Eph signaling controls mitotic spindle orientation and cell proliferation in neuroepithelial cells

A precise regulation of mitotic spindle orientation is crucial during development and adult tissue homeostasis. It determines cell fate specification and tissue architecture in the context of asymmetric and symmetric cell division, respectively (Morin and Bellaïche, 2011; Lu and Johnston, 2013; Williams and Fuchs, 2013). The positioning of the spindle during cell division involves autonomous and nonautonomous mechanisms. However, while the intrinsic factors that control spindle orientation have been extensively studied over the past decades, our knowledge about the extrinsic signals that modulate this process and their link with the intrinsic spindle orientation machinery remains limited (Werts and Goldstein, 2011; Williams and Fuchs, 2013).

Regarding the autonomous mechanisms, intrinsic polarity cues linked to the cell cortex converge on astral microtubule-associated motor complexes, these exerting the pulling forces that orientate the spindle (Williams and Fuchs, 2013). The core components of the spindle orientation machinery are few and well conserved, differing slightly depending on the cell type and the mode of cell division (Morin and Bellaïche, 2011). For example, in asymmetrically dividing Drosophila melanogaster neuroblasts (NBs), the apical proteins Par-6, Par-3 (Bazooka, Baz, in Drosophila), and atypical PKC (aPKC) associate with Inscuteable (Insc). The Par–Insc complex recruits Partner of Insc (Pins)/LGN to the apical pole, where it binds the Gαi subunit anchored to the membrane, orchestrating the orientation of the spindle along the apico–basal axis of cell polarity (Lu and Johnston, 2013). This process requires the PDZ protein Canoe (Cno)/Afadin, which after being phosphorylated by the Warts/LATS1-2 kinase, binds to the Pins/LGN^TPRs domain and displaces Insc bound to the same region (Yu et al., 2000; Wee et al., 2011; Keder et al., 2015). Cno/Afadin then facilitates the recruitment of Discs Large1 (Dlg1; Keder et al., 2015) to the Pins/LGN^Linker domain (Johnston et al., 2009) and of Mushroom body defect (Mud)/NuMA, which also binds to the Pins/LGN^TPRs domain, displacing Cno/Afadin (Bowman et al., 2006; Izumi et al., 2006; Siller et al., 2006; Speicher et al., 2008; Keder et al., 2015). In turn, Dlg1 and Mud/NuMA associate with Khc-73 and the Dynein–Dynactin complex, respectively, which bind to astral microtubules anchoring the spindle and promoting the pulling forces required for its correct orientation (Johnston et al., 2009). In symmetrically dividing epithelial cells, the spindle is oriented parallel to the plane of the tissue (planar division). In this process, the intrinsic regulators Pins/LGN, Gαi, and the Mud/NuMA–Dynein complex are also found at the core of the spindle orientation machinery (Morin and Bellaïche, 2011; Bergstralh et al., 2013a). However, it was recently shown that Pins/LGN is not required for the planar orientation of the spindle in epithelial cells of Drosophila wing discs, and that only the Mud/NuMA–Dynein–Dynactin complex is critical for this process (Bergstralh et al., 2016).

Regarding the nonautonomous mechanisms, extrinsic mechanical cues have long been implicated in coordinating spindle orientation (Hertwig, 1884; Morin and Bellaïche, 2011; Nestor-Bergmann et al., 2014). In this context, the actomyosin network is an important link between the external forces and mitotic spindle positioning (Severson and Bowerman, 2003; Goulding et al., 2007). The Wnt-activated planar cell polarity pathway has been widely implicated in spindle orientation in different systems (Gong et al., 2004; Saburi et al., 2008; Castanon and González-Gaitán, 2011). This pathway impinges directly on the spindle orientation machinery by interacting with the conserved intrinsic spindle regulator Mud/NuMA, in both Drosophila and zebrafish (Ségalen et al., 2010). Other extrinsic cues also affect spindle orientation in vertebrates, such as Cadherin-mediated intercellular signaling or the FGF/Ras/ERK signaling pathway, although the downstream mechanisms that directly link them to the spindle orientation machinery are poorly understood (den Elzen et al., 2009; Castanon and González-Gaitán, 2011; Tang et al., 2011; Žigman et al., 2011). In this regard, a direct interaction between E-cadherin and LGN (Pins in Drosophila) has recently been identified in cultured cells (Gloerich et al., 2017). Likewise, a few years ago Semaphorin-Plexin signaling was identified as a novel extrinsic regulator of spindle orientation during kidney development and repair, as well as in the mouse spinal cord. This signaling controls the activity of Cdc42, a known regulator of spindle alignment (Arbeille et al., 2015; Xia et al., 2015).

Like Semaphorin-Plexin, there has been much interest in the role of Ephrin–Eph signaling during axon guidance (Cramer and Miko, 2016; Kania and Klein, 2016). However, this pathway also regulates different developmental processes, as well as adult tissue homeostasis and carcinogenesis (Batlle and Wilkinson, 2012; Lisabeth et al., 2013; Wilkinson, 2014; Jungas et al., 2016; Kania and Klein, 2016). A large family of ∼14 Eph receptors and 8 Ephrin ligands has been identified in vertebrates, whereas Drosophila has only 1 Eph receptor and 1 Ephrin ligand (Scully et al., 1999; Bossing and Brand, 2002). Both Eph tyrosine kinase receptors and their Ephrin ligands are membrane-bound proteins triggering cell–cell contact–mediated signaling, either through the receptor (forward signaling) or the ligand (reverse signaling; Lisabeth et al., 2013; Kania and Klein, 2016). This reverse signaling through the Ephrin intracellular domain can affect cell junctions, cell–cell adhesion, and ultimately tissue architecture (Jones et al., 1998; Chong et al., 2000; Lee et al., 2008; Lee and Daar, 2009; Arvanitis et al., 2013). Ephrin reverse signaling has also been shown to regulate the balance between proliferation and differentiation in the neural progenitor cells of the mammalian cerebral cortex, favoring the maintenance of the progenitors in detriment to their differentiation (Qiu et al., 2008). Ephrin B1–dependent forward EphA4 signaling has also been implicated in promoting progenitor proliferation in the developing cerebral cortex (North et al., 2009). However, a role for EphA receptors in inducing the differentiation of mammalian neural progenitor cells in vitro and in vivo has also been proposed (Aoki et al., 2004; Laussu et al., 2014).

Here, we uncover a novel function for Ephrin–Eph intercellular signaling as a new extrinsic cue controlling mitotic spindle orientation in the symmetrically dividing neuroepithelial cells in the Drosophila optic lobe. This function relies on aPKC activity–dependent myosin II regulation, which influences the architecture of the neuroepithelium (NE) and the cortical distribution of core components of the spindle orientation machinery. In addition, we identify a requirement for Ephrin–Eph signaling to regulate proliferation in this NE through the Rho signaling–dependent inhibition of the phosphoinositide-3-kinase (PI3K)–Akt pathway.

Having initially detected a widespread, punctate distribution of the Eph receptor in the Drosophila optic lobe NE (Fig. 1, A and B; and Fig. S1), we were interested to analyze a potential function of Eph signaling in the development of this tissue. Given that polarity is an intrinsic characteristic of epithelial cells, we started by analyzing the distribution of well-conserved epithelial polarity cues in the optic lobe NE of Eph^x652-null homozygote mutants at the late third instar larval (L3) stage. Specifically, we tested the apical marker Patj; the markers of the subapical region (equivalent to the tight junction region in vertebrates) Par-6 and aPKC; adherens junctions (AJs) proteins such as E-cadherin (Shotgun, Shg, in Drosophila), Par-3 (Baz in Drosophila), and Cno/Afadin; the basolateral proteins Dlg1 and Scribble (Scrib); and Myospheroid (Mys), the β subunit of the integrin dimer that accumulates in the basal domain of epithelial cells (Fig. 1, C–Q). Significantly, we detected an anomalous apical enrichment of Dlg1 in the neuroepithelial cells of Eph^x652 mutants (Fig. 1, C–D′), while the general apico-basal distribution of the rest of the proteins tested did not appear to be altered. Nevertheless, the local distribution and thresholds of these proteins revealed a disturbed NE architecture. For example, less Shg/E-cadherin and Cno/Afadin accumulated at AJs in Eph^x652 mutants (Fig. 1, K, K″, L, and L″–N), and the apical localization of Patj and aPKC, as well as the basal distribution of Mys/β integrin subunit, were less uniform than in control neuroepithelia (Fig. 1, E–H′ and O–P′). An apical disruption of the NE in Eph^x652 mutants was also very apparent in some cases (Fig. 1, I–J′ and K–L″). Similar phenotypes, or milder in the case of Dlg1 apical expansion, were observed in other Eph mutants, such as a kinase dead mutant, as well as following Ephrin^RNAi down-regulation (Fig. S2). We also noticed that while the cell nuclei lie at different positions in control neuroepithelia, in a pseudostratified organization, at this late L3 stage the cell nuclei tended to align at the same level in Eph^x652 neuroepithelia (simple organization). Consequently, mutant neuroepithelia generally looked narrower than those of controls (Fig. 1, R–T). Hence, Eph signaling is necessary to maintain the correct architecture of the NE.

Given that the cortical distribution of actomyosin helps model cell shape and size, we assessed the distribution of both actin and the regulatory light chain of the nonmuscle myosin II (Spaghetti squash [Sqh] in Drosophila) in Eph neuroepithelia. At the early L3 stage, the distribution of both actin and myosin II (Sqh) was already compromised in Eph mutants, at least in part explaining the defects in the architecture of the NE detected at late L3 on this mutant background (Fig. 2, A–B″). Intriguingly, we observed that the subapical region, well defined between the AJs and the most apical cell region in control neuroepithelial cells, was collapsed in Eph mutant neuroepithelia, and Shg (E-cadherin)⁺ AJs were in close contact or intermingled with the apical region (Fig. 2, A–B″). In addition, aPKC, normally localized to the subapical region, was completely apical in Eph mutant neuroepithelia (Fig. 2, C–D″). Finally, to determine the involvement of myosin II (Sqh) in these Eph mutant phenotypes, we looked at a phosphomutant form of Sqh (Sqh^A20A21, hereafter Sqh^AA; Jordan and Karess, 1997) at the early L3 stage. Again, some disruption of the subapical region was detected in neuroepithelial cells, although aPKC in myosin II mutant (Sqh^AA) neuroepithelia was not so apically expanded as in Eph mutants (Fig. 2, E–E″). From these observations, we conclude that Eph signaling is required to establish the correct optic lobe NE architecture, at least partially through myosin II regulation.

An intact actomyosin cell cortex is emerging as an important requirement for a correct mitotic spindle orientation (Luxenburg et al., 2011; Nakajima et al., 2013; Zheng et al., 2013; Lancaster and Baum, 2014; di Pietro et al., 2016). Given that we found alterations in both the actin and myosin II (Sqh) distribution in Eph optic lobe neuroepithelia (Fig. 2, A–B″), we analyzed the mitotic spindle orientation in Eph mutants. We performed this analysis at the early L3 stage, a moment at which cells divide parallel to the plane of the NE in WT optic lobes (Rujano et al., 2013). While in control neuroepithelia most spindles were positioned in parallel (0°–30° angle), in Eph^x652 homozygotes most of them were oriented perpendicular or at an oblique angle (>30°) to the plane of the NE (Fig. 3, A and B). Similar phenotypes to that of Eph^x652 mutants were found in neuroepithelia specifically expressing either Eph^RNAi or Ephrin^RNAi, as well as in Eph^KD (kinase dead) mutant neuroepithelia (Fig. 3, C–E). We also observed defects in the position of the nucleus in neuroepithelial cells during mitosis at this early L3 stage. As mentioned above, WT optic lobe neuroepithelial cells arrange their nuclei at different positions along the apico–basal axis. This is due to the INM (Kosodo, 2012; Rujano et al., 2013) that shifts cells closer to the apical surface to divide (Rujano et al., 2013; Fig. 3 F). In early L3 Eph and Ephrin^RNAi mutants, we observed mitotic cells at random positions along the apico–basal cell axis (Fig. 3, B, C, E, G, and H). Spindle orientation failed more often in mitotic cells at basal positions than at apical positions in Eph^x652 mutant neuroepithelia, while in controls there was a similar percentage of defects in apical and basal cells (Fig. 3 I). This result suggests that the basal position of mitotic cells contributes to the effects on spindle orientation failures in Eph^x652 mutants. However, mitotic cell position is not the only driver of the phenotype observed, as correctly located apical cells in mutant neuroepithelia still showed a higher proportion of spindle defects than control apical cells (Fig. 3 I). Thus, Eph regulates mitotic spindle orientation and INM in the optic lobe NE.

To examine how Eph regulates the orientation of the mitotic spindle and given the altered myosin II distribution in Eph mutant neuroepithelial cells (Fig. 2), we first analyzed the phenotype of the phosphomutant form of myosin II (Sqh^AA; Jordan and Karess, 1997), identifying mitotic spindle alterations in the optic lobe neuroepithelial cells similar to those found in Eph mutants (Fig. 4 A). In mammals, both Eph receptors and ephrin ligands can activate the RhoA small GTPase, which in turn induces the activity of its downstream effector RhoA-associated kinase (ROCK; Wahl et al., 2000; Noren and Pasquale, 2004; Lisabeth et al., 2013). Activated ROCKs phosphorylate the myosin II light chain in both mammals and Drosophila, inducing myosin–actin interactions and therefore myosin ATPase activity (Amano et al., 1996; Winter et al., 2001; Fig. 4 B). The myosin II (Sqh) defects we observed in Eph mutant neuroepithelia at the early L3 stage (Fig. 2, A–B″) were maintained at late L3 (Fig. 4, C–D′″). Hence, we assessed whether Eph acts through a ROCK to regulate myosin II (Sqh) in our system, finding clear alterations in myosin II (Sqh) localization in both Rho1 (Drosophila RhoA) and Rok (Drosophila ROCK) mutants (Fig. 4, E–G′″). We reasoned that a phosphomimetic form of the myosin II (Sqh^E20E21, hereafter Sqh^EE; Winter et al., 2001) that bypasses Eph/Rok-dependent phosphorylation would be correctly located in Eph mutants. However, strong defects in myosin II Sqh^EE localization were still observed on an Eph mutant background, the same phenotype found when Ephrin was specifically down-regulated in the optic lobe NE (Fig. 4, H–J′″). Thus, Rok is not sufficient to activate myosin II in the absence of Eph.

In addition to Eph/Rho kinase–dependent phosphorylation, aPKC activity is essential to make myosin fully functional in some contexts (Wang and Riechmann, 2007; Kishikawa et al., 2008). Thus, we analyzed the distribution of active myosin II (Sqh^EE) on an aPKC^CAAXDN (kinase dead) mutant background, again finding clear alterations in its distribution (Fig. 5, B–C′″). This result suggested that Eph signaling in the optic lobe NE might act through aPKC, which in turn impinges on myosin functionality (Fig. 5 A). To test this possibility, we expressed a constitutively activated form of aPKC (aPKC^ΔN) in Eph mutant neuroepithelia, rescuing the apical distribution of active myosin II (Sqh^EE; Fig. 5, D–E′″). Moreover, the defects in the subapical region observed in Eph mutants at early stages (Fig. 2, C–D″) were also rescued by aPKC^ΔN on this active myosin II (Sqh^EE) genetic background (Fig. S3). Thus, Eph is required for aPKC activation in the optic lobe NE, an activity that is essential for myosin II (Sqh) full functionality. We next checked whether the inactivation of aPKC in Eph mutants, and the consequent impairment of myosin II functionality, was responsible for the failures in mitotic spindle orientation observed in Eph mutants. The expression of the kinase-dead form of aPKC (aPKC^CAAXDN) in neuroepithelial cells provoked mitotic spindle alterations similar to those found in Eph and myosin II (Sqh^AA) mutants (Fig. 5 F). In addition, aPKC^CAAXDN mutants also showed defects in the INM process (Fig. 5 G). Moreover, both the mitotic spindle misorientation and the INM phenotypes of Eph mutants were rescued by aPKC^ΔN (Fig. 5, H and I). Hence, Eph controls mitotic spindle orientation and INM through aPKC activity–mediated myosin II regulation.

The activity of aPKC was impaired in Eph mutants, in turn affecting the distribution of myosin II (Sqh). Given that aPKC is involved in linking cortical polarity with the machinery required for mitotic spindle orientation in some epithelia (Hao et al., 2010; Durgan et al., 2011; Guilgur et al., 2012), we wondered whether the localization of these spindle regulators would be affected at mitosis in Eph mutants. We first examined Dlg1, which is required for spindle orientation in different cell types (Bellaïche et al., 2001; Johnston et al., 2009; Bergstralh et al., 2013b) and whose distribution was drastically altered in the neuroepithelial cells of the Eph^x652 optic lobes at late L3 (Fig. 1, C–D′). While Dlg1 always presented a basolateral distribution at metaphase in control mitotic cells (n = 11) at early L3, in Eph^x652 mitotic mutant cells Dlg1 was expanded apically in six of nine metaphases analyzed (Fig. 6, A–B′). Moreover, Dlg1 distribution was similarly affected in the neuroepithelial cells of myosin II (Sqh^AA) mitotic mutants (five cells with defects of eight metaphases analyzed; Fig. 6, C and C′). In addition, constitutively active aPKC (aPKC^ΔN) on an active myosin (Sqh^EE) and Eph mutant background rescued to a great extent the normal Dlg1 distribution. Neither aPKC^ΔN nor Sqh^EE alone rescued the Dlg1 phenotype in Eph mutants, further supporting the relevance of both aPKC and Rok downstream of Eph to fully activate myosin II (Sqh; Fig. S4). To directly test the role of Dlg1 in the orientation of the mitotic spindle in the optic lobe NE, we analyzed dlg1¹⁸ homozygotes, detecting significant spindle orientation defects at early L3 (Fig. 6 D).

We also wanted to analyze the localization of Mud/NuMA in Eph mutants, one of the most conserved spindle orientation regulators in different contexts and species (di Pietro et al., 2016). In control neuroepithelial cells, Mud was located at lateral membranes at pro-metaphase, and it was strongly enriched at centrosomes (Fig. 6, E–E′″). No apparent differences in this Mud enrichment at centrosomes were observed in Eph^x652 or myosin II (Sqh^AA) mutant mitotic neuroepithelial cells at this stage, although there was a significant loss of Mud at the membrane. Indeed, in contrast to control cells (n = 7), Mud became mostly cytoplasmic on these mutant backgrounds in all pro-metaphase cells analyzed (n = 4 and n = 11, respectively; Fig. 6, F–G′″). We reasoned that the mislocalization of Mud could contribute to the spindle phenotype observed in Eph^x652 and myosin II mutants (Sqh^AA). In fact, mud down-regulation in neuroepithelial cells produced significant defects in spindle orientation compared with control cells (Fig. 6 H). Hence, both Dlg1 and Mud mislocalization in Eph mutants could account for the spindle phenotype observed in these mutants.

Cno and its vertebrate homologue Afadin regulate mitotic spindle orientation in different cell types both in vivo (Speicher et al., 2008; Rakotomamonjy et al., 2017) and in cell culture (Johnston et al., 2013; Carminati et al., 2016). In addition, Cno is required for the cortical recruitment of Mud/NuMA and Dlg1 in mitotic NBs (Speicher et al., 2008; Keder et al., 2015). Thus, we wondered whether Cno localization would also be affected in mitotic Eph^x652 optic lobe neuroepithelial cells. While in control neuroepithelia Cno is always enriched at AJs (n = 6 mitotic cells; Fig. 6, I and I′), in Eph^x652 mutants the cortical distribution of Cno was altered (i.e., it was either apically expanded or showed cortical gaps) in 7 of 10 mitotic cells analyzed (Fig. 6, J and J′). A similar phenotype was observed in 8 of 10 myosin II mutant (Sqh^AA) mitotic cells (Fig. 6, K and K′). In addition, we found that the down-regulation of cno in optic lobe neuroepithelial cells caused a misorientation of mitotic spindles (Fig. 6 L). Finally, we assessed the localization of both Dlg1 and Mud in mitotic neuroepithelial cells in which Cno was down-regulated. Dlg1 was apically expanded in 10 of 16 cno mitotic mutant cells (Fig. 6, M–N′), while the localization of Mud to lateral membranes was strongly diminished at all pro-metaphases analyzed (n = 6) relative to control cells (n = 7; Fig. 6, O–P′). We also observed an enrichment of Mud at what seemed to be the AJs in some cno mutant neuroepithelia (Fig. 6, P and P′). To confirm this, we used Shg (E-cadherin) as a marker of the AJs, and in two of eight cno mutant neuroepithelia (from eight different brains), Mud was clearly enriched at AJs (Fig. 6, Q–R″). In control neuroepithelia, Mud was occasionally observed at AJs along with Shg (E-cadherin), yet it was never clearly enriched in them (n = 6 brains; Fig. 6, Q–R″). Hence, we conclude that the restriction of Cno to the AJs of neuroepithelial cells in normal conditions might contribute to confine Dlg1 and Mud to lateral membranes.

The optic lobes in Eph^x652 brains were consistently larger than in WT brains (Fig. 7, A–C). In fact, compared with control brains, Eph^x652 optic lobe neuroepithelia displayed overgrowth at the mid-late L3 stage (Fig. 7, D–F), which was accompanied by an increase in the rate of neuroepithelial cell proliferation in Eph^x652 mutants (Fig. 7, G–I) along with a decrease in apoptosis at early L3 (Fig. 7, J, M, and P). At mid-late L3 stage, the levels of apoptosis were similar to those in control neuroepithelia (Fig. 7, K, N, and L–P), although the rate of cell proliferation was still higher in Eph^x652 mutants (Fig. 7, Q–T). Eph receptors are functionally linked to Rho small GTPases (Wahl et al., 2000; Noren and Pasquale, 2004; Lisabeth et al., 2013) and impaired Rho signaling has in turn been associated with an activation of the PI3K–Akt signaling pathway, a major regulator of cell proliferation (Lawlor and Alessi, 2001; Scheid and Woodgett, 2001; Luo et al., 2003; Li et al., 2005; Hopkins et al., 2014; Chang et al., 2016). We found that the expression of PI3K in the optic lobe NE was associated with an overproliferation phenotype similar to that of Eph^x652 mutants (Fig. 8 A). Hence, we tested the levels of Akt activation in Eph^x652 mutant brains. Compared with control brains, Eph^x652 mutant brain extracts revealed higher levels of phospho-Akt1 (pAkt1) in Western blots (WBs), as also seen in Eph^KD (kinase dead) mutant extracts (Fig. 8 B). Moreover, the overgrowth phenotype in Eph^x652 neuroepithelia was rescued by specifically expressing Akt1^RNAi or PI3K^RNAi in this tissue (Fig. 8, C and D). The spindle phenotype in Eph^x652 neuroepithelial cells was also partially rescued by down-regulating Akt1 signaling, although many vertical divisions (>30°) still persisted (i.e., 43.5% of the total divisions, n = 92, compared with 61.2% in Eph^x652 mutants, n = 80, and 17.3% in control divisions, n = 75; Fig. 8 E; see also Discussion). Finally, to determine whether Rho signaling was involved in this Akt1 regulation, we analyzed both neuroepithelial cell proliferation as well as the levels of pAKT1 in Rok^CATKG kinase dead and Rok^RNAi mutants, which displayed similar phenotypes to those shown by Eph mutants (Fig. 8, F and G). Moreover, a constitutively activated form of Rok (Rok^CAT) suppressed the overgrowth phenotype showed by Eph^x652 mutants (Fig. 8 H). We conclude that Eph signaling regulates cell proliferation in the optic lobe NE through activation of Rok, which in turn inhibits PI3K–Akt1 signaling pathway (see also below).

Different intrinsic and, to an apparently much lesser extent, extrinsic mechanisms tightly coordinate the correct orientation of the spindle during epithelial cell division (Morin and Bellaïche, 2011; Werts and Goldstein, 2011; Lu and Johnston, 2013; Williams and Fuchs, 2013). This process is essential to maintain epithelial morphology and to prevent hyperplasia (McCaffrey and Macara, 2011; Pease and Tirnauer, 2011). In fact, robust mechanisms have evolved to avert the consequences that a failure to correctly align the spindle would cause (Guilgur et al., 2012; Nakajima et al., 2013; Bergstralh et al., 2015).

Here we reveal Ephrin–Eph signaling to be a novel extrinsic mechanism that controls spindle orientation, influencing the regulation of intrinsic cues in the Drosophila optic lobe NE. Moreover, the loss of Eph signaling entails an up-regulation of the PI3K–Akt1 pathway and an ensuing increase in cell proliferation in this tissue. Both events seem to be independently regulated by Eph signaling, although both processes depend on Rok activation by Eph. From this point onwards, spindle orientation relies on Eph signaling–dependent aPKC activity, whereas cell proliferation is controlled by Eph/Rok-dependent inhibition of PI3K–Akt1 signaling pathway.

As a result of the data obtained, we propose a working model in which both reverse and forward Eph signaling would operate between optic lobe neuroepithelial cells (Fig. 9). On one hand, reverse Eph-ephrin signaling at the level of the subapical region would activate aPKC. Indeed, the existence of such signaling was previously shown to regulate tight junctions in Xenopus embryos through Par-6, which as part of the Par complex (Par-6/aPKC/Cdc42) ultimately leads to aPKC activation (Lee et al., 2008). Intriguingly, we observed a collapse of the subapical region (the equivalent in Drosophila of vertebrate tight junctions) in the Eph mutant optic lobe neuroepithelial cells. Moreover, a constitutively activated form of aPKC on an Eph mutant background rescued this collapse in the subapical region. In addition, constitutively active aPKC rescued the distribution of phosphomimetic myosin II (Sqh^EE), as well as the spindle orientation and deficient INM found in Eph mutant neuroepithelia. It is worth noting that constitutively active aPKC rescued the localization defects of Dlg1 in dividing Eph mutant cells in combination with active myosin II (Sqh^EE), but not in the presence of inactive myosin II (Sqh^WT; Fig. S4). This result strongly supports the requirement of both aPKC and Rok to make myosin II fully functional. myosin II mutants (Sqh^AA) had defects in the subapical region and in the spindle orientation similar to those found in Eph mutants, and failures in INM have previously been described when the apico-basal distribution of myosin II is altered (Rujano et al., 2013). Hence, myosin II seems to lie at the central regulatory node of all these processes. Interestingly, Ephrin B1 reverse signaling regulates apical adhesion of neural progenitors at the ventricular zone of the mouse developing cortex. In this context, Ephrin B1 induces adhesion of apical progenitors to the extracellular matrix by promoting the apical localization of integrin β1 (Arvanitis et al., 2013). Given that β integrin/Mys localizes basally in the Drosophila optic lobe NE, Ephrin reverse signaling would act, in principle, through aPKC activation to organize the apical/subapical region in this NE. Nevertheless, the decreased levels of β integrin/Mys we detected in Eph mutant neuroepithelial cells might also lead to defective adhesion to the extracellular matrix in this system.

On the other hand, PI3K–Akt1 signaling would be inhibited by forward Ephrin–Eph signaling through Rho1/Rok activation. Indeed, there was an increase in pAkt1 in Rok mutant brain extracts, as well as in cell proliferation in Rok mutant optic lobe neuroepithelia. Moreover, a constitutively activated form of Rok suppressed the Eph mutant overproliferation phenotype. Activation of Rho GTPases and the ensuing activation of Rho kinases downstream of Eph receptors has been well documented, especially in neuronal growth cones (Wahl et al., 2000; Shamah et al., 2001; Lisabeth et al., 2013). Our results, along with previously published data, suggest that Rok acts upstream of PI3K–Akt1 (Sordella et al., 2003; Furukawa et al., 2005; Chang et al., 2016). In fact, ROCK phosphorylates and activates the Phosphatase and tensin homologue PTEN, an inhibitor of PI3K–Akt signaling pathway, in mammalian cells. Phosphorylated PTEN inhibits the recruitment of Akt by PI3K and thus represses proliferation (Li et al., 2005; Hopkins et al., 2014). The down-regulation of Akt1 on an Eph mutant background rescued the Eph overproliferation phenotype, and intriguingly, it also partially rescued the Eph spindle phenotype. Akt is involved in microtubule interactions and spindle orientation in some contexts (Buttrick et al., 2008; Buttrick and Wakefield, 2008). Based on our results, Eph would inhibit Akt signaling in normal circumstances, and hence, in principle Akt would not be required for spindle alignment in optic lobe neuroepithelial cells. As Akt was overactivated in Eph mutants, this could interfere with spindle microtubules and contribute to the defects in spindle orientation. In such a scenario, Akt down-regulation could partially rescue the spindle alignment defects, as observed. However, in this situation many more vertical divisions were still detected, while the rescue of the spindle phenotype of Eph mutants by aPKC activation was much more robust (see Results). Along with the inhibition of Akt by Eph detected in WT neuroepithelial cells, these observations strongly suggest that Eph regulating myosin II through aPKC activation is the main driver orienting the spindle in this context.

Finally, the loss of Eph signaling in the optic lobe NE altered the localization of known intrinsic mitotic spindle regulators in other cell types, including Dlg1, Mud/NuMA, and Cno/Afadin, the loss of which also led to spindle orientation defects in this NE. aPKC activity, which was disrupted in Eph mutants, is required for the correct orientation of the spindle in some epithelia. For example, aPKC phosphorylates Pins, excluding it to the apical membrane and favoring its lateral localization in Drosophila imaginal disc epithelia (Guilgur et al., 2012). In NBs and S2 cells, Pins binds to Cno, Mud, and Dlg1 for orchestrating the correct alignment of the spindle (Bowman et al., 2006; Izumi et al., 2006; Siller et al., 2006; Speicher et al., 2008; Johnston et al., 2009; Wee et al., 2011). Hence, it is possible that aPKC may not phosphorylate and exclude Pins from the apical membrane in Eph mutant neuroepithelial cells. This might explain, at least in part, the apically expanded domains of Dlg1 and Cno in Eph mutants. However, alterations to the actomyosin cortex, which is compromised in Eph mutants, must also be directly responsible for the mislocalization of these spindle regulators, particularly as myosin II (Sqh^AA) mutants showed similar phenotypes to Eph mutants. In addition, Mud was not detected apically in Eph mutants. In fact, Mud was enriched in the cytoplasm, in detriment to its lateral distribution at pro-metaphase, in both Eph and myosin II (Sqh^AA) mutants. This lateral distribution of Mud was also compromised by down-regulating cno in the NE. Intriguingly, in these circumstances Mud was enriched at AJs in some of the larval brain neuroepithelia analyzed. Dlg1 was apically expanded in most dividing cno mutant cells. Thus, this cno mutant phenotype would suggest that Cno restriction to AJs might help confine Mud and Dlg to the lateral cortex in normal neuroepithelia. However, further work will be necessary to clarify this issue. Moreover, given the general conservation of spindle regulators in different cell types and organisms, it will be interesting to determine whether Eph signaling performs a similar function in other epithelia, both in Drosophila and in vertebrates, to modulate mitotic spindle alignment.

All the fly stocks used were from the Bloomington Drosophila Stock Center (BDSC) and the Vienna Drosophila Resource Center (VDRC), unless otherwise stated: yw (Speicher et al., 2008); Eph^x652 (Boyle et al., 2006); Eph^KD (Dr. Richard E. Dearborn, Albany College of Pharmacy and Health Sciences, Albany, NY); dlg1¹⁸ (BDSC: 36279); UAS-Eph-Myc (Scully et al., 1999); UAS-Eph^RNAi (VDRC: KK105139); C855a-Gal4 (BDSC: 6990); GFP-Sqh^WT, GFP-Sqh^EE, and GFP-Sqh^AA (Dr. Renata Basto, Institute Curie, PSL Research University, CNRS, Paris, France); UAS-Rho1^RNAi (BDSC: 9909); UAS-Rok^RNAi (BDSC: 28797); UAS-Rok^CAT-KG (BDSC: 6670); UAS-Rok^CAT (BDSC: 6668); UAS-aPKC^CAAXDN (Dr. Sol Sotillos, CABD, CSIC/JA/Universidad Pablo de Olavide, Seville, Spain); UAS-aPKC^ΔN (BDSC: 51673); UAS-aPKC^RNAi (VDRC: KK105624); UAS-cno^RNAi (VDRC: GD7769); UAS-Akt1^RNAi (VDRC: KK103703); UAS-PI3K^RNAi (VDRC: GD38985); UAS-mud^RNAi (BDSC: 38190); UAS-Dp110^CAAX (BDSC: 25908); and Gal80^ts (BDSC: 7108). The GAL4xUAS crosses were performed at 29°C except for the Tub-GAl80^ts; C855a-Gal4; Eph^x652 × GFP-Sqh^EE (or GFP-Sqh^WT) aPKC^ΔN; Eph^x652 cross, which was incubated for 3 d at 18°C and 1 d at 29°C (Fig. 5, H and I; and Figs. S3 and S4), and 2 d at 18°C followed by 3 d at 29°C (Fig. 5 E) before dissection. Similarly, the Tub-GAl80^ts; C855a-Gal4 × UAS-PI3K^CAAX (Fig. 8, A and G) and Tub-GAl80^ts; C855a-Gal4; Eph^x652 × UASPI3K^RNAi; Eph^x652 (Fig. 8 D) crosses were incubated for 2 d at 18°C and 3 d at 29°C before dissection.

Larval brains were dissected out in PBS and fixed with 4% PFA in PBT (PBS and Triton X-100 0.1%) for 20 min at room temperature with gentle rocking. For immunostaining with the rabbit anti-Cno antibody, the brains were fixed using the heat and methanol method (Miller et al., 1989) with small modifications. Briefly, dissected brains were incubated for 20 s at 80°C in buffer H (70 mM NaCl and 0.1% Triton X-100), followed by 5-min incubation on ice. The solution was then changed for PBT and incubated for another 5 min on ice. Embryos in Fig. S1 (D and E) were also fixed using the heat and methanol method; in this case, after the 5-min incubation on ice in buffer H, embryos were passed to heptane and treated with the same volume of methanol for devitelinization. From this point, both fixed brains and embryos were incubated in PBT for at least 30 min and then in PBT-BSA for 1 h before incubation with the corresponding primary antibody. For embryo staining with the anti-Eph antiserum (1:1,000; see below), a biotin anti-rabbit (Vector Laboratories) and the VectaStain Elite ABC HRP kit (Vector Laboratories) were used.

The primary antibodies used in this study were rabbit anti-Eph (1:500; see below), rat anti-DE-cadherin (1:100; Developmental Studies Hybridoma Bank [DSHB]), rabbit anti-PatJ (1:2,000; Dr. Elisabeth Knust, Max-Planck Institute for Molecular Cell Biology and Genetics, Dresden, Germany), rabbit anti-PKCζ (1:100; Santa Cruz Biotechnology, sc-216), rabbit anti-Par6 (1:2,000; Dr. Juergen Knoblich, Institute of Molecular Biotechnology of the Austrian Academy of Sciences, Vienna, Austria), rabbit anti-Baz (1:1,000; Dr. Andreas Wodarz, Institute for Anatomy University of Cologne Medical School, Koln, Germany), rabbit anti-Cno (1:400; Speicher et al., 2008), mouse anti-Dlg1 (1:100; DSHB), rabbit anti-Mud (1:400; Dr. Fumio Matsuzaki, RIKEN Center for Developmental Biology, Chuo-ku, Kobe, Japan), mouse anti-Mys (1:100; DSHB), rabbit anti-Scrib (1:2,000; Dr. Chris Doe, Institute of Neuroscience, Howard Hughes Medical Institute, University of Oregon, Eugene, OR), guinea pig anti-Dpn (1:2,000; Rives-Quinto et al., 2017), rabbit anti-cleaved dDcp1 (1:100; Cell Signaling Technology, 9578), mouse anti-PH3 (1:2,000; Millipore, 05-806), rabbit anti-PH3 (1:1,000; Millipore, 06-570), rabbit anti-CycE (1:100; Santa Cruz Biotechnology, 33748), rabbit anti-Cnn (1:400; Dr. Thomas C. Kaufman, Indiana University, Bloomington, IN), and mouse anti-γTub (1:100; Sigma-Aldrich, T5326). Secondary antibodies conjugated to fluorescent dyes were obtained from Molecular Probes or Jackson ImmunoResearch, and they were used at a dilution of 1:200. Cell nuclei were stained with DAPI (Vector Laboratories) and F-actin with Alexa Fluor 633–conjugated phalloidin (Invitrogen).

Fluorescence images were recorded by using a Leica upright microscope DM-SL with Spectral Confocal acquisition software (Leica). Images were taken with an HCX Plan Apochromat 63×/1.32–0.6-NA oil confocal scanning objective. Fluorescent images for Fig. 1 (B–D, R, and S), Fig. 6 (E–G and O–R), and Fig. S4 were recorded by using an Inverted Leica laser-scanning spectral confocal microscope TCS SP2 (Leica). Images were taken with 20×/0.7-NA Imm/Corr (Fig. 1 B) or 63×/1.4-NA oil (inset in Fig. 1 B; Fig. 1, C, D, R, and S; and Fig. S4) scanning objectives. Fluorescent imaging was performed at 21°C. Details of fluorochromes used in individual experiments are described in the immunostaining procedures. The distance between focal planes in each Z-stack was 0.8 µm (unless otherwise stated). The micrographs in the figures represent single focal planes from confocal Z-stacks except Fig. 6 (E–G, I–K, and M–P) and Fig. 7 (G and H), which represent the sum of 2–4 (Fig. 6) or 10 (Fig. 7) focal planes from the Z-stack (the same number for the different genotypes in a given experiment) and that were generated with the Z project tool of the open source Fiji software. For image analysis, mutant and control specimens from the same experiment were processed in parallel. Samples were analyzed within the same work session when possible, using exactly the same acquisition parameters in the confocal microscope.

Embryos in Fig. S1 (D and E) were imaged on a Zeiss Axio Imager A1 microscope with a 40×/1.3-NA oil objective lens (Carl Zeiss) and with a CCD camera (AxioCam HRc; Carl Zeiss) controlled by AxioVision software (Carl Zeiss). All images were assembled using Adobe Photoshop CS3.

To generate an anti-Eph antibody, two rabbits were injected with the synthetic peptide containing amino acids 994–1,022 of the Drosophila Eph intracellular domain (LTTRPSPESDGNHILDGQRGQNIFISTDLC, described in Scully et al., 1999). After three immunizations, the animals were bled, and the resulting sera were validated by immunofluorescence, immunohistochemistry, and in WBs (Fig. S1).

10 brains from third-instar larvae were dissected out in PBS, homogenized by repeated pipetting in fresh lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 0.1% SDS, 1 M EDTA, 1% Triton X-100, 1% sodium deoxycholate, protease inhibitors, 1 mM NaF, 100 mM Na₃VO₄, 2 mM PMSF, and Complete Protease Inhibitor Cocktail from Roche) and incubated on ice for 15 min. The protein extracts were centrifuged at 6,000 g at 4°C for 10 min and 5× Laemmli buffer with 0.1 M DTT was added to each sample, which was heated at 95°C for 10 min. After centrifugation at maximum speed (8,000 g) for 1 min, the protein extracts were resolved by SDS-PAGE. From the whole-protein extract, the material corresponding to five brains was loaded for Akt WBs (Fig. 6) or two brains for Eph WBs (Fig. S1). The Spectra Multicolor High Range Protein Ladder (Fermentas) was used as a molecular weight marker, and the PVDF filters were probed with the rabbit anti-pAkt (Ser473, 1:500; Cell Signaling Technology, 9271), rabbit anti-pan Akt (C67E7, 1:200; 4691), rabbit anti-Eph (1:3,000), mouse anti–α-tubulin (1:5,000; Sigma-Aldrich, T6199), or rabbit anti-Myc (1:5,000; Abcam, Ab9106), which were detected with appropriate HRP-coupled secondary antibodies.

Third-instar larval brains were dissected out in Shields and Sang M3 insect medium and incubated in fresh M3 medium containing 10% FBS and 40 µM EdU at RT for 1 h. The medium containing EdU was removed, and the brains were rinsed thrice with PBS. The brains were fixed with 4% PFA in in PBT (0.1% Triton X-100) at RT for 20 min and blocked with 0.1% BSA in PBT (0.3% Triton X-100). The Click-iT reaction was performed according to the manufacturer’s instructions (Life Technology, C10640), and the slides were then immunostained for E-cadherin to label neuroepithelial cells and cyclin-E and mounted in Vectashield mounting medium.

Spindle angles were calculated using ImageJ, measuring the angle between a line linking both centrosomes and the planar apical surface. These measurements frequently required correction in the xy plane to allow both spindle poles to be present in a single Z-plane. To avoid overcorrection, only mitoses with centrosomes in the same or in two consecutive Z-planes were considered for quantification. The dataset was analyzed using a χ² test with Yates correction.

The distance of the mitotic cell to the neuroepithelial apical surface was estimated by normalizing the distance from the metaphasic plate to the apical surface in relation to the total width of the NE (delimited by E-cadherin immunolabeling). At least 52 mitoses from no fewer than six different brains were quantified for each experimental condition (see corresponding figures and figure legends for details).

The optic lobe was defined by E-cadherin immunolabeling and measured using ImageJ. The Z-focal plane with the largest area was selected in each brain, and the data from the mutant was normalized with respect to the average of the control brains.

The size of the NE was estimated by measuring the depth of the corresponding confocal Z-stack. Each Z-stack (in micrometers) started from where the neuroepithelial cells appear and continued in an antero-posterior direction, terminating when the outer proliferation center (OPC) disappeared.

Cleaved-Caspase (Dcp1) immunostaining was performed to measure cell death. E-cadherin was used to visualize the OPC NE, and only those Dcp1-positive cells within the OPC or in direct contact with it were counted. The values shown are the mean of the total number of Dcp1-positive cells per OPC.

EdU-positive cells in the NE were measured in frontal brain sections every 3 µm to ensure that the same cell was not counted twice. The area analyzed was defined by anatomical references, starting when neuroepithelial cells appear and following an antero-posterior direction, sampling five sections for each brain hemisphere. The final number of proliferating neuroepithelial cells was the mean number of EdU-positive neuroepithelial cells per hemisphere, and at least seven hemispheres from a minimum of four brains were analyzed for each experimental condition.

Band intensity was quantified in the WBs using Fiji software, and the experiments were repeated at least three times. A minimum of six brains were used in each experiment, and the cells labeled in the cell death and cell proliferation assays were counted by eye.

For statistical analysis, the data were first analyzed using the Shapiro–Wilk test to determine whether the sample followed a normal distribution. The data with a normal distribution were analyzed using the unpaired two-tailed Student's t test for two-sample comparison, after performing an F test for variances, or with one-way ANOVA (with Tukey’s post hoc correction) for comparing three or more independent samples simultaneously. Data that did not follow a normal distribution were analyzed using a Mann–Whitney U test or a χ² proportion test in the case of the spindle angles (i.e., the proportion of populations with vertical divisions [>30°] were compared between the control and each of the mutant conditions). A Z test to compare two proportions was used in Fig. 7 T. The specific test used, the sample size (n), the SEM, or the SD of the population and the P value is indicated in each figure and/or figure legend; ****, P ≤ 0.0001; ***, P ≤ 0.001; **, P ≤ 0.01; and *, P ≤ 0.05.

Fig. S1 shows the specificity of the anti-Eph antibody both by immunofluorescence (in larval NE)/immunohistochemistry (in embryos) and by WB. Fig. S2 shows Eph^KD (kinase dead) and Ephrin^RNAi mutant phenotypes affecting polarity cue local distribution and NE architecture similar to those observed in Eph^x652-null mutants. Fig. S3 shows how a constitutively activated form of aPKC is able to rescue the collapse in the subapical region observed in Eph^x652-null mutant neuroepithelia. Fig. S4 shows that both active myosin (Sqh^EE) and aPKC are required to suppress the Dlg1 apical expansion phenotype observed in Eph^x652 dividing mutant cells.

We thank R. Basto, T. Chihara, R. Dearborn, T. Kaufmann, J. Knoblich, E. Knust, F. Matsuzaki, S. Sotillos, J.B. Thomas, A. Wodarz, the Bloomington Drosophila Stock Center at the University of Indiana, the Vienna Drosophila Resource Center, and the Developmental Studies Hybridoma Bank at the University of Iowa for kindly providing fly strains and reagents.

Our laboratory was supported by the Spanish grants from the Ministerio de Economía y Competitividad (BFU2012-33020 and BFU2015-64251) and by the European Regional Development Fund. The Instituto de Neurociencias in Alicante is a “Severo Ochoa” Center of Excellence.

The authors declare no competing financial interests.

Author contributions: M. Franco contributed to the design of experiments, carried out all the experiments, and analyzed data; A. Carmena designed the experiments, analyzed data, and wrote the manuscript.