Prostate cancer (PCa) is the most common cancer among men worldwide, and has a high incidence and associated mortality rate, despite the recent advances in treatment1. PCa is an androgen-driven disease in which tumor growth is triggered via ligand-mediated signaling through the androgen receptor (AR)2. Although androgen deprivation therapy (ADT) can effectively control tumor growth, the majority of the patients eventually develop an androgen-insensitive form of PCa known as castration-resistant prostate cancer (CRPC), a hallmark of advanced disease3. Most CRPC cases involve the AR signaling pathway. Novel antineoplastics targeting the AR axis have been developed and approved (e.g., enzalutamide and abiraterone). These drugs prolong the survival of patients with advanced PCa4,5,6; however, androgen receptor pathway inhibitors (ARPI) can induce the development of neuroendocrine prostate cancer (NEPC) and aggressive variant PCa (AVPC). NEPC is highly aggressive and has a poor prognosis. It is characterized by the expression of neuroendocrine (NE) markers, such as chromogranin A (CGA) and synaptophysin (SYN/SYP)7,8,9. These cancer cells can evade ARPI treatment and proliferate, eventually leading to the development of drug resistance. Moreover, the exact mechanism driving NEPC differentiation remains unclear, partly due to the limited availability of cell and tumor models of NEPC10. Therefore, it is necessary to establish a platform for preclinical drug evaluation to assess the treatment response.

Furthermore, there is an urgent need for mice or cell models that recapitulate the heterogeneity and progression of NEPC; these models are presumed to provide essential information for clarifying the mechanisms underlying the development of NEPC, information that would enable the treatment of NEPC patients10. In this review, we summarize the development of NEPC models and strategies for their evaluation, with a special focus on the development of PCa and associated heterogeneity. Additionally, we provide an overview of the regulatory network contributing to NEPC. Furthermore, we propose a new strategy for the evaluation of NEPC models—for use in experimental research—to improve the accuracy of model classification, thus providing an ideal tool for research on PCa development and drug screening.

Strategies for the establishment of NEPC models

An ideal NEPC model effectively recapitulates the clinical characteristics, molecular diversity, and cellular heterogeneity and can be used for research on the transdifferentiation mechanisms11,12; however, methods used for the generation of this type of model differ widely, and studies are therefore limited by a general lack of standardization.

Establishment of a mouse model of NEPC

The characteristics of implanted PCa specimens, i.e., hormone-independent or small cell morphology, should be accurately reflected in the host mouse. NEPC cell lines and fresh tissues and organoid cultures from patients with NEPC are commonly used as NEPC models13,14,15,16; however, the direct establishment of NEPC animal models is limited. In contrast, it is easy to establish a hormone-naive PCa (HNPC)/CRPC model, and these models can therefore serve as a starting point for the establishment of NEPC models. ADT and endocrine treatment regimens are increasingly being used for inducing neuroendocrine differentiation (NED) in HNPC/CRPC models13,17,18. This model has advantages such as the successful recapitulation of NEPC; it has shaped our knowledge of NEPC, particularly the characteristics of its progression and associated stages. To date, the representative NEPC models include the LuCaP series, KuCaP series, LTL-series, and MDA PCa series13,19,20,21; of these, the LTL-331R patient-derived xenograft (PDX) model and 42D/FENZR model, established by the Living Tumor Laboratory (LTL) and by Bishop et al., respectively, are widely used. In these two models, the NEPC phenotype is obtained by employing different induction methods. The LTL-331R patient-derived xenograft (PDX) model is derived from LTL331, a typical adenocarcinoma model, generated using castration. LTL-331R undergoes growth arrest in response to castration and exists in a dormant state for at least 38 weeks. After prolonged androgen deprivation, the PDX model displays spontaneous androgen-independent outgrowth and irreversible NEPC transformation20. The 42D/FENZR model is derived from the LNCaP CDX model with enzalutamide resistance. This model recapitulates the entire process of transdifferentiation, i.e., from HNPC to NEPC. First, the cell line-derived xenograft (CDX) model of HNPC is established by injecting LNCaP cells into intact male athymic mice to produce subcutaneous tumors. Thereafter, these mice are castrated; the resulting mice model relapsed CRPC. Eventually, the NEPC model is established by gavage with enzalutamide for three generations17. These models have been irreplaceable for studies on the emergence of NEPC variants under the selective pressure of ARPI.

The TRAMP model is one of the most commonly used transgenic PCa models. Gingrich used the transgenic technology to transfer the monkey virus 40 T antigen (SV40-T-Ag) into mice; in this technology, the expression of SV40-T-Ag is driven by the rat probasin promoter, resulting in the specific induction of PCa development22. Approximately 80% of TRAMP mice will develop into NEPC expressing NE marker and metastasize to lymph nodes, lungs, bones, kidneys, and adrenal glands after 12 weeks of castration23. This may be related to the pathway involving the tumor suppressor genes P53 and Rb that are blocked by monkey virus 40 (SV40). The TRAMP model has been recognized as a mouse model for NEPC and is used to study the effects of Icaritin, aqueous metabolites, 2,3,7,8-tetrachlorodibenzo-p-dioxin, and germline genetic variation24,25. Moreover, the TRAMP model provides effective tools to explore the mechanism underlying the transformation of “AR-driven” to “non-AR-driven” PCa phenotypes, similar to those used by Bishop et al. to explore the correlation between the POU-domain transcription factor BRN2 and NE markers18.

Establishment of an NEPC cell model

Despite extensive clinical and basic research on NEPC, the overall research progress remains slow due to the lack of suitable cell models. In 1994, Whitesell et al. proposed that cyclic AMP (cAMP) could induce the generation of neuroendocrine phenotypes in LNCaP cells, thereby launching a new era of cell model research26. Subsequently, researchers have established different induction strategies17,20,21,25,26,27,28,29,30,31,32,33,34,35 (Table 1). Each strategy has its own advantages and disadvantages and should be implemented based on the experimental requirements. For example, the hypoxia induction strategy might be exploited for all applications in which proliferation is required, whereas androgen sensitivity may be required to achieve transdifferentiation in the culture system34.

Table 1 Overview of construction methods for NEPC cell models.

NEPC evaluation methods

Clinical characteristics

The incidence of treatment-emergent NEPC (t-NEPC) is increasing in this era of novel AR inhibitors, and the frequency of these t-NEPC tumors is much higher than that of de novo NEPC7; however, no consensus is available on the pathological definition of t-NEPC in the clinic. In 2004, the World Health Organization (WHO) established a histomorphologic classification of PCa with NED, thereby highlighting the pathological features of NEPC for the first time36. In 2013, the Prostate Cancer Foundation (PCF) classified NEPC in more detail, thus illustrating the uniqueness of NED in the context of PCa37. In 2019, lymphocyte infiltration around the tumor was proposed for diagnosing patients with NEPC. Williams et al.38 evaluated NEPC specimens and found that NED may aggravate immune cell infiltration, making cold tumors “cooler.” In a prospective study of a human cohort, Zhang et al.39 screened 94 t-NEPC samples from biopsy specimens of 231 CRPC patients based on the histopathological and immunohistochemical evaluations of SYP, CGA, CD56, AR, and prostate-specific antigen (PSA). The extent of SYP expression was highest among the three NE markers (84.0%), and most patients were single- or double-positive for NE markers (rather than being positive for all markers). Moreover, Zhang et al. proposed the following two conditions for the diagnosis of t-NEPC: (1) at least one NE marker with diffuse (>50%) positivity and (2) morphological features of typical small-cell carcinoma, even if the sample is negative for all NE markers. Therefore, clinicopathological and immunohistochemical analyses are used for diagnosing NEPC. In addition, before the diagnosis of t-NEPC, changes may be observed in the pathological and molecular characteristics of the patients. For example, in certain cases, the disease progresses rapidly when PSA levels are low, liver metastasis is detected shortly after the detection of elevated lactate dehydrogenase (LDH) levels, and “non-AR-driven” characteristics appear immediately after the loss of TP53 or RB140. These results indicate that the expression of these “predictive” factors should be tested earlier for the early diagnosis of NEPC.

Cell models

NE cells are typical components of the human prostate and are present at a low frequency. In the early stage of PCa, NE cells—present in numerous epithelial cells—are present but do not contribute substantially to disease progression. This explains the rare occurrence of de novo NEPC (1%); however, the frequency of NEPC increases by 25% after treatment WITH ADT or ARPI41,42. NCI-H660, a typical well-characterized small cell cancer cell line with a unique morphology, was originally used in studies describing the effects of atrial natriuretic peptide (ANP) on lung cancer; in 2000, it was first used to study PCa10,43. When the morphological analysis of epithelial cells reveals a spear morphology, compact cell bodies, and extended fine branches, cells undergo differentiation into neuroendocrine lineages. The greater number of neurites per cell and the average neurite length, the more obvious the NE trend44. However, a proportion of NEPC cell lines form 3-dimensional (3D) multicellular spheroids, even under 2-dimensional (2D) culture conditions. These cells express high levels of multiple NE markers10; however, a recent study reported that CGA is expressed in both CRPC and NEPC45. This lack of specificity makes it difficult to distinguish NEPC based on the expression of NE markers. Therefore, the identification of alternative markers is urgently needed.

Recently, the comprehensive use of multiple markers for evaluating the veracity of NEPC cell models has gradually increased. Bishop et al.18 observed NEPC induced by enzalutamide in PSA (compared to PSA+ enzalutamide-resistant CRPC cell lines), indicating that PSA is not required for NEPC recurrence. Expression of various genes, such as Aurora Kinase A (AURKA), MYCN, and EZH2, which are specifically overexpressed in human NE prostate tumors where they induce the expression of NE markers46, is markedly upregulated in NEPC stem cells (CSCs). Bhagirath et al.47 developed an “extracellular vesicles (EV)-miRNA classifier” that is able to robustly differentiate between “CRPC-NE” and “CRPC-Adeno.” They identified thrombospondin 1 (TSP1) as a specific biomarker and proposed that a miRNA panel and TSP1 can be used as a novel, noninvasive tool to identify NEPC and guide treatment decisions. Accordingly, the strategy of diagnosing NEPC using novel, highly specific molecular biomarkers is promising.

Moreover, the proliferation and invasion abilities of tumor cells are positively correlated with the degree of malignancy; however, the NEPC cell model does not exhibit this biological feature48. For example, NCI-H660 cells exhibit a slow growth rate in vitro10. This indicates that the NEPC cell model remains in a quiescent state for a long period in vitro. Presumably, NED in common prostatic adenocarcinoma exclusively occurs in the G0 phase of the cell cycle, inducing G0 arrest. NED is lost when the tumor cells re-enter the cell cycle48. Similarly, prostatic epithelial cells characterized by the expression of CGA consistently lack the expression of proliferation-associated Ki 67, thereby clearly indicating that neoplastic NE cells do not proliferate48. Wang et al.49 reported that even in the absence of inducing factors, a majority of NEPC cells remain in a quiescent state, continue to exhibit elevated NE marker levels, and maintain their typical NE morphological appearance. Surprisingly, the growth rate in vivo differs from that in vitro. Such as, the NCI-H660 cell line can form NEPC xenografts exhibiting rapid progression when grafted in immunocompromised mice50.

Responsiveness to drugs is another potential evaluation method. The “non-AR-driven” characteristics make NEPC cells nonresponsive to ARPI and sensitive to platinum-based chemotherapy or other targeted drugs, such as AURKA inhibitor and anti-EGFR mTOR51,52. Therefore, analyses of cell dynamics and biological functions can be used as an auxiliary strategy for evaluating the NEPC cell models.

Mouse models

The establishment of PCa animal models can be traced back to 1968. Paulson et al.53 successfully generated a model of human prostatic malignancy by transforming hamster prostate tissue with SV40; however, due to an insufficient understanding of NEPC, early in vivo models have mostly focused on the castration resistance stage and rarely demonstrate NE features. Nevertheless, researchers have reclassified various models, such as Rotterdam PC models, as NEPC models54. The criteria for model evaluation have become increasingly detailed.

NEPC displays androgen-independent characteristics9. When NEPC tumors are implanted into intact, testosterone-treated, and castrated male mice, the tumor growth rate is not affected by the androgen levels20. Moreover, NEPC tumor cells present a small cell morphology with a high nuclear to cytoplasmic ratio and a “salt and pepper” structure. A high nuclear-cytoplasmic ratio and indistinct cell borders are characteristic features, along with high mitotic rate and apoptotic bodies revealed by H&E staining37. Furthermore, ERG FISH and serological tests are potential first lines of approach. For instance, plasma CGA/SYP levels are related to the volume of xenograft tumors; these tumors are repopulated with NEPC cells, despite low or moderate rising serum PSA levels18,37.

In particular, the NEPC growth rate in vivo is remarkably faster than that observed in vitro. The in vivo model not only requires adaptation to the low androgen environment during the induction process but also gradually acquires the ability to proliferate10. During the comparison of the P1–P5 generations of LnNE tumors, the tumor doubling time was shortened, and levels of NE markers were upregulated during NEPC development. These results indicate that transdifferentiation and acceleration of proliferation occur in parallel. The difference in NEPC proliferation in vivo and in vitro may be related to the following factors. (1) Although model proliferation is not dependent on AR, the molecules secreted by NE cells (e.g., 5-hydroxytryptamine, bombesin, calcitonin, somatostatin, and parathyroid hormone-related peptides) may promote growth37,48. This hypothesis is supported by a recent study by Wu, who demonstrated that monoamine oxidase A (MAOA) produced by tumor microenvironmental stromal cells promotes the development of PCa via the IL-6/STAT3 paracrine signaling pathway55. (2) Proliferation-related molecular changes play a pivotal role in regulation in addition to changes in lineage plasticity and NED during NEPC differentiation and tumor repopulation. NEPC cells must outgrow other PCa cells for clonal expansion and tumor repopulation56. This process may be aided by the expression of NEPC-specific proliferation factors such as PEG10. Increased PEG10 expression in the LTL331R model and 42D/FENZR model confers increased growth under the selective pressure of ARPI57,58. (3) The upregulated expression of antiapoptotic genes during NEPC induction may explain the replacement of adenocarcinoma cells with AR-negative cells and their survival in response to androgen deprivation and chemotherapy48.

Among the reported NEPC animal models, PDX act as optimal models for oncological research. Although a low success rate and long incubation period pose a challenge, this model preserves the heterogeneity of the primary tumor as well as the microenvironment11. Table 2 summarizes the main characteristics of the NEPC PDX model13,20,59,60,61,62,63,64, such as chromosomal changes, including an 8p (broken arm) deletion, MPRSS2-ERG gene fusion, tumor suppressor gene deletion, and proto-oncogene upregulation, all of which are consistent with the patient characteristics. Evaluation of the pathology, genomic features, and biological function of the model provide a crucial basis for an accurate NEPC diagnosis.

Table 2 Characteristics of NEPC PDX model obtained by induction.

Novel method to evaluate the NEPC model

The aforementioned studies have demonstrated that NEPC is actively responding to these conventional assessment methods. In addition, recent studies have also proposed some new methods, such as, including integrated NEPC scores65, ghrelin (GHRL) polypeptide probes targeting growth hormone secretion receptor (GHSR)66, circulating tumor DNA (ctDNA)17, and long noncoding RNA (lncRNA)67, for the more accurate discovery of NEPC. These methods are presently used or are actively being refined. For instance, the integrated NEPC score proposed by Beltran et al. has successfully screened four NEPC PDX models from the ENZ-T1 and ENZ-T2 models65. The integrated NEPC score provides an excellent method for diagnosing NEPC; however, it is insufficient for samples lacking RNA sequencing data. In this review, we summarized the NEPC literature and proposed a comprehensive model evaluation method applicable to cell and animal models (Table 3). Cell model evaluations mainly focus on morphological characteristics, phenotypic characteristics, molecular biology, and genome characteristics; animal model evaluations focus on tumor characteristics, histopathological characteristics, and phenotypic characteristics. In fact, not all assessed samples will reveal responses in these test items. Thus, a cumulative total points system was developed as summarized in Table 3. Most NEPC models display a positive score for NE markers, negative scores for PSA, and morphological features of typical small-cell carcinoma in microscopic13,14,64. Thus, a two-point score was observed for these three items. The change in NEPC-related gene expression acts as powerful evidence for the NEPC model; hence, a maximum score of 4 points is assigned. The remaining features are indispensable and score 1 point. The distribution of these scoring points demonstrated relative precision. Although other factors may reveal a strong impact, we presume that the NEPC model tumors may be directly related to moderate or rising NEPC scores. A published normative score is obtained when a response to the current test item is obtained, and the model is identified as an NEPC model when the total points are ≥5. This score provides an objective approach for NEPC model evaluation and improves the accuracy of judgment.

Table 3 Model evaluation for neuroendocrine prostate cancer.

Changes in NEPC-related genes and molecular pathways

Relationships between specific genes and the occurrence of NEPC are gaining immense interest in research. Thus, a suitable NEPC model reflecting the clinical characters of patients with PCa will be used to clarify the origin of NEPC and the transformation mechanism. The results of substantial basic and preclinical research indicate that NEPC models provide a basis for detailed analyses of the discovery of NEPC-related genes and molecular pathways68. In 2021, Kaarijarvi et al. have elaborated on the molecules and pathways contributing to neuroendocrine phenotype in PCa68. Similarly, we further analyzed the changes in NEPC-related genes and molecular pathways from a new perspective. The androgen–AR axis is specific and necessary18, whereas, PCa cells possess lineage plasticity and could differentiate into other lineages. In particular, PCa cells may undergo a dedifferentiation process, lose AR and PSA, become reprogrammed into an intermediate stem cell state, undergo transdifferentiation, and acquire NE characteristics69. Therefore, we discussed the major changes in genes and pathways related to the formation of NEPC from different perspectives (Fig. 1).

Fig. 1: Proposed pathways related to NEPC.
figure 1

AR-related pathways, mutations, or deletions of TP53 and RB1, SRRM4, and the tumor microenvironment all contribute to NEPC. First, AR-related pathways can be divided into two types (A, B). WLS, SOX2, MUC1-C, and BRN2 expression can be suppressed by AR by directly binding to the androgen response element (ARE) of AR. In detail, WLS promotes WNT5A ligand secretion and WNT5A promotes NE marker expression via the ROR2/PKCσ/ERK pathway. MUC1-C can induce NE marker expression via the MUC1-C/MYC/BRN2/SOX2 pathway and promotes NED via SOX2 (C). Other AR-related genes, FoxA1, EZH2, and N-myc, mediate the expression of NE markers by physical interactions with AR. N-Myc binds directly to AR to inhibit the AR signaling pathway, whereas EZH2 can enhance the formation of the N-Myc/AR/EZH2- prc2 complex. Second, SRRM4 can selectively splice REST and trigger NEPC. Moreover, the hypoxic tumor microenvironment can induce the expression of IL-6, HIF1α, EZH2, and ONECUT2 and promote neuroendocrine transdifferentiation. Paracrine pathways play a role in NED, including IL-6 can induce NED in PCa cells via PPAR γ and ADRP; GHRH-induced stimulation of NED involves EGFR/HER2 transactivation.

Androgen–AR-target gene signal axis/AR axis

Particularly, AR plays a pivotal role in the progression of PCa. HNPC rarely develops into NEPC after activation of the androgen–AR signaling axis. Under the selective pressure of ADT or ARPI, the downregulation of AR expression or activity leads to a high incidence of NEPC41. Cell experiments and mouse models also indicate that the androgen–AR-target gene signaling axis contributes to the occurrence of NEPC. First, AR inhibitors, such as enzalutamide, can release NEPC lineage differentiation genes, which are inhibited by AR, such as BRN218, HMG-box gene 2 (SOX2)70, and MUC1-C71, thereby blocking the effect of the signaling axis including androgen–AR target genes on the progression of PCa (Fig. 1). In enzalutamide-resistant mouse models, Bishop et al.18 revealed that the major neural transcription factor BRN2 is suppressed by AR via directly binding to the androgen response element (ARE) upstream of the transcription start site. BRN2 mainly triggers the transdifferentiation from CRPC to NEPC and maintains the NE phenotype by interacting with SOX2; however, an analysis of RNA sequencing data from multiple sources by Beltran et al. and Akamatsu et al. revealed that SOX2 is an effective suppressor of adenocarcinoma-specific gene expression during the progression of NEPC. Li et al.70 confirmed that SOX2 markedly inhibits the expression of specific adenocarcinoma master regulators (MRs), such as FOXA1, NKX3-1, and AR as well as AR target genes as well as interacts with BRN2 to induce the expression of NE markers and promote NEPC development18.

Moreover, interactions between AR and secreted proteins have been reported. For example, AR inhibits the transcription of WLS, which induces the formation and proliferation of NEPC via the ROR2/PKCσ/ERK pathway72. Mucin 1 (MUC1-C) is overexpressed in several cancers. A reciprocal interaction between AR and MUC1-C was first identified by demonstrating that AR occupies the MUC1 promoter and represses MUC1 transcription in LNCaP cells73. Yasumizu et al.71 clarified that MUC1-C induces NED via the MUC1-C/MYC/BRN2/SOX2 pathway, instead of the AR–BRN2 axis. Furthermore, various transcription factors upstream of AR regulate NE transdifferentiation, such as EZH2 and N-Myc74. N-Myc can directly bind to target genes to inhibit transcription and AR signaling pathways, whereas EZH2 can promote the formation of the N-Myc/AR/EZH2-prc2 complex. Follow-up studies have confirmed that EZH2 functions independently of the histone modification substrate H3K27me375.

These findings emphasize the importance of preclinical research and the establishment of an NEPC model of ARPI resistance that reproduces the process of transdifferentiation from AR+ to AR. Notably, these genes exhibit multiple, complex roles. For example, MAOA not only induces the stemness of PCa but also promotes NED55,76. Thus, the occurrence and development of NEPC involve a complex molecular regulatory network, and more detailed research in this area is urgently needed.

Stem cell differentiation

Whether NEPC originates from the transdifferentiation of adenocarcinomas or oncogenic mutation of normal neuroendocrine cells remains controversial. Complex, heterogeneous mechanisms, and various genetic pathways are involved48,77. A prevailing hypothesis suggests that NEPC is the result of adenocarcinomas undergoing lineage conversion (transdifferentiation) to the neuroendocrine lineage52. Lineage plasticity enables the adaptation and survival of tumor cells in the presence of ARPI78,79. Despite the lack of direct evidence that NEPC arises from the transdifferentiation of adenocarcinoma cells via a CSC intermediate, lineage plasticity may contribute to the acquisition of the CSC phenotype69. Indeed, these CSCs may be induced into various lineage types depending on the cell microenvironment79. In other words, PCa cells undergo the important step “Stem cell differentiation.”

Several studies have confirmed the roles of genomic, genetic, or epigenetic dysregulation in this process, for example, RB1, TP53, and PTEN deletion, and MYCN and AURKA overexpression65,69,80,81. A comprehensive genomic analysis of NEPC has revealed that the loss of RB1 and the mutation or deletion of TP53 occur together more commonly in NEPC tumors (~50%) than in prostatic adenocarcinoma (~14%)69,80. Indeed, the combined loss of function of these genes promotes lineage plasticity and NED. Silencing RB1 and TP53 in an antiandrogen-sensitive LNCaP-AR human prostate cancer cell line causes a rapid phenotypic shift from AR-dependent luminal epithelial cells to AR-independent basal and neuroendocrine cells65. These findings suggest that the loss of these genes may trigger lineage plasticity, combined with epigenetic dysregulation to form an epigenetic environment similar to that of stem cells.

The tumor microenvironment, including immunity, inflammation, and hypoxia, also contributes to the lineage plasticity mechanism82,83,84,85,86. As a pro-inflammatory cytokine, IL-6 plays a crucial regulatory role in the tumor microenvironment of PCa52. The level of MAOA in tumor stromal fibroblasts increases, and paracrine Twist1/IL-6/STAT3 promotes PCa stemness and expression of CSC markers55. IL-6 produced by tumor-associated macrophages provides a suitable microenvironment for CD44-induced CSC and drives the epithelial-mesenchymal transition and NED plasticity changes82,83. Wu et al.82 found that CD44 + is more conducive to the establishment of an immunosuppressive microenvironment based on the cell experiments and tumor models. The hypoxic tumor microenvironment also influences cell plasticity. Recent evidence suggests that hypoxia can induce the expression of IL-6 and CSC marker genes such as Nanog, Oct4, and EZH284,85. A direct connection between hypoxia and a neuroendocrine phenotype, via the interaction between HIF1α and One cut domain family member 2 (ONECUT2), has been described87. ONECUT2 can regulate the degree of binding of HIF1α to chromatin and activate SMAD3, a cofactor of HIF1α, thereby regulating hypoxia-related pathways. The synergy between ONECUT2 and the hypoxic tumor microenvironment suppresses androgen signaling and promotes neuroendocrine transdifferentiation of PCa. Additionally, ONECUT2 drives NEPC via PEG10, FOXA1, and PTEN69.

Similarly, some drugs trigger the phenotypic transformation in PCa. As aforementioned, cAMP induces the appearance of neurosecretory cell-like dense nuclear granules and NE markers29. Therefore, analyses of the source of NEPC from the perspective of CSCs further supports the relationship between lineage plasticity and NED.


According to the cell microenvironment, CSC goes through the next step: cells may be induced into various lineage types. The neuroendocrine lineage occupies an absolute advantage over the other various lineages. NED allows PCa cells to survive under selective pressure and contributes to treatment resistance69. This resistance includes common ARPI resistance as well as chemotherapy resistance. Wang et al.86 found that a deficiency in Nei endonuclease VIII-like 3 (NEIL3) can upregulate the expression of neuroendocrine-related genes and promote chemotherapy resistance, thereby revealing the potential mechanisms linking therapeutic resistance and NED.

Serine/arginine repetitive matrix 4 (SRRM4) promotes alternative splicing and the fusion of neural-specific exon inclusions88. Moreover, studies have reported an NEPC-specific RNA splicing signature that is predominantly controlled by SRRM432. The overexpression of SRRM4 promotes NED via epigenetic modifications, including its downstream target genes: histone acetyltransferase or demethyltransferase, such as MEAF6 and BHC80. Furthermore, SRRM4 functions by affecting other NEPC driver genes. For example, it selectively splices REST into the inactive form REST4. Lee et al.89 reported a novel mechanism by which SRRM4 transforms HNPC/HNPC cells into t-NEPC xenografts via a pluripotency gene network containing SOX2. These studies highlight the complex signaling mechanism driving tumor progression to NEPC.

Recently, paracrine mechanisms have emerged as key regulators of PCa progression and metastasis. Extracellular vesicles possess abundant specific information and are able to mediate intercellular communication and affect the surrounding microenvironment, thereby increasing interest in the prospective use of extracellular vesicles as specific markers90. Bhagirath et al. reported that EVs play a pivotal role in mediating NED states in advanced PCa. These vesicles act as vehicles for exchanging neuronal transcription factors between heterogeneous populations of tumor cells. BRN4 and BRN2 are selectively released in EVs, promoting NED and engendering a transmitted enzalutamide resistance90. Furthermore, NED can be mediated by the release of exosomes from CRPC cells exposed to IL-691. Growth hormone-releasing hormone (GHRH) can act as a paracrine factor that increases the percentage of neurite-bearing cells and protein levels of neuron-specific enolase via activation of calcium channels and EGFR/HER2 transactivation92. Collectively, these studies indicated that NE cells could produce and secrete peptide hormones and growth factors in a paracrine manner in order to promote NED. Therefore, the paracrine pathway may occupy an increasingly important position in the study of the NEPC mechanism. Notably, the aforementioned studies provide strong evidence for the close relationship between stem cell differentiation and the emergence of a neuroendocrine phenotype. This underscored transdifferentiation can be further explored to explain the origin and molecular mechanism of NEPC.

Limitations and future perspectives of NEPC

Despite substantial advances in NEPC, various aforementioned shortcomings and limitations of recent studies need to be addressed in future research.

(i) The origin of NEPC remains unclear. To improve the efficacy of NEPC treatment and develop individualized treatment plans, it is necessary to explore the pathogenesis more deeply. Although there are two standard hypotheses to explain the origin of NEPC, the origin of PCa remains controversial and requires in-depth studies. Advances in preclinical research and bioinformatic approaches may facilitate this area of research.

(ii) Lack of a standard model of NEPC in vivo. The PDX model maintains the heterogeneity of the primary tumor. Specific genomic abnormalities and genetic changes in patients with PCa are common in PDX transplanted tumors13; however, compared with other cancer types, there are few PDX models of PCa, and these do not effectively recapitulate disease progression. For example, the existing PDX models are static models of various stages or exhibit partial clinical transformation; therefore, they cannot completely recapitulate the clinical transformation process from HNPC to CRPC or NEPC. Therefore, it is necessary to construct a model for paired comparisons between stages within the same specimen for comparative genotype and phenotype analyses and to explore the mechanism underlying heterogeneity. Additionally, the lack of standard modeling approaches contributes to the low success rate of NEPC PDX models. Simulating clinical surgical castration may be an important means to induce CRPC in PDX models. New AR-targeted drugs may promote the neuroendocrine transdifferentiation of epithelial cells and induce NEPC18,84.

(iii) Research on targeted therapies based on NEPC models is progressing slowly. Patients with NEPC do not respond to ADT, and long-term androgen administration will drive the gradual transformation of adenocarcinoma components toward neural components. Similarly, the newly developed AR-targeted systemic therapy has weak effects. Platinum-based chemotherapy or combinations of targeted therapies with small molecule inhibitors (such as MAOA inhibitors) may be a promising strategy.


The establishment of NEPC models in a more timely and accurate manner will provide a basis for detailed analyses of the occurrence of PCa and provide insights into the causes of antiandrogenic drug resistance. In particular, it will facilitate research on the mechanism underlying heterogeneous transformation as well as the identification of novel target drugs for NEPC. In summary, numerous studies have confirmed the positive role of the NEPC model with respect to research on the progression and molecular mechanism underlying PCa. We have proposed a novel NEPC model assessment method (Table 3); and judge the sample as NEPC when the score ≥5; this is a simple, fast and accurate method. In the future, this assessment method may be widely used for effectively extending the survival of cancer patients.

In summary, additional studies are needed to elucidate the gradual progression to NEPC in detail. It is particularly important to establish more representative preclinical models to facilitate clinical trials aimed at preventing, slowing, or even blocking NED.