Adaptive epithelial-mesenchymal bi-directional transition: A key invasive plasticity for tumor metastasis

Summary

Epithelial-mesenchymal transition (EMT) is a reversible cellular program that allows epithelium-derived cells to acquire mesenchymal phenotypes, whereas mesenchymal-epithelial transition (MET) enables mesenchymal cells to regain epithelial properties. The capacity for EMT and MET, known as epithelial-mesenchymal plasticity (EMP), represents an indispensable mechanism of invasive plasticity for tumor progression during the spatiotemporal invasion-metastasis cascade. In this review, we elucidate the pivotal role of EMP in tumor metastasis and treatment resistance, as well as the latest understanding of the molecular mechanisms that regulate this cellular characteristic. We further summarize the methodology and technology for evaluating EMP and discuss the clinical translational potential of targeting EMP-related signaling pathways as a personalized anti-metastasis therapeutic strategy.

Graphical abstract

Molecular biology; Cancer

Introduction

Epithelial-mesenchymal transition (EMT) is a reversible process in which epithelial cells lose their tight junctions and polarity while gaining mesenchymal characteristics.¹ EMT program naturally occurs during early embryonic development and processes associated with later development and maturation. It is also essential for maintaining physiological homeostasis in mature tissues, such as physiological wound healing.¹ The opposite process of EMT, mesenchymal-epithelial transition (MET), enables cells with high phenotypic plasticity to restore their epithelial phenotype, usually after prior EMT.²

The primary cause of cancer-related mortality is the dissemination of malignant cells to distant organs.^3,4 In most cases, distant metastases predict treatment failure.^3,4,5 Tumor cell invasion and metastasis are intricately linked to EMT, and this process allows tumor cells to lose their intercellular adhesion and acquire higher migratory and invasive capabilities, enabling them to breach the basement membrane and enter lymphatic or hematogenous circulation.^6,7,8 Upon reaching distant organs, these disseminated tumor cells revert to an epithelial phenotype through activation of the MET program. The regained epithelial phenotype facilitates metastatic colonization and the subsequent proliferation of secondary tumors at remote sites.^9,10 An essential feature of human cancers in the spread of primary tumor cells to secondary sites is their phenotypic plasticity, known as epithelial-mesenchymal plasticity (EMP).¹¹ The EMT and MET programs rarely operate in a strictly binary manner and should instead be viewed as a continuum of stages in tumor cells.¹⁰ In this regard, EMP represents a key feature of aggressive tumor cells and encompasses the mutual transition among epithelial, mesenchymal, and hybrid E/M phenotypes.¹² In addition to invasion and metastasis, EMP is also closely related to chemoradiotherapy resistance, another common cause of treatment failure in human cancers.^11,13,14

Using single-cell RNA sequencing (scRNA-seq) and proteomic analysis, a significant characteristic of the EMT program in metastatic tumor cells has been revealed; specifically, the transition from an epithelial to a mesenchymal state is generally incomplete, sufficiently indicating that EMT is not a point-to-point, singular, or linear process.^15,16,17,18 Thus, elucidation of the signaling pathways underlying EMP will not only facilitate a comprehensive understanding of the molecular mechanisms governing tumor invasion, metastasis, and chemotherapy resistance but also establish a scientific foundation for exploring novel strategies for the development of targeted cancer therapy.^10,12,19,20 In this review, we propose that EMP should be viewed as an adaptive response to intense survival pressures. This adaptive invasive plasticity enables tumor cells to overcome survival challenges across different spatiotemporal dimensions during invasion and metastasis. We highlight recent advances in the clinical assessment of EMP profiles, discuss their therapeutic potential, and summarize recent clinical trials targeting EMP.

Emerging role of EMP in tumor progression

EMP for invasive plasticity

It is well established that cancer cell plasticity is a prerequisite for EMT and MET during tumorigenesis, development, and metastasis.²¹ The effect of cell differentiation on tumor development varies depending on the distinct phenotypic state of the cells. Different states of EMT transition are associated with diverse phenotypic functions. Moreover, EMT and MET programs enable the adaptive transition of cancer cells into various phenotypic states spanning the epithelial-to-mesenchymal spectrum, which represents the most invasive mechanism of plasticity in aggressive tumor cells.^22,23,24 In other words, EMT and MET represent fundamental manifestations of the phenotypic plasticity of tumor cells, and they, in turn, serve as the cornerstone of tumor invasion, metastasis, and resistance to treatment.²⁵

EMP allows tumor cells to maintain a hybrid E/M state, enabling them to retain both epithelial and mesenchymal traits. This plasticity permits dynamic transitions between phenotypes, allowing tumor cells to adapt to the evolving tumor microenvironment (TME) during metastasis.^10,11 Rather than existing as binary states, EMT is increasingly recognized as a continuous and dynamic program. Transcriptome integration analysis has shown that cells in the intermediate E/M state exhibit a hybrid phenotype combining high migratory and high proliferative capacities.²⁶ Tumor cells in this hybrid state are particularly aggressive.^18,27 Acquisition of the hybrid E/M phenotype is crucial for tumorigenicity in basal-like breast cancer.²⁴ P-cadherin promotes tumor invasiveness by facilitating the hybrid E/M phenotype in high-grade serous carcinoma of the fallopian tube and ovary.²⁸ Additionally, the hybrid EMT state confers resistance to ferroptosis in colorectal cancer cells via histone acetylation and WNT/β-catenin signaling.²⁹ Graphical Abstract illustrates the bidirectional EMTs involved in the adaptation to changes in the invasive microenvironment during the multistep process of tumor metastasis. The EMP-driven bidirectional transitions represent adaptive responses to external survival pressures, regulated by intrinsic genetic and epigenetic mechanisms.

EMP for tumor cell stemness

The presence of histological heterogeneity is accompanied by the heterogeneous expression of various markers among cancer cells, known as intratumor heterogeneity. Additionally, there is significant heterogeneity between tumors that develop in different patients with the same type of cancer, which is referred to as intertumoral heterogeneity.^30,31 Subsets of cancer cells, known as cancer stem cells (CSCs), can self-renew, reproduce, and even enhance the heterogeneity.³² CSCs represent a unique pool of highly tumorigenic tumor cells that retain some characteristics of normal epithelial stem cells.³³ The ability to undergo asymmetric division, self-renewal, and generate more differentiated progeny defines stemness. EMP can be viewed as a key driver of stemness-associated phenotypic transitions.³⁴ Moreover, the diverse trajectories within the high-dimensional EMP and CSC landscapes may exhibit distinct patterns of coupling.³⁵

CSCs can be generated by EMT in breast cancer cells.³⁶ Tumor cells in the intermediate E/M state exhibit remarkable plasticity and can acquire stem cell characteristics.^37,38,39 Ovarian cancer cells in the hybrid E/M state can differentiate into multiple cell lineages and exhibit self-renewal ability.⁴⁰ Activation of Twist1, an EMT transcription factor (EMT-TF), can induce stem cell-like characteristics in a subset of mammary epithelial cells.⁴¹ Signaling molecules in the local microenvironment, such as transforming growth factor β (TGF-β), promote higher expression of stem cell markers in lung cancer cells with a hybrid E/M phenotype.⁴² Notably, the cancer cells in the fully terminal EMT state show lower expression of stem cell-related genes, whereas the cells in the early or partial EMT states retain higher stemness and the ability to initiate metastasis.¹⁸

Circulating tumor cells (CTCs) spread in the peripheral blood, and they originate from primary or metastatic lesions.⁴³ CTCs generated through EMT have characteristics of CSCs.⁴⁴ Those from breast cancer exhibit characteristics associated with both CSCs and EMT and undergo phenotypic changes.^45,46,47 EMT not only enhances invasive capabilities but also endows CTCs with characteristics resembling those of CSCs.^48,49,50 In dedifferentiated tumor cells, the properties of EMT are combined with stem-like characteristics, thereby establishing the concept of “migrating CSCs” as the fundamental mechanism underlying metastasis. In addition, EMT facilitates the conversion of epithelial tumor cells into CSCs with high metastatic potential, which drives tumor invasion and metastasis.⁵¹

EMP for therapy resistance

Resistance to therapy is the main cause of tumor treatment failure. Therapeutic resistance can be classified as primary (intrinsic) or secondary (acquired) resistance, the latter arising during tumor progression as a result of genetic or epigenetic variations.⁵² The secondary resistance reflects the adaptive evolution of tumor cells, occurring alongside the accumulation of intratumoral heterogeneity. EMT is a central driver of the phenotypic diversity that underlies the development of secondary resistance in tumor cells.^13,48 Accordingly, EMP-driven resistance is a coordinated, multifactorial process that involves not only remodeling of tumor cell properties but also immune evasion.

EMP promotes drug resistance by enhancing drug efflux, partly through the regulation of efflux transport proteins. The expression of ATP-binding cassette (ABC) transporter is upregulated in the drug-resistant tumor cells undergoing EMT, facilitating active drug efflux.⁵³ EMT-TFs, such as Twist and Snail, further contribute to chemotherapy resistance by regulating ABC transporter expression.⁵⁴ In addition, EMP enhances DNA damage repair capacity, enabling tumor cells to survive chemotherapy- or radiotherapy-induced genotoxic stress. Ataxia telangiectasia mutated (ATM), a serine/threonine kinase essential for DNA damage sensing and repair, is hyperactivated in radioresistant breast cancer cells due to upregulation of zinc finger E-box binding homeobox 1 (ZEB1), an EMT-TF.^55,56 Similarly, integrin beta-1 promotes radiation resistance by enhancing DNA repair and activating YAP1-induced EMT in non-small cell lung cancer (NSCLC).⁵⁷ Moreover, Ras homolog family member J, preferentially expressed in EMT-associated cancer cells, mediates chemotherapy resistance by regulating replication stress responses and activating the DNA damage response.⁵⁸

The dysregulation of signaling molecules involved in immune modulation represents a crucial mechanism by which EMT facilitates treatment resistance through the evasion or suppression of immune responses.^48,59,60 Immune checkpoints play a vital role in regulating the immune system and preventing an overactive immune response from destroying normal cells. Tumor cells can exploit this mechanism to suppress cytotoxic T cell activity by expressing immune checkpoint proteins, thereby evading immune surveillance and immune-mediated attack.⁶¹ EMP-mediated immune resistance is primarily characterized by upregulation of immune checkpoint molecules. Consistent with this, multiple immune checkpoint proteins were found to be elevated in lung adenocarcinoma exhibiting an EMT-associated phenotype.⁶² The inhibition of Snail-induced EMT effectively suppresses tumor-specific immunosuppression in malignant melanoma and ovarian cancer, indicating that EMT is also involved in tumor immunotherapy resistance.^63,64,65 Further, EMT is associated with the suppression of crucial immune cells in the tumor immune response, and this process correlates with increased expression of targetable immune checkpoints such as cytotoxic T lymphocyte-associated antigen 4 (CTLA-4), T cell immunoglobulin and mucin domain-containing protein 3 (TIM-3) in NSCLC.⁶⁶

The plasticity of tumor cells partially accounts for intratumor heterogeneity, as they exhibit varying degrees of phenotypic switching between states of drug sensitivity and resistance.⁶⁷ Treatment-resistant cancer cells are generally more aggressive than treatment-sensitive cells owing to their higher degree of plasticity. EMP in tumors is most likely characterized by a hybrid E/M phenotype that potentially correlates with stemness and tumor heterogeneity.^68,69 The prevalence of partial EMT is significantly higher in tumors than in normal tissues, and tumor cells at different stages along the EMT continuum exhibit distinct contributions to chemotherapy resistance.⁷⁰ Moreover, tumor heterogeneity increases during disease progression because of genomic instability and the selective pressure exerted by drug therapy.⁷¹ Treatment-resistant tumor cells more commonly occur in the mesenchymal state than in the epithelial state.^72,73,74 Abrogation of the effect of the EMT-TFs, such as ZEB1, effectively overcomes resistance to chemotherapy, suggesting that targeting EMT-TFs is a feasible treatment strategy.⁷⁵ As mentioned above, EMT-TFs play a central role in treatment resistance through multiple mechanisms, including drug efflux, activation of DNA damage responses, and immune checkpoint regulation. Therefore, targeting EMT-TFs represents a promising strategy for overcoming therapeutic resistance.

Despite these studies, there is a paucity of in vivo evidence of the mechanisms underlying EMT-induced treatment resistance. The primary cause of this is the challenge of monitoring transient and reversible phenotypic switches throughout EMT. By employing a mesenchymal-specific Cre-labeled fluorescent indicator switching system to track the EMT lineage in a mouse model, breast cancer cell lung metastasis was found to predominantly comprise non-EMT tumor cells that retain an epithelial phenotype.⁷⁶ Utilizing such an in vivo experimental system is expected to provide an in-depth understanding of the mechanism underlying EMP-mediated drug resistance, which is conducive to providing a theoretical basis for improving clinical outcomes in patients receiving chemotherapy.

EMP for metabolic reprogramming

EMP is profoundly implicated in the metabolic reprogramming of tumor cells. Key EMT-TFs, such as Snail and ZEB1 family members, promote aerobic glycolysis, glutamine breakdown, and lipid metabolism reprogramming by regulating the expression of metabolic genes.^77,78 This metabolic reprogramming allows tumor cells to meet the energetic demands required for adaptations to the TME. Snail regulates glucose flux toward the pentose phosphate pathway, thereby supporting cancer cells to survive under metabolic stress.⁷⁹ In turn, aberrant metabolic intermediates promote EMP through epigenetic modifications.^80,81 Cells with a hybrid E/M phenotype exhibit a mixed metabolic state characterized by the coexistence of glycolysis and oxidative phosphorylation, conferring metabolic flexibility. This adaptability allows tumor cells to survive fluctuating TME conditions and resist chemotherapy-induced apoptosis.^82,83 Metabolic remodeling also sustains EMP-conferred stem cell-like characteristics, enabling self-renewal, dormancy, and resistance to radiotherapy and chemotherapy.⁸⁴ Moreover, metabolic adaptation permits tumor cells to bypass the oncogenic pathways inhibited by targeted therapies, while metabolic byproducts further remodel the immunosuppressive TME, contributing to resistance to immunotherapy.⁸⁵

Key signaling pathways that regulate EMP and the associated molecular mechanisms

Backbone networks for regulating EMT/MET (the reception and initiation cascade)

The classical EMT-inducing signaling pathways, including TGF-β, Wnt, and Notch, are well established. These pathways act as cellular “antennas” that enable cells to sense and integrate microenvironmental cues, thereby initiating the EMT program in response to changes in external conditions. Among them, the TGF-β/SMAD pathway functions as the most potent “master switch” for EMT induction. Upon ligand binding, the activated TGF-β receptor phosphorylates SMAD proteins, which then translocate into the nucleus and induce the expression of EMT-TFs, such as Snail and ZEB.^86,87,88 Notch signaling not only directly promotes EMT but also cooperates with TGF-β/SMAD signaling to regulate ZEB expression.⁸⁹ The Wnt/β-catenin pathway facilitates mesenchymal transformation by stabilizing EMT-TFs and repressing epithelial gene expression.^90,91,92 In addition, this pathway is modulated by non-coding RNAs that regulate β-catenin stability, thereby maintaining the dynamic balance between EMT and MET.^90,93,94 Together, these signaling pathways form the backbone network governing EMT/MET transitions.

Core transcriptional regulatory network (decision-making and execution cascade)

The core transcriptional regulatory network functions as the “central processor” and “molecular executor,” responsible for determining phenotypic fate and implementing EMT programs. Its central components are EMT-TFs, including members of the Snail, ZEB and Twist families.⁹⁵ Once activated, these transcription factors repress epithelial gene expression by directly binding to their promoters, while simultaneously activating mesenchymal gene expression.^96,97,98 Conversely, microRNA (miRNA)-mediated feedback loops act as negative regulators that help maintain the epithelial phenotype. For instance, the miR-200, miR-34, and miR-186 families form reciprocal regulatory circuits with ZEB, Snail1, and Twist1, respectively.^99,100,101 Notably, the ZEB/miR-200 and Snail/miR-34 feedback loops are themselves regulated by TGF-β signaling.^88,102 Moreover, combinations of different miRNAs can exert synergistic effects at the systems level to regulate EMP.¹⁰³ Multiple miRNA binding sites and RNA degradation dynamics are critical for stabilizing hybrid EMT states and sustaining the phenotypic continuum.¹⁰⁴ The coordinated RNA consumption process can trigger the intermediate EMT state in the absence of transcriptional feedback.¹⁰⁴ All of these transcriptional regulatory networks jointly regulate the plasticity for phenotypic transition by controlling EMT/MET in tumor cells.

MET-transcription factors (MET-TFs)

The hybrid E/M phenotype observed during the EMT process was once considered to be temporary and “unstable.” However, it is now recognized that various molecules function as “brakes” to delay the transition from the hybrid E/M state to a complete mesenchymal state. These molecules include Ovol-like zinc finger 2 (OVOL2), Krüppel-like factor 4 (KLF4), grainyhead-like transcription factor 2 (GRHL2), and E74-like ETS transcription factors 3 and 5 (ELF3 and ELF5), which are known as MET-TFs.^25,105 The transcription factor OVOL expands the range of the hybrid E/M phenotype by inhibiting EMT and driving MET in combination with the ZEB/miR-200 feedback loop.^106,107 Silencing GRHL2 leads to increased CpG methylation and nucleosome remodeling in ovarian cancer cells, which is similar to the changes observed in EMT progression. This suggests that GRHL2 inhibits EMT by stabilizing the intermediate phases of EMT/MET.¹⁰⁸ Additionally, GRHL2 maintains the epithelial state by interacting with the ZEB1/miR-200 feedback loop, modulating transcription and modifying histones.^109,110 KLF4 negatively regulates the EMT program by inhibiting the transformation of cell states from epithelial to mesenchymal through multiple parallel pathways.¹¹¹ ELF5 suppresses EMT by downregulating Snail2 at a transcription level in breast cancer cells.¹¹² ELF3 was reported to negatively regulate EMT in bladder cancer cells.¹¹³ These molecules collectively contribute to stabilizing the hybrid E/M phenotype and participate in regulating EMP.

Epigenetic modifications

Cellular phenotype transition is regulated by epigenetic mechanisms through DNA methylation, histone modification, and chromatin remodeling. These epigenetic modifications are implicated in the EMT program through the interplay with EMT-TFs.¹¹⁴ ZEB1 reduces CDH1 expression and transports DNA methyltransferase 1 (DNMT1) to the promoter region, thereby maintaining the methylation status of the CDH1 promoter in breast cancer cells.¹¹⁵ Snail downregulates E-cadherin by interacting with lysine-specific demethylase 1, and recruits this histone demethylase to promoters of the epithelial genes.¹¹⁶ ATP-dependent chromatin remodeling complexes, such as the SWItch/Sucrose Non-Fermentable (SWI/SNF), control the transcription of key EMT-related genes by regulating chromatin accessibility. ZEB1 interacts with the SWI/SNF-type chromatin remodeling protein Brahma-related gene 1 to inhibit E-cadherin and induce the EMT process.¹¹⁷ Notably, such epigenetic modifications can, in turn, impede EMT progress. For instance, DNMT1 maintains the E/M hybrid state by methylating the promoter of Snail in hepatocellular carcinoma cells.¹¹⁸ Expressions of Snail and Slug are epigenetically regulated by DNA methylation during EMT/MET.¹¹⁹ The loss of FAT atypical cadherin 1 leads to the inactivation of enhancer of zeste homolog 2 and promotes the expression of SRY-box transcription factor 2, thereby facilitating the hybrid E/M state.¹²⁰ The ten-eleven translocation family enzymes maintain the epithelial phenotype through demethylation in ovarian cancer cells.¹²¹ Overall, the epigenetic regulation of EMT exhibits significant reversibility and spatiotemporal diversity.

Common EMT-TFs, MET-TFs, and regulatory molecules are listed in Table 1. Taken together, the key signaling pathways and molecular mechanisms regulating EMP are multilevel and highly complex, involving a variety of signaling molecules, TFs, and epigenetic regulators.¹²² The key signaling pathways and molecular mechanisms are summarized in Figure 1.

Table 1.

EMT and MET transcription factors and the regulated molecules

TFs	Classification	Upregulated molecules	Downregulated molecule	Reference
Snail1	EMT-TFs	N-cadherin, collagen, ZEB1/2, Twist	E-cadherin, claudins	Peinado et al.123 Batlle et al.96 Tong et al.124
Twist1	EMT-TFs	N-cadherin, Fibronectin, integrinα5	E-cadherin, claudins	Vesuna et al.97 Hong et al.125
ZEB1 and ZEB2	EMT-TFs	N-cadherin, MMPs	E-cadherin, ZO1	Peinado et al.123 Sánchez-Tilló et al.117
GRHL2	MET-TFs	E-cadherin, claudin3/4, miR-200b/c	ZEB1/2, Snail, Twist	Mooney et al.110 Frisch et al.126
OVOL2	MET-TFs	E-cadherin, claudins	N-cadherin, Vimentin, ZEB1/2, Twist, Snail	Wu et al.127 Wang et al.128 Hong et al.129
KLF4	MET-TFs	E-cadherin, claudins	snail, twist	Subbalakshmi et al.111 Zhou et al.130
ELF3 and ELF5	MET-TFs	E-cadherin, claudin3/4, ZO1	snail, MMP9	Subbalakshmi et al.131 Gondkar et al.113

Abbreviation: EMT, epithelial-mesenchymal transition; MET, mesenchymal-epithelial transition; TFs, transcription factors; EMT-TFs, EMT transcription factors; MMPs, matrix metalloproteinases; GRHL2, grainy head-like 2; MET-TFs, MET-transcription factors; OVOL2, zinc-finger OVOL-like 2; KLF4, krüppel-like factor 4; ELF3, transcription factor E74-like factor 3; ELF5, transcription factor E74-like factor 5.

Figure 1.

Key signaling pathways that govern epithelial-mesenchymal plasticity (EMP) and the associated molecular mechanisms

The TGF-β, Wnt/β-catenin, and Notch signaling pathways co-regulate EMP. TGF-β signaling activates the epithelial-mesenchymal transition nuclear transcription factors (EMT-TFs) by stimulating the phosphorylation of SMAD2/3 and SMAD4, thereby inducing the expression of EMT-promoting genes. Activation of the Notch receptor releases NICD and facilitates the formation of the NICD/CSL/MAML complex, which activates EMT-related genes by modulating EMT-TFs. The activation of Wnt leads to the accumulation of β-catenin, which binds TCF/LEF transcription factors in the nucleus to promote EMT. Integrins mediate the signaling transduction between tumor cells and the extracellular matrix (ECM), enabling perception and adaptation to the changes in the regional microenvironment through epithelial-mesenchymal bi-directional transition.

Clinical approaches for the assessment of EMP

Liquid biopsy

Both EMT and MET are highly dynamic processes in disseminated tumor cells that occur during invasion into the vascular system, survival in circulation, and the formation of metastases in distant organs.¹³² EMT alters the morphology, physical properties, intercellular junctions, molecular markers, and cytoskeletal organization of CTCs.¹³³ Accordingly, the simple utilization of epithelial antigen-based assays may lead to the omission of aggressive subpopulations of CTCs with mesenchymal phenotypes. Disseminated pancreatic cancer cells that exhibit mesenchymal phenotypes and have stem cell properties were identified before the primary tumor lesion was detected in a mouse model, and they successfully colonized the liver.¹³⁴ Moreover, disseminated tumor cells have been observed in the bone marrow of patients with ductal carcinoma in situ.¹³⁵ Optimized methodologies for capturing CTCs enable their molecular characterization, especially the EMP-associated phenotypic transition. Such an assessment of EMP can advance our understanding of the intricate cascade involved in tumor invasion and metastasis.¹³⁶

The term “liquid biopsy” refers to a highly sensitive blood test capable of detecting single or clustered tumor cells among one billion normal hematopoietic cells.¹³⁷ CTC analysis is considered a real-time “liquid biopsy,” as this technology can monitor the evolution of blood metastasis at the single-cell level.¹³⁷ Profiling the EMT status of CTCs is advantageous for revealing the mechanism underlying EMT in human cancer metastasis.¹³⁶ CTCs exist as single cells or cell clusters consisting of several dozen tumor cells.^138,139,140 Clustered CTCs were captured, and the association between the risk of tumor metastasis and the distribution of mesenchymal CTCs was identified using the CanPatrol CTC-enrichment technique in nasopharyngeal carcinoma (NPC).¹⁴¹ A significant correlation between mesenchymal CTCs and disease progression has also been demonstrated through the serial monitoring of CTCs in breast cancer.¹⁴² Furthermore, the presence of key regulators of the EMT program in CTCs supports the crucial role of EMT in the hematogenous dissemination of breast cancer cells.¹⁴² Finally, a phenotypic classification of CTCs indicates that E/M hybrid-predominant CTC distribution is associated with unfavorable disease-free survival and distant metastasis-free survival in progressive NPC.¹⁴³ However, assessing epithelial and mesenchymal markers alone does not accurately capture true phenotypic plasticity, as this approach fails to reflect the evolutionary trajectories of disseminated tumor cells. Other components of liquid biopsy, including circulating tumor DNA (ctDNA), tumor-derived exosomes and other extracellular vesicles (EVs), exosomal miRNAs, and EMP-related proteins, can provide additional insights into the genomic features underlying plasticity and the signaling pathways involved in EMP regulation.^144,145,146 Nevertheless, EMP evaluation based on liquid biopsy is largely influenced by sample heterogeneity, and these assessments can only reflect information from the sampling site. Moreover, such analyses cannot reveal real-time functional behaviors, such as invasive capacity or therapeutic resistance.^147,148 In addition, the extremely low abundance of analytes such as ctDNA, together with unavoidable losses during technical preprocessing, often compromises the efficiency and sensitivity of liquid biopsy approaches.^149,150 Consequently, achieving dynamic and functional assessment of tumor cell plasticity using liquid biopsy remains a substantial challenge.

Single-cell multi-omics analysis

Cancer cell plasticity refers to the dynamic transition between distinct cellular states that enhances tumor heterogeneity and facilitates disease progression. Characterizing inter-tumor and intratumor heterogeneity offers insights into tumor metastasis and guides personalized treatment strategies. Single-cell multi-omics technology has significant advantages in analyzing the cellular composition and molecular characteristics, presenting an opportunity to unveil the heterogeneous state of individual cells, including EMP, a prototypical illustration of cancer cell plasticity.¹⁵¹

Examining the transcriptomic landscape using high-throughput scRNA-seq enables in-depth transcriptome analysis, reclassification of CTC subtypes, and improved detection of rare new subgroups.¹⁵² The latest scRNA-seq analyses and mathematical models have revealed that the intercellular communication between the epithelial and mesenchymal states is crucial for tumor heterogeneity.¹⁵³ Cells undergoing EMT promote the spread of epithelial cancer cells through mechanical conduction, thereby facilitating the disintegration of tumor spheres.¹⁵⁴ A comprehensive understanding of intercellular communication is imperative for elucidating EMP and characterizing tumor heterogeneity at both the population and subpopulation levels. Combining spatial transcriptomics and scRNA-seq technology systematically reveals the dynamic remodeling process of the TME in both spatial and temporal dimensions.¹⁵⁵ Highly multiplexed scRNA-seq analysis enables reconstruction of the dynamic path of EMT phenotypic transitions and facilitates the identification of key regulatory nodes and molecular events during EMT and MET processes.¹⁵ Single-cell lineage tracing provides a powerful approach for assessing EMP, revealing that tumor cells occupy a continuous spectrum of EMT states and that metastatic potential peaks in late hybrid EMT states.¹⁵⁶

Various cells in the tumor ecosystem exist in hybrid or intermediate states. Tumor cells in the E/M hybrid state play crucial roles in invasion, metastasis, and drug resistance. For example, the acquisition of a hybrid state is crucial for tumorigenesis in basal breast cancer cells.²⁴ Moreover, a single-cell lineage analysis revealed that metastatic cancer cells exhibit a hybrid EMT state.¹⁵⁶ Accordingly, single-cell proteomic techniques are instrumental in exploring tumor heterogeneity and elucidating the signaling pathways that drive tumor progression. For example, mass cytometry (CyTOF) has been used to classify human breast cancer into luminal subfractions.¹⁵⁷ Non-targeted single-cell proteomic analysis dissects tumor tissue into individual cells and can hopefully be translated into clinical applications.¹⁵⁸ The transitional and continuous nature of the EMT and MET spectra requires single-cell omics analysis to reveal the heterogeneity and dynamics of the EMP regulatory epigenome, decipher the pathways that regulate EMT plasticity in different contexts, guide the future development of the transformation and application of the concept of EMP in diseases, and provide a new perspective for understanding EMP.¹⁵⁹ However, fully capturing the dynamic and reversible cellular biological processes underlying EMT/MET remains highly challenging. Table 2 summarizes the methods currently used to assess EMP, highlighting their respective measurements, characteristics, and limitations.

Table 2.

Approaches for evaluating EMP

Technical methods	Measurement contents	Measurement attributes	Main limitations	Reference
Immunohistochemistry	expressions, locations, and distributions of specific proteins in the tumor tissues	static	fail to capture the dynamic transformation process or reflect the reversibility	Sukswai et al.160 Hofman et al.161 Zeisberg et al.162
Phenotypic analysis of CTCs	expressions of epithelial and mesenchymal markers in the isolated and enriched CTCs	static	only reflects the state at the sampling, and the low capture efficiency	Yu et al.142 Boya et al.163
Single-cell RNA sequencing	mapping tumor cells on the EMT continuum by profiling various related genes	dynamic	single-time-point data fail to directly prove the trajectory of the transformation	Lawson et al.164 Feng et al.165 Winkler et al.166
Lineage tracing	genetically marking the epithelial cells and tracking the obtained mesenchymal profiles in their descendants	dynamic and with temporal continuity	poor applicability of clinical samples due to the complex operation	Simeonov et al.156 Kretzschmar et al.167
Spatial transcriptomics	measuring gene expression while preserving the spatial positions	static but retains spatial information	dynamic information is insufficient, and the resolution is limited	Robles-Remacho et al.168 Malagoli et al.169

Abbreviation: EMP, epithelial-mesenchymal plasticity; CTCs, circulating tumor cells; EMT, epithelial-mesenchymal transition.

Targeting EMP as a potential therapeutic

EMP not only empowers tumor cells to switch their functional phenotype according to the need for invasion and metastasis but also significantly contributes to cancer treatment resistance. Tumor cells in the E/M hybrid state have a heightened capacity to adapt to the TME.¹⁷⁰ GRHL2 and OVOL2, recognized as MET-TFs, exert inhibitory effects on mesenchymal-related molecule expression while promoting the MET program, thereby stabilizing the hybrid E/M state, facilitating the formation of CTC clusters, and ultimately leading to distant metastasis.^105,171 The knockout of GRHL2 and OVOL2 results in tumor cells transitioning from an incomplete E/M state to a complete mesenchymal transition state.¹⁰⁵ Targeting MET-TFs is a potential therapeutic strategy for disrupting the hybrid E/M phenotype and preventing metastasis. miRNAs regulate EMT and MET by interacting with EMT-TFs.¹⁷² The feedback loops involving ZEB and miR-200, as well as Snail1 and members of the miRNA-34 family, regulate cell fates among different tumor cells with EMP.¹⁷³ Therefore, these miRNAs can be used as therapeutic targets to prevent tumor metastasis and treatment resistance.^174,175 The crosstalk between EMT-TFs and MET-TFs, along with non-coding RNAs, collectively regulates EMP in tumor cells. Targeting EMP could therefore serve as a novel approach to impede tumor metastasis and suppress resistance to chemotherapy.¹²

Currently, targeted therapies for EMP are principally focused on core regulatory pathways, EMT-TFs, and the TME. Vactosertib, a TGF-β type I receptor inhibitor, can alleviate radiation-induced EMT and tumor stem cell characteristics in breast cancer.¹⁷⁶ Combining the TGF-β receptor II inhibitor SHR-1701 with chemotherapy has shown promising efficacy in treating unresectable stage III NSCLC.¹⁷⁷ The safety and tolerability of histone deacetylase inhibitors regarding epigenetic modifications have been tested in phase I/II clinical trials.^178,179 Metabolic inhibitors can be used to antagonize EMP-driven metabolic reprogramming. Telaglenastat, a glutaminase 1 inhibitor, is currently undergoing clinical trials for the treatment of metastatic melanoma, renal cell carcinoma, and other diseases.¹⁸⁰ Isocitrate dehydrogenase 1 (IDH1) participates in metabolic processes and is closely related to epigenetic regulation.¹⁸¹ Using the IDH1 inhibitor ivosidenib to treat patients with advanced cholangiocarcinoma carrying IDH1 mutations has been shown to provide clinical benefits.^182,183 Inhibiting MET is also an effective therapeutic strategy. It restores tumor cells’ sensitivity to standard treatments by inhibiting functional mesenchymal-specific proteins and cell plasticity, while inducing re-differentiation or trans-differentiation.¹⁸⁴ In the United States, the focal adhesion kinase (FAK) inhibitor defactinib has been approved for use in combination therapy for low-grade serous ovarian cancer.^185,186 APG-2449, an orally active FAK inhibitor, has demonstrated promising safety, pharmacokinetics, and efficacy in patients with NSCLC who have not previously received treatment with tyrosine kinase inhibitors or who have developed resistance to second-generation ALK inhibitors.¹⁸⁷ In addition, panitumumab, an EGF receptor inhibitor, has been approved for use in the treatment of refractory colon cancer.¹⁸⁸ The miRNAs that regulate EMP are expected to be ideal tools and targets for novel therapeutic approaches. However, miRNA-based therapies remain in their early stages, with no candidates yet having entered phase III clinical trials or received approval from the US Food and Drug Administration.¹⁸⁹ A comprehensive risk assessment of miRNA therapies is of the utmost importance. Given that the EMT/MET process is highly dynamic and reversible, leading to tumor cell heterogeneity, it is difficult to achieve a long-term response with a single targeted treatment. Therefore, a comprehensive treatment plan should be developed that addresses multiple aspects of EMP, including the use of multiple targets in combination with concurrent radiotherapy and chemotherapy.⁸ Table 3 shows the most recent clinical trials of therapies targeting EMP.

Table 3.

Clinical trials for the treatments targeting EMP

Principle of the therapy	Therapeutic targets	Drugs	Phase of the trials/Stage	Types of cancers	Clinical outcomes	Reference
Targeting the key signaling pathways regulating EMP	Wnt/β	WNT974	Ib/II	metastatic colorectal cancer	phase II was not initiated due to low safety	Tabernero et al.190
	NOTCH2/3	Tarextumab	I/II	metastatic pancreatic cancer	poor therapeutic effect	Hu et al.191
	TGF-βR1	Vactosertib	Ib/II	desmoid tumors	improve PFS	Ahn et al.192
		Galunisertib	II	rectal cancer	increase CRR	Yamazaki et al.193
	EGFR	Panitumumab	marketed	colorectal cancer	improve PFS	Modest et al.188
	TGF-βR2	SHR-1701	II	NSCLC	good therapeutic effect	Zhou et al.177
Metabolic and epigenetic inhibitors	GLS	Telaglenastat (CB-839)	I/II	metastatic melanoma, renal cell carcinoma, NSCLC	poor therapeutic effect, good tolerance	Gouda et al.180
	IDH1	Ivosidenib	III	cholangiocarcinoma	improve PFS and OS	Abou-Alfa et al.182 Zhu et al.183
	HDAC	Vorinostat	I/II	metastatic squamous cell carcinoma	partially improve ORR	Borcoman et al.178
		Mocetinostat	I/II	lung cancer	low ORR but with long-lasting efficacy	Johnson et al.194
Targeting ECM and inhibiting MET	FAK	Defactinib	marketed	low-grade serous ovarian cancer	improve ORR	Kabirian et al.185 Blair et al.186 Banerjee et al.195
		APG-2449	I	NSCLC	acceptable tolerance	Ma et al.187
	EGFR, c-MET	Amivantamab	III	EGFR-mutated advanced NSCLC	improve PFS	Cho et al.196

Abbreviation: EMP, epithelial-mesenchymal plasticity; PFS, progression-free survival period; CRR, complete response rate; NSCLC, non-small cell lung cancer; GLS, glutaminase; IDH1, isocitrate dehydrogenase 1; OS, overall survival; HDAC, histone deacetylase; ORR, objective response rate; ECM, extracellular matrix; FAK, focal adhesion kinase; MET, mesenchymal-epithelial transition.

Conclusion

Distant metastasis is the primary contributor to cancer-related mortality, and chemotherapy resistance is detrimental to clinical prognoses. The process of EMT or MET is not a singular event, but rather a dynamic process in which tumor cells exist in various intermediate states along the EMT axis. Tumor cells in these intermediate states exhibit E/M hybrid functional phenotypic characteristics. The capacity for epithelial-mesenchymal bidirectional transition enables the dissemination of tumor cells to adapt to changes in the invasive microenvironment during the multi-step process of metastasis. On this basis, EMP determines the invasiveness of tumor cells and their responsiveness to treatment and is therefore closely correlated with clinical outcomes. Unfortunately, the incomplete inhibition of EMT may result in the arrest of tumor cells at the E/M hybrid phenotype phase, thus potentially promoting stemness and resistance to treatment and increasing the risk of distant metastasis.⁸ Uncovering the signaling pathways that modulate and maintain EMP and identifying key regulatory molecules could lead to the development of novel and effective therapeutic strategies for preventing EMP-driven tumor progression and treatment resistance.

The controversy over whether EMT is strictly required for metastasis continues. This is because EMT is not a binary state, but a continuum. In this regard, a complete EMT program may not be necessary for tumor metastasis. In particular, when distant metastasis is considered to be the spatial migration of the entire tumor cell population rather than a single tumor cell, complete EMT does not seem to be required for every tumor cell. Findings from a recent study indicated that metastatic cells retain an epithelial phenotype and that EMT contributes to therapy resistance.¹⁹⁷ Similarly, this phenotypic transition reflects adaptive EMP under treatment-related survival stresses, and the EMT program is dynamic and incomplete.

The integration of EMP-aware diagnostics and therapies into cancer care pathways remains a major challenge. First, achieving effective EMP evaluation requires a shift from static assessments to dynamic monitoring. Combining traditional immunohistochemistry with multi-timepoint single-cell sequencing of tumor cell subpopulations may serve as a standard approach for assessing EMP. Second, EMP scoring could provide a quantitative indicator of metastatic risk and treatment resistance. However, the unified criteria for classifying EMP, based on molecular marker detection, gene expression profiling, DNA methylation scoring, and other metrics, have not yet been established. Third, reversing the EMT program may represent a promising strategy to counter EMP-driven tumor progression.¹⁹⁸ Nevertheless, incomplete EMT inhibition by manipulating MET-TFs can trap tumor cells in a hybrid E/M state, potentially enhancing stemness, promoting therapy resistance, and increasing the risk of distant metastasis.⁸ The safety and efficacy of EMP-targeting drugs must therefore be rigorously evaluated in well-designed clinical trials.

Acknowledgments

This work was supported by the Guangxi Natural Science Foundation (2023GXNSFAA026179 and 2023GXNSFAA026033), the National Natural Science Foundation of China (82560501 and 82073004), the First-class discipline innovation-driven talent program of Guangxi Medical University granted to J.Z., the Key Talent Program of Guangxi Zhuang Autonomous Region (Bagui Young Excellence Talents) awarded to J.W., and the Guangxi Medical and Health Key Discipline Construction Project. We would like to thank Editage (www.editage.cn) for English language editing.

Author contributions

Conceptualization: J.W. and J.Z.; data curation: Y.L., L.H., Y.H., B.C., Q.F., X.L., Y.D., J.Y., F.L., Y.L., J.W., and J.Z.; funding acquisition: J.W. and J.Z.; visualization: Y.L., Y.H., and J.W.; writing – original draft: Y.L., L.H., and J.W.; writing – review and editing: Y.L., L.H., J.W., and J.Z. All the authors approved the submission of the manuscript.

Declaration of interests

The authors declare no conflict of interest.