Besides its aggressive behavior and poor response to treatment, another major feature of pancreatic ductal adenocarcinoma (PDAC) is perineural invasion (PNI) is PNI, which is defined as cancer cells surrounding at least 33% of the epineurial, perineural, and endoneurial space of the nerve sheath. PNI is present in virtually all patients ^[1][2][10,17]. This is a significant difference from other solid tumors in which PNI is less common ^[3][18]. PNI can be quantified using scoring systems and is primarily based on its extent and frequency (Figure 1). Interestingly, in PDAC, a significant positive correlation was found between the extent of PNI and patient survival ^[3][4][5][18,19,20]. PNI is an independent risk factor for the development of R1 resection and tumor recurrence ^[1][4][10,19] and an important predictor of metastatic spread along the neuronal compartment ^[2][6][8,17]. For instance, in a retrospective study, Takahashi et al. showed how micrometastases can be frequently found in the neuronal compartment of healthy pancreatic sections of PDAC-bearing patients ^[2][17]. This event, termed ‘intrapancreatic extratumoral perineural invasion (NEX) phenomenon’, was found in more than 50% of patients undergoing curative surgery. The occurrence of the NEX phenomenon was positively correlated with NI, whilst overall survival was negatively correlated with NEX, as NEX+ patients had a significantly worse survival compared to NEX- (350 vs. 1042 days). All patients who survived PDAC without tumor recurrence were NEX- ^[2][17].

In addition, PNI also has an impact on neuronal architecture. Specifically, extensive nerve fiber hypertrophy and elongation or sprouting of nerve fibers have been usually correlated with PNI occurrence ^[7][8][21,22]. Nerve fiber hypertrophy appears to be a major contributor to the development of cancer-associated pain in PDAC. Approximately 80% of patients develop cancer-associated pain during the progression of the disease ^[7][8][21,22]. Even with modern analgesic therapies, cancer-associated pain cannot always be controlled, thus prompting the quest for novel therapeutic targets. Interestingly, two retrospective studies showed that neoadjuvant therapy significantly reduced the rate of PNI ^[9][10][23,24] As the role of neoadjuvant therapy is currently not well established, this is an interesting observation ^[11][25]. For instance, Barbier et al. showed that neoadjuvant chemoradiation significantly reduced the rate of PNI from 93 to 43% ^[9][23]. However, neoadjuvant chemoradiation also prevented more than half of patients from receiving resection due to cancer progression. Consequently, it is unsurprising that neoadjuvant chemoradiation did not improve overall survival compared to direct resection ^[11][25]. Whether neoadjuvant therapy has a positive impact on patient survival or on the development of cancer-associated pain, e.g., as part of an extended multimodality therapy concept, remains to be elucidated.

2. Molecular Mechanisms of Cancer–Neuronal Crosstalk and Targeted Therapies

2.1. Neurons Are Attracted to Pancreatic Ductal AdenocarcinomaAC Cells

Many mediators are involved in this close cancer–neuronal bidirectional crosstalk. However, our understanding of this complex interaction is still very preliminary. Currently, it is known that PDAC cells secrete mediators that promote nerve invasion into tumor tissue ^[12][15]. Cancer cells activate physiological mechanisms which in homeostasis are necessary to provide organ innervations and nerve regeneration. Among the identified mediators, neurotrophins, particularly nerve growth factor (NGF), have been best studied (Figure 24A) ^[12][13][15,78]. During embryonic development, tissues that need to be innervated secrete NGF. Along the NGF gradient, neurons eventually innervate the target organ. In a healthy pancreas, NGF is barely detectable. However, as early as the PanIN stage, the amount of NGF doubled in PDAC cells. In the PDAC stage, the amount of NGF in tumor cells can increase up to seven-fold. Overexpression and secretion of NGF attracts neurons leading to neurite outgrowth into tumor tissue. In human tissue, a clear correlation has been reported between NGF overexpression and an increased extent of NI. Thus, it is not surprising that NGF concentration in tumor tissue is associated with an increased metastasis rate and increased probability of R1 resection. Microdissection studies demonstrated that both PDAC cells and nerves produce NGF and both express the corresponding receptors, so that a reciprocal interaction may be established over the time course of tumor disease ^[13][78]. NGF binds to two receptors: the high-specificity trkA receptor and the low-specificity p75NTR. Upon binding of NGF to the high-specificity receptor trkA, activation of MEK and MAPK pathways occurs, which promotes proliferation and suppresses apoptosis ^[14][79]. Binding to the low-affinity p75NTR inhibits proliferation and induces apoptosis ^[14][79]. In this regard, patients with a high trkA receptor have significantly reduced survival. Patients with high p75NTR expression showed significantly prolonged survival. Interestingly, NGF receptor expression was shown to be one of the most important predictive parameters in PDAC in a multivariable model ^[15][35]. The crucial role of NGF in this bidirectional crosstalk in PDAC has also been demonstrated in preclinical models. For instance, in vitro, knockdown of NGF or its receptors trkA and p75NTR, respectively, could limit the proliferation and migration of PDAC cells (Figure 2A). Also, inhibition of the NGF—trkA axis resulted in decreased migration of Mia PaCa2 cells toward DRG neurons and decreased neurite outgrowth ^[1][10]. Similarly, Lei et al. tested the following hypothesis using a robust siRNA (gold nanocluster-associated delivery of siRNA of NGF; GNC-siRNA) for NGF knockdown, which was used both in vitro and in vivo ^[16][54]. Application of GNC-siRNA reduced proliferation of Panc1 cells and inhibited migration of these tumor cells in a migration chamber. When Panc1 cells were co-cultured with DRG neurons, neurite outgrowth was directed towards Panc1 cells. Once Panc1 cells were pretreated with GNC-siRNA, the extent of neurite outgrowth decreased compared with untreated Panc1 cells. GNC-siRNA was also examined in vivo. Three PDAC mouse models were used (subcutaneous model, orthotopic model and patient-derived xenograft). After NGF knockdown in tumor cells, the amount of neurite outgrowth into tumor tissue was significantly reduced and tumor growth was reduced by approximately 50%. Then, depending on the tumor model, reduction of the metastatic rate varied from a discrete reduction to the virtually complete absence of distant metastasis. Thus, inhibition of the NGF axis represents a promising therapeutic option ^[16][54]. Human anti-NGF antibodies are already available and used in clinical trials. As NGF also plays an important role in the development and maintenance of pain in chronic inflammation, NGF antibodies have so far been used mainly in clinical trials on painful osteoarthritis ^[17][80]. The side effects mainly observed include headache, paresthesias and hypoesthesias ^[17][18][80,81]. Compared to aggressive chemotherapies, these side effects are considered acceptable. To date, no clinical trial for use in pancreatic cancer has been initiated, but this may soon follow due to the promising preclinical data. Specifically, the biweekly application of anti-NGF antibodies had favorable effects in LSL-Kras+/G12D; LSL-Trp53+/R172H; Pdx1-Cre (KPC) animals ^[19][55]. The development of PDAC was suppressed, the rate of PNI was reduced by 40% and the macrometastases were not detectable (vs. 30% in sham-treated KPC animals). Inhibition of the TrkA receptor also appears to be effective, although human trk inhibitors can cross the blood–brain barrier and thus the side effects to be expected might be more critical ^[19][55]. Nevertheless, first clinical studies in solid tumors, including pancreatic cancer, have been initiated (Table 1). NGF also activates multiple pathways within the tumor that increase cellular invasiveness such as nuclear factor kappa–light-chain-enhancer of activated B-cells (NF-κB) ^[20][82]. In vitro data demonstrate that inhibitors of the NF-κB pathway inhibit NGF-mediated PNI and neural outgrowth. In vivo, inhibition of the NF-κB pathway leads to a reduction in neurotrophin expression, nerve density and PNI ^[20][82]. Inhibition of the NF-κB pathway was achieved on the one hand by NF-κB modulation plasmids and on the other hand by Triptolide and its water-soluble prodrug Minnelide. Triptolide is a diterpenoid triepoxide, which is extracted from the Chinese herb Tripterygium wilfordii. Among the over 300 ingredients of Tripterygium wilfordii, Triptolide is the most important bioactive component. The effect of triptolide is manifold ^[28][29][90,91]. On the one hand, it induces apoptosis in pancreatic cancer cells and elicits its antitumor activity through super-enhancer disruption to re-program cellular crosstalk ^[23][85]. On the other hand, it increases the cytotoxicity of various chemotherapeutic agents and inhibits the NF-κB pathway ^[20][29][30][82,91,92]. Therefore, it remains to be confirmed whether inhibition of the NF-κB pathway by triptolide leads to inhibition of cancer–neuronal crosstalk or whether other triptolide-induced pathways are responsible for the observed effects. Nevertheless, preliminary promising results on triptolide in PDAC are already available. For instance, in an orthotopic PDAC model, it was shown that the combined administration of Minnelide and paclitaxel significantly reduced tumor growth and improved survival ^[31][93]. In this model, the life expectancy in untreated animals was 13 days. After administration of paclitaxel or Minnelide as monotherapy, survival was significantly improved up to 21 days. By combined administration of Minnelide and paclitaxel, all animals were still alive after more than 6 weeks. Overall, compared to untreated animals, combined therapy reduced the rate of distant metastases by up to 90%. Based on these data, several clinical trials in PDAC (and other tumor entities) have recently been initiated (Table 1) ^[32][94].

Drug	Target	Current Status	Effect	Reference
Anti-NGF antibody or siRNA	NGF	Mouse model	Reduced neural outgrowth, reduced cancer growth, reduced metastasis	[16][19][54,55]
Trk-Inhibitor	Trk (unselective) or trkA (selective)	Phase I–II	NA	NCT03556228, NCT05046847, NCT02097810, NCT02568267, NCT04879121
Plerixafor	CXCR4 antagonist	Phase I–II	NA	NCT03277209, NCT02179970, NCT04177810
miR-383	ROBO3	Mouse model	Reduced cancer growth and metastasis	[33][56]
Minnelide	Several effects, including Inhibition des NF-κB	Phase I–II	NA	NCT03117920, NCT05557851, NCT03129139, NCT04896073
Propranolol	Beta adrenergic receptor	Phase II	NA	NCT03838029, NCT05451043
Celiac ganglion ablation	Denervation	Phase II–III	Reduction in cancer-associated pain, but no significant prolonged survival	[34][57]
Splanchnicectomy	Denervation	Phase III	Reduction in cancer-associated pain, prolonged survival only in presence of cancer-associated pain before intervention	[35][58]
Bethanechol	Muscarinic receptor M1	Phase I–II	NA	NCT03572283, NCT05241249

2.2. Pancreatic Ductal AdenocarcinomaAC Cells Are Attracted to Neurons

Chemokines are signaling molecules that exert the attraction and directional movement of leukocytes, as well as other cell types, including endothelial and epithelial cells. They are central mediators in the migration of cells, making it not surprising that several chemokines are involved in cancer–neuronal crosstalk ^[36][16]. Fractalkine (CX3CL1) is a well-studied chemokine released by neurons (Figure 2B) ^[37][95]. In the central nervous system, CX3CL1 plays an important role in neuron–glia crosstalk. Several PDAC cell lines express the corresponding receptor CX3CR1, which enables them to demonstrate PNI ^[37][95] (Figure 2B). CX3CR1 is not detectable in healthy human pancreatic tissue, whereas it is highly expressed in human PDAC tissue. Thereby, CX3CR1 expression correlates with the extent of NI, as well as local and early tumor recurrence ^[37][95]. The downstream mechanisms of CX3CR1 activation in PDAC need to be identified. However, it has been shown in prostate and breast cancer that the activation of the receptor leads to subsequent activation of several pathways, including PI3K/AKT, Raf/MEK/ERK and JAK/STAT, among others ^[38][96]. E6011 is a humanized IgG2 monoclonal antibody against human fractalkine, which has been used in phase II trials in inflammatory diseases such as rheumatoid arthritis. Clinical studies in PDAC have not yet been performed ^[21][83]. CXCL12 is also released from neurons and binds to the receptor CXCR4 (and CXCR7) on PDAC cells (Figure 24B) ^[21][83]. The activation of the CXCL12/CXCR4 axis significantly increased PDAC cells’ PNI and promoted neurite outgrowth. The expression of CXCR4 correlated highly significantly with the presence of NI in human specimens ^[39][97]. Inhibition of this pathway in vivo inhibited tumor growth and invasion of the sciatic nerve ^[39][97]. In an orthotopic PDAC mouse model, CXCR4 inhibition (using plerixafor) resulted in a significantly improved response to gemcitabine chemotherapy ^[40][98]. In human specimens, high CXCR4 expression but not high CXCR7 expression was associated with reduced overall survival ^[41][99]. While some gastric and colorectal carcinoma cells can produce CXCL12 themselves, the quantitative largest source in PDAC is derived from cancer-associated fibroblasts (CAFs) ^[42][100]. Physiologically, pancreatic stellate cells produce CXCL12 after tissue damage. In PDAC, cancer cells encourage CAFs to produce CXCL12 by secreting TNF-alpha, TGF-beta and others ^[42][100]. CXCL12 binding enhances PDAC proliferation via Akt, ERK and MEK signaling ^[43][101]. CXCL12 binding also increases treatment resistance through increased expression of pro-survival proteins Bcl-2, BclxL and Notch1 and inactivation of BAD ^[42][100]. In addition to its influence on cancer–neuronal crosstalk and tumor growth, the CXCL12/CXCR4 axis is also important for immune cell migration and possibly plays a role in modulating the response to immunotherapy ^[42][44][100,102]. Plerixafor is a selective CXCR4 antagonist approved to mobilize hematopoietic stem cells into the peripheral blood for collection and autologous transplantation in patients with non-Hodgkin lymphoma and multiple myeloma. Therefore, it is not surprising that CXCL12-CXCR4-axis also plays an important role in immune cell migration ^[42][44][100,102]. Based on the following findings, preliminary clinical studies are planned or underway. This will show the impact of the CXCL12–CXCR4 axis on cancer–neuronal crosstalk, tumor growth and response to immunotherapy. Another prominent group of proteins secreted by nerves is the glial-cell-derived neurotrophic factor family of ligands (GDNF). The GDNF family ligands include GDNF, neurturin, artemin and persephin. GDNF family members bind to glycosylphosphatidylinositol anchor-linked GDNF family receptor alpha 1 (GFRα1), which recruits rearranged during transfection (RET) receptor tyrosine kinase for dimerization ^[22][84]. GDNF is highly expressed in the peripheral and central nervous system and stimulates the development, survival and differentiation of neuronal cells ^[22][84]. RET is expressed on several PDAC cell lines ^[22][84]. Neurons from mice deficient in GDNF had a reduced ability to attract cancer cells ^[12][15]. It has been shown that PNI is dependent on neuronal GDNF secretion and GDNF coreceptors RET and GFRα1 expressed in human PDAC cells (Figure 2C) ^[22][84]. High RET expression in human PDAC was established as a negative prognostic parameter ^[6][8]. Interestingly, PNI was blocked by treatment with PYP1, a potent RET inhibitor, showing how targeting nerves could serve as a potential mechanism to decrease PDAC growth ^[12][15]. After GDNF stimulated the RET mitogen-activated protein kinase cascade, it facilitated matrix degradation by producing more matrix metallopeptidase 2 (MMP2), MMP9 and MMP14, thereby accelerating cell invasiveness ^[6][8]. Monoclonal antibody-GDNF fusion protein was tested in parkinsonian monkeys. Focal pancreatic acinar to ductular metaplasia with transition to pancreatic intraepithelial neoplasia 1B (PanIN-1B) lesions were detected in several animals ^[45][103]. Whether this was due to the specific antibody fusion protein or to GDNF inhibition itself remains to be elucidated. Recently, protein neuroligin 1 (NLGN1) was shown to promote cancer cell invasion and migration along nerves ^[46][104]. Functionally, NLGN1 was shown to exert its effect in cooperation with GDNF (NLGN1-GDNF cooperation). Therefore, NLGN1 inhibition is a promising target for clinical trials. Slit glycoproteins (Slit) and their roundabout receptors (Robo) are guide molecules in neuronal development, axon guidance, glial migration and angiogenesis ^[47][48][105,106]. Especially in PDAC, this pathway seems to play an important role in cancer–neuronal crosstalk and metastasis (Figure 24D). Thus, the expression of SLIT2 is reduced in PDAC tissue ^[23][85]. SLIT2 binds to ROBO1, which is expressed on neurons. Overexpression of SLIT2 in deficient PDAC cells reduces attraction by DRG neurons and thus PDAC migration along outgrowing neurons is inhibited ^[23][85]. Interestingly, even though the migration of PDAC cells towards neurons was reduced, their general motility was not affected. In vivo, restored SLIT2 expression was shown to reduce metastasis and vascularization ^[23][85]. Nevertheless, these findings remain to be confirmed, as the opposite effect was also observed for SLIT2-ROBO1 ^[49][107]. Inhibition of the WNT/β-catenin pathway may thus be the main mechanism by which the Slit/Robo pathway inhibits pancreatic cancer growth. Specifically, inhibition of the WNT/β-catenin pathway by Slit2/Robo signaling enhances the formation of β-catenin and E-cadherin complexes, increasing tumor cell adhesion and inhibiting tumor invasion and migration, thereby improving patient prognosis ^[24][86]. Interestingly, high expression of ROBO3, a known inhibitor of ROBO1/2 signaling, was associated with shorter survival in a cohort of 142 PDAC patients undergoing pancreatectomy with curative intent ^[33][56]. ROBO3 increases with clinical grade of PDAC and promotes cancer cell growth and metastasis in vitro and in vivo ^[50][108]. Robo3 can activate the WNT/β-catenin pathway, thus promoting pancreatic cancer growth and invasion ^[24][86]. Thus, the Slit-Robo pathway represents a very interesting target in PDAC. For example, miR-383 was identified as a suppressor of ROBO3 ^[50][108]. Besides its role in cancer–neuronal crosstalk, the SLIT-ROBO pathway is involved in other mechanisms of tumor growth and metastasis ^[24][86]. Another axon guidance molecule is Semaphorin 3D, which was found to be overexpressed in tumorigenesis ^[25][87]. Migration and PNI of PDAC is increased via secretion of Semaphorin 3D by pancreatic cells and activation of Plexin D1 on neurons (Figure 24E). Knockdown of Semaphorin 3D and loss of neural Plexin D1 reduces neurite outgrowth and metastasis in vivo ^[25][87]. Furthermore, high mRNA expression of Semaphorin 3D and Plexin A1, another molecule central to semaphorin signaling, were both associated with poor patient survival ^[24][86]. Secretion of Semaphorin 3D is increased by AnnexinA2. Annexin A2 is a metastasis-associated protein in PDAC that has been shown to be essential for the metastatic growth in genetically engineered spontaneous pancreatic tumor-producing KPC mice. By comparing tumor cells of AnnexinA2 wild type vs. knock-out KPC mice, Semaphorin 3D and PlexinD1 were among the most differentially expressed genes. Thus, disrupting the AnxA2/Sema3D/PlexinD1 signaling appears to be a promising therapeutic strategy for further clinical trials ^[25][87]. The complex signaling mechanisms of Semaphorin 3D are only partially understood. Interestingly, in non-PDAC cancer cells, the prometastatic activity of Semaphorin 3E is mediated by transactivation of PlexinD1-associated Erb2b (Figure 24E) ^[26][88]. Considering the PDAC environment is characterized by poor nutrition supply, it is becoming evident that PDAC cells are dependent on neurons to receive external growth stimuli ^[51][109]. Specifically, neurons secrete mediators stimulating cancer growth and supplying oxygen and nutrients ^[36][52][53][16,110,111]. For instance, amino acids are essential for PDAC growth. Serine is a conditionally essential amino acid that represents the second most abundant amino acid found in human proteins and can be secreted by neurons ^[27][89]. Serine is necessary for several metabolic pathways, highlighting its role in terms of tumor growth and survival in cancers ^[54][112]. For instance, human PDAC cell lines are completely dependent on exogenous serine (Figure 24F). If neurons are prevented from producing serine, tumors grow approximately 40% smaller ^[51][109]. Interestingly, after serine deprivation, PDAC cells massively overexpress NGF in order to recruit more neurons to meet its nutritional needs ^[51][109].