The early insect embryo develops as a multinucleated cell distributing the genome uniformly to the cell cortex. Mechanistic insight for nuclear positioning beyond cytoskeletal requirements is missing. Contemporary hypotheses propose actomyosin-driven cytoplasmic movement transporting nuclei or repulsion of neighbor nuclei driven by microtubule motors. Here, we show that microtubule cross-linking by Feo and Klp3A is essential for nuclear distribution and internuclear distance maintenance in Drosophila. Germline knockdown causes irregular, less-dense nuclear delivery to the cell cortex and smaller distribution in ex vivo embryo explants. A minimal internuclear distance is maintained in explants from control embryos but not from Feo-inhibited embryos, following micromanipulation-assisted repositioning. A dimerization-deficient Feo abolishes nuclear separation in embryo explants, while the full-length protein rescues the genetic knockdown. We conclude that Feo and Klp3A cross-linking of antiparallel microtubule overlap generates a length-regulated mechanical link between neighboring microtubule asters. Enabled by a novel experimental approach, our study illuminates an essential process of embryonic multicellularity.
The nucleus relocates within the cell boundary in response to cell function (Bone and Starr, 2016; Gundersen and Worman, 2013). Aberrant nuclear positioning has been linked to failure of fundamental processes such as early embryo development, cell differentiation, cell migration, polarity determination, and homeostasis (Almonacid et al., 2015, 2019; Levy and Holzbaur, 2008; Minc et al., 2011; Neelam et al., 2015; Starr and Han, 2002). In mononuclear cells, cytoskeletal elements mechanically connect the nucleus to the cell cortex and act as the reference system for positioning (von Dassow et al., 2009; Pécréaux et al., 2016). One exception are large eggs in which cytoskeletal links between the nucleus and the distant cell cortex are not achieved (Wühr et al., 2009). Conversely, a multinucleated cell (coenocyte) undergoing nuclear proliferation has to disseminate positional information to each additional nucleus and requires a mechanism that adjusts the distance between neighboring nuclei (Gibeaux et al., 2017; Manhart et al., 2018). The early embryo of Drosophila melanogaster is both large and multinucleated but exhibits a surprising positional regularity of hundreds of nuclei perturbed by cycles of metasynchronous nuclear divisions (Foe and Alberts, 1983). During the first seven cycles, the nuclei spread axially from the anterior to the posterior end of the syncytial embryo and occupy the entire volume of the embryo (Baker et al., 1993). During nuclear cycles 7–9, most nuclei migrate to the embryo cortex, where they undergo additional rounds of division as they are anchored and prepared for cellularization (Lecuit and Wieschaus, 2000). Adequate numbers of nuclei and their proper positioning at the cortex determine cell size (Callaini et al., 1992) and precision of developmental patterning (Petkova et al., 2019), and are a result of regular distribution of ancestor nuclei during the preceding developmental phase (Kao and Megraw, 2009; Megraw et al., 1999; Vaizel-Ohayon and Schejter, 1999). The mechanisms required for maintaining uniform internuclear distances are not understood.
Drug inhibition and mutagenesis suggest that actomyosin-mediated cortical contractions drive cytoplasmic streaming and transport the nuclei predominantly along the longer axis of the embryo (Callaini et al., 1992; von Dassow and Schubiger, 1994; Deneke et al., 2019; Hatanaka and Okada, 1991; Royou et al., 2004; Wheatley et al., 1995). However, large-scale transport of cytoplasm explains neither how a uniform distribution emerges nor how nuclei are kept separate. Conversely, astral microtubules from neighboring nuclei may interact and generate repulsive force by motor binding and sliding antiparallel overlaps (Baker et al., 1993), which is reminiscent of the spindle midzone model explaining spindle elongation during anaphase B (Fu et al., 2009; Khmelinskii et al., 2009; Scholey et al., 2016). The effector Klp61F, a homotetrameric, bipolar kinesin-5, binds two overlapping microtubules and, when microtubules are antiparallel, slides them outward, reducing microtubule overlap length (Cheerambathur et al., 2013; Reinemann et al., 2017; Tao et al., 2006). Fascetto (Feo) is the Drosophila homologue of the Ase1p/Prc1/MAP65 family of homodimeric nonmotor MAPs that preferentially binds antiparallel microtubule overlaps (Bieling et al., 2010; Subramanian et al., 2010; Vernì et al., 2004). It accumulates at the spindle midzone from anaphase to telophase upon cyclin B degradation and controls the binding affinity of molecular motors in the spindle midzone (Hu et al., 2012; Khmelinskii et al., 2009; Kwon and Scholey, 2004; Wang et al., 2015; Zhu et al., 2006). One of these motors is Klp3A, a kinesin-4 homologue and inhibitor of microtubule dynamics (Bieling et al., 2010; Bringmann et al., 2004; Kwon et al., 2004; Subramanian et al., 2013; Williams et al., 1997). Prc1 and kinesin-4 are sufficient to form a stable microtubule overlap in vitro (Bieling et al., 2010). Kinesin-5 can reduce the overlapping, antiparallel microtubules cross-linked by Prc1 in vitro (Subramanian et al., 2010), which was proposed to contribute to force balance in the spindle midzone during anaphase B (Scholey et al., 2016).
Here, we investigated whether these three proteins are required for nuclear separation, lending support to an aster–aster interaction model (Baker et al., 1993) which has been reconstituted in Xenopus laevis egg extract (Nguyen et al., 2014, 2018). We performed a combination of gene knockdown, micromanipulation, and perturbation by exogenous protein addition in embryo explants to enable previously unachieved time-lapse visualization of nuclear and cytoskeletal dynamics.
Feo localization confirms antiparallel microtubule overlaps between asters of nonsister nuclei
Molecular cross-linking between astral microtubules of neighboring nuclei during the preblastoderm embryo stage has largely been unexplored because of optical constraints in live imaging. Using an extraction method to generate embryo explants from individual preblastoderm embryos (de-Carvalho et al., 2018) expressing Klp61F::GFP and Feo::mCherry and injected with Alexa Fluor 647–labeled Tubulin (Fig. 1 A), we visualized the localization of Klp61F and Feo to study their binding to spindle microtubules (Fig. 1 B). Klp61F::GFP localized at the microtubule-organizing centers (MTOCs), the metaphase spindle, and the spindle midzone in anaphase, as described previously for the nuclear divisions at the blastoderm stage (Heck et al., 1993; Sharp et al., 1999; Tao et al., 2006; Cheerambathur et al., 2008, 2013; Video 1). Furthermore, during anaphase B and telophase, we observed Klp61F::GFP-decorated microtubules intercalating with those from the neighboring aster, raising the possibility of antiparallel alignment of these astral microtubules forming an overlap zone to which kinesin-5 binds. On the other hand, Feo::mCherry exhibited weak localization to the metaphase spindle but strong localization to the spindle midzone during anaphase B and telophase (Fig. 1 B, arrows), as previously described for blastoderm division cycles (Wang et al., 2015). Strikingly, Feo also localized as small foci to the region between the nuclei (Fig. 1, B and C, arrowheads), thus reporting the presence of antiparallel microtubule overlaps to which Feo binds with higher affinity than individual microtubules. In vitro, microtubule overlaps that are decorated by Feo homologues are length controlled through microtubule stabilization and sliding by kinesin-4 (Bieling et al., 2010). Thus, the signal of Feo along microtubule overlaps should have a consistent length for a given concentration or activity of Feo and Klp3A. Accordingly, we measured the length of Feo::mCherry signal foci during anaphase B (Fig. 1 D). Because individual microtubules were not resolved, we measured the orientation of the signal foci in the context of where microtubules were growing and radially emanating from the MTOCs at the spindle pole. In anaphase and telophase, the four nuclei emerging from any two neighboring spindles defined four MTOCs and, thus, four possible combinations of astral microtubule interaction (Fig. 1 C, right). We measured the angle between the long axes of the signal foci and the closest connecting line between two MTOCs (Fig. 1 E). deviated little from 0, supporting the notion that Feo reports microtubule overlaps along the shortest path between neighboring asters. Altogether, in explants from preblastoderm embryos, the relative position of nuclei and the length of astral microtubules led to the formation of short, antiparallel, Feo-decorated microtubule overlaps. Furthermore, as a consequence of Feo cross-linking astral microtubules, a mechanical connection was established that may control the distance between neighboring asters and their associated nuclei.
During the last four syncytial nuclear cycles at the cortex, actin-based pseudo-compartments that drive membrane invagination—a physical barrier that is assembled and disassembled in every division cycle (Kotadia et al., 2010; Mavrakis et al., 2009a)—are thought to guarantee nuclear separation. Surprisingly, time-lapse confocal imaging of live embryos expressing Klp3A::GFP and Feo::mCherry (Fig. 1 F and Video 2) and injected with Alexa Fluor 647–Tubulin, revealed Feo colocalizing with Klp3A at the spindle midzone (arrows) and spot-like signals between neighboring spindles (arrowheads) in anaphase and telophase. On one hand, this observation suggests the combined activity of Feo and Klp3A, whereby Feo binding to microtubule overlaps recruits Klp3A to the overlap (Bieling et al., 2010; Subramanian et al., 2013). On the other hand, the signal foci indicate that antiparallel microtubule overlaps occur between neighboring nonsister nuclei across actin furrows and membrane invaginations. With consecutive division cycles and increasing nuclear packing, the foci maintain size but become more frequent (Fig. S1 A). Triangulation analysis revealed that foci appear at locations equidistant to neighboring spindle poles, in the center of the aster–aster overlap zone (Figs. 1 G and S1 B), and their size is comparable to that observed in explants (Fig. S1 C). In embryos expressing Klp61F::GFP and Feo::mCherry, the GFP signal is excluded at the spindle midzone, where Feo::mCherry localizes strongly. Although the Klp61F concentrates at the spindle poles, it does not show distinct localization at the aster–aster interaction zone (Fig. 1 H and Video 3). Altogether, these observations led us to question the current paradigm that actin pseudo-compartments prevent microtubule cross-linking between neighboring asters or nuclei. We hypothesize from this localization data that the microtubule-based mechanical connection plays a decisive role in nuclear positioning in preblastoderm- and early-blastoderm–stage embryos. Since microtubules are a prerequisite, we perturbed microtubule dynamics with low doses of the depolymerization drug nocodazole. We supplemented this drug after explant deposition, using a fine micropipette and buffer conditions described elsewhere (Telley et al., 2013). A concentration of 4 µM sustained spindle assembly and chromosome segregation but abolished daughter nuclei separation (Video 4). The microtubule perturbation caused local aggregation of dividing nuclei. We also observed chromosomes fusing and forming larger nuclei. Those nuclei that were more distant from the drug injection site still separated consistently, which we attribute to a diffusion gradient in drug concentration. Overall, this experiment demonstrates that nuclear separation depends on microtubules and supports a role of Feo and Klp3A in nuclear positioning.
Knockdown of Feo, Klp3A, or Klp61F leads to defective nuclear delivery to the embryo cortex
We wanted to understand the functional implication of the microtubule-binding proteins that localize between nonsister nuclei and if they are required for correct nuclear delivery to the cortex. To this end, we perturbed the protein levels of Feo or Klp3A using an RNAi approach and upstream activating sequence (UAS)–Gal4 expression in the germline (Staller et al., 2013). We also targeted Klp61F for phenotypic comparison. Using different available fly lines (TRiP), we expressed RNAi against these genes individually in the developing Drosophila oocyte (Fig. S2 A), while expressing Jupiter::GFP, a microtubule reporter (Morin et al., 2001), and H2Av::RFP, a chromatin reporter (Schuh et al., 2007). We exploited the expression kinetics of V32–Gal4 to drive the UASp–RNAi constructs with peak in late oogenesis to prevent undesirable defects during stem cell division. The efficiency of knockdown was measured at the RNA level using a quantitative PCR approach (Table 1 and Materials and methods). Fertilization in embryos inhibited for Feo, Klp3A, or Klp61F expression was similar to that of control embryos. However, we were unable to determine the exact cycle number when nuclei arrived at the cortex in knockdown embryos. Of note, the interval of division cycles occurring at the cortex and in the embryo explants remained unaltered when compared with controls. One RNAi construct against feo (TRiP 28926) did not show any phenotype. Under all other knockdown conditions, we observed nuclei arriving later on average: ∼45 min in knockdown condition versus ∼15 min in controls, following a 45-min egg laying period. In knockdown embryos, nuclei were irregularly distributed at the cortex, exhibiting clustering and sometimes missing entirely at the posterior end, in contrast to the regular distribution seen in the control RNAi embryo (Figs. 2 A and S2 B). The nuclear density is reduced after knockdown as compared with the control (Fig. 2 B) but exhibited considerable variability between embryos, which we attributed to variability in UAS-Gal4–mediated expression of RNAi between individual embryos. Reassuringly, in feo RNAi embryos, we observed a reduction (below detection) of fluorescence intensity of Klp3A::GFP at the midzone and between neighbor asters (Fig. S2 C and D). This confirms Klp3A being downstream of Feo binding to microtubule overlaps (Bieling et al., 2010).
Our nuclear positioning analysis revealed embryos with larger areas lacking nuclei (Fig. S3 A) and with anatomically eccentric (Fig. S3 B) and asymmetric (Fig. S3 C) nuclear distribution. Overall, RNAi against feo resulted in larger distribution changes than RNAi against klp61f (Fig. S3) despite similar average internuclear distance (Fig. 2 C). The distribution of first-neighbor internuclear distance was broader and skewed toward longer distances for knockdown conditions, reflecting clusters and larger unoccupied areas, while RNAi against klp3a gave on average the strongest phenotype (Figs. 2 C and S3). The irregularity in nuclear position at the cortex increased as the cortical nuclear cycles progressed (Video 5). Occasionally, in feo knockdown embryos, we observed sister nuclei fusing after mitosis and nonsister nuclei moving toward each other, leading to fusion of the spindles and overcondensed chromatin. This likely occurs in preblastoderm cycles as well, leading to fewer nuclei arriving at the cortex and contributing to nonuniformity. Conversely, fusion was never seen in controls.
In summary, the activity of all three microtubule-associated proteins is required in the preblastoderm embryo for correct delivery of nuclei to the embryo cortex. However, kinesin-5 is required for spindle assembly (Heck et al., 1993; Sawin et al., 1992), and thus the phenotype can emerge due to assembly defects rather than postmitotic nuclear separation. Because depletion of Klp61F led to a milder phenotype despite high knockdown efficiency, and because of the distinct localization of Feo and Klp3A in the aster–aster interaction zone (Fig. 1, F and G), we followed up on the role of the latter two genes in maintaining internuclear distance.
Developmental reset ex vivo reveals failure in nuclear distribution upon RNAi knockdown
Our analysis of nuclear distribution during knockdown in the embryo suggests that Feo and Klp3A are involved in nuclear delivery to the cortex. However, our knockdown approach in vivo has two drawbacks that can lead to misinterpretation: (1) the three proteins play a role in spindle midzone function, and their depletion may affect chromosome segregation in anaphase; and (2) the RNAi expression occurs chronically during late oogenesis. Thus, the irregular distribution of nuclei during cortical migration can be due to early sister chromatid separation errors, leading to missing nuclei in the embryo center and exponentially fewer in subsequent division cycles. Alternatively, inefficient nuclear separation following fertilization can lead to spindle fusion and mitotic errors. To circumvent the inability to detect accumulated effects, we performed time-lapse imaging of nuclear division cycles in embryo explants from preblastoderm embryos that were inhibited for either Feo or Klp3A protein expression. Because these explants contained only a few dividing nuclei, we followed their distribution, or the failure thereof, while mimicking the very beginning of preblastoderm embryo development. We tracked individual nuclei undergoing division cycles and registered the distribution and any fusion events between sister and nonsister nuclei (Fig. 3 A and Video 6). Importantly, the time-lapse observation of nuclear divisions allowed us to determine if nuclear distribution changes arise from reduced separation or form mitotic failures and arrest, which has different consequences for the delivery of nuclei to the embryo cortex. In explants from control embryos, nuclei divide and distribute regularly in the entire explant (Fig. 3 A, left, white dashed circle) until a saturated nuclear distribution is reached and occasional mitotic failures in the subsequent cycle are observed. The nuclear density at saturation is comparable to nuclear cycle 10 in the intact embryo (1,800–2,000 nuclei/mm2; Foe and Alberts, 1983), corresponding to an internuclear distance of ∼25 µm (hexagonal approximation). Strikingly, the nuclei from feo- and klp3a-knockdown embryos also divide consecutively. The average distance between sister nuclei and between nonsister nuclei was lower in the test RNAi as compared with the control (Fig. 3, B and C). However, the nuclear position after mitotic separation was maintained in the feo RNAi, while knockdown of klp3a led to frequent spindle fusion at a comparable nuclear density and accumulation of mitotic failure. Interestingly, spindle length decreased upon depletion of feo (Fig. 3 A, middle), but we did not observe a significant decrease in spindle length upon klp3a depletion as reported earlier, most likely due to inefficient knockdown as compared with deletion (Williams et al., 1995). In summary, the inhibition of Feo expression leads to reduced nuclear separation between sister nuclei and incomplete occupation of nuclei within the explant, a hallmark of the unoccupied spaces in the blastoderm embryo (Fig. 2). However, while a knockdown of feo sustains mitotic divisions, klp3a knockdown produces a spindle fusion phenotype. It is conceivable that the reduction of Klp3A protein causes microtubule overlap overgrowth, and despite cross-linking by Feo and other MAPs, these long overlaps allow for a net attractive force to build up between spindles (Gatlin et al., 2009). However, we cannot exclude that the partial knockdown of klp3a already causes mitotic defects related to spindle assembly and chromosome alignment. Because of these confounding effects, we decided to focus on Feo exclusively.
Displacement of nuclei is rescued in control but not in feo RNAi embryo explants
To test the model of an astral microtubule cross-linker–based separation mechanism for nonsister nuclei, we took advantage of the amenability of embryo explants for mechanical manipulation and designed an acute perturbation approach. We asked how Feo relocalizes when the distance between two interphase nuclei is manually reduced. Finally, we asked whether, under a feo-knockdown condition, nuclei could adjust their position when brought in close proximity before division. To address these questions, we performed contact micromanipulation and changed the positions of two nonsister nuclei that were exiting mitosis (Fig. 4 A, magenta; and Video 7), during the nonmitotic phase, which is amenable to manipulation (Telley et al., 2012). Following manipulation of two nuclei (Fig. 4 B, yellow), nuclei were typically in interphase, when Feo is not expected to localize due to cycle regulation. Nonetheless, we registered strong localization of Feo::mCherry in anaphase and telophase of the next cycle at midzones of all spindles, and strongly in the region between the manipulated nuclei (Fig. 4 B, bottom, asterisk), while asters from distant, nonmanipulated nuclei did not recruit the microtubule cross-linker detectably (Fig. 4 B, arrowhead). Next, we quantified nuclear separation of two neighboring nuclei dividing into four daughter nuclei by determining the four final positions (Fig. 4 C), arranging these positions as a quadrilateral; aligning, annotating, and overlaying them in a common coordinate system (Fig. 4, D and E); and calculating area (Fig. 4 F) and lateral distances (Fig. 4, G and H). We performed these measurements under the control RNAi condition for nuclei in a large unoccupied cytoplasmic space, in a saturated space where several nuclei have spread through the entire explant, and in a crowded explant representing one more division cycle after saturation; this differentiated comparison takes into account nuclear density dynamics in cycling explants, as described in the previous section (Fig. 3). We found that the area of nuclear separation after manipulation is lower than in the nonmanipulated and saturated space but indifferent from the crowded control (Fig. 4 F). The manipulated nuclei divided and separated their daughter nuclei at ∼15 µm, while the distance between nonsiblings was maintained at ∼25 µm (Fig. 4, G and H; Telley et al., 2012), phenocopying the minimal separation seen in crowded explants, for which distance maintenance is challenging. Interestingly, these separation distances are similar to what was reported for the blastoderm embryo (Kanesaki et al., 2011). Finally, we performed the manipulation of nuclear position in feo-knockdown explants expressing Jupiter::GFP and H2Av::RFP. In these experiments, after manipulation, the daughter nuclei moved toward each other rather than apart, forming nuclear aggregates (Fig. S4) and phenocopying the treatment with nocodazole as presented earlier (Video 4). The separation of siblings was approximately the nuclear diameter (∼7 µm; Fig. 4 G, dashed line) and the separation of nonsiblings was ∼10 µm (Fig. 4 H). We conclude that acute repositioning of nuclei is detected by the separation machinery, as reported by Feo, and is counteracted to prevent spindle fusion or aggregation of nuclei. In other words, Feo is required to prevent nuclear collisions.
Nuclear separation in the syncytium requires astral microtubule cross-linking by Feo
Feo is a dimer and, in vitro, has high affinity for binding two antiparallel microtubules (Bieling et al., 2010; Subramanian et al., 2013, 2010). In this function, Feo can generate a repulsive mechanical link—an apparent stiffness—that prevents concentric movement and eventual contact of neighboring nuclei. This model predicts a lower repulsion stiffness in the presence of a monomeric construct of Feo, which binds to the same microtubule lattice binding site as the full-length dimer but does not cross-link the antiparallel microtubules. We expect that this molecular competition leads to weaker cross-linking and can be measured as shorter internuclear distance, irregular separation, or frequent nuclear contacts. Thus, we designed two synthetic protein expression constructs, one containing the full feo coding sequence (sFeoFL) and the other lacking the N-terminal, putative dimerization domain (sFeoN; Subramanian et al., 2013), and fused them to a C-terminal GFP and a His6 tag sequence (Fig. 5, A and C). By design, the truncated construct should have unaltered microtubule-binding affinity, while dimerization and thus cross-linking are abolished. Proteins were expressed in Escherichia coli, affinity-purified, and dialyzed into embryo explant–compatible buffer (Telley et al., 2013; Fig. S5 A). Full-length Feo::GFP forms dimers, as was reported for vertebrate homologues (Bieling et al., 2010; Subramanian et al., 2013, 2010), while the truncated protein is dimerization deficient and forms monomers or weak dimers, as shown in a native-PAGE (Fig. S5 B). When the full-length protein was injected at nanomolar concentration into embryos, GFP signal localized at the central spindle (Fig. 5 B, arrow). As in transgenic embryos (Fig. 1, F and G, arrowheads), we also detected small foci of green fluorescence between neighboring nuclei (Fig. 5 B, arrowheads), suggesting that the purified protein and the transgenic construct localize identically and under cell cycle control, with fluorescence disappearing in interphase and reappearing during telophase of the following cycle (Video 8). Though less strong, the truncated protein construct also localized at the spindle midzone, however not in the cycle following injection but in the subsequent division cycle (Fig. 5 D, Fig. S5 C, and Video 9). The lower intensity suggests that the truncated construct competes weakly with endogenous Feo. Injection of full-length Feo::GFP maintained regular nuclear delivery to the embryo cortex, while injection of truncated FeoN::GFP caused unnatural spindle contacts and nuclear separation defects (Fig. S5 C, middle; and Video 8). Furthermore, when the full-length protein was injected into feo RNAi embryos, the defective nuclear distribution was rescued to a large extent (Fig. S5 D). Nuclei arrived at the embryo cortex more symmetrically between anterior and posterior ends (Fig. 5 E), in a less skewed distribution (Fig. 5 F), and with more uniform internuclear distance (Fig. 5 G) as compared with mock-injected feo RNAi embryos. Owing to the variability of injection and knockdown efficiency, we fully recovered nuclear density to a normal level in two embryos and significantly increased nuclear density in the remaining five embryos (Fig. 5 H). Notably, the injected protein pool was stable for ≥90 min, throughout several division cycles. In summary, we show that a GFP-tagged full-length protein construct localizes correctly and rescues the gene knockdown in the germline. We conclude that it is functionally identical to the endogenous protein that is maternally deposited in the egg and stable during syncytial development.
Finally, having designed and purified the truncated and full-length proteins with identical procedures, we asked how nuclear separation changes upon excess of the dimerization-deficient Feo protein, added at 100–200 nM final concentration to wild-type embryo explants containing one or two nuclei. As a control condition, we injected the full-length protein at the same final concentration into embryo explants, and despite this perturbation, the explant supported normal nuclear separation and distribution (Fig. 5 I, left). Conversely, adding the truncated protein construct worsened nuclear separation considerably after chromosomes segregated. Here, in contrast to the control condition, nuclei did not occupy the entire explant space after consecutive divisions. The short internuclear distance led to unnatural chromosome aggregation and fusion and eventually to mitotic failure. Nuclear separation of two neighboring nonsister nuclei, as measured by the quadrilateral area defined by their position, was significantly smaller than in control divisions in the presence of full-length Feo protein (Fig. 5, J and K). We conclude that microtubule cross-linking by Feo generates a repulsive mechanical link between microtubule asters. Thus, it lies at the heart of nuclear separation maintenance during the multinucleated one-cell stage of Drosophila embryo development.
A cornerstone of embryonic development is the formation of a polarized epithelium. Plants and many invertebrates achieve this developmental stage with a unicellular embryo undergoing nuclear proliferation followed by cellularization, a specialized form of cytokinesis (Hehenberger et al., 2012; Lecuit and Wieschaus, 2000). Recently, the molecular building blocks and morphogenetic characteristics of cellularization have also been identified as part of the life cycle of a nonanimal eukaryote (Dudin et al., 2019). The offspring of Sphaeroforma arctica arises from nuclear proliferation, compartmentalization, and plasma membrane invagination, generating a proto-epithelium from which newborn cells detach. These observations support the hypothesis that epithelia evolutionarily predate animals (Dickinson et al., 2012). We propose that correct compartmentalization and generation of uninuclear offspring necessitates robust nuclear separation. If warranted true, then a separation mechanism must have coevolved with the origin of epithelia and was essential for the emergence of multicellularity.
Nuclear proliferation in a coenocyte poses a new challenge: how does the cell safeguard the separation and prevent contact of nuclei while their number increases? Two solutions seem plausible. On one hand, the cell may control the division axes and separate daughter nuclei along linear paths that do not cross. On the other hand, the cell may constrain internuclear distance independently of separation trajectories. Consider two nuclei that are about to divide and separate their progeny along the spindle axis (Fig. 6 A). In a 3D space, none of the daughter nuclei may collide unless the spindle axes are both coplanar and nonparallel. Typically, nuclei migrate only 10–15 µm away from the original spindle center before dividing again (Telley et al., 2012). This geometric constraint reduces configurations that produce colliding trajectories in a 2D topology to ∼40% of all possible spindle axis orientations, so that axes intersect at an angle between 0° (collinear) and 70° (Fig. 6 B). Adding complexity, spindles in a network with optimal packing face a number of neighbors (6 in 2D and 12 in 3D; Fig. 6 C). Thus, a synchronously dividing spindle network will inevitably produce colliding trajectories of daughter nuclei. It is therefore necessary that, instead of controlling division axes, the cell controls nuclear proximity independently of the relative orientations they divide (Fig. 6 B). This enables the syncytial embryo to divide hundreds of nuclei synchronously and distribute them to any unoccupied position. Here, we demonstrate a molecular mechanism that responds to short internuclear distances in the syncytium with microtubule-dependent repulsion. Each nucleus is associated with a radial array of microtubules nucleated by the centrosome, which duplicates and forms the two spindle poles in the next division. First, however, this microtubule aster guides nuclear migration and grows large enough to encounter microtubules from neighboring asters that migrate as well. This encounter leads to interdigitation of the microtubule plus ends (antiparallel overlaps) and forms binding sites for cross-linking proteins. Our data shows that Feo, the Prc1 homologue in Drosophila and antiparallel microtubule cross-linker, plays a central role in defining a minimal internuclear distance in the syncytial Drosophila preblastoderm embryo.
Vertebrate Prc1 is a microtubule-binding protein with high turnover kinetics and ≥28 times higher affinity for antiparallel microtubule overlaps than for single microtubules (Bieling et al., 2010). This biochemical property, together with fluorescent labeling, renders Prc1 homologues reliable reporters for microtubule aster overlaps in live-cell imaging assays (Nguyen et al., 2014). Prc1 cross-linking antiparallel microtubules generates a high-affinity binding site for the motor protein kinesin-4 (Kif4/Xklp1/Klp3A) at the overlap (Bieling et al., 2010; Kurasawa et al., 2004). In Xenopus eggs and early embryos, where asters are unusually large, Prc1 and Kif4A define the radial organization and dynamics of microtubules and prevent invasion of neighboring asters by antiparallel microtubule cross-linking (Nguyen et al., 2018). The same protein module is responsible for recruiting cytokinesis signaling complexes and formation of the cleavage furrow (Nguyen et al., 2014). In vitro, in addition to maintaining a stable overlap length, coactivity of Prc1 and Xklp1 causes buckling of overlapping microtubules, which are immobilized at their minus ends (Bieling et al., 2010). In a sliding assay of taxol-stabilized microtubules in which microtubules in solution and glass-immobilized microtubules form pairs cross-linked by Prc1, the antiparallel pairs of microtubules are slid apart by Kif4 (Wijeratne and Subramanian, 2018). This is reminiscent of plus end–directed sliding of kinesin-5 (Eg5; Kapitein et al., 2005) and explains the requirement of Prc1 orthologues for spindle elongation in several species (Khmelinskii et al., 2009; Schuyler et al., 2003; Vukušić et al., 2021; Wang et al., 2015; Zhu et al., 2006). Indeed, plus end–overlapping microtubules have an apparent mechanical stiffness that is governed by molecular friction and motor activity (Forth et al., 2014; Wijeratne and Subramanian, 2018). An assembly of tens of such microtubule pairs generates sufficient mechanical resistance against compressive forces in the nanonewton range, enough to keep two spherical organelles of 5–8-µm diameter attached to the microtubule minus ends well separated (Lele et al., 2018). Thus, modular upscaling of a single pair into overlapping radial arrays illustrates how the stabilization and sliding mechanism of a Feo- and Klp3A-decorated antiparallel microtubule pair produces repulsion between two syncytial nuclei.
Feo::GFP or Feo::mCherry expressed in the transgenic line, or supplemented as purified protein, exhibited focal fluorescence signals in the blastoderm embryo and in the explant from preblastoderm embryos. Here, we showed that the length of these signal foci is surprisingly short and uniform. According to in vitro data, and neglecting any regulation other than affinity and stabilization activity for the underlying microtubule overlap to maintain such a short length, the concentration of kinesin-4 in the cytoplasm must be at least one magnitude in excess of Feo (Bieling et al., 2010). Moreover, knockdown of feo by RNAi abolished the signal of Klp3A::GFP below detection, thus considerably reducing the bound fraction of Klp3A at the central spindle. In the embryo, while confirming their already established localization at the spindle midzone (D’Avino et al., 2007; Kwon et al., 2004; Page and Hawley, 2005; Wang et al., 2015; Williams et al., 1995), we recorded Klp3A::GFP signal colocalizing with Feo::mCherry in areas between neighboring spindle asters. However, we could not clearly assess the localization of Klp3A in explants from preblastoderm embryos because of the low signal intensity. A single copy–tagged Klp3A construct expressed with the endogenous promoter failed to provide sufficient signal, and we decided to work with overexpression constructs (Sarov et al., 2016). This indicates that the microtubule overlap–bound fraction of endogenous Klp3A is comparatively small despite the molar excess in the cytosol as derived from overlap length. Together, these observations point at a protein interaction network localized at antiparallel microtubule overlaps that is sensitive to small changes of Feo. As Feo binds microtubule overlaps independently (Bieling et al., 2010), the phenotypes in intact embryos and in explants can arise due to disproportionate Klp3A perturbation downstream of Feo.
In summary, our live-cell microscopy data from blastoderm embryos and preblastoderm embryo explants support a “central spindle model” built from individual pairs of microtubules cross-linked and length-regulated by Feo and Klp3A (Bieling et al., 2010), and potentially through stabilization/sliding activities of Klp3A (Vukušić et al., 2021). More importantly, we show how overlapping microtubules in the aster–aster interaction zone (Nguyen et al., 2014) form midzone-analogous cytoskeletal assemblies that persist throughout blastoderm development. This is particularly intriguing given that, at the embryo cortex from cycle 10 onward, actin-based pseudo-furrows are thought to form precellular compartments that prevent nuclear contact (Karr and Alberts, 1986; Kellogg et al., 1988; Lecuit, 2004; Mavrakis et al., 2009b). In the early blastoderm cycles, however, this compartmentalization may not yet be efficient enough to safeguard nuclear separation, and astral microtubule cross-linking persists as the dominant mechanism. This interpretation is further supported by an earlier observation in mutants of the maternal-effect gene sponge, embryos of which do not form actin caps and pseudo-furrows in blastoderm stage but depict a homogeneous nuclear distribution in cycles 10 and 11 (Postner et al., 1992).
Feo is essential for central spindle assembly and cytokinesis in somatic cells, containing two Cdk phosphorylation sites (Vernì et al., 2004). Feo, like Prc1 in human cells and Ase1p in fission yeast, is under cell cycle control and undergoes phosphorylation-dependent localization, from low-intensity decoration of metaphase spindle microtubules to a strong localization at the central spindle in anaphase and telophase (Khmelinskii et al., 2009; Polak et al., 2017; Subramanian et al., 2013; Wang et al., 2015; Zhu et al., 2006). In the present work, we showed that the focal localization of Feo and Klp3A between neighboring nuclei is in synchrony with central spindle localization. It is in this phase of the division cycle that expanding spindles and separating nuclei cause a large spatial perturbation to the positional distribution (Kanesaki et al., 2011; Lv et al., 2020). Thus, a dual role for Feo under cell cycle control emerges: while it targets the central spindle at anaphase onset, forming the spindle midbody, it also binds to astral microtubule overlaps in a phase during which collision prevention is most needed.
In Drosophila embryos, spindle elongation at anaphase B is powered by the sliding activity of Klp61F (Brust-Mascher et al., 2009). Following the mechanism proposed by Baker et al. (1993), and because Klp61F is a candidate cross-linker and slider of overlapping astral microtubules, we performed RNAi knockdown in the germline. Inhibition of Klp61F expression led to lower density and nonuniform delivery of nuclei to the embryo cortex, confirming its essential role during preblastoderm development. However, owing to the established role of Klp61F in spindle assembly, the RNAi phenotype can emerge because of multiple chromosome segregation failures that were undetectable in the preblastoderm embryo. Here, the embryo explant assay overcomes an experimental limitation and enables time-lapse image acquisition of uni- or binuclear explants undergoing multiple divisions. Consequently, we could confirm that klp61f knockdown led to more frequent division failures rather than shorter nuclear separation. Still, Klp61F and Feo can functionally cooperate in cross-linking astral microtubules because both proteins recognize and bind to microtubule pairs, although with different preference for microtubule orientation (Bieling et al., 2010; Kapitein et al., 2005; Kellogg et al., 2016; Subramanian et al., 2010). In human cells, Prc1-dependent Kif4A motor activity and the microtubule sliding by Eg5 are redundant for spindle elongation during anaphase (Vukušić et al., 2021). Interestingly, in Drosophila, while Feo modulates binding and localization of Klp61F at the spindle midzone in anaphase, Klp61F cannot functionally rescue the absence of Feo (Wang et al., 2015). Presumably, Ase1p/Prc1/Feo binding to microtubule overlaps creates a protein binding hub for motors and regulators (Bieling et al., 2010; D’Avino et al., 2007; Hu et al., 2012; Khmelinskii et al., 2009; Sasabe and Machida, 2006; Subramanian et al., 2013). This property has not been demonstrated for kinesin-5 orthologues. Together, the collection of evidence from us and others suggests that Klp61F is not at the core of astral microtubule-driven nuclear separation.
Lastly, the reader may wonder how astral microtubule overlap cross-linking by Feo and Klp3A defines the internuclear distance metric, leading to a distribution of syncytial nuclei with high regularity. In an earlier study, Telley et al. (2012) showed that microtubule aster size varies throughout the nuclear division cycle, reaching a maximum of 11 ± 3 µm in telophase. Herein, the aster size represents the length distribution of microtubules which, for dynamic microtubules with nongrowing minus ends, is well approximated with an exponential distribution (Howard, 2001). We assume that two microtubules from neighboring asters grow at least to average length, overlap with their plus ends, and are collinear. If the stabilized overlap length is ∼1 µm, then the total length from centrosome to centrosome is on average 21 ± 4 µm. Considering that a centrosome is ∼1 µm in size, and that a nucleus in late telophase is 5 ± 1 µm in diameter, the total distance between the centers of neighboring nuclei is 28 ± 4 µm. This estimate is in good agreement with the internuclear distance distribution measured from center to center of each nucleus (Fig. 3 C), the minimal nonsibling internuclear distance in extract (Fig. 4 H), and earlier reported separation distances of daughter nuclei (Telley et al., 2012). Thus, the short antiparallel overlap length of microtubules from neighboring asters and the microtubule length distribution are sufficient to explain the geometry of nuclear distribution in the Drosophila syncytial embryo.
Materials and methods
Rearing of flies for general maintenance was done as previously described (Stocker and Gallant, 2008). The following fly lines were used to make recombinants or trans-heterozygotes: Jupiter::GFP (Bloomington Drosophila Stock Center [BDSC] 6836), Jupiter::mCherry (generated by and obtained from Nick Lowe in D. St. Johnston’s laboratory, The Gurdon Institute, Cambridge, UK), pUbq>Spd2::GFP (gift from M. Bettencourt-Dias, Instituto Gulbenkian de Ciência), H2Av::RFP (BDSC 23650), Feo::GFP (BDSC 59274), Feo::mCherry (BDSC 59277), Klp61F::GFP (BDSC 35509), Klp3A::GFP (Vienna Drosophila Resource Center 318352), RNAi targeting feo (BDSC 28926 and 35467), RNAi targeting klp3a (BDSC 40944 and 43230), RNAi targeting klp61f (BDSC 33685 and 35804), RNAi targeting mcherry (BDSC 35785), and UASp–GFP (BDSC 35786).
Knockdown experiments were performed using the TRiPGermline fly lines for RNAi in germline cells (Perkins et al., 2015). The UAS-hairpin against a gene of interest was expressed using V32–Gal4 (gift from M. Bettencourt-Dias) at 25°C. The expression profile of V32–Gal4 in the oocyte was assessed by dissecting ovaries of flies expressing UASp–GFP at 25°C and comparing GFP expression at different developmental stages with fluorescence microscopy.
Sample preparation and extraction
Embryos were collected from apple juice agar plates mounted on a fly cage. They were dechorionated in 7% sodium hypochlorite solution, aligned, and immobilized on a clean coverslip using adhesive dissolved in heptane and covered with halocarbon oil (Voltalef 10S). Extraction of cytoplasm from individual embryos and generation of explants were performed on a custom-made microscope as previously described (de-Carvalho et al., 2018; Telley et al., 2013).
To measure the transcript levels of feo, klp3a, and klp61f, total RNA was extracted following standard procedures (PureLink RNA Mini Kit; Ambion) from embryos collected after 40 min of egg laying. A cDNA library was made from Oligo(dT)12–18 as described in the manufacturer's protocol (Transcriptor First Strand cDNA Synthesis Kit; Roche). Quantitative PCR was performed using Quantifast SYBR Green PCR Kit (204052) and QuantiTect Primers for feo (QT00919758) in feo RNAi (35467 and 28926), klp3a (QT00497154) in klp3a RNAi (40944 and 43230), and klp61f (QT00955822) in klp61f RNAi (35804 and 33685). actin (QT00498883) was used as a housekeeping gene control. Results are presented in Table 1.
Purification of sFeoFL::GFP and sFeoN::GFP
The full coding sequence of the feo gene fused to a C-terminal GFP tag, synthesized, and codon optimized by NZYTech, and is referred to herein as sFeoFL::GFP. The DNA was cloned into the vector pET-21a containing a C-terminal His6-tag using restriction enzymes NheI/XhoI and transformed into E. coli Rosetta cells. The coding sequence of the feo gene without the initial 73 N-terminal residues fused to a C-terminal GFP tag, referred here as truncated sFeoN::GFP construct, was amplified from the synthesized sFeoFL::GFP construct and recloned into the pET-21a vector using restriction enzymes XhoI/BamHI. Both proteins were produced by induction with 0.5 mM IPTG at 25°C. After 4 h of incubation, the cells were harvested and resuspended in lysis buffer (100 mM K-Hepes, pH 7.4, 500 mM NaCl, 10% glycerol, 0.1% Triton X-100, and 3 M urea, supplemented with protease inhibitors [Roche] and 100 U of DNase type I [NZYTech]). The cells were lysed using a digital sonifier (SLPe; Branson) at 70% amplitude with six pulses of 30 s on/30 s off and clarified by centrifugation at 30,000 g for 45 min at 4°C. For purification of the truncated construct, the supernatant was loaded onto a 5-ml HiTrap Chelating HP (GE Healthcare), charged with 0.1 mM NiCl2, equilibrated with wash buffer (100 mM K-Hepes, pH 7.4, 500 mM NaCl, 10% glycerol, 40 mM imidazole, and 1 mM 2-mercaptoethanol), extensively washed with this wash buffer, and eluted with elution buffer (100 mM K-Hepes, pH 7.4, 500 mM NaCl, 10% glycerol, 500 mM imidazole, and 1 mM 2-mercaptoethanol) throughout a gradient of 6 column volumes. For purification of the full-length construct, the supernatant was loaded onto a 1-ml HiTrap TALON crude (GE Healthcare), charged with 50 mM CoCl2, equilibrated with wash buffer (100 mM K-Hepes, pH 7.2, 500 mM NaCl, 10% glycerol, 5 mM imidazole, 1 mM 2-mercaptoethanol), extensively washed with this wash buffer, and eluted with elution buffer (100 mM K-Hepes, pH 7.2, 500 mM NaCl, 10% glycerol, 150 mM imidazole, and 1 mM 2-mercaptoethanol), throughout a gradient of 20 column volumes. Fractions containing the protein of interest were pooled, and the buffer was exchanged into embryo explant–compatible buffer (100 mM K-Hepes, pH 7.8, 1 mM MgCl2, and 100 mM KCl) using a PD-10 desalting column (GE Healthcare) and concentrated using a 50K MWCO Amicon Ultracentrifugal filter (Merck). The purifications were performed using the ÄKTApurifier protein purification system (Cytiva), and the chromatographic profile of both proteins was followed by measuring the absorbance at 280, 254, and 488 nm in the UV-900 monitor. The size exclusion method resulted in Feo constructs strongly associated with an unknown contaminant at ∼50 kD. The concentration of each construct was estimated as ∼50% of the total measured protein concentration based on band analysis of SDS-PAGE. Total protein concentrations were measured with a NanoDrop2000 UV-Vis spectrophotometer (Thermo Fisher Scientific).
SDS-PAGE, native-PAGE, and blotting with quantification
Polyacrylamide gels for electrophoresis and Coomassie staining were made with 15% SDS. Molecular mass was estimated by linear regression of log10(molecular mass in daltons) as a function of the migration distances (centimeters) of the ladder (161-0375; Bio-Rad). To determine the dimerization of sFeoFL::GFP-His6 and sFeoN::GFP-His6, a Western blot was generated using a PVDF membrane (Bio-Rad) from a 4–12% gradient SDS and a native polyacrylamide gel side by side using a mouse anti-GFP antibody (11814460001; Roche) at 1:500 dilution. The molecular mass of the protein samples was estimated using a linear regression of log10(molecular mass in daltons) as a function of the migration distances (centimeters) of the ladder (MB090; NZYTech); for the SDS gel, a regression was performed on the entire ladder; for the native gel, only the higher molecular mass ladder (100–245 kD) was considered. See Fig. S5.
Addition of exogenous purified proteins
Purified porcine Tubulin (T240; Cytoskeleton) was labeled with Alexa Fluor 647 (Invitrogen, Thermo Fisher Scientific) following a published protocol (Hyman, 1991) and injected into embryos or explants at 0.3–0.8 mg/ml. Freshly purified sFeoFL::GFP-His6 and sFeoN::GFP-His6 were injected at 2 mg/ml in EC buffer in embryos or explants. This concentration was a result from a series of titrations over a magnitude of different concentrations assessing phenotype or rescue. For embryos, the injected volume assumed a spherical shape with diameter D ≈ 0.018 mm, resulting an injection volume of 3.05 × 10−6 mm3. The average length and width of the embryo are 0.5 and 0.2 mm, respectively (Markow et al., 2009). Assuming an ellipsoid geometry for the embryo, its volume is ∼10−2 mm3. Thus, the final concentration of injected protein after equilibration in the entire embryo was 5–6 nM. For explants, both protein constructs were added to explant cytoplasm at 1:200 (vol/vol), resulting in a final concentration in the cytoplasm of 100–200 nM. Importantly, such an excess of full-length Feo protein preserved nuclear divisions and distribution.
Treatment of embryo explants with nocodazole
Nocodazole (M-1404, Sigma Aldrich) was dissolved in DMSO and diluted to 200 µM in electrical conductivity (EC) buffer at pH 7.8, with a concentration of DMSO in EC buffer of 2% (vol/vol). This buffer mixture was added to explants at ∼1:50 (vol/vol), estimated by measuring the diameters of the buffer droplet in the explant and the explant itself, calculating their area, and scaling according to area change (Telley et al., 2013). The final concentration of nocodazole in the explant was ∼4 µM. Control experiments were conducted by adding the buffer mixture without the drug to explants at similar ratios.
Image acquisition, processing, and analysis
Transmission light microscopy images were obtained with a 10× 0.25-NA objective and the polarizer and analyzer of the microscope were in crossed configuration. Time-lapse confocal fluorescence Z stacks were acquired on a Yokogawa CSU-W1 spinning disk confocal scanner with 488-, 561-, and 640-nm laser lines. Images of whole embryos were acquired with a 40× 1.3-NA oil immersion objective. Images of embryo explants were acquired with a 60× 1.2-NA or a 40× 1.15-NA water immersion objective. Images were recorded with an Andor iXon3 888 EMCCD 1024 × 1024 camera with 13-µm pixel2 size and 2× magnification in front of the camera.
Image processing (i.e., making Z-projections, image cropping, image down-sampling, and video generation) was performed in Fiji (Schindelin et al., 2012). Whole-embryo images for knockdown experiments were obtained by pairwise stitching using a plugin in Fiji.
The fluorescence signal of Feo::GFP in explants was analyzed with the line profile tool in Fiji. First, images of dividing nuclei during anaphase or telophase were filtered with a Gaussian kernel Spot-like signals located between nonsibling nuclei were identified and, where spots were noncircular, a line was drawn along the longer axis. The angle of the line relative to the image coordinate system was recorded, and an intensity profile was generated. Profiles were aligned relative to the position of highest intensity and averaged. For each image, an intensity profile from a location void of microtubule signal was generated to obtain the background and the SD of Feo::GFP intensity. Finally, the size of the spot was determined by calculating the width of the curve where the intensity was higher than two times SD of the background. The angle of every profile line was transformed relative to the closer of the two axes that connect the centrosome of one nucleus with the centrosome of the two neighboring sister nuclei (Fig. 1 C), termed angle A probability density plot from all measured angles was generated in Matlab. Voronoi segmentation was performed by determining the position of the spindle poles manually in Fiji and providing these coordinates to the voronoi function implemented in Matlab.
The measurement of internuclear distances in embryo explants was performed manually in Fiji. The precision of distance measurement was ±0.12 µm as determined by repeated measurement (intraoperator variability). The intensity profile plots of Klp3A::GFP in the Feo RNAi background were obtained using the line profile tool in Fiji, by drawing a line between daughter nuclei in the red (H2Av::RFP) channel and generating an intensity profile plot in the green channel, aligning these profiles according to the peak intensity, and averaging profiles from different locations and embryos.
Unless otherwise mentioned, we plotted the distribution of raw data or cumulative probabilities. Sample size (n) and the number of experimental repeats (N) are reported in the figure legends. A Wilcoxon rank-sum test was performed with Matlab starting with a significance level of
Online supplemental material
Fig. S1 provides a detailed analysis of the localization of Feo between neighbor nuclei along subsequent division cycles. Fig. S2 provides a visualization of the GAL4 expression pattern during oogenesis driving the RNAi knockdown, images from embryos expressing two alternative RNAi constructs, and evidence for the delocalization of Klp3A at the spindle midzone when Feo is knocked down. Fig. S3 provides the full dataset of nuclear positions in embryos for the control and test RNAi, together with various distribution parameters. Fig. S4 shows images from an explant before and after micromanipulation, generated from an embryo under Feo knockdown conditions. Fig. S5 provides SDS-PAGE and native-PAGE gels of the full-length and truncated Feo protein construct, time-lapse images from an embryo after injection of the truncated construct FeoN::GFP, and the full dataset of nuclear positions in embryos where endogenous Feo is depleted and the phenotype is partially rescued by injection of full-length Feo::GFP. Video 1 shows fluorescence of Feo, Klp61F, and microtubules in a cycling embryo explant. Video 2 shows fluorescence of Feo, Klp61F, and microtubules in a cycling embryo. Video 3 shows fluorescence of Klp3A, Feo, and microtubules in a cycling embryo. Video 4 shows fluorescence of microtubules and chromatin in a cycling embryo explant after nocodazole treatment. Video 5 shows fluorescence of microtubules and chromatin in cycling embryos under RNAi knockdown. Video 6 shows fluorescence of microtubules and chromatin in cycling embryo explants under RNAi knockdown. Video 7 shows a micromanipulation experiment and subsequent time lapse of fluorescence of Feo. Video 8 shows fluorescence of purified sFeoFL::GFP protein injected in an embryo. Video 9 shows fluorescence of purified sFeoN::GFP protein injected in an embryo.
We thank members of the Telley laboratory for fruitful discussions, Jonathon Scholey (University of California, Davis, Davis, CA) for constructive comments on the manuscript and for fly stocks, and Thomas Surrey for discussions throughout the project. We thank the staff of the Fly Facility, the Advanced Imaging Facility, and the Technical Support Service at the Instituto Gulbenkian de Ciência.
Transgenic fly stocks were obtained from the Vienna Drosophila Resource Center and Bloomington Drosophila Stock Center (National Institutes of Health P40OD018537). We acknowledge financial support provided by Fundação Calouste Gulbenkian, European Commission Seventh Framework Programme Marie Curie Career Integration Grant to I.A. Telley (PCIG13-GA-2013-618743), Human Frontier Science Program Young Investigators’ Grant to I.A. Telley (RGY0083/2016), and a doctoral fellowship from Fundação para a Ciência e a Tecnologia (SFRH/BD/52174/2013) to O. Deshpande, and a project grant (PTDC/BIA-BQM-31843/2017) supporting D. Vieira. We acknowledge Programa Operacional Regional de Lisboa (LISBOA-01-0145-FEDER-007654) supporting Instituto Gulbenkian de Ciência’s core operation, Consortium for Genetically Tractable Organisms (CONGENTO; LISBOA-01-0145-FEDER-022170) supporting the Fly Facility, and Portuguese Platform of BioImaging (PPBI-POCI-01-0145-FEDER-022122) supporting the Advanced Imaging Facility, all co-financed by Fundação para a Ciência e a Tecnologia (Portugal) and Lisboa2020, under the PORTUGAL2020 agreement (European Regional Development Fund).
The authors declare no competing financial interests.
Author contributions: O. Deshpande, J. de-Carvalho, and I.A. Telley conceived and designed the project. O. Deshpande and I.A. Telley designed experiments, and O. Deshpande performed them with occasional support from J. de-Carvalho. D.V. Vieira and O. Deshpande designed, purified, and characterized the protein constructs. I.A. Telley designed and assembled the instrument for explant generation and manipulation. I.A. Telley designed the image analysis procedures. O. Deshpande, J. de-Carvalho, and I.A. Telley analyzed the image data. O. Deshpande, J. de-Carvalho, D.V. Vieira, and I.A. Telley prepared the figures. O. Deshpande and I.A. Telley wrote the draft manuscript, which was edited by all authors.
J. de-Carvalho and D.V. Vieira contributed equally to this paper.