Division orientation depends on mitotic spindle positioning, that is generally attained in metaphase and sometime corrected in telophase (Morin and Bellaïche, 2011; Lough et al., 2019). In many epithelial systems, division orientation follows the Hertwig’s rule, according to which the spindle aligns along the long axis of the dividing cell (Hertwig, 1884). To which extent spindle alignment to the long cell axis is guided by mechano-sensing pathways responding to compressional cues exerted by neighbouring cells, or it is contributed by cytoskeletal forces exerted by MT motors is still debated. Elegant studies in MDCK (Madin–Darby Canine Kidney) Extra-Cellular-Matrix-free (ECM-free) monolayers “in suspension” showed that the division orientation occurs along the longest cell axis and is instructed by the interphase geometry (Wyatt et al., 2015). In these cells, components of the force generators complexes including NuMA and Gαi accumulates at cortical polar sites. Consistently, studies in Xenopus epithelia indicate that cells divide according to interphase cellular shape that is defined by three-cell junction distribution, where LGN and E-cadherin accumulates (Nestor-Bergmann et al., 2019). Collectively, this evidence suggests that in mammalian epithelial cells interphase shape drives force generators distributions to orchestrate divisions along the longest cell axis. Notably, these findings in vertebrate cells are consistent with previous observations in Drosophila tissues (Bosveld et al., 2016), although do not seem to apply to the development of Drosophila follicular epithelium at early-stage egg chambers (Finegan et al., 2019).

Several studies elucidated the molecular mechanisms of orientation, in which a fundamental role is played by Gαi/LGN/NuMA proteins, an evolutionarily conserved ternary complex. Gαi is the subunit of heterotrimeric G-proteins that localizes at the plasma membrane, LGN acts as a molecular scaffold, and NuMA is the mitotic dynein-adaptor involved in MT-pulling force onset. The majority of studies addressing the mechanistics of orientation were conducted in adherent cells in isolation, such as HeLa cells (Du et al., 2001; Du and Macara, 2004; Zheng et al., 2010; Kiyomitsu and Cheeseman, 2012; Kotak et al., 2012; Gallini et al., 2016; Pirovano et al., 2019; Takayanagi et al., 2019), or in a monolayers of MDCK cells, in which the spindle axis aligns parallel to the substratum in an integrin-dependent manner (Reinsch and Karsenti, 1994; Tuncay et al., 2015; Chishiki et al., 2017; Lázaro-Diéguez and Müsch, 2017).

In metaphase, the Gαi/LGN/NuMA complex localizes at the plasma membrane above the spindle poles (Du et al., 2001; Du and Macara, 2004; Kotak et al., 2012; Gallini et al., 2016; Pirovano et al., 2019; Zheng et al., 2010; Machicoane et al., 2014) and recruits the MT motor dynein/dynactin (Kotak et al., 2012; Okumura et al., 2018; Woodard et al., 2010) (Figure 1A). Exploiting the minus-end directed movement of dynein, cortically localized dynein motors generate pulling forces on astral MTs branching from the spindle poles that in metaphase contribute to spindle placement (Théry et al., 2007; Kiyomitsu and Cheeseman, 2012; Kotak et al., 2012). Notably, ectopic recruitment of NuMA to the cell cortex by optogenetic techniques is necessary and sufficient to orient the spindle, while cortical targeting of dynein is not sufficient to generate enough pulling forces to place the spindle (Fielmich et al., 2018; Okumura et al., 2018), implying that the activity of MT motors requires a defined spatial cortical organization. In line with these findings, recent studies revealed that not only the levels of NuMA/dynein/dynactin motors present at the cortex, but also their spatial distribution plays a role in the onset of effective MT-pulling forces (Pirovano et al., 2019; Renna et al., 2020).

In mitosis Gαi proteins are uniformly enriched at the cell cortex, but only a GDP-loaded pool of Gαi (Gαi^GDP) accumulates above the spindle poles and is the one that selectively binds to LGN (Du et al., 2001; Willard et al., 2004; Mochizuki et al., 1996). The recruitment of LGN at the cortex by Gαi is controlled by GAPs (GTPase activating proteins) and GEFs (Guanine Exchange Factors) that tune the GTP-state of Gαi. An important Gαi GEF implicated in spindle placement is Ric-8A, which appears to play a key role in targeting LGN to the cortex (Chishiki et al., 2017; Woodard et al., 2010). In metaphase, LGN is spatially restricted to the cortical side facing the spindle poles by direct binding to Gαi (Zheng et al., 2010) (Figure 1A). Lateral recruitment of LGN and in turn NuMA/dynein motors promotes planar spindle orientation (Zheng et al., 2010). Notably, the conformation of LGN depends on its binding partners (Du and Macara, 2004; Pan et al., 2013): in the unliganded form LGN is kept in an inhibited conformation by intra-molecular interactions between the N-terminal TPR domain and the C-terminal GoLoco region. Cooperative binding of the four GoLoco motifs to cortical Gαi^GDP molecules recruits LGN to the cortex and induces a conformational change releasing the TPR domain that in turn associates with NuMA. These events result in the enrichment for NuMA/dynein/dynactin motors to specific cortical sites and onset of astral MT-pulling forces (Du and Macara, 2004; Pan et al., 2013) (Figure 1A). Notably, the TPR domain of LGN interacts not only with NuMA but also with Afadin and E-cadherin in a mutually exclusive manner, with functional implications that will be discussed below. NuMA shares the domain structures with other dynein-activator proteins (Kiyomitsu and Boerner, 2021), including a hook domain and a CC1-like box motifs, both at the N-terminus, responsible for the binding to dynein and dynactin. After a central 1500-residue long coiled-coil, NuMA codes for a C-terminal region binding to LGN, microtubules, as well as to the plasma membrane in anaphase (Kotak et al., 2013; Seldin et al., 2013; Carminati et al., 2016; Seldin et al., 2016; Pirovano et al., 2019; Renna et al., 2020). All these diverse functionalities of NuMA C-terminus contribute to spindle placement during mitotic progression, partly modulated by mitotic kinases’ phosphorylation (Lechler and Mapelli, 2021).

In 2D systems such as polarized MDCK cells grown in monolayer, cells divide with the spindle axis aligned to the substratum by planar symmetric divisions that generate two daughter cells remaining in the same monolayer (Reinsch and Karsenti, 1994; Tuncay et al., 2015; Lázaro-Diéguez and Müsch, 2017). In this setting, spindle alignment is maintained by astral MTs captured by cortical cues localized at the lateral domains of the dividing cell, including cell-cell adhesion molecules (Gloerich et al., 2017; Lázaro-Diéguez and Müsch, 2017). Additional information has been obtained in more physiologically relevant 3D models, such as cysts.

The most common cells used to study oriented divisions in cysts are MDCK and Caco-2 (human colon adenocarcinoma) cells that, when plated on a substrate that mimics the ECM such as matrigel, grow as monolayered spheres by planar divisions occurring with the spindle axis perpendicular to the apico-basal polarity (Zegers et al., 2003; Jaffe et al., 2008). A cyst is characterized by a central lumen and a surrounding monolayer of polarized cells (Zegers et al., 2003) (Figure 1B). Notably, lumen formation in MDCK- or Caco-2-derived cysts relies on spindle orientation, as opposed to cysts obtained from MCF10A cells (human breast immortalized cells) where lumen forms by anoikis, i. e apoptosis of inner cells after a full sphere is formed (Debnath et al., 2002). After the first division, MDCK single cells have been shown to form an apical membrane initiation site (AMIS) between the two daughter cells, in the position where the midbody was located (Overeem et al., 2015), that will later become the lumen of the nascent cyst (Rodriguez-Boulan and Macara, 2014) (Figure 1B left). Cells composing the mature cyst have two types of domains: the apical side facing the central lumen where the PAR (partitioning defective) family proteins localize, and the baso-lateral domain where adhesion proteins such as integrins are in contact with the ECM, and where adjacent cells are in contact with each other by adherens (AJ) and tight junctions (TJ) (McCaffrey and Macara, 2011) (Figure 1B right). Importantly, each of these membrane domains is key for the localization of spindle orientation proteins instructing planar divisions including Gαi, NuMA and LGN (see below) (Overeem et al., 2015; Nakajima, 2018). In Table 1 we summarized the proteins involved in spindle orientation with their function, localization and defects occurring upon depletion.

One of the first proteins to be implicated in planar divisions in cysts was the GTPase Cdc42, whose depletion in Caco-2 cells results in multi-lumen cysts due to spindle misorientation (Jaffe et al., 2008). In MDCK cells, Cdc42 has been shown to be activated by the two GEFs Tuba, regulating cell-cell junctions and Cdc42 apical localization (Qin et al., 2010; Otani et al., 2006), and Intersectin-2, implicated in endocytosis and in the mitotic Cdc42 targeting at centrosomes (Rodriguez-Fraticelli et al., 2010; Okamoto et al., 1999; Hussain et al., 2001). Planar spindle orientation is also mediated by the apically-localized polarity complex composed by Par3, Par6 and the kinase aPKC (Figure 1B). Several studies in 3D systems have shown that depletion of Par3 leads to mislocalization of the kinase aPKC (Hao et al., 2010; Zheng et al., 2010; Durgan et al., 2011; Vorhagen and Niessen, 2014), which phosphorylates LGN on Ser401 to exclude it from the apical side ensuring its localization at the lateral cortex, possibly by direct association with the baso-lateral protein Dlg-1 (Saadaoui et al., 2014). An intriguing role has been described for the Par1b/MARK2 kinase that in MDCK cells monolayer with high Rho activity promotes LGN/NuMA recruitment at the lateral site and planar divisions with the spindle axis aligned to the substratum. Conversely, in hepatocytes, that in addition to apico-basal polarity also organize a lateral lumen for the development of bile canalicular networks and have reduced Rho activity, Par1b prevents NuMA/LGN lateral recruitment causing tilted spindles and asymmetric partitioning of the lateral lumen among daughter cells (Lázaro-Diéguez et al., 2013; Slim et al., 2013).

In addition to their cohesive role, also some junctional proteins have been shown to be involved in spindle orientation in cysts, including the Junctional adhesion molecule-A (JAM-A), Afadin (AF6), E-Cadherin and Dlg-1 (Discs large homolog 1) (Figure 1B). In MDCK cysts, JAM-A activates Cdc42 and PI(3)K (Phosphatidylinositol 3-kinases), generating a gradient of PtdIns(3,4,5)P3 enriched at the cortex area facing the spindle poles, which is required for correct localization of dynein/dynactin and for spindle orientation (Toyoshima et al., 2007; Tuncay et al., 2015). Consistently, JAM-A was shown to activate Cdc42 also in progenitors of the developing cerebral cortex this way contributing to spindle orientation (Fededa et al., 2016).

The actin-binding protein Afadin, localized at adherent junctions, mediates planar spindle orientation in Caco-2 cyst by recruiting LGN to the lateral cortex via direct interactions with the LGN-TPR domain (Carminati et al., 2016; Gao et al., 2017; Bonucci et al., 2020) (Figure 1B). Consistently, studies conducted in MDCK cysts (Gao et al., 2017), hepatocyte cells (Bonucci et al., 2020), and murine neuro glia (Rakotomamonjy et al., 2017) show that Afadin is crucial for spindle orientation as its depletion leads to an aberrant spindle placement (Lough et al., 2019; Carminati et al., 2016; Gao et al., 2017; Bonucci et al., 2020; Rakotomamonjy et al., 2017). In MDCK cysts, planar cell divisions also rely on the interaction between the intra-cellular domain of E-cadherin and LGN-TPR domain (Gloerich et al., 2017; Hart et al., 2017) (Figure 1B). As with Afadin, NuMA competes also with E-cadherin for LGN binding (Zhu et al., 2011; Carminati et al., 2016; Gloerich et al., 2017; Hart et al., 2017). This suggests that Afadin and E-cadherin might be needed for the initial LGN targeting at the cortex, when NuMA is still in the nucleus, and that these interactions dissociate later in mitosis. An alternative explanation envisions that the cortical Gαi^GDP-bound pool of LGN cycles between different mitotic binding partners associating with its TPR domain, including NuMA, Afadin and E-cadherin, in order to coordinate mechano-sensing junctional cues with spindle orientation and mitotic progression. Future live-imaging studies will clarify whether this is the case.

In addition to this role, E-cadherin was shown to be important for maintenance of cell polarity and spindle orientation in prostate epithelia by interacting with LGN, NuMA and Scrib at the lateral sites of mitotic cells, this way preserving correct apico-basal polarity, planar cell divisions and tissue integrity. Consistently, conditional loss of E-cadherin during murine prostate development leads to disorganized epithelia observed in early state prostate tumorigensis (Wang et al., 2018). Spindle orientation functions have been reported also for the baso-lateral polarity protein Dlg-1, that belongs to the membrane-associated guanylate kinase (MAGUK) family and is required for adherens junction formation and maintenance (Su et al., 2012) (Figure 1B). In HeLa cells, in MDCK cysts and in the chick neuroepithelium, Dlg-1 promotes spindle orientation by binding to the phosphorylated LGN protein (Saadaoui et al., 2014; Saadaoui et al., 2017; Porter et al., 2019), fully in line with was previously shown in Drosophila epithelial systems (Morin and Bellaïche, 2011; Pietro et al., 2016). In turn, the correct localization of Dlg-1 is influenced by other factors including Gαi (Saadaoui et al., 2014) and the tumor suppressor protein CASK (calcium/calmodulin-dependent serine protein kinase) (Porter et al., 2019). The binding of Dlg-1 to CASK and Gαi is key to direct LGN to restricted cortical regions before metaphase, and ultimately to target LGN and NuMA-dynein appropriately (Saadaoui et al., 2014; Saadaoui et al., 2017; Porter et al., 2019).

Another polarity protein affecting LGN cortical recruitment is the suppressor APC domain containing 2 (SAPCD2), that has been shown to interact with Gαi/LGN complexes to orchestrate mitotic spindle orientation in MDCK cyst and in mouse retina (Chiu et al., 2016). Specifically, SAPCD2 binding to the close conformation of LGN restricts LGN/NuMA accumulation at the lateral site providing a mechanism to balance the proportion of planar and vertical divisions, and hence the symmetric or asymmetric outcome of retinal progenitor mitosis (Chiu et al., 2016).

We already reported the relevance of the Gαi GEF Ric-8A for spindle orientation in HeLa cells (Woodard et al., 2010). Recent work highlighted a role for Ric-8A in tight junction formation in MDCK cysts and in LGN recruitment to the lateral cortex by generation of a localized Gαi-GDP pool promoting planar cell divisions (Chishiki et al., 2017) (Figure 1B).

Beside junctional and polarity proteins, the actin cytoskeleton, as well as actin and microtubule-binding proteins, contribute actively to spindle orientation (Pietro et al., 2016), as described in invertebrate systems such as Drosophila neuroblasts (Kunda and Baum, 2009) and HeLa cells (Pietro et al., 2016; Rizzelli et al., 2020). However, the role of actin in planar division and cystogenesis is less clear. In MDCK cysts, the microtubule-associated protein IQGAP1, localized at the basal site, participates to MTs dynamics and promotes planar spindle orientation by interacting with the MT plus-ends and by targeting NuMA laterally (Bañón-Rodríguez et al., 2014) (Figure 1B). Notably, in HeLa cells the interaction between IQGAP1 and Cdc42 has been shown to allow the binding of Cdc42 to the actin-binding protein MISP (Mitotic Interactor and Substrate of PLK1) implicated in spindle positioning (Zhu et al., 2013; Cadart et al., 2014; Vodicska et al., 2018). MISP associates to members of the ERM (Ezrin, Radixin and Moesin) protein family, that connects the mitotic acto-myosin cortex to the plasma membrane, in this way assisting the correct localization of NuMA at the cortex for correct spindle positioning (Hebert et al., 2012; Zhu et al., 2013; Machicoane et al., 2014; Kschonsak and Hoffmann, 2018) (Figure 1B).

Studies of oriented divisions in cysts provided great insights into the crosstalk between orientation pathways and epithelial polarity. However, cysts of immortalized cell lines do not entirely recapitulate the cell diversity and the signaling response of epithelial tissues in vivo (Lancaster and Knoblich, 2014; Clevers and Tuveson, 2019).

Tissue organoids, especially murine intestinal organoids, are becoming a relevant model to study division orientation in a more physiological setting. Organoids are model systems that recapitulate not only the morphology of the organ but also the cellular composition, from stem cells to differentiated lineages (Clevers, 2013; Sato and Clevers, 2013). Methods to grow, manipulate genetically and image intestinal organoids have been first established in the Clevers lab (Sato and Clevers, 2013; Sato et al., 2009), whose work revealed that the organoids grown from intestinal epithelial cells form crypt and villi-like domains mirroring the morphology of the intestinal epithelium, with an analogous composition and distribution of cell types (Sato and Clevers, 2013; Sato et al., 2009) (Figure 1C). These studies revealed that in intestinal organoids the proliferating cells reside at the bottom of the crypt, close to the stem cell niche compartment constituted by non-dividing Paneth cells, that generate a Wnt3 gradient decreasing along the crypt axis (Figure 1C). Intestinal stem cells (ISCs) divide symmetrically moving toward the apical side of the monolayer that faces the organoid lumen, with the metaphase plate perpendicular to the apical side (Figure 1C). These ISC apical mitosis retain a connection to the basal site, and hence to the ECM, through an actin cable (Carroll et al., 2017; McKinley et al., 2018) (Figure 1C) that is essential for daughter cells to move back to the basal side of the monolayer upon cytokinesis (Carroll et al., 2017). As a matter of fact, the use of intestinal organoids to study oriented division is still in its infancy, contributed mainly by descriptive imaging experiments and a few mechanistical studies investigating the molecular mechanisms of mitosis. Little is known on molecules executing oriented divisions in organoid, but it is plausible that the same set of polarity and junctional proteins important for correct cystogenesis is implicated in division orientation also in these systems, with molecular details that remain to be explored.

Ablation of Dlg-1 from the murine intestinal crypts has been shown to result in misoriented divisions of the intestinal stem cells with a consequent delay in cell migration from the crypts bottom to the villi that promotes tumorigenic events (Young et al., 2019). Similarly, depletion from the murine crypts of the protein Tacc3, which is involved in MT crosslinking and stabilization of the Aurora-A dependent kinetochore-microtubules attachment (LeRoy et al., 2007; Burgess et al., 2015), blocks proliferation (Yao et al., 2016). Interestingly, knock-out of Tacc3 from intestinal organoids derived from APC (Adenomatous polyposis coli) mutated mice, models for colorectal cancer (Merenda et al., 2020), increases chromosome misalignment and hypomorphic mitotic spindles, leading to prolonged mitosis or mitotic arrest (Yao et al., 2016), to a certain extent mimicking what observed in vivo. Both findings open the possibility to target specific mitotic spindle proteins for chemotherapeutic therapy. In conclusion, although organoids hold the potential to allow more insightful analyses on the orientation pathways and their relevance for morphogenesis and disease, more studies are required to elucidate the molecular mechanisms accounting for oriented divisions in these systems.

As discussed, oriented divisions are important for the regulation of epithelial morphogenesis and homeostasis. Consistently, their deregulation has been associated to several pathological conditions such as cancer, microcephaly, and developmental defects (Gillies and Cabernard, 2011; Nakajima, 2018; Lechler and Mapelli, 2021). However, not always the causal relationship between misorientation and diseases is clear. In vivo studies revealed that spindle misorientation is oftentimes corrected or is embryonic lethal (Nakajima, 2018; Lechler and Mapelli, 2021).

In murine hepatic epithelial cells in vivo, spindle misorientation leads to detachment of epithelial sheets from nephron epithelial tubules (Gao et al., 2017). Similarly, in stem cell systems, misorientation alters the balance between symmetric and asymmetric divisions resulting in defective changes in architecture and functioning. This has been documented for neuroepithelial progenitors during murine cortical development, in which misorientation leads to the expansion of the radial glial compartment with a delay in neurogenesis (Fededa et al., 2016).

Tissues have developed different mechanisms to rescue the damage that a misoriented spindle can cause, that have been first discovered in Drosophila and still await to be confirmed in mammalian tissues. The first mechanism impinges on the ability of epithelial tissue to reintegrate cells that after misoriented cytokinesis are misplaced above the epithelial layer (Bergstralh et al., 2015; Lough et al., 2019). As described for intestinal organoids (Carroll et al., 2017), in Drosophila imaginal disc the dividing cells have an actin protrusion that keeps them in connection to the basal side of the monolayer and assists the appropriate repositioning of daughters after cytokinesis (Nakajima et al., 2013). Parallel studies showed that also adhesive molecules, such as Fasciculin-2/3 and neuroglian, play a role in reintegrating in the epithelial layer the cells misplaced above the follicular epithelium due to orientation defects (Bergstralh et al., 2015; Cammarota et al., 2020). In Drosophila imaginal discs, evidence was provided that upon misorientation, one of the two daughter cells loses connection with the basal side and is displaced in the lumen (Nakajima et al., 2013). In the absence of re-integration, the misplaced cells can encounter two different fates: it either remains in the wrong position, where proliferation causes morphological defects (Dekanty et al., 2012; Nakajima et al., 2013; Poulton et al., 2014), or it undergoes apoptosis due to lack of survival signals (Nakajima et al., 2013; Poulton et al., 2014). Whether any of these mechanisms for misorientation correction is in place in vertebrate epithelial tissues remains an interesting open question.