Intracellular organization in cell polarity – placing organelles into the polarity loop

Cell polarity is involved in many aspects of cell and developmental biology and is of fundamental importance for processes as diverse as cell division, differentiation, proliferation, morphogenesis and cell motility (Campanale et al., 2017; Guzmán-Herrera and Mao, 2020; Piroli et al., 2019). Cell polarity can be described as a robust maintenance of cellular asymmetries and their functional exploitation by the cell, for instance during directed cell migration (Fig. 1). These asymmetries are found at different levels – from asymmetries in protein or lipid levels (Fig. 1A,D) to higher-order protein complexes and organelles (Fig. 1B,C) (Campanale et al., 2017; Woodham and Machesky, 2014). Although these asymmetries can either be induced by external cues or can arise spontaneously through internal symmetry breaking events (Abu Shah and Keren, 2014; Martin, 2015; Vendel et al., 2019), the presence of numerous feedback loops between different molecules inside cells enables long-lasting subcellular patterns and allows their productive use. Since feedback between different molecules is at the basis of self-organization in biology (Wedlich-Söldner and Betz, 2018), cell polarity exemplifies the self-organizing property of living systems. In that sense, cell polarity can be thought of as an emergent property of interconnected cellular asymmetries. In this Review, we exemplify the well-studied asymmetry in polarity complexes and their downstream signaling pathways as well as the asymmetric architecture of the actin cytoskeleton to highlight that the asymmetry provided by intracellular organelles and lipid trafficking routes is an indispensable element of cell polarity. We then discuss several recent lines of evidence showing how the coupling between organelle positioning and other intracellular processes can support cell polarity establishment and maintenance.

One prominent asymmetry in cells is found at the level of the distribution of three broadly conserved polarity complexes: partitioning defective PAR (CDC42–PAR3–PAR6–aPKC), Crumbs (Crb–PALS–PATJ) and Scribble (Scrib–Dlg–Lgl) complexes (note that mammals have more than one isoform of some of these components) (Campanale et al., 2017). Details on these polarity complexes have been extensively reviewed elsewhere (Peglion and Goehring, 2019; Pichaud et al., 2019), and it is becoming clear that far from being stable complexes, they show dynamic behavior allowing their segregation in space, adaptation to changing context, and, importantly, that they participate in diverse polarity networks (Peglion and Goehring, 2019). The cortical tethering of polarity complexes is often considered to initialize polarization events in a linear cascade, yet, a dynamic maintenance of polarity is advantageous for flexible and rapid cellular adjustments (Peglion and Goehring, 2019; Pichaud et al., 2019). Indeed, several proteins of the polarity complexes have transmembrane domains [such as Crumbs (Crb)] or interact with membrane lipids through membrane-binding adaptor proteins (such as Par3 and Cdc42) (Peglion and Goehring, 2019), and thus, their localization and/or accumulation is dynamically regulated by membrane trafficking.

In particular, dynamic maintenance of polarity can be achieved through the cycling of proteins between distinct conformational states (Peglion and Goehring, 2019). Pioneering work in yeast has demonstrated that the dynamic switch between different conformations of the Rho GTPase Cdc42, which is part of the PAR polarity complex, is key to polarity establishment in this organism (Marco et al., 2007; Slaughter et al., 2009). Rho GTPases are small membrane-bound signaling proteins of the Ras superfamily, whose evolutionary conserved characteristic is to cycle between an ‘on’, GTP-bound state, and an ‘off’, GDP-bound state. The three canonical Rho GTPases, Rac1, Cdc42, and RhoA, are well characterized to build polarized gradients of protein activities along the perimeter of cells (de Beco et al., 2018; Etienne-Manneville and Hall, 2001; Pertz, 2010). For instance, in polarized mesenchymal migrating cells, Rac1 and Cdc42 are active at the front, whereas RhoA is mainly active at the rear (Fig. 1A) (de Beco et al., 2018; Etienne-Manneville and Hall, 2001; Pertz, 2010). To sustain the asymmetric activities of GTPases, their inevitable dispersion must be balanced by active processes such as differential immobilization, local activation or deactivation, and directed transport (Martin, 2015; Woodham and Machesky, 2014). Several mechanisms have been characterized in detail for Cdc42: the intrinsic GTP–GDP cycle of Cdc42 together with its activators [guanine exchange factors (GEFs)], (Bruurs et al., 2017), and deactivators, including gtpase-activating proteins (GAPs) and guanine nucleotide dissociation inhibitors (GDIs) that locally immobilize and activate or deactivate Cdc42 (Klünder et al., 2013; Woods et al., 2016), constitutes a minimal system that can sustain Cdc42 asymmetries. In addition to this intrinsic regulation, the asymmetry in Cdc42 activity is amplified through feedback loops that are based on transport along the actin cytoskeleton (Marco et al., 2007): in their active forms, Rho GTPases recruit and activate a large variety of effectors mostly involved in the remodeling of the actin cytoskeleton (de Beco et al., 2018; Phuyal and Farhan, 2019). Conversely, actin-dependent transport of Cdc42 is used to concentrate Cdc42 at specific regions of the plasma membrane (PM) (Layton et al., 2011). Mathematical modeling and experimental work in yeast have shown that this positive feedback is sufficient for polarity to emerge (Freisinger et al., 2013; Slaughter et al., 2009). It has been proposed that combining several mechanisms for accumulation through positive feedback ensures the reliability and spatial precision in cell polarization (Freisinger et al., 2013). Additionally, coupling of several cellular mechanisms through feedbacks has the advantage that distinct shapes of Cdc42 distributions can be tuned to be exploited for distinct functions such as vegetative growth or mating of yeast cells (Slaughter et al., 2009).

The finding of a positive feedback between Cdc42 and actin is interesting because the actin cytoskeleton constitutes a prominent feature of cell polarization. Its dynamic properties can induce symmetry breaking events (Abu Shah and Keren, 2014), and the various mechanisms for actin cytoskeleton remodeling lead to strong asymmetric distributions of the network (Fig. 1B). Typically, actin nucleation and branching are initiated downstream of Cdc42 and Rac1, which recruit nucleators of actin-related protein (Arp) 2/3 complex-dependent branched actin, such as the Wiskott–Aldrich syndrome protein (WASp) and the WAVE complex, respectively (Molinie and Gautreau, 2018). Large linear stress fibers are formed downstream of RhoA (Lawson and Ridley, 2018). Of note, the global asymmetric remodeling of actin induces a flow that is transmitted to the entire cell cytoplasm: the actin retrograde flow (Gardel et al., 2008). It is mediated by actin polymerization and myosin II-dependent contraction (Bugyi and Carlier, 2010; Burnette et al., 2011). Yet, although actin is under the regulation of polarity complexes (Campanale et al., 2017; Elsum et al., 2012; Phuyal and Farhan, 2019), actin tracks – in combination with motor proteins of the myosin family – are employed to mediate the accumulation of Cdc42 at the polarizing tips in yeast, amplifying cell asymmetries through feedback (Fig. 2, arrow 1). In mammals, polarized actin flow has been shown to control the localization of polarity factors (Maiuri et al., 2015). Moreover, surprising observations on the rearward movement of the nucleus during cell polarization have uncovered that the actin retrograde flow drags organelles with the cytosolic flow, which is facilitated by transmembrane actin-associated nuclear (TAN) lines (reviewed in Calero-Cuenca et al., 2018) and therefore leads to the polarized positioning of organelles (Fig. 1C). In addition, the small GTPase Cdc42 has been shown to control the orientation of organelles itself through the microtubule cytoskeleton (for a review, see Haga and Ridley, 2016), to which all intracellular organelles are dynamically connected through motor proteins; in vertebrate cells, Cdc42 activation leads to capture and stabilization of microtubules at the cell cortex, where they interact with dynein. This minus-end-directed motor pulls on them to reorient the microtubule-organizing center (MTOC) towards the protruding front (Hendricks et al., 2012; Manneville et al., 2010). Because endomembranes, such as the Golgi complex, the endosomal recycling compartment or the endoplasmic reticulum (ER) are closely linked to the MTOC, they reorient together, leading to an asymmetrical positioning of organelles (Egea et al., 2015; Hehnly et al., 2010). This highlights that cellular asymmetries at different levels help to maintain each other through feedback (Fig. 2). This facilitates collective interactions and the emergence of a stable cell polarity. In the next sections, we will review how an asymmetrical positioning of organelles can support cell polarity establishment and maintenance by providing diverse opportunities for feedback.

Intracellular organelles are not randomly positioned in cells but are found at reproducible locations in a given geometry and cell type (Schauer et al., 2010). As part of the endomembrane system, they constitute an intracellular compass that asymmetrically positions the trafficking machinery along the polarity axis of the cell (Fig. 1C). It has long been recognized that the vectorial orientation of the trafficking pathways facilitates the directed, vectorial delivery of polarity proteins, including Cdc42, towards specific regions of the PM, contributing to a feedback for the amplification of protein asymmetries (Fig. 2, arrow 2) (Phuyal and Farhan, 2019).

The Golgi complex is the major organelle in which newly synthesized proteins are post-translationally modified for sorting and intracellular trafficking. The biosynthetic, anterograde pathway through the Golgi is primordial to the secretion of extracellular matrix components (e.g. collagen) and the trafficking of transmembrane and membrane-bound proteins to the PM (e.g. integrins or PAR proteins, Fig. 3, point 1) (Tortosa and Hoogenraad, 2018). The Golgi complex is positioned close to the MTOC through dynein-mediated transport of Golgi-ministacks towards the MTOC, which is partly regulated by the GTPase Rab6 (Olenick and Holzbaur, 2019). The repositioning of the Golgi complex to the cell front is an effective mean to deliver newly synthesized proteins to defined sites of the cell cortex, contributing to their local accumulation (Fig. 3, point 1) (Fourriere et al., 2019; Millarte and Farhan, 2012; Xing et al., 2016). Interestingly, the Golgi complex has the capacity to nucleate microtubules in a centrosome-independent manner, which sustains polarized secretion and directed cell migration (Efimov et al., 2007; Ríos et al., 2004). In addition to the anterograde pathway, the Golgi complex is a main hub for the retrograde pathway, which allows proteins coming from the PM, which have been endocytosed in endosomes, to traffic to the Golgi complex to be either further transported to the ER or re-sorted for secretion (Johannes and Wunder, 2011). For example, retrograde Golgi traffic is used to relocalize β1-integrins from previous sites of function to the protruding edge of highly polarized migratory cells (Fig. 3, point 2) (Shafaq-Zadah et al., 2016). Interestingly, a similar mechanism has also been found recently for the retrograde traffic and polarized secretion of the adapter molecule linker for activation of T cells (LAT) towards the immune synapse for T lymphocyte activation (Carpier et al., 2018).

In addition to the Golgi, the endosomal recycling compartment is closely linked to the MTOC through Rab11-dependent dynein recruitment and transport to the MTOC (Olenick and Holzbaur, 2019). Thus, its repositioning to the cell front also facilitates the accumulation of PM proteins that have been endocytosed. An example is the Rab11-dynein-mediated regulation of the polarized distribution of the Par3 polarity protein in the Drosophila melanogaster oocyte (Fig. 3, point 2) (Jouette et al., 2019). Par3 is endocytosed from the posterior PM, relying on Rab5 and the PM lipid phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] to be transported towards the anterior cortex via the Rab11 compartment through MTs and dynein motor-dependent transport. Interestingly, Rab11 (Rab11A and Rab11B forms in mammals) and its isoform Rab25 (also called Rab11C) have been widely implicated in the context of cell polarity loss during cancer development (reviewed in Kelly et al., 2012).

Polarized delivery can also occur from the lysosomal compartment. Particularly in immune cells, such as B cells, in which lysosomes are strongly associated with the MTOC, polarized secretion of lysosomes to the immune synapse regulates antigen processing and presentation functions (Yuseff et al., 2011). Of note, it was recently shown that polarized delivery of lysosomes was dependent on the exocyst complex protein Exo70 (also known as EXOC7) (Sáez et al., 2019), which interacts with Cdc42 in yeast and mammalian cells (see below). Polarized secretion of lysosomes can also be observed in osteoclasts, which are large multinucleated cells exquisitely adapted to resorb bone matrix (Ng et al., 2019).

More globally, members of Rab GTPases, which control the identity of a large variety of endomembranes, have been shown to participate in polarity establishment (Parker et al., 2018). In addition to Rab6 (Rab6A and Rab6B) and Rab11, Rab8 (Rab8A and Rab8B) functions downstream of Rab11 through the exocyst complex for polarized secretion of Par3 and activation of Cdc42 at the apical surface (Fig. 3, point 3) (Bryant et al., 2010). Inversely, the polarized delivery of the exocyst complex has been shown to be regulated by Cdc42 and Par3 polarity proteins that directly bind to Exo70 for exocyst docking (Ahmed and Macara, 2017; Yamashita et al., 2010). Another Rab GTPase that is implicated in polarity is Rab35, which physically tethers intracellular vesicles containing the lumen-promoting factor podocalyxin, the atypical protein kinase C (aPKC), Cdc42 and Crumbs3 to the cell division cleavage site in 3D epithelia (Fig. 3, point 3) (Klinkert et al., 2016). Rab35 also recruits Cdc42 and Rac1 to sites of filopodium and lamellipodium formation in Drosophila immune cells (Shim et al., 2010). In epithelial cells, Rab35-dependent regulation is important for lumen formation and apico-basal polarity establishment after cytokinesis (Klinkert et al., 2016).

Although intracellular trafficking pathways are numerous, in all examples above they are globally used to collect proteins and re-polarize them through the transits of anisotropic intracellular compartments. The simultaneous use of numerous trafficking pathways provides the cell with many possibilities to provide feedback on the actions of Cdc42 or polarity complexes. Thus, the unifying mechanisms by which intracellular compartments contribute to directionality rely on their nature to serve as pipelines for transport. In addition to the translocation of proteins, the endomembrane compass can be used for directed lipid transport, which will be discussed in the next section.

Transport of lipids by the endomembrane system contributes to bulk lipid accumulation. For instance, membrane flows to the leading edge of a migrating cell provide lipids for localized PM extension for its protrusion activity (Fig. 3, point 4). This is best evidenced during amoeboid cell migration, which is independent of adhesion and whose directionality is powered by membrane flow to the cell front (O'Neill et al., 2018). This is, for instance, achieved by internalization and recycling of the PM, often regulated by the GTPase Rab5 (Rab5A, Rab5B and Rab5C) (Jouette et al., 2019; Malinverno et al., 2017; O'Neill et al., 2018). Of note, Cdc42 inactivation has been found to decrease the rate of membrane flow to the cell surface due to reduction of exocytosis. Interestingly, Cdc42 inactivation could be rescued by increased recycling after overproduction of the previously mentioned GTPase Rab11 (Mohammadi and Isberg, 2013).

In addition, the endomembrane system supports selective lipid enrichment (Fig. 3, point 5). Because organelles are of different lipid composition, their anisotropic positioning leads to an asymmetric redistribution of lipids, such as phosphoinositide lipids (Fig. 1D). This asymmetric distribution of lipids in cell membranes feeds back on the activity of RhoGTPases, because their GEFs contain pleckstrin homology (PH) domains, which have high binding affinity for specific phosphoinositide lipids such as phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃], typically found at the leading edge of cells (Fig. 1D; Senju and Lappalainen, 2019). Active RhoGTPases such as Rac1, which forms clusters at the PM, can themselves also directly interact with PI(3,4,5)P₃, helping to shape and sustain its signaling gradient (Fig. 2, arrow 3; Remorino et al., 2017). Similarly, the distribution of phosphatidylserine has been shown to be polarized and required for optimal Cdc42 targeting and activation during cell division and mating of Saccharomyces cerevisiae, which requires vectorial delivery of phosphatidylserine to specific sites of the PM, and local exclusion by actin and myosin 1-mediated endocytosis (Fairn et al., 2011; Haupt and Minc, 2017; Sartorel et al., 2018). Thus, intracellular compartments can be regarded as dynamic carriers for non-selective and selective lipid transport.

Whereas membrane delivery through vesicular transport carriers has been considered as the traditional resource of lipids, there is increasing evidence that non-vesicular lipid transfer at membrane contact sites significantly contributes to lipid distribution (Balla et al., 2019; Stefan et al., 2017). Membrane contact sites are domains in which two membranes of different compartments are in close proximity (∼20 nm) but do not fuse. They function in the transport of lipids (including cholesterol, phosphatidylserine and ceramides, (Balla et al., 2019) and Ca²⁺ (Muallem et al., 2017). These tight contacts are present between different membranous organelles, but particularly the ER has been shown to have tight contact with all organelles and the PM (Balla et al., 2019). ER–PM contacts have recently been shown to restrict vesicular exocytosis in yeast (Fig. 3, point 6), which was required for polarized morphogenesis (Ng et al., 2018). At the same time, high Cdc42 activity promoted vesicular exocytosis at the growing zone and prohibited local ER–PM contact site formation. This spatially segregated negative feedback between membrane contact site formation and Cdc42 promotes robustness of the formation of confined exocytic domains for polarized cell morphogenesis in yeast (Ng et al., 2018). It will be interesting to see whether similar mechanisms exist in mammalian cells. However, membrane contact sites are hotspots of lipid transport and therefore could also enhance directed membrane flow to specific parts of the cells. This raises questions about a more-prominent role of membrane contact sites and particularly the ER in polarization.

Although directed transport is key to cell polarity establishment, the orientation of the endomembrane compass does not seem to be constant. For instance, whereas mesenchymal cells migrate with endomembranes facing the protruding front, lymphatic cells of the immune system migrate with rear-positioned endomembranes, which has led to a debate on the role of intracellular organization in polarity (Natividad et al., 2018; Orlofsky, 2019; Pouthas et al., 2008; Zhang and Wang, 2017). Importantly, organelles are not only involved in directional flux, but can also function as mobile platforms or units that locally activate signaling pathways (Phuyal and Farhan, 2019). In particular, endosomes have been shown to locally regulate signaling, giving rise to the concept of the ‘signaling endosome’ (Scita and Di Fiore, 2010; Villaseñor et al., 2016). A frequent signaling cascade is the activation of RhoGTPases at endosomes that drive actin remodeling events; for example, in the signaling downstream of Rab5A on early endosomes (Fig. 3, point 7), Rab5A spatially restricts Rac1 activation on polarized cell protrusions (Malinverno et al., 2017). It has been proposed that the strength of local feedback loops, such as the one between Rac1 and F-actin, shapes the patterns of signal transduction waves and therefore defines the architecture of the excitable network (Miao et al., 2017, 2019). A particular feature of organelles is that they have a variable amount of surface that is in contact with the cytosol (surface-to-cytosol ratio) and membrane curvature (Iversen et al., 2015). Interestingly, the surface-to-cytosol ratio has been shown to alter the probability of activation for PAR proteins and RhoGTPases in cells (Schmick and Bastiaens, 2014). Similarly, changes in membrane curvature can lead to the recruitment of curvature-sensitive actuators (e.g. I-BAR domain-containing proteins) that help to activate GTPases (Begemann et al., 2019). Thus, organelles could provide specific biophysical properties to locally enhance polarity signaling. Conversely, actin cytoskeleton remodeling downstream of RhoGTPases produces forces that deform membranes and induce local changes in membrane curvature (Lämmermann and Sixt, 2009) that could feedback on lipid distributions. Therefore, organelles could provide an alternative feedback mechanism for polarity maintenance through local actin remodeling (Fig. 2, arrow 4).

Another interesting mechanism for actin remodeling at organelles is the Ca²⁺-dependent activation of actin-interacting proteins, as revealed during dendritic cell (DC) migration (reviewed in Sáez et al., 2018), where localized release of Ca²⁺ from organelles, as a cofactor of calmodulin, induces the contractility of actin-dependent motors of the myosin family. For instance, it has been shown that fast and persistent motility in dendritic cells (DC) depends on Ca²⁺ release from the ER (Fig. 3, point 8) (Solanes et al., 2015). In addition to the ER, which is largely recognized for its role as Ca²⁺ storage compartment, other membranous organelles can play similar roles in localized Ca²⁺ release. Indeed, in mature DCs it has been shown that Ca²⁺ release via the transient receptor potential mucolipin 1 channel (TRPML1, also known as MCOLN1) from the rear-positioned lysosomes triggers local myosin IIA contractility for fast and directional migration of mature DCs (Bretou et al., 2017; Vargas et al., 2016). A similar mechanism seems to be at work in mammalian cells (Melchionda et al., 2016; Nguyen et al., 2017). It is interesting that Ca²⁺ transport takes place at intracellular membrane contact sites; therefore, increasing our knowledge on their cellular positioning might provide further insights into the mechanisms contributing to the regulation of organelle positioning.

In summary, intracellular organelles and their associated trafficking provide feedback for the maintenance of asymmetric cellular distributions of proteins, lipids as well as higher order networks, such as the actin cytoskeleton. Feedback is achieved through directed, vectorial delivery and selective local activation.

The many examples of interconnections between intracellular processes highlight the fact that cell polarity cannot be reduced to the function of a few selected molecular players; rather, cell polarity emerges as an outcome of the collective interactions of all its individual components. Cell polarity can be broken down by removing a specific critical element, but its establishment and maintenance requires the integrity of the whole molecular network. Evidence is increasing that the tight coupling between the molecular and architectural elements of the cell (Fig. 2), which happens on many different levels, is key for robust maintenance of cell polarity. Intracellular organelles and their associated trafficking routes show strong asymmetrical distributions and are indispensable elements for cell polarity. It is thus time to consider them as an equal part of the polarity network.

The advantages of a dynamic maintenance of polarity are robustness and adaptability (Freisinger et al., 2013). This provides great flexibility for regulation in response to many physiological conditions. Intracellular trafficking pathways show great plasticity and thus are perfectly suited to provide the cell with numerous possibilities to dynamically maintain polarity. The tight integration of lipids and proteins to defined organelles and their interconnection with the actin cytoskeleton allows the use of various feedback loops for polarity establishment in order to fine-tune cellular asymmetries for distinct functions. It can be tentatively speculated that the observed variability of endomembrane positioning during cell migration supports cellular asymmetries through alternative mechanisms: whereas organelles might sustain feedback loops based on vectorial transport of adhesion molecules when endomembranes are at the front during mesenchymal migration, they might play a more important role to sustain feedback loops based on actin cytoskeleton contractility when endomembranes are at the back during amoeboid migration. In this context, the positioning of endomembranes relative to cell boundaries or the nucleus (Natividad et al., 2018; Pouthas et al., 2008; Zhang and Wang, 2017) will be less important. Using the same machinery to dynamically maintain cell polarity for alternative functions is key to natural selection and evolution of living organisms.

In yeast, mathematical models based on reaction–diffusion systems combined with experimental evidence have started to investigate the quantitative contribution of separate feedback mechanisms (Freisinger et al., 2013; Slaughter et al., 2009). Similar studies are lagging behind in mammalian cells. A particular challenge is to measure the contribution of different trafficking pathways for polarity and to integrate them into quantitative models; it is, for instance, not known how much the recycling or secretion pathways contribute to localized Cdc42 activity in a given cell type. Indeed, models that include several trafficking routes have not yet been constructed, limiting our understanding of the role of distinct organelles in polarity establishment. Also, the fact that trafficking pathways show great plasticity complicates quantitative models. Another challenge is the difficulty to define a readout that can be used in different polarity models to quantitatively compare, for instance, polarity maintenance during migration and lumen formation. A major challenge in the understanding of cell polarity is the multiscale nature of collective interactions (Fig. 2), in which the GTPase cycle of one protein is intervened with the dynamics of macromolecules, such as the actin cytoskeleton or cellular organelle compartments. In the future, quantitative models that bridge different scales and levels of complexity need to be established to fully understand the emergence of polarity.

Our knowledge of the mechanisms underlying cell polarity is mostly based on classical approaches that apply long-term perturbation, such as depletion or downregulation of individual cell components. However, these methods have two main disadvantages. First, the intrinsic adaptability of the collective interactions complicates the interpretation of phenotypes. Often, these collective interactions are driven to an aberrant functioning point that does not inform us on the fine spatiotemporal regulation happening in the wild-type situation. Second, these perturbations are global and permanent, whereas cell polarity relies on the coupling of asymmetric and dynamic processes. Recent advances in molecular biology have provided the tools to measure and manipulate molecular pathways involved in polarity with unprecedented spatiotemporal precision. These include gene editing, which enables the study of proteins at their endogenous levels, and, in particular, the use of well-controlled minimal systems (see Box 1), which allow the manipulation of cell adhesion, as well as light-controlled protein activation through optogenetics (see Box 2). We believe that the particular tools presented in Boxes 1 and 2 could provide a basis to specifically dissect properties of the individual components and their collective interactions during polarity establishment and maintenance.

We conclude that cell polarity should be thought of as an emergent property from collective interactions of individual components rather than the consequence of the (de)activation of a particular molecular player. In addition to dissecting feedback mechanisms that support asymmetric cellular distributions in detail, emerging interdependencies need to be uncovered. Unsurprisingly, these are centered around the intracellular endomembrane compass and include the crosstalk between the actin cytoskeleton and lipid distributions (Fig. 2, arrow 5) as well as the role of lipid distribution in organelle biogenesis, morphology and positioning (Fig. 2, arrow 6). The challenge for future research is to identify and better quantify these multiscale collective interactions, and to understand what are the sufficient conditions for cell polarity to emerge.

We thank Bruno Goud for critical reading of the manuscript.