Emerging interplay of cytoskeletal architecture, cytomechanics and pluripotency

Metazoan development requires the amplification of cell numbers and the commitment of appropriate cells to specialized tissue-specific roles. Populations of distinct types of stem cells arise at various developmental stages, forming pools of uncommitted cells capable of self-renewal while being poised for transitions to more specialized cell types upon proper inducement. Embryonic stem cells (ESCs), as representative pluripotent stem cells (PSCs), were initially isolated from the inner cell mass (ICM) of mouse or human blastocysts at pre-implantation stage (Evans and Kaufman, 1981; Martin, 1981; Thomson et al., 1998), with demonstrated ability to differentiate into the three germ layers as well as trophoblasts (Amita et al., 2013; Io et al., 2021; Xu et al., 2002). In contrast, somatic (or adult) stem cells are partially committed and can be differentiated into more limited tissue-specific lineages. Somatic stem cells are found in various tissue-specific niches in adults and function to replenish specialized cells during normal physiological turnover or recovery from injury (Nelson et al., 2009; Uccelli et al., 2008). Examples include haematopoietic stem cells, which generate cells of the circulatory system, or mesenchymal stem cells (MSCs) that can differentiate into bone, cartilage and cardiomyocytes, to name a few (Nelson et al., 2009). Whereas PSC populations are typically exhausted during embryonic development, differentiated somatic cells such as fibroblasts can be reprogrammed to dedifferentiate and reacquire pluripotency by judicious perturbation of key transcription factors (Park et al., 2008; Takahashi et al., 2007; Takahashi and Yamanaka, 2006; Wernig et al., 2007; Yu et al., 2007), yielding induced pluripotent stem cells (iPSCs).

Stem cells have captivated the attention of both the fundamental and translational research communities due to their central roles in developmental morphogenesis and homeostasis, and because of their vast potential in tissue engineering, regenerative medicine, drug discovery and disease modelling (Nelson et al., 2009). MSCs have been shown to be particularly amenable to microenvironmental and biomaterial engineering, as their mechanosensitivity is driven by the cell adhesion apparatuses and actomyosin contractility (see Glossary; Isomursu et al., 2019; Saidova and Vorobjev, 2020; Sun et al., 2012; Yim and Sheetz, 2012), and hence can be manipulated by a broad range of experimental techniques. Meanwhile, PSCs have served as a powerful platform for interrogating the complex gene regulatory networks that drive embryonic development (Romito and Cobellis, 2016), and their capacity for differentiation and self-organization can be harnessed to generate higher-order cellular organization, such as formation of organoids and gastruloids (Shao et al., 2017a,b; Simunovic and Brivanlou, 2017; Sozen et al., 2018; Zheng et al., 2019). Notably, a clear divergence between pluripotent and somatic stem cells in terms of general cell biology, cell morphology and differentiation factors is apparent (Nelson et al., 2009; Yim and Sheetz, 2012). However, the underlying molecular basis that underpins the distinctive cell mechanics and mechanotransduction in PSCs has not been fully understood. It is increasingly recognized that fundamental aspects of PSC biology can diverge significantly from those of differentiated or transformed tissue cells that traditionally serve as the workhorses of mechanistic cell biology (Arnold et al., 2022; Chowdhury et al., 2010b; Fan et al., 2018; Pillarisetti et al., 2011; Poh et al., 2010; Xia et al., 2019a,b). In this Review, we focus our attention on emerging findings that describe how PSCs may possess distinctive cellular architecture and cellular mechanics, and how they may therefore mobilize cellular machinery in an ‘unorthodox’ fashion to carry out their cellular functions.

PSCs include both naturally derived stem cells and iPSCs obtained from nuclear reprogramming of adult somatic cells (Park et al., 2008; Takahashi et al., 2007; Takahashi and Yamanaka, 2006; Wernig et al., 2007). Two well-known types of naturally derived PSCs are ESCs and epiblast stem cells (EpiSCs). These are extracted from pre- and post-implantation embryos, respectively (Brons et al., 2007; Evans and Kaufman, 1981; Martin, 1981; Tesar et al., 2007). Note that although mouse and human PSCs share many common characteristics, there exist important distinctions, which are discussed further below.

Pluripotency encompasses multiple interrelated states that progress in conjunction with implantation and has been reviewed in detail elsewhere (Atlasi and Stunnenberg, 2017; Hackett and Surani, 2014; Nichols and Smith, 2009; Shahbazi et al., 2017; Weinberger et al., 2016). Briefly, mouse ESCs (mESCs) derived from the ICM of pre-implantation blastocysts show unbiased naïve pluripotency, with unrestricted potential to generate all somatic cells and germ lines, and can easily re-enter embryonic development to produce chimeric embryos when incorporated into epiblast (Hackett and Surani, 2014; Weinberger et al., 2016). By contrast, mouse EpiSCs (mEpiSCs) from post-implantation embryos correspond to ‘primed pluripotency’. They exhibit a limited contribution to chimeras, while still being capable of differentiating into the three germ layers. Importantly, these two distinctive pluripotent states display morphological, transcriptional and epigenetic differences (Atlasi and Stunnenberg, 2017).

Human ESCs (hESCs) have been derived from the pre-implantation ICM of supernumerary embryos (Thomson et al., 1998). Due to ethical issues, no equivalent derivations from post-implantation human embryos have been reported. hESCs are distinct from mESCs, resembling more closely the primed pluripotency states, as they have a primed-like morphology comparable to that of mEpiSCs and exhibit global DNA hypermethylation. They express naïve pluripotency markers such as KLF17 and DPPA3 at low levels, whereas primed and/or lineage-associated genes (FGF5, LEFTY1) are expressed at relatively high levels (Hackett and Surani, 2014; also see Box 1). In addition, similar to mEpiSCs, maintenance of hESCs depends on fibroblast growth factor (FGF) and Activin–Nodal signalling pathways, rather than leukaemia inhibitory factor (LIF; see Glossary) (Dahéron et al., 2004; Vallier et al., 2005). Significant efforts have been made to obtain naïve hESCs. Expression of the naïve markers NANOG and KLF2 has been shown to reset human pluripotency to the naïve state, with inhibition of ERK mitogen-activated protein kinases (MAPKs) and protein kinase C (PKC) helping to retain this rewired naïve status in a transgene-independent manner (Takashima et al., 2014).

Cell shape, cell mechanical properties and cell responses to mechanical stimulations are strongly dependent on the collective organization and dynamics of cytoskeletal polymers. In PSCs, the initial characterization of unusual nuclear mechanics (Pagliara et al., 2014; Pajerowski et al., 2007) has more recently been accompanied by studies that delineate the unusual cytoplasmic mechanics of these cells (Table 1). Owing to space constraints, our discussion here is focused on the actin cytoskeleton, which is the primary contributor to cytoplasmic mechanics in most vertebrate contexts (Fletcher and Mullins, 2010). Actin filaments produce directed forces by polymerization and depolymerization to elicit cell shape changes or mediate internal cellular remodelling. Their network architectures are controlled by various regulatory proteins, including nucleation-promoting factors to initiate filament formation, capping proteins to cease filament growth, depolymerizing factors and severing factors to disassemble filaments, as well as crosslinkers to organize and strengthen higher-order networks. Internal or external mechanical forces can influence these regulatory proteins, either directly or via multiple signalling pathways, which in turn regulate the cytoskeletal filament organization (Fletcher and Mullins, 2010).

Mechanical properties of a cell, such as stiffness (see Glossary), can be characterized by techniques such as magnetic twisting cytometry (MTC; Chowdhury et al., 2010b), atomic force microscopy (AFM; Pillarisetti et al., 2011) or acoustic tweezing cytometry (ATC; Fan et al., 2018). Measurements made using these techniques have revealed striking differences in the cytomechanical properties of PSCs (both mESCs and hESCs) compared to those of differentiated cells (Tables 1 and 2). For example, as measured by MTC, the elastic modulus of mESCs (cultured in serum/LIF; see Glossary) is approximately an order of magnitude smaller (Chowdhury et al., 2010b) than that of differentiated counterparts. Interestingly, local cyclic stress applied by MTC through integrin ligand (RGD peptide)-tethered beads elicits cell protrusion in mESCs but not in differentiated cells, an effect that has been attributed to F-actin density and deformation ability of the cytoskeleton (Chowdhury et al., 2010b).

MSC differentiation is known to be dependent on substrate stiffness (Engler et al., 2006). Consistently, for MSCs, cell spreading increases with increasing substrate stiffness (Solon et al., 2007). For mESCs, the contribution of substrate stiffness is more complex. The cell spreading area of single mESCs in serum/LIF does not increase with increasing substrate stiffness, unlike in mESC-derived differentiated cells (Chowdhury et al., 2010b). MTC measurements have demonstrated that although the basal traction force (see Glossary) of single mESCs is elevated with increased substrate stiffness, the apical stiffness is not altered (Poh et al., 2010). Interestingly, when mESCs are grown as colonies in serum/LIF, both basal tractions and apical stiffness increase with the substrate stiffness. This has been attributed to the roles of E-cadherins and cell–cell junctions signalling in PSCs, although the underlying signalling crosstalk is not well understood (Chowdhury et al., 2010a). Furthermore, a soft substrate has been shown to contribute to self-renewal and pluripotency of mESC colonies in serum/LIF, which retain pluripotency even after 5 days of LIF withdrawal on a soft substrate (0.6 kPa, similar to the intrinsic mESC stiffness), in contrast to the rapid differentiation after LIF withdrawal for colonies on a rigid substrate (Chowdhury et al., 2010a). When cultured in 2i conditions supplemented with LIF (2i/LIF; see Glossary and Box 1), colonies of ground-state mESCs were found to be insensitive to stretching of the substrate; however, removal of 2i/LIF makes them susceptible to stretching and results in broad transcriptional changes (Verstreken et al., 2019). These findings suggest that mESCs are less responsive to classic mechanobiological perturbations, such as substrate rigidity and stretching, that elicit strong responses in differentiated cell types, leading to the notion that mESCs might be relatively mechanorefractory (Poh et al., 2010; Verstreken et al., 2019; Xia et al., 2019b). Cell–extracellular matrix (ECM) adhesions and the cortical actin cytoskeleton have thus far been implicated in contributing to such unique cytomechanical properties, based on recent studies, although it is likely that additional unidentified factors are involved (Table 2).

The contribution of integrins and associated proteins to PSC biology is well appreciated in the context of self-renewal and differentiation signalling, as has been reviewed previously (Vitillo and Kimber, 2017). However, the contribution of integrin to cytomechanics (see Glossary) has been largely characterized in adherent differentiated cells such as fibroblasts, where integrin and associated proteins form prominent focal adhesions (FAs; see Glossary) to link the intracellular actin cytoskeleton to substrate ECM. FAs in such adherent migratory cells act as anchoring points to backstop membrane retraction and promote protrusion at the leading edge, whereas in stationary cells, they serve as anchorage to maintain cell morphology (Gardel et al., 2010; Morgan et al., 2007). Importantly, FAs sense the mechanical signals from either actomyosin-generated forces or external forces exerted through or by surrounding ECM, undergoing characteristic maturation (see Glossary; Gardel et al., 2010). The mechanoresponse of FAs involves a dynamic remodelling of the compositions and structural organizations of FA proteins to further elicit downstream signalling cascades. Thus, FAs play indispensable roles in cell migration, ECM remodelling and substrate rigidity sensing (Bershadsky et al., 2003; Xia and Kanchanawong, 2017). As characterized by super-resolution microscopy, FAs contain three major structural layers along the vertical (z) axis. An integrin-signalling layer (ISL), which is closest to the plasma membrane, contains paxillin and focal adhesion kinase (FAK), while an intermediate force-transduction layer (FTL) consists of talin-1 and vinculin. The uppermost actin-regulatory layer (ARL) is comprised of actin-associated proteins, such as zyxin, vasodilator-stimulated phosphoprotein (VASP) and α-actinins (Kanchanawong et al., 2010).

The mechanobiology of FAs in PSCs has recently been investigated in mouse and human PSCs by our group, as well as by others. We found that when certain strains of mESCs are cultured as single cells, FAs are less prominent, restricted to the cell periphery and significantly smaller in area (<2 μm²) compared to the classical prominent FAs in fibroblasts (which are >4 μm²; as determined based on paxillin localization) (Xia et al., 2019b) (Figs 1B and 2A). Interestingly, however, these FAs exhibit markers of mature FAs, including vinculin, zyxin and α-actinins, and undergo assembly and disassembly in response to myosin II contractility. Moreover, transcriptomic analysis has shown that PSCs share a common reduction in the expression of LIM-domain proteins, a family of adaptor proteins that have been implicated in FA mechanosensitivity (Smith et al., 2014; Sun et al., 2020). Taken together, based on a common definition of FA maturation (Gardel et al., 2010), these findings imply that FAs in single mESCs could perhaps be described as partially mature, with maturation markers present, but with diminutive sizes and likely only low levels of FA-mediated signalling compared to that in differentiated somatic cells (Xia et al., 2019b).

By contrast, in human iPSC (hiPSC) colonies, a subset of unusually prominent FAs (each with an area of ∼4 µm²), termed cornerstone FAs, are observed at the corner cells of the colony (Narva et al., 2017; Stubb et al., 2019) (Fig. 2B,D). Since FAs are generally connected to bundled F-actin, such differences in observed FA morphology compared to that of single mESCs is consistent with a different F-actin organization in iPSC colonies, whereby the corner cells display thick, fence-like actin stress fibres that run along the colony edge and terminate at the cornerstone FAs (Narva et al., 2017). Such actin fences are reminiscent of contractile ventral stress fibres (see Glossary) found in adherent differentiated cells, with prominent colocalization of myosin IIA, phosphorylated myosin regulatory light chain (MLC) proteins and α-actinins (Naumanen et al., 2008). The actin fence and cornerstone FAs in hiPSC colonies have been shown to play crucial roles in maintaining a compact colony morphology and the pluripotency of the interior cells (Narva et al., 2017). Treatment with the Rho-associated kinase (ROCK; here referring to both ROCK1 and ROCK2) inhibitor Y-27632 results in disassembly of both the actin fence and the cornerstone FAs, significantly reducing expression of the pluripotency marker SOX2 (Narva et al., 2017). Consistent with this, when hiPSCs are cultured on micropatterned substrates to constrain the attachment area of FAs, distortions in the cornerstone FAs and actin fence organization are observed in conjunction with a more rapid loss of pluripotency markers than is observed for cells on unpatterned substrates (Stubb et al., 2019). Upon induction of differentiation using retinoic acid, cells lose the cornerstone FAs and actin fence configuration, adopting a more typical FA–actin organization (Narva et al., 2017). While these results establish a link between pluripotency regulation and adhesion–cytoskeleton architecture, it should be noted that in the colony context, cell–cell interactions are expected to have important mechanical and signalling roles (Chowdhury et al., 2010a; Li et al., 2010), and their exact involvement remains to be dissected in further studies.

At the nanoscale, FAs in differentiated cells exhibit a multilayer architecture consisting of an ISL, FTL and ARL, as described above (Bertocchi et al., 2013; Kanchanawong et al., 2010; Liu et al., 2015; Xia and Kanchanawong, 2017). Given the atypical morphology of FAs in PSCs, a salient question is therefore whether the differences in FA morphology correspond to differences in the FA nanoscale architecture. Super-resolution microscopy studies from both our group and that of Johanna Ivaska have revealed that for FAs in both mESCs (Xia et al., 2019a) and hiPSCs (Stubb et al., 2019), the ISL–FTL–ARL stratification is largely conserved, with talin-1 adopting an oriented organization, linking the cytoplasmic domain of integrins and the ISL with F-actin and the ARL; however, in mESCs, the overall vertical span of talin-1 appears to be much more compact (∼10–15 nm) compared to that in hiPSCs (∼30 nm), likely due to the reduced contractility associated with smaller FAs (Xia et al., 2019b). Interestingly, there is a notable variation in the molecular orientation of vinculin, a key mechanotransducer protein that crosslinks talin-1 with actin, in both mESCs and hiPSCs (Stubb et al., 2019; Xia et al., 2019b). Earlier studies of FAs in various differentiated cells found that vinculin could be localized and redistributed in FAs through interactions with different binding partners, such as paxillin, talin-1 and actin, before adopting an oriented organization with an elevated C terminus (Case et al., 2015). In contrast, the inverted vinculin configuration of an elevated N terminus has been observed both for mESCs on laminin (Xia et al., 2019b) and for hiPSCs (Stubb et al., 2019). Additionally, the large cornerstone FAs in hiPSCs are laterally segregated, with integrin β5, talin-1 and KANK (here referring to both KANK1 and KANK2) enriched at the FA rim, in contrast to the more homogeneous distribution of other proteins (Stubb et al., 2019). At present, it remains unclear which molecular factors regulate this differential nanoscale organization, although for mESCs, the inverted vinculin configuration has only been observed in cells cultured on laminin, but not fibronectin and gelatin, implicating differential integrin signalling as a possible mechanism (Xia et al., 2019b). All in all, these studies indicate that while FAs in PSCs may retain a common core ‘chassis’ as in differentiated cells, distinct signalling pathways likely operate that may serve to modulate the organization and, presumably, the functions of key adaptor and signalling proteins. Further studies will be needed to mechanistically define the molecular and structural linkages between the pluripotency states and the adhesion–cytoskeleton architecture.

Whereas the actin cytoskeletal organization in adherent differentiated cells, as well as MSCs, typically consists of multiple types of actin-based structures, including lamellipodia, filopodia and stress fibres, actin organization in naïve pluripotent mESCs is characterized by an actin cortex (see Glossary), containing very limited bundled actin at the cell edge, and an absence of dorsal stress fibres (see Glossary; Boraas et al., 2016; Chalut and Paluch, 2016; Xia et al., 2019a,b). Using super-resolution microscopy, our group revealed that the actin cytoskeleton in mESCs (cultured in serum/LIF) primarily consists of sparse, isotropic meshworks (Xia et al., 2019a) (Fig. 1B). In addition, a minimal dependence of the network architecture on myosin II was observed, likely because the pore size of the loose meshwork is larger than the 300-nm dimension of myosin II filaments, thus impeding any myosin II-mediated contractility. We also showed that the cortical actin meshwork organization of mESCs is regulated by a tripartite network of Arp2/3, formins and capping proteins, which promotes the formation of local transient radial assemblies of actin and associated proteins termed ‘asters’, which are hot spots of actin polymerization that are interspersed among the cortical meshwork (Xia et al., 2019a). Formin-driven long actin filaments appear to determine the cortex mechanics, as inhibition of formin by SMIFH2 resulted in reduced cortical stiffness. These findings suggest that the asters may contribute to the unusually low elastic modulus of naïve mESCs by limiting formin-dependent actin polymerization, hence giving rise to sparse cortical networks that are sterically recalcitrant to myosin II-mediated crosslinking and contraction. In turn, the sparse actin network architecture of mESCs might account for the unusual cytomechanical properties, such as low cell stiffness and attenuated responses to substrate rigidity. Intriguingly, upon exit from pluripotency, the sparse actin cortex of mESCs appears to transform rapidly into prominent stress fibres, which likely involves concerted changes in the transcriptional regulation of multiple genes (Xia et al., 2019a). However, further detailed analysis will be needed to fully characterize this process.

Beyond mESCs, the roles of actin cytoskeletal organization have been investigated mostly in the context of human PSC (hPSC) differentiation. For example, latrunculin A treatment has been shown to significantly promote the generation of pancreatic β cells from hPSCs, both in vitro and in vivo (Hogrebe et al., 2020), whereas inhibiting actin polymerization or actomyosin activity enhances endodermal differentiation but reduces mesodermal differentiation (Boraas et al., 2018). Given the significant extent of actin reorganization observed during early embryonic development (Lim and Plachta, 2021), it is likely that many salient aspects of how actin contributes to PSCs remain to be discovered, especially given that the role of actin remodelling in MSC lineage differentiation has been well documented (Khan et al., 2020; Saidova and Vorobjev, 2020). For example, actin polymerization favours the osteogenic lineage (Sonowal et al., 2013) or chondrogenesis (Lim et al., 2000), whereas actin depolymerization promotes adipocyte differentiation (Chen et al., 2018).

Adherent differentiated cells, together with MSCs, are known to robustly respond to mechanobiological stimuli such as a rigid substrate by mobilizing the growth of cell–ECM adhesions and cytoskeletal contractility, resulting in upregulation of cytomechanical properties, such as cell spreading area, actin cytoskeletal organization and cell stiffness (Gardel et al., 2009; Solon et al., 2007) (Fig. 1A). Over the past decades, such mechanoresponses have been widely considered as key mechanobiological phenotypes. As discussed above, recent findings suggest that such a mechanoresponse paradigm is not necessarily applicable to PSCs. Indeed, naïve mESCs are considered to be relatively mechanorefractory, showing attenuated sensitivity and responses to mechanical stimuli such as substrate rigidity and cell stretching (Chowdhury et al., 2010b; Poh et al., 2010; Verstreken et al., 2019; Xia et al., 2019b). Importantly, mESCs become mechanosensitive to mechanical signals upon exiting the ground state, in conjunction with broad changes in transcriptional profiles (Verstreken et al., 2019). While the questions of how and why naïve mESCs maintain the mechanorefractory status remain very much open, here we summarize recent findings that delineate the linkage between PSC fate and cytomechanical properties.

The contribution of transcription factor networks to pluripotency regulation has been extensively studied over the past two decades, but surprisingly, the question of how pluripotency is regulated by cell shape and cell mechanics has only begun to be addressed recently. Back-to-back studies (Bergert et al., 2021; De Belly et al., 2021) have reported that during exit from pluripotency induced by removal of 2i/LIF, naïve ground-state mESCs undergo a morphological transition from a round morphology to a spread morphology in concert with a drastic reduction in membrane tension. Indeed, the maintenance of naïve mESCs in a high-membrane-tension state by various means, including overexpression of constitutively active ezrin and a chemical biology approach, severely hinders their exit from naïve pluripotency (Bergert et al., 2021). The underlying mechanism appears to involve a β-catenin (also known as CTNNB1)- and RhoA-mediated decrease in the phosphorylation of ERM proteins (ezrin, radixin and moesin) (De Belly et al., 2021). ERM proteins together with myosin I (mainly MYO1B) are implicated in coupling the actin cortex to the plasma membrane, as their depletion has been shown to reduce membrane tension and thus permit pluripotency exit (De Belly et al., 2021). Here, exit of pluripotency is accompanied by an increase in RAB5A-dependent endocytosis of FGF signalling components, which activates ERK signalling and directs exit from the pluripotency status, thus identifying ERK signalling as the major pluripotency pathway responding to changes in membrane tension (De Belly et al., 2021). It should be noted, however, that a reduction in membrane tension alone is not sufficient to drive pluripotency exit, suggesting that membrane tension may thus serve as a mechanical checkpoint that gates exit from naïve pluripotency (De Belly et al., 2021).

Membrane tension is known to pleiotropically affect numerous cellular events, including endocytosis (Riggi et al., 2019). Consistently, clathrin-mediated endocytosis (CME), which is important for the internalization of E-cadherin (CDH1) and TGF-β receptor type 1 in mESCs, has been shown to regulate pluripotency (Narayana et al., 2019). Knocking down the clathrin heavy chain (CLTC) in mESCs results in loss of pluripotency in conjunction with reduced E-cadherin and increased TGF-β and ERK signalling (Narayana et al., 2019). CLTC knockdown also leads to reorganization of the actin cytoskeleton in mESCs, promoting formation of actin stress fibres and increasing cell stiffness (Mote et al., 2020). However, restoring cell softness by treatment with the actin polymerization inhibitors latrunculin A or cytochalasin D is insufficient to recover pluripotency (Mote et al., 2020), suggesting that CME might be indispensable for pluripotency maintenance signalling beyond the regulation of cell mechanics; however, the exact mechanisms remain to be uncovered.

Unlike mESCs, which can proliferate and remain pluripotent in isolation, hESCs and hiPSCs suffer from poor survival after their dissociation into single cells (Chen et al., 2010; Narva et al., 2017; Ohgushi et al., 2010; Watanabe et al., 2007). Historically, the vulnerability to apoptosis after cell detachment and low cloning efficiency of individualized cells has hindered applications such as clonal isolation after gene manipulation. The addition of the ROCK inhibitor Y-27632 has been shown to remarkably reduce dissociation-induced death (Watanabe et al., 2007). Indeed, Y-27632 treatment helps maintain pluripotency marker expression (including OCT3/4, which is also known as POU5F1, and SSEA4, which is also known as ST3GAL2), even after 30 passages, and improves cloning efficiency (Watanabe et al., 2007). Importantly, cells treated with Y-27632 can still form teratomas and differentiate into ectoderm, mesoderm and endoderm precursors. Furthermore, ROCK inhibition also protects hESCs from apoptosis during serum-free suspension culture (Watanabe et al., 2007).

The molecular mechanisms underlying dissociation-induced apoptosis appear to involve overactive actomyosin contractility, which results in hESC blebbing and reduced viability (Chen et al., 2010). When hESCs are replated as single cells, most cells form pronounced blebs and start dying at ∼6 h (Chen et al., 2010). Cell blebs typically result from unbalanced stresses and eventually terminate in cell death if a homeostatic balance is not restored (Charras et al., 2005). Thus, when myosin II is inhibited with blebbistatin, blebbing is inhibited and cell adhesion and spreading are improved, increasing cell survival and cloning efficiency (Chen et al., 2010). Additionally, the silencing of three MLC-family genes (MRLC1, MRLC2 and MRLC3, also known as MYL9, MYL12B and MYL12A, respectively) also reduces cell death after dissociation. Consistent with this, inhibition of ROCK, upstream of myosin IIA, using either Y-27632 or siRNA-mediated silencing, suppresses MLC phosphorylation, reduces blebbing and improves cell viability (Chen et al., 2010). The underlying signalling pathway was further mapped in another study that identified Abr, a Rho guanine-nucleotide-exchange factor (GEF) protein with a functional Rac GTPase-activating protein (GAP) domain, as the master regulator of dissociation-induced apoptosis and actomyosin contractility in hESCs (Ohgushi et al., 2010). Abr activity gives rise to the so-called ‘Rho high, Rac low’ condition that is responsible for the initial hyperactivation of actomyosin, resulting in apoptosis (Ohgushi et al., 2010). These effects can be abrogated by depletion of Abr and can be recapitulated by artificial Rho or ROCK activation in conjunction with Rac inhibition (Ohgushi et al., 2010).

Taken together, although these studies suggest multiple intervention points for contractility in hESCs or hiPSCs, the long-term effects of these treatments vary in their severity. For example, whereas pharmacological disruption of actin using cytochalasin D (an inhibitor of barbed-end polymerization) and swinholide A or mycalolide B (which severs actin filaments) inhibits blebbing and improves cell survival in the short term (24 h), colony formation is not improved over a longer timescale (5 days) (Chen et al., 2010). These effects could be due to the ubiquitous roles of actin in essential cellular functions beyond cell contractility. Notably, actomyosin contractility is also essential for the cleavage furrow and cytokinesis; thus, long-term blebbistatin treatment inhibits cell division (Chen et al., 2010). Fortunately, long-term treatment with ROCK inhibitors has been found to support the increased initial survival and cloning efficiency by sufficiently suppressing the initial blebbing and cell death without compromising mitotic cell division (Chen et al., 2010; Watanabe et al., 2007).

PSCs are maintained in vitro as colonies, whose morphology (shape, compactness, integrity) has been widely used to assess the quality of ESC culture (Li et al., 2010). For example, during iPSC reprogramming from somatic cells, the presence of compact colonies with tight cellular association is a simple but reliable readout for successful iPSC conversion (Takahashi and Yamanaka, 2006). Here, we primarily focus on the mechanical aspects of colony organization, though it should be noted that both mESC and hESC colonies largely rely on cell–cell junctions mediated by E-cadherin, whose involvement in PSC signalling has been reviewed elsewhere (Pieters and Van Roy, 2014).

In contrast to the deleterious effects of hyperactive actomyosin on apoptosis of single hESCs, actomyosin contractility is crucial for both mESC and hPSC colony maintenance (Du et al., 2019; Narva et al., 2017; Rosowski et al., 2015; Stubb et al., 2019). F-actin and myosin IIA (but not myosin IIB) are enriched at the mESC colony boundary, effectively forming a colony-spanning three-dimensional (3D) supracellular actomyosin cortex (3D-SAC; Fig. 2C) (Du et al., 2019). Use of laser ablation to estimate tension from the recoil rate has revealed high tensile forces in the 3D-SAC, and 3D-SAC disruption leads to rapid colony collapse within several minutes as well as loss of pluripotency, as assayed by reduced OCT3/4 and NANOG expression and reduced alkaline phosphatase staining (Du et al., 2019). Likewise, local inhibition of myosin II using two-photon-mediated crosslinking of azidoblebbistatin (Képiró et al., 2012, 2015) also results in flattened colonies and a rapid decrease in expression of pluripotency markers (Du et al., 2019). Taken together, these results indicate that the 3D-SAC is indispensable for maintaining colony morphology and pluripotency, analogous to the cornerstone FA and actin fence arrangement observed in hiPSCs discussed above (Narva et al., 2017; Stubb et al., 2019). Stronger actomyosin contractility has also been observed at the edge of hESC colonies, as indicated by a higher traction force magnitude compared to that in the colony interior (Kim et al., 2021; Rosowski et al., 2015; Xue et al., 2018). The divergence in cell shape and cytoskeletal contractility between peripheral and central cells appears to involve the canonical SMAD-dependent bone morphogenetic protein (BMP) signalling pathway, in which SMAD phosphorylation activates gene transcription to regulate embryogenesis (Kopf et al., 2014). During embryonic development, cells are required to identify their positions within the embryo and differentiate accordingly to generate well-defined forms and patterns (Chan et al., 2017). When hPSCs are cultured on confined islands using microcontact printing, a difference in cell shape and contractility is observed between the central and peripheral cells, and the patterning of neuroepithelial and neural plate border cells is induced, mimicking early neurulation in vivo (Xue et al., 2018). High peripheral actomyosin contractility appears to be a major mechanism that helps demarcate the colony boundary; however, whether signalling from other adhesion receptors, such as cadherin, is involved remains to be determined.

Apicobasal polarity, the differential distribution of cytoskeletal and membrane-associated proteins, is indispensable for symmetry breaking, patterning and morphogenesis of PSCs (Kim et al., 2021; Simunovic and Brivanlou, 2017). hPSC colonies exhibit position-dependent apical structures and functions (Kim et al., 2021). Since peripheral cells can be stained more readily by lipophilic dyes, central and peripheral cell populations can be sorted and subjected to RNA sequencing analysis, which has revealed high expression of actin-related genes in the peripheral population. Indeed, actin dynamics appear to be crucial to establish the differential central–peripheral patterning, as treatment with Y-27632 has been observed to disturb the distribution of apical markers (Kim et al., 2021).

In the embryonic development stage from pre- to post-implantation, epiblast cells undergo polarization as a consequence of integrin β1 signalling from extra-embryonic tissues (Shahbazi, 2020; Shahbazi and Zernicka-Goetz, 2018). Following polarization, epiblast cells transform into a rosette-like structure and undergo lumenogenesis to generate the pro-amniotic cavity (Shahbazi and Zernicka-Goetz, 2018). Impairment of amniotic cavity formation disrupts gastrulation, lineage specification and developmental progression (Shahbazi, 2020). Strikingly, a transition of pluripotency from the naïve to primed status is a pre-requisite for lumenogenesis, as maintaining mESCs in the naïve 2i/LIF condition prevents amniotic cavity formation (Shahbazi et al., 2017; Shahbazi and Zernicka-Goetz, 2018).

The 3D culture of PSCs in 3D Matrigel can mimic lumenogenesis in vitro, serving as a powerful experimental platform. Two-cell cysts of hESCs can form in Matrigel, and these cysts exhibit the representative apical markers ezrin and PKC in a region that is surrounded by early endosomes and the Golgi (Taniguchi et al., 2015). The cysts collapse after 5 days, but the hESCs maintain their pluripotency. Actin polymerization by the formin mDia1 (also known as DIAPH1) has been implicated in formation of the lumen, as treatment with the formin inhibitor SMIFH2 decreases lumen formation, whereas transfection with a constitutively active mDia1 mutant dramatically increases lumen formation. Similarly, treatment with the Arp2/3 inhibitor CK666 also leads to decreased lumen formation, implying that actin branching is crucial for lumenogenesis. In contrast, increased actomyosin contractility appears to suppress lumen formation, highlighting the critical importance of maintaining a fine balance between structures and mechanics during early embryogenesis (Taniguchi et al., 2015).

MSCs from adult tissues have been investigated extensively for their mechanical responses following a broad spectrum of perturbations. A vast literature on MSCs, which has been reviewed previously (Saidova and Vorobjev, 2020), thus provides valuable background insights for the fledgling research into PSC mechanobiology. Briefly, ECM properties, physical cues and chemical signalling can all regulate the proliferation and differentiation of MSCs. Cell shape and lineage commitment of MSCs can be controlled by shear stress, matrix stiffness, surface adhesiveness and surface topography, even without specific chemical signals. A wide array of cytoskeletal components, ranging from the primary cilium and FAs to actin, microtubules and intermediate filaments, participates in mechanotransduction in MSCs, while their differentiation is regulated by sophisticated signalling networks, including RhoA–ROCK, AKT–ERK and Hippo–YAP/TAZ signalling (Fan et al., 2019; Saidova and Vorobjev, 2020).

Compared to the extensive studies on MSC differentiation (Guilak et al., 2009; McBeath et al., 2004; Saidova and Vorobjev, 2020), the involvement of mechanical cues in PSC differentiation is much more nuanced (Sun et al., 2014; Xue et al., 2018). Substrate rigidity is an important mechanical factor that influences neural differentiation of hPSCs. In a comparison with glass substrates and rigid polydimethylsiloxane (PDMS) micropost arrays (PMAs), soft PMAs have been shown to not only accelerate neural induction but also dramatically increase the purity and yield of functional motor neurons from hPSCs (Sun et al., 2014). However, a key distinction appears to be that while mechanical cues are sufficient for MSC differentiation in a wide variety of contexts (Gao et al., 2010; McBeath et al., 2004), stringent soluble factors at specific time points are still required for PSC differentiation (Sun et al., 2014). In other words, whereas MSCs can be considered to be ‘primed’ for mechanodifferentiation, PSC differentiation requires a more extensive and multi-step signalling package. Indeed, these features of PSCs have stimulated recent research about direct conversions, in which differentiated cells are transdifferentiated into different specialized cell types without first reverting to the pluripotent state (Jopling et al., 2011; Mollinari et al., 2018). Examples include transdifferentiation of B cells into macrophages upon expression of the transcription factors CEBPα and CEBPβ (also known as CEBPA and CEBPB, respectively; Xie et al., 2004), induction of murine pancreatic α cells to form β cells for replacement (Thorel et al., 2010), as well as the conversion of mouse fibroblasts into functional neurons by expression of ASCL1, BRN2 (also known as POU3F2) and MYT1L in vitro (Vierbuchen et al., 2010).

The recent findings discussed here highlight how PSC mechanobiology can significantly deviate from better characterized systems, such as adult stem cells and differentiated cells. Mechanotransduction in PSCs mediates key pluripotency and differentiation processes and operates via distinct molecular mechanisms that are still not fully understood. We speculate that the mechanobiological signature of PSCs (including their unique cellular architecture and cytomechanics, as well as the mechanorefractory behaviours of naïve mESCs) likely correlates with the physicochemical parameters that are compatible with pluripotency maintenance, such as the soft nuclear mechanics of PSCs (Pajerowski et al., 2007). However, the extent to which such mechanobiological phenotypes correspond to the causes or effects of pluripotency is currently unclear.

PSCs have been studied extensively, in large part due to their fundamental importance and vast potential applications. Nevertheless, there remains a large gap in our understanding of the interplay between mechanical cues and pluripotency. Further investigation and re-examination of mechanobiological paradigms in the context of PSCs are needed. We anticipate that new findings will come from technological advances in probing cellular mechanotransduction, especially in terms of the ability to recapitulate desirable properties in a precisely controlled manner. For example, opto- and chemo-genetic tools can enable direct control of molecule-specific force transmission (Wang et al., 2019; Yu et al., 2020), while super-resolution microscopy techniques (Bertocchi et al., 2017; Xia et al., 2019b) and various molecular tension sensors could be used to report on mechanosensitive activity of crucial factors (Acharya et al., 2017; Liu et al., 2020; Wang and Wang, 2009). Likewise, further developments in biomaterials or microengineered substrates for long-term culture of PSCs, organoids and embryoid systems are necessary and logical next steps for interrogating the mechanobiology of PSCs (Labouesse et al., 2021; Zheng et al., 2019).

We gratefully acknowledge Diego Pitta de Araujo for generating the graphical illustrations.