Reconceiving Perineural Invasion in Cutaneous Squamous Cell Carcinoma: From Biological to Histopathological Assessment

Cutaneous squamous cell carcinoma (cSCC) is the second most prevalent form of non-melanoma skin cancer, accounting for approximately 20% of all cutaneous malignancies on a global scale [1, 2], with an annual incidence of about 100 new cases per 100,000 individuals in the USA [3, 4]. Most cSCCs are diagnosed without evidence of further dissemination, allowing radical removal, and have excellent prognostic results with a 5-year survival exceeding 90% [5]. Although most cSCCs fall within the low-risk category, high-risk variations can have metastatic rates as high as 37% [6]. Among the risk factors of tumor-related recurrence, metastasis, and disease-specific mortality [7], several studies have highlighted perineural invasion (PNI) as an indicator of heightened malignant potential and tumor-specific mortality [8]. PNI entails the identification of the proximity of tumors to nerve fibers, encompassing invasion of nerve sheaths across their three layers [9]. The underlying mechanism of PNI involves a mutual attraction between the tumor and nerves, facilitated by humoral factors, such as neurotransmitters, cytokines, and growth factors [9‒11]. However, the precise pathogenesis of PNI remains predominantly undisclosed, and therapies targeting nerve invasion are presently lacking.

The involvement of nerve structures in cancer has multiple significant nuances, which is why it is important to distinguish different subtypes of neural involvement. PNI can be classified into two categories: clinical PNI (cPNI) and incidental PNI (iPNI). cPNI is defined as evidence of spread along large-caliber nerves with clinical evidence, or radiological demonstration, with magnetic resonance imaging recognized as the most sensitive imaging technique for detection [12]. iPNI is defined as invasion of nerves identified by histological examination and represents small caliber nerve involvement. It is estimated that between 60% and 70% of patients with PNI present with incidental findings [13], as the perineurium is a protective multilayered barrier, so cancer cells more easily invade smaller nerves with a thinner perineurium [14]. In this regard, a recent trend divides perineural tumor spread into two different processes: PNI and perineural spread (PNS). The former is a process of small microscopically identified peripheral nerves in the immediate vicinity of the invasive neoplasm and is related to iPNI. The latter involves larger, typically called central, nerves, and it is more frequently associated to cPNI. The two processes undergo different molecular processes [15]. PNI detection by the histopathological assessment of whole histological samples remains a major challenge and is subject to a wide variability of reported incidence rates that, for cSCC, range between 2.5 and 14% [16]. The high variance can be explained in part because PNI can involve both small and large nerves with sporadic distribution patterns in sometimes very large tissue samples, making accurate pathological assessment a time-consuming and tedious task. PNI in cSCC is also more frequently observed in male patients, recurrent and facial tumors, tumors with poorly differentiated histology, deep tumor extension, and desmoplasia [17, 18]. The current staging guidelines indicate PNI assessment as a binary variable based on presence versus absence. However, more detailed studies of PNI-related histological features could further refine staging systems for a better patients’ prognostic stratification.

The identification of molecular/cellular mechanisms that drive PNI has progressed consistently over the past few years. To date, several molecular pathways have been implicated in cSCC PNI. As a highly intricate and dynamic process, PNI involves a series of bidirectional signaling interactions between the tumor and various nerve components, such as Schwann cells and neurons. Schwann cells play a predominant role as glial cells in the peripheral nervous system, serving distinct functions as both myelinating and nonmyelinating subtypes [19]. Recent investigations in different types of tumors underline that Schwann cells have the potential to promote PNI. This is accomplished by dedifferentiation, migration toward the cancer site, dispersal of cancer cells, and the conveyance of cancer cells back to nerve regions [20, 21]. More precisely, cSCC can release neurotrophic factors into the surrounding environment, where they are recognized by Schwann cells and neurons. Subsequently, upon detection, Schwann cells and neurons may secrete additional neurotrophic or other molecular factors, initiating downstream events that promote both cancer cell invasion (shown in Fig. 1) and neurite outgrowth [22‒24].

Neurotrophic factors, including nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), and glial cell-derived neurotrophic factor, are increased in some human solid tumors, including cSCC [25, 26]. Several investigations have reported high expression levels of NGF in the perineural space, which, by binding to the high-affinity receptors, TrkA and p75NTR, modulate cell survival, differentiation, proliferation, migration, and invasion of malignant cells, supporting their role in PNI [27‒29]. BDNF, by acting on the tropomyosin receptor kinase B (TrkB) receptor, also contributes to neuronal survival, morphogenesis, and plasticity [30]. By transcriptional profiling and immunohistochemical study, a positive correlation between BDNF and TrkB mRNA expression has been reported in cSCC specimens when compared to normal mucosa [31, 32]. Cancer cells have been shown to exhibit a preferential migration toward Schwann cells and, upon contact, intercalation and mingling occurred between the 2 cell types. The potential role of TrkB and Schwann cells in PNI correlated with the enhancement of the intercalation of cancer cells and Schwann and cell-associated cancer cell dispersion along the nerve [33]. Neural cell adhesion molecules (NCAMs), alongside other adhesion molecules, by acting as a cell surface glycoprotein, enhanced the interaction between cancer cells and nerves, facilitating adhesion with Schwann cells and migration of cancer cells, particularly along nerve fibers [34]. Schwann cells can migrate toward cancer, or promote cancer cell migration toward nerves, and invade in a manner that relies on the cell surface expression of NCAM [35]. Different reports have shown increased expression of NCAM in both cancer cells and nerves, a feature that was associated with a higher likelihood of PNS of cSCC tumors [36]. By immunohistochemistry, increased NCAM expression was frequently found in cSCC tumors with PNS compared to tumors without, thus supporting a role of NCAM in cell-cell adhesion to promote the attachment of cancer cells to nerve fibers, facilitating the migration of such tumors along neural pathways [37].

In addition to neurotrophic signaling and changes in the expression in cell adhesion molecules, other signaling pathways influence PNI in cSCC, including changes in the extracellular matrix composition around nerves that can provide a conducive environment for cancer cell invasion. Matrix metalloproteinases and other proteases participate in extracellular matrix remodeling, allowing cancer cells to break through barriers and promote metastatic spread of aggressive cSCC [38]. Neurotransmitters and neuropeptides released by nerves can also influence cancer cell behavior, promoting migration and invasion. Nerve-derived galanin can activate its receptor in cancer cells and initiates a crosstalk between nerve and cancer by inducing tumor cells to secrete pro-inflammatory mediators and neuropeptides, thus promoting PNI [39]. Furthermore, the immune microenvironment around nerves modulates PNI. Macrophages and T-cell lymphocytes may promote the invasion process. Disruption of perineurium resulting from infiltrated cancer cells initiates an inflammatory cytokine cascade, giving rise to a peculiar cellular and biochemical microenvironment surrounding the nerve, identified as the perineural niche [40]. Composed of various cellular elements, including inflammatory mediators like cytokines and chemokines, the perineural niche affects neural tracking, and attracts and activates additional immune cells, thereby fostering the invasion process.

In the complex landscape of the tumor microenvironment, stromal cells and fibroblasts emerge as major contributors to PNI. Stromal cells play multifaceted roles, influencing the surrounding milieu through their diverse secretory functions and interactions [41]. Fibroblasts, known for their involvement in tissue remodeling, contribute to the dynamic changes within the microenvironment by producing neurotrophic factors [42]. Collectively, interplay between the cellular/molecular components of the tumor microenvironment, in the context of cancer progression, is essential in identifying effective markers that contribute to PNI promotion in cSCC.

The current standards for the classification of cSCC are described in the American Joint Committee on Cancer (AJCC) and the Brigham and Women’s Hospital (BWH) systems [43, 44]. AJCC staging of cSCC is applicable limited to tumours of head and neck (HN) origin, and includes tumor diameter, depth of invasion, involvement invasion of underlying structures, and PNI in nerves ≥0.1 mm diameter for primary tumors and incorporates nodal and systemic involvement [43]. According to the College of American Pathologists reporting protocol [45], pathologists are required to report PNI along with key histological pathological characteristics of the tumor, particularly tumor size and depth of invasion as well as additional prognostic variables such as the histological grade of differentiation, the presence of desmoplasia and the assessment of margins status [43]. In fact, confirmed high-risk microscopic prognostic features in cSCC include PNI [46‒50], together with tumor thickness >2 mm [51], Clark level IV or V invasion [52], primary site on the ear, lip, temple, and cheek [47‒49, 51], and poor differentiation [53, 54]. Specifically, PNI of deep dermis located nerves ≥0.1 mm in diameter has been demonstrated to correlate with higher disease-specific death rates [55]. In the AJCC Cancer Staging Manual 8th edition, cSCC are classified primarily based on their diameter, with any invasion into deep structures leading to classification as T3 [43]. Conversely, the BWH classification considers the number of accompanying risk factors to determine the T classification [44]. This system integrates four high-risk attributes, in addition to bone invasion, into the classification: poor differentiation, PNI (initially of any caliber, and ≥0.1 mm in the modified BWH staging system), diameter ≥2 cm, and invasion beyond subcutaneous tissue [44].

Microscopically, three connective tissue layers compose the nerve sheath: the innermost endoneurium, which encircles individual nerve fibers, made by axons and associated Schwann cells; the perineurium, surrounding individual nerve fascicles and constituted by of endothelial cells lined by basal lamina; and the outermost epineurium, which unites several nerve fascicles together to form larger nerve trunks [56]. Nerves are part of the tumor microenvironment, together with fibroblasts, vascular endothelial, smooth muscle cells, and immune cells establishing mutual interactions [57].

The histological definition of iPNI is ambiguous, and its assessment remains controversial. The first statement, provided in 1985 by Batsakis et al. [58], defined iPNI as tumor cell invasion in, around, and through peripheral nerves. Dunn et al. [59] proposed the presence of cytologically malignant cells in the perineural space of nerves as satisfactory diagnostic criteria for iPNI, adding that, in equivocal cases, the observation of total or near-total circumferential involvement is supportive, as is the presence of perineural tracking in tangential sections and intraneural involvement. Liebig et al. [11] characterized iPNI as the presence of tumor cells within any of the 3 nerve sheath layers, expanding iPNI evaluation to include two different morphological phenotypes of nerve involvement: the first pattern (type A) is recognized when tumor cells are located within the peripheral nerve sheath and infiltration into the 3 nerve sheath layers can be clearly identified; the second pattern (type B) is attributed when tumor cells are seen in close proximity to the nerve and involve at least 33% of sheath circumference, whereas the intraneural invasion is used when tumor cells are noted inside the internal endoneurium. However, in large nerves, such as those involved in PNS of head and neck cutaneous squamous cell carcinoma (HNcSCC) malignancy, the layers are well defined and PNS occurs within the true perineural space. In this circumstance, Brown [60] proposed a more precise definition to recognize the type of spread based on the anatomical location involved, by defining tumor spread in large nerves as showing epineural invasion, iPNI, and/or intraneural (endoneurial) invasion. Although some authors consider intraneural invasion a subtype of iPNI that must be specified in the pathological report for its prognostic relevance, there is currently insufficient evidence in the literature that demonstrates prognostic differences to justify this distinction, which is why to date intraneural invasion and iPNI are considered overlapping terms in cSCC [61].

Most commonly, iPNI in cSCC microscopically appears within the perineural layers, creating an onion skin-like architecture. Observing the presence of iPNI at low power, the nerve frequently seems to be enlarged and cellular. A lymphoid infiltrate surrounding the involved nerve may also be noted and it can be sometimes prominent, and partially hiding the tumor. At higher power, malignant cells more frequently show a typical epithelioid morphology; however, spindled features can be seen as well, especially in a desmoplastic scenario. Pathologists may be aware of several histological findings reported as mimickers of iPNI in cSCC. According to Hassanein et al. [62], peritumoral fibrosis (PF), defined as the presence of concentric layers of fibrous tissue surrounding and/or surrounded by tumor formations, represents an insidious benign entity that may be mistaken for iPNI (shown in Fig. 2a, b). Interestingly, the authors also noted a strong association between PF and iPNI, encouraging the search for iPNI when PF is highlighted. In the context of a previous malignancy, reparative perineural proliferation may represent another important confounding factor in the assessment of iPNI, as regenerating nerves in a healing surgical wound could reveal prominent proliferation of the perineurium, and this event can mimic iPNI (shown in Fig. 2c, d). The benign nature of this condition can be confirmed by showing negative immunohistochemical staining in the spindle cells for S100 and cytokeratins but positive staining with epithelial membrane antigen, as displayed in normal perineural cells.

Another histological finding that could be misinterpreted as malignant iPNI includes epithelial sheath neuroma (ESN), a benign dermal lesion characterized by multiple enlarged peripheral nerve fibers unsheathed by mature squamous epithelium and occasionally surrounded by lymphoplasmacytic and histiocytic perineurial inflammation or intraneural mucin accumulation [63]. However, identification of iPNI in cSCC, due to the almost indistinguishable cytopathology of ESN, can be detected by the absence of associated in situ or invasive carcinoma and by the nerve fiber localization in ESN, which are typically found in the reticular dermis. cSCC uncommon multinodular growth patterns can also mimic intraneural invasion (shown in Fig. 2e–j). Other benign histological mimics of iPNI should always be considered, including ganglion cells resembling the cells of well-differentiated cSCC, intraneural blood vessels with prominent endothelial cells, and reactive perineural cell hypertrophy, which may be observed near iPNI or as a reaction to inflammation around tumor cells. In most cases, accurate attention to cytoarchitectural features can help to identify these confounding findings, since malignant cells commonly show hyperchromatic nuclei, irregular enlarged nucleoli, pleomorphism in nuclear size and shape, and sometimes pathognomonic presence of mitoses and/or apoptotic bodies. In particularly challenging cases, immunohistochemical stains may be necessary.

Current staging systems indicate iPNI as a pathological binary variable for its evaluation, since no further qualitative or quantitative definitions of histological features related to iPNI are required in pathology reports. A possible approach that considers different measures of iPNI is represented in Figure 3. A recent retrospective study by Totonchy et al. [64] was aimed at identifying other potential histological prognostic features related to iPNI in cSCC, such as the number and depth of nerves involved, the extension of iPNI beyond the bulk of the tumor mass, and the degree of nerve sheath involvement. According to this study, nerve diameter and number of affected nerves were significantly associated with adverse outcome. Different degrees of nerve sheath involvement are detailed in Figure 4. However, debate persists in the literature whether the diameter of nerve involvement and the number of involved nerves are primary and independent risk factors for poor outcome, as a recent study found both significantly associated with survival [14]. Moreover, further data are required to establish a standard dimensional cutoff definition of enlarged nerve, considering the wide variability of the measurement system used, such as maximum cross-sectional area or diameter [55, 65]. Regarding the number of nerves involved, several studies have reported the association between “extensive” involvement and clinical outcome in HNcSCC [14, 66‒69]. Interestingly, Massey et al. [69] recently explored iPNI measurement comparing 4 assessments of iPNI in cSCC, their associations with poor outcomes, and implications for their inclusion in the staging system of BWH. More specifically, they evaluated a large retrospective cohort of 140 patients diagnosed with cSCC, taking into account four iPNI features: nerve caliber, number of involved nerves per section, iPNI maximal depth, and iPNI location with respect to tumor. However, only involvement of multiple nerves was associated with a poorer outcome. iPNI of 5 or more distinct nerves, called extensive PNI (ePNI), was found to be independently associated with local recurrence, disease-specific death, and any poor outcome [69]. A revised BWH staging system with substitution of ePNI for large-caliber iPNI seemed to result in an improved area under the curve and test characteristics compared with current BWH staging criteria that use nerve caliber as the measure of iPNI, suggesting that ePNI should be considered for inclusion in cSCC tumor staging. The reported additional measures of iPNI in cSCC and their prognostic correlations are summarized in Table 1.

Proper assessment of iPNI can be unresolved in the daily practice, and hematoxylin and eosin (H&E) in deeper sections, and/or immunohistochemistry (S100 and keratin), could be mandatory to establish iPNI in equivocal and challenging cases, as shown in Figure 2g–j. Regarding the usefulness of immunohistochemistry for iPNI assessment in cSCC, Frydenlund et al. [70] compared the incidence of iPNI in cSCCs of HN versus non-head and neck (non-HN) areas using a double immunostaining (DIS) protocol, with S-100 for nerve and p63 for nuclear labeling of tumoral cells. The DIS protocol for iPNI detection was compared to detection by H&E alone. Review of H&E sections revealed iPNI in 6 (11%) of 57 cases from the HN and 3 (6%) of 53 cases from non-HN areas. Using DIS, iPNI was detected in 13 (23%) of 57 cases from the HN and 8 (15%) of 53 cases from non-HN areas. Thirteen cases of iPNI were detected with DIS that were not seen on H&E, representing an increase of 2.33 times. Despite all these precautions, iPNI determination by the histopathological examination of full histological slides remains a substantial challenge and this reflects the wide variance of reported incidence rates in cSCC [71] and the time-consuming and arduous procedure. To tackle this issue, several recent studies have focused on computational approaches to extract nerves and iPNI from histologically stained whole-slide images, utilizing deep learning networks or artificial-intelligence-based classifiers [72‒74] and exploring further methods to enhance precision, providing additional measures of iPNI. Lee et al. [75] conducted a pilot study aimed at developing a deep learning-based human-enhanced tool, called domain knowledge enhanced yield (Domain-KEY) algorithm, for identifying iPNI in digital slides, reaching a mean diagnostic accuracy as high as 97.5% versus traditional pathology. Li et al. [74] applied a trained model for nerve segmentation to both prostate cancer and HN cancer slides. In particular, a computational approach was proposed to extract nerves and iPNI from whole digital slides. Comparisons were then made for segmentations with and without the proposed domain adaptation on whole slide histopathology images from “The Cancer Genome Atlas” (TCGA) database, and improvements were observed in the HN cohort. Although a computer-assisted diagnosis appears feasible, the limitations to these studies are the small sample sizes and the lack of independent validation in larger clinical cohorts [71].

In recent years, PNI studies have increasingly focused on the identification of new biomarkers related to pathogenetic mechanisms underlying PNI, with the aim of contributing both to a more precise prognostic stratification and to constitute potential targets for the development of future therapies. Until now, stratification tools have had a limited impact on clinical practice and management, particularly among high-risk cSCC patients. These uncertainties underline the importance of detecting PNI-related biomarkers with both prognostic and predictive significance. PNI is currently thought to occur by invasion, as the result of a reciprocal and dynamic association process between tumor cells and nerve components [15]. Several neurotrophic agents have been shown to be involved, including NGFs, BDNFs, and other neurotrophins [76]. More recently, Wysong et al. [77] developed a prognostic 40 gene expression profile (GEP) test, stratifying patients with high-risk cSCC into three classes based on metastasis risk, revealing a positive predictive value of 60% for the highest-risk group. Eviston et al. [78] evaluated the GEP of 45 cases of HNcSCC with PNI, performing a tailored gene panel for sensitivity and specificity analysis. The case cohort was stratified into three groups (extensive, focal, and non-PNI) based on predefined clinicopathological criteria. The performed analysis showed significantly distinct GEP in HNcSCC with ePNI, due to up- and downregulation of more than 140 genes. A restricted 10-gene panel was also associated to ePNI detection. However, the retrospective nature of this study does not allow prediction of the onset of clinically significant PNI, as paired biopsy and resection specimens would be necessary to assess whether there is a role for this tool in preoperative evaluation in order to obtain a more accurate prognostic stratification.

Warren et al. [79] focused on the expression analysis of PNI specimens with an emphasis on mutations affecting p53 activation. The results of the analysis at the protein level showed signatures of gene expression representative of activation of p53 in tumors with PNI compared to tumors without, along with other alterations. Immunohistochemical staining of p53 showed HNcSCC with cPNI to be more likely to exhibit a diffuse overexpression pattern, with no tumors showing normal p53 staining. However, DNA sequencing of HNcSCC samples with cPNI did not highlight any significant difference in mutation number or position compared to samples negative for PNI. In 2020, an interesting focus on the expression of melanoma antigen family A, 3 (MAGE-A3) at the mRNA level in cSCC with PNI was proposed [80]. MAGE-A3 expression is known to be related to marked cell proliferation and mediate fibronectin-controlled cancer progression and metastasis in many tumors, such as lung cancer [81], diffuse large B-cell lymphoma [82], and gastric cancer [83]. In a cohort of 24 patients with cSCC, upregulated expression of MAGE-A3 emerged in poorly differentiated cSCC with PNI, suggesting a role of this biomarker in PNI and cancer progression [80]. However, larger studies are needed to validate the prognostic significance of MAGE-A3 in cSCC. Zilberg et al. [84] studied somatic mutations associated with adverse histopathological features in a cohort of 24 high-risk HNcSCC and their relevance, according to currently available clinical and preclinical targeted therapeutic agents. Somatic missense mutations in the fibroblast growth factor receptor 2 (FGFR2) were seen exclusively in patients with histological evidence of PNI. Of these, FGFR2 p.N549K and p.M536I are validated targetable activating mutations, suggesting further treatment options for these patients. Although many efforts have been made to increase knowledge of biomarkers associated with PNI, larger studies are required to provide prognostic and potentially predictive results.

iPNI, a recognized negative prognostic factor in several types of cancer, including cSCC, is a dynamic process involving reciprocal tropism between tumors and nerves that exhibits specific patterns across different tumors, depending on anatomical location, nerve density, and invasiveness levels. Several critical points need further exploration, starting with the assessment of iPNI. Several recent studies have been aimed at trying to identify other potential histological prognostic features associated with iPNI in cSCC, such as the number and depth of involved nerves with significant correlations with prognosis, and it may lead to refine the current staging guidelines. Digital pathology may aid in exploring additional methods to enhance precision of additional measures of iPNI, contributing to precision medicine. In addition, knowledge of new biomarkers may be beneficial for the personalization of treatment.

The authors have no conflicts of interest to declare.

This study was supported by grants from Associazione Italiana per la Ricerca sul Cancro (AIRC) under IG 2020 - ID 24503 (RN).

Filippo Nozzoli and Romina Nassini: writing – original draft; Francesco De Logu: data and figure curation; Martina Catalano and Giandomenico Roviello: review and editing; and Daniela Massi: conceptualization, formal analysis, methodology, validation, and writing – review and editing.