Neuroplastic Changes Occur Early in the Development of Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is associated with significant morbidity, mortality, and pain that can significantly affect quality of life and survival time (1). Both PDAC-related pain and local tumor spread to retroperitoneal structures are thought to be related to perineural tumor invasion (1). Tumor invasion of intrapancreatic nerves facilitates local and distant tumor spread and exposes nerves to a complex inflammatory milieu. Inflammatory mediators and neurotrophic factors identified in resected PDAC specimens, mostly from patients with advanced disease, are known to produce nerve hypertrophy, enhance excitability, and promote perineural invasion (2–6). Furthermore, these neuroplastic changes correlate strongly with the severity of PDAC-associated pain (1, 6).

Neuroplastic changes and their sequelae are thought to be consequences of pancreatic cancer because of factors released in the tumor. However, a recent study of neuroplastic changes related to the development of prostate cancer suggests that the peripheral nervous system plays an early, active role in tumor development and invasion (7). Furthermore, neuroplastic changes in the pancreas have been described in pancreatitis (8, 9) and we hypothesized that similar changes may occur during pancreatic intraepithelial neoplasia (PanIN)–only stages of PDAC development as well. To address this issue, we examined the time course of neuroplastic changes throughout PDAC progression in a genetically engineered mouse model of PDAC to determine whether changes in pancreatic nerves begin during precancer stages. Our results demonstrate that changes in innervation and sensory neuron properties parallel disease progression and begin before the appearance of cancer, suggesting that a better understanding of early changes in the peripheral nervous system is important for our understanding of PDAC biology.

The KPC mice express a conditional mutant Kras allele (LSL-Kras^G12D), a conditional Trp53 allele with LoxP sites (p53^Lox/+), and p48-Cre (p48-Cre), as previously described (10–13). Mice with LSL-Kras^G12D; p53^+/+, LSL-Kras^G12D; p53^lox/+, LSL-Kras^G12D; p53^lox/lox genotypes or either conditional allele alone were used as age- and sex-matched littermate controls. To visualize tumor cells of pancreatic origin, PDAC mice and littermate controls were crossed with the ROSA-LSL-tdTomato reporter strain [B6.Cg-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J; The Jackson Laboratory, Bar Harbor, ME] to produce KPCT mice and tdTomato controls. KPC and KPCT mice of both sexes were analyzed at 4 time points during tumor development: 3 to 4 weeks, 6 to 8 weeks, 10 to 12 weeks, and greater than 16 (>16) weeks.

KPC mice were weighed weekly to monitor sickness, although no significant difference in weight between KPC mice and age- and sex-matched littermates was observed (data not shown). Although age provided an approximation of disease progression, significant variability in tumor size, tumor location, and disease severity was observed among KPC mice. Disease progression and sickness severity were also assessed using a mouse hunching scale adapted from a previous study of pancreatic acinar carcinoma in the mouse (14). Animals were scored weekly starting at post-natal day 28 using the following criteria: 0, no detectable sickness-related behavior; 1, slight notch visible in the animals' back, near the shoulders; 2, a noticeably hunched posture and mild piloerection; 3, moderately hunched posture and increased piloerection; and 4, severe hunching, piloerection, and very limited voluntary movement. The average age that a hunching score of 1 was first detected in KPC mice was 10 weeks, which correlates with the PanIN stage of tumor development described above. A hunching score of 2 or 3 correlates with significant sickness and was detected as early as 19 weeks and as late as 26 weeks, at which point animals have significant tumor burden.

All animals were housed in the Association for Assessment and Accreditation of Laboratory Animal Care-accredited Division of Laboratory Animal Resources at the University of Pittsburgh with ad libitum access to water and food. Animals were cared for and used in accordance with guidelines of the Institutional Animal Care and Use Committee at the University of Pittsburgh.

Tissues were harvested from KPCT mice and tdTomato controls perfused with 4% paraformaldehyde in 0.1 M phosphate buffer. Pancreata/tumors were embedded in OCT compound and sectioned on a cryostat at 30-μm thickness. Celiac ganglia and the spinal cord with associated dorsal root ganglia (DRG) from one KPCT mouse with tumor encasement of the spinal cord were embedded in 10% gelatin in 0.1 M phosphate buffer and sectioned on a sliding microtome at 20-μm thickness. Tissue sections were washed, blocked (5% normal horse serum and 0.25% Triton X-100 in 0.1 M phosphate buffer), incubated in primary antibody overnight at room temperature, washed, incubated in secondary antibody (donkey anti-rabbit Cy2 1:500; Jackson ImmunoResearch) for 2 hours at room temperature, washed, dehydrated, and coverslipped with DPX mounting media (pancreas sections) or Vectashield mounting media (celiac ganglion and spinal cord). Primary antibodies used were rabbit anti-protein gene product 9.5 (PGP 9.5; 1:1,000, UltraClone), rabbit anti-tyrosine hydroxylase (1:200, Cell Signaling Technology), and rabbit anti-activating transcription factor 3 (ATF3; 1:200, Santa Cruz Biotechnology). Sections of pancreas were also hematoxylin and eosin stained and coverslipped with DPX mounting medium. Sections were viewed and photographed on a LEICA DM 4000B microscope (Leica Microsystems) using Leica Application Suite software.

Changes in pancreatic innervation were quantified in sections of pancreas from KPCT mice >16 and 10 to 12 weeks of age and tdTomato controls (n = 4 for each group). Nerve fibers were visualized using an antibody for PGP 9.5, as described above, and at least 40 images were captured per animal, spanning across approximately 36 sections (about 1-mm thickness). Images were captured at ×20 magnification and nerve fibers were traced and analyzed using the ImageJ plugin, NeuronJ (15). The total number of fibers, average fiber density (fibers/mm²), and average fiber length were calculated for each animal. For each parameter, statistical significance between cancer and control groups was determined using t test comparisons (SPSS Statistics, IBM Corporation).

To assess pain-related behavior in KPC mice, open-field exploratory behavior was analyzed at time points ranging from 7 to 31 weeks of age. As previously described (8, 9), mice were placed in plexiglass boxes and their activity in the horizontal and vertical planes was measured photoelectrically at a 0.75-cm spatial resolution for a period of 15 minutes. TruScan software (Coulbourn Instruments) analyzed movements, speed, distance traveled, time spent moving. Both horizontal movement along the floor of the behavior arena and vertical movement in which animals were extending upward were measured. The monitoring period was divided into 3 blocks of 5 minutes and data for each movement parameter were analyzed using a linear mixed effects model with age, genotype, and time block treated as fixed effects and each individual animal treated as a random effect to account for intra-animal correlation. All analyses were performed using the R package lme4 and P values for the fixed effects were based on likelihood ratio tests. Data from the second time block are presented as representative.

Animals were deeply anesthetized with 0.1 to 0.2 cc ketamine/xylazine and either the pancreas/tumor was removed and immediately homogenized in 2 mL TRIzol reagent (Invitrogen) or mice were transcardially perfused with 0.9% saline and DRG 9 to 12 were removed bilaterally and frozen on dry ice. RNA was TRIzol/chloroform extracted, precipitated in isopropanol, washed with 75% ethanol, resuspended in RNase-free water, and further purified using RNeasy columns (Qiagen). RNA from DRG was isolated using RNeasy columns and reconstituted in RNase-free water. All samples were treated with DNAse (Invitrogen) and 300 ng to 1 μg was reverse-transcribed using Superscript II (Invitrogen). SYBR green-labeled PCR amplification was performed using an ABI 7000 real-time thermal cycler controlled by Prism 7000 SDS software (Applied Biosystems) and threshold cycle (C_t) values were recorded as a measure of initial template concentration. Primers for nerve growth factor (Ngf), brain-derived neurotrophic factor (Bdnf), glial cell line–derived neurotrophic factor (Gdnf), neurturin (Nrtn), and artemin (Artn), their respective receptors, neurotrophic tyrosine kinase receptor, type 1 and type 2 (Trka and Trkb), and glycosylphosphatidylinositol-linked receptor α1, α2, and α3 (Gfrα1, Gfrα2, and Gfrα3), the nociception-related genes transient receptor potential cation channel vanilloid 1 (Trpv1), transient receptor potential cation channel subfamily A member 1 (Trpa1), sodium voltage-gated channel 1.8 (Na_v. 1.8), potassium voltage-gated channel 4.3 (K_v 4.3), calcitonin-related polypeptide α (Cgrp), and neurokinin1 (Nk1), and the injury marker, ATF (Atf3) are listed in Supplementary Table S1. Two housekeeping genes, ribosomal protein L13A (Rpl13a) and glyceraldehyde 3-phosphate dehydrogenase (Gapdh) were used for normalization (Supplementary Table S1). Additional housekeeping genes were tested using primers from a mouse housekeeping gene primer set (MHK-1; RealTimePrimers.com; data not shown). Relative fold change in RNA was calculated using the ΔΔC_t method (16) with Rpl13a as a reference standard for pancreas samples and Gapdh as a reference standard for DRG samples. Significance was determined using Mann–Whitney U tests (SPSS Statistics, IBM Corporation).

Transcriptional profiling of human tumor–derived cell lines (MiaPaCa2 and Panc1) and mouse tumor-derived cell lines (Kpc1 and Kpc2) was carried out. Mouse cell lines were derived from tumors from 2 different KPC mice (p48-Cre; LSL-KRAS^G12D; p53^lox/lox) and cultured in RPMI media with 10% FBS and 1% penicillin/streptomycin (Pen/Strep; Invitrogen). MiaPaCa2 cells [American Type Culture Collection (ATCC), CRL-1420] were cultured in MEM with 10% FBS, 2.5% horse serum, and 1% Pen/Strep. Panc1 cells (ATCC, 1469) were cultured in MEM with 10% FBS and 1% Pen/Strep. Cell lines were grown to confluence and RNA was extracted using TRIzol reagent. 1 μg of DNAsed RNA was reverse-transcribed and PCR amplified using GoTaq DNA polymerase (Qiagen) using primer sets listed in Supplementary Table S1 (murine) and Supplementary Table S2 (human).

Neuroplastic changes associated with PDAC were studied using genetically engineered mice that express a conditional mutant Kras allele and a Trp53 allele with LoxP sites under control of a pancreas-specific promoter driving Cre, p48-Cre. Although disease course was variable among KPC and KPCT mice, PanIN lesions, and fibrosis typically were present by 6 to 8 weeks of age, more advanced PanIN lesions, and more extensive fibrosis were evident by 10 to 12 weeks, and ductal adenocarcinoma was observed in most mice by 16 weeks of age (Supplementary Fig. S1). This time course of tumor development is consistent with prior reports using related mouse strains (11, 13).

Neuroplastic changes are common in sections of resected tumor from patients with PDAC (1) and these changes in innervation frequently include perineural tumor invasion (17–22). To determine if similar changes in innervation and perineural invasion (PNI) occur in the mouse model of PDAC, we used the pan-neuronal marker PGP 9.5 to examine the density and distribution of pancreatic nerve fibers in KPCT mice (Fig. 1). At 6 to 8 weeks of age, KPCT pancreata exhibited focal areas of hyperinnervation (Fig. 1E–H) not present in control pancreata (Fig. 1A–D). These areas were restricted to relatively small regions of the pancreas containing fibrosis, acinar atrophy, and/or PanIN lesions, whereas innervation in histologically normal pancreas was similar to that of controls. In the pancreata of KPCT mice at 10 to 12 weeks of age, multifocal PanIN lesions ranging from PanIN 1a to PanIN 3 were seen and hyperinnervation accompanied this expansion of pancreas pathology (Fig. 1I–L).

Large intrapancreatic nerve bundles were observed in PDAC pancreata/tumors at >16 weeks of age (Fig. 2). Hypertrophied nerves were often associated with areas of fibrosis and acinar cell atrophy (Fig. 2D, H, L). PGP+ fibers were visualized extending from atrophied/aggregated islets (Fig. 2A–D), near ducts, and vasculature (Fig. 2E–H), and adjacent to areas of PanIN lesions or tumor cells (Fig. 2I–L). Although intra-pancreatic PNI was not definitively observed, areas of focal fibrosis with hyperinnervation often featured branching of PGP9.5⁺ fibers near tdTomato⁺ cells or PGP9.5⁺ fibers appearing to engulf tdTomato⁺ cells, suggesting a mutual tropism between the two cell types (Fig. 2C).

PGP+ fibers in the pancreas were traced using NeuronJ software to quantify the changes in pancreatic innervation observed in KPCT mice at 10 to 12 and >16 weeks of age. Total number of fibers (mean ± SD) traced in the pancreas/tumor of KPCT mice at 10 to 12 weeks (570.5 ± 151.4) and >16 weeks (585.3 ± 118.6) of age was significantly increased compared with tdTomato controls (348 ± 52.39; P = 0.032 and 0.011, respectively). Similarly, the average fiber density in the pancreas/tumor of KPCT mice at 10 to 12 weeks (3.828 ± 1.043 fibers/mm²) and >16 weeks (3.943 ± 0.7990 fibers/mm²) was significantly increased compared with tdTomato controls (2.275 ± 0.4203 fibers/mm²; P = 0.03 and 0.0086, respectively). Consistent with the observed hypertrophied pancreatic afferents described above, the average fiber length in KPCT mice at 10 to 12 weeks (54.72 ± 8.021 μm) and >16 weeks (45.61 ± 5.977 μm) was also significantly increased compared with tdTomato controls (33.38 ± 4.69 μm; P = 0.0037 and 0.018, respectively).

Using KPCT mice allowed tracking of tdTomato⁺ cells of pancreatic origin in local and distant sensory and sympathetic nerve ganglia. No tdTomato⁺ cells were ever visualized in the celiac ganglion of tdTomato control mice (Supplementary Fig. S2). However, in one KPCT animal with advanced disease tdTomato⁺ cells invaded into the celiac ganglion (Fig. 3), which was surrounded by a large tumor metastasis. In this case many celiac ganglion neurons were ATF3⁺, a marker of nerve injury only rarely expressed in control celiac ganglion cells (Fig. 3D–F, arrow in Fig. 3D). In another KPCT case, pancreas-derived cells invaded the dorsal T11 and T10 root ganglia (Fig. 4C and E). This migration of tumor cells into the DRG was associated with a metastatic tumor that surrounded the spinal cord at levels T9 to T12, the levels of spinal cord that normally innervate the pancreas (Fig. 4A; ref. 23). Some pancreas-derived cells in the DRG assumed complex morphology, exhibiting processes conducive to cell-to-cell adhesion or chemotaxis (inset in Fig. 4E). Despite the large number of tdTomato⁺ cells present in the KPCT DRG, relatively few cells in the ipsi- or contralateral T10 and T11 DRG were ATF3⁺ (not shown).

Prior studies have found that pancreatic nerve hypertrophy is associated with changes in sensory neuron gene expression in mouse models of acute and chronic pancreatitis (8, 9). To determine if similar changes are associated with the development and progression of PDAC, the relative level of mRNAs encoding genes related to nociception, neurogenic inflammation, and nerve injury were assessed in DRG 9 to 12 of KPC mice at ages 10 to 12 and >16 weeks. Similar to findings reported in mouse models of acute and chronic pancreatitis, expression of Trpa1 and Trpv1 was significantly increased in DRG of KPC mice >16 weeks of age (1.74- and 1.36-fold, respectively, P = 0.013 and 0.043; Table 1). A nonsignificant trend toward increased Trpa1 expression was also measured in 10- to 12-week-old KPC mice (1.38-fold, P = 0.059; Table 1). In rodent models of pancreatitis, TRPV1 and TRPA1 channels are implicated in neurogenic inflammation and increased release of CGRP and NK1 neuropeptides from neural terminals into the pancreas (8, 9, 24–28). In DRG of KPC mice at 10 to 12 and >16 weeks of age, Cgrp mRNA expression is significantly increased 1.36- and 1.27-fold, respectively (P = 0.043 and 0.029; Table 1). In contrast, Nk1 expression in the DRG was not significantly different between KPC and control mice at 10 to 12 weeks and only a nonsignificant trend toward increased expression was measured at >16 weeks of age (1.44-fold, P = 0.081; Table 1). Similarly, expression of Atf3 was not significantly different between control and KPC DRG at 10 to 12 weeks, but a nonsignificant trend toward increased expression was evident at >16 weeks of age (1.92-fold, P = 0.059; Table 1). In addition, mRNAs encoding the voltage-gated ion channels Na_v1.8 and K_v4.3 were also not significantly different in the DRG of control and KPC mice at either age (Table 1).

Having observed neuroplastic changes in KPC mice similar to those described in human PDAC, we hypothesized that KPC mice would develop pain related to this condition as well. To measure the impact of PDAC on mouse behavior, we analyzed open-field exploratory activity using photoelectric monitoring throughout disease progression. The change over time in measures of horizontal exploratory activity, as animals moved along the floor of the behavior arena, was not significantly different between cancer and control mice (Fig. 5). In contrast, there was a significant difference in the change over time in measures of vertical exploratory activity between KPC and control mice (Fig. 5). Vertical movements generally decreased over time in KPC mice but not in controls (moves, P = 0.0311; time, P = 0.0006; distance, P = 0.0099; entries, P = 0.0014; Fig. 5). Vertical exploratory activity measurements included postural movements such as rearing and more subtle movements in which the mouse dorsiflexed its neck. One aspect that these vertical movements have in common is the elongation of the trunk, which conflicts with the hunched postures adopted by these animals as the disease progresses and with the hunching described in a mouse model of pancreatic acinar carcinoma (14, 29).

Previous studies have associated increased neurotrophic growth factor expression with PDAC-induced neuronal hypertrophy, PNI, and PDAC-related pain (2–6, 18, 19, 28, 30, 31). To determine if changes occur in neurotrophic factors and their receptors during tumor development and progression, pancreas RNA from KPC and control mice was analyzed at 3 to 4, 6 to 8, 10 to 12, and >16 weeks of age. There were no significant differences in expression of Ngf, Bdnf, Gdnf, Artn, Nrtn, or their receptors between KPC and control mice at 3 to 4 weeks of age (Fig. 6). Gfrα2 was significantly increased (3.74-fold; P = 0.008) in the pancreas of KPC mice at 6 to 8 week of age (Fig. 6). In the 10- to 12-week age group (when advanced PanIN lesions are present), Ngf, Trkb, Nrtn, and Gfrα2 were significantly increased (2.25-, 2.18-, 1.39-, and 3.42-fold, respectively; P = 0.007, 0.012, 0.028, and 0.028) in the pancreas of KPC mice (Fig. 6). In the pancreas of KPC mice >16 weeks of age, Ngf, Trka, Gdnf, and Gfrα2 were significantly increased (3.6-, 29.9-, 7.11-, and 4.02-fold, respectively; P = 0.03, 0.001, 0.023, and 0.024; Fig. 6). In contrast, Gfrα3 expression in the pancreas of >16-week-old PDAC mice was significantly decreased (4.30-fold; P = 0.009; Fig. 6).

For the analyses above, gene expression was normalized to the housekeeping gene Rpl13a, which did not significantly change in KPC mice at 3 to 4, 6 to 8, and 10 to 12 weeks of age. However, expression of Rpl13a was increased 3-fold in the pancreas of KPC mice >16 weeks of age, and the expression of 7 other housekeeping genes was increased 5- to 15-fold (Supplementary Table S3). It should also be mentioned that if mRNA expression is normalized based on the starting cDNA concentration, larger changes in gene expression are apparent for all growth factors and receptors, except Gfrα3 (Fig. 6). Thus, regardless of the method used, significant changes in growth factor and receptor mRNA occur, indicating that as cancer progresses, the pancreas produces a milieu that resembles the pro-growth environment experienced by the peripheral nervous system during development (32, 33).

A variety of cell types, including tumor cells, infiltrating immune cells, and cells in the stromal compartment of the tumor, could each contribute to the observed changes in growth factor and growth factor expression in the pancreas of KPC mice. Previous studies of human tumor cell lines have demonstrated that many express a variety of growth factors and receptors, suggesting that PDAC cells represent a significant source and target of released intrapancreatic neurotrophic factors. To address the potential contribution of the tumor cells to changes in growth factor expression, RNA from 2 murine PDAC cell lines derived from KPC mice (Kpc1 and Kpc2) was analyzed to assess tumor-specific neurotrophic factor and neurotrophic factor receptor expression. Kpc1 and Kpc2 expressed Ngf, Artn, Gdnf, Nrtn, and Gfrα2. Kpc2 additionally expressed Gfrα3 and Gfrα1 (Supplementary Table S4). This expression profile is quite similar to the human tumor cell lines, Panc1 and MiaPaCa2, which express ARTN, BDNF, TRKA, GFRα3, and GFRα2 (Supplementary Table S4). Panc1 also expressed NGF, GFRα1, and TRKB whereas MiaPaCa2 expressed GDNF (Supplementary Table S4).

The histologic progression from PanIN to PDAC in KPC mice has been shown previously to closely model histological changes observed in the human disease (12). As such, these mice provide an opportunity to study changes in the nervous system occurring at precancer stages of the disease, which would be difficult to study in patients with PDAC. Here we demonstrate that neuroplastic changes associated with PDAC begin at the histologic precancer stage, suggesting an active role of the nervous system in disease progression. These early neuroplastic changes are likely to contribute to the development of PDAC-related pain and may play a significant role in tumor progression, similar to what has been described in prostate cancer (7). Furthermore, the presence of nerve hypertrophy in the pancreas provides an anatomical substrate for neurogenic inflammation and metastases.

Tumor–nerve interactions have been widely described in human PDAC (17–21, 31). Previous studies have shown that intrapancreatic PNI is present in up to 100% of PDAC cases (17–21) and hypertrophied nerve bundles have also been described (1). In this study, similar changes in pancreas/tumor innervation were observed in KPCT mice >16 weeks old. We hypothesized that PDAC-related changes in pancreas innervation observed in advanced stages of human and mouse PDAC are not simply a consequence of the tumor, but represent a reciprocal relationship between the cancer and the peripheral nervous system that begins early in tumor development. As early as 6 to 8 and 10 to 12 weeks of age, areas of hyperinnervation were found in the pancreas of KPCT mice, suggesting that early changes in the microenvironment can, in fact, affect pancreatic nerves.

Although intrapancreatic PNI is difficult to detect in KPCT mice at any age, invasion of local and distant extra-pancreatic nerve ganglia was found. Invasion of extrapancreatic nerve plexuses has been described in PDAC and the presence of tumor invasion of extrapancreatic nerve plexuses is significantly correlated with decreased survival rate following tumor resection (18, 21, 22). The celiac ganglion of one KPCT mouse was encased by a large tumor metastasis and tdTomato⁺ cells were present inside the ganglion. This suggests that tumor cells migrated from the pancreas along the sympathetic nerves that synapse in the celiac ganglion or the sensory afferents that run through it, a mode of tumor spread that may occur in human PDAC as well. Similarly, tdTomato⁺ cells were located inside the T10 and T11 DRG in a different KPCT mouse in which tumor encasement of the spinal cord was grossly visible, again demonstrating the potential for tumor cell migration along sensory afferents. How tumor cells invade and move along peripheral nerve fibers is not well understood. It is likely however, that a variety of secreted factors in the environment and expression of genes in tumor cells such as adhesion molecules play a role.

In human PDAC, pancreatic nerve hypertrophy and PNI are also associated with PDAC-related pain (1, 19, 28), suggesting that nerves in the pancreas are both damaged by the invading tumor cells and sensitized by changes in the tumor microenvironment. In KPC mice, pain-like behaviors were observed with increasing age. That PDAC-related pain is most evident after the progression from PanIN to PDAC is similar to previous observations made in a genetic mouse model of acinar carcinoma (14, 29). Although decreased exploratory behavior could be because of other factors unrelated to pain such as general malaise associated with illness severity, previous studies of acute and chronic pancreatitis pain have shown that pancreatitis-related changes in exploratory behavior can be reversed by morphine treatment (8, 9), which supports a role for pain in the observed behavioral changes. Furthermore, the selective inhibition of vertical, but not horizontal, movement is also suggestive of abdominal discomfort. However, it is also true that abdominal pain in KPC mice could be related to mechanical distention of the abdomen because of ascites or somatic pain associated with peritonitis. In this study, only a minority of mice displayed signs of significant abdominal ascites at the time of testing, but future studies will be necessary to better address the specific contribution of pancreas-related pain to decreased vertical exploratory behavior.

We and others have shown that neurotrophic factors drive sprouting and sensitization of primary sensory afferents and sympathetic neurons in adult systems (34–40). Many of these factors and receptors have also been implicated in human PDAC (2–4, 6, 30, 41). Increased NGF and ARTN is significantly correlated with nerve hypertrophy in human PDAC (2, 3) and increased expression of NGF and TRKA in PDAC tissue correlates with patient reports of pain (6, 19, 30). In KPC mice >16 weeks of age, changes in neurotrophic factor and receptor expression were observed similar to those described in human PDAC. Furthermore, given that both human and murine PDAC cell lines express a variety of neurotrophic factors and receptors, in patients, tumor cells themselves (and not associated tissues like blood vessels or immune cells) probably represent an important source of these molecules. Therefore, neurotrophic factor signaling in PDAC likely plays a prominent role in promoting tumor growth and spread, in addition to affecting pancreatic afferent growth and sensitization.

In KPCT mice, the presence of altered innervation in the pancreas at 6 to 8 and 10 to 12 weeks of ages suggests that neurotrophic factor-rich, “pro-growth” microenvironments are also present at precancer stages of disease progression. Although detection of changes in growth factor or growth factor receptor expression was limited in KPC mice at 6 to 8 weeks, at 10 to 12 weeks, when more pancreas lobules contain pathologic changes, the expression of several growth factors and receptors was significantly increased. At precancer time points, neurotrophic factors and receptors may be produced by tumor precursor cells as well as other components of the microenvironment such as infiltrating immune cells. Thus, changes in neurotrophic factor signaling in the pancreas before the development of cancer could affect the progression from PanIN to PDAC through direct action on tumor precursor cells and/or via growth factor–mediated changes in pancreatic innervation.

Neuroplastic changes frequently described in PDAC are similar to changes observed in pancreatitis. Pancreatitis is associated with increased growth factor expression (42), hypertrophied nerve bundles (1, 42), and sensitized pancreatic afferents, all of which are thought to increase pancreatitis-related pain (1). In rodent models of pancreatitis, pancreatic sensory afferents upregulate nonspecific cation channels, such as TRPV1 and TRPA1, and demonstrate hypersensitivity (8, 9, 26). Activation of TRPV1 in sensitized pancreatic afferents can drive neurogenic inflammation in the pancreas through release of peptides such as CGRP and NK1, and importantly, blocking this process attenuates pain and inflammation (8, 9, 24–28). In the DRG of KPC mice >16 weeks of age, expression of Trpv1, Trpa1, and Cgrp is increased and accompanied by a trend toward increased Nk1 expression, suggesting that a similar process of peripheral afferent sensitization and neurogenic inflammation occurs in PDAC. At 10 to 12 weeks of age, a trend of increased Trpa1 expression and increased Cgrp expression was also measured. Although we would expect to see changes in the DRG similar to what has been described in pancreatitis at precancer stages of PDAC progression, the number of DRG neurons in the T9 to T12 ganglia that innervate the pancreas is relatively small. And based on the limited distribution of hyperinnervation at early time points, only a subset of pancreatic afferents may be affected early in the disease. Therefore, when analyzing whole DRG gene expression, it is not surprising that we did not detect statistically significant changes of nociception-related molecules until later in the disease process.

Chronic pancreatitis increases the risk of malignancy (43, 44) and pancreatic inflammation has been shown to promote disease progression and metastases in mouse models of PDAC (13, 45–48). Studies from our laboratories and others show that the peripheral nervous system can drive pancreatic inflammation (8, 9, 24–28). This raises the possibility that pancreatic nerves play an active role in PDAC development and progression. Even a small population of sensitized pancreatic afferents at early, pretumor stages could drive neurogenic inflammation similar to what has been described in animal models of pancreatitis, thereby contributing to and even accelerating the progression from PanIN to PDAC.

Ablation of pancreatic innervation has previously been performed in patients with PDAC, with the hypothesis that pancreatic denervation would inhibit pain and improve survival (49, 50). These and other studies have demonstrated at least some reduction in pain following chemical splanchnicectomy or celiac plexus block in patients with unresectable pancreatic cancer. However, the effects on survival time were mixed, with one study demonstrating increased survival time in patients following chemical splanchnicectomy (49) and the other reporting no effect on survival post-celiac plexus block (50). Patients studied in both trials had advanced pancreatic cancer and the results of this study suggest that if pancreas denervation was performed at an earlier time point (ideally at the pre-cancer stage) a significant inhibitory effect on tumor growth may have been seen. Although it is difficult to perform such a study because it is not possible to routinely identify noncystic, precancerous PanIN lesions in patients, targeting pancreatic innervation in patients with early, resectable disease could yield benefits in terms of both delayed onset of pain and increased survival time.

In conclusion, early changes in neurotrophic factor expression and pancreatic innervation are important not only for the subsequent development of PDAC-related pain, but also for driving disease progression from premalignant stages to cancer via sensory afferent sensitization and neurogenic inflammation. Given that prominent PNI and tumor–nerve interactions have been described in a number of malignancies, early interventions targeting the peripheral nervous system represent a novel tumor treatment strategy for a variety of cancers, including PDAC.

No potential conflicts of interest were disclosed.

Conception and design: R.E. Stopczynski, K. Bielefeldt, K.M. Albers, B.M. Davis

Development of methodology: R.E. Stopczynski, B.M. Davis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.E. Stopczynski, D.J. Hartman, H. Ying, R.A. DePinho, B.M. Davis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.E. Stopczynski, D.P. Normolle, D.J. Hartman, J.J. DeBerry, K. Bielefeldt, A.D. Rhim, B.M. Davis

Writing, review, and/or revision of the manuscript: R.E. Stopczynski, D.J. Hartman, A.D. Rhim, K.M. Albers, B.M. Davis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.M. Davis

Study supervision: J.J. DeBerry, K.M. Albers, B.M. Davis

The authors thank C. Diges and C. Sullivan for their technical assistance.

This work was supported by NIH CA177857 (B.M. Davis and A.D. Rhim), NIH NS050758 (B.M. Davis), NIH NS075760 (B.M. Davis), NIH NS033730 (K.M. Albers), NIH DK063922 (R.E. Stopczynski), NIH DK088945 (A.D. Rhim), and the Shirley Hobbs Martin Memorial Fund (B.M. Davis).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.