Cancer-Associated Neurogenesis and Nerve-Cancer Cross-talk

Mounting evidence suggests that cancer prognosis is associated with intratumoral neural infiltration. This phenomenon is most commonly observed in cancers that arise in highly innervated organs, including nearly all pancreatic cancers, 80% of head and neck cancers, 75% of prostate cancers, and 33% of colorectal cancers (1). Studies on patient tumor samples have revealed that intratumoral nerve density is associated with increased metastasis, morbidity, and mortality. Consistently, the presence of nerve fibers is an independent prognostic factor for overall survival in pancreatic ductal adenocarcinoma (PDAC; refs. 1–4), gastric carcinoma (5, 6), biliary tract tumors (7), head and neck cancer (8–10), and cervical cancer (11, 12) and an indicator for recurrence risk in pancreatic cancer (13), prostate cancer (11, 14), gastric cancer (15), and colorectal cancer (16–21). Tumor innervation may play an important direct role in facilitating metastasis, as tumor-associated nerves may extend into the central nervous system and cultivate premetastatic niches (19, 22). Tumor innervation may also affect patients' quality of life by causing pain, paresthesia, numbness, and paralysis. The purpose of this review is to help clinicians and researchers gain a deeper mechanistic understanding of nerve-cancer cross-talk. Of note, this nerve-cancer cross-talk is distinct from perineural invasion, the process by which cancer cells disseminate through lymphatic vessels within the perineural space (23, 24).

Despite the known impact of denervation in reducing cancer growth, investigation of nerve–cancer interaction has been slow (25–27). However, knowledge regarding nerve-cancer cross-talk has been greatly advanced by a few landmark discoveries that have elucidated three key mechanisms by which tumors regulate nerves: axonogenesis, neurogenesis, and neural reprogramming (Fig. 1A).

Following their observation that murine dorsal root ganglia form neurite outgrowths toward prostate cancer cells (28), Ayala and colleagues (29) investigated the symbiotic relationship between nerves and cancer. Through two- and three-dimensional reconstructions of entire prostates, Ayala and colleagues uncovered and confirmed cancer-related axonogenesis (the enlargement of nerves or increase in nerve density) and demonstrated an association between prostate cancer and neurogenesis (an increased number of neurons). These findings were supported by the observation that the axon-guidance molecule semaphorin 4F (S4F) is highly expressed—and may be secreted—by DU-145 cells when cocultured with neurons, and S4F induced neurite sprouting and increased neurite length in neurons compared with controls. Furthermore, a reduction of S4F via siRNA transfection decreases neurite outgrowth. Together, these findings demonstrated, for the first time, that cancer cells produce known neurotropic molecules capable of driving axonogenesis.

Around the same time as this discovery, the revelation that chronic stress promoted angiogenesis and malignant cell growth through sympathetic β-adrenergic activation (30) caused researchers to further understand the mechanistic effects of stress on tumor growth. A closer investigation of the role of sympathetic β-adrenergic activation in tumor progression led to the landmark demonstration that tumor progression is promoted by nerve stimulation (31). In this study, surgical or pharmacologic denervation of both parasympathetic and sympathetic nerves led to prostate tumor suppression in mice. Genetic deletion of sympathetic β₂- and β₃-adrenergic receptors in stromal cells also prevented early tumor progression. In contrast, parasympathetic stimulation contributed to later tumor progression, invasion, and metastasis through pharmacologic or genetic disruption of the muscarinic 1 receptor. In addition, clinical sympathetic and parasympathetic nerve densities were greatest in patient tumors and surrounding tissues, respectively, and were associated with poor clinical outcomes (32). This study suggests that both sympathetic and parasympathetic signaling may serve as potential therapeutic targets and that cancer-related neurogenesis may drive tumor progression.

In addition to studying cancer's role in axonogenesis, Ayala and colleagues and Magnon and colleagues proposed that cancer may drive neurogenesis, the outgrowth of neural progenitors to the tumor. Mauffrey and colleagues (33) expounded on this hypothesis, demonstrating that neural progenitor cells expressing the neural stem cell marker doublecortin (DCX+) migrate from neurogenic regions of the brain's subventricular zone to tumorous and metastatic niches via the bloodstream, differentiating into noradrenergic, and mature neuronal phenotypes. Furthermore, DCX+ cells were observed in greater numbers in high-risk versus low-risk human prostate cancer specimens. DCX+ cell depletion also decreased incidence of neoplastic lesions and increased tumor growth in tumors with the addition of DCX+ neural progenitor cells. Additional studies to identify the mechanisms by which DCX+ progenitor cells migrate will be important next steps to confirm the conclusion of this study, cancer-related neurogenesis.

Cancer is also capable of forming de novo neurons from cancer stem cells. Lu and colleagues (34) showed that cancer stem cells derived from patients' gastric and colorectal carcinomas were able to differentiate into both tyrosine hydroxylase (TH)-producing sympathetic and vesicular acetylcholine transporter-producing parasympathetic neurons. These neurons, in turn, were able to reciprocally communicate with cancer cells within xenografts to facilitate tumor growth. Knockdown of the neuron-generating capacity of these stem cells by MAP2 inhibited tumor xenograft growth, thereby underscoring the importance of these de novo nerves to cancer progression.

Concurrently with Mauffrey and colleagues study, Amit and colleagues (35) made a surprising discovery regarding cancer: reciprocal neural cross-talk. Specifically, the authors observed exosome-induced neural reprogramming—a phenomenon typically only observed during development—by tumors and showed that cancer-derived extracellular vesicles (EV) play a role in cancer-related axonogenesis. They discovered that genetically aberrant, p53-knockout or -mutant (p53^C176F and p53^A161S) oral cavity squamous cell carcinoma (OCSCC) cells release EVs that promote neuritogenesis in dorsal root ganglia. This release was dependent on Rab27A and Rab27B, GTPases both necessary for EV release (35, 36–39). Analysis of miRNA array revealed a decrease in the p53-deficient, cancer-derived EV packaging of miR-34a and miR-141. The knockdown or antagonism of miR-34a in p53^WT cancer cells produced similar EVs to those seen in p53-deficient cancer cells. Furthermore, Amit and colleagues were the first to report that daily intratumoral injections of p53^WT OCSCC EVs can suppress noradrenergic neurogenesis. Decreased levels of miR-34a not only promoted the neuritogenesis of sensory nerves, but also induced transdifferentiation of these nerves into noradrenaline-producing adrenergic nerves, which are commonly enriched in head and neck tumors and have been shown to promote tumor growth (38). To demonstrate that neural reprogramming, rather than outgrowth of existing adrenergic nerves, drove this process, the authors performed lingual denervation prior to OCSCC implantation in mice, and found lower intratumoral tyrosine hydroxylase-positive nerve density than sham controls. In contrast, global chemical sympathectomy before OCSCC implantation in mice did not affect tumor growth. Similarly, TH+ nerve density in patients with OCSCC was associated with lower overall survival rates and thus it may potentially serve as an independent prognostic marker. Overall, these results demonstrate a novel mechanism by which cancer cells induce nerve density and initiate adrenergic neurogenesis.

The mechanisms that drive axonogenesis through nerve-cancer cross-talk have not yet been fully elucidated. However, evidence described in this section indicates that axonogenesis is stimulated by cancer cells' release of neurotrophic growth factors and EVs containing altered miRNA levels. Reciprocally, nerves release neurotransmitters that stimulate cancer growth, possibly through either bona fide or pseudotripartite synapses (resembling a tripartite synapse, in which a synapse consisting of two neurons is closely associated with a third nonneural cell). Together, tumor-associated, stromal, endothelial, and immune cells are nurtured and primed for angiogenesis and inflammation (Fig. 1B).

EVs play an important role in establishing a tumor microenvironment that can positively regulate tumor initiation and metastasis. EVs, including exosomes, which are assembled and released through a transcriptionally driven process, are released from nearly every cell type and carry complex cargos, consisting of DNA, RNA, miRNA, transfer RNA, long noncoding RNA, lipids, and proteins, across long distances (38–40). As described below, cancer-derived exosomes carry different cargo than do noncancer-derived exosomes and signal inflammation, angiogenesis, and axonogenesis.

Exosomes extracted from human head and neck cancer samples and from murine oropharyngeal squamous cell carcinomas induce increased neurite outgrowth in PC12 cells (a rat pheochromocytoma cell line) compared with exosomes derived from control plasma and tissue. In contrast, the pharmacologic exosome inhibitor GW4869 and genomic knockout of Rab27A and Rab27B, genes that control the exosome secretion pathway, decrease neurite outgrowth (40). High concentrations of at least one exosome cargo, the axon-guidance molecule ephrin B1, drive axonogenesis, also seen in cervical cancer cell lines (41). In addition, neuritogenic effects have been seen when PC12 cells treated with cancer-derived exosomes were used to recruit sensory nerves to the tumor microenvironment. These results demonstrate that cancer-derived EVs promote tumor innervation.

It is likely that additional signals packaged in exosomes drive axonogenesis. For example, cancer commonly presents with aberrant regulation of gene expression by miRNAs, which bind to mRNA to reduce its translation or increase its degradation. This binding to mRNA regulates many cellular pathways and transcription factors, such as those involved in proliferation, differentiation, and apoptosis, all of which may be corrupted in cancer. Thus, the genomic aberrations, epigenetic changes, or mutations of miRNA that affect its processing or activity may nurture a microenvironment conducive to tumorigenesis or metastasis (42–44). As such, miRNAs can be classified as oncoMIRs or tumor-suppressor miRNAs (45, 46). In support of this concept, Amit and colleagues revealed that a combination of stimulatory miR-21 and miR-324 with scramble miRNA increased axonogenesis more than did a cocktail including inhibitory miR-34a (35). This suggests the presence of cancer-orchestrated signaling to other cells in the tumor microenvironment through specific exosome packing and release of miRNAs.

In addition, the promotion of tumor progression through tumor-derived EVs and dysregulated miRNA has been demonstrated by a number of groups studying various cancer types. In lung cancer, low levels of miR-100-5p have been shown to increase levels of mTOR (47). In breast cancer, modified EVs and miR-23a, miR-222, miR-452, and miR-24 alter their respective targets—Sprouty2, PTEN, APC4, and p27—to confer drug resistance (48, 49). Tumor-derived EVs can also provide nutrients to the tumor and mediate tumor-stem cell and tumor-progenitor cell communication (50–52). Thus, there are abundant opportunities to investigate the roles of other molecules within tumor-derived EVs in regard to nerve-cancer cross-talk.

Cancer cells increase secretion of neurogenic factors promoting axonogenesis, while nerves increase expression of the complementary receptors. Deregulation of NGF, which is responsible for the survival, differentiation, and neurite outgrowth of neurons, has been implicated in a number of cancer types that express the tyrosine-receptor kinases (Trk) TrkA, TrkB, and TrkC and the p75 neurotrophin receptor (53).

These receptors are also expressed on nerves, and p75 neurotrophin receptor has recently been reported to act as a chemoattractant for cancer cells (54). Indeed, cancer-derived NGF drives neurite growth and cancer proliferation and migration and is correlated with nerve-cancer cross-talk (1, 23, 55–63). One notable study demonstrated an association between expanded enteric nerves and increased NGF expression in gastric cancer and the NGF/Trk signaling regulation of Dclk1+ tuft-cell coordinated cross-talk between nerves and gastric cancer (64–66). Migration of pancreatic cancer cells toward dorsal root ganglia was reduced when NGF signaling was blocked through NGF knockdown or NGF-neutralizing antibodies (67, 68). Similarly, in breast cancer cells, there is a correlation between nerve fibers and NGF expression, and breast cancer cells are capable of driving axonogenesis in PC12 cells, a process partly reversed by blocking NGF (57).

The expression not only of neurogenic factors, but also their cognate receptors, influences nerve-cancer cross-talk (23). NGF's receptor, nerve growth factor receptor (NGFR), is expressed across multiple cancer types; for example, it is expressed in luminal breast cancer in rare, basal-like cells resistant to antiestrogens (58). NGFR inhibits p53 activity within tumor cells in a negative feedback loop across multiple tumor types. This process is central to maintaining melanoma stem cells in vitro and melanoma growth in vivo (59). Through it, NGF signaling from nerves via NGFR expression on cancer stem cells may drive cancer stem cell renewal and proliferation.

Like NGFR, L1 cell adhesion molecule (L1CAM), a surface receptor central to proper cell adhesion and migration during neural development, is highly upregulated in several different tumor types. It promotes cancer cells' motility and invasiveness, leading to cancer metastasis and chemoresistance and radioresistance (60, 61). Recently, it has been found that L1CAM expression, in conjunction with CD133 expression, defines a cancer stem cell population in glioma and ovarian cancer. In in vivo ovarian cancer models, L1CAM expression elicited cancer stemness in several ways, for example, through the enhancement of spherogenicity, the tumor take rate, cancer cells' self-renewal capacity, and tumor growth (61).

Like NGF, brain-derived neurotrophic factor is induced by noradrenergic signaling to stimulate axonogenesis through Trk receptors (69, 70) and is associated with the promotion of angiogenesis and increased tumor cell proliferation (71). Another NGF, NT-3, is overexpressed in PDAC and its nerves, and NT-3 inhibition led to decreased growth of PDAC in a murine xenograft model (63, 64, 72).

Glial cell–derived neurotrophic factor (GDNF) is robustly expressed in human PDAC and is significantly correlated with neural invasion, and associated with an increase in pain levels (73–75). The cancer-promoting effects of GDNF, including neural invasion, are likely mediated through GDNF receptors including GDNF family receptors α1–3 and through RET, initiating downstream activation of the RAS/ERK, MAPK, JNK, and PI3K/Akt signaling pathways (74, 75). GDNF is also known to promote integrin expression, activate matrix metalloproteinase (MMP)-9, and increase NFκB, which plays a significant role in nerve adhesion and invasion (76–79).

Neurturin and artemin also signal through RET and promote innervation (80). They are also highly expressed in PDAC and are associated with tumor invasiveness and nerve alteration, respectively (3, 23, 81, 82). Artemin has been found primarily in the hypertrophic nerves of PDAC tissue samples, as determined through Western blotting and IHC (83).

Finally, when upregulated, the axon-guidance molecule netrin-1 acts as an oncogene in a number of cancer types (84, 85) and promotes gastric cancer cell navigation along sensory dorsal root ganglia cells and sciatic nerve invasion in vitro (86). In contrast, in PDAC, restoration of pathologically decreased expression of the axon-guidance molecule, Slit2, to normal levels reduces metastasis and neural invasion (87). While these studies reveal promising therapeutic targets, further investigation will be important to translate their feasibility in clinical application.

Communication between nerves and brain cancer cells has recently been characterized as taking place via either bona fide chemical synapses, as in glioblastomas, or via pseudotripartite synapses as in breast-to-brain metastases (B2BM). Gene expression of synaptic markers has been found in primary glioma cells (88). Neuron–glioma interaction via bona fide chemical synapses have electrophysiologic properties similar to those seen in α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) signaling (88, 89). Spontaneous excitatory postsynaptic currents from AMPA receptor–expressing glioma cells demonstrates a functional neuroglioma synapse (89). It was also determined that these synapses do not result from a direct connection to neurons. Rather, tumor microtubules are found in glioma tissue and facilitate gap-junction signal transmission (88, 90). In addition, the optogenetic stimulation of glioma cells promote glioma growth and proliferation. Furthermore, it has been demonstrated that activated glioma increases neuronal hyperexcitability in the tumor microenvironment in a reciprocal interaction (88).

Similarly, activation of glutamate ligand receptors, N-methyl-D-aspartate receptors (NMDAR), facilitates breast cancer metastasis to the brain (22). Previous reports have demonstrated tumor growth via NMDARs in neuroendocrine and ductal pancreatic cancers, and this growth is thought to result from autocrine secretion. Breast cancer transcriptional signatures have also implicated NMDARs in metastasis and NMDAR subunits have high expression levels in B2BM cell lines. It was also discovered that the extracellular glutamate in the B2BM tumor microenvironment was not sufficient to substantiate autocrine signaling. B2BM cells express adhesion molecules, such as postsynaptic density protein 95, that form pseudosynapses between nonneuronal cells and axons and contribute to astrocytic synaptogenesis. Nondisruptive B2BM processes establish pseudotripartite synapses to access glutamate. The knockdown of NMDAR subunits resulted in fewer brain tumors and increased subject brain metastasis-free survival but did not affect their primary tumor burdens or lung metastases, which were rescued with the reexpression of NMDAR subunits. Together, these finding describe how tumors can coopt neural signaling in the brain to promote tumor progression.

Neurotransmitters are signaling molecules that allow neurons to communicate with other neurons or cells. Because various cancers and components of their stroma express corresponding neurotransmitter receptors, the induction of a protumor microenvironment can be supported by the peripheral nervous system. Autonomic nervous system responses promote tumor progression. Sympathetic stimulation is involved in the early stages of carcinogenesis and angiogenesis, while parasympathetic stimulation promotes invasion and metastasis. The sympathetic nervous system is also involved in immunomodulation and cancer-associated neuropathic pain.

Sympathetic nerve innervation has been demonstrated in human tumors and murine tumor models. Increased levels of noradrenaline have been observed in solid tumors and implicated in stress-mediated tumor progression. Furthermore, noradrenergic receptors are widely distributed in various cancer types, and modulating them has known effects on tumorigenesis and progression, described below.

The growth of a new vascular network, angiogenesis, is marked by proangiogenic molecules (such as IL8, TNFα, VEGF, TGFα, TGFβ, angiogenin, platelet-derived growth factor, and FGF), the levels of which indicate tumor aggression and are important factors in prognostic outcomes (90–93). Peripheral neurons participate in vascular organization during development and wound repair and contribute to tumor angiogenesis (94–96).

The catecholamines noradrenaline and dopamine have opposing roles in angiogenesis; respectively, they stimulate and inhibit vascular networks. Noradrenaline stimulates angiogenesis by signaling increased VEGF expression in tumor-associated macrophages in primary mammary tissues and amplifying the expression of cytokines known to stimulate angiogenesis (34, 97–99). Noradrenergic signaling in a β₂-adrenergic and β₃-adrenergic knockout in a xenogeneic orthotopic prostate cancer model in immunodeficient mice resulted in reduced tumor-associated vascular density (100). Furthermore, β₂-adrenergic knockout in a spontaneous transgenic prostate cancer mouse model also exhibited decreased vascular density, migration, and branching compared to that seen in controls. Dopamine, on the other hand, downregulates the VEGFR -2–mediated signaling pathway, diminishing proliferation and migration in colon cancer cell lines and impairing tumor growth in mouse models of gastric cancer and ovarian cancer (97, 101–105). These studies indicate that tumor innervation promotes angiogenesis and neovascularization of the tumor microenvironment.

As previously mentioned, surgical and pharmacologic sympathectomy decreases prostate tumor growth and lung metastasis (33, 106). Protumorigenic properties have been observed in Rv1 and LNCaP prostate cancer cell lines that express α_1A-adrenergic receptors and in PC3 and DU-145 cells lines that express α_1B- and α_1D-adrenergic receptors. Blocking β₂- and β₃-adrenergic signaling was found to arrest tumor growth and angiogenesis, and it was later shown that β₂-adrenergic signaling activates the angiogenic switch by inhibiting the expression of mitochondrial cytochrome c oxidase assembly factor 6, which decreases oxidative phosphorylation (aerobic glycolysis; ref. 100). It has also been shown that noradrenergic signaling indirectly stimulates angiogenesis by stimulating M2 macrophages to secrete VEGF, which was inhibited in mice with propranolol (106). Together, these findings suggest that stress-induced sympathetic signaling drives angiogenesis and tumor progression.

Like dense, intratumoral sympathetic nerves, enhanced adrenergic signaling may drive primary or secondary resistance to cytotoxic chemotherapies. Eng and colleagues (107) noted that pancreatic cancer biology and treatment responses in their mouse colonies were heavily dependent on temperature, and that the sympathetic cold stress response enhanced noradrenaline and activated β-adrenergic receptors, driving resistance to cisplatin and paclitaxel (108). In head and neck cancer cells treated with cisplatin, the half-maximal inhibitory concentration of cisplatin was strikingly increased in oral cancer cells cocultured with neurons compared with oral cancer cells cultured alone. In addition, oral cells treated with a β2-adrenergic receptor blocker (ICI 118,551) demonstrated significantly diminished cell viability 48 hours after treatment with cisplatin (unpublished data). Similarly, blockading of β2-adrenergic receptor (ADRB2) and NGF pathways improved gemcitabine efficacy in KrasG12D p53R172H; Pdx1-Cre pancreatic cancer models (72). In a recent study by Chen and colleagues (108), signaling through upregulated β2 receptors on cervical cancer cells led to upregulation of Sirt1; inactivation of its target, p53; and downregulation of p53 target genes. As a result, these tumor cells were resistant to doxorubicin-induced p53 acetylation. Additional work needs to be done to fully determine how nerves influence cancer response and resistance to chemotherapy and other treatment modalities.

In addition to stimulating tumors, parasympathetic signals can also suppress tumor progression. In gastric, prostate, and breast cancers, parasympathetic signals serve as specific markers for tumors, and the corresponding receptors are expressed in gastric, pancreatic, lung, cervical, and colon cancer cells (70, 109–111).

Vagotomies performed in patients with gastric cancer and mice with pancreatic cancer have increased tumor progression (91, 112). However, the activation of muscarinic cholinergic receptor 1 (Chrm1) reduces tumor incidence, perturbs cancer cell signaling pathways through the suppression of EGFR/MAPK and PI3K/AKT, and suppresses cancer stem cells (113–116). Cholinergic stimulation also prevents colorectal cancer progression by inducing trefoil factor 2 secretion by memory T cells and suppresses breast cancer by reducing PD-1 expression levels in CD4⁺ and CD8⁺ lymphocytes (117, 118). Furthermore, cholinergic deprivation increases macrophage influx and TNFα production, promoting cancer progression (106). These studies suggest a contribution of parasympathetic signaling in tumor suppression.

Conversely, cholinergic promotion of tumorigenesis and metastasis has been demonstrated in prostate and gastric cancers, respectively (32, 91). Chrm1 expression in prostate cancer stromal cells is essential for metastasis (34, 70). Other studies have indicated that Chrm3 mechanisms drive the protumor effects of the parasympathetic nervous system. Chrm3 activation in gastric cancer induces Wnt–β-catenin signaling downstream of YAP (69, 91, 116). The Wnt–β-catenin signals expand cancer stem cells and induce NGF's promotion of nerve innervation in gastric cancer (91, 116). In small-cell lung carcinoma, Chrm3 activates MAPK/Akt signaling (114). In colon cancer, Chrm3 activates ErbB receptors downstream of the MMP-7 adhesion molecules that promote tumor cell invasion (115). Converse to the previous section, these findings suggest parasympathetic signaling can promote tumor progression, depending on the type and stage of tumor.

Stress, as manifested by enhanced sympathetic signaling and glucocorticoid production, contributes to many disorders, both benign and malignant. Although acute bouts of stress may enhance a CD4⁺ T cell– and B cell–mediated immune response, chronic repeated stress diminishes the immune response and renders important immune effectors anergic. Because many innate and adaptive immune cell types express or upregulate neural receptors including adrenergic receptors, it follows that immune cells are sensitive to neural signaling during oncogenesis.

We are now beginning to understand that the neuroregulation of inflammation plays a critical role in cancer, as inflammatory changes in nerves are observed in the early stages of some cancers (106, 119). In breast cancer, CD11b⁺F4/80+ macrophages have been shown to infiltrate the tumor parenchyma upon pharmacologic sympathetic activation of adrenergic receptors and to contribute to a 30-fold increase in metastasis to the lymph nodes and lung. The β-adrenergic receptor blocker propranolol reversed this effect (119). CD11b⁺F4/80+ macrophages also secrete higher levels of GDNF in pancreatic adenocarcinoma cells compared with resting endoneurial macrophages, and a CCR2-deficient model of perineural invasion of cancer cells with reduced recruitment and activation of tumor-associated macrophages showed reduced CD11b⁺ F4/80+ macrophages and nerve invasion (74). Similarly, neuropathy of the sympathetic nervous system induced leukemic bone marrow infiltration in an acute myelogenous leukemia mouse model through the β₂-adrenergic receptor expressed in stromal cells (120). In another study, splenic vagal denervation suppressed cytotoxic T cells and promoted carcinogenesis (118). It has also been shown that noradrenaline stimulates IL6 production and activates macrophages and other stromal cells in the tumor microenvironment (94). Macrophages are recruited by cancer cells through cancer cell–secreted colony stimulating factor and release GDNF to promote cancer migration and nerve invasion (74). Consistently, β₃-adrenergic signaling induces ovarian cancer cells to secrete brain-derived neurotrophic factor, which ultimately signals axonogenesis and plays a role in switching macrophages and neutrophils into immunocompetent M1 and N1 types and in recruiting and maintaining hemopoietic and mesenchymal stem cells (110, 111).

Nerves can also interact with immune cells in an immune tolerizing fashion, allowing for tumor evasion and escape. Sensory neurons have been shown to attract myeloid-derived suppressor cells in melanoma, promoting an immune-tolerant, protumor milieu (121). Within prostate cancer, regions rich in autonomic tumor-infiltrating nerves express high levels of the immune checkpoint protein programmed cell death ligand-1 (PD-L1). This inhibitory immune checkpoint ligand, which binds and promotes anergy of cytotoxic CD8⁺ T cells expressing the cognate receptor, PD-1. Regions high in PD-L1–expressing nerves were inversely correlated with those expressing CD8⁺ T cells, and high density of PD-L1 nerves was associated with recurrence (122). Also supporting cross-talk between the autonomic nervous system and T-cell function, genetic parasympathetic stimulation of tumor decreased PD-L1 expression in tumors and PD-1 expression in T cells, and enhanced CD8⁺/Treg (regulatory T cell) ratios (109). In an analysis of breast cancer specimens, sympathetic nerve density was associated both with high expression of immune checkpoints and poor prognosis. Ultimately, a connection between nerve cross-talk and immune effectors may have important ramifications for the future of immune checkpoint blockade and other immunotherapies, but warrants further investigation.

Although advances in cancer treatment have improved patient outcomes, we still have much to learn about how tumors interact with their microenvironment. We are now beginning to realize that, in some solid tumors, infiltrating nerves and catecholaminergic signaling may play an important role in tumor initiation and progression. A deeper understanding of the mechanistic basis of cancer progression, specifically in regard to nerve-cancer cross-talk, will reveal new therapeutic targets and allow the repurposing of existing treatments, including neuromodulatory therapies to slow or stop cancer progression, or to be used in conjunction with chemotherapies or immunotherapies. As a greater understanding of nerve-cancer crosstalk and the neuro-immune axis emerges, new antineurogenic targets hold tremendous potential as novel opportunities for treating cancer (Table 1).

M. Amit reports grants from NIH/NCI during the conduct of the study. No disclosures were reported by the other authors.

The authors would like to thank Eyal Kimhi for artistic work and the University of Texas MD Anderson Cancer Center's Scientific Editing Services and Research Medical Library for assistance in preparing this article.

Silverman's training and projects are supported by NIH/NCI grant F30 CA228258 (to D.A. Silverman), the Cullen Endowment for Higher Education, the John J. Kopchick Fellowship, and the American Legion Auxiliary Fellowship in Cancer Research. G.A. Calin is the Felix L. Haas Endowed Professor in Basic Science. Work in G.A. Calin's laboratory is supported by NIH/NCATS grant UH3TR00943-01 (to G.A. Calin) through the NIH Common Fund, Office of Strategic Coordination (OSC), the NCI grants 1R01 CA182905-01 and 1R01CA222007-01A1 (to G.A. Calin), an NIGMS 1R01GM122775-01 grant (to G.A. Calin), a Team DOD (CA160445P1 to G.A. Calin) grant, a Chronic Lymphocytic Leukemia Moonshot Flagship project, a CLL Global Research Foundation 2019 grant, a CLL Global Research Foundation 2020 grant, and donor support through Jaffer Ajani and the Estate of C.G. Johnson. M. Amit's work and laboratory is supported by the MDACC Head and Neck Cancer Moonshot and NIH/NCI R37 CA242006-01A1, to M. Amit.