Perspectives in collective cell migration – moving forward Free

Collective cell migration is a widespread phenomenon, playing a crucial role in diverse processes such as the formation of early embryonic tissues (Ewald et al., 2008; Rupp and Kulesa, 2007; Solnica-Krezel, 2005; Sutherland et al., 1996), the healing of epithelial wounds (Farooqui and Fenteany, 2005; Poujade et al., 2007) and the relentless march of metastatic cancer cells (Aceto et al., 2014; Astin et al., 2010; Friedl et al., 1995; Hegerfeldt et al., 2002; Nabeshima et al., 2000). This type of migration is a cooperative journey where cells move as a cohesive unit, maintaining contact and influencing each other as they forge a path through new territories (Rorth, 2009). Maintaining cell–cell interactions during migration may facilitate the efficient coordination of cell dispersal to the appropriate locations whilst preserving tissue integrity during remodelling.

Collectively migrating cells can move together as sheets, such as epithelial monolayers (Poujade et al., 2007), or as loosely connected streams, such as neural crest cells or invasive cancer cells (Carmona-Fontaine et al., 2008; Nabeshima et al., 2000; Rupp and Kulesa, 2007). They can also migrate as small groups of cells, as is seen with Drosophila border cells and metastatic cancer cell clusters (Friedl et al., 1995; Montell et al., 1992), or they can form complex three-dimensional (3D) structures, including blood vessels, tubes, glands and ducts, through sprouting or branching morphogenesis (Ewald et al., 2008; Gerhardt et al., 2003; Ghabrial and Krasnow, 2006; Larsen et al., 2006; Sainson et al., 2005). Which mode of collective cell migration cells use is determined by the morphology and strength of their cell–cell interactions and their environmental context. Collectively migrating epithelial cells maintain stable adherens junctions (AJs), mediated by intercellular adhesion of cadherins, that mechanically couple the cells into a single collective. Mesenchymal cells, however, only maintain transient AJs and migrate more independently within loose streams or chains of cells. Nevertheless, the formation of transient cell junctions is required to maintain the directionality of movement during collective mesenchymal migration (Arboleda-Estudillo et al., 2010).

The stereotypical migration of a single cell is initiated by front–rear polarisation whereby the cell establishes a leading edge and a trailing edge, allowing it to orient its movement in response to various stimuli (Lauffenburger and Horwitz, 1996; Ridley et al., 2003). This is followed by the extension of protrusions at the leading edge as flat, sheet-like lamellipodia and/or as slender, rod-like filopodia, driven by the polymerisation of actin filaments. These protrusions adhere to the substrate or extracellular matrix (ECM) via integrins, which scaffold onto multimolecular protein complexes called focal adhesions (FAs), anchoring the cell and allowing it to generate traction. At the rear of the cell, force generated by myosin motor proteins sliding actin filaments relative to each other (actomyosin contraction) pulls the trailing edge forward. The coordinated turnover of the front and rear matrix adhesion complexes allows the contracting rear of the cell to move forward while the leading edge progresses (Lauffenburger and Horwitz, 1996).

Similar principles apply during collective migration; however, the continued coupling of collectively migrating cells through AJs enables a larger-scale supracellular polarisation that establishes front–rear asymmetry within the group of cells (Fig. 1). Two morphologically and functionally distinct cell types are created: leader cells and follower cells. Leader cells exhibit asymmetric adherent interactions, forming FAs at their advancing fronts and cadherin-based AJs at their cell–cell interfaces. This asymmetry of cell–cell coupling is sufficient to induce front–rear polarisation of the cytoskeleton and cytoplasm of the leader cells (Dumortier et al., 2012; Dupin et al., 2009), while the follower cells that trail behind the leader cells experience more symmetric cell–cell coupling through AJs with their surrounding cells. Using a range of model systems (see Box 1), researchers have begun to unravel the complex mechanisms of collective migration, shedding light on the vital roles of cell–cell interactions, mechanical forces, and the self-generation and sensing of chemokine gradients by cells.

To attain a more complete understanding of the dynamic cell behaviours underlying collective migration, the next steps involve achieving a better understanding of the interaction between migrating cells and their tissue surroundings, quantitative analysis of cell behaviours, and measurements of the mechanical properties of cells moving in complex environments. We met to discuss the challenges awaiting this field at The Company of Biologists Workshop ‘Collective Cell Migration: From In Vitro to In Vivo’, which took place in 2023. Following the main themes of the Workshop, this Perspective article discusses our understanding of collective cell migration in development, tissue morphogenesis and cancer metastasis, as well as emerging mechanisms that regulate this process. We discuss model organisms used to answer different questions surrounding collective cell migration and highlight how mathematical and biophysical modelling can help us to better understand more physical and mechanical aspects. Finally, we discuss the remaining challenges in the field and emphasise the need for interdisciplinary collaborations to address them.

Early studies of collective cell migration focused largely on the interactions between migrating cells, the cellular machinery they use for migration and external chemical signals such as morphogen gradients (Aman and Piotrowski, 2008; Carmona-Fontaine et al., 2008; Ewald et al., 2008; Farooqui and Fenteany, 2005; Ghabrial and Krasnow, 2006; Montell, 1999; Nabeshima et al., 2000). More recent work is revealing that the extracellular environment is not merely a substrate on which cells migrate but also provides physical cues essential for efficient and directional collective cell migration. Spatial cues such as confinement, geometric constraints and tissue topography direct collective cell movement by physically restricting the space available for effective migration (Dai et al., 2020; Szabo et al., 2016). Mechanical cues such as tissue stiffness can control the timing and directionality of collectively migrating cells (Barriga et al., 2018; Shellard and Mayor, 2021). The topological and rheological properties of variable ECM surrounding collectively migrating cells provide directional and spatiotemporal cues (Elosegui-Artola et al., 2023; Sapudom et al., 2015; Sarker et al., 2019). Together, the interplay of physical cues in the tissue environment helps to guide migrating cell collectives towards their destination.

The collective migration of cells underpins many morphogenetic processes during embryonic development. At the cellular level, the contractile forces generated by actomyosin are transmitted through adhesion complexes, facilitating their propagation to neighbouring cells or the ECM. This enables the coordinated movement of cells and the translation of individual cell-generated forces into global changes in tissue shape. An excellent example of this is the collective migration of the border cells in the egg chamber of the Drosophila embryo. The border cells are a small cluster of cells that delaminate from the anterior of the egg chamber, extend actin-rich protrusions and migrate through a network of large nurse cells to reach the oocyte (Fig. 2A). Early research investigated the mechanisms that specify border cell identity and how they regulate the cell–cell adhesion, polarity and actomyosin dynamics of border cells (Montell et al., 2012). However, the critical role of physical constraints from surrounding tissues is now coming to light. Successful migration of border cells depends on withstanding compressive forces from the neighbouring nurse cells (Aranjuez et al., 2016). In response to these forces, the border cells increase actomyosin contractility at the periphery of the cluster to regulate supracellular shape and migration efficiency. The compressive forces exerted by the nurse cells may in turn be regulated by the stiffness of the basement membrane that encapsulates the egg chamber (Molina López et al., 2023). Changing the stiffness of the basement membrane by altering its composition affects the cortical tension of the nurse cells, and the protrusion dynamics and migration speed of the border cells.

Furthermore, the central path the border cells migrate along through the egg chamber is determined by the topography and packing of the surrounding nurse cells (Alsous et al., 2018; Dai et al., 2020). The geometry of the egg chamber means that junctures between three or more nurse cells are concentrated towards the centre, leaving larger junctures than in the two-cell interfaces found at the egg chamber periphery. This creates an energetically more favourable medial path where the border cells can extend protrusions into the pre-existing space without having to break as many adhesion bonds between nurse cells. Taken together, these studies show how the physical properties of the cellular environment direct the collective migration of the border cells, ensuring their arrival at the oocyte.

The physical properties of the surrounding tissue also direct the collective migration of neural crest cells during Xenopus morphogenesis. Pre-migratory cephalic neural crest cells reside within the dorsal neural tube and only initiate their collective migration away from the neural tube when the underlying mesodermal tissue stiffens (Barriga et al., 2018). Culturing these cells on stiff substrates increases the expression of the mesenchymal marker N-cadherin (also known as cadherin-2), suggesting that stiffening of the mesoderm can promote a neural crest epithelial-to-mesenchymal transition (EMT). This mesodermal stiffening therefore acts as a morphogenetic timer, coordinating changes in tissue mechanics due to cellular movements in gastrulation with the onset of neural crest migration. In response to mesodermal stiffening, the neural crest cells dynamically decrease their stiffness, potentially maintaining an optimal ratio of cell:substrate stiffness for collective migration (Marchant et al., 2022). Once initiated, the direction of neural crest collective cell migration is determined by a combination of chemical and physical cues (Shellard and Mayor, 2021). The cephalic neural crest cells migrate in well-defined streams due to geometric constraints from the surrounding tissues that are non-permissive to their migration (Szabo et al., 2016). Stromal cell-derived factor-1 (SDF1), a well-established chemoattractant, is expressed in the cranial placodes, which causes neural crest cells to ‘chase’ them (Theveneau et al., 2013). As the neural crest cells reach the placodes, the placodal cells ‘run’ using an N-cadherin-dependent repulsion mechanism. However, this chemical signal must be reinforced by a synergistic mechanical signal for effective migration (Shellard and Mayor, 2021): the neural crest cells interacting with the placodes soften the placodal tissue using an N-cadherin-dependent mechanism. This generates a stiffness gradient in the placode, which moves with the neural crest such that the collectively migrating cells pursue a receding region of higher stiffness. In this way, mechanical changes in the surrounding tissue environment control both the timing and directionality of neural crest collective migration.

We have gained some understanding of the cellular aspects of collective cell migration, such as the importance of cell–cell interactions in cell polarisation (Mayor and Etienne-Manneville, 2016) and the interplay between cell–cell and cell–substrate adhesions that generate mechanical forces within a group of cells on a two-dimensional (2D) substrate (Ladoux et al., 2016). However, fully comprehending how cells mechanically interact with their surroundings in intricate 3D environments or in vivo, and how these interactions influence migratory cell behaviours, remains a significant challenge.

Progress has been made in understanding some of the functional interactions between cancer cells and cancer-associated fibroblasts (CAFs), a heterogeneous group of activated fibroblasts that interact with cancer cells. CAFs can interact with cancer cells directly through secreted molecules or cell adhesions, and/or indirectly by remodelling the ECM and encapsulating tumours in vivo (Sahai et al., 2020). Combining in vivo genetic and physical manipulation with traction force microscopy to measure compressive forces exerted by CAFs in vitro has revealed how CAFs compress intestinal cancer cells in an actomyosin-dependent manner, effectively restricting cancer cell proliferation (Barbazan et al., 2023). Force-dependent feedback is also important for tumour cells that have been shed from a primary tumour and are circulating in the bloodstream (circulating tumour cells, CTCs) to escape the blood vessel and seed new metastatic tumours. In the zebrafish larva, CTCs can stably adhere to endothelial cells in vascular regions with lower shear forces, as low flow favours deposition of fibronectin on the lumen of vessels, creating conditions permissive for CTC–endothelial cell interactions (Osmani et al., 2019). These reports highlight how mechanical feedback mechanisms from the surrounding in vivo environment are essential to allow or inhibit tissue invasion. Nevertheless, it remains challenging to measure mechanical quantities in deep tissue in vivo and to visualise cell–matrix interactions.

The ability of cells to transition between epithelial and mesenchymal states is essential for the migration and organisation of cells into tissues during morphogenesis, but it also enables the metastatic spread of cancer cells. EMT was initially considered a binary process in which a mature apico-basally polarised and tightly adherent epithelial cell fully transforms into a loosely connected and migratory mesenchymal state. However, more recent work has demonstrated that EMT occurs in varying degrees, leading to a range of intermediate cell types with both epithelial and mesenchymal traits (Campbell and Casanova, 2016; Haerinck et al., 2023). The term epithelial–mesenchymal plasticity has been proposed to describe cells that have mixed epithelial and mesenchymal features, and the ability to move between hybrid states along the epithelial–mesenchymal spectrum (Yang et al., 2020). Single-cell RNA-sequencing data suggest a continuum with innumerable partial EMT states that are highly context-specific and possess few shared changes in gene expression (Cook and Vanderhyden, 2020).

Mesenchymal-to-epithelial transition (MET) is often considered the inverse of EMT, whereby a cell re-establishes apico-basal polarity and increases cell–cell adhesion to form an organised epithelium. During development, multiple epithelial tissues are formed by MET, including primitive endoderm, somites, the epicardium and the renal epithelium (Pei et al., 2019). Although critical for embryogenesis, MET is also proposed to drive the colonisation of secondary sites by cancer cells that transition towards an epithelial state to form a new tumour. However, a growing body of work suggests that MET is not merely the restoration of mechanisms that were dismantled during EMT, but instead results in distinct epithelial states. For example, MET of the Drosophila endoderm to form the midgut requires downregulation of EMT-promoting transcription factors (Campbell et al., 2011), but this alone is not sufficient. Secretion of laminin from the surrounding mesoderm is also necessary to initiate repolarisation for MET (Pitsidianaki et al., 2021). Furthermore, the repolarisation process does not rely on re-expression of the key factors that determine polarity in the embryonic ectoderm from which the tissue is derived. Instead, an alternative mechanism based on the integrin adhesion complex is employed (Chen et al., 2018), implying that the cells have acquired an epithelial state that differs from their previous state.

During EMT, cell polarity changes from apico-basal to front–rear. This loss of apico-basal polarity is a key event associated with the disruption of cell and tissue architecture in epithelial cancers (Peglion and Etienne-Manneville, 2024). Furthermore, the loss of apico-basal polarity itself can be an initiating event in EMT. Downregulation of the apical polarity protein Crumbs drives a loss of apico-basal polarity and fragmentation of AJs, which triggers a partial EMT and collective cell migration during Drosophila endoderm formation (Campbell et al., 2011). Depletion of the basolateral-identity protein septin 9, or mislocalisation of the apical-identity protein podocalyxin, results in the acquisition of mesenchymal features in 3D cell culture models, increasing collective invasion into the surrounding stroma (Cai et al., 2023; Roman-Fernandez et al., 2023).

Repression of the epithelial marker E-cadherin (also known as cadherin-1 in vertebrates, DE-cadherin in Drosophila) is often considered a key step in EMT (Huber et al., 2005; Onder et al., 2008); however, during embryo development, cells become migratory while still preserving E-cadherin-dependent intercellular adhesion, as seen in many collectively migrating cell populations. The migration of both the Drosophila posterior midgut and the border cell cluster in the egg chamber requires the presence of E-cadherin to maintain migratory cell clusters. When E-cadherin is specifically depleted in border cells, it leads to cluster disassembly (Cai et al., 2014). Similarly, disruption of E-cadherin in migrating posterior midguts causes the detachment of mesenchymal cells from one another (Campbell and Casanova, 2015). E-cadherin-dependent intercellular adhesion is also required for the collective migration of zebrafish mesendoderm and Xenopus cranial neural crest (Dumortier et al., 2012; Huang et al., 2016). Maintaining E-cadherin in these cells might not only enable their collective migration but could also confer protection against apoptosis as a result of cellular detachment from the ECM (anoikis) (Bergin et al., 2000; Fouquet et al., 2004; Hatzold et al., 2021).

In cancer, EMT is often partial, resulting in tumour cells with characteristics of both epithelial and mesenchymal cells (Shibue and Weinberg, 2017; Simeonov et al., 2021). Such cellular phenotypic plasticity may enhance the capacity of a cell to respond to changing microenvironments, and a mixed population of tumour cells in different hybrid states could possess the highest capacity to negotiate varying microenvironments during metastasis. Indeed, cell populations undergoing a partial EMT have a higher metastatic potential than solely epithelial or mesenchymal cells (Pastushenko et al., 2018), or mixed populations of cells in a stable epithelial or mesenchymal state (Kroger et al., 2019), reaffirming the clinical importance of incomplete transitions between these two states. Interestingly, E-cadherin expression can even be higher in metastatic sites than in the primary tumour in breast cancer patients (Kowalski et al., 2003). Moreover, although E-cadherin knockout promotes cell invasion in 3D organoid assays, it also reduces proliferation and cell survival in an in vivo mammary cancer mouse model (Padmanaban et al., 2019). This is surprising given the current dogma of E-cadherin as a repressor of invasive capabilities and suggests that E-cadherin expression controls a delicate balance between pro-migratory and pro-survival signalling that is essential for metastatic spread. In orthotopic xenograft mouse models, metastatic cells lacking E-cadherin show increased levels of reactive oxygen species in the mitochondria, as E-cadherin is important for upregulation of the de novo serine synthesis pathway in breast cancer cells (Lee et al., 2023 preprint). This provides metabolic precursors for glycolysis and the tricarboxylic acid (TCA) cycle, as well as resistance to oxidative stress, making it critical for E-cadherin-positive breast cancer cells to successfully metastasise. These recent studies reveal previously unappreciated roles for E-cadherin beyond adhesion during EMT.

More broadly, the growing understanding of partial EMT suggests new avenues for therapeutic intervention in cancer metastasis. Traditional therapies that target EMT can inhibit the migration of cells away from the primary tumour but may inadvertently promote secondary tumour formation. A promising new strategy is to target MET and drive the transdifferentiation of mesenchymal tumour cells into a different, harmless cell fate (Ishay-Ronen et al., 2019). Alternatively, targeting cells that have undergone a partial EMT and fixing their position along the EMT–MET spectrum might prevent the plasticity they require to respond to changing microenvironments and successfully metastasise.

As outlined above, collective migration modes are thought to depend on cell adhesion to generate traction forces. Within motile cell clusters, leader cell behaviour is reinforced by mechanical and chemical cues from the microenvironment, with retraction of the trailing edge driven by contractility in the follower cells (Shellard et al., 2018). Clusters of tumour cells, either cell lines or cells derived directly from patient abdominal fluid (ascites) and xenografts, can migrate independently of adhesion via a recently described ‘collective amoeboid’ mode (Pages et al., 2022). During single-cell amoeboid migration, cell movement is powered by persistent and coordinated flows of actomyosin to the rear of the cell that generate active friction against the substrate to propel the cell forward (Liu et al., 2015; O'Neill et al., 2018). However, cell clusters do not show persistent rearward actomyosin flow, either at the single-cell level or coordinated across the cluster (Pages et al., 2022). Instead, cluster movement is propelled by uneven friction forces that result from cell shape changes driven by varying fluctuations in actomyosin flow. This process is referred to as ‘polarised jiggling’ (Pages et al., 2022). This new mode of motility gives insight into metastatic spread in tissues where matrix-independent environments exist, such as malignant ascites in the peritoneal cavity, or where cells do not express appropriate adhesion receptors.

As cancer cells metastasise, they move through environments that lack ECM, such as lymphatic and haematogenous routes. There is evidence that metastatic potential is increased by tumour cells circulating as clusters, rather than as individual cells (Fig. 2B) (Aceto et al., 2014; Cheung and Ewald, 2016; Luo et al., 2014). Our understanding of the sequence of events preceding the formation of circulating clusters and following their arrest in the vasculature of distal tissues is lacking (Giuliano et al., 2018). It is currently not known how clusters are formed and enter the vasculature, although hypoxia (Donato et al., 2020) and cell jamming are implicated (Haeger et al., 2014). Cell jamming occurs when cells become densely packed and lose the ability to move freely, leading to a solid-like tissue state (Park et al., 2015). Unjamming of epithelial tumours, a tissue-level mechanical event where cells transition from a solid-like state to a fluid-like state, triggers collective cell migration (Malinverno et al., 2017). In models of breast cancer, transition to the fluid-like state results in nuclear envelope ruptures that induce a cytosolic DNA response, as well as activation of cGAS–STING pro-inflammatory signalling to drive a metastatic cell state (Frittoli et al., 2023). This could drive type I interferon signalling in CTC clusters (Gkountela et al., 2019) to recruit neutrophils to clusters, accelerating metastatic seeding (Szczerba et al., 2019). The mechanisms underlying how tumour cell clusters arrest and exit the vasculature are poorly understood, although the ability to maintain a hybrid EMT state is thought to be important for seeding success (Simeonov et al., 2021). Future live-cell imaging studies will be necessary to discern how collective cell migration mechanisms contribute to these steps of metastasis. The imaging compatibility of model systems such as zebrafish and Drosophila, and the shared pathophysiology of these models and human cancers, will be key to progressing our understanding of these relatively rare events (Campbell et al., 2019; Peralta et al., 2022).

When migrating collectively in vivo, cells must sometimes push through their surroundings to reach their destination. This can be an exhausting, energy-consuming task for leader cells at the forefront of a migrating collective, and it is becoming clear that cells adjust their metabolic responses to cope with the demand. Insights from single-cell migration during developmental morphogenesis are revealing how the heightened energy requirements associated with invasion can be sensed and responded to at the level of the cellular collective in both developmental morphogenetic processes and in cancer cell invasion. For example, when Drosophila macrophages invade the embryonic epithelium, they boost their mitochondrial oxidative phosphorylation activity by upregulating key metabolic enzymes and enhancing translation of mitochondrial proteins (Emtenani et al., 2022). The resulting increase in energy production enables the leading macrophages to overcome the surrounding tissue resistance and forge a path for infiltration. Energy adaptation is coordinated between collectively migrating breast cancer cells as they invade through the ECM in in vitro spheroid and ex vivo organoid invasion models (Zhang et al., 2019). Leader cells creating a path for invasion require a higher ATP:ADP ratio and take in more glucose than follower cells. However, once a leader cell has depleted the energy available to it, it is usually replaced by a follower cell that has a higher ATP:ADP ratio, which resumes the invasion (Vilchez Mercedes et al., 2021; Zhang et al., 2019). Enhanced ATP production extends the life of leader cells, whereas glucose deprivation shortens it. When invasion is more energy demanding, as in denser collagen environments, the frequency of leader–follower transitions increases.

At the subcellular level, mitochondria are repositioned to the leading edge of invading cells to supply energy for the cytoskeletal remodelling required to advance. In Caenorhabditis elegans, a specialised anchor cell uses invadopodia to break down and penetrate the basement membrane during development of the reproductive system (Hagedorn et al., 2013). To meet the energetic demands of this process, mitochondria and glucose transporters are trafficked to the invading forefront of the cell to expedite ATP production. This coordinated effort of energy import, processing and generation is intricately connected with the mechanisms that drive anchor cell invasion (Garde et al., 2022; Kelley et al., 2019). The dynamic adaptation of cellular metabolic activity to meet tissue-specific energy requirements increasingly appears to be a widespread phenomenon during developmental morphogenesis (Cao et al., 2023 preprint; Dingare et al., 2023 preprint). Invading cancer cells also exhibit subcellular spatiotemporal organisation of energy production. 5′ adenosine monophosphate-activated protein kinase (AMPK)-dependent mitochondrial trafficking to the leading edge of ovarian cancer cells regulates their ability to migrate and metastasise (Cunniff et al., 2016). Recent reports identify AMPK as a ‘mechano-metabolic’ sensor in melanoma invasion, coupling matrix-influenced cytoskeletal dynamics to energy production (Crosas-Molist et al., 2023). Similarly, in collectively migrating lung cancer cells, leader cells exhibit a more peripheral distribution of mitochondria, whereas the mitochondria in follower cells typically stay near the cell nucleus (Commander et al., 2020). In models of confined migration, narrower tracks elicit a higher use of energy than wider compartments, directly linking migration decisions to the physical parameters of the cell surroundings (Zanotelli et al., 2019). Therefore, interfering with the ability of cancer cells to regulate energy production in response to environmental demands (for example, through pharmacological inhibition of AMPK) might be an important avenue for delaying or preventing metastatic invasion.

As it is not always possible to experimentally perturb a system, close interdisciplinary collaboration between biologists and theorists using in silico modelling has become vital to achieve a quantitative understanding of collective migration phenomena. Collectively migrating cells often exhibit emergent properties that are difficult to predict. Combining mechanistic modelling and simulations with observed cell biology enables us to test hypotheses beyond what is achievable using pure experimental wet-lab approaches. The goal of mathematical modelling is to replicate cellular behaviours observed during wet-lab experiments. The models are then utilised to decipher underlying mechanisms and to pinpoint which components involved in cell migration are crucial and how they work together to define the response of both individual cells and groups of cells. This type of combined approach has recently demonstrated how dynamic deposition and remodelling of the ECM promotes robust collective migration of neural crest cells in the chick embryo (Martinson et al., 2023), and how the environment of migrating Drosophila border cells influences the optimal cluster size by impacting on migration speed (Cai et al., 2016).

A variety of different modelling frameworks have been applied to collective cell migration, ranging from individual-based models at the cellular level to higher-level partial differential equation models, enabling analysis of the system from different perspectives (Alert and Trepat, 2020; Schumacher et al., 2016; Te Boekhorst et al., 2016). Modelling the properties or behaviour of small groups of interacting cells allows the prediction of emergent multicellular behaviours. Observing the behaviours of pairs of cells in an experimental ‘cell collider’, where pairs of migrating cells are confined within a micropattern of two adhesive squares separated by a thin bridge, has allowed an interacting equation of motion to be inferred (Bruckner et al., 2021). This has revealed that non-cancerous cells exhibit repulsion and effective friction, whereas cancerous cells exhibit attraction and an unanticipated anti-friction interaction that could potentially play a role in tissue fluidisation during tumour invasion (Kang et al., 2021; Palamidessi et al., 2019).

At the single-cell level, models describing one-dimensional migration of cells along annular or linear tracks have demonstrated how polarisation of individual cells can coordinate migration of the collective (Jain et al., 2020; Ron et al., 2023). By incorporating the effects of contact inhibition of locomotion and cryptic lamellipodia on intracellular actin flow and polymerisation, these models can mimic and shed light on cellular behaviours observed during morphogenesis in vivo.

At the other end of the spectrum is a more ‘top-down’ approach of observing tissue-scale behaviours and inferring the underlying rules, interactions and feedback loops governing that behaviour. Concepts developed in the field of soft-matter physics are now being applied to understand morphogenesis and collective cell migration. Considering cells and tissues as active nematic systems in which cells exhibit self-propulsion and align with each other, resulting in fluid-like behaviour with emergent collective motion and non-equilibrium dynamics, provides insight into how cellular properties and the physical geometry of the cellular microenvironment controls the coordination of multicellular movements. As the density of spindle-shaped cells increases in a colony of mouse fibroblasts cultured in 2D monolayers, the cells become aligned into a high-density nematic state with regions of common orientation (Fig. 3A) (Duclos et al., 2014). These domains of nematically ordered cells are separated by topological defects that become trapped in the tissue structure and behave as self-propelled particles (Fig. 3B,C) (Duclos et al., 2017). Experimentation combined with in silico modelling has demonstrated that cell movement within the collective is determined by the interplay between cell shape and density, the strength of cell–cell interactions, and environmental cues such as confinement geometry and ECM deposition (Balasubramaniam et al., 2021; Duclos et al., 2017; Saw et al., 2018). Increases in cell density and maturation of cell–cell and cell–substrate adhesion can lead to slowing of the collective cell movement over time, until the system transitions from an unjammed fluid-like state to a jammed solid-like state (Angelini et al., 2011; Bi et al., 2015; Garcia et al., 2015; Sadati et al., 2013). These transitions in tissue fluidity have been proposed to underlie diverse processes in embryo morphogenesis and cancer invasion (Ilina et al., 2020; Malinverno et al., 2017; Mongera et al., 2018; Palamidessi et al., 2019; Petridou et al., 2019).

Within a monolayer, cells concentrate and accumulate at +1/2 topological defects (Fig. 3C), where the nematic alignment is locally lost, leading to high local compressive stress that can drive cell extrusion and apoptosis (Saw et al., 2017). If a layer of ECM covers the monolayer, cells at the head of the defect can collectively migrate over the cells at the tail of the defect, which migrate in the opposite direction (Sarkar et al., 2023). This marks the beginning of a transformation of the tissue from a simple 2D layer into a 3D criss-crossed bilayer (Fig. 3D). The ECM itself may also be organised into a nematically aligned structure that can shape tissues and provide ‘highways’ for metastatic cell migration (Chevalier et al., 2021; Kim et al., 2021). Studying the nematic ordering of biological systems has provided a new framework to explain how cells collectively migrate in fluid-like packs, streams and swirls, or switch to more quiescent solid-like states, enabling forces at the cellular level to shape and organise tissues.

A primary challenge in studying collective cell migration is the automated segmentation of cells imaged in three dimensions within complex tissues over time. These cells are often heterogenous in shape and size, making it challenging for algorithms to accurately identify and delineate cell boundaries. The segmentation process is also often further complicated by the cells in 3D tissues being closely packed, with high nuclear:cytoplasmic ratios, and engaging in frequent neighbour exchanges (Gómez et al., 2021). Being able to track cell migration, membrane and nuclear shape, and subcellular features is crucial for understanding the mechanisms governing collective cell behaviour in 3D tissues. Improvements in computational cell segmentation can be achieved by optimising the acquisition, handling and analysis of images. Development of new live-cell dyes or stains that are tissue permeant with bright, stable fluorescence would enable clearer images while helping to also minimise phototoxicity. New approaches for managing large datasets generated using current technologies would make the task of segmentation less cumbersome. However, a growing need remains for computational algorithms and tools for automated cell segmentation and tracking that are robust and adaptable enough to handle the inherent variability and heterogeneity present in biological samples. Future developments in these tools will greatly facilitate the investigation of collective cell migration.

Another major challenge in the field is the measurement of intra-, inter- and extra-cellular forces in 3D multicellular systems. A comprehensive understanding of the mechanical forces influencing cell migration and how they are modulated at different spatial and temporal scales is crucial for deciphering the principles governing collective cell migration. Although many approaches are now available to measure the mechanical forces impacting cell migration in vitro (Nunes Vicente et al., 2023), it is difficult to recapitulate the spatiotemporal combination of forces that arise in vivo due to changes in tissue topography, geometry and stiffness over time. Fluorescent force sensors have been applied to measure cellular tension distribution in tissues in Xenopus (Hirano et al., 2018; Yamashita et al., 2016), zebrafish (Lagendijk et al., 2017) and mice (Tao et al., 2019). Although it is well established that cells collectively migrating on a 2D substrate exert traction forces via cell–matrix adhesions, visualising these adhesions and measuring traction stresses in 3D environments in vivo remains problematic (Roca-Cusachs et al., 2017). This has been successfully attempted using the collectively migrating zebrafish lateral line primordium, where traction stresses at the posterior end of the cluster are mediated by an integrin–talin–actin–basement membrane coupling (Yamaguchi et al., 2022). However, the lateral line migrates immediately underneath the surface of the embryo, and measuring forces and visualising cell–matrix interactions in deep tissue in vivo remains out of reach in most model systems. Intravital imaging of fluorescent biosensors is beginning to reveal how cells collectively migrate in vivo during developmental processes (Nobis et al., 2017) and cancer metastasis (Murphy et al., 2021b; Vennin et al., 2017). Further advances in imaging resolution, the development of more sensitive and accurate biosensors, and the pairing of different combinations of biosensors to simultaneously study both migrating cells and their tissue environment will provide a more comprehensive understanding of the mechanisms underlying collective cell migration.

Linking microscopic to mesoscopic scales is essential to understand how cellular and subcellular processes contribute to collective cell migration across larger tissue regions. Bridging these scales requires the integration of experimental and theoretical approaches, including the development of multiscale mathematical models and quantitative analysis of experimental data. Theorists in the field of collective cell migration require more experimental data to validate and refine their models. The integration of experimental and theoretical approaches is crucial for understanding the mechanisms underlying collective cell migration and identifying new targets for intervention in pathological conditions such as cancer metastasis. Incorporating cell polarity within theoretical models of collective cell migration remains challenging. Developing models that account for cell polarity in three dimensions will provide a more comprehensive understanding of the processes governing collective cell migration. To develop a unifying biophysical understanding of collective cell migration, closer collaboration between biologists, physicists and mathematicians is required.

We thank The Company of Biologists for funding the 2023 Workshop on Collective Cell Migration: From In Vitro to In Vivo, and all Workshop participants for discussions. We apologise to the colleagues whose work we were unable to cover here due to space limitations.