Research Article

Autonomic Nerve Development Contributes to Prostate Cancer Progression

See allHide authors and affiliations

Science  12 Jul 2013:
Vol. 341, Issue 6142, 1236361
DOI: 10.1126/science.1236361

Structured Abstract


Cancer cells usurp the healthy tissue microenvironment to promote their survival, proliferation, and dissemination. The role of angiogenesis, the formation of new blood vessels, in solid tumor growth is well established. Whether neurogenesis, the formation of new nerve fibers, likewise contributes to tumor development and progression remains unclear. Here, studying mouse models and human tumor samples, we examined the role of the autonomic nervous system in prostate cancer growth and dissemination.

Embedded Image

The parasympathetic nervous system promotes prostate cancer dissemination in mice. Image shows bone metastases (arrowheads) detected by Na18F-PET scanning of Hi-Myc mice, a model of prostate cancer. Such metastases are not detected when the Hi-Myc mice are genetically deficient in muscarinic cholinergic receptor type 1, an essential signaling component of the parasympathetic branch of the autonomic nervous system.


To track tumor growth and dissemination, we studied (i) mice bearing PC-3 prostate tumor xenografts that expressed luciferase and (ii) transgenic mice expressing the c-Myc oncogene under the control of the probasin promoter (Hi-Myc mice), which develop prostatic intraepithelial neoplasia that progresses to invasive adenocarcinoma. Tumors were monitored by bioluminescence, positron emission tomography (PET), and histological analyses. Sympathetic (adrenergic) and parasympathetic (cholinergic) nerve functions were assessed using chemical or surgical neural ablation, pharmacological agonists or antagonists, and genetically engineered mice. We also determined the adrenergic and cholinergic nerve densities in radical prostatectomy tissues from a cohort of 43 patients with prostate cancer.


Quantitative bioluminescence and immunofluorescence analyses, combined with histological examinations, revealed that sympathetic adrenergic nerve outgrowth was critical in the early phases of cancer development. Prostate tumor xenografts developed poorly in mice that had been pretreated by chemical or surgical sympathectomy of the prostate gland, or when stromal β2- and β3-adrenergic receptors were genetically deleted. Prostate tumors were also infiltrated by parasympathetic cholinergic fibers that promoted cancer dissemination. Cholinergic-induced tumor invasion and metastasis in mice were inhibited by pharmacological blockade or genetic disruption of the stromal type 1 muscarinic receptor. Quantitative confocal microscopy analysis of radical prostatectomy specimens from patients with low-risk (n = 30) or high-risk (n = 13) prostate adenocarcinoma revealed higher overall nerve densities in high-risk tumors relative to low-risk tumors. Adrenergic fibers were increased in normal prostate tissues surrounding the human tumors, whereas cholinergic fibers infiltrated the tumor tissue. Higher densities of adrenergic and cholinergic nerve fibers were associated with poor clinical outcome, including higher preoperative levels of prostate-specific antigen (PSA), extension beyond the prostatic capsule, and biochemical recurrence.


These results suggest that the formation of new nerve fibers within and around prostate tumors can alter tumor behavior. The autonomic nervous system appears to exert dual functions in prostate cancer: Sympathetic neonerves promote early stages of tumorigenesis, whereas parasympathetic nerve fibers promote cancer dissemination. Conceivably, drugs targeting both branches of the autonomic nervous system could provide therapeutic benefit.

Cancer Hits a Nerve

Solid tumors sculpt their microenvironment to maximize their growth and metastatic potential. This concept is illustrated most famously by tumor angiogenesis, a process whereby tumors induce the growth of new blood vessels to boost their supply of oxygen and blood-borne nutrients. Magnon et al. (p. 10.1126/science.1236361; see the Perspective by Isaacs) now highlight the important contribution made by another microenvironmental component—developing autonomic nerve fibers—to tumor growth and metastasis. In mouse models of prostate cancer, surgical or chemical destruction of sympathetic nerves prevented early-stage growth of tumors, whereas pharmacological inhibition of parasympathetic nerves inhibited tumor dissemination. In a small study of human prostate cancer specimens, the presence of a high density of nerve fibers in and around the tumor tissue was found to correlate with poor clinical outcome. These results raise the possibility that drugs targeting the autonomic nervous system may have therapeutic potential for prostate cancer.


Nerves are a common feature of the microenvironment, but their role in tumor growth and progression remains unclear. We found that the formation of autonomic nerve fibers in the prostate gland regulates prostate cancer development and dissemination in mouse models. The early phases of tumor development were prevented by chemical or surgical sympathectomy and by genetic deletion of stromal β2- and β3-adrenergic receptors. Tumors were also infiltrated by parasympathetic cholinergic fibers that promoted cancer dissemination. Cholinergic-induced tumor invasion and metastasis were inhibited by pharmacological blockade or genetic disruption of the stromal type 1 muscarinic receptor, leading to improved survival of the mice. A retrospective blinded analysis of prostate adenocarcinoma specimens from 43 patients revealed that the densities of sympathetic and parasympathetic nerve fibers in tumor and surrounding normal tissue, respectively, were associated with poor clinical outcomes. These findings may lead to novel therapeutic approaches for prostate cancer.

Several lines of evidence have linked the nervous system to tumor growth and progression. Migration of tumor cells along nerves—a process termed perineural invasion—correlates with poor prognosis in certain epithelial cancers, including prostate cancer (13). In vitro coculture experiments have shown that sensory neurons from the dorsal root ganglion promote prostate cancer cell proliferation (4). Work with in vivo model systems suggests that neural stimulation may increase tumor incidence and the number of metastases (58). In addition, recent retrospective clinical data suggest that breast, melanoma, or prostate cancer patients taking β-blockers have lower recurrence rates and mortality (913).

Whether nerve fibers infiltrate tumors and alter their behavior has not been examined in detail. The prostate stroma is abundantly innervated by the sympathetic and parasympathetic branches of the autonomic nervous system, controlling growth and maintenance of the prostate gland (14, 15) (fig. S1). Here, we tested the hypothesis that along with neoangiogenesis (16), prostate tumors are infiltrated by autonomic neonerves that affect cancer development and dissemination.


Altered Prostate Cancer Development After Sympathetic Nerve Ablation

To develop a mouse model of prostate cancer that would allow longitudinal monitoring of tumor growth and metastasis by bioluminescence—a noninvasive imaging method correlating with tumor volume in vivo (17)—we injected PC-3 human prostate cells stably expressing luciferase (PC-3luc) into the ventral prostate of immunodeficient nude Balb/c (nu/nu) mice (Fig. 1, A and B). After 11 weeks of tumor development, a time sufficient to allow nerve development and tumor cell–prostate microenvironment interactions, we examined sections of intracapsular tumor and surrounding healthy prostate tissues by histology. This analysis revealed tumor-infiltrating sympathetic fibers [adrenergic, identified by tyrosine hydroxylase (TH) staining] arising from normal prostate tissue, as well as intratumor parasympathetic fibers [cholinergic, identified by vesicular acetylcholine transporter (VAChT) staining; Fig. 1C and fig. S2]. We confirmed the neural specificity of TH and VAChT immunofluorescence analyses by coexpression of the neuron-specific cytoskeletal subunits of neurofilament-L (NF-L) or neurofilament-H (NF-H) (Fig. 1D), which mark newly formed and mature nerve fibers, respectively, in the fetal rat (18). Tumor tissues contained significantly more NF-L than NF-H staining (fig. S3), and many fibers expressed both neural markers with sprouting NF-L+ neurites (Fig. 1D and fig. S3). These results suggest that the tumor recruits newly formed nerves in the stroma.

Fig. 1 Sympathetic nervous system (SNS) controls tumor engraftment in mice.

(A) Examples of in vivo bioluminescence imaging of Balb/c nu/nu males mice 11 weeks after injection of PC-3luc cells into the ventral prostate without development of metastasis (left) and with distant metastases (right). (B) Top left: Ex vivo imaging of the prostate tumor. Top right: Representative H&E-stained section showing the intracapsular tumor surrounded by healthy prostate tissues. Bottom, left to right: Metastases within the intestine, lung, and liver. (C) Representative immunofluorescence staining of TH (left) and VAChT (right). (D) NF-L tumor nerves in green costained with NF-H (DAPI, blue; NF-H, red). Note the developing NF-L+ branches (arrows) arising from a double-positive fiber. (E) Serial quantification of bioluminescence intensities in tumors within the prostate, weeks 1 to 11, after xenografting in mice denervated with 6OHDA (n = 8) and in a PBS-treated group (n = 11). Inset shows quantification of pelvic lymph node invasion and metastases at week 11. (F) Real-time quantification of bioluminescence of PC-3luc tumor cells injected into prostate glands denervated of hypogastric nerves (HGNx; n = 5) or sham-operated (n = 8). (G) Serial in vivo bioluminescence analyses evaluating the growth of PC-3luc cells in the prostate of Adrβ2−/− (n = 10), Adrβ3−/− (n = 10), and Adrβ2−/−Adrβ3−/− (n = 9) nu/nu mice compared to Adrβ2+/−Adrβ3+/− and Adrβ2/+/+Adrβ3+/+ nu/nu controls (n = 24) of the same background. (H) Ex vivo quantification of bioluminescence in PC-3luc prostate tumors (left) and in lymph nodes and distant metastases (right) of the same mice shown in (G). *P < 0.05, **P < 0.01. Scale bars, 10 μm. Error bars indicate SE.

To assess the functional role of the sympathetic nervous system (SNS), we ablated adrenergic nerves by injecting 6-hydroxydopamine (6OHDA) into the tumor-bearing mice. 6OHDA treatment specifically destroyed TH+ neural fibers located in the basal neural layer underneath the prostate epithelium, without affecting VAChT+ parasympathetic fibers surrounding epithelial cells (fig. S4, A and B). 6OHDA-induced sympathectomy prevented the development of tumors in the prostate, suggesting a critical role for sympathetic neural activity in tumor engraftment (Fig. 1E). In control experiments, we found that 6OHDA was not toxic to tumor cells in vitro (fig. S4C), but chemically sympathectomized mice exhibited increased rates of epithelial cell apoptosis in the tumor-free prostate (fig. S4, D and E). Whereas distant metastases were detected 11 weeks after injection of PC-3luc tumor cells in control mice, no detectable metastasis was observed in sympathectomized mice, likely because of the impaired tumor development at the orthotopic site (Fig. 1E, inset). To confirm these results and to ascertain whether sympathetic signals were locally delivered in the tumor microenvironment, we surgically cut the hypogastric nerves, which carry selectively sympathetic fibers into the prostate gland, prior to orthotopic injection of tumor cells (Fig. 1F and fig. S5). Surgical denervation markedly inhibited tumor development, whereas tumors in sham-operated nerve-intact mice grew exponentially from week 4. These results suggest that SNS signals are critical at early stages of prostate tumor development in this xenogeneic mouse model.

Recent studies have revealed that the SNS, a major pathway for stress-induced signals, enhances tumor growth through the β2-adrenergic receptor (Adrβ2) expressed on tumor cells (7, 19) and also regulates the hematopoietic stem cell niche in the bone marrow via Adrβ2 and Adrβ3 expressed in the stroma (20, 21). On the basis of possible parallels between the behavior of the hematopoietic and cancer stem cell niches, we investigated the role of adrenergic signals in tumor engraftment by crossing nu/nu mice with mice that were genetically deficient in the β2-, β3-, or both β2- and β3-adrenergic receptors. Whereas tumor development in the prostate was slightly delayed in mice lacking a single adrenergic receptor, it was severely compromised in Adrβ2−/−Adrβ3−/− mice (Fig. 1G). In addition, tumor cell dissemination into the lymph nodes and distant organs was significantly reduced in the double knockouts (Fig. 1H). Although β-blockers have been suggested to inhibit metastases by antagonizing Adrβ2 expressed by tumor cells (19), we speculate that the reduced tumor volume at the orthotopic site may explain the reduction in metastasis in the Adrβ2−/−Adrβ3−/− mice. Control experiments revealed that these observations were not specific to PC-3luc cells, as tumor development and dissemination were also impaired when LNCaP human prostate cancer cells were injected into the Adrβ2−/−Adrβ3−/− mice (fig. S6). Thus, expression of both the Adrβ2 and Adrβ3 receptors in the microenvironment appears to be critical for tumor development in this mouse model.

To explore the role of sympathetic innervation in tumorigenesis in a genetic model of prostate cancer, we evaluated the effect of chemical or surgical sympathectomy on tumor progression in Hi-Myc transgenic mice (22). These mice selectively overexpress cMyc in the prostate (under the control of the probasin promoter), which leads to the complete penetrance of mouse prostatic intraepithelial neoplasia (PIN) from postnatal week 2, progressing to invasive adenocarcinomas within 3 to 6 months (Fig. 2, A and B). Relative to vehicle-treated control mice, the incidence of PIN was reduced by 83% in mice that had been treated as 2-day-old neonates by chemical sympathectomy (P < 0.001; Fig. 2C). Similar results were observed after chemical or surgical sympathectomy of young adult (1-month-old) mice, although the reduction in PIN incidence was only ~25%. Sympathectomy had no effect on tumorigenesis when it was performed on Hi-Myc mice that were 2 months of age or older (Fig. 2C). Tumor invasion was also not affected by ablation of the SNS (Fig. 2D). Consistent with a role of adrenergic signals in PIN formation, immunostaining of prostate sections from tumor-bearing Hi-Myc mice revealed that adrenergic fibers develop predominantly around healthy or PIN acini, with fewer sympathetic nerves near the tumor (fig. S7). One month after surgical sympathectomy of 1-month-old mice, we observed a marked increase in the number of apoptotic neoplastic epithelial cells in PIN (Fig. 2E). Together, these data support the notion that adrenergic signals play an important role in prostate tumorigenesis.

Fig. 2 SNS nerve ablation inhibits prostate tumorigenesis in Hi-Myc transgenic mice.

(A) H&E-stained prostate sections from 4-month-old Hi-Myc transgenic mice illustrating (left to right) normal prostate acinus, mouse prostate intraepithelial neoplasia (PIN) delineated by the fibromuscular stroma, and invasive cancer zone. (B) Timeline for cMyc-induced tumor progression and experimental protocols. Hi-Myc mice were chemically sympathectomized at day 2 or at 1 month after birth, following the neonate protocol, with low or high dose of 6OHDA or the adult1 protocol, respectively, and killed at 2 months of age. For adult2 and adult3 protocols, mice were chemically or surgically sympathectomized at 2 or 5 months after birth, respectively, and killed 1 month after denervation (i.e., 3 or 6 months after birth, respectively). (C and D) Effect of systemic (6OHDA, n = 19 neonates, n = 26 adults) or local HGNx (n = 11) denervation of the SNS on the prevalence of PIN (C) or invasive cancer zones (D) in Hi-Myc transgenic mice. Denervation of 2-day-old neonates or at 1 month after birth, but not later, significantly reduced the percentage of PIN. Data were analyzed from 10 sections per animal. (E) Left: Quantification of apoptotic TUNEL+ cells in HGNx mice versus controls 1 month after surgery. Right: Illustrative TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) of apoptotic cells in PIN from a surgically sympathectomized mouse. Data were obtained from five fields per animal (n = 6). *P < 0.05, ***P < 0.001. Scale bars, 10 μm. Error bars indicate SE.

Regulation of Tumor Invasion by Cholinergic Parasympathetic Signaling

Because we detected the infiltration of VAChT+ fibers from the parasympathetic nervous system (PNS) in a subset of prostate tumors from animals that exhibited metastases, we next evaluated the possible role of the PNS in cancer progression. Postganglionic PNS neurons activate muscarinic cholinergic receptors on the effector organ (23). We thus profiled the expression of the five known muscarinic receptor genes (24) in the mouse prostate gland and human prostate cancer cell lines. PC-3 cells largely express Chrm3 (cholinergic receptor, muscarinic 3), whereas Chrm5 was the predominant receptor expressed in DU145 and LNCaP cells (fig. S8 and table S1). Acetylcholine was previously reported to induce proliferation of prostate cancer cell lines in vitro via Chrm3 (25). In contrast, consistent with prior studies within human prostate tissue (26, 27), we found that Chrm1 was expressed at high levels in the healthy prostate gland of mice (fig. S8).

Accordingly, we explored whether parasympathetic activity in tumors might affect tumor progression through Chrm1 expressed in the prostate tumor microenvironment. To investigate this, we orthotopically treated PC-3luc tumor-bearing mice with carbamoylcholine chloride (carbachol), a nonselective Chrm agonist, and tested the role of cognate muscarinic receptors with pharmacological antagonists. Carbachol treatment significantly enhanced tumor cell invasion of pelvic lymph nodes that drain the prostate gland (Fig. 3A and fig. S9). Local tumor cell dissemination was mediated by a muscarinic receptor, because this phenomenon was inhibited by a nonselective muscarinic antagonist (scopolamine) (Fig. 3A). Treatment with pirenzepine, a Chrm1-specific antagonist, also inhibited lymph node invasion (Fig. 3A). Cardiovascular hemodynamics were not significantly altered in mice treated with carbachol (table S2), suggesting that the effects on tumor behavior were not due to nonspecific cardiovascular alterations. Tumors from carbachol-treated mice exhibited higher proliferation (Ki-67+) indexes, but in vitro treatment of PC-3luc cells with carbachol did not stimulate proliferation, which suggests that the effect was not tumor cell–autonomous (fig. S10).

Fig. 3 Parasympathetic nervous system (PNS) controls prostate tumor invasion in mice.

(A) Ex vivo quantification of bioluminescence of xenografts in the prostate gland (black) or lymph nodes (red) at week 5 with muscarinic receptor agonist [carbachol (Carb)] or antagonists [scopolamine (Sco) or pirenzepine (PZP)]; n = 4 to 9 Balb/c nu/nu males per group. (B) Quantification of bioluminescence of xenografts in the prostate gland (black) or lymph nodes (red) of carbachol-treated nu/nu Chrm1−/− mice (n = 9) compared to nu/nu Chrm1+/− and nu/nu Chrm1+/+ control littermates (n = 11). (C) Serial in vivo analyses of PC-3luc cell engraftment and growth in nu/nu Chrm1−/− mice (n = 7) and nu/nu Chrm1+/− or nu/nu Chrm1+/+ (n = 7) controls. *P < 0.05; error bars indicate SE.

Contribution of Stromal Chrm1 in Tumor Invasion and Metastasis

To explore whether tumor cholinergic signals were mediated by stromal Chrm1 expression, we crossed Chrm1−/− mice with nu/nu mice and implanted PC-3luc cells into the prostate of the nu/nu Chrm1−/− and nu/nu Chrm1+/+ progeny. Carbachol-induced tumor cell spreading to lymph nodes was markedly reduced when the prostate microenvironment was deprived of Chrm1 (Fig. 3B). Deficiency in Chrm1 signaling did not affect tumor growth at the implantation site within the prostate (Fig. 3C). This, combined with the increased proliferation of tumor cells in the prostate (fig. S10B), suggests that Chrm1 may affect the dissemination of proliferative tumor cells.

To evaluate the effect of cholinergic agonists on prostate cancer dissemination, we assessed Hi-Myc transgenic mice deficient or sufficient in Chrm1 expression (Fig. 4, A and B). Carbachol treatment of 3-month-old transgenic mice significantly increased the incidence of PIN (Fig. 4C, left) and accelerated the progression of these neoplastic lesions to invasive carcinoma 1 month later (Fig. 4C, right). In addition, carbachol treatment significantly increased the tumor proliferative index (Fig. 4D). Treatment of Hi-Myc mice with pirenzepine or genetic deletion of Chrm1 completely inhibited carbachol-induced malignant progression (Fig. 4, C and D). To confirm the role of stromal Chrm1 expression in prostate cancer progression, we implanted c-Myc+ Chrm1−/− prostate acini into the dorsal lobe of healthy Chrm1+/+ nude mice. In the resulting chimeric prostate, only the donor prostate tissue could develop cancer and only the recipient tissue could respond to Chrm1-mediated signals (Fig. 4E, left box). Treatment of engrafted mice with carbachol switched the tumor behavior to an invasive phenotype with stromal Chrm1-mediated disruption of the basement membrane and proliferation of tumor epithelial cells (Fig. 4E, top row). Thus, these studies demonstrate that cholinergic signals transduced in the tumor stroma by the type 1 muscarinic receptor promote prostate cancer invasion in two mouse models of prostate cancer.

Fig. 4 Chrm1 promotes prostate cancer progression in Hi-Myc transgenic mice.

(A) Timeline for Hi-Myc–induced tumor progression and therapeutic schedule. (B) H&E-stained section of the prostate from an animal treated by carbachol; boxed area shows higher magnification of the invasive cancer zone. (C) Percentage of neoplastic acini (left) and number of invasive cancer zones per prostate section (right) after treatment by carbachol [treatment 2 from (A)], pharmacological inhibition of Chrm1 [treatment 5 from (A)], or genetic disruption of Chrm1 in the cMyc+ Chrm1−/− mice. (D) Quantification of Ki-67 staining in treatment groups 1, 2, and 5. (E) Histological analyses of prostate chimeric tissues. cMyc+ Chrm1−/− prostate tissues were grafted in the dorsal lobe of nu/nu Chrm1+/+ prostate glands (left). Tissues from mice treated with carbachol (top row) or carbachol + PZP (bottom row) were harvested 10 weeks after surgery. H&E-stained sections of cMyc+ Chrm1−/− grafts show invasive cancer areas at late (1) or early (2) stage of development in the carbachol-treated group. PIN lesions were maintained when Chrm1 was blocked by PZP prior to carbachol injection (4, 5). Immunofluorescence staining for laminin-α2 (light blue) shows disruption of basement membranes (2, 3) and Ki-67+ (red) cancer cells (3) in carbachol-treated chimeras relative to the intact basement membrane in the PZP + carbachol–treated group (5, 6) (DAPI, dark blue). **P < 0.01, ***P < 0.001. Scale bars, 100 μm [(B) and (E)]. Error bars indicate SE.

Ex vivo bioluminescence quantification of metastases to distant organs (see Fig. 1B) also revealed a factor of ~6 increase in tumor cell dissemination in carbachol-treated mice relative to the control group (Fig. 5, A to C). Metastases were reduced in mice in which type 1 muscarinic receptor function was pharmacologically or genetically ablated (Fig. 5, A to C), leading to improved survival of the animals (Fig. 5, D and E). We confirmed these results in older (18- to 24-month-old) Hi-Myc transgenic mice. Positron emission tomography (PET) imaging using [18F]fluoro-2-deoxy-d-glucose (18FDG) showed significantly more spontaneous soft tissue metastases in c-Myc+ Chrm1+/+ mice relative to c-Myc+ Chrm1−/− animals (Fig. 6, A and B). By contrast, FDG uptake in primary tumor sites did not differ between the two strains (Fig. 6, A and B). Moreover, bone metastases traced by [18F]sodium fluoride (Na18F) were detected in c-Myc+ Chrm1+/+ mice but not in Hi-Myc mice lacking Chrm1 (Fig. 6, C to E). These data suggest that muscarinic signals promote cancer metastasis.

Fig. 5 Cholinergic signal contributes to prostate cancer metastasis in xenografted mice through the type 1 muscarinic receptor Chrm1.

(A and B) Ex vivo quantification of bioluminescence from distant organs at week 5 after selective pharmacologic (A) or genetic (B) inactivation of Chrm1; n = 7 mice per group. (C) Representative ex vivo bioluminescence imaging of intestinal metastases induced by carbachol with (right) or without (left) Chrm1 inhibition. (D) Kaplan-Meier curves depicting the survival of metastatic nu/nu Chrm1+/+ mice treated with carbachol (n = 14) by comparison to nonmetastatic carbachol-treated nu/nu Chrm1−/− animals (n = 6; P = 0.0005, log-rank test) or nu/nu Chrm1+/+ mice treated with PZP + carbachol (n = 9; P = 0.03, log-rank test). (E) Representative images of mice from (D) at different time points. *P < 0.05. Error bars indicate SE.

Fig. 6 Chrm1 controls prostate cancer metastasis in Hi-Myc transgenic mice.

(A) Tumor volume (left) and maximum standard uptake value (SUVmax) (right) of 18FDG+ tumors in the prostate (P) or metastases (M) to lung or in paraspinal lymph nodes obtained by PET scanning of 18- to 24-month-old cMyc+ Chrm1+/+ or cMyc+ Chrm1−/− mice; n = 5 to 10 mice per group. (B) 18FDG+ images of a cMyc+ Chrm1+/+ cancer-bearing mouse (left) with abnormal soft-tissue pelvic uptake (dotted area 2) and lung uptake (dotted area 1, with necropsy photograph confirming spontaneous lung metastases and matched H&E-stained section with boxed area at higher magnification). Note the absence of 18FDG+ spots in the lung or pelvis of a control cMyc Chrm1+/+ wild-type mouse (right). (C) SUVmax (top) and number of bone metastases (bottom) in spine quantified by Na18F-PET scanning. (D) Representative images of a bone metastatic cMyc+ Chrm1+/+ mouse (left) injected with Na18F; boxed area shows higher magnification of a front view of thorax (right, dotted red box). (E) H&E-stained section of T12/L1 vertebrae (top) of the boxed area 2 shown in (D); higher magnification (bottom) shows prostate tumor metastasis in bone confirmed by consecutive pan-cytokeratin (panCK)–stained section of epithelial cells in green (DAPI, blue). *P < 0.05, ****P < 0.0001. Scale bars, 50 μm for H&E and fluorescence images, 10 mm [(B) and (D), left] and 1 mm [(D), right] per interval for PET images. Error bars indicate SE.

Autonomic Neural Network in Human Prostate Cancer

To evaluate the relevance of autonomic innervation in human cancer, we retrospectively analyzed nerve fiber densities in prostatectomy tissues from 43 treatment-naïve patients with prostate cancer. These patients had low-risk cancer [prostate-specific antigen (PSA) level <10 ng/ml, Gleason score <7, and disease stage T1c or T2a; n = 30] or high-risk cancer [PSA level ≥10 ng/ml, or Gleason score ≥7, or disease stage ≥T2b; n = 13] (28, 29), and they had been treated at the VA Medical Center in Durham, North Carolina (Fig. 7 and table S3). Quantification of nerve fiber densities (TH+, VAChT+, NF-L+, and NF-H+) was performed in a blinded fashion without prior knowledge of clinical or pathological stage and clinical outcome. Nerve-specific staining for NF-L or NF-H revealed increased fiber densities within tumor areas, and also in normal prostate tissues surrounding the cancer in high-risk patients (fig. S11). Further individual assessment of each branch of the autonomic nervous system showed that TH+ adrenergic fibers densely innervated normal prostate tissues surrounding the tumor (Fig. 7, A to C), whereas VAChT+ cholinergic fibers were largely restricted to the tumor (Fig. 7, D to F). In addition, high nerve densities were associated with a higher tumor proliferative index [Fig. 8A and fig. S12; Pearson correlation coefficients = 0.529 and 0.649 for TH and VAChT, respectively (P < 0.001)]. These data are consistent with an aberrant, but orchestrated, recruitment of sympathetic and parasympathetic nerve fibers related to the malignancy.

Fig. 7 Human high-risk prostate adenocarcinomas are rich in adrenergic and cholinergic nerve fibers.

(A) Quantification of immunostained TH+ neural areas in low-risk (n = 30) and high-risk (n = 13) human prostate adenocarcinomas. Representation of TH+ nerve densities per field in normal and Gleason grade 3, 4, or 5 tumor areas. Each bar represents the averages for a patient. (B) Representative images showing the thick disorganized TH+ adrenergic neural network in normal tissues surrounding cancer in a high-risk patient (left, TH+, red; DAPI+, blue; with matched H&E section) and boxed area showing higher magnification of infiltrating fibers. By contrast, Gleason 4 invasive adenocarcinoma displays fewer discrete nerves (right). (C) Average TH+ fiber densities in both normal (blue) and cancer (red) tissues of low-risk (Lo) and high-risk (Hi) patients. (D) Quantification of cholinergic VAChT+ nerves in normal and tumor tissues in the same patients as in (A). (E) Immunofluorescence images of VAChT+ nerves (red; DAPI+, blue) in Gleason 3 (left) and Gleason 4 (right) tumor areas, with matched H&E-stained fields and boxed areas showing higher magnification. (F) Average VAChT+ fiber densities compiled as in (C). For (A) and (D), each bar represents average nerve densities of a patient obtained from 10 fields per Gleason grade or per normal area, field surface = 0.15 mm2. Scale bars, 50 μm. ***P < 0.001. Error bars indicate SE.

Fig. 8 Density of nerve fibers in human prostate cancer specimens correlates with tumor aggressiveness.

(A) Left: Representative immunofluorescence image of Ki-67+ nuclei (green; DAPI+, blue; NF-H+ fibers, red) in a Gleason 4 cancer area. Right: Average proliferative indexes in normal (blue) or cancer (red) areas. Bars represent average proliferative indexes obtained from 10 fields per Gleason grade or per normal area per patient, field surface = 0.15 mm2. Scale bar, 50 μm. ***P < 0.001. (B) Left: Recurrence-free survival of patients with high (>2000 μm2/field) and low (<2000 μm2/field) adrenergic nerve densities. Right: Recurrence-free survival of patients with high (>300 μm2/field) and low (<300 μm2/field) cholinergic nerve densities. Error bars indicate SE. (C) Schematic illustration showing how prostate cancer initiation and metastasis may be regulated by dual neural mechanisms. Nerve fibers from the sympathetic nervous system (SNS) deliver norepinephrine (NE) from nerve terminals, which acts on β2- and β3-adrenergic receptors (Adrβ2, Adrβ3) expressed on stromal cells, promoting the survival of cancer cells and the initial development of the tumor. Nerve fibers from the parasympathetic nervous system (PNS) also invade tumors, delivering acetylcholine (Ach), which promotes tumor cell proliferation and egress to lymph nodes and distant organs through the type 1 muscarinic receptor (Chrm1) expressed on stromal cells.

We next assessed whether there was a relationship between nerve density and the clinical progression of the disease. The number of TH+ fibers in normal areas or VAChT+ fibers in cancer was found to associate positively with preoperative levels of PSA, as determined by a linear regression model (P = 0.0007 and P = 0.02 for TH and VAChT, respectively). Bivariate association revealed that TH+ nerve density and overall nerve densities of the normal tissue were positively correlated with time to biochemical recurrence (P = 0.0039, P = 0.0095, and P = 0.0454 for NF-L, NF-H, and TH, respectively) and to tumor spreading outside of the prostate (P < 0.0001 for each). In addition, VAChT+, NF-L+, and NF-H+ fibers in cancer were significantly associated with extraprostatic extension (P < 0.0001 for each). Cox and logistic regression models confirmed these associations (tables S4 and S5). Statistical significance associated with fiber densities was not sustained in multivariable analysis when clinical variables including PSA levels and Gleason score were adjusted. This finding may relate to the association of fiber densities with these clinical variables and/or the small sample size that limits the possibility of detecting the independent prognostic value of the biomarkers. However, specific thresholds of TH+ nerve areas per field of normal tissue (>2000 μm2 per field; Fig. 8B) or VAChT+ nerve areas in tumor tissue (>300 μm2 per field; Fig. 8B) at diagnosis were associated with higher recurrence rates. These preliminary results suggest that nerve density assessment merits exploration as a possible predictive marker of prostate cancer aggressiveness (28).


Prior studies have described a process termed perineural invasion in which tumor cells grow and migrate along native nerve fibers. This process is associated with a poor prognosis, possibly because the nerves provide survival signals to the tumors (30) or provide a gateway toward hematogenous spread (31). The present results uncover a distinct phenomenon, possibly analogous to angiogenesis (16), where the tumor itself is infiltrated by a network of newly developed autonomic nerve projections that regulate cancer initiation and progression.

Studying mouse models, we have found complementary functions for the two branches of the autonomic nervous system: Adrenergic fibers from the SNS, acting through stromal β2- and β3-adrenergic receptors, play an important role in the initial phases of cancer development by promoting tumor cell survival, while cholinergic fibers of the PNS play predominant roles in tumor cell invasion, migration, and distant metastases through stromal Chrm1-mediated signals (Fig. 8C). This idea is supported by the innervation patterns of human prostate cancer, where sympathetic fibers accumulate in normal tissues and infiltrate the tumor edge, whereas parasympathetic fibers frankly infiltrate tumor tissues. These results are consistent with recent epidemiological data suggesting that β-blocker intake is associated with improved survival of prostate cancer patients (13, 14). It is notable that clinically used β-blockers are largely selective to the β1 adrenoreceptor, and thus our results suggest that these clinical studies may underestimate their potential benefits. Although further studies will be required to dissect the molecular events linking tumor “neurogenesis” to cancer progression, our data raise the tantalizing possibility that drugs targeting both branches of the autonomic nervous system may be useful therapeutics for prostate cancer.

Materials and Methods

Balb/c nu/nu (B6.Cg-Foxn1nu+/−) and Hi-Myc mice [FVB-Tg(ARR2/Pbsn-MYC)7Key (22)] were obtained from Charles River laboratories and the National Cancer Institute, respectively. Balb/c heterozygous nude mice were intercrossed with Adrb2tm1Bkk/J−/− and Adrb3tm1Lowl/J−/− mice or with Chrm1tm1Stl−/− mice obtained from the Jackson Laboratory. FVB-Tg(ARR2/Pbsn-MYC)7Key Chrm1tm1Stl−/− and respective controls were also generated by intercrossing the two strains.

All in vivo experiments were approved by the Animal Care and Use Committee of Mount Sinai School of Medicine and Albert Einstein College of Medicine. Human tumors were induced by orthotopic surgical implantations of 105 PC-3luc cells into 6- to 8-week-old Balb/c nu/nu mice. Ten days after cell injection (day 0), the animals were randomized into the different groups and received the appropriate drugs as indicated. 6OHDA or vehicle was injected at day 0 (100 mg/kg) and day 2 (250 mg/kg). In other experiments, 2 × 105 PC-3luc cells were orthotopically injected into nu/nu Adrβ2tm1Bkk/J+/+Adrβ3tm1Lowl/J+/+, nu/nu Adrβ2tm1Bkk/J−/−, nu/nu Adrβ3tm1Lowl/J−/−, and nu/nu Adrβ2tm1Bkk/J−/−Adrβ3tm1Lowl/J−/− mice. In selected experiments, 2 × 106 LNCaP-luc cells were injected in nu/nu Adrβ2tm1Bkk/J+/+Adrβ3tm1Lowl/J+/+ and nu/nu Adrβ2tm1Bkk/J−/−Adrβ3tm1Lowl/J−/− mice. For the transgenic model, 1-, 2-, and 5-month-old Hi-Myc mice (Adult1, Adult2, and Adult3, respectively) were injected with 6OHDA or surgically sympathectomized according to protocols described above and killed 30 days later. Neonates Hi-Myc littermates were injected with 6OHDA at day 2 (lo,100 mg/kg; hi, 200 mg/kg), day 4 and 6 (lo, 100 mg/kg; hi, 400 mg/kg), at day 8 (lo, 100 mg/kg; hi, 800 mg/kg) and at day 9 (lo, 100 mg/kg) after birth and killed 30 or 60 days later.

For experiments on the PNS, 15 days after tumor cell injection, animals received carbamoylcholine chloride (Sigma) at 250 (day 0), 300 (day 1), 350 (day 2), 500 μg/kg per day (day 3) [every 12 hours, 8 divided doses intraperitoneally (i.p.) in saline] alone or in combination with scopolamine hydrobromide (Sigma, 1 mg/kg) (32) or pirenzepine dihydrochloride (Sigma, 6 mg/kg) (33). A second cycle was administered at week 4 and mice were killed 1 week later (5 weeks after graft). For experiments using nu/nu Chrm1tm1Stl+/+ and nu/nu Chrm1tm1Stl−/− animals, mice were killed at week 6 after grafting. For survival study, grafted mice were treated as described above from week 4, every 2 weeks for up to 25 weeks or until death. Disease progression was monitored by bioluminescence scanner. For Hi-Myc model, 3-month-old mice were injected with carbachol alone or in combination with pirenzepine for 4 days after the protocol described above. One week later, mice received a second round of treatment and then were killed at month 4. For cMyc+ acini implantation, 2-month-old cMyc+ Chrm1tm1Stl−/− acini were implanted into 6-week-old nu/nu Chrm1tm1Stl+/+ recipients. After 5 weeks, mice were treated with two cycles (week 6 and 8) of carbamoylcholine chloride alone or combined with pirenzepine dihydrochloride as described above. Mice were killed at week 10.

Cell Culture

PC-3 cells stably transfected with the luciferase gene (PC-3luc, gift from J. Blanco, CSIC, Barcelona, Spain) were grown in F12-Glutamax medium supplemented with 10% fetal bovine serum (FBS), bicarbonate sodium (1.5 g/liter), and G418 (500 mg/ml; Invitrogen). LNCaP cells expressing luciferase (Xenogen; Caliper Life Sciences, Hopkinton, MA) were grown according to manufacturer’s recommendations in RPMI medium (ATCC#30-2001) supplemented with 10% FBS and 1% penicillin/streptomycin (Gibco).

Bioluminescence Imaging

In vivo and ex vivo bioluminescence imaging was performed and analyzed using an IVIS imaging system 200 series (Xenogen). Bioluminescent signal was induced by i.p. injection of d-luciferin [150 mg/kg in phosphate-buffered saline (PBS)] 8 min before in vivo imaging. For ex vivo imaging, d-luciferin (300 mg/kg) was injected 7 min before necropsy. Organs of interest were immersed in a solution of d-luciferin at 150 mg/ml (17).

Histology and Immunofluorescence

Upon killing, mouse prostate tissues were immersed in OCT medium. Frozen sections (thickness 5 μm) were stained with hematoxylin and eosin (H&E). For immunofluorescence, prostate sections were fixed with acetone or methanol and incubated in H2O2 to quench endogenous peroxidase. Nonspecific binding was blocked with goat serum in BSA solution and an avidin/biotin blocking kit (Vector Laboratories, Burlingame, CA) (20). Sections were incubated with a rabbit antibody to TH (Millipore, Billerica, MA), or antibody to VAChT (Phoenix Pharmaceuticals, Burlingame, CA), or antibodies to NF-L (Millipore) or NF-H (Abcam, Cambridge, MA) followed by secondary biotinylated goat antibody to rabbit immunoglobulin G (IgG) (Vector). Signal was amplified by Vectastain Elite ABC Kit (Vector) and visualized by Tyramide Signal Amplification kit for TRITC (PerkinElmer). For proliferative cell quantification, sections were incubated with a rabbit polyclonal antibody to Ki-67 (Abcam) and then Alexafluor568-conjugated goat antibody to rabbit IgG (Molecular Probes). For apoptotic cell quantification, prostate sections were fixed with 4% paraformaldehyde and stained with the Mebstain Apoptosis Kit according to manufacturer’s recommendations (MBL International, Woburn, MA). Basement membranes were stained with a rat polyclonal antibody to α2laminin (Abcam) followed by Alexafluor647-conjugated goat antibody to rat IgG (Molecular Probes). For bone staining, samples were fixed and decalcified prior to embedding in paraffin. Bone sections were stained with a pancytokeratin antibody from Sigma. Human prostate tissues were previously fixed and embedded in paraffin as part of routine care at the Durham VA Medical Center. Blocks were serially sectioned (thickness 5 μm) and H&E staining was performed using standard procedures. For immunofluorescence analyses, sections were deparaffinized with xylene and rehydrated through graded alcohol washes followed by antigen retrieval in sodium citrate buffer following manufacturer recommendations (Vector). Primary antibodies (rabbit antibody to TH, chicken antibody to NF-H, rabbit antibody to NF-L from Millipore, rabbit antibody to VAChT from MBL International) were incubated overnight, followed by amplification steps only for TH and VAChT stainings, as described above. For NF-L or NF-H stainings, slides were blocked in goat serum and BSA solution and then subsequently incubated with goat antibody to rabbit IgG or to chicken IgG, respectively. For proliferative index, sections were incubated with an antibody to human Ki-67 (Vector) followed by the amplification steps described above.

Bright-field images were captured and collected with a Zeiss axioplan2 microscope (Zeiss MicroImaging, Thornwood, NY) and with a Q-imaging MP3.3 RTV color camera controlled by Zeiss AxioVision software. Full Hi-Myc prostate sections were captured with a Zeiss Axioplan2IE and a Zeiss AxioCamMRc camera controlled by Zeiss AxioVision software equipped with a motorized stage that automates montage acquisition and stitching for high-resolution images of large areas.

Fluorescence images were captured and analyzed using a Axio Examiner.D1 microscope (Zeiss) equipped with a Yokogawa CSU-X1 confocal scanner head with four-stack laser system (405-, 488-, 561-, and 642-nm wavelengths). Images were obtained as three-dimensional (3D) stacks scanning through the whole thickness of the tissue using Coolsnap HQ digital camera (Ropert Scientific, Munich) and analyzed using Slidebook software (Intelligent Imaging Innovations, Denver).

Human Prostate Samples

Preexisting human formalin-fixed paraffin-embedded radical prostatectomies were obtained for staining after internal review board approval by the Durham VA Medical Center. Patient characteristics including age, race, and dates of surgery are shown in table S3. All patients underwent primary radical prostatectomy and had histologically confirmed and clinically localized prostate cancer [stage T1-T2NxM0 in the tumor-node-metastasis classification system according to the American Joint Committee on Cancer (34)]. For each patient, data on the primary tumor (clinical and pathological stages and Gleason grade) were recorded, as well as preoperative PSA levels. PSA recurrence was defined as a single PSA value at >0.2 ng/ml, two values at 0.2 ng/ml, or secondary treatment for a rising PSA. Patients treated with adjuvant therapy with an undetectable PSA (n = 6) were censored as nonrecurrent at the time of adjuvant treatment. Recurrence might be local or distant, although no metastasis has been documented thus far in this cohort of patients. Median PSA follow-up among men without recurrence was 57 months. Extraprostatic extension was defined as disease involving one or more of extracapsular, bladder neck, or seminar vesicle extension. Surgically resected primary tumors were paraffin-embedded and serially sectioned (thickness 5 μm). For each block, a section was stained with H&E to evaluate tissue viability, to localize normal areas among cancer, and to map the different Gleason grade areas. For each patient, the whole histological section was analyzed by two independent pathologists from the Durham VA Medical Center and Albert Einstein School of Medicine to define the Gleason grade. Assessment of nerve densities was conducted blind, without knowledge of histological diagnoses and clinical data. For each patient (n = 43), consecutive sections were stained for TH, VAChT, both NF-L and NF-H, or Ki-67 as described above to quantify adrenergic, cholinergic, and total autonomic nerve fiber densities or cell proliferation, respectively, in prostate tumor areas (Gleason grade from 3 to 5) and in remaining normal prostate tissues surrounding cancer areas. For each marker defined above, the average of 10 representative fields (one field = 0.15 mm2 for Zeiss 20×/1.0 NA objective) was calculated from normal areas and for each tumor grade (when present) captured as described above. A total of 3989 z-stack images were acquired and converted in 2D maximum projections that were digitally analyzed with the Slidebook to quantify nerve fiber areas and Ki-67+/DAPI+ cell areas per field (DAPI, 4′,6-diamidino-2-phenylindole). For these analyses, masks were drawn to delimit precisely contours of nerve fibers or Ki-67+ or DAPI+ cells. The surface areas of TH+, VAChT+, NFL-L+, or NF-H+ nerve fibers and the ratios between Ki-67+ and DAPI+ cells were then established.

Positron Emission Tomography (PET) Imaging

All mice were imaged after 12 hours of fasting and anesthetized throughout the whole procedure with 1.5% isoflurane-O2 mixture. Animals were injected with 400 μCi of either [18F]fluoro-2-deoxyglucose (FDG) or Na18F, in 0.1 ml of normal saline, into the tail vein. Image acquisition started 50 min after injection, using an Inveon Multimodality scanner (Siemens, Malvern, PA) with the PET module, which provides 12.7-cm axial and 10-cm transaxial active field of view. The PET scanner has no septa and acquisitions were performed in 3D list mode. A reconstructed full width at half maximum (FWHM) resolution of <1.4 mm was achievable in the center of the axial field of view. List mode acquisition of data was performed to permit dynamic reframing for kinetic evaluation of radiotracer uptake. After each acquisition, data were sorted into 3D sinograms, and images were reconstructed using a 2D-ordered subset expectation maximization algorithm. Data were corrected for decay, deadtime counting losses, random coincidences, and the measured nonuniformity of detector response, but not for attenuation or scatter. Analyses were performed using IRW dedicated software (Siemens). All images were inspected visually in a rotating 3D projection display to examine for interpretability and image artifact. Regions of interest were drawn around areas of pathologic uptake. Successive scrolling through 2D slices (each 1.2 mm thick in the axial images) permits measurement of a radioactivity within defined volumes. Corrected counts per cc within this volume divided by the counts per gram of total body mass of injected radioactivity determines the standardized uptake value (SUV); SUVmax is the maximum value of SUV within each tumor volume. The total number of bone spots uptaking Na18F has also been determined for each mouse.

Assessment of Hemodynamics and Cardiac Function

To evaluate the effect of the cholinergic agonist on blood flow, mice were anesthetized and prepared for intravital microscopy as described (35) 1 hour after carbachol injection. Centerline red blood cell velocities were measured for 10 venules and 10 arterioles in two independent experiments. Wall shear rates (γ) were calculated according to Poiseuille’s law for a Newtonian fluid, γ = 2.12(8Vmean)/Dv or Da, where Dv and Da are the venular or arterial diameters, the mean blood flow velocity (Vmean) was estimated as VRBC/1.6, and 2.12 is a median empirical correction factor obtained from actual velocity profiles measured in microvessels in vivo (35). The blood flow rate was calculated from the formula VmeanπD2/4.

Cardiac function was assessed by echocardiography using the Vevo 2100 ultrasound imaging system, in which mice were treated for four consecutive days with saline or carbachol according to the protocol described above (n = 4 per group). For imaging, animals were anesthetized with a mixture of O2/1.5% isoflurane and then positioned ventral side up on the platform of the imaging system. ECG signal and respiratory rate were captured through the electrode pads on the advanced physiological monitoring unit and transmitted to the Vevo system for monitoring. Cardiac examinations were performed in 2D images using the parasternal long axis (PLAX) view in B-mode with a 1MS550D 40-MHz probe. Two cineloops (300 frames per cineloop) were recorded per animal and then analyzed on two diastoles and two systoles per animal. The endocardial stroke volume and endocardial diastolic volume were determined to calculate the ejection fraction.

Proliferation Assay

Experiments were carried out as described (36).

RNA Extraction and qRT-PCR

Gene expression levels were analyzed from RNA extracted, using TRIzol solution (Invitrogen), from PC-3, DU145, LNCaP cell lines or prostates from Balb/c nu/nu mice by quantitative real-time reverse transcription polymerase chain reaction (qRT-PCR), as described (20, 21). Primer sequences are shown in table S1.

Statistical Analyses

For mouse studies, all values are reported as means ± SEM. Statistical significance for three or more groups was assessed by a nonparametric one-way analysis of variance (Kruskal-Wallis), followed by an unpaired Mann-Whitney test. Significance was set at P < 0.05. The Kaplan-Meier method was used for survival curve analysis, and the log-rank (Mantel-Cox) test was used to determine the statistical significance of difference between survival curves using Graphpad Prism 5 software.

For human studies, we examined bivariate associations between each of the markers and the outcome variables. For biochemical tumor recurrence and extraprostatic extension (yes or no), a Wilcoxon ranked sum test was used. Pearson correlations between markers were calculated to examine the association with proliferative indexes. The association between TH or VAChT expression and time to recurrence was determined using a Kaplan-Meier plot and statistical significance assessed by log-rank test. Regression models were used to further estimate the magnitude of the association: A linear regression model was used to estimate the association of nerve densities in normal tissue and cancer separately as well as jointly with preoperative PSA levels; a Cox proportional-hazards model was used to estimate hazard ratios with 95% confidence intervals (CIs) for the association between the same predictors with time to biochemical tumor recurrence; and a logistic regression model was used to estimate odds ratios and 95% CIs for the association between the same predictors with tumor invasion (yes or no). Several steps of multivariate analysis were also conducted in the regression models with initial adjustments for age and/or race (characterized as black versus otherwise) and then PSA score and Gleason stage. Significance level was set at P < 0.05. Statistical tests were two-sided and conducted using SAS 9.1 software.

Supplementary Materials

Figs. S1 to S12

Tables S1 to S5

References and Notes

  1. Acknowledgments: We thank N. Mall, J. Harding, and R. Basu for tissue sectioning and H&E staining; Y. Zhou for assistance with bioluminescence scanning; E. Fine, L. Jelicks, and W. Koba for help with microPET scanning at the Gruss Magnetic Resonance Research Center at Einstein; W. Guo for comments on the manuscript; and H. Ostrer for advice on human samples. Supported by U.S. Department of Defense Idea Development award W81XWH-07-1-0165 and NIH grants DK056638, HL069438, and HL097819. C.M. was the recipient of a fellowship from the Fondation pour la Recherche Médicale, France. Albert Einstein College of Medicine and the authors (C.M., P.S.F.) have filed a patent application (WO 2012071573 A2) relating to the use of adrenergic and muscarinic receptor antagonists for cancer therapy.
View Abstract

Stay Connected to Science

Navigate This Article