Rab8 directs furrow ingression and membrane addition during epithelial formation in Drosophila melanogaster

The ingression of a plasma membrane furrow is an essential cell shape change that occurs in all eukaryotic organisms. Furrow ingression is also an obligatory step in animal cells during cytokinesis. Understanding how cells are able to dynamically remodel plasma membrane topologies and form a stable furrow in vivo is an unresolved question. The early Drosophila embryo forms thousands of furrows, with some being generated in as little as 4 min (Holly et al., 2015). Fly embryos make both transient and permanent furrows in the course of development. Following fertilization and the fusion of the male and female pronuclei, the zygotic nucleus undergoes 13 rounds of replication in the absence of cell division to generate a single-celled embryo with ∼5000 nuclei (Hartenstein, 1993). The first nine rounds of replication occur deep in the yolk of the embryo; however, at cycle 10, nuclei migrate to the periphery and begin organizing the formation of cytokinetic-like furrows that serve to separate mitotic nuclei in the syncytium. These furrows are transient, forming during each syncytial cycle where they compartmentalize and anchor mitotic spindles, before then regressing. Embryos undergo four rapid rounds of furrow formation (cycles 10-13), followed by a final fifth round of furrow ingression that results in the permanent packaging of nuclei into individual cells and the formation of an epithelium through a process known as cellularization.

Furrow ingression during cellularization shares mechanistic similarities with the earlier syncytial furrows and requires the function of filamentous actin (F-actin), microtubule, and membrane trafficking networks. A contractile actomyosin array directs the basal movement of a bulbous furrow canal, while new membrane delivery is targeted apically (reviewed by Schejter and Wieschaus, 1993; Mazumdar and Mazumdar, 2002; Lecuit, 2004). Microtubules are essential during the early, or slow, phase of cellularization and appear to provide the directional cue that orients the cellularization network (Foe et al., 2000). After the slow phase of cellularization, the network shifts into a fast phase of rapid furrow ingression and membrane addition is targeted to an apicolateral location (Lecuit and Wieschaus, 2000).

A puzzling aspect of furrow formation in the Drosophila embryo has been the absence of identified dynamic membrane compartment behaviors that are directed towards the furrow. The plasma membrane grows ∼30-fold during cellularization, and requires new membrane addition for its growth. Although some furrow membrane is derived from the flattening of apical microvilli, both Golgi and recycling endosomal (RE) pathways have been shown to be required for furrow ingression, and membrane-marking experiments have indicated regions where directed membrane addition is likely to occur (Lecuit and Wieschaus, 2000; Sisson et al., 2000; Pelissier et al., 2003; Riggs et al., 2003; Fabrowski et al., 2013; Figard et al., 2013). However, markers for Golgi and RE compartments, such as Lava lamp (Lva) and Rab11, do not display an association with, or move in a polarized fashion towards, the developing furrow (Sisson et al., 2000; Pelissier et al., 2003; Riggs et al., 2003).

The Rab family of small GTPase proteins are key mediators of membrane trafficking and cytoskeletal dynamics. When in an active GTP-bound state, Rab proteins regulate membrane compartment behaviors through their association with tethering and trafficking effectors (Pfeffer, 2005; Grosshans et al., 2006; Horgan and McCaffrey, 2012), and mutations in Rab proteins are associated with a variety of diseases and developmental disorders (Baskys et al., 2007; Mitra et al., 2011; Recchi and Seabra, 2012; Hardiman et al., 2012). At the plasma membrane, Rab proteins have been shown to bind to the vesicular exocyst tethering complex. During lumen formation and ciliogenesis in mammalian cells, Rab8 binds to the Sec15 subunit of the exocyst complex (Bryant et al., 2010; Knödler et al., 2010; Feng et al., 2012), and the Rab11 RE protein can directly associate with the exocyst subunits Sec5 and Sec15 (Zhang et al., 2004; Beronja et al., 2005; Jafar-Nejad et al., 2005; Langevin et al., 2005; Wu et al., 2005). A newly emerging theme in Rab protein function is that Rab proteins and membrane trafficking pathways can also act through the cytoskeleton (Riggs et al., 2003; Hattula et al., 2006; Cao et al., 2008; Zhang et al., 2009; Peränen, 2011). A complete collection of YFP-tagged constitutively active and dominant-negative, as well as wild-type, lines of all known Drosophila Rab proteins has been constructed (Zhang et al., 2007), making it possible to screen for Rab proteins that are required for early embryonic development. An additional collection of large genomic fragment-driven Rab Gal4s permits the systematic profiling of the cellular and subcellular expression of Rab GTPases (Chan et al., 2011).

Within the Rab family, Rab8 has been identified as associating with late exocytic vesicles. This protein has been demonstrated to direct polarized trafficking events during ciliogenesis, neuronal development, regulated secretion and cytokinesis (Yoshimura et al., 2007; Nachury et al., 2007; Huber et al., 1995; Chen et al., 2001; Kaplan and Reiner, 2011), possibly through the ability to reorganize actin and microtubules (Peränen et al., 1996). Defects in Rab8 function have been implicated in trafficking diseases, including retinal degeneration, cell polarity defects and irregular tissue formation (Moritz et al., 2001; Satoh et al., 2007; Huber et al., 1995). Here, we show that Rab8 demonstrates dynamic behaviors that are specifically associated with furrow ingression in the early Drosophila embryo. Rab8 transitions from an intracellular punctate localization to an association with the apicolateral plasma membrane, and furrow ingression is abolished when Rab8 function is compromised. We also use CRISPR/Cas9 techniques to engineer a GFP:Rab8 fusion at the endogenous Rab8 locus. Our results suggest a model in which Rab8 cooperates with cytoplasmic Rab11 and Golgi stores to direct the targeted membrane delivery that is required for furrow ingression via interactions with the exocyst tethering complex.

We performed an initial screen to determine which of the Drosophila Rab proteins is involved in early developmental processes. Either GDP dissociation-deficient, dominant-negative (DN) or GTPase-deficient, constitutively active (CA) Rab proteins were expressed in embryos and scored under oil for tissue-level defects in early embryonic development (Fig. S1). From this screen, we found that expression of Rab8 DN produces the strongest disruption of early embryogenesis and resulted in embryos with an uneven epithelium and a loss of later gastrulation movements (data not shown), suggesting that Rab8 is required in early morphological processes.

These results led us to examine the dynamics of Rab8 in early embryos. Embryos expressing YFP:Rab8 driven from a genomic Rab8-Gal4 were imaged using a spinning disk microscope. Prior to the movement of nuclei to the embryonic periphery at cycle 10, Rab8 exists in large, relatively static puncta (Movie 1). However, once furrow formation is initiated in the syncytial divisions, Rab8 displays striking behaviors (Fig. 1A, Movie 2). As the syncytial cycle proceeds, Rab8 puncta become more numerous (Fig. 1A,A′). Rab8 then transitions to an association with the plasma membrane, and the cytoplasmic punctate population becomes depleted (Fig. 1A″). The association with the cell periphery is rapid and occurs within a few minutes (Fig. 1A-A″), which then persists through the remainder of the division cycle (Fig. 1A‴).

To identify how changes in Rab8 populations are correlated with the cell cycle, we analyzed Rab8 localization in fixed embryos stained for YFP:Rab8 and DNA (Fig. 1B-E). During interphase of syncytial cycles 10-13, Rab8 exists in a punctate population (Fig. 1B), but then transitions to an association with the plasma membrane at the onset of prophase and mitosis (Fig. 1C). These structures persist through the mitotic division (Fig. 1D) and return to a punctate population at the onset of the next interphase (Fig. 1E). These results suggest that Rab8 behaviors are cell cycle regulated and undergo a transition from cytoplasmic membrane compartments to the cell periphery that correlates with the time period when syncytial furrow formation occurs.

Similar to the earlier formation of syncytial furrows, Rab8 shows a transition from punctate structures to a hexagonal, cell periphery-associated array during cellularization. At early cellularization stages, Rab8 exists in punctate structures (Fig. 1F, Movie 3). As cellularization proceeds, these puncta become more numerous (Fig. 1F′,G). Rab8 then makes a transition to a tight peripheral array (Fig. 1F″). The cytoplasmic puncta become depleted and the association with the plasma membrane persists throughout the remainder of cellularization (Fig. 1F‴,G). This transition from punctate structures to a peripheral association also represents a transition from the slow phase (Fig. 2A,B) to the fast phase (Fig. 2C,D) of cellularization, which is marked by the F-actin furrows reaching the basalmost portion of the nuclei (Fig. 2C). During the fast phase, furrow ingression rapidly increases and membrane addition is targeted to an apicolateral location (Lecuit and Wieschaus, 2000). Rab8 localizes to this area of rapid membrane addition (Fig. 2C,D). Additionally, the similarity in Rab8 behavior during both syncytial divisions and cellularization suggests that both these processes might use Rab8-dependent machinery to initiate furrow formation.

In order to determine whether the dynamics displayed in UASp-YFP:Rab8 embryos accurately reports the behaviors of endogenous Rab8 protein, we utilized CRISPR-HDR to insert eGFP into the Drosophila Rab8 genomic locus (Baena-Lopez et al., 2013; Bassett et al., 2013; Gratz et al., 2013; Gokcezade et al., 2014). An HDR donor plasmid was constructed by cloning the 1.5 kb of genomic DNA upstream (5′) of the Rab8 ATG start site into the pBluescript KS plasmid, followed by the coding sequence for eGFP and a 1.5 kb 3′ homology arm starting at the beginning of the Rab8 coding sequence. Two guide RNAs (gRNA) were designed that cut 41 bp 5′ and 16 bp 3′ of the Rab8 ATG start site. The HDR donor plasmid and gRNAs were injected into embryos and successful CRISPR-mediated insertions were identified by fluorescence microscopy and genomic PCR (Fig. 2E-L, Fig. S2A,B). The generated CRISPR GFP:Rab8 line is homozygous viable and fertile, indicating that CRISPR GFP:Rab8 is fully functional. Insertion of eGFP at the genomic Rab8 locus reveals that endogenous Rab8 behaviors are similar to those of the UASp-YFP:Rab8 transgene: Rab8 initially localizes to a mobile, cytoplasmic punctate population and transitions to an association with the plasma membrane during the formation of both syncytial and cellularization furrows (Fig. 2E-L).

Both fixed and live imaging of Rab8 structures in the early embryo demonstrate dynamic behaviors and a strong association with the ingression of membrane furrows. This led us to examine the effects of Rab8 knockdown on furrow formation. We performed knockdown experiments using RNAi targeted against Rab8. Double-stranded (ds) RNA was introduced by microinjection into an otherwise wild-type (WT) background and embryos were scored for defective phenotypes (Fig. 3A). Three different dsRNAs directed against Rab8 were used, and all dsRNAs demonstrated a severe disruption in embryonic development (Fig. 3). Importantly, the majority of dsRNA-injected embryos arrested at stages requiring furrow ingression; 40% of embryos injected with SNAPDRAGON-designed Rab8 dsRNA arrested during syncytial divisions (Fig. 3A, n=89), while 56% arrested during cellularization (Fig. 3A).

We also used the endogenous CRISPR GFP:Rab8 to address Rab8 function by co-expressing either deGradFP or GFP TRiP shRNAi in a homozygous CRISPR GFP:Rab8 background (Fig. S3). Again, a deep disruption of development occurred in the early embryonic stages during which furrow formation is highly active, with 72% of deGradFP-expressing embryos at 18°C arresting during the syncytial divisions (Fig. S3A). It is interesting to note that deGradFP expression produces a stronger defect than GFP TRiP; a similar result has been observed by Nagarkar-Jaiswal et al. (2015).

To examine the early arrest defects more closely, dsRNA was introduced by microinjection into embryos expressing Histone:RFP and GFP:actin-binding-domain of Moesin (MoeABD). This allowed the visualization of chromosomal behaviors and the ingressing furrows concurrently during development. As compared with control-injected embryos, Rab8 RNAi had a dramatic effect on the formation of furrows. In control-injected [Rhodopsin 3 (Rh3) dsRNA] embryos, syncytial furrows form and ingress during cycles 10-13 to prevent chromosomal mixing and provide spindle anchor points during nuclear divisions (Fig. 3B). In Rab8 knockdown embryos, no observable furrow formation occurred and fusions between adjacent nuclei were observed (Fig. 3C). The cortical F-actin that would normally associate apically with ingressing furrows is completely disrupted in these severely affected embryos (Fig. 3C). Nuclei aggregated following aberrant division cycles and subsequently dropped from their subcortical positions (Fig. 3C). Similar defects in furrow formation were observed in CRISPR GFP:Rab8 embryos expressing deGradFP or GFP TRiP shRNAi (Fig. S3B,C). These knockdown experiments demonstrate that Rab8 has a vital role in directing the formation and ingression of furrows in the early embryo.

Previous reports have shown the functional requirement of REs and the Golgi during furrow ingression in the early embryo (Sisson et al., 2000; Pelissier et al., 2003; Riggs et al., 2003; Cao et al., 2008). Given the transition that Rab8 exhibits from cytoplasmic puncta to an association with the plasma membrane, we posited that Rab8 might link internal membrane stores with delivery to the cell surface. This led us to determine whether Rab8 structures are coincident with the RE or Golgi. Immunostaining against CRISPR GFP:Rab8 and Rab11, an RE marker, show that these proteins are largely present in distinct puncta during early cellularization (Fig. 4A-C). However, a small percentage (∼20%) of cytoplasmic puncta show colocalization of Rab8 and Rab11 (Fig. 4A-C, Fig. S4). This is seen in the distance plots as a small increase in Rab8 at Rab11 maxima and vice versa (Fig. 4B,C). This coincidence increases as embryos enter fast phase, a period of rapid membrane addition (Fig. 4D, Fig. S4) (Lecuit and Wieschaus, 2000). During this period, Rab11 puncta increase in size and accumulate in an apical position above the nuclei. Rab8 often appears to colocalize in small subdomains of larger Rab11 structures (Fig. 4D). Indeed, when only larger Rab11 structures (10-26 pixels, pixel size=0.056 µm) are quantitated, the number of larger Rab11 aggregates that possess a subdomain of coincident Rab8 rises to 64% (Fig. S4). These results suggest that there might be a transient interaction between Rab11 and Rab8 compartments, and that this interaction is more prevalent during mid to late cellularization stages when cytoplasmic membrane stores are mobilized.

Rab8 also forms a distinct compartment from the Golgi. Rab8 vesicle populations show little association with Golgi compartments during early cellularization (Fig. 4G-I, Fig. S4). However, again during the fast phase of cellularization, the percent coincidence with the Golgi marker Lva rises from 1-2% to ∼11% (Fig. 4J-L, Fig. S4). This can be seen in the distance plots as a mild depletion of Rab8 compartments distant from Lva maxima, and the mild increase in Rab8 compartments near Lva maxima (Fig. 4K,L). Once again, the slight colocalization between Rab8 puncta and the Golgi would be consistent with a possible transient interaction between Rab8 and the Golgi that occurs with higher frequency in embryos undergoing the fast phase of cellularization.

The transition of Rab8 to the cell periphery during furrow ingression and the association of Rab8 with internal membrane stores, such as Rab11, and the Golgi led us to ask if Rab8 is responsible for delivering the membrane that is necessary for furrow ingression. To examine whether Rab8 regulates membrane addition during furrow ingression, we performed Rab8 RNAi in a Gap43:mCherry background. Gap43:mCherry is a membrane marker that consists of a 20 amino acid peptide containing a double palmitoylation sequence that directs the mCherry protein to furrows and the apical plasma membrane (Martin et al., 2010; Holly et al., 2015) (Fig. 5A, Movie 4). Upon knockdown of Rab8, trafficking of Gap43:mCherry to the cell surface is severely disrupted (Fig. 5B, Movie 5). Gap43:mCherry remains in cytoplasmic puncta and does not traffic to the cell surface when Rab8 function is disrupted (Fig. 5B). These results suggest that Rab8 indeed functions as a regulator of exocytic trafficking of membrane during furrow formation.

To examine how the GTP state of Rab8 directs its activity, we examined the localization of GTPase or GDP dissociation-deficient mutants. We hypothesized that GTP-bound (CA) Rab8 represents the active population and would associate with those compartments in which Rab8 is directing membrane dynamics. Likewise, GDP-bound (DN) Rab8 would most likely represent inactive pools of the protein. We performed live imaging and immunostaining against these activity-deficient mutants of Rab8 in order to examine how the GTP state affects the localization of Rab8 in relation to other compartments involved in furrow ingression.

We expressed CA or DN versions of YFP-tagged Rab8 in embryos to examine changes in protein behavior over time. Immunostaining revealed that Rab8 CA (GTPase deficient) forms similar structures to Rab8 WT (compare Fig. 6A,B with Fig. 2C,D). However, the association with the cell periphery arises earlier, during the slow phase of cellularization, as opposed to during the fast phase in WT embryos (compare Fig. 6B with Fig. 2B). Immunostaining of Rab8 DN (GDP dissociation deficient) revealed large aggregates that localize to the apical cytoplasm (Fig. 6C). Very low levels of Rab8 DN can also be seen at the plasma membrane when stained embryos are imaged at high gain settings (Fig. 6D).

Effects of GTPase activity on Rab8 behavior were also examined in whole-mount embryos expressing either Rab8 CA or Rab8 DN over time. Again, Rab8 CA showed similar behaviors to WT. During early cellularization, Rab8 CA exists in punctate populations before transitioning to the cell periphery, where it persists until the end of cellularization (Fig. 6E-E‴). Expression of Rab8 DN in whole embryos shows the progressive appearance of apical cytoplasmic aggregates, which grow and persist throughout the cellularization process (Fig. 6F-F‴). The above results suggest that active, GTP-bound Rab8 localizes to puncta and associates with the plasma membrane, whereas inactive, GDP-bound Rab8 is cytoplasmic.

We also examined how GTPase activity affects the localization of Rab8 in relation to the RE and Golgi. Rab8 CA forms compartments distinct from both the RE and Golgi (Fig. 6G,J). Rab8 DN also does not appear to associate with RE or Golgi compartments (Fig. 6H,I,K,L). However, Rab11 compartments are enlarged in embryos expressing Rab8 DN when compared with those in Rab8 CA and WT (Fig. 6H,I). This enlargement is apparent both in apical planes (Fig. 6H) and more basal sections (Fig. 6I). These results would be consistent with a requirement for Rab8 in the trafficking of membrane from the RE to the ingressing furrow, and further support a possible interaction between the two compartments.

As the exocyst complex is required for targeted membrane addition during furrow formation in the early embryo (Murthy et al., 2010; Holly et al., 2015) and Rab8 can bind to the Sec15 subunit of the exocyst during lumen formation and ciliogenesis (Bryant et al., 2010; Knödler et al., 2010; Feng et al., 2012), we asked if exocyst complex function is required for Rab8 dynamics in the early embryo. RNAi against Sec5 was performed in a YFP:Rab8 background and imaged over time. These experiments revealed a failure of Rab8 to transition from cytoplasmic puncta to a peripheral association (Fig. 7B-B‴), as compared with controls (Fig. 7A-A‴), suggesting that Rab8 might utilize the exocyst to properly localize to the plasma membrane.

The RE has previously been implicated as being required for membrane addition during cellularization (Pelissier et al., 2003). We performed RNAi against Rab11 to ask if disruption of the RE results in changes in Rab8 protein behavior. Similar to Sec5 knockdown, disruption of Rab11 led to a loss of Rab8 transitioning to an association with the plasma membrane (Fig. 7C-C‴), consistent with Rab8 requiring Rab11 as a source of membrane or a Rab11-recruited GEF activity. Interestingly, when RNAi of Rab8 is performed in a YFP:Rab11 background, compared with controls (Fig. 7D), Rab11 puncta are present (Fig. 7E) but appear to grow to become slightly enlarged over time (Fig. 7E″,E‴, yellow arrowhead). Taken together, these data suggest that Rab8 requires both exocyst complex and RE function in order to traffic to the cell surface. It is possible that Rab8 requires the exocyst as a targeting and tethering cue in order to properly associate with the plasma membrane, while utilizing Rab11 as a source of membrane that permits Rab8 activity and trafficking to the cell cortex. The slight enlargement of RE puncta in Rab8 DN and Rab8 RNAi backgrounds also strengthens the possibility that Rab8 functions to mediate the delivery of membrane from RE stores to the plasma membrane during furrow ingression.

Here, we identify Rab8 as a crucial component of a trafficking pathway that directs furrow ingression in the early Drosophila embryo. Rab8 demonstrates striking dynamic behaviors in the early embryo, existing in small punctate structures that then transition to an association with the plasma membrane. This transition correlates with the ingression of syncytial furrows as well as the onset of the fast phase during cellularization, suggesting that a common membrane trafficking mechanism is required during each of these processes. We have also previously observed that syncytial furrows ingress at a similar rate as that reported for the fast phase (Holly et al., 2015). Peripheral Rab8 is specifically found in an apicolateral region on cellularizing furrows, which is the region of the furrow to which membrane addition is known to occur (Lecuit and Wieschaus, 2000; Sisson et al., 2000; Pelissier et al., 2003), and furrow formation fails when Rab8 function is disrupted. Additionally, when Rab8 function is disrupted, exocytic delivery to the cell surface fails and cargo proteins remain in cytoplasmic puncta. These results suggest a model in which Rab8 directs targeted membrane addition to furrows, which permits furrow ingression (Fig. 8).

Previous studies have shown that membrane addition is required during cellularization and is provided through RE and Golgi trafficking pathways (Sisson et al., 2000; Pelissier et al., 2003; Riggs et al., 2003). However, a confounding aspect of these known membrane pathways is that they do not display an association with, or movement to, ingressing furrows. Immunostaining against either the RE or Golgi and Rab8 indicated that Rab8 largely exists in distinct compartments. However, quantification of colocalization revealed that Rab8 displays a small, but reproducible, coincidence with the RE and Golgi (10-20% of compartments quantified during fast phase), and Rab8 appeared to mark subdomains of larger Rab11 compartments. It is therefore possible that these populations have transient interactions with each other. Indeed, Rab trafficking pathways often function through compartments that change their identity by dissociating with one Rab and acquiring another (Rink et al., 2005), and Rab11 has been demonstrated to bind and stimulate a Rab8 GEF, Rabin8, during ciliogenesis (Knödler et al., 2010). We also observed that disrupting the function of Rab11 resulted in the failure of Rab8 to transition from the cytoplasmic puncta to the plasma membrane, and that Rab8 puncta appeared to be stationary and inactive. Conversely, when Rab8 function was disrupted, either in Rab8 DN embryos or by Rab8 RNAi, a mild enlargement of the Rab11 RE was observed. These results would be consistent with Rab8 serving to connect membrane cytoplasmic stores to the ingressing furrow (Fig. 8B).

Rab8-directed membrane trafficking to the plasma membrane might occur through its binding and tethering of vesicles to the exocyst complex (Fig. 8C). During lumen formation and ciliogenesis in mammalian cells, Rab8 binds to the Sec15 subunit of the exocyst complex (Bryant et al., 2010; Knödler et al., 2010; Feng et al., 2012). This function and interaction might be preserved during cellularization, as the Sec5 subunit of the exocyst complex is required for cellularization, and Sec5 localizes to the apicolateral region of the cellularizing furrow in the area where membrane addition and Rab8 localization occur (Murthy et al., 2010). Indeed, disruption of Sec5 by RNAi leads to a failure of Rab8 to transition to a cortical association. RalA GTPase function, an important factor in directing exocyst complex assembly, is also required for Rab8 localization to the cell cortex (Moskalenko et al., 2002, 2003; Fukai et al., 2003; Holly et al., 2015). Additionally, disruption of the Exo84 exocyst subunit in the fly embryo leads to an enlargement of Rab11 REs, which resembles the defects observed in Rab8 DN-expressing embryos (Blankenship et al., 2007). In summary, Rab8 might direct vesicular traffic from the RE/Golgi to exocyst tethering complexes on the ingressing furrow (Fig. 8).

Syncytial furrow formation and the cellularization network are, essentially, specialized forms of cytokinetic furrows (Schejter and Wieschaus, 1993; Sullivan and Theurkauf, 1995). It is therefore interesting to contrast our findings with those for classical cytokinetic furrows. It is clear that ingressing furrows in both cytokinetic cells and the early fly embryo require membrane trafficking pathways as part of the furrowing mechanism (Lecuit and Wieschaus, 2000; Albertson et al., 2005; Prekeris and Gould, 2008). However, syncytial furrow formation is reported to be independent of Myosin II function, whereas cytokinetic furrows and cellularization require the activation of Myosin II motor proteins (Warn et al., 1990; Kiehart, 1990; Young et al., 1991; Royou et al., 2004). Rab8 localizes to, and is required for, both syncytial furrows and the cellularization network. It is therefore possible that rapid membrane addition through a Rab8-dependent mechanism may provide an important portion of the additional surface area that the deeply ingressing furrows in the early embryo require. Further, it might be that targeted membrane addition provides some of the force for the ingression of syncytial furrows that lack Myosin II function, although it is also clear that membrane ingression in the early embryo additionally requires the function of the F-actin cytoskeleton (Sokac and Wieschaus, 2008a,b; Yan et al., 2013; Reversi et al., 2014). A requirement for Rab8 function at cytokinetic furrows has been less clear, although it has been shown that Rab8 vesicles are transported along microtubules to the midbody (Pohl and Jentsch, 2008), and siRNAs directed against RAB8 that produce binucleate cells were identified in a systematic RNAi screen for proteins that are required for mitosis in human cells (Neumann et al., 2010). It is interesting to note that several Rab proteins (RAB13, RAB15, RAB36) that exhibit high similarity in either identity or function to Rab8 exist in the human genome but do not appear to be present in Drosophila (Chan et al., 2011). A focus for further studies will be a teasing apart of the contributions of Rab8-dependent membrane trafficking and F-actin and Myosin II function in directing cleavage furrow ingression.

Flies were maintained at 25°C by standard procedures. UAS transgenic flies were crossed to matαTub-Gal4VP16 67C;15 (D. St. Johnson, Gurdon Institute, Cambridge, UK) maternal driver females or Rab8>Gal4 [Fig. 1; Bloomington Drosophila Stock Center (BDSC)] and second generation embryos were analyzed. UAS-deGradFP; CRISPR:GFP:Rab8 or UAS-GFP TRiP; CRISPR:GFP:Rab8 were crossed to matαTub-Gal4VP16 67C; CRISPR:GFP:Rab8 and second generation embryos were analyzed.

Fly stocks: UASp-YFP:Rab8, UASp-YFP:Rab8T22N, UASp-YFP:Rab8 Q67L, UASp-YFP:Rab11, sqh-Histone:RFP, UASp-deGradFP 58740, UAS-GFP TRiP Valium 20 41555 (all BDSC); sqh-GFP:MoesinABD (Kiehart et al., 2000); sqh-mCh:MoesinABD (Millard and Martin, 2008); and Gap43:mCherry (Martin et al., 2010).

To create transgenic lines of CRISPR GFP:Rab8, a homology-directed repair (HDR) donor plasmid was generated comprising the genomic DNA 1.5 kb upstream of the Rab8 ATG start site cloned into pBluescript KS (Addgene), followed by the coding sequence of GFP and the 1.5 kb 3′ of the Rab8 ATG start site. In total, 3 kb of the genomic Rab8 locus was cloned and inserted into the pCasper4 plasmid (Drosophila Genomics Resource Center, stock #1213), along with the 717 bp coding sequence of eGFP, using standard PCR and cloning protocols to create the HDR donor plasmid. Two gRNA sequences that serve to guide the Cas9 nuclease to specific genomic locations were designed according to Gokcezade et al. (2014) and inserted into pU6-BbsI-gRNA (AddGene stock #45946). All constructs were sequence verified (UC Denver Core Sequencing).

gRNAs (Table S1; 100 ng/µl) and GFP:Rab8 plasmid (150 ng/µl) were co-injected into Cas9-expressing flies (y, nos-Cas9, w; BDSC, #54591) by BestGene. Surviving adults were crossed to Sp/Cyo; Dr/Tm3 flies in a series of crosses to create possible insertion lines. Embryos from these individual lines were tested for fluorescence using a spinning disk confocal microscope, and insertion into the genome was verified by PCR (see below). Successful insertions were relatively rare: 270 lines were scored before recovery of CRISPR GFP:Rab8. Cloning primers for HDR donor GFP:Rab8 are listed in Table S1.

PCR primers (see Table S1) to verify correct insertion of GFP into the Drosophila genome via CRISPR were designed in the following manner. The 5′ primer was targeted against a genomic region outside of that which was cloned into the HDR donor GFP:Rab8 plasmid (1.7 kb upstream of Rab8 start codon), while the 3′ primer was targeted against a region within the GFP coding sequence. Thus, a PCR product could only be produced when eGFP has been integrated into the genomic Rab8 locus. Three primer pairs were designed in this manner (NIH, Primer-BLAST) and PCR was performed in both CRISPR GFP:Rab8 flies and control OreR flies.

Confocal images were acquired on an Olympus Fluoview FV1000 confocal laser scanning microscope with 40×/1.35 NA or 60×/1.42 NA objective for fixed specimens. Time-lapse imaging was performed on a CSU10b Yokogawa spinning disk confocal from Zeiss and Solamere Technologies Group with 40×/1.3 NA or 63×/1.4 NA objective. Embryos were imaged after dechorionation and placement on a gas-permeable membrane in Halocarbon 27 oil. Live imaging was performed using exposure settings of 150-500 ms and images were acquired at either fast (<1/s) or slow (1/30 s) rates (see figure and movie captions).

Embryos were dechorionated in 50% bleach solution and fixed for 1 h 15 min at the interface of heptane and 3.7% formaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) before being manually devitellinized and stained with Hoechst (1:500; Sigma), Alexa 546 phalloidin (1:200; Molecular Probes), rabbit anti-GFP (1:1000; Invitrogen, A11122), mouse anti-GFP (1:100; Molecular Probes, A11120), rabbit anti-Lva (1:250; Sisson et al., 2000), rabbit anti-Rab11 (1:1000; Satoh et al., 2007) or mouse anti-Sec5 (1:35; Murthy et al., 2010). Secondary antibodies conjugated to Alexa 488, Alexa 546 or Alexa 568 were used at 1:500 (Molecular Probes). Embryos were mounted in Prolong Gold with DAPI (Molecular Probes).

Confocal and spinning disk images were edited using Adobe Photoshop. Channels for fixed images were uniformly leveled for optimal channel appearance. Time-lapse images were leveled with Photoshop to show optimal protein populations, or were processed with identical settings to permit comparisons of protein levels.

Automated quantitation was performed as follows. In each color channel, local intensity maxima were detected using a wavelet-based detection method as used by Loerke et al. (2009), which is based on à trous detection first proposed by Olivo-Marin (2002). For each channel, a Euclidian distance map was then calculated with respect to the detected maxima. Using this distance map in combination with the raw intensity images, we interpolated the average intensity-versus-distance function (using Gaussian smoothing with σ=0.5 pixels) from all pixels in the image. This interpolation was performed for the intensity and the distance map from the same channel (e.g. red intensity versus distance from red maxima), but also for the combination of intensity and the distance map from the other channel (e.g. red intensity versus distance from green maxima). In addition, we calculated a distance map for randomly placed maxima positions (which was performed in practice by switching the x and y coordinates of the true maxima) that removes all spatial correlation of the intensity signal. The error bars shown in the figure represent the standard deviation of cell-to-cell variation of the intensity function for a given condition.

The associated P-value is that of a two-sided t-test between the other channel intensity function and the random position intensity function (both at distance zero), i.e. the P-value tells us whether the measured colocalization of the other channel intensity function is significantly different from a spatially random (i.e. uncolocalized) intensity distribution. A minimum of ten individual images collected from at least five different embryos were analyzed for each stage and immunostain.

Distance maps can give a probabilistic determination of coincidence but cannot distinguish objects and give a segmented analysis of percent colocalization other than on a pixel-by-pixel basis. We therefore additionally used manual identification of individual Rab8, Rab11 and Lva puncta. Manual quantification of individual objects and coincidence between Rab8 vesicle populations and Rab11 RE and Golgi populations was performed using Adobe Photoshop. Quantification was performed by selecting puncta between 2 and 26 pixels (pixel size=0.056 µm) in size in set areas of 5×5 grids. The selected puncta were then overlaid with the opposing channel and colocalizing puncta counted. Average colocalization was found by performing a weighted average calculation from a minimum of six individual images collected from three different embryos and standard deviations were plotted.

Manual vesicle counts of YFP:Rab8 during syncytial divisions or cellularization stages were performed in the same manner as the colocalization counts described above. Still frames from the onset of a syncytial stage/slow phase, prior to cortical transition, first appearance of cortical array and persistence of cortical array were chosen and all vesicles were counted from a 512×512 pixel image.

Primers for siRNA treatments were chosen to represent independent regions of Rab8 (5′UTR) or through the use of the SNAPDRAGON RNAi design program that bioinformatically selects against off-target effects (Drosophila RNAi Screening Center). siRNA primers for Rh3, Rab11 and Sec5 were also designed using the SNAPDRAGON system (termed SNAP). dsRNA was made using a Megascript T7 Transcription Kit (Ambion) and purified using Qiagen RNeasy columns. Final concentration was determined with a NanoDrop ND1000 spectrophotometer.

Embryos were prepared in the same manner as for live imaging and then glued to a coverslip using heptane glue. Embryos were dehydrated for ∼12 min, and then injected with siRNA against Rh3 (13 µM), Rab8 (550 bp, 13 µM; 5′UTR, 11 µM; SNAP, 13 µM), Rab11 (13 µM) or Sec5 (18 µM).

Embryos were collected on apple juice agar plates, and development was followed and scored under Halocarbon 27 oil on a Stemi 2000 Zeiss dissecting microscope with transillumination. Pooled data from multiple trials were summed, and RNAi scoring data (Fig. 3) were tested for statistical significance using a two-dimensional contingency table with a χ² test with α=0.05. Tests were performed in two ways: (1) all data were tested to determine if changes in the proportions of phenotypes seen were significant, which resulted in P<0.0001; (2) both RNAi treatment groups were tested against the Rh3 control treatment to determine if the changes in the proportions of phenotypes were significant in relation to control proportions. All tests gave P<0.0001.

We thank the generous colleagues who provided antibodies and fly lines: Hugo Bellen (Houston, TX, USA), Don Ready (West Lafayette, IN, USA), John Sisson (Austin, TX, USA), Adam Martin (Cambridge, MA, USA), Daniel Kiehart (Durham, NC, USA), Tom Millard (Manchester, UK) and Tom Schwarz (Cambridge, MA, USA). We gratefully acknowledge the work of Cayla Jewett in pioneering CRISPR and deGradFP techniques in the lab. Thanks are due to Dinah Loerke (Denver, CO, USA), Alan Zhu (Cleveland, OH, USA) and the J.T.B. lab for suggestions and critical reading of the manuscript.