Mechanisms Underlying the Initiation and Termination of Collective Cell Migration: Perspectives for Understanding Development and Cancer Metastasis

ABSTRACT

Collective cell migration is fundamental to developmental processes and disease progression. Despite extensive study, the field lacks a unifying framework for how collective cells initiate and terminate their migration. While these processes have traditionally been explained for individual cell migration by epithelial‒mesenchymal transition (EMT) and mesenchymal‒epithelial transition (MET), these models do not fully recapitulate the complex features of collective cell migration. In this review, I explore the distinct mechanisms by which groups of cells initiate and terminate collective migration, highlighting in vivo examples such as gastrulation and neural crest formation in vertebrates, lateral line migration in zebrafish, and tracheal branch and border cell migration in Drosophila. I also discuss collective cell migration in cancer metastasis. I focus on how the initiation and termination of collective migration are regulated, emphasizing the regulatory pathways and unique features. Clarifying these mechanisms will guide hypothesis‐driven discovery and inform strategies to modulate collective cell behaviors in development, regeneration, and metastasis.

How collective cell migration starts and stops is a fundamental question but remains poorly understood. This review discusses the distinct mechanisms in in vivo models and cancer metastasis that define the regulatory pathways and collective features governing the initiation and termination of collective cell migration.

1. Introduction

In the context of biological activities, cells frequently move from one place to another by dynamically initiating and terminating interactions with their neighbors and the surrounding environment [1, 2, 3]. This dynamic interplay is essential for normal developmental processes as well as the progression of various diseases [4]. Although cells can migrate individually, collective cell migration—in which cells move cohesively in clusters, strands, sheets, or streams—is prevalent in numerous developmental contexts and pathological conditions, including cancer metastasis [5, 6, 7].

Understanding the precise mechanisms by which cells initiate and terminate migration under physiological conditions is crucial for advancements in fields such as tissue engineering and regenerative medicine, offering potential therapeutic strategies for repairing damaged tissues and organs. Conversely, these mechanisms can also be hijacked by pathological cells, particularly in cancer, where they contribute significantly to metastasis, a complex, multistep process involving the detachment of tumor cells, migration through diverse environments, and colonization at distant sites; this process remains a major obstacle in cancer treatment [8, 9].

Historically, epithelial‒mesenchymal transition (EMT) and its reverse, mesenchymal‒epithelial transition (MET), have been central to understanding how epithelial cells initiate and terminate their movement [10, 11]. According to the classic EMT model, cells detach and migrate as individual entities, severing their original epithelial connections. However, collectively, cells may not require the full EMT process to initiate movement as they maintain cell‐cell contact. For example, recent research on cancer metastasis has revealed that collective migration might be mechanistically different and potentially more efficient in specific cancer types, facilitating metastasis up to 100 times more effectively than individual cell migration [7, 12, 13]. Emerging evidence indicates that EMT is not necessarily a prerequisite for metastasis in multiple cancer types [12, 14, 15, 16].

Despite extensive studies on collective cell migration, significant knowledge gaps remain regarding the regulation of the initiation and the subsequent re‐establishment of cellular connections at new locations. How cells coordinate during these dynamic processes are still largely unknown. Central questions include how leader and follower identities are specified, stabilized and exchanged; how collective cells maintain intercellular coupling and communication. In this review, I critically evaluate current insights into the mechanisms governing the initiation and termination of collective cell migration across in vivo models—vertebrate gastrulation and neural crest migration, zebrafish lateral line formation, and Drosophila tracheal branching and border cell migration—and draw parallels to carcinoma collectives in metastasis. My aim is to identify conserved versus context‐specific mechanism and to articulate a unifying start‐stop framework for collective cell migration.

2. EMT and MET: Classic Models to Explain How Epithelial Cells Initiate and Terminate Migration

Epithelial cells normally maintain a stationary structure characterized by stable cell‒cell adhesion, defined apical–basal polarity, and an organized cytoskeletal architecture. EMT in epithelial cells involves the loss of cell polarity and adhesion properties and the acquisition of migration and invasion capabilities; through this process, epithelial cells transform into mesenchymal cells. Conversely, MET refers to the reverse transition from mesenchymal cells back to epithelial cells, in which cells regain cell adhesion and polarity [10, 17, 18, 19, 20, 21]. In Table 1, I summarize the key features of EMT and MET, including characteristics, transcription factors, signaling pathways and gene expression changes involved.

TABLE 1.

Key features of classic epithelial‒mesenchymal transition (EMT) and mesenchymal‒epithelial transition (MET) models.

Characteristics of EMT and MET
EMT	MET
Loss/decrease of cell adhesion: Epithelial cells lose cell‒cell adhesions and apical polarity and undergo cytoskeletal changes	Gain/increase of cell adhesion: Mesenchymal cells upregulate E‐cadherin expression and regain cell‒cell adhesions
Increased motility: Cells acquire increased motility, invasiveness, and resistance to apoptosis	Decreased motility: The cytoskeleton reverts to a state that supports an epithelial phenotype
Cells secrete enzymes that degrade the extracellular matrix (ECM)	Cells modify the ECM to support epithelial structures
Adoption of mesenchymal phenotypes: Cells assume mesenchymal phenotypes, developing front–rear migratory polarity and exhibiting stem cell‐like properties	Adoption of epithelial phenotypes: Cells regain apical polarity and fully assume an epithelial phenotype
Key transcription factors in EMT
Twist, Snail, Slug/Snail2, ZEB1, and ZEB2
Signaling pathways inducing EMT
TGF‐β pathway activates SMAD proteins, induces expression of EMT‐related genes Wnt pathway induces expression of EMT‐related genes through beta‐catenin Notch pathway induces expression of EMT‐related genes, such as Snail EGF pathway induces expression of EMT‐related genes through PI3K/AKT and MAPK pathways Hedgehog pathway induces expression of EMT‐related genes, such as Snail and Snail2
Key gene expression changes
EMT	MET
Downregulated: E‐cadherin and cytokeratin Upregulated: N‐cad, vimentin, fibronectin, matrix metalloproteinases (MMPs), and stem cell markers (CD44 and Sox2)	Downregulated: N‐cadherin, vimentin, fibronectin, and matrix metalloproteinases (MMPs) Upregulated: E‐cadherin and cytokeratin

One in vivo example of EMT is vertebrate gastrulation (Figure 1a), a critical developmental event during early embryogenesis [22, 23, 24]. Gastrulation transforms the simple epithelial structure of the blastula into the more complex, multilayered gastrula, forming the three primary germ layers: the ectoderm, mesoderm, and endoderm. In amniotes, including mammals, birds, and reptiles, mesodermal precursor cells within the epiblast undergo EMT. The Nodal/TGFβ, Wnt, bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) signaling pathways activate and regulate the EMT process [25, 26, 27], during which the cells lose their epithelial characteristics and acquire mesenchymal traits, enabling them to ingress individually through the primitive streak and migrate extensively. This EMT‐driven migration during gastrulation has been well studied in mouse and chick embryos [28, 29]. Interestingly, gastrulation also includes a preceding MET phase. Before mesodermal and endodermal specification, embryonic cells undergo MET, forming the pluripotent epithelial epiblast (Figure 1a, before migration). This structured epithelial tissue subsequently serves as the origin of migratory mesodermal cells through EMT (Figure 1a, initiation) and undergoes MET to form the parietal endoderm (Figure 1a, termination), revealing a precisely coordinated MET to EMT to MET sequence that is critical for embryonic morphogenesis.

FIGURE 1.

In vivo examples of epithelial‒mesenchymal transition (EMT) and mesenchymal‒epithelial transition (MET). (a) Gastrulation in the mouse embryo. Following fertilization, the mouse egg enters cell division through cleavage. By embryonic day 3.75 (E3.75), the trophectoderm (pink) has acquired a fully polarized epithelial structure. By E4.0, MET is initiated in primitive endodermal cells (green), sorting them from the inner cell mass (ICM). By E4.5, the primitive endoderm expands along the mural trophectoderm and undergoes partial EMT at the leading edge. By E4.75, cells in the ICM start reorganizing their intercellular contacts to acquire apicobasal polarity. At E5.0, MET is fully executed to form the parietal endoderm. Adapted from [21]. (b) Neural crest cell migration in the chick embryo (original illustrations). During delamination, neural crest cells undergo partial EMT: cells weaken their adhesions, lose apical–basal polarity and acquire polarized motility to delaminate. Neural crest cell migration contributes to the development of a variety of tissues and organs.

Another example of EMT is the migration of neural crest cells in the chick embryo (Figure 1b). The neural crest comprises a unique multipotent stem cell population in vertebrates that originates from the ectoderm. The neural crest forms at the boundary between the neural plate (which later becomes the central nervous system) and the adjacent nonneural ectoderm (future epidermis) (Figure 1b, before migration). As the neural plate folds inward to form the neural tube, neural crest precursor cells are initially integrated into the neuroepithelium at the dorsal neural folds. Multiple transcription factors and signaling pathways regulate neural crest cell migration [30, 31]. Migration is initiated when these premigratory neural crest cells undergo EMT and delaminate from their epithelial layer (Figure 1b, initiation and [32, 33, 34]). After delamination, the neural crest cells migrate extensively throughout the developing embryo, eventually settling at diverse anatomical sites (Figure 1b, termination). At these locations, they differentiate into a wide array of cell types, including peripheral neurons, glial cells, and melanocytes, and contribute to the formation of structures such as the craniofacial cartilage.

Gastrulation and neural crest migration establish EMT and MET as canonical mechanisms for starting and stopping cell migration and highlight context‐dependent MET at destinations. However, classic models incompletely explain collective cell migration, where cell‐cell coupling persists, hybrid epithelial‐mesenchymal states are common, and leader‐follower roles must be specified.

3. Redefining EMT: The Concepts of Partial and Intermediate EMT

Recent studies have expanded the EMT paradigm to include a spectrum of intermediate states between epithelial and mesenchymal phenotypes. The concepts of “partial EMT” and “intermediate EMT” recognize that cells often adopt hybrid or transient epithelial/mesenchymal states rather than undergoing full epithelial–mesenchymal or mesenchymal–epithelial transitions [20, 35, 36, 37]. Such states allow for greater plasticity and functional diversity, particularly in contexts involving collective cell migration.

For example, during neural crest migration (Figure 1b, initiation), cells retain partial epithelial characteristics, including cadherin‐based contacts, while expressing mesenchymal markers; this intermediate configuration is often referred to as E‐cadherin/N‐cadherin heterotypic adhesion. Time‐lapse microscopy of delaminating neural crest cells in chick embryos revealed that motility is acquired prior to the downregulation of cell–cell adhesion, suggesting that the loss of adhesion is not a prerequisite for delamination [38]. Similarly, in frog and zebrafish embryos, the physical separation of neural crest cells is largely driven by traction forces that pull cells away from intercellular contact sites rather than by the complete dissolution of adhesive junctions [39].

In amniote gastrulation, definitive endoderm precursors undergo transient EMT followed by MET as they integrate into an epithelium formed from the primitive endoderm, resulting in partial EMT (Figure 1a, initiation and [23]).

In both classic and partial EMT, hallmark features include the dismantling of apicobasal polarity, the downregulation of tight and/or adherens junctions, and the acquisition of front–rear migratory polarity [40]. These transitions are largely driven by transcription factors such as Twist, Snail, Slug (Snail2), ZEB1, and ZEB2, which repress the expression of epithelial genes, particularly E‐cadherin (Table 1).

Partial and intermediate EMT are being increasingly recognized across developmental and disease contexts. Quasimesenchymal states—characterized by partial cell–cell adhesion and collective cell motility—have been observed in the endoderm and mesoderm cells of Drosophila melanogaster, zebrafish, frog, and mouse embryos, as well as in the neural crest cells of zebrafish, chick, and frog embryos [39, 41, 42, 43, 44, 45, 46, 47]. Additionally, in kidney development, MET is essential for nephrogenesis, as mesenchymal cells derived from the intermediate mesoderm become organized epithelial nephron structures [21]. In cancer metastasis, EMT facilitates tumor cell invasion and dissemination. Upon reaching metastatic sites, many disseminated cells undergo MET to establish epithelial colonies [8, 36, 48, 49].

These examples highlight the redefined EMT as a highly adaptable morphogenetic mechanism capable of driving both individual and collective migratory behaviors depending on the developmental stage and tissue context.

4. Collective Versus Individual Cell Migration: Unique Features of Collective Cell Migration

While the EMT model helps explain how certain groups of cells initiate and terminate collective migration, it does not fully recapitulate the complexity of all collective cell behaviors. Collective migration involves unique features that extend beyond the EMT framework. A defining characteristic of collective migration is the need for cells to coordinate dynamically with one another—at both the initiation and the termination of migration—to function as a cohesive unit.

Multilevel polarity, the maintenance of connections, and coordinated cell behaviors are key distinguishing features of collective migration. (i) Multilevel polarity refers to the fact that in many cases, migrating collectives not only exhibit front–rear polarity as a group but also retain apicobasal polarity as individual cells or a group. (ii) The maintenance of connections between cells, such as adherens junctions mediated by cadherins (especially E‐cadherin), frequently occurs during collective migration, enabling structural cohesion and mechanical communication among the collective. (iii) Coordinated cell behaviors require a collective response to guidance cues that depends on the dynamic coordination between leader cells and follower cells. Cells must act as a coordinated group during the dynamic transitions involved in the initiation and termination of collective migration. The regulation of these features during the initiation and termination of collective cell migration remains incompletely defined.

Here, I discuss the collective features and regulation across three well‐studied in vivo models—lateral line primordium cell migration in the zebrafish embryo, branching morphogenesis in the developing Drosophila tracheal system, and border cell migration in the Drosophila ovary—and draw parallels to collective invasion and colonization cancer metastasis.

5. Zebrafish Lateral Line Primordium Cell Migration

The lateral line of anamniote vertebrates is an organ used to detect local water flow and weak bioelectric fields [50]. The zebrafish lateral line develops from a migrating primordium (Figure 2a), a compact cluster of approximately 100 cells, that initially forms posterior to the otic vesicle [51]. This primordium migrates posteriorly along the horizontal myoseptum toward the tail tip, periodically depositing small sensory structures called neuromasts (Figure 2a, termination and [52]). The posterior lateral line primordium emerges between 19 and 25 h postfertilization; cells of the lateral line undergo proliferation and move along the embryonic axis, and clusters of 20–30 cells are sequentially deposited as neuromasts at regular intervals of every 5–7 somites [53]. Typically, five to six primary neuromasts (L1–L6) form along the trunk and tail, with an additional two to three terminal neuromasts created at the tail tip following fragmentation of the primordium [53].

FIGURE 2.

In vivo models of collective cell migration (original illustrations). (a) Lateral line primordium cell migration in zebrafish. The signaling pathway network shown in the left panel is adapted from [57]. (b) Dorsal branching morphogenesis in the Drosophila tracheal system. (c) Border cell migration in the Drosophila ovary.

5.1. Before Migration: Primordium Patterning

Prior to migration, the primordium becomes clearly polarized and is divided into distinct leading and trailing regions. The anterior region (leading zone) has a loosely organized, “pseudomesenchymal” appearance, whereas the posterior region (trailing zone) consists of organized rosette‐shaped cell clusters known as protoneuromasts (Figure 2a, before migration and [52, 53]). Cells within protoneuromast rosettes display apical–basal polarity, apical constriction, and basal nuclear positioning, which are typically features of epithelial cells in stable structures. This polarity is regulated by cell polarity proteins such as Lethal giant larvae 1 and 2 (Lgl1 and Lgl2), which are critical for the formation and stabilization of these rosette structures [54]. Cells remain in close contact via cell‒cell junctions during migration. Specifically, E‐cadherin and N‐cadherin are both expressed in the primordium but are specifically localized within protoneuromasts (Figure 2a, initiation and [55]). All cells in the primordium need to sense the attractant molecule and adhere to each other to coordinate their movements [55].

Chemokine signaling through the receptors cxcr4b and cxcr7b contributes significantly to primordium polarity and directed migration. The chemokine ligand cxcl12a (SDF1a) is expressed along the horizontal myoseptum and delineates the migratory path for the primordium. Additionally, interactions between the Wnt/β‐catenin and FGF signaling pathways shape regional patterning within the primordium. Specifically, FGF signaling primarily promotes rosette formation, whereas Wnt/β‐catenin signaling modulates FGF activity (Figure 2a, before migration). Furthermore, Delta and Fgf ligands are expressed by central cells within each rosette and are essential for hair cell progenitor specification via lateral inhibition and subsequent hair cell differentiation (reviewed previously in [51, 53, 56, 57]).

5.2. Migration Initiation: Response of Cells to Cxcl12a and Partial EMT

The onset of lateral line primordium migration coincides closely with the expression of the chemokine receptor cxcr4b in leading‐region cells and its ligand cxcl12a along the migratory path (Figure 2a, before migration and [58, 59]). Although chemokine signaling via cxcr4b/cxcl12a is critical for sustained directional migration, studies have indicated that migration initiation might precede chemokine‐dependent cues. Observations in cxcr4b and cxcl12a mutants indicate that initial migration still occurs, albeit to a limited extent, suggesting that additional signals may also contribute to the initiation of migration [50, 58].

Recent studies have proposed that the transcription factor Snail1b, which is triggered by cxcl12a and expressed specifically in leading‐region cells, may facilitate the initiation of migration through a partial EMT‐like process [60]. Thus, the initiation of migration involves both chemokine signaling and potential partial EMT‐mediated loosening of epithelial connections, enabling cells to initiate posterior movement.

5.3. Migration Termination: Neuromast Deposition and Cessation of Migration

Primordium migration terminates through a highly coordinated deposition process. As the primordium migrates caudally, the most mature protoneuromast rosettes positioned at the trailing end gradually terminate forward movement [52, 53]. Once incorporated into stable epithelial rosettes, these trailing cells lose cohesive interactions with cells remaining in the primordium and are deposited along the horizontal myoseptum as neuromasts. Cells that are not fully integrated into rosettes settle in between as interneuromast cells, which serve as progenitors for the generation of additional neuromasts (Figure 2a, termination and [61]).

The termination of migration and the timing of deposition are tightly controlled by chemokine receptor cxcr7b, the expression of which progressively expands throughout the protoneuromast rosette (Figure 2a, termination). When this occurs, cxcr7b‐positive cells terminate migration and are subsequently deposited [62]. Moreover, cell proliferation, which is regulated by Wnt and FGF signaling within the primordium, directly influences the timing and spacing of neuromast deposition, ensuring appropriate lateral line morphogenesis [62].

The mechanisms of some migration‐related cell behaviors remain to be explained, including the differential adhesion among the cells in a proneuromast rosette within the primordium and the lack of adhesion to continuously deposited interneuromast cells. A decrease in N‐cadherin expression in interneuromast cells may facilitate continuous deposition [55].

Analogous to mechanisms identified in primordial germ cells and other migratory cell systems, a uniform and elevated Cxcl12a concentration near the tail tip may act as a signal to terminate further cell migration [50, 56]. Mosaic analyses also revealed that FGF signaling is critical for maintaining the correct timing and spacing of neuromast formation and deposition [63].

Ultimately, these sequential cycles of rosette maturation, chemokine signaling, and controlled proliferation ensure that the primordium systematically terminates migration and deposits precisely spaced neuromasts along the developing zebrafish lateral line.

In summary, zebrafish lateral line migration illustrates how collective cells start their movement via leader‐follower polarization, chemokine‐guided bias, and partial EMT‐like adhesion tuning, and stop through cxcr‐7b‐mediated neuromast deposition. Together, these mechanisms clarify the regulatory logic of start‐stop control in collective cell migration.

6. Branching Morphogenesis in the Drosophila Tracheal System

The Drosophila tracheal system is a highly branched tubular network responsible for delivering oxygen to tissues, and this system serves as an established model for studying collective cell migration (reviewed in [64]). During embryogenesis, the tracheal system originates from 10 pairs of ectodermal placodes in body segments [65]. These placodes are specified by the transcription factors Trachealess (trh) and Ventral veinless (vvl) and require downstream FGF signaling to establish tracheal cell identity [66, 67, 68, 69]. Once specified, the tracheal placodes invaginate to form compact epithelial sacs [65]. From these sacs, tracheal cells extend and migrate in six stereotyped directions to generate primary branches [70, 71, 72]. The FGF ligand Branchless (Bnl), expressed in adjacent tissues, acts as a crucial chemoattractant [71]. Its receptor, Breathless (Btl/FGFR), is expressed in tracheal cells and mediates directional migration [73, 74]. At the end of migration, specialized fusion cells at the tips of specific branches meet and undergo anastomosis, forming continuous lumens and completing the gas transport network [75]. Specifically, the collective migration of dorsal branches toward the dorsal epidermis, which connects the tracheal tubes by the end of embryonic development, has been well studied [76, 77, 78, 79, 80, 81].

6.1. Before Migration: Definition of Leaders and Followers

Prior to dorsal branch outgrowth, leader (tip cell) and follower (stalk cell) identities are specified (Figure 2b, before migration). Btl signaling is selectively activated in two prospective tip cells at the leading edge [70]. Genetic mosaic analyses have shown that Btl activity is necessary only in tip cells to drive migration [82]. Ectopic expression of activated Btl is sufficient to reprogram a follower into a tip cell, initiating branch outgrowth [81, 83]. Positive feedback reinforces Btl transcription in tip cells, solidifying their identity [70]. Initial broad induction of Delta expression by Bnl/FGF and Wingless signaling is subsequently refined by Notch‐mediated lateral inhibition, ensuring the formation of Delta‐high tip cells and neighboring stalk cells [84, 85, 86, 87].

Thus, the selective activation of the Bnl/FGF signaling pathway distinctly divides tracheal cells into two populations: tip cells—ERK‐active cells that directly respond to and migrate toward Bnl signals, maintaining their leading position throughout the subsequent migration process—and stalk cells that follow the tip cells.

6.2. Migration Initiation: FGF‐Driven Tip Cell Motility

Dorsal branch migration is initiated as tip cells respond to localized Bnl/FGF expression (Figure 2b, initiation). Tip cells extend filopodia and lamellipodia, driven by active Btl signaling, and migrate toward the FGF source [72, 88]. The trailing stalk cells remodel cell–cell junctions and undergo cell intercalation to elongate the branch, forming narrow unicellular tubes supported by E‐cadherin turnover [89, 90]. This process illustrates a case of leader–follower coordination, a feature unique to collective cell migration.

6.3. Migration Termination: Fusion Cell‐Mediated Branch Fusion

Migration terminates when one of the two tip cells at opposing dorsal branches adopts a fusion cell identity (Figure 2b, termination), a process regulated by Wingless signaling. These cells sequentially express the transcription factors escargot (esg) and dysfusion (dysf), which are required for their specialized role [78, 79, 84, 85, 91]. Meanwhile, FGF downstream ERK signaling decreases in fusion cells but increases in adjacent terminal cells; terminal cells differentiate into blistered (DSRF)‐positive cells and flatten along the epidermis to form terminal branches [92, 93]. This transition from initial FGF‐mediated migration to terminal differentiation is regulated by esg, which represses ERK activity specifically in future fusion cells [80].

The physical process of migration termination occurs through branch fusion or anastomosis. After epidermal closure, the fusion cells, which are positioned at the basal surface, migrate to the dorsal midline. These cells extend filopodia to contact their counterparts from opposing segments. E‐cadherin accumulates at these interfaces, driven by esg‐dependent synthesis, and promotes selective adhesion [79]. While fusion cell filopodia transiently interact with other cells, stable contact is selectively maintained between opposing fusion cells, although the mechanism underlying the specificity of this interaction is still not fully understood. After establishing contact, fusion cells initiate de novo lumen formation through Golgi‐directed secretion, connecting adjacent branches into a continuous tubular network [77].

In summary, tracheal branching in Drosophila exemplifies a precisely coordinated program of collective cell migration with clear initiation and termination phases that occur independently of EMT. Initiation is mediated by selective Bnl/FGF signaling in leader cells, while migration terminates through the differentiation of fusion cells that enable tube connectivity and lumen continuity. This system continues to offer powerful mechanistic insights into epithelial morphogenesis and collective cell migration.

7. Border Cell Migration in the Drosophila Ovary

Border cells are a group of cells in the Drosophila ovary that undergo collective migration (reviewed in [94]). The Drosophila ovary consists of egg chambers at different stages of development. Each egg chamber comprises 15 large nurse cells and an oocyte, known as the germline cells, all of which are nestled within a nurturing envelope of somatic follicle cells. During stage 9, in the anterior of the egg chamber, a select group of follicle cells, that is, the border cells, form a cluster that includes two centrally positioned polar cells (Figure 2c, before migration). As the border cells round up and project dynamic protrusions, they detach from the extracellular matrix (ECM) and initiate migration (Figure 2c, initiation). This meticulously coordinated process is termed delamination. Then, the border cell cluster (4–8 border cells with 2 polar cells) migrates toward the oocyte, covering a distance of approximately 150 µm in 4–6 h and reaching its destination—the oocyte—by stage 10 (Figure 2c, termination). The border cells then align dorsally with the oocyte nucleus, establishing stable interactions with the oocyte membrane. Meanwhile, centripetal follicular cells (cfcs) commence their inward migration, converging on the same region and interacting with the border cells. This collective movement culminates in “neolamination” and the formation of a micropyle, a crucial passage for sperm entry during fertilization [95]. This final orchestrated series of events—border cell‒oocyte and border cell‒centripetal cell interactions—is integral to the successful completion of border cell migration [96].

7.1. Before Migration: Border Cell Specification

In stage 8, border cell fate is established through the Janus kinase/signal transducer and activator of transcription (JAK–STAT) signaling pathway (Figure 2c, before migration and [97, 98]). Two polar cells positioned at the anterior end of egg chamber secrete the ligand Unpaired (Upd). Follicular cells surrounding these polar cells express varying levels of the STAT receptor; the cells with the highest level of STAT activation adopt a border cell fate [97, 98]. Subsequently, 4–8 specified border cells activate the downstream transcription factor slow border (slbo), which further triggers the expression of downstream target genes [95]. The border cells then round up and form a cohesive cluster around the polar cells [97, 98]. Importantly, this cell cluster remodels while maintaining epithelial polarity and ultimately detaches from the anterior follicular epithelium to initiate migration.

7.2. Migration Initiation: Border Cell Cluster Detachment From the Anterior End of the Egg Chamber

Border cell migration initiates when the cluster detaches from the anterior follicular epithelium of the egg chamber during the process of delamination (Figure 2c, initiation). PVF1 and EGF ligands (Karen and Spitz) act as guidance cues to initiate border cell delamination. PVF1–PVR signaling in border cells occurs through myoblast city (Mbc, a Rac activator) to induce Rac signaling, which then controls F‐actin accumulation [99]. PVF1, together with the ecdysone pathway, regulates the distribution of E‐cadherin in border cells via the ecdysone receptor (EcR) Taiman (Tai) [100, 101].

During delamination, one or two border cells (leading cells) lead migration, while others (follower cells) retract their extensions and follow. E‐cadherin plays a central role in communicating the direction of migration from leading cells to follower cells [102]. Critically, the polarity and epithelial characteristics of the cluster are maintained throughout migration [103]. Prior to detachment, border cell and polar cell apical surfaces contact the germline, lateral surfaces adhere to each other, and basal surfaces adhere to the basement membrane that surrounds the egg chamber. As they detach, protrusions extend from lateral surfaces, and the cells retain their shared apicobasal polarization [103, 104, 105]. Additionally, as the cluster pulls away from the basement membrane and anterior follicular cells, it undergoes a 90° rotation such that its apical surface is oriented approximately perpendicular to the direction of migration. Appropriate levels of apical proteins, including Crumbs (Crb) and atypical protein kinase C (aPKC), are essential for maintaining polarity and cohesion. Disruption of apical proteins such as PAR3 or PAR6 causes fragmentation of the cluster [104, 106].

7.3. Migration Termination: Border Cell Cluster Attachment to the Oocyte Surface

This migration is completed by stage 10, at which point the border cell cluster reaches the oocyte (Figure 2c, termination). Upon reaching the oocyte, the cluster rotates another 90°, enabling the apical surface of the border cells to dock onto the oocyte membrane. Typically, one border cell makes initial lateral contact, after which each border cell establishes apical contact with the oocyte to form a stable interface [96].

The mechanisms that control border cell–oocyte interactions are largely unknown. My previous work demonstrated that Innexins, gap junction‐forming proteins [107, 108, 109] are required for stable border cell–oocyte interactions. Colleagues and I revealed the channel‐independent function of Innexins, which regulate microtubules to maintain the stability of the border cell–oocyte interface. The loss of Innexin2 (Inx2), Inx3, or Inx4 disrupts this interface, leading to dislocation of the border cell cluster and impaired interaction with cfcs, ultimately hindering neolamination [96]. Identifying additional regulators of neolamination and delineating how microtubules control the dynamic border cell behaviors will further clarify how collective cells terminate migration.

In summary, border cells provide a compact, genetically tractable model for studying the initiation and termination of collective cell migration independently of EMT. After being specified via JAK‐STAT signaling, border cells delaminate from the anterior of egg chamber with preserved epithelial polarity. They migrate in‐between the nurse cells and stop the migration via docking to oocyte surface. These features align with the collective start‐stop framework and enable cross‐system comparisons.

8. Cancer Metastasis

Metastasis—the spread of cancer cells from a primary tumor to distant organs—is a leading cause of cancer‐related mortality. To metastasize, tumor cells must (i) detach from the primary mass and invade adjacent tissue, (ii) intravasate into blood or lymphatic vessels, (iii) survive transport and immune attack in the circulation, (iv) extravasate into a distant tissue, and (v) colonize a new site to form a secondary tumor [8]. Although these steps are often compared to normal developmental migration, metastasis is substantially more complex because of the phenotypic heterogeneity within tumors. Phenotypic heterogeneity refers to the coexistence of multiple distinct subpopulations of cancer cells within a single tumor, each differing in morphology, gene expression, metabolism, proliferation rates, and invasive behaviors [8]. This heterogeneity results from both genetic variation and nongenetic factors such as local microenvironmental cues, epigenetic changes, and stochastic cell‐state transitions. The presence of phenotypically diverse cells complicates the metastatic process, as different subpopulations may contribute to different steps—some acting as leaders that guide invasion, others providing proliferative capacity, and still others remaining relatively quiescent until colonization at secondary sites [8].

8.1. Collective Invasion and Termination of Migration

While the classical paradigm emphasizes single‐cell dissemination, mounting evidence shows that collective invasion, that is, the migration of cohesive multicellular groups that retain cell–cell junctions, is common in many types of carcinoma, including breast, prostate, head and neck, and colorectal cancer [7, 12]. In collective invasion, E‐cadherin–mediated adhesions are frequently maintained, enabling clusters that range from a few cells to large sheets to move through the stroma as mechanically coupled units [7, 12, 110].

The mechanisms that initiate and terminate metastatic migration remain incompletely defined. The EMT/MET framework offers a useful perspective [10, 36, 49, 111, 112, 113, 114, 115]. In many tumors, collective invasion initiates when a subset of leader cells at the invasive front acquires partial EMT traits—gaining motility and front–rear polarity while retaining sufficient epithelial adhesion to keep the group intact. Leader cells typically upregulate motility‐associated transcription factors, increase the expression of ECM‐remodeling enzymes such as MMP2 and MMP9 to forge paths through the stroma, and display a strongly polarized cytoskeleton with actin‐rich lamellipodia and filopodia oriented toward chemo‐ and haptotactic cues. Follower cells retain stronger epithelial features and generally higher E‐cadherin expression to preserve cohesion and coupling with leaders via adherens junctions to transmit forces and guidance cues. They contribute to proliferative expansion, confer structural integrity to migrating clusters, and enhance survival during dissemination. In distant organs, successful colonization often correlates with MET or partial MET, including the reactivation of epithelial programs, strengthening of junctions, and restoration of apical–basal polarity to support proliferative outgrowth. Notably, the expression levels of key molecules vary across cancers, including the absolute levels and spatial distribution of E‐cadherin and the expression of EMT‐regulating transcription factors (e.g., Snail, Slug, Twist, and ZEB1/2); consequently, the extent of EMT at the initiation of migration and the completeness of MET at the termination of migration are context dependent. Leader–follower roles can also be exchanged dynamically during invasion, although how these identities are maintained—or how they change—during colonization remains less well defined.

Multiple classes of cues shape both the initiation and termination of collective metastasis. Initiation cues include microenvironmental growth factors such as TGF‐β, FGF, and EGF that induce partial EMT programs and specify leader cell behaviors [116, 117, 118]; mechanical inputs, including matrix stiffening and interstitial flow, which polarize the cytoskeleton and protrusions at the invasive front to drive directed movement [119, 120, 121]; paracrine stromal loops in which cancer‐associated fibroblasts and tumor‐associated macrophages supply chemokines (e.g., Cxcl12 and CCL‐2) that bias leader activation and sustain invasion [122, 123]; and hypoxia, which stabilizes HIF‐1α to increase the expression of EMT‐associated transcription factors and proteases and ultimately increase invasion [124]. Conversely, colonization cues include niche‐driven epithelialization—arrival in a microenvironment whose ECM composition and stiffness favor epithelial organization, thereby promoting MET; loss of motility signals through dissipation of chemotactic gradients or depletion of motility‐inducing growth factors, which reduces protrusive activity and halts migration; adhesive crosstalk, in which productive contacts with resident epithelial or stromal cells provide adhesion cues and growth signals that stabilize clusters and foster outgrowth; and transcriptional/epigenetic reset, in which repression of Snail/Slug/ZEB programs and induction of epithelial regulators such as GRHL2 and OVOL2 re‐establish junctions and polarity, reinforcing MET [125, 126].

Currently available evidence suggests that the balance between EMT and MET is dynamic rather than binary—metastatic cells can switch states multiple times in response to evolving environmental cues [36]. Importantly, MET at secondary sites is not always complete; many micrometastases retain hybrid epithelial/mesenchymal traits that can facilitate further dissemination or confer resistance to therapy. Finally, leader identity can be stable (a distinct subpopulation) or plastic (followers assume leadership when leaders are lost), reflecting the phenotypic diversity of tumors. This division of labor allows collective groups of cells to combine the invasive potency of leader programs with the survival and growth advantages of a cohesive epithelium.

9. Conclusion

9.1. A Start‐Stop Framework of Collective Cell Migration

Collective migration is distinguished from single‐cell migration by supracellular coordination. Cells undergoing collective migration move as mechanically and chemically coupled cohorts that preserve junctional continuity while redistributing forces and guidance cues across the group. Across systems reviewed here, I propose three principles: (i) Prepatterned leader–follower roles: Before migration initiates, leader–follower identities are typically specified by patterning cues (e.g., growth factors, chemokines, lateral inhibition circuits, and tissue mechanics); however, in some cases, such as in metastasis, these roles may remain plastic, allowing followers to assume leadership if conditions change. (ii) Maintenance of cohesion: Adhesions (typically mediated by intermediate cadherins) are maintained or adjusted, which permits traction without loss of cohesion. (iii) Distributed polarity and force transmission: Leader and follower cells establish front–rear and apical‒basal polarity; cells relay forces through cell–cell junctions and the actomyosin network to move as a unit. These principles support the control of collective cell migration initiation and termination across contexts.

9.2. Open Questions

Although researchers have expended much effort to understand how collective cells initiate and terminate migration, much remains to be learned.

How does collective migration terminate? Compared with initiation, termination processes are more diverse and context dependent. The molecular “off‐switches” that terminate cell motility remain to be delineated across tissues. For example, in cancer metastasis, the combinations of niche ligands, mechanical thresholds, and metabolic states that commit a cohort to re‐epithelialization, fusion, or dormancy are largely unknown.

How are leader and follower specified and maintained/changed? What transcriptional, mechanical, and metabolic circuits specify—or eliminate—leader identity, and on what timescales can roles be changed in vivo? The regulation of these processes is still largely unknown.

How do individual cells in the group undergoing collective migration communicate with each other? Cells can communicate with each other via gap junctions, tunneling nanotubes, exosomes, and mechanically transmitted signals. What are the relative contributions of these communications to collective decision‐making? Does blocking specific channels disrupt cohesion, guidance, or colonization differently in the context of development versus that of cancer?

Addressing these questions will move the field from descriptive models toward predictive, actionable control of collective migration; for example, in cancer, it may be possible to prevent the initiation of collective migration (e.g., via strategies aiming to block leader induction or the stroma paracrine loop) or block the termination of migration that is required for colonization.

Author Contributions

G.M. conceived the article, reviewed literature, developed the conceptual framework, prepared the figures, and wrote and revised the manuscript.

Conflicts of Interest

The author declares no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.