Polarization of the epithelial layer and apical localization of integrins are required for engulfment of apoptotic cells in the Drosophila ovary

ABSTRACT Inefficient clearance of dead cells or debris by epithelial cells can lead to or exacerbate debilitating conditions such as retinitis pigmentosa, macular degeneration, chronic obstructive pulmonary disease and asthma. Despite the importance of engulfment by epithelial cells, little is known about the molecular changes that are required within these cells. The misregulation of integrins has previously been associated with disease states, suggesting that a better understanding of the regulation of receptor trafficking could be key to treating diseases caused by defects in phagocytosis. Here, we demonstrate that the integrin heterodimer αPS3/βPS becomes apically enriched and is required for engulfment by the epithelial follicle cells of the Drosophila ovary. We found that integrin heterodimer localization and function is largely directed by the α-subunit. Moreover, proper cell polarity promotes asymmetric integrin enrichment, suggesting that αPS3/βPS trafficking occurs in a polarized fashion. We show that several genes previously known for their roles in trafficking and cell migration are also required for engulfment. Moreover, as in mammals, the same α-integrin subunit is required by professional and non-professional phagocytes and migrating cells in Drosophila. Our findings suggest that migrating and engulfing cells use common machinery, and demonstrate a crucial role for integrin function and polarized trafficking of integrin subunits during engulfment. This study also establishes the epithelial follicle cells of the Drosophila ovary as a powerful model for understanding the molecular changes required for engulfment by a polarized epithelium. Summary: Apical integrin localization, mediated by polarized and directed trafficking, is crucial for proper engulfment by epithelial cells.


INTRODUCTION
Epithelial cells, such as the retinal pigment epithelium (RPE) and bronchial epithelial cells, are required for engulfment on a daily basis. Engulfment in an epithelial layer requires substantial changes, all of which must be tightly regulated. However, very little is known about the molecular changes required for an epithelial layer to initiate engulfment. Integrins are αβ heterodimeric receptors that are crucial for phagocytosis in both professional and non-professional phagocytes in mammals (Finnemann and Silverstein, 2001;Sexton et al., 2001;Hsieh et al., 2012;Flannagan et al., 2014). Mammals have 18 αand 8 β-subunits, but only use one specific α-subunit, αν, for engulfment by professional and non-professional phagocytes. Interestingly, in Caenorhabditis elegans, two different α-subunits are used for engulfment (Hsu and Wu, 2010;Hsieh et al., 2012). In Drosophila, there are five αand two β-subunits. Recently, the integrin alpha-PS3 (αPS3)/βν heterodimer was shown to be required in professional phagocytes for the engulfment of bacteria and apoptotic cells (Shiratsuchi et al., 2012;Nonaka et al., 2013), but a role for integrins in non-professional phagocytes in Drosophila has not yet been demonstrated.
One example of an epithelium that is required for engulfment is the mammalian retinal pigment epithelium (RPE), cells of which engulf photoreceptor outer segments via their apical surface (Kevany and Palczewski, 2010). ανβ5, the integrin heterodimer associated with phagocytosis, is apically localized in RPE cells (Finnemann et al., 1997;Nandrot et al., 2008), despite basal localization of integrins in most cells. This suggests that heterodimer localization might be important for function. However, it is currently unknown how integrin heterodimers are asymmetrically localized within an engulfing cell.
One possibility is that cellular polarity is important for integrin trafficking, as it often is in migrating cells. The RPE cells are highly polarized throughout development and engulfment (Marmorstein, 2001), suggesting that this might be the case. Interestingly, the epithelial follicle cells (FCs) in the Drosophila ovary are also highly polarized (Tanentzapf et al., 2000;Morais-de-Sa et al., 2010;Fletcher et al., 2012) and engulf apoptotic debris via their apical side (Giorgi and Deri, 1976;Mazzalupo and Cooley, 2006;Tanner et al., 2011;Etchegaray et al., 2012). A second possibility is that integrin heterodimers are trafficked in a directed fashion. We tested both of these possibilities in this study and found that both play a role in engulfment and integrin trafficking.
The Drosophila ovary serves as a model of inducible engulfment by epithelial cells, which can provide insight into how a phagocytic state is activated in non-professional phagocytes Tanner et al., 2011;Etchegaray et al., 2012;Pritchett and McCall, 2012). Here, we demonstrate that integrins are required for engulfment by FCs and we show how they are apically trafficked in an adherent, epithelial cell layer. We found that integrin heterodimer localization and function is largely directed by the α-subunit, and apical localization of the αPS3/ integrin beta-PS (βPS) heterodimer is required for engulfment. We found that many of the genes required for integrin trafficking in migratory cells are also required during engulfment, suggesting that migrating and engulfing cells might share common machinery. Moreover, our findings suggest that the Drosophila ovary might serve as an excellent in vivo model for integrin trafficking and function within the polarized RPE cells: our results indicate a high degree of similarity between these two tissues. Thus, the information gained here about the molecular changes within an engulfing epithelium might provide valuable insight into treatment for diseases such as macular degeneration and retinitis pigmentosa.

RESULTS
The integrin subunits αPS3 and βPS become apically enriched in engulfing FCs Starvation-induced cell death occurs specifically during stages 7-9 of oogenesis (Giorgi and Deri, 1976), and five phases of death have been defined based on the progressive condensation and fragmentation of apoptotic germline-derived nurse cell nuclei (Etchegaray et al., 2012) (Fig. 1A-D). In healthy egg chambers, nurse cell chromatin is dispersed (Fig. 1A), but becomes disorganized in very early dying, or phase 1, egg chambers (Fig. 1B). By phase 3 of death, the nurse cell chromatin has become highly condensed in balls (Fig. 1C) and, by phase 5 of death, most nurse cell chromatin has been removed (Fig. 1D). Few hemocytes circulate within the ovary (King, 1970) and engulfment of germline debris is carried out by adjacent epithelial FCs (Giorgi and Deri, 1976).
We previously showed that the engulfment receptor Draper and the JNK pathway are required for engulfment by FCs (Etchegaray et al., 2012). In other systems, Draper and its orthologs work in parallel to other engulfment genes, including integrins (Nagaosa et al., 2011;Hsieh et al., 2012;Kim et al., 2012;Shiratsuchi et al., 2012). To investigate whether integrins participate during engulfment in the ovary, we analyzed the expression of integrin subunits during engulfment utilizing previously made antibodies (Brower et al., 1984;Wada et al., 2007;Nagaosa et al., 2011). It has previously been shown that αPS1, αPS2 and βPS are expressed throughout oogenesis, whereas αPS3, αPS4 and αPS5 become transcriptionally upregulated during late oogenesis (Dinkins et al., 2008). We found that the expression pattern and intensity of αPS1, αPS2 and βν did not change during engulfment (data not shown), but αPS3 and βPS increased strikingly. In healthy mid-stage (stages 7-9) egg chambers, αPS3 was not detected on the FCs (Fig. 1A′) and βPS was present on all surfaces (Fig. 1A″) (Dinkins et al., 2008). Both subunits increased specifically on the apical surface of FCs in very early dying, phase 1, egg chambers ( Fig. 1B′,B″). We quantified the intensity of αPS3 and βPS throughout the cell , and found that the intensity of αPS3 increased more than twofold on the apical surface ( Fig. 1E) of phase 1 egg chambers, whereas there was only a 1.2-fold increase on the basal surface (Fig. 1G). Both αPS3 and βPS continued to increase on the apical surface as engulfment proceeded ( phases 3-5), but also started to increase on the basal surface and within the cytoplasm to a lesser extent (Fig. 1C′,C″,D′,D″,E-G). The colocalization of these two subunits suggests that αPS3 and βPS form a heterodimer on the apical surface during engulfment.
The integrin heterodimer αPS3/βPS is required specifically in FCs during engulfment To determine whether αPS3 and βPS were required in the FCs for engulfment, we knocked down αPS3 and βPS using RNA interference (RNAi) and confirmed that there was effective knockdown using antibody staining ( Fig. 1H-K). We expressed dsRNA against integrin subunits using a GAL4 driver (GR1-GAL4) that begins to be expressed in mid-oogenesis (Etchegaray et al., 2012), allowing for integrin expression and establishment of a normal follicular epithelium in early oogenesis. When well-fed, loss of αPS3 and βPS in FCs did not disrupt normal development of the egg chambers. Loss of βPS resulted in round late-stage egg chambers with short dorsal appendages, as previously shown (Bateman et al., 2001), but the chorion formed normally, indicating that FCs functioned normally under these conditions.
To visualize engulfment, flies expressing αPS3 and βPS dsRNA under the control of GR1-GAL4 were starved and ovaries were stained with cellular markers (Fig. 2) (Etchegaray et al., 2012;Sarkissian et al., 2014). Whereas healthy egg chambers were normal with RNAi knockdown ( Fig. 2A,D,G), loss of either subunit resulted in defective FC enlargement and engulfment in dying egg chambers compared to controls ( Fig. 2B-C,E-F,H-I). Specifically, FCs failed to enlarge and engulf during phase 3, and FCs disappeared by phase 5, leaving behind unengulfed germline cytoplasm and highly condensed nurse cell chromatin, resembling draper mutants (Fig. 2K-L). As previously described (Etchegaray et al., 2012), we quantified the percentage of unengulfed germline using membrane markers to measure the area of the germline and the total egg chamber (Fig. 2M), and found that integrin knockdowns had engulfment defects similar to draper mutants. As another

Clinical issue
Epithelial cells, such as retinal pigment epithelium cells and bronchial epithelial cells, continuously clear dead cells and debris from tissues. Failure of this engulfment process can lead to or exacerbate debilitating conditions such as retinitis pigmentosa, macular degeneration and asthma. However, little is known about the molecular changes that are required for an epithelial layer to initiate engulfment. To date, much of the research undertaken to understand engulfment has focused on identifying the genes required for engulfment. For example, it is wellestablished that integrins are αβ heterodimeric receptors that are required for phagocytosis in both professional and non-professional phagocytes in mammals. However, the pathways required for the regulation of engulfment receptors in vivo remain poorly understood.

Results
In the Drosophila ovary, germline debris produced by starvation-induced cell death at specific stages during oogenesis is engulfed by adjacent epithelial follicle cells. Here, the authors use this model of inducible engulfment by epithelial cells and the powerful genetic tools available in Drosophila to study the regulation of engulfment receptors in vivo. They show that integrins are required specifically on the apical surface of the epithelial follicle cells of the Drosophila ovary for engulfment. Through the use of RNA interference (RNAi) screens, they identify a requirement for polarization in engulfment and integrin trafficking, and show that genes involved in directed trafficking of integrins are also required for engulfment. Notably, many of these genes have no previously reported role in engulfment, but have been shown to have a role in migration.

Implications and future directions
Together, these results identify several genes required for integrin trafficking and regulation in an engulfing epithelial layer and suggest that engulfing and migrating cells share common trafficking machinery. These findings also establish the Drosophila ovary as a powerful model that shares many characteristics with mammalian retinal pigment epithelium cells for studying the molecular changes within an engulfing epithelium. Interestingly, genes involved in vesicle transport have been identified as candidate disease genes in retinitis pigmentosa, and the pathology of this condition is often attributed to the mislocalization of rhodopsin within photoreceptor cells. However, these findings suggest that some of the transport genes implicated in retinitis pigmentosa might instead be required for receptor trafficking within the engulfing epithelial layer. Further studies on the regulation of engulfment in the Drosophila ovary should improve our understanding of diseases in which engulfment by epithelial cells is defective. measurement of engulfment, we used an antibody raised against cleaved caspase Dcp-1, which marks the germline of dying egg chambers and engulfed apoptotic debris (Sarkissian et al., 2014) ( Fig. 2A-L). We used this antibody to specifically count the discrete number of engulfed particles within the FCs. Healthy egg chambers did not label for cleaved Dcp-1 ( Fig. 2A,D,G,J), but control dying or βPS in healthy or dying egg chambers. All data are mean±s.e.m. For E-G, at least four egg chambers were quantified for each phase and a total of 63 were quantified over the five phases. One-way ANOVA and Bonferroni-Holm post-hoc tests were performed: ***P<0.005, **P<0.01. Scale bar: 50 µm. egg chambers showed particle uptake beginning in phase 1, which increased and plateaued by phase 3 (Fig. 2N). However, the loss of draper, αPS3 or βPS resulted in defective particle uptake when compared to controls (Fig. 2N). This defect is most clearly seen in the images showing only the Dcp-1 channel (Fig. 2B′,E′,H′,K′). By phase 4, draper mutants contained more particles than the integrin knockdowns ( Fig. 2N), suggesting that both draper and integrins are required for uptake but draper might also have defects in phagosome maturation, whereas integrins do not. Indeed, Draper and its ortholog in C. elegans, Ced-1, have been shown in multiple cell types to play a role in cell corpse processing (Kurant et al., 2008;Yu et al., 2008;Lu et al., 2011;Chen et al., 2013;Han et al., 2014;Evans et al., 2015). We also found that draper was not required for enrichment of αPS3/βPS ( The localization and function of αPS3/βPS is largely directed by αPS3 Given the polarized localization of integrins in our system, we manipulated expression levels of αPS3 and βPS to determine which subunit dictates the heterodimer localization. Overexpression of αPS3 in FCs led to a robust increase in levels of the protein in whole ovaries (Fig. 3M) and a premature twofold apical enrichment of αPS3/βPS in healthy egg chambers (Fig. 3A,A′,D,D′), equivalent to phase 1 control egg chamber enrichment (Fig. 3P). Overall, αPS3expressing egg chambers engulfed normally (Fig. 3E,O), but often terminated with lingering germline material owing to premature FC migration (Fig. 3F,N). During oogenesis, some of the anterior FCs normally migrate toward the posterior, beginning in stage 9 (King, 1970). In many egg chambers overexpressing αPS3, all of the anterior FCs migrated to the posterior, leaving part of the egg chamber without FCs (Fig. 3E, arrow). Despite the lingering germline, the number of engulfed particles indicated that the egg chambers overexpressing αPS3 were not engulfment-defective (Fig. 3O).  (Etchegaray et al., 2012) are normal when healthy (J) but have little to no particle uptake in phase Surprisingly, we found that egg chambers overexpressing βPS were engulfment-defective ( Fig. 3K-L,N,O) and had considerable basal enrichment of βPS in healthy (Fig. 3J) and dying (Fig. 3K,K′) egg chambers. Throughout engulfment, overexpression of βPS prevented apical enrichment of αPS3 until phase 4 (Fig. 3P), suggesting that overexpression of βPS inhibits αPS3/βPS apical trafficking. Together, these results suggest that the αPS3 subunit determines the apical localization and function of the heterodimer, and basal localization of βPS has an inhibitory effect on engulfment. Consistent with this, loss of αPS3 was associated with normal levels of βPS within healthy egg chambers (Fig. 1H), but βPS did not become apically enriched in dying egg chambers (Fig. 1I,I′). In addition, βPS dsRNA -expressing egg chambers had no detectable αPS3 or βPS in the FCs (Fig. 1J-K′), suggesting that αPS3 expression is dependent on βPS. Because of the importance of αPS3 in directing the localization and function of the heterodimer, we screened candidate genes that might regulate αPS3 localization.
The maintenance of a polarized epithelium is required for engulfment Given the apical enrichment of integrins during engulfment, we hypothesized that cell polarity would be essential for proper engulfment by the FCs. Atypical protein kinase C (aPKC), one of the factors that determines the apical domain of the FCs, was maintained throughout engulfment ( Fig. 4A-C), suggesting that FC apical polarization is needed. To determine whether FC polarity was required for the engulfment of germline material, we performed FCspecific knockdowns of several apical factors that regulate cell polarity: aPKC, baz, par-6 and crb. Again, we used GR1-GAL4, so that proper polarity would be established in the follicular epithelium in early oogenesis, and polarity would only be disrupted beginning in mid-oogenesis. When each knockdown line was well-fed, flies produced healthy stage-10 egg chambers with normal yolk deposition, showing that oogenesis was not grossly affected. However, when the lines were starved, some defects were seen. aPKC dsRNA -, baz dsRNA -and par-6 dsRNA -expressing healthy egg chambers showed a reduced amount of aPKC and some egg chambers had a few gaps between FCs, but otherwise developed At least three egg chambers were quantified for each genotype, phase and quantification method in V,W. 51 egg chambers were quantified for the control, 35 for aPKC dsRNA and 30 for crb dsRNA . All data are mean±s.e.m. One-way ANOVA and Bonferroni-Holm post-hoc tests were performed: ***P<0.005, *P<0.05, n.s., non-significant. Scale bar: 50 µm. normally (Fig. 4D,G,J, arrows). A premature-migration defect was also observed in par-6 dsRNA -expressing healthy egg chambers (starved or well-fed), where anterior FCs migrated towards the posterior early in stage-8 egg chambers, forming a double layer (Fig. 4J, arrowhead). crb dsRNA -expressing healthy egg chambers showed normal aPKC localization with no developmental defects (Fig. 4M).
In degenerating egg chambers, however, pronounced engulfment defects were observed in knockdowns of all of these apical factors (aPKC, baz, par-6 and crb; Fig. 4E,F,H,I,K,L,N,O). Phase 3 dying , which is later lost (R). (S-U) Quantification of αPS3 in the apical and basal regions, and within the cytoplasm, for control, aPKC dsRNA and crb dsRNA egg chambers. The high temperature (29°C) might inhibit normal αPS3/βPS enrichment somewhat compared to 25°C (Fig. 1). All data are mean±s.e.m. At least three egg chambers were quantified for each genotype and phase in S-U. 62 egg chambers were quantified for the control, 38 for aPKC dsRNA and 38 for crb dsRNA . One-way ANOVA and Bonferroni-Holm post-hoc tests were performed: ***P<0.005, **P<0.01, n.s., non-significant. Scale bar: 50 µm. egg chambers expressing aPKC dsRNA , par-6 dsRNA or crb dsRNA showed reduced aPKC and defective FC enlargement (Fig. 4E,K,N, arrows). Phase 3 egg chambers expressing baz dsRNA were variable, but often showed increased but mislocalized aPKC (Fig. 4H), suggesting that Bazooka might inhibit aPKC as it does during apical constriction (David et al., 2013). In each knockdown line, the FCs that maintained any apical aPKC localization were able to enlarge. This demonstrates a need for aPKC for both the polarization of the FCs and the enlargement of the FCs during engulfment. All late dying (phase 5) egg chambers from each knockdown line showed a loss in aPKC localization, a loss of FCs, and large amounts of germline material remaining (Fig. 4F,I,L,O). This indicates that the knockdown of any one of the apical factors (aPKC, baz, par-6 or crb) results in defective engulfment. Loss of aPKC and crb resulted in the strongest and most consistent engulfment-defective phenotype, so we analyzed these mutants further. Based on antibody staining, aPKC dsRNA produced a partial knockdown (Fig. 4A-F), but crb dsRNA produced a very strong knockdown ( Fig. S3A-D). Both aPKC dsRNA and crb dsRNA showed strong defects in the percentage of unengulfed germline and the number of engulfed particles (Fig. 4V,W). We also knocked down the two molecular motors that maintain cell polarity, Dynein and Kinesin. We verified the FC-specific knockdowns using antibody staining and found that both were strong, yet incomplete, knockdowns ( Fig. S3E-L′). Absence of the Dynein heavy chain, Dhc64C, led to engulfment defects, but only after phase 3 (Fig. 4Q,R). Dhc64C dsRNA -expressing phase 5 egg chambers showed a loss in FCs, resulting in a large amount of lingering germline material. Interestingly, the Kinesin heavy chain (Khc) knockdown did not show engulfment defects and had phase 5 egg chambers with proper FC enlargement and clearance of germline material (Fig. 4T,U). Within the FCs, Dynein transports material along the microtubules towards the apical membrane and Kinesin transports material toward the basal membrane (Horne-Badovinac and Bilder, 2008). This suggests that apical molecular transport is essential for engulfment. Taken together, these findings show that apical-basal FC polarity establishment and maintenance is required for engulfment to occur properly in epithelial cells, identifying a previously undescribed role for cell polarity.

RESEARCH ARTICLE
Cell polarization is required for asymmetric integrin localization during engulfment Because FC polarization was required for engulfment, we screened the cell-polarity knockdown lines to see whether they affected the apical enrichment of αPS3 and βPS compared to controls ( Fig. 5A-C). We found that, when cell polarity genes were disrupted, βPS expression was normal in healthy egg chambers (Fig. 5D,G,J,M). Although αPS3/βPS increased during engulfment, neither subunit localized asymmetrically on the apical surface of the FCs (Fig. 5E,E′,H,H′,K,K′,N,N′). Instead, αPS3/βPS had a variable and uneven localization to multiple surfaces and throughout the cytoplasm (Fig. 5S-U). This variability was seen between egg chambers and between cells within individual egg chambers (Fig.  S3M-R). When molecular transport was disrupted with the knockdown of Dhc64C, αPS3 and βPS were also mislocalized in the dying egg chambers (Fig. 5Q-R). This indicates that FC polarization promotes the asymmetric enrichment of the αPS3/βPS integrin heterodimer during engulfment. Without proper polarity, αPS3/βPS becomes mislocalized to all surfaces of the cell, suggesting that polarization sets up the directionality of integrin transport. Within these egg chambers, when αPS3 was basally localized, the FCs often did not enlarge or engulf, suggesting that basal localization of αPS3 can inhibit engulfment. This can be seen in egg chambers that had variable αPS3 localization in FCs that are side by side. Only the FCs with normal apical enrichment enlarged (Fig. 5, zooms). We also examined the localization of another known engulfment receptor, Draper, in the cell-polarity knockdown lines. We found that Draper enrichment was defective in FCs only in the aPKC knockdown line (Fig. S4C-D), perhaps owing to the importance of aPKC as a signaling component without the remainder of the apical complex (Baz and Par-6) (Kim et al., 2009). The other knockdown lines did not have noticeable defects in Draper localization (baz dsRNA shown as an example in Fig. S4E-F). This indicates that aPKC and cell-polarity genes might be necessary for the asymmetric localization of multiple engulfment receptors during engulfment.

Migrating and engulfing cells utilize much of the same integrin trafficking machinery
Cell polarization is often required for integrin localization in migratory cells . Interestingly, migration in border cells and during dorsal closure requires much of the engulfment machinery (Dinkins et al., 2008;Brock et al., 2012;Ellis et al., 2013), suggesting that common molecular processes are shared between migration and engulfment. We found it striking that the same α-subunit is used for both engulfment and migration in Drosophila and in mammals (Dinkins et al., 2008;Shiratsuchi et al., 2012;Nonaka et al., 2013;Law et al., 2015;Tsirmoula et al., 2015). Thus, given the link between cell polarity and integrins in migratory cells, and the fact that αPS3 is required in migratory and engulfing cells, we screened for engulfment and integrin-enrichment defects in knockdowns of several genes that are associated with migration and/or integrin signaling. We found that several genes associated with transport along microtubules and Golgi vesicle formation were required for engulfment and integrin enrichment: Syntaxin 5, short stop and Clasp (Fig. 6D-L). Loss of these genes did not disrupt βPS expression in healthy egg chambers (Fig. 6D,G,J) and had some initial enrichment of αPS3 during engulfment. As engulfment proceeded, however, αPS3 did not apically enrich to control levels (Fig. 6B,B′,E,E′,H,H′,K,K′). Loss of Syntaxin 5, short stop or Clasp resulted in engulfment-defective egg chambers (Fig. 6F,I,L).
Surprisingly, we also found that talin and Pinch, genes associated with integrin activation and downstream signaling (Honda et al., 2013), were required for integrin enrichment (Fig. 6M-R). Loss of talin resulted in increased basolateral βPS expression in healthy egg chambers (Fig. 6M, inset) and a lack of αPS3 enrichment on any (U-W) Quantification of αPS3 and βPS in the apical and basal regions, and within the cytoplasm. All data are mean±s.e.m. At least three egg chambers were quantified for each genotype, phase and quantification method in S-W. For S, 94 egg chambers were quantified for the control, 44 for Syntaxin 5 dsRNA and 30 for talin dsRNA . For T, 55 egg chambers were quantified for the control, 21 for Syntaxin 5 dsRNA and 16 for talin dsRNA . For U-W, 45 egg chambers were quantified for the control, 40 for Syntaxin 5 dsRNA and 28 for talin dsRNA . One-way ANOVA and Bonferroni-Holm post-hoc tests were performed: ***P<0.005, **P<0.01, *P<0.05, n.s., non-significant. Scale bar: 50 µm.

RESEARCH ARTICLE
Disease Models & Mechanisms (2015) 8, 1603-1614doi:10.1242 surface in dying egg chambers (Fig. 6N,N′). Loss of Pinch resulted in normal integrin expression in healthy egg chambers (Fig. 6P) and some initial αPS3 enrichment, but no further enrichment as engulfment progressed (Fig. 6Q,Q′). Loss of either talin or Pinch resulted in engulfment-defective egg chambers (Fig. 6O,R). Loss of Syntaxin 5 and talin had the strongest and least variable defects in engulfment, so we analyzed their phenotypes further. Antibodies were available against Talin, so we tested the knockdown of talin and confirmed that Talin protein was undetectable in talin dsRNA -expressing egg chambers (Fig. S4G-J). Both Syntaxin 5 and talin knockdowns had strong defects in terms of unengulfed germline and number of particles engulfed (Fig. 6S,T). In particular, the loss of talin was extremely strong and resulted in almost no engulfment or apical αPS3 enrichment (Fig. 6T-W). In contrast to the polarity knockdown lines, neither Syntaxin 5 nor talin knockdowns showed mislocalization of αPS3, but they failed to show apical enrichment ( Fig. 6U-W). We examined the expression of Draper in Syntaxin 5 dsRNA and talin dsRNA egg chambers as well, and found no detectable difference in Draper enrichment (data not shown), suggesting that these genes are specific for integrin trafficking. Several other genes we tested had engulfment defects but no effect on integrin enrichment (Table S1), including Grasp, Pkd1 and Arf6; most of these are required for integrin endocytosis in other systems, suggesting that endocytosis does not affect integrin enrichment in FCs.

DISCUSSION
Here, we have investigated integrin function and regulation within engulfing cells. We have shown here that a specific α-subunit is essential for directing the apical localization and function of the integrin heterodimer during engulfment in non-professional phagocytes. We found that αPS3/βPS becomes specifically enriched on the apical surface during engulfment, which suggests that there might be similarities between the epithelial FCs and other polarized epithelial cells that function as phagocytes. αPS3 and βPS are also required for engulfment in the FCs, which suggests that, like in mammals, the same α-subunit functions in engulfment in both professional and non-professional phagocytes and can pair with multiple β-subunits to do so. The loss of cell-polarity genes in the FCs resulted in engulfment defects and integrin-enrichment defects, suggesting that polarization of the cell is required for integrin trafficking in engulfing cells. Cell polarity and integrin trafficking have been linked specifically at the leading edge in migratory cells, and we found here that other genes associated with migration, Golgi orientation, microtubule stabilization and integrin activation are also required for apical αPS3/βPS trafficking and engulfment. To our knowledge, we show the first evidence that βPS, aPKC, baz, par-6, crb, Dhc64C, Syntaxin 5, Clasp and Pinch are required for engulfment.
We propose the following model ( Fig. 7) for integrin trafficking in engulfing cells: cell-polarity proteins polarize the microtubules, and Shot and Clasp bind to microtubules and each other, trafficking αPS3/βPS to the apical surface. Clasp might also bind to the Golgi  and reorient it so that newly synthesized αPS3/ βPS is shuttled to the apical surface. Once at the surface, the integrin heterodimer signals through Talin and Pinch, which are required to promote increased αPS3/βPS on the apical surface. Talin might also be required for integrin release from the Golgi, as has been shown in other systems (Martel et al., 2000). Both Talin and Pinch might be required within the cell for integrin trafficking to the apical surface, for downstream integrin signaling and/or for stabilization at the apical surface.
We found it striking that the same α-subunit is required for both migration and engulfment: αPS3 in Drosophila and αν in mammals (Dinkins et al., 2008;Shiratsuchi et al., 2012;Nonaka et al., 2013;Law et al., 2015;Tsirmoula et al., 2015). So how are migration and engulfment tied together? Are the mechanisms shared in all migrating and engulfing cells, or only those cells that are already equipped to handle both processes? Epithelial cells are capable of either migration or engulfment when appropriate. Integrins are known for their role in adherence, migration and cancer cell invasion, and regulation of their trafficking and function might play a crucial role in regulating cellular function. It is already known that turnover and regulation of integrin heterodimers determines the Our results suggest that cell-polarity proteins polarize the microtubules; Shot and Clasp bind to microtubules and shuttle αPS3/βPS to the apical surface. Clasp might also bind to the Golgi and reorient it so that newly synthesized αPS3/βPS is shuttled to the apical surface. Once at the surface, the integrin heterodimer can signal through Talin and Pinch.

RESEARCH ARTICLE
Disease Models & Mechanisms (2015) 8, 1603-1614doi:10.1242 speed and direction of a migrating cell, and that, if they are not properly regulated, integrins can contribute to cancer cell invasion (Morse et al., 2014). Migration genes are required for engulfment in professional phagocytes that migrate to the site of the dead cell but, given that FCs do not need to migrate because they are already adjacent to the dying germline, our results suggest that the migration machinery might be integral to the engulfment process itself. Several types of epithelial cells are crucial for engulfment on a daily basis in organs such as the retina and bronchial tubes. Without proper clearance in these and other organs, debilitating conditions can develop. However, not much is known about how epithelial cells undergo the molecular changes necessary for engulfment. Thus far, we have identified two engulfment receptors, Draper and αPS3/βPS, and the downstream signaling components of the JNK pathway, as being required for engulfment by the FCs. In addition, we have shown that integrins become enriched specifically on the apical surface of the FCs, and we identified several genes required for integrin trafficking, suggesting that the ovary provides a valuable model for studying integrin trafficking and function within a polarized epithelium. Our findings show that the Drosophila ovary serves as an excellent model for studying the pathways governing engulfment within epithelial cells.

Fly strains and manipulations
All strains were reared on standard cornmeal molasses fly food at 25°C unless otherwise indicated. For starvation experiments, adult flies were conditioned in new vials containing fly food supplemented with freshly made yeast paste for 1.5-2 days and transferred to apple juice agar vials overnight (Sarkissian et al., 2014). When flies are deprived of nutrients overnight (approximately 16-20 h), egg chambers in all five phases of death can be found (Etchegaray et al., 2012). All strains were obtained from Harvard TRiP (Ni et al., 2008), the Bloomington Stock Center or Vienna Drosophila Resource Center unless otherwise indicated (Table S1). G89 (G00089), a GFP gene trap from FlyTrap (Morin et al., 2001;Etchegaray et al., 2012) was recombined with the FC-specific driver GR1-GAL4 (Trudi Schüpbach, Princeton University, NJ, USA; Goentoro et al., 2006;Etchegaray et al., 2012) and was crossed to all UAS lines. Some dsRNA lines were lethal, so GR1-GAL4 was combined with tubulin-GAL80 ts and flies were reared at 18°C. Progeny with tubulin-GAL80 ts were transferred to 29°C for 2 days to inactivate GAL80. All dsRNA crosses shown were transferred to 29°C while conditioning and starving. All other lines were reared, conditioned and starved at 25°C. The puc-lacZ line was puc A251.1 ,ry/ TM3 (Martín-Blanco et al., 1998;Etchegaray et al., 2012). drpr Δ5 (draper −/− ) (Freeman et al., 2003;Etchegaray et al., 2012) was provided by Estee Kurant (Technion-Israel Institute of Technology, Haifa, Israel) and the UAS-βPS lines (Schöck and Perrimon, 2003) were provided by Frieder Schöck (McGill University, Quebec, Canada).

Western blot staining and quantification
To prepare tissue for the western blot, ten ovaries were homogenized in 150 µl dH 2 O using a motorized pestle before boiling. 10 µl of each sample was loaded into a precast 10% polyacrylamide gel to separate proteins. Separated protein was then transferred onto nitrocellulose membranes. Membranes were immunoprobed with αPS3 (1:2000) and lamin 84.12 (1:10,000), followed by LiCor fluorophore-conjugated secondaries (1:20,000). Blots were imaged and quantified using Image Studio.

Engulfment quantification
We quantified the percentage of unengulfed germline as previously described (Etchegaray et al., 2012). At least three egg chambers were analyzed for each phase and genotype. To measure uptake directly, we used an antibody raised against cleaved Dcp-1 to mark engulfed particles (Sarkissian et al., 2014), and at least three egg chambers were quantified for each phase ( phases 0-4) and genotype. P-values were determined using oneway ANOVA and a Bonferroni-Holm post-hoc test.

Integrin enrichment quantification
To quantify the localization of αPS3 and βPS, lines were drawn from apical to basal through nine cells of each egg chamber and a line plot of the intensity measurements was generated (Fig. S1). The last two microns on either end of the cell were designated as the apical or basal regions. The peak values were identified and the remainder of the cell (cytoplasmic) was averaged; these values were averaged and graphed. These values were normalized to the average intensity from healthy control egg chambers imaged on the same slide at the same settings. At least three egg chambers for each phase and genotype were quantified ( phases 1-4). P-values were determined using one-way ANOVA and a Bonferroni-Holm post-hoc test.