervous system affects cancer growth. For instance, although preclinical evidence strongly suggests an oncogenic role of glutamate, glutamate receptor regulators have not yielded positive results in prospective clinical trials thus far (table 2). Although the exact roles of electrical input remain to be fully elucidated, current understanding of electrical networking is consistent with the overall shift in perspective that tumours are driven not only by cell-intrinsic processes but also the intimately connected components of its surrounding microenvironment.

Management of neuropathic pain by targeting the sensory nervous system

Patients with cancer frequently report intense pain at primary and metastatic sites. This has been associated with the presence of perineural invasion and is thought to result in part from the actions of neurotrophins and neuroregulatory proteins (table 1). NGF, for instance, activates sensory afferents from the dorsal root ganglia and has been most directly implicated in pancreatic cancer pain (figure 2 and table 1, rows 3 and 28–29). However, although cancer pain is often treated with the goal of symptom control as opposed to disease eradication, targeting neuropathic pain might simultaneously promote cancer control as well.

Given the morbidity of cancer pain and the fact that the upstream paracrine signalling proteins modulating pain are also known to have pro-tumour growth effects, these neuronal signalling pathways represent attractive therapeutic targets. A subset of patients with advanced cancer experience severe and refractory pain that might in part be explained by these underlying neurotrophic mechanisms, which are not targeted by commonly used analgesics such as opioids. Novel therapeutic strategies tailored to these tumour–nerve signalling dynamics thus have the potential to both control neuropathic cancer pain and confer anti-cancer benefit. Indeed, tanezumab, a monoclonal antibody against NGF, is currently being evaluated in a phase 3 study for patients with painful bone metastases (table 2). Additional ongoing strategies include targeting the TRPV1 receptor, a downstream effector of NGF that facilitates nociception in sensory neurons. Moreover, TRPV1⁺ sensory nerves promote the growth of pancreatic intraepithelial neoplasia (PanIN) organoids. Denervation of sensory neurons in mice using resiniferatoxin, a high-affinity capsaicin analogue, significantly impaired tumour progression in vivo and is currently in phase 1 testing for patients with severe or refractory pain from advanced cancer (table 2).¹⁰¹

Although cancer itself often causes neuropathic pain, so can treatments. Chemotherapy-induced peripheral neuropathy is a common iatrogenic cause of neuropathic pain most frequently associated with certain platinums, taxanes, and vinca alkaloids.¹⁰² The dorsal root ganglion lacks an efficient blood–brain barrier,¹⁰³ is prone to neurotoxic damage, and is believed to account for many of the sensory symptoms associated with chemotherapy-induced peripheral neuropathy.¹⁰⁴ Dorsal root ganglion sensory neurons also play a role in the initiation and growth of certain cancers. In mouse models, previous work revealed perineural invasion, inflammation, and neuronal damage in the nervous system as early as the PanIN-2 stage (table 1, row 3). Ablation of sensory neurons using neonatal capsaicin injections prevented perineural invasion and neuronal damage, suggesting that sensory neurons transmit inflammatory signals from oncogenic Kras-induced neoplasia to the CNS. Taken together, one could postulate that perhaps cytotoxic damage to peripheral sensory nerves has beneficial antineoplastic effects through disruption of tumour–nerve interactions but also results in morbidity through chemotherapy-induced peripheral neuropathy. This hypothesis evokes parallels with the association between immune-related adverse effects caused by immune checkpoint inhibitors and improved treatment response¹⁰⁵ and the correlation between mild-to-moderate chronic graft-versus-host disease after allogeneic haematopoietic stem-cell transplantation and lower risk of relapse.¹⁰⁶

Future therapeutic perspectives and challenges

As the tumour microenvironment has become increasingly appreciated as a major driver of cancer biology, the nervous system has emerged as a crucial component. The progress in both preclinical and clinical cancer neuroscience has increased exponentially in recent years and enhanced our mechanistic understanding of tumour–nerve crosstalk such that molecularly-targeted strategies are now poised for impactful clinical implementation. Three broad questions that remain incompletely answered include (1) whether the tumour–nerve axis can be successfully targeted with sufficient specificity for broad clinical benefit, (2) how neural regulation can synergistically integrate into existing treatment regimens, and (3) whether histopathological or other companion diagnostics can identify patients who are most likely to benefit from these therapeutic strategies.

This first question is the subject of multiple ongoing prospective clinical trials targeting components of the tumour–nerve axis (table 2). These trials also reflect emerging efforts from joint academic–industry partnerships, including companies such as Divide and Conquer and Cygnal Therapeutics, which are likely to grow as the underpinnings of tumour–nerve crosstalk continue to be illuminated.⁷⁵ In glioblastoma, for instance, therapeutic strategies are underway to inhibit GAP43, a regulator of synaptic plasticity and membrane tube outgrowth;⁷⁵ meanwhile, there are also attempts at modulating peripheral nerve signalling to cancer cells and the immune system.⁷⁵

Beyond therapeutic development, the opportunity for repurposing existing drugs that regulate the nervous system and typically have a tolerable adverse effect profile presents further opportunities. These efforts have taken advantage of both the on-target and off-target effects of these drugs. For example, antiepileptics and neurotransmitter regulators have at times shown promising effects in preclinical and retrospective clinical studies. Perhaps the most clinically vetted neuroregulatory strategy in cancer to date has been β-blockade, which has been evaluated in a range of tumour types including breast, ovarian, and colorectal. It should be noted, however, that the conclusions of retrospective studies in aggregate have been mixed, with some citing benefit and others not. At the same time, this ambiguity should be interpreted with caution, as there might be numerous reasons for a lack of clear improvement, including therapeutic window concerns, inherent limitations of retrospective analyses, and receipt of concurrent therapies in these studies. Importantly, these results also raise questions on how therapeutics targeting the tumour–nerve axis should be optimally utilised with existing treatment strategies. For example, neuroregulatory drugs might have limited efficacy as a monotherapy, but might synergise with treatments such as chemotherapy or radiotherapy. Given the known mechanistic role of nerves, including serving as nutrient sources, immune modulators, and growth stimuli, neural regulation might offer the ability to augment the response to immunotherapy, increase cytotoxic effects with concurrent radiation, or additionally target specific neural signalling pathways that are also commonly altered in cancer (table 1).

Several challenges remain in applying cancer neuroscience therapeutics to the clinic, including the need to improve our mechanistic understanding of the processes driving these pathological processes. Improvements to diagnostic and research techniques are helping to achieve this goal, such as use of single cell and spatial omics technologies. Clinically, little is known about how tumour–nerve dialogue affects and is affected by established treatments, including surgery, radiotherapy, chemotherapy, immunotherapy, and other targeted therapies. Furthermore, given that many tumour–nerve interactions invoke signalling pathways that exist in normal tissue function or development, the ability for therapies that target these interactions to exhibit tumour specificity might limit the therapeutic window. Lastly, although treatment strategies targeting the tumour–nerve axis are largely directed at cell-extrinsic processes, they probably act in concert with cell-intrinsic oncogenic events to promote cancer growth. As such, therapeutically targeting tumour–nerve interactions might be best used in combination with other local or systemic treatments. For example, the preclinical finding that adrenergic stress constrains anti-tumour immunity and abscopal responses after focal irradiation (table 1, row 37) suggests that the combination of β blockade and radiotherapy could be worth exploring further.

Effective patient selection for therapeutic targeting of the tumour-nerve axis requires identification of reliable biomarkers and companion diagnostics. Currently, the most histologically apparent feature suggesting a tumour’s direct involvement with the nervous system is perineural invasion, which has variable impact on clinical management depending on cancer type and regional paradigms. Perineural invasion itself is a poor prognostic factor across many cancers.^4,107

In practice, perineural invasion status is often integrated with other clinical and histopathological risk factors to determine adjuvant treatment selection, including radiotherapy and or chemotherapy.⁴ It should be noted, however, that although perineural invasion associates with poorer clinical characteristics in many cancers,^4,107 it remains unclear whether perineural invasion itself actively contributes to tumour pathogenesis or whether it simply demarcates an aggressive subset of tumours with a penchant for crosstalk with the nervous system. Furthermore, the relatively limited clinical integration of perineural invasion as a biomarker is in part due to (1) a lack of standardisation in criteria to report perineural invasion, (2) suboptimal tissue sampling and limitations of commonly used staining techniques, and (3) incomplete understanding of the biological and therapeutic implications of perineural invasion. However, akin to how PD-L1 status can guide immunotherapy use, the presence of perineural invasion might eventually serve as a predictive biomarker to guide patient selection for therapies targeting tumour–nerve crosstalk, but the utility of this strategy needs to be rigorously established. Recent efforts to improve detection sensitivity and standardisation of perineural invasion include a pan-cancer perineural invasion gene expression signature developed using machine learning-based algorithms and transcriptomic data.¹⁰⁷ Additional diagnostic developments such as an assay for tumour serine biosynthesis and content for pancreatic cancer (table 1, row 34) might also be clinically informative.

The nascent field of cancer neuroscience is growing rapidly, with parallel efforts in uncovering the mechanistic underpinnings of tumour–nerve interactions and developing novel therapies through collaborative academic and commercial efforts. Lessons learned from the tumour microenvironment suggest that disrupting tumour–nerve dialogue could eventually become a mainstay of clinical oncology, akin to anti-angiogenic and immunomodulatory therapies nowadays.

Search strategy and selection criteria.

Relevant preclinical studies were identified in PubMed with the search terms “cancer” and specific cancer sites (eg, “prostate”, “pancreatic”, “breast”) in combination with “neuroscience”, “nerve(s)”, or specific neural signalling components such as “glutamate”, “norepinephrine”, “dopamine”, “cholinergic”, and “GABA”. Relevant clinical studies were identified by searching using these same terms on ClinicalTrials.gov. No date restrictions were applied to either search; the last search of both databases was done on Nov 2, 2021. Only articles in English were reviewed. Publications corresponding to trials included in this Review were identified via search of National Clinical Trial numbers on PubMed. The final reference listforthe preclinical studies was generated on the basis of novelty, impact, and relevance to cancer neuroscience mechanisms and therapeutic avenues. The final reference list forthe clinical studies was based on sample size and relevance.