Roles of the Actin Cytoskeleton and Cell Adhesion in Tissue Morphogenesis

Abstract

The generation of organismal form (i.e., morphogenesis) arises from forces produced at the cellular level. In animal cells, much of this force is produced by the actin cytoskeleton. Here, we review how mechanisms of actin-based force generation are deployed during animal morphogenesis to sculpt organs and organisms. Furthermore, we discuss how cytoskeletal forces are coupled through cell adhesions to propagate across tissues, and cases where cytoskeletal force or adhesion is patterned across a tissue to direct shape changes. Together, our review highlights a conceptual framework to reflect our current understanding of animal morphogenesis and provides perspectives on future opportunities of study.

Introduction:

A fundamental goal of developmental biology is to determine how the cells of the embryo generate the exquisite structures of the adult body. This process, morphogenesis, or “the creation of ordered form”, has intrigued scientists for centuries because it represents the ‘nuts and bolts’ mechanism of how our bodies, and the bodies of other animals, are constructed – how cells move to generate new structures, how embryonic tissues morph into organs. The more we learn about morphogenesis, the closer we come to knowing how our bodies are built, which is of vital importance for human health¹. Congenital malformations resulting from defects in morphogenesis are the leading cause of infant mortality in the United States, and pose a significant risk for children of all ages^2-5.

From its outset, the field of experimental embryology has been a quest to link the movements and shape changes of cells within the embryo to the generation of adult body form. Wilhelm His observed neural tube closure in the chick embryo and theorized that a mechanical process driven by the mitotic divisions and motility of the cells could be responsible for the folding of this tissue⁶. Wilhelm Roux expanded on this idea with his concept of ‘developmental mechanics’, in which he merged cellular descriptions of developmental processes with experimental manipulations to infer causal relationships and identify ‘active’ components of embryonic tissues⁷. The works of His and Roux, and the later thinking of D’arcy Thompson, marked a transition from the view that an external ‘vital force’ sculpted embryonic tissues, to the recognition that morphogenetic processes rely on quantifiable physical forces generated by embryonic cells⁸. While the underlying molecular mechanism was unknowable in the 19^th century, we now know that the actin cytoskeleton, a meshwork of filamentous actin (F-actin) and various accessory proteins, including the molecular motor non-muscle myosin type II (Myo-II), is a key force generating machine that powers cell movements⁹. Furthermore, we know that these mechanisms are evolutionarily conserved and shared by all animals (and likely beyond)^9-11.

In this review, we examine mechanisms by which animal cells change their shape using the actin cytoskeleton, and how cell shape changes are coordinated to restructure tissues. Because tissue morphogenesis results from forces generated from the molecular to tissue scales, it can be broken down into several components, including: (1) the molecular mechanisms of force generation and how they act locally within the cell to produce shape changes, and (2) the physical mechanisms that connect or adhere the cells of a tissue. The nature and patterns of intercellular linkages in a tissue must also be considered, including (a) the mechanical integration of force generating machines and adhesion, and (b) the spatial and temporal organization of the force generating machines and how they connect across a tissue.

Lastly, other cytoskeletal components, such as microtubules and intermediate filaments, also contribute significantly to morphogenetic processes. The interaction between these cytoskeletal systems and the actomyosin cytoskeleton is an exciting and active area of research. However, our main focus here is on morphogenetic mechanisms which utilize the actin cytoskeleton, because this seems to be the predominant mode of force generation in animal morphogenesis⁹. For a detailed overview of microtubule-dependent mechanisms of morphogenesis and intermediate filaments, we recommend other excellent reviews^12-14,16.

Cytoskeletal mechanisms of cell shape change:

Cells utilize a variety of mechanisms to generate the force necessary to change shape, but these transformations can be placed in a simple framework in which cells push or pull in order to expand or contract, respectively. These transformations can be organized in one or two dimensions, such as modifying the shape of a single cell edge/interface or of a cellular surface, or in three dimensions, such as changes to multiple cellular surfaces or a change in volume. Here, we will provide a brief overview of actin cytoskeletal force-generating mechanisms associated with pushing and pulling forces, and then discuss how these mechanisms are deployed in a few ‘case studies’ of tissue remodeling during development.

Molecular mechanisms of pushing and pulling:

Pushing forces are primarily generated by F-actin network polymerization. F-actin is a semi-flexible polymer that forms from the controlled polymerization of monomeric actin subunits. Actin has an intrinsic polarity, such that monomers preferentially add to the growing (plus or barbed) end (Fig. 1A) (For a detailed review on mechanisms of actin polymerization and turnover; refer to^15,17). When filament growth is oriented towards the plasma membrane, such as in the lamellipodium of a migrating cell, actin monomer addition can push the cell edge forward (Fig. 1B)^15,18,19. This generates a modest pushing force at the molecular level, on the order of ~1 pN per polymerizing filament when measured in vitro (Table 1).

Figure 1: Molecular mechanisms of pushing and pulling:

(A) polymerization of F-actin by the preferential addition of actin monomers to the plus end creates a small pushing force (~1.3 pN). (B) In a branched actin network at the leading edge of the cell, the force of actin polymerization is sufficient to push the plasma membrane forward. (C) pulling forces are produced by the action of Myo-II minifilaments (pink) pulling against opposing F-actin filaments to generate tension (~4 pN). (D) polarized contraction of Myo-II (red) within an actomyosin network can pull cell junctions (green), or other points of connection to the membrane, inwards and contract a surface of a cell.

Table 1: magnitude of forces generated at molecular, cellular, and tissue scales.

values are approximations based on the literature, and provided to allow for ballpark comparison as opposed to a definitive reference (for a more detailed accounting of forces at the cellular level, please see Ananthakrishnan and Erlicher, 2007 ¹⁹⁷).

	Molecule	Stress type	Force (pN)	Citation(s):
molecules	Myo-II	tension (pulling)	3 – 5	25,26
	F- actin (polymerization)	compression (pushing)	1 – 2	191,192
Within the cell			Force (pN μm−2)
	Lamellipodium (keratinocytes and fibroblasts in culture)	compression	1000 – 2500	34,35
	Actomyosin cortex (embryos)	tension	100 - 500	36,37
	Actomyosin cortex (cell culture)		600 – 800	38
	Traction at focal adhesion (Fibroblast)		1000 – 2000	193
	Tug across a single cell-cell contact (adherens junction)		500 - 1000	39,194
Between cells in tissue culture	Cell Type
	Epithelial	tension	3000 – 4000	41
	Epithelial (during collective migration)		300 – 1000	195
	Mesenchymal		1000 – 1500	41
Between cells in embryos	Species
	Drosophila	tension	100	40
	Zebrafish		20 - 60	103
	Xenopus		5	196
	Mouse		1600	41

The core mechanism for producing pulling forces is mediated by F-actin and Myo-II. In its active, phosphorylated state, Myo-II oligomerizes into a bipolar mini-filament that can bind to opposing actin filaments and “walk” along them in an ATP-dependent manner (Fig. 1C)^20-22. This myosin motor activity produces actomyosin contractility: movement of the myosin head domain slides F-actin filaments in opposing directions, shortening the total length of the actomyosin assembly (for details of actomyosin mechanics, see^22-24). At the molecular level, a single myosin molecule generates a force on the order of several piconewtons (~3-5pN)^25,26. In the cell, actomyosin networks can generate a pulling force that tugs on points of connection between the cytoskeleton and integral membrane proteins, such as those at cell junctions. Actomyosin assemblies can be organized in one dimension, such as in a bundle or fiber (e.g., stress fibers) that can shorten or bear stress^27,28. It can also be organized in two dimensions (2D), as in the lamella of a migrating cell or the surface of an epithelial cell (i.e., apical, basal, lateral). On such 2D surfaces, actomyosin is organized as a meshwork that can draw the cell edges towards the cell center through contraction (Fig. 1D)^19,29-32.

At the scale of cells and tissues, single-molecule forces can compound significantly to produce larger forces, but the extent to which this happens depends on tissue context. For instance, the force of actin polymerization is compounded within a branched F-actin network composed of many growing filaments, such that forces can be more than three orders of magnitude greater than a single filament across an entire protruding cell surface (~1 nN μm⁻²)^33-35. Similarly, contractility of a larger actomyosin network containing many Myo-II mini-filaments (dozens to hundreds) generates tension in the cell cortex that is approximately two orders of magnitude greater than the force generated by a single Myo-II mini-filament (~100 - 800 pN μm⁻²)^36-40. This tension is largely dependent on actomyosin contractility, because inhibition of myosin activity decreases cortical tension, and increasing contractility or inhibiting Arp2/3-mediated F-actin polymerization increases tension³⁸. Tissue-level tension can be an order of magnitude greater than cellular tension (thousands of pN per square micron)^39-41, but levels of tension can vary greatly between tissues composed of different cell types, and between model organisms: tension measurements range over three orders of magnitude when comparing between species (~5 – 4000 pN μm⁻²; Table 1). Thus, it is clear that many hundreds or even thousands of motors work together to generate force in a given cell. However, because these force generating mechanisms (actin polymerization and actomyosin) are oriented and exist within a network with complex architecture, the relationship between force and motor number is not simple.

F-actin polymerization, depolymerization, and contraction happen continuously in cells, even when cells are not actively changing shape. Homeostatic levels of actomyosin contractility and actin polymerization and depolymerization (actin turnover) are part of the normal state of the actomyosin cortex²⁴, and occur regularly in the cortical F-actin cytoskeleton of multiple cell states (epithelial and mesenchymal), independent of developmental signaling or mechanical cues from cell-cell or cell-ECM adhesion^42,43. There is inherent antagonism between contractility and turnover – maintaining cytoskeletal network cohesion requires a certain amount of actin turnover to prevent Myo-II contractility from fragmenting the actomyosin network, however, actin turnover can dissipate stress generated through Myo-II contractility^44-48. Shape change occurs when these processes are up- or down-regulated locally outside of the range of this typical ‘resting state’ of normal cytoskeletal dynamics, such that a force imbalance results.

Actin-dependent force generation is evolutionarily ancient and broadly required across all animals and is also present in our unicellular eukaryotic relatives. Actins and myosins are not unique to animals; they are ancient, pan-eukaryotic protein families that are found in other multicellular eukaryotes, including plants⁴⁹, and are particularly diverse in the Holozoa (animals plus our closes unicellular relatives)⁵⁰. Actomyosin-based contractility mechanisms likely arose before the advent of animal multicellularity – they appear to underlie a collective contractile mechanism for colony morphogenesis in choanoflagellates, the unicellular protists most closely related to animals, suggesting they were present in our common ancestor¹¹. The mechanisms for cell shape changes based on F-actin polymerization are also quite ancient. The Excavate protist, Naegleria, which shared a common ancestor with humans more than 1 billion years ago, exists predominantly in an amoeboid state that completely lacks a microtubule cytoskeleton, and relies on a branched actin network regulated by many of the same actin regulators found in our own cells for protrusive motility and phagocytosis⁵¹. If, as this evidence suggests, actin-based mechanisms of cell shape change are part of the common inheritance of animals, a key outstanding question is how the fantastic variation of forms seen between species are created from a conserved set of cytoskeletal effectors. It is perhaps the evolution of complexity in localizing these processes to different areas of the cell and their coordination between cells that has generated the vast complexity in tissue form from a common set of cytoskeletal components. In the next section, we will explain how localized contraction and expansion mediate morphogenesis. We will not explain the particulars of each individual system but focus on the broader geometrical framework for understanding how tissues change shape.

Localized contraction and expansion in epithelial morphogenesis:

Epithelial tissues are widespread in animals and have polarity⁵². Epithelial cells have multiple surfaces that can contract or expand, such as apical (lumen facing), basal (ECM facing), and lateral (between neighboring cells). Furthermore, epithelial cells can exhibit a vectorial polarity with respect to the epithelial plane, called planar cell polarity⁵³. Because of the intrinsic polarity in the plasma membrane components and underlying actin cortex, proteins can be differentially localized and/or activated to induce distinct apical, basal, or lateral domain behaviors.

Collective contraction or expansion of apical or basal surfaces induces epithelial curvature. Apical and basal constriction converts epithelial cell shape from columnar to wedged, which changes local tissue curvature when happening in a population of cells (Figure 2, tissues). Apical constriction is implicated in mammalian intestinal crypt invagination⁵⁴, vertebrate lens placode invagination⁵⁵, Drosophila mesoderm and endoderm invagination⁵⁶, and Drosophila salivary gland invagination⁵⁷. Basal constriction induces the opposite curvature with respect to the apical-basal axis of an epithelium, generating the midbrain-hindbrain boundary folds in zebrafish⁵⁸ and optic cup in zebrafish⁵⁹. Expansion can also induce curvature. In the Drosophila wing disc, local basal cortex relaxation results in basal expansion that induces inward epithelial bending independently of apical constriction⁶⁰. Thus, like differential expansion of one metal on a bimetallic strip changing strip curvature⁶¹, apical or basal constriction/expansion changes epithelial curvature by changing the length of one surface relative to the other. For this mechanism to be effective, the apical and basal surfaces must be mechanically coupled through lateral edges. In addition to apical constriction, contractility along lateral edges is necessary for gastrulation in Drosophila, and in ascidians^62,64,87.

Figure 2: Shape change by contraction and expansion:

(tissues, top) polarized contraction or expansion on apical or basal surfaces can generate tissue folding. (surfaces, center) Shape change of individual cellular surfaces can also alter tissue geometry, as in basal extrusion or apical emergence. (edges, bottom) Coordinated shape change of individual edges or cellular interfaces can drive tissue movements, such as the convergent extension movements that elongates the Drosophila germ band (red). Contraction of lateral interfaces can also shorten edges between neighboring cells along the apical-basal axis, promoting tissue folding, as is seen in the Drosophila wing disc epithelium (right).

Constriction or expansion of individual epithelial cell surfaces also changes tissue architecture (Figure 2, surfaces). For example, unbalanced apical contractility (i.e., a cell is more contractile than its neighbors) can induce basal epithelial cell extrusion, such as during C. elegans gastrulation^29,63, Drosophila neuroblast ingression^65,66, and Drosophila dorsal closure⁶⁷. In cell culture, epithelial cell extrusion results when apical tension relaxes relative to neighboring cells, suggesting that the force imbalance is critical for a cell to leave the epithelium⁶⁸. Consistent with the force balance argument, cell-autonomous actin-based protrusion causes apical emergence (i.e. the opposite of extrusion), as is seen as multiciliated cells are added to an epithelium in Xenopus ^69,70. Thus, apical/basal contraction or expansion have distinct effects depending on whether they occur at the single cell or population level.

Lateral surfaces between two cells (i.e., bicellular junctions) can contract or expand leading to cell rearrangements. When these rearrangements are planar cell polarized they lead to convergence and extension movements that elongate tissues, such as in the vertebrate neural plate^71,72 and the Drosophila germband⁷³ (Figure 2, edges). Planar polarized Myo-II activation can contract junctions^71,74-77,80 – elevating tension along distinct junctional interfaces in the Drosophila germband^78,79. Mediolaterally polarized basolateral protrusive activity in the cells at junctional vertices cooperates with junctional contractility to shrink these edges⁸¹. When a bicellular junction contracts to create a 4-cell interface, expansion must occur to create a new bicellular junction orthogonal to the ‘old’ junction. In cell culture, Rac-mediated protrusions expand junctional interfaces⁸² and actin polymerization-mediated pushing has been observed to counteract contractility and expand epithelial junctions in vivo⁸³. In addition to pushing forces, contractile forces can promote junction growth by pulling at the poles of the new junction during Drosophila germ band extension^84,85. Thus, convergence and extension movements illustrate how contractility and protrusive forces in a local cell neighborhood cooperate to elicit a complicated morphogenetic movement. Lateral edges can also contract along the apical-basal axis, which shortens cells and promotes tissue folding, as is in the Drosophila wing and leg discs^60,86 (Figure 2, edges).

Indeed, morphogenetic processes are often quite complex, with multiple changes happening simultaneously within the cell or in different regions of the embryo. Contraction and expansion can coincide in space or time, sometimes with synergistic effects. For example, constriction on one side of the epithelium is often associated with expansion of the other side. In Drosophila mesoderm invagination, basal expansion closely follows apical constriction and is associated with the invagination⁸⁷. Additionally, inhibition of contractility in neighboring ectoderm cells is important for gastrulation, as it allows these cells to soften, stretch, and move in order to accommodate the cell shape changes associated with mesoderm invagination^88,89. During C. elegans gastrulation, cell ingression is assisted by neighboring cells pushing with Arp2/3-mediated actin protrusions⁹⁰. Mitotic cell entry has also been shown to be associated with apical relaxation, which can promote apical constriction and invagination of neighboring contractile cells^37,91. Thus, understanding morphogenetic events requires investigating both local properties of the tissue undergoing shape change, and the properties of the surrounding environment because both can contribute to movement.

Cell Adhesion:

The formation of robust mechanical contacts to other cells and to the extracellular environment enables cytoskeletal force transmission to power tissue shape changes. The two main adhesive structures involved in actomyosin force propagation are adherens junctions, which mediate cell-cell adhesion, and focal adhesions, which mediate cell-ECM adhesion (Figure 3). The core adhesion receptors in the adherens junction are cadherins, which are composed of extracellular calcium-dependent adhesion domain repeats that homotypically interact with cadherins on adjacent cells⁹². In mammalian embryos, loss-of-function mutations in cadherins disrupt cohesion and compaction of the early embryo, and result in failures in later morphogenetic events, including gastrulation, neurulation, and organogenesis^93-95. The principal adhesion receptors in focal adhesions are integrins, which are heterodimers containing an α- and β-subunit, that interact with a variety of ECM components via their large extracellular domains^96,97. In addition to serving as critical signaling centers that are important for cell survival and cell polarity, cell-ECM adhesions are also crucial for morphogenesis, such as during zebrafish optic cup morphogenesis and mouse neural tube closure^98-100.

Figure 3: Structure of junctions that link adhesion to the actomyosin cytoskeleton:

Adherens junction (A), and focal adhesions (B) share common organizational principles that facilitate dynamic attachment to the F-actin cytoskeleton, including stratification into a similar set of functionally distinct compartments that enable: (1) extracellular force transduction, (2) adhesion regulation, (3) intracellular force transduction, and (4) F-actin regulation.

There are physical constraints to the load-bearing capacity of adhesion molecules. Estimates of the maximum rupture forces of individual adhesion molecules are variable across the literature but are generally on the order of hundreds of piconewtons, compared to the ~5 pN generated by a single Myo-II mini-filament (Table 2). Adhesive strength varies between individual adhesion receptors, and therefore also varies between cell types expressing different complements of receptors. For example, the interaction between two E-cadherin molecules is, on average, an order of magnitude stronger than the N-cadherin:N-cadherin interaction (~200 vs. 40 pN)¹⁰¹. Similar differences exist between different combinations of α- and β-integrin subunits and ECM components (Table 2). Corresponding differences in adhesive capacity are seen between germ layers in the zebrafish embryo and could reflect either differences in adhesion proteins or different organization/concentration of the same adhesion protein (Table 3)¹⁰². These differences in adhesion strength are important, as they can alter the mechanical properties of a tissue^103,104, and drive morphogenetic movements, such as cell sorting^102,105. They also suggest that different tissues have inherently different tension-bearing capacities, which is useful to consider when comparing different morphogenetic processes.

Table 2: Examples of unbinding forces required to break or rupture bonds within mechanical connections between the actomyosin cytoskeleton and adhesion complexes.

(this represents only a partial list; for more extensive information, please see Weisel et al. 2003, and Rocha-Cusachs et al. 2012) ^132,210.

	Molecular interaction	Stress type	Rupture force (pN)	Citation:
Adhesions	α5β1 integrin : fibronectin	Tension	60 – 100	198,199
	α2β1 : collagen		100 – 160	200,201
	E-cadherin : E-Cadherin		70 - 200	101,202
	N-Cadherin : N-Cadherin		30 – 40	101
Adaptors	Talin : Actin		2	203
	α-catenin : Actin		5-10	146
	Vinculin : Actin		4-8	148
Cytoskeleton	Actin : Myosin		5 - 15	204
	Actin : Actin (within F-actin)		400 – 600	205,206
		Torsion	100 – 300	205
		Compression	0.16 – 4	207-209

Table 3: Rupture forces required to break adhesions in cells from different germ layers of a vertebrate embryo.

Organism	Germ layer	Rupture force (pN μm−2)	Citation
Zebrafish	Endoderm	400 – 600	102
	Mesoderm	700 – 1500
	Ectoderm	1500 – 3000

Adhesions can mature or strengthen due to structural changes in adhesion receptors in response to force or due to prolonged physical contact. Many protein binding interactions exhibit a ‘slip bond’ behavior, in which bond lifetimes decrease under tension. In contrast, many adhesion protein interactions are strengthened (i.e., become longer lived) under tension, a phenomenon called ‘catch bond’ behavior. The extracellular domains of E-cadherin can bind in two distinct conformations, X-dimers or strand-swapped dimers. These conformations differ in their response to tensile force: X-dimers behave like catch bonds, whereas strand-swap dimers behave like slip bonds¹⁰⁶. The catch bond effect of the X-dimer is activated at forces greater than ~20 pN, which would require the coordinated activity of multiple Myo-II motors connected to a single cadherin. This suggests that cadherin heterodimer conformation responds to local tensile forces generated by actomyosin within a tissue, switching from X-dimer catch bonds that grip strongly under load, progressing to form more robust strand-swap dimers that have a high affinity in the absence of force. Integrins also exhibit catch bond behavior that is dependent on Myo-II contractility, as well as the stiffness of the underlying ECM. When tension is applied across the α₅β₁ integrin heterodimer, the extracellular integrin headpiece shifts to an activated conformation that can bind fibronectin for longer durations^107,108. In addition, integrin adhesion force increases with substrate stiffness, as increased stiffness allows for interaction of the integrin headpiece with an additional synergy site adjacent to the primary binding site in fibronectin¹⁰⁷.

Organization of adhesive junctions:

The strength and stability of adhesion also depends on the organization of cell junctions. Studies at the nanometer scale have shown that both cell-cell^109-111 and cell-matrix¹¹² adhesions are composed of smaller clusters of adhesion proteins, on the order of 50 – 100 nm, that represent a modular unit of organization that may be a general feature of cell adhesion receptors (for in-depth reviews on cadherin and integrin clustering, see^113,114). Receptor clustering stabilizes junctions and can increase the force bearing capacity of adhesions¹¹⁵. While clustering is not necessary to establish adhesions, it locally increases adhesion receptor density, and, thereby, promotes adhesion formation¹¹⁶. Both cadherin and integrin heterodimer cluster formation are driven by multiple factors, including ligand binding, extracellular domain cis-interactions, and cytoplasmic interactions with cytoskeletal linking proteins like α-Catenin and Talin^115,117-120. Integrin clusters have up to a 6-fold increase in tension threshold compared to individual integrin heterodimers¹¹⁷, and clustered E-cadherin also has a more robust mechanical connection to the actin cytoskeleton that can resist higher tensile forces¹¹⁵. Interestingly, not all adhesion molecules within an adhesion cluster are loaded equally – in a focal adhesion cluster, there is heterogeneity of force loading on integrin heterodimers, with a majority experiencing only weak forces (~1-10pN), and a subpopulation experiencing more substantial loads¹²¹. Similarly, within a cadherin cluster, only a subset of cadherins (~50%) have adhesive trans interactions and can propagate force between cells¹²². This implies that most adhesion receptors are experiencing approximately single-molecule levels of force, but more additive forces are channeled through a subset of adhesion molecules within a cluster.

Super-resolution microscopy has also demonstrated that in addition to the two-dimensional organization of adhesion complexes into clusters, there is stratification relative to the plasma membrane. This stratification has been thoroughly demonstrated in integrin adhesions^123-125, and more recently for cadherin adhesions¹²⁶. Together, these data suggest a general model for the organization of actin-linked adhesion complexes (Figure 3): (1) an extracellular adhesion layer, which has cis interactions governing cluster formation, and a mechanosensitive element that promotes stronger adhesions under force; (2) a membrane-proximal signaling layer, that also contributes to clustering, activation of adhesion, and regulation of turnover; (3) a force transduction and cytoskeletal adaptor layer, that contains a second mechanosensitive element to meter interactions with F-actin; and (4) an F-actin regulatory layer that contains a variety of actin-binding proteins that modulate the state of the cytoskeleton (bundling, polymerization, nucleation, etc.). Importantly, while layers 1 and 2 contain components that are distinct between cell-cell and cell-ECM adhesions, the layers 3 and 4 are partially shared by both, with many proteins (Vinculin in layer 3, and Zyxin, VASP, α-Actinin, etc. in layer 4) localizing to adherens junctions and focal adhesions^123,126,129.

Coupling of force generation to adhesions:

Robust mechanical connections between adhesions and the cytoskeleton are essential for morphogenesis^127,128,130. When adhesion is compromised, the point of failure is often not breakage of extracellular adhesions, but instead rupture of the connection between adhesion complexes and the actin cortex¹⁰². This link is comprised of a ‘core’ set of adaptor proteins capable of bridging the cytoplasmic tails of adhesion receptors and F-actin. For cadherin, these adaptor proteins are α- and β-Catenin¹³¹, and for integrin, a major adaptor protein is Talin¹³². Each of these linking proteins are broadly conserved^133-135, and inhibition or mutation of them produce severe defects in adhesion^96,131. In both adherens junctions and focal adhesions, there is a larger repertoire of interacting ‘adhesome’ proteins that play redundant, modulatory, or supporting roles in facilitating linkage, on the order of 125 adhesome proteins for cadherins, and 200 for integrins^136-138. Characterizing the composition, organization, and force dependence of the network of proteins mediating cytoskeletal linkage to adhesions will continue to be an exciting area of research with implications for morphogenesis.

The linkage between the cytoskeleton and the adhesions is dynamic and can be actively regulated during development. A general model that has emerged is the concept of a molecular clutch, in which junctional adaptor proteins operate in an analogous manner to the clutch of a mechanical engine – when the clutch is ‘engaged’, pushing or pulling forces generated by the actin cytoskeleton are physically coupled to cell adhesions, resulting in force propagation to neighboring cells or the ECM^139,140. This mechanical ‘clutch’ model for the regulated coupling between an adhesive complex and the actin cytoskeleton originated from detailed study of focal adhesions and actin filament movement, where it was shown that variable traction stress relates to the lamellipodial actin retrograde flow^141,142. However, a regulated attachment between the actin cortex and adherens junctions was also shown for apically constricting cells in C. elegans gastrulation²⁹.

Part of the underlying mechanism of the molecular clutch appears to be force-sensitive elements within the structure of adaptor proteins. For example, when force is applied to Talin, it induces a shift in structural conformation that exposes a binding site for the recruitment of Vinculin, which reinforces its interaction with F-actin^118,124,143. A similar force-dependent conformational switch exists in α-Catenin, which controls vinculin recruitment at adherens junctions^144,145. The α-Catenin : F-actin interaction operates as a catch bond, increasing in binding affinity under force, suggesting multiple mechanisms for mechanosensation and linkage adaptation may be operating simultaneously¹⁴⁶. Indeed, many adaptor proteins have recently been shown to exhibit catch bonds, including Talin¹⁴⁷ and Vinculin¹⁴⁸, which suggests that mechanosensation may be a general property of adaptor proteins involved in cytoskeletal-junctional linkage. Some of these proteins, such as Vinculin and Afadin may enable strengthening of this connection in response to force^149-151. In addition, actin turnover can strengthen this connection. In Drosophila gastrulation, F-actin turnover promotes stable connection of the contractile machinery to junctions during apical constriction¹⁵². In the Drosophila pupal wing epithelium, F-actin turnover also recruits additional factors, such as Canoe/Afadin, to strengthen the connection between actomyosin and the junction¹⁵³.

The properties and mechanisms of cytoskeleton-junctional connectivity can vary between different regions of the cell, between tissues, and between organisms. Adhesome composition can vary with cell type and between adhesion molecule subtypes and can depend on tension. For example, the N-cadherin adhesome in cardiomyocytes is distinct from the E-Cadherin adhesome of epithelial cells¹⁵⁴. Proteomics approaches have also revealed that as many as 400 proteins bind to integrin adhesions in a force dependent manner and may be involved in mechanotransduction¹⁵⁵. Characterizing the composition of junctional adhesomes under different mechanical states will be critical to our understanding of force propagation in morphogenesis. The recruitment of additional adaptor proteins, such as Vinculin or p120-catenin, has been shown to modulate the total force transmitted between adhesion complexes and actomyosin¹⁵⁶, as well as the stability of junctional complexes by regulating rates of adhesion receptor endocytosis¹⁵⁷.

Macroscopic patterning of force generation in morphogenesis:

Because cytoskeletal systems are connected between cells, the macroscopic organization of cytoskeletal components, and how forces are patterned across an entire tissue is critical to morphogenesis. Similar to how morphogens set up a gradient of signaling activity to specify unique cell fate, signaling gradients can also specify unique force generating properties in cells across a tissue, which can promote morphogenesis. As opposed to cases of uniform contractility (Fig. 4A), spatial gradients of Myo-II activity (Fig. 4B) can create an imbalance of forces within a tissue that enhance tissue deformation. In addition, the actomyosin cytoskeleton can form structures that are interconnected across tissues (i.e., supracellular), sometimes including hundreds or even thousands of cells.

Figure 4: Spatial patterning of contractility:

(A) When contractility is uniform forces between cells are balanced, and deformation is limited. In contrast, when contractility is organized in a gradient (B), force imbalance can allow for increased deformation and tissue folding. (C) In chick hindgut development, an FGF signaling gradient patterns graded contractility and cell movement to produce tissue folding: endoderm anterior to the caudal intestinal portal (CIP) receives graded levels of FGF, which induces a corresponding gradient in contractility levels. (D) In Drosophila gastrulation, the dorsal-ventral patterning cascade (left) produces a gradient of Myo-II activity. A graded pattern of the transcription factor, twist, on the ventral side of the embryo (center) proceeds a gradient of Myo-II activity that powers invagination of the mesoderm (right, red).

Recent evidence suggests that force imbalances may be genetically patterned by the same signaling mechanisms involved in cell fate specification. Fibroblast growth factor (FGF) signaling stimulates cell motility and is required for axis elongation of the chick embryo¹⁵⁸. FGF signaling also stimulates actomyosin contractility and apical constriction in the zebrafish lateral line and the chick otic vesicle^159-161. In the developing chick hindgut, a gradient of FGF signaling establishes a contractility gradient that is required for the polarized collective cell migration that forms the hindgut (Fig. 4C)¹⁶². In this case, hindgut cells move from low tension to high tension, towards the source of highest FGF signal, and this movement, in conjunction with cell contraction, generates a fold in the endoderm that gives rise to the hindgut. Contractility gradients have also been observed in other systems, such as Drosophila gastrulation (Fig. 4D). During Drosophila gastrulation, the transcription factor twist controls mesoderm differentiation and its target genes that induce apical actomyosin activity are expressed in a gradient around the ventral midline^163-165. This gene expression gradient results in a gradient of apical constriction and myosin activity that is highest at the midline of the invagination and decreases with distance from the midline^{56,163,164,166}. Drosophila endoderm invagination also exhibits a wave of apical constriction, but in this case the spatial patterning results from mechanotransduction. In the endoderm, there is an integrin-mediated anchorage of cells at the primordium edge and a mechanical relay in which cells tug and stretch their neighbors inducing them to apically constrict^167,168.

Actomyosin cytoskeletal structures can also be organized at larger spatial scales within the embryo. In many morphogenetically active tissues, junctional actomyosin is oriented and linked into supracellular ‘cables’ or meshworks stretching over many cell diameters that serve important functions in tissue-scale processes such as wound healing and morphogenesis¹⁶⁹. F-actin cables or rings form in a variety of morphogenetic processes in diverse organisms, such as during mouse neural tube closure and eyelid closure^170,171, chick amniogenesis and lens placode invagination ^172,173, zebrafish epiboly and rhombomere boundary formation^174-176, and Drosophila dorsal closure, compartment boundary formation, and germband extension^79,177,178, which suggests that supracellular actomyosin cables may be an evolutionarily conserved mechanism of transmitting forces at tissue scales. However, there is some debate as to the function of supracellular actomyosin structures – do they contract or constrict structures, just at larger scales? Or do they have alternative functions, such as modulating tissue mechanical properties?

An intuitive model for the function of supracellular actomyosin networks is to generate tissue-scale contractility in order to constrict or shorten surfaces or edges. For example, in the ‘purse string’ model of dorsal closure in the Drosophila embryo, a supracellular actomyosin cable was thought to constrict in order to ‘zip’ the epidermis closed (Fig. 5A)^177,179. However, other evidence has argued against this model, instead suggesting that the actin cable creates tension to straighten tissues edges, to allow for uniform tissue closure^180-182. Actomyosin cables have been shown to play a similar role in straightening edges during the formation of compartment boundaries in multiple tissues within the Drosophila embryo – Myo-II contractility at these boundaries creates a local increase in tension that biases junctional rearrangements to suppress transient mixing between compartments (Fig. 5B)^178,183,184. A similar phenomenon also occurs at compartment boundaries in the developing zebrafish brain, where actomyosin cables refine boundaries between rhombomeres¹⁷⁶. In the early Drosophila embryo, planar polarized Myo-II is organized into a supracellular “ribbon” that functions as a ‘denoising’ mechanism to ensure morphogenetic precision in creating the cephalic fold that separates future head structures from the rest of the embryo (Fig. 5C)¹⁸⁵. Similarly, in Drosophila gastrulation, actomyosin organized into a supracellular meshwork, as opposed to a cable, promotes directional tissue stiffening and robust folding during mesoderm invagination (Fig. 5D)¹⁸⁶. In the Drosophila wing disc, multicellular actomyosin cables form in a tension-dependent manner in order to stiffen the tissue and prevent mechanical stress from degrading tissue integrity¹⁸⁷. Taken together, these findings suggest that supracellular actomyosin structures may operate macroscopically to generate tissue stiffness and/or confer robustness, and do not always function to contract tissue surfaces.

Figure 5: Supracellular actomyosin structures in embryos:

In Drosophila embryos, supracellular actomyosin cables (green) are seen during dorsal closure (A) and cases of compartment boundary formation, such as between parasegments in the early embryo (B). One function of actomyosin cables is to generate tension to straighten tissue edges and prevent mixing across boundaries (B, inset). (C) A similar phenomenon is seen in the Zebrafish hindbrain, where actomyosin cables promote boundary formation between the developing rhombomere brain segments and inhibit cell mixing. Supracellular actomyosin structures can also confer proof-reading or robustness to morphogenetic processes, such as the actomyosin meshwork present during Drosophila gastrulation (D). Here, the meshwork (green) creates directional tissue stiffness along the anterior-posterior axis, which resists bending along the long axis of the embryo, and supports folding and internalization of the mesoderm (red).

How do such extensive cytoskeletal structures form? In the Drosophila pupal dorsal thorax and wing disc, evidence suggests that supracellular actomyosin structures are, in part, organized by the adhesion apparatus, but it is unclear whether this is a general organizing principal across tissues^27,104. During Drosophila germ band extension, contractile actomyosin flows are biased towards sites of increased adhesion, which amplifies asymmetries in junctional contractility and generates anisotropic cell deformation¹⁸⁸. Alternatively, the actomyosin cytoskeleton itself could respond to extrinsic forces or constraints that are present in the tissue. Indeed it has been shown in numerous developmental contexts that actomyosin fibers are assembled or stabilized in response to stretch or tension ^79,187. More work is necessary to determine how these structures form, and interpret their function and importance for tissue morphogenesis.

Conclusions and future directions:

Overall, it is clear that organisms liberally utilize cytoskeletal contraction and protrusion and a multitude of mechanisms to polarize force-generating machines to generate the wondrous diversity of organismal form. In addition to thinking about morphogenesis as a sum of cellular building blocks, there is clear importance to understanding how cells interconnect and transmit force across a tissue and how this is patterned at a macroscopic level. Understanding how cytoskeletal structures link between cells and how forces are transmitted from cell to cell in a tissue will be a fruitful area of further investigation. In addition to force generation, supracellular cytoskeletal structures are also likely to affect the mechanical state of tissues in a developing embryo, such as whether it behaves like a solid or liquid. Indeed, recent evidence suggests that tissue fluidity can be modulated by multicellular actomyosin patterns, and that this is important for gap closure in the context of wound healing and normal developmental processes^189,190. Further studies linking supracellular actomyosin structures to tissue material properties in diverse systems will be necessary to unravel this connection.

It will be important to couple this macroscopic view of morphogenesis to higher resolution visualization of the machines that carry out force generation. Some outstanding questions are: 1) how is cytoskeletal filament alignment, which determines force polarity, controlled in a developing tissue? and 2) what is the importance of dynamic cytoskeletal behaviors, such as waves, pulses, and flows? We find the coupling of the macro-scale tissue analysis to the nano- and micro-scale organization within cells to be one of the most exciting future opportunities that will advance our understanding of morphogenesis.