Prominent crosstalks between perineural invasion and immunosuppressive tumor microenvironment in pancreatic ductal adenocarcinoma

Abstract

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies worldwide, with a 5-year overall survival of about 12%. Perineural invasion (PNI), documented in most surgical specimens, is a common hallmark of PDAC that drives cancer-associated pain, recurrence, and resistance to therapy. Far from a passive histological finding, PNI reflects an active, bidirectional program shaped by dynamic crosstalk between tumor cells, nerves, and the immunosuppressive tumor microenvironment. Cancer-associated fibroblasts, tumor-associated neutrophils, tumor-associated macrophages, and myeloid-derived suppressor cells each make distinctive contributions to the immunosuppressive perineural niche. We further delineate how PNI facilitates neural route metastasis, drives chemotherapy resistance through neurotrophic and immune mechanisms, and reprograms cellular metabolism in ways that concurrently support tumor invasion and impair anti-tumor immunity. Finally, we highlight promising therapeutic strategies targeting the neural-immune axis, including β-adrenergic blockade, neurotrophic receptor inhibition, immune checkpoint combination approaches, and metabolic interventions. These strategies may hold transformative potential for the management of PNI-positive PDAC.

Key Words: Pancreatic ductal adenocarcinoma; Perineural invasion; Tumor immune microenvironment; Metastasis; Chemotherapy

Core Tip: Perineural invasion is a bidirectional program in pancreatic ductal adenocarcinoma. It is sustained by crosstalks between tumor cells, nerves, and the immunosuppressive tumor microenvironment. The cellular compartments in tumor microenvironment collectively establish an immunosuppressive perineural niche. This niche enables neural dissemination, chemotherapy resistance, and cancer-associated pain. Adrenergic and cholinergic signaling amplifies this immunosuppression. Schwann cells actively guide tumor migration along nerve trunks. Therapeutic strategies targeting the neural-immune axis hold transformative potential for pancreatic ductal adenocarcinoma management.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDAC) is the main histopathological type of pancreatic cancer. It represents one of the most lethal malignancies worldwide with a 5-year overall survival rate around 12%[1]. According to GLOBOCAN 2022, pancreatic cancer brings approximately 511000 new cases and 467000 deaths annually. It ranks as the twelfth most common cancer but the sixth leading cause of cancer-related mortality globally[2]. PDAC is highly lethal by virtue of its covert and aggressive biology. Patients typically present with nonspecific symptoms, most commonly abdominal or back pain, unintentional weight loss, and painless jaundice. These symptoms often emerge only after locoregional invasion or distant metastasis[3]. PDAC represents a disease of multifactorial etiology. Known risk factors of PDAC comprise cigarette smoking, obesity, type 2 diabetes, chronic pancreatitis, and excessive alcohol use[4]. Approximately 5% of PDAC cases are attributable to inherited genetic predisposition, including germline mutations in BRCA2, BRCA1, PALB2, ATM, CDKN2A, and DNA mismatch repair genes[5]. In contrast, oncogenic KRAS mutations are the defining somatic driver. Hereditary risk assessment is now advised in current guidelines for every newly diagnosed patient[6]. Its dismal prognosis is also attributable to resistance to chemotherapy, which remains the cornerstone in PDAC treatment[7]. Unlike others malignancies, where management strategies have expanded greatly, PDAC has seen little progress. Targeted therapies and immunotherapies have largely failed to demonstrate meaningful benefits in PDAC patients. PDAC treatment is hence mainly confined to cytotoxic chemotherapy. In the metastatic setting, FOLFIRINOX (oxaliplatin, irinotecan, leucovorin, fluorouracil) and gemcitabine combined with nab-paclitaxel constitute the first-line regimens of choice. However, the survival benefit over gemcitabine monotherapy is only modest[8,9].

PDAC is a devastating diagnosis. Median overall survival remains below 12 months for metastatic PDAC patients, and the majority of patients experience disease progression within months of treatment[10]. For resectable PDAC, modified FOLFIRINOX serves as the recommended adjuvant regimen after surgery. However, the recurrence rate is high[11]. Unrelenting pain is a clinically prominent feature of PDAC, which severely impairs patient quality of life and frequently serves as the initial presenting symptom[12]. This reflects the predilection of PDAC to invade nerves, which is known as perineural invasion (PNI)[13]. Studies have documented PNI in about 80% of PDAC surgical resection specimens without neoadjuvant chemoradiation, underscoring its prevalence in this disease[14].

Recent research has illuminated several aspects of PNI in PDAC that extend beyond its established roles in pain generation and local spread. First, cancer cells and stromal cells appear capable of reprogramming tumor metabolism within the perineural niche[15]. Second, and of direct clinical relevance, PNI is increasingly recognized as an active contributor to chemotherapy resistance, a particular challenge given that gemcitabine-based regimens remain the cornerstone of PDAC treatment[16]. The TME of PDAC harbors a coordinated network of immunosuppressive cellular components that sustain PNI. Cancer-associated fibroblasts (CAFs) generate a dense desmoplastic stroma and secrete neurotrophic factors that support tumor-nerve crosstalks[17]. Tumor-associated macrophages (TAMs) are enriched around invaded nerves, where they produce glial cell line-derived neurotrophic factor (GDNF) to drive RET-mediated neural invasion and reciprocally activate Schwann cells[18,19]. Tumor-associated neutrophils (TANs) and myeloid-derived suppressor cells (MDSCs) promote immunosuppression around the perineural niche through arginase-1-mediated arginine depletion, effectively excluding cytotoxic T cells[20,21]. Autonomic signaling through sympathetic nerves and parasympathetic nerves is superimposed on this cellular network, reinforcing both immunosuppression and neural remodeling[22]. In this review, we systematically dissect how these neural and immune components converge at the perineural niche to sustain PDAC progression, and highlight emerging therapeutic strategies that target this integrated neural-immune axis.

PNI IN PDAC

PNI is defined as the presence of tumor cells within any of the three layers of the nerve sheath, including the epineurium, perineurium, or endoneurium. Tumor cells tracking along nerve trunks irrespective of sheath penetration are also classified as PNI[23]. First systematically described in head and neck carcinomas, PNI is now a well-characterized pathological feature associated with poor prognosis across multiple cancer types including prostate, colorectal, biliary, and pancreatic cancers. PNI is not a bystander phenomenon of cancer. Indeed, recent studies demonstrate that it is an active process driven by crosstalks between tumor cells, neural elements, and the surrounding TME[24]. Recent research has shown three additional facets of PNI in PDAC besides pain. First, PNI has been implicated in facilitating metastasis as an alternative route[25]. Second, PNI appears to reprogram tumor metabolism, altering metabolism in the perineural niche[26]. Third, PNI is increasingly recognized as a contributor to drug resistance, such as chemotherapy resistance.

The peripheral nervous system comprises the somatic and autonomic nerves. They are both recognized as active participants in tumor. The pancreas is innervated by an autonomic network derived from the celiac plexus[27]. This neural network, which regulates exocrine and endocrine pancreatic function, is profoundly remodeled during PDAC development. A recent study shows this remodeling process as a multistage program encompassing neurogenesis, neurite outgrowth, axonogenesis, and synapse-like structure formation, each of which is influenced by distinct tumor-derived neurotrophic cues[28]. The nerve sheath includes epineurium, perineurium, and endoneurium. The perineurium is the outside barrier. Its disruption by invading cancer cells enables tumor spread along the nerves for a long distance[29]. The perineural space provides a physical route for dissemination. In addition, this is a nutrient-rich and immunologically privileged space that is suitable for cancer cell growth. Studies have revealed strikingly divergent roles for the sympathetic and parasympathetic branches of the autonomic nervous system (ANS) across cancer types. For example, Magnon et al[30] demonstrated in prostate cancer that sympathetic innervation promotes early tumor development, whereas parasympathetic innervation drives tumor metastasis. This established a dichotomy that has since been explored in other solid tumors including PDAC.

In PDAC, adrenergic signaling through β-adrenergic receptors on tumor cells activates the cyclic AMP and extracellular signal-regulated kinase pathways, thereby supporting tumor progression[31]. In addition, chronic stress-induced catecholamine release has been linked to accelerated pancreatic tumor growth in preclinical models. β-blocker use has been associated with improved outcomes in patients with PDAC[32]. Autonomic signaling appears to exert context-dependent effects. Studies indicate that intact vagal activity restrains pancreatic tumor growth, whereas its disruption through vagotomy accelerates tumor progression and worsens survival[33]. The context-dependent neural effects are mediated, in part, through modulation of the TME. β-adrenergic signaling could suppress macrophage tumor necrosis factor-alpha production, contributing to an anti-inflammatory shift in cytokine output[34]. These findings identify the ANS as an upstream regulator of the TME in PDAC, with PNI serving as a key neuroimmune interface (Figure 1). Oncogenic KRAS mutations are present in most PDAC cases and represent the dominant initiating genetic event in pancreatic carcinogenesis[35]. In genetically engineered PDAC mouse models, multiple neurotrophic factors and their receptors are upregulated in the pancreas as early as the pancreatic intraepithelial neoplasia stage. Neurotrophic factors and their receptors are important contributors in early phase of PDAC development[36]. The perineural niche further supports invasion through recruitment of myeloid cells that degrade neural barriers.

Figure 1 Bidirectional crosstalk between autonomic nerves and key cellular components of the perineural tumor microenvironment in pancreatic ductal adenocarcinoma. Nerve terminals release norepinephrine, neuropeptides, and acetylcholine to modulate tumor-associated macrophages, tumor-associated neutrophils, myeloid-derived suppressor cells, cancer-associated fibroblasts, and cancer cells, collectively promoting immunosuppression, neural remodeling, and perineural invasion. Reciprocally, cellular compartments and associated Schwann cells secrete neurotrophic factors including nerve growth factor, brain-derived neurotrophic factor, glial cell line-derived neurotrophic factor, and artemin that sustain axonal sprouting and reinforce the pro-invasive perineural niche. TAMs: Tumor-associated macrophages; TANs: Tumor-associated neutrophils; MDSCs: Myeloid-derived suppressor cells; CAFs: Cancer-associated fibroblasts; NGF: Nerve growth factor; BDNF: Brain-derived neurotrophic factor; GDNF: Glial cell line-derived neurotrophic factor; MMP: Matrix metalloproteinase; NETs: Neutrophil extracellular traps; PGE2: Prostaglandin E2; TNF-α: Tumor necrosis factor-α.

Inflammation and PNI

Inflammation has for long been implicated in PDAC development. That said, it’s not a clear-cut relationship. So is the relationship between inflammation and PNI. Chronic pancreatitis is a well-established risk factor for PDAC and PDAC can manifest as pancreatitis in some cases. In both PDAC and chronic pancreatitis, cytotoxic T lymphocytes, macrophages, and mast cells coexists within PNI spaces and correlates with pain[37]. Inflammatory mediators like prostaglandins, bradykinin, and other cytokines can directly activate peripheral nerves[38]. This inflammatory milieu might contribute to tumor evasion. In addition, cancer-associated pain in PDAC is closely linked to inflammation. Pancreatic stellate cells (PSCs) activated by TAMs can increase matrix metalloproteinase (MMP)-2, nerve growth factor (NGF) expression to increase PNI[39]. NGF then sensitizes primary nociceptors by engaging tropomyosin receptor kinase A (TrkA)-coupled phospholipase C signaling, which depletes membrane phosphatidylinositol 4,5-bisphosphate and disinhibits transient receptor potential vanilloid 1[40].

Together, NGF-driven neuroinflammation may generate pain and sustain a feedforward loop in which activated sensory neurons release substance P and calcitonin gene-related peptide into the peritumoral tissue, promoting local inflammation. The inflammatory cytokine interleukin (IL)-6 deserves particular mention given its pleiotropic roles in both inflammation and PNI[41]. Leukemia inhibitory factor (LIF), a member of the IL-6-gp130 cytokine family, is secreted by stromal cells in the tumor microenvironment (TME) and activates Janus kinase (JAK)-signal transducer and activator of transcription 3 signaling in Schwann cells, promoting their migration and differentiation as well as plasticity, thereby driving neural remodeling[42]. Neurotransmitters and neuropeptides in the TME directly affect cancer cell behavior and immune function. Norepinephrine and epinephrine released from sympathetic nerve terminals activate β2-adrenergic receptors on PDAC cells, promoting proliferation and acinar-to-ductal metaplasia. Adrenergic signaling increases NGF in PDAC cells. This promote axonogenesis and sympathetic nerve growth[22].

Acetylcholine is the primary parasympathetic neurotransmitter. The engagement of acetylcholine and its receptors modulates PDAC tumorigenesis and progression[43]. Neuropeptide Y and substance P also influence PDAC development[44]. Pain is the main clinical manifestation of PNI in PDAC. It reflects structural and functional changes in pancreatic nerves remodeled by cancerous cells. Histological studies show increased fiber diameter and increased neural area in PNI sites[45]. This is also driven by neurotrophic factors, which lead to axonal sprouting and recruitment of neural progenitor cells[46]. In mouse models, NGF drives pathological sprouting of sensory and sympathetic nerve fibers and neuroma formation[47]. Mechanistically, neural injuries like PNI might lead to an increase in functions of nociceptive pathways, which contribute to pain[48]. Clinically, patients with painful, inoperable pancreatic cancer may benefit from early celiac plexus neurolysis, which could relieve severe pain and moderate opioid consumption[49].

Neurotrophic factors in PNI

NGF plays a central role in PNI of PDAC. NGF is a member of the neurotrophin family, which signals primarily through the TrkA and the p75 neurotrophin receptor[50]. In PDAC, NGF and TrkA are overexpressed, and their elevated expression correlates with increased PNI and pain in PDAC patients[51]. NGF binding to TrkA activates downstream signaling cascades, including phosphoinositide 3-kinase-AKT pathway[52]. In addition to NGF, glial cell-derived neurotrophic factor (GDNF)-RET signaling has also been implicated in PNI of PDAC. Mechanistically, GDNF secreted by peripheral nerves acts as a chemoattractant that drives pancreatic cancer cells toward neural structures[53]. Similarly, artemin, another member of the GDNF family, signals through the GFRα3/RET receptor complex and is markedly overexpressed in PDAC tissues, with protein levels increased up to 30-fold compared with normal pancreas. Artemin localizes to nerves, arterial walls, and cancer cells, and promotes pancreatic cancer cell invasion and chemotaxis in a dose-dependent manner[54].

CAFS AND PNI

CAFs are the dominant stromal cell type in PDAC. The desmoplastic nature of PDAC is largely attributed to CAFs. They form a heterogeneous population with distinct functional roles in tumor[55]. CAFs have been classified into inflammatory CAFs (iCAFs) and myofibroblastic CAFs, which exhibit distinct spatial distributions within the PDAC stroma. Myofibroblastic CAFs localize adjacent to tumor cells and are shaped by transforming growth factor-beta (TGF-β) signaling, whereas iCAFs reside away in the stroma and are induced by tumor-secreted IL-1 through JAK/STAT activation[56]. CAFs, largely derived from activated PSCs, secrete NGF and other neurotrophic factors that contribute to neural remodeling and PNI in pancreatic cancer. ICAFs shape the immune composition of the perineural niche. They secrete IL-6 and LIF, factors that profoundly remodel the TME. In addition, CAFs reprogram sympathetic nerve transcriptional programs. Schwann cells, the glial cells that myelinate peripheral nerve fibers, actively interact with CAFs in the PDAC TME. These cells drive CAFs to more malignant iCAFs[57]. Direct contact between Schwann cells and PDAC cells guides tumor cells into the perineural space, which is dependent on neural cell adhesion molecule 1[58]. Schwann cells intercalate between cancer cells, forming tumor-activated Schwann cell tracks[59] (Figure 2).

Figure 2 Activated Schwann cells serve as central orchestrators of perineural invasion in pancreatic ductal adenocarcinoma. They secrete neurotrophins including nerve growth factor, brain-derived neurotrophic factor, and glial cell line-derived neurotrophic factor to drive neural remodeling, while matrix metalloproteinase-2, matrix metalloproteinase-9, and L1-cell adhesion molecule degrade the perineural extracellular matrix to facilitate tumor cell entry. Schwann cell-derived CCL2, CXCL12, and interleukin-6 further recruit immune cells and guide tumor cell migration along nerve trunks toward the celiac plexus. NGF: Nerve growth factor; BDNF: Brain-derived neurotrophic factor; GDNF: Glial cell line-derived neurotrophic factor; NT-3: Neurotrophin-3; IL: Interleukin; TGF-β: Transforming growth factor-β; TNF-α: Tumor necrosis factor-α; PNI: Perineural invasion; ECM: Extracellular matrix; MMP: Matrix metalloproteinase; CAM: Cell adhesion molecule.

IMMUNE CELL INTERACTIONS WITH PNI IN PDAC TME

Interactions between PNI and TANs

TANs have also been identified as important participants in PDAC progression. Due to the limits in isolation and enrichment, neutrophils have long been viewed as short-lived innate immune effectors. Advances in single-cell technologies have since profiled neutrophils as highly plastic cells that can adopt various phenotypes within the TME[60]. In PDAC, TANs constitute a significant component of the immune infiltrate, and represent independent prognosticators of poor overall survival and disease-free survival[61]. TANs in cancer have been conceptually divided into anti-tumorigenic N1 and pro-tumorigenic N2 phenotypes. The balance between these states is determined by the cytokine milieu of the TME[62]. In cancers, the dominant immunosuppressive environment strongly favors N2 polarization. Key drivers include TGF-β, IL-10, which endow N2 TANs with pro-angiogenic, pro-invasive, and immunosuppressive properties[60]. N2 TANs infiltration is also correlated with more regulatory T cell infiltration and less CD8+ T cell infiltration[61].

Mechanistically, N2 TANs and myeloid cells express high levels of arginase-1, depleting local arginine and suppressing T cell receptor signaling in nearby cytotoxic lymphocytes[62]. N2 TANs also produce neutrophil extracellular traps (NETs), which are web-like chromatin structures decorated with antimicrobial proteins, that impair natural killer (NK) cell and T cell function by obstructing contact[63]. NET formation has been shown to be associated with poor adjuvant chemotherapy outcomes in PDAC[64]. TANs might contribute to neural remodeling and PNI through multiple mechanisms. Given that neutrophil-derived MMP-9 and MMP-8 have been shown to modify extracellular matrix components and activate TGF-β within tumor stroma, it is plausible that similar proteolytic activities exist within the perineural niche[65]. Neutrophil elastase degrades E-cadherin on PDAC cells, promoting epithelial-to-mesenchymal transition (EMT) and enhancing their migratory capacity, which may in turn facilitate tumor cell spread toward neural elements[66].

Activated neutrophils release hepatocyte growth factor, oncostatin M, and pro-angiogenic factors including vascular endothelial growth factor-A and angiopoietin-1[67-69]. These factors might promote axonal sprouting, increase vascular permeability at the invasive front, and sustain the neurotrophic environment that maintains PNI. For example, hepatocyte growth factor/c-Met pathway promotes PNI through mammalian target of rapamycin/NGF axis in PDAC[70]. NETs contain a toxic cargo of histones, neutrophil elastase, cathepsin G, and myeloperoxidase[71]. Their deposition around nerve bundles may contribute to the proteolytic environment that facilitates PNI. NET-derived components have been shown to injure peripheral nerves and contribute to mechanical hyperalgesia in models of neuropathy. Analogous mechanisms may operate in the perineural niche of PDAC, where inhibition of NETs might reduce PNI-associated pain through improving peripheral microcirculation[72]. NETs induce proinflammatory cytokines through multiple mechanisms[73]. This inflammatory state might induce neural remodeling and promote further PNI. Pharmacological disruption of NET formation through IL-17 inhibition has shown effects against PDAC in preclinical models[74].

Interactions between PNI and TAMs

TAMs represent one of the most abundant immune cell populations within the PDAC TME, and their density correlates with poor prognosis[75,76]. In PDAC, TAMs are predominantly polarized toward the M2-like immunosuppressive phenotype, characterized by high expression of CD68, CD163, and CD204[77]. Their density within the tumor correlates significantly with PNI frequency, lymph node metastasis, and shortened overall survival. The perineural niche is particularly enriched in TAMs, which contribute to neural remodeling and the immunosuppressive microenvironment surrounding invaded nerves[18]. The selective enrichment of TAMs in perineural regions reflects a complex chemotactic signaling network. CCL2, produced abundantly by PDAC cells, is the dominant monocyte chemoattractant driving TAMs recruitment[78]. Colony-stimulating factor 1 signals through CSF-1R on monocytes also drive TAMs recruitment within the PDAC stroma[79]. Neural elements actively participate in TAMs recruitment through neurotransmitter release. Sympathetic nerve-derived catecholamines have been shown to induce TAMs recruitment through monocyte chemotactic protein 1 in other cancers[80]. Substance P, released from activated sensory afferents, binds NK-1 receptor expressed on PDAC cells and promotes their migration, invasion, and perineural spread[81,82]. Given that NK-1 receptor is also expressed on macrophages, substance P may also contribute to local TAM recruitment and polarization within the perineural niche, although direct evidence in PDAC is currently lacking[83]. In addition, the CXCL12/CXCR4 axis is a well-characterized driver of stromal-immune crosstalk in PDAC. CXCL12 is predominantly produced by PSCs, and its blockade has shown therapeutic potential in combination with immune checkpoint inhibition[84]. In this study, spatial analysis showed closer proximity between TAMs and T cells. Once recruited to the perineural niche, TAMs function as important producers of neurotrophic factors. TAMs surrounding nerves invaded by PDAC cells express high levels of GDNF, which acts on phosphorylation of RET to drive PNI[18]. GDNF binding to the RET complex activates downstream extracellular signal-regulated kinases signaling in PDAC cells. In addition, co-culture of macrophages with fibroblasts markedly enhances stromal secretion of LIF, which could activate JAK/signal transducer and activator of transcription 3 signaling in Schwann cells and ganglia neurons[18].

TAMs also contribute to neural remodeling indirectly through reciprocal crosstalk with Schwann cells. Mechanistically, TAMs activate Schwann cells via the basic fibroblast growth factor/phosphoinositide 3-kinase/AKT/c-myc/glial fibrillary acidic protein pathway. Schwann cells secrete IL-33 to recruit macrophages into the perineural milieu and facilitate the M2 polarisation of macrophages[19]. Beyond driving neural remodeling, TAMs enforce local immune evasion at the perineural niche through multiple signaling pathways. TAMs in PDAC express high levels of arginase-1, which suppress CD8+ T cell activation[85]. TAMs also secrete tumor necrosis factor-alpha, which upregulates programmed death-ligand 1 expression in cancer cells, driving T cell exhaustion[86].

Interactions between PNI and MDSCs

MDSCs are a heterogeneous population of immature myeloid cells that represent one of the most potent mechanisms of immune evasion in PDAC[87]. MDSCs are broadly divided into polymorphonuclear MDSCs and monocytic MDSCs. Both subsets are markedly expanded in PDAC patients[88,89]. Their abundance correlates significantly with disease stage and poor clinical outcomes in PDAC patients, and has been associated with other markers of aggressive tumor biology including PNI[21,90]. Granulocyte-macrophage colony-stimulating factor produced by oncogenic KRAS-expressing PDAC cells is the dominant driver of MDSC expansion. It acts on myeloid progenitors to promote generation of MDSCs[91,92]. Vascular endothelial growth factor, IL-1β and IL-4 within the PDAC TME cooperate with granulocyte-macrophage colony-stimulating factor to sustain this expansion[93]. Sympathetic signaling may further contribute to MDSCs recruitment, as norepinephrine released from sympathetic terminals has been shown to increase MDSC frequency and survival through β2-adrenergic receptor signaling in murine tumor models[94].

PDAC cells express β2-adrenergic receptor, raising the possibility of a neural-tumor-MDSC axis that may exist[95]. Once recruited to the PDAC perineural niche, MDSCs might exert their immunosuppressive functions through multiple mechanisms that converge to impair antitumor T cell responses. First, arginase-1 expressed by MDSCs depletes local L-arginine, preventing T cell receptor zeta-chain expression required for T cell response[96]. Second, nitric oxide produced by MDSCs mediates CD8+ T-cell suppression[97]. Finally, MDSCs additionally promote regulatory T cell differentiation through secretion of TGF-β and IL-10[98].

CROSSTALKS BETWEEN PNI AND TME IN PDAC METASTASIS

Besides hematogenous and lymphatic dissemination, PNI provides another alternative metastatic route for metastasis. In PDAC, cancer cells exploit the autonomic nerve network in and around pancreas and the final destinations are celiac plexus and retroperitoneum. Centripetal spread along autonomic nerve trunks has also been documented in surgical specimens. In such cases, tumor cells have been detected within nerve fascicles located far from the primary lesion[99]. The appearance of PNI in the celiac plexus is correlated with invasion of margins, demonstrating higher malignancy[100]. The TME actively enables this neural dissemination. First of all, TAMs, along with other myeloid-derived immunosuppressive populations, contribute to maintaining the immunosuppression at perineural niche[101]. EMT further equips perineural tumor cells with upregulated programmed death-ligand 1, rendering them more resistant to immune surveillance[102]. Targeting EMT-driving pathways represents a promising therapeutic strategy. For example, combined cyclin-dependent kinase 4/6 and BET inhibition has been shown to synergistically suppress PDAC growth and EMT by modulating glycogen synthase kinase 3β-mediated Wnt/β-catenin signaling[103]. It offers a potential avenue to disrupt the metastatic competence of perineural tumor cells.

ARTIFICIAL INTELLIGENCE IN CANCER PNI RESEARCH

Artificial intelligence (AI) has been transforming the detection and assessment of many malignancies including PDAC. Histopathological identification of PNI is costly in time and labor. Interindividual variability and limit of tissue sectioning planes may stymie recognition of neural invasion. Deep learning-based new methods have improved accuracy in the assessment of Gleason score in prostate cancer specimens[104]. Borsekofsky et al[105] developed a PDAC-specific PNI detection algorithm trained on 260 manually annotated nerve and tumor regions from scanned whole-slide images of 6 PDAC cases. In a cohort of 59 patients, this algorithm raised PNI detection from 52% to 81% of cases while reducing pathologist review time to an average of 24 seconds per case.

Beyond detection, AI models have been used to integrate histological, genomic, and clinical data to predict PNI status preoperatively from computed tomography (CT) and magnetic resonance imaging (MRI). Radiomics-based machine learning pipelines extract quantitative imaging features from pancreatic protocol CT scans that correlate with PNI confirmed on surgical pathology. For example, Sun et al[106] extracted 851 radiomic features from preoperative contrast-enhanced CT images of 167 PDAC patients and developed a combined radiomic model that achieved area under the receiver operating characteristic curve values of 0.945 and 0.881 in the training and validation cohorts for predicting PNI status. In a larger multicenter study, Yu et al[107] developed a fully automated deep learning model trained on preoperative CT scans from 1065 PDAC patients across two hospitals to predict extrapancreatic PNI. The model achieved area under the curve values of 0.83 in external test cohort. AI-predicted PNI status was also independently associated with patient survival. These models may assist in surgical planning and patient stratification prior to resection. Within the TME, spatial transcriptomics combined with AI-driven cell type deconvolution has begun to map the cellular architecture of the perineural niche at single-cell resolution. For example, Chen et al[108] performed integrated single-cell RNA sequencing and spatial transcriptomics on 62 samples from 25 PDAC patients, revealing that NLR family pyrin domain containing 3+ macrophages and cancer-associated myofibroblasts specifically surround invaded nerves in highly PNI tumors, while tertiary lymphoid structures co-localize with non-invaded nerves in low PNI tissues.

CURRENT STATUS AND FUTURE PERSPECTIVES

PDAC is refractory in nature. Understanding PNI in PDAC will bring new insights in this regard. PNI is not as a passive histological finding. Indeed, PNI is an active, multicellular program shaped by crosstalk between tumor cells, neural compartments, and the TME. Despite this mechanistic progress, meaningful clinical translation remains limited. Existing experimental models fail to fully recapitulate the immune and stromal complexity of the human perineural niche, and this gap constrains the preclinical evaluation of neural-immune targeting strategies. Several therapeutic directions show promise (Figure 3). β-adrenergic blockade is in clinical trials. The phase II randomized PROSPER trial evaluated perioperative propranolol combined with the cyclooxygenase-2 inhibitor etodolac in resectable PDAC patients undergoing pancreatoduodenectomy. Although the trial was prematurely closed due to slow recruitment, the treatment arm showed a markedly lower rate of distant recurrence (11.1% vs 54.5% in placebo) and prolonged median disease-free survival (16.4 months vs 11.3 months).

Figure 3 Schematic illustration of targeting the neural-immune axis in perineural invasion-positive pancreatic ductal adenocarcinoma. Neural-targeted therapies including β-adrenergic blockade, tropomyosin receptor kinase A/nerve growth factor inhibition, and RET/glial cell line-derived neurotrophic factor inhibition suppress sympathetic-driven immunosuppression, axonal sprouting, and Schwann cell-directed tumor invasion. Combined stromal-myeloid targeting and immunotherapy disrupts the immunosuppressive perineural niche by depleting myeloid-derived suppressor cells and tumor-associated neutrophils, activating dendritic cell, and stimulating CD8+ T cells at the invasive neural front. TrkA: Tropomyosin receptor kinase A; NGF: Nerve growth factor; PNI: Perineural invasion; TAM: Tumor-associated macrophage; ECM: Extracellular matrix; GDNF: Glial cell line-derived neurotrophic factor; CAF: Cancer-associated fibroblast; PDAC: Pancreatic ductal adenocarcinoma.

Neurotrophic receptor inhibition targeting the NGF-TrkA and GDNF-RET axes has demonstrated efficacy in preclinical models and warrants prospective evaluation in PNI-positive disease. Selpercatinib, a selective RET inhibitor, has received tissue-agnostic FDA approval and demonstrated a 54.5% objective response rate in the RET fusion-positive pancreatic adenocarcinoma subgroup (n = 11) of the LIBRETTO-001 basket trial, although RET fusions are rare in PDAC (< 1%), limiting broader applicability[109]. The limits of these trials include small trial size and lack of PNI-specific patient stratification. AI-driven pathology analysis is beginning to enable reproducible, quantitative assessment of PNI extent and perineural immune composition from routine diagnostic slides. Realizing the clinical potential of these advances will require prospective trial designs that stratify patients by PNI status and perineural immune phenotype. Integration of AI-derived neural-immune biomarkers into patient selection represents a tractable near-term objective. Progress will depend on sustained interdisciplinary collaboration between oncologists, neuroscientists, immunologists, and computational biologists united by a shared recognition that the neural dimension of PDAC is not peripheral to its biology but central to it.

CONCLUSION

PNI in PDAC is not a passive histological finding. It is sustained by coordinated crosstalks between tumor cells and the TME. CAFs, TANs, TAMs, and MDSCs each make different contributions to the perineural niche. Together, they create an environment that enables neural dissemination, therapeutic resistance, and cancer-associated pain. The ANS including adrenergic and cholinergic signaling renders the perineural space sheltered for cytotoxic immunity. This understanding might bring therapeutic strategies that target the neural-immune in a combinatorial way. AI-driven tools for perineural immune profiling can also facilitate research in PNI-positive PDAC. Future research should put more emphasis on new technologies such as spatial transcriptomics, single-cell RNA sequencing, and AI-assisted image analysis. These new technologies will greatly improve the landscape of PDAC management. Targeting PNI in PDAC holds considerable beneficial potential.

ACKNOWLEDGEMENTS

We would like to thank the kind support from the residency training base of pathology in Hangzhou First People’s Hospital and Director Qi Shen of Hangzhou Cancer Hospital for this work.