The phosphoinositide phosphatase Sac1 regulates cell shape and microtubule stability in the developing Drosophila eye

Undifferentiated epithelial cells are patterned and specified during development to yield highly ordered tissues composed of multiple cell types. Patterning and differentiation within an epithelium are driven in part by cell surface adhesion molecules that promote intercellular interactions. Defects in cell-cell adhesion lead to developmental abnormalities and contribute to disease progression. For example, the mammalian Irre cell recognition module (IRM) adhesion proteins NEPH1 (KIRREL) and nephrin (NPHS1) are required in the kidney for development and maintenance of the filtration barrier or slit diaphragm (Donoviel et al., 2001; Kestilä et al., 1998; Ruotsalainen et al., 1999). Despite their importance in animal development and physiology, little is known about how trafficking and delivery of cell surface adhesion molecules such as IRM proteins is achieved.

The adult Drosophila eye contains ∼750 individual units called ommatidia. Prior to pupariation, each ommatidium consists of eight photoreceptor (PR) cells and four cone cells, surrounded by undifferentiated interommatidial cells (IOCs) (Bao and Cagan, 2005; Cagan and Ready, 1989a,b). During the first half of pupal eye development, 0-42 h after puparium formation (APF), IOCs undergo dynamic morphogenetic changes to give rise to two primary pigment cells (1°pc), six secondary pigment cells (2°pc), three tertiary pigment cells (3°pc) and three bristles, arranged in a honeycomb lattice. These cells support and optically isolate individual ommatidia.

Specification and organization of Drosophila retinal cells requires IRM protein function (Araujo et al., 2003; Bao and Cagan, 2005; Bao et al., 2010; Reiter et al., 1996). 1°pc express the nephrin homologs Hibris (Hbs) and Sticks and stones (Sns), whereas IOCs express the NEPH1 homologs Roughest (Rst) and Kin of Irre (Kirre, also called Dumbfounded) (Bao and Cagan, 2005; Bao et al., 2010). Between 24 and 30 h APF, all four proteins localize to the plasma membrane (PM) of these cells at the 1°pc:IOC border, and heterophilic binding of Rst and Kirre with Hbs and Sns is needed for specification and morphogenesis of 2°pc and 3°pc (2°/3°pc) (Bao and Cagan, 2005; Bao et al., 2010; Reiter et al., 1996). Mutations that affect cell surface accumulation of Rst result in rough eyes and reduced levels of pigmentation as a result of defects in 2°/3°pc differentiation (Araujo et al., 2003; Reiter et al., 1996; Wolff and Ready, 1991).

Phosphoinositides, or phosphatidylinositol (PI) phosphates (PIPs), regulate essential cellular processes such as membrane trafficking and actin cytoskeletal organization. In the canonical PIP pathway, PI is phosphorylated by PI 4-kinases (PI4Ks) to generate PI 4-phosphate (PI4P) (Fig. 1A), which is a precursor for other PIPs, including PI 4,5-bisphosphate [PI(4,5)P₂], and serves as a potent signaling molecule, for example by recruiting effectors to the Golgi body to promote membrane trafficking (De Matteis et al., 2013; Delage et al., 2013; Santiago-Tirado and Bretscher, 2011; Tan and Brill, 2014). Downregulation of PI4P in specific membrane compartments also drives cellular events. For example, in budding yeast (Saccharomyces cerevisiae), PI4P dephosphorylation is required for targeted delivery of cargo to the PM by the exocyst complex (Ling et al., 2014; Mizuno-Yamasaki et al., 2010). Drosophila has three PI4Ks, including a single type II enzyme (PI4KII) that generates PI4P at the trans-Golgi network (TGN) and on endosomes (Burgess et al., 2012). PI4P levels are kept in check by the phosphoinositide phosphatase Sac1 (Fig. 1A), a conserved transmembrane protein present in endoplasmic reticulum and Golgi membranes (Forrest et al., 2013; Foti et al., 2001; Guo et al., 1999; Nemoto et al., 2000; Rivas et al., 1999; Whitters et al., 1993).

We previously showed that Sac1 is required for normal eye development in Drosophila (Wei et al., 2003b). Sac1 is essential for viability in Drosophila (Wei et al., 2003a,b). Flies transheterozygous for a hypomorphic allele and a null allele exhibit rough eyes with necrotic patches and drown in the food soon after eclosing (Wei et al., 2003b). Here, using the hypomorphic allele, which we show is temperature sensitive (ts), we demonstrate that Sac1 plays a crucial role in patterning the retinal epithelium. IOCs of Sac1^ts flies cultured at 25°C exhibit a dramatic increase in PI4P levels and decreased levels of PM PI(4,5)P₂. Although Rst is present at the cell surface of these IOCs, fixation on ice ex vivo results in re-distribution of Rst to enlarged structures containing the exocyst complex subunit Sec8, suggesting a microtubule defect. Indeed, Sac1^ts IOCs contain sparse, disorganized microtubules (MTs) that disappear when fixed on ice. Our results thus identify a novel link between Sac1, PIP levels and microtubule stability in the developing Drosophila eye. Because Sac1 is conserved, our findings suggest a possible role for MT regulation by Sac1 in human development and disease.

We discovered that a hypomorphic allele of Sac1, isolated in a screen to identify lethal mutations in the 61D-61F region of Drosophila chromosome III (Wei et al., 2003b), was temperature sensitive (ts) (Sac1^ts; previously called l(3)2107^L4H). The Sac1^ts mutation causes an amino acid substitution (G382R) within the Sac phosphatase domain, six amino acids upstream of the catalytic motif (Fig. 1B). To determine whether the G382R (TS) mutation affects Sac1 protein stability, Flag-tagged wild-type (WT), phosphatase-reduced (PR; C391S) and TS versions of Drosophila Sac1 were expressed in HeLa cells (Fig. 1C). All three constructs yielded similar amounts of Sac1 protein when cells were grown at 25°C; however, when grown at 37°C, cells expressing the TS construct had nearly undetectable levels of Sac1. To examine the catalytic activity of Sac1, phosphatase assays were performed on Flag-tagged Sac1 proteins affinity purified from HeLa cells. WT Drosophila Sac1 and human SAC1 (SACM1L) exhibited comparable activity in vitro, whereas PR Drosophila Sac1 had significantly reduced phosphatase activity, but retained some residual activity compared with phosphatase dead (PD) human SAC1 (Fig. S1; see Materials and Methods). Despite being expressed at equivalent levels at 25°C, Flag-tagged Sac1-TS could not be purified in sufficient quantities for phosphatase assays. Together, these data suggest that Sac1^ts encodes an unstable protein.

When raised at 18°C (permissive temperature), Sac1^ts mutant flies appeared morphologically normal; however, when raised at 23.5°C or 25°C, Sac1^ts mutants had reduced viability (adult flies died 1-3 days post-eclosion) and exhibited a rough eye phenotype (Fig. 2). To characterize the eye defect, we performed scanning electron microscopy (SEM) on WT and Sac1^ts mutant eyes. Ommatidia from WT (Fig. 2A-C) or from Sac1^ts flies raised at 18°C (Fig. 2D-F) were organized in smooth hexagonal arrays with regularly spaced bristles. In contrast, Sac1^ts mutants raised at 23.5°C displayed mildly rough eyes with missing or misplaced bristles (Fig. 2G-I). Sac1^ts mutants raised at 25°C exhibited a more severe phenotype, including rough eyes with missing bristles, gaps between ommatidia, and abnormal ommatidial shape (Fig. 2J-L). To understand the cellular basis of the observed defects, we used transmission electron microscopy (TEM). In WT ommatidia, 2°/3°pc were well organized, forming a continuous layer between adjacent PR cell clusters (Fig. 2S,T). In contrast, in Sac1^ts mutants, 2°/3°pc appeared dysmorphic and discontinuous, resulting in fused ommatidia where cells from adjacent PR clusters were in direct contact (Fig. 2U-X, arrows). These defects were more severe in Sac1^ts flies raised at 25°C (Fig. 2W,X, arrows), and were not observed in Sac1^ts flies raised at 18°C (not shown).

To determine whether Sac1 exerts its effect on eye development through regulation of PI4P, we examined genetic interactions with PI4KII (also known as PI4KIIα) (Burgess et al., 2012). By SEM, PI4KII null mutants displayed morphologically normal eyes, with the exception of occasional duplicated bristles (Fig. 2M-O). At the ultrastructural level, PI4KII mutants showed minor defects in 2°/3°pc morphology (Fig. 2Y,Z); occasionally, PR cells from adjacent ommatidia were in direct contact because of discontinuous 2°/3°pc (Fig. 2Y, arrow). The Sac1^ts rough eye phenotype and 2°/3°pc morphology defects, as well as the minor defects in PI4KII mutants, were suppressed in Sac1^ts PI4KII double mutants (Fig. 2P-R,AA,BB), indicating that the two enzymes function antagonistically, consistent with a model in which they regulate a common pool of PI4P important for normal eye development.

To further our understanding of the 2°/3°pc morphology defects in Sac1^ts mutants, we examined the adherens junction protein Armadillo (Arm; β-catenin) as a marker for IOC patterning. In WT pupal retinas at 24 h APF, 1°pc were specified and IOCs were uniform and organized single file (Fig. 3A,D,D′). By 30 h APF, 2°/3°pc began to differentiate and each vertex was occupied by a single cell (Fig. 3B,F,F′). By 42 h APF, cells took on their characteristic shapes and the ommatidial lattice was complete (Fig. 3C,H,H′). At 24 h APF, Sac1^ts mutant retinas exhibited normal 1°pc; however, IOCs were neither uniform in shape nor organized single file (Fig. 3E,E′, arrow). At 30 h APF, Sac1^ts IOCs were irregular, bristles were misplaced and multiple cells were present at some vertices (Fig. 3G,G′, arrows). At 42 h APF, the Sac1^ts ommatidial lattice was disorganized; bristles were disrupted and cells appeared enlarged, with defects in cell number and organization (Fig. 3I,I′, arrows). Discs large (Dlg; Dlg1) staining confirmed the Sac1^ts IOC patterning defects (Fig. S2). Notably, distribution of Arm and Dlg appeared to be unaffected in Sac1^ts mutants, indicating that apical junctions were normal. Ommatidial mispatterning scores (OMS) (Johnson and Cagan, 2009) were calculated to quantify the IOC defect (Fig. 3O). WT eyes had an OMS of 0.3±0.11 errors per ommatidium (n=78), whereas Sac1^ts mutant eyes had a significantly higher OMS of 3.3±0.32 errors per ommatidium (n=78, P<1×10⁻¹⁴). In contrast, PI4KII eyes largely resembled WT (Fig. 3M,O; OMS=0.6±0.12, n=78). Hence, Sac1 is required for proper IOC patterning during pupal eye development.

To confirm that the observed defects were due to Sac1 loss of function and increased PI4P levels, we performed rescue experiments and genetic interaction studies. Sac1^ts IOC patterning defects were rescued by a ubiquitously expressed WT mCherry-Sac1 transgene [mCh-Sac1(WT)] (Fig. 3J,J′,O; OMS=0.4±0.11, n=79), but not by expression of a phosphatase-reduced transgene [mCh-Sac1(PR)] (Fig. 3K,K′,O; OMS=2.7±0.24, n=79). In addition, Sac1^ts IOC defects were suppressed by PI4KII (Fig. 3N,N′,O; OMS=0.4±0.12, n=77) and enhanced by expression of mCh-PI4KII (Fig. 3L,L′, arrowhead; OMS not calculated owing to poor survival). Expression of mCh-PI4KII in Sac1^ts also resulted in abnormal cone cell clusters and disorganized 1°pc (Fig. 3L,L′), which were rarely observed in Sac1^ts alone (Fig. 3I,I′). Together, these results indicate that the catalytic activity of Sac1 is important for its function in retinal development and suggest that retinal patterning is highly sensitive to elevated levels of PI4P.

To investigate the consequences of the defects in IOC patterning in Sac1^ts, we examined a previously described reporter line (BA12-lacZ) that provides a readout for 2°/3°pc fate. WT BA12-lacZ flies exhibit nuclear β-gal in 2°/3°pc and bristle cells, as previously described (Araujo et al., 2003), whereas Sac1^ts BA12-lacZ 2°/3°pc nuclei exhibited reduced β-gal-expression at 42 h APF (Fig. S3). Thus, Sac1^ts mutants exhibit defects in determining 2°/3°pc fate.

To determine whether Sac1^ts leads to increased PI4P levels in vivo, we fused mRFP to the pleckstrin homology (PH) domain of mammalian FAPP1 (PLEKHA3) (mRFP-PH-FAPP), which serves as a coincidence detector that binds Arf1 and PI4P at the Golgi (Dowler et al., 2000; Godi et al., 2004). Compared with WT, Sac1^ts mutant IOCs displayed increased accumulation of mRFP-PH-FAPP on intracellular organelles (Fig. 4A-B′), with a significant increase in the number (Fig. 4G; n=12, P<1×10⁻⁵) and size (Fig. 4H; n=12, P<1×10⁻⁹) of PH-FAPP-positive puncta per ommatidium, indicating elevated levels of PI4P. Enlarged PH-FAPP-positive puncta were adjacent to and partially overlapping with the cis-Golgi marker Lava lamp (Lva) in Sac1^ts, suggesting that Sac1 loss resulted in increased Golgi/TGN PI4P (Fig. 4B,B′). A mild but significant increase was observed for the number of Golgi bodies marked by Lva per ommatidium in Sac1^ts mutants (Fig. 4G; n=12, P<1×10⁻³); however, the size of Lva-positive puncta was largely unchanged (Fig. 4H).

Sac1 can affect other PIPs in addition to PI4P (Foti et al., 2001; Guo et al., 1999; Nemoto et al., 2000). For example, Sac1 can dephosphorylate PI3P, PI5P and PI(3,5)P₂ in vitro. In addition, loss of Sac1 in budding yeast results in reduced levels of PI(4,5)P₂ (Foti et al., 2001; Guo et al., 1999; Hughes et al., 2000). To determine the effect of Sac1 on other PIPs, we examined fluorescent markers for PI3P (mCh-2xFYVE; Gillooly et al., 2000) and PI(4,5)P₂ (PLCδ-PH-GFP; Stauffer et al., 1998; Raucher et al., 2000). mCh-2xFYVE-positive endosomes appeared morphologically similar in WT and Sac1^ts retinas at 24 h APF (Fig. 4C-D′), indicating that PI3P levels were unaffected in mutant IOCs. In contrast, Sac1^ts mutants displayed decreased levels of PLCδ-PH-GFP at the IOC PM (Fig. 4E-F′), with the average intensity of PLCδ-PH-GFP at the PM reduced to 70% of WT (Fig. 4I; n=30, P<1×10⁻⁵). To quantify the relative defect, we normalized the ratio of PLCδ-PH-GFP to F-actin (Tan et al., 2014), which appeared largely unchanged in Sac1^ts retinas (Fig. 4F,F′). The PI(4,5)P₂: F-actin ratio was reduced to 67% of WT (Fig. 4J; n=30, P<1×10⁻⁵). Together, these data suggest that PI(4,5)P₂ but not PI3P levels are reduced and that PI4P accumulates in the Golgi/TGN of Sac1^ts IOCs.

Because the eye defects in Sac1^ts mutants resembled mutations in roughest (rst), we reasoned that Sac1 might regulate localization of Rst or other IRM proteins. To assess the distribution of Rst and its paralog Kirre, we examined WT and Sac1^ts retinas at 24 h APF by immunofluorescence. In these experiments, fixation was performed on ice to preserve tissue integrity. In WT, Rst and Kirre accumulated at the 1°pc:IOC border, as previously reported (Fig. 5A,A′) (Araujo et al., 2003; Bao et al., 2010; Reiter et al., 1996). However, in Sac1^ts we observed a dramatic depletion of Rst at the border (Fig. 5B,B′). Quantification of protein intensities at the 1°pc:IOC border revealed a significant decrease in Rst in Sac1^ts mutants (Fig. 5I; 46% of WT, n=30, P<1×10⁻²⁰), whereas Kirre showed only a slight decrease (Fig. 5I; 88% of WT, n=30, P<0.05).

Rst accumulation at the 1°pc:IOC border requires its binding partner Hbs (Bao and Cagan, 2005; Bao et al., 2010), as well as DE-Cadherin (DE-Cad) (Cordero et al., 2007; Grzeschik and Knust, 2005). To determine whether Rst depletion is a secondary consequence of an effect on these regulators, we examined Hbs and DE-Cad (Shg) at 24 h APF. In Sac1^ts retinas fixed on ice, Hbs accumulated normally at the 1°pc:IOC border (Fig. 5C-D′,K) and the average ratio of Rst:Hbs at 24 h APF was 47% of WT (Fig. 5L; n=30, P<1×10⁻⁴). DE-Cad enrichment at the 1°pc:IOC border was mildly reduced in Sac1^ts retinas fixed on ice (Fig. 5E-F′,M; 80% of WT, n=30, P<0.05). However, the ratio of Rst:DE-Cad was still substantially decreased (Fig. 5N; 65% of WT, n=30, P<1×10⁻⁴). Hence, an effect on Hbs or DE-Cad cannot account for the severe defect in Rst accumulation.

At 30 h APF, Kirre, Hbs and DE-Cad all appeared to be unaffected in Sac1^ts, whereas Rst was still significantly decreased at the 1°pc:IOC border (Fig. S4A-H). Additionally, Notch (another regulator of Rst) (Gorski et al., 2000; Grzeschik and Knust, 2005) was unaffected in Sac1^ts at 30 h APF (Fig. S4I,J). Thus, the reduction in apical Rst is neither a consequence of disrupting known Rst regulators nor due to a developmental delay in Sac1^ts.

Surprisingly, in a separate experiment in which retinas were fixed at room temperature (RT), depletion of Rst at the border in Sac1^ts compared with WT was not observed (Fig. 5G,H). Quantification confirmed that Rst and Kirre intensities at the border were unaffected in Sac1^ts retinas fixed at RT (Fig. 5O; 95% and 102% of WT, respectively, n=30). The average ratio of Rst:Kirre in Sac1^ts was 49% of WT when fixed on ice (Fig. 5L; n=30, P<1×10⁻²⁰), compared with 93% of WT when fixed at RT (Fig. 5P; n=30). The observed difference was not due to sensitivity of the anti-Rst antibody to cold temperature, as only the fixation step was altered (see Materials and Methods). Furthermore, the effect of cold fixation on Rst distribution was specific to Sac1^ts, as WT retinas fixed on ice or at RT and stained in parallel (i.e. in the same experiment) exhibited no difference in Rst intensity at the 1°pc:IOC border (Fig. S5; n=90).

Although Rst appeared to accumulate normally in Sac1^ts retinas fixed at RT, it remained possible that the protein could be enriched in secretory vesicles at the cell cortex rather than in the PM. To determine whether Rst is inserted properly at the cell surface in Sac1^ts, we stained IOCs for exofacial Rst. Retinas at 24-30 h APF were incubated with anti-Rst (which recognizes the extracellular domain) and anti-Lava lamp (Lva, which marks the cis-Golgi; a control for membrane permeabilization) prior to RT fixation (see Materials and Methods). Under these conditions, Rst was present at the cell surface in WT and Sac1^ts retinas and Lva staining was not observed (Fig. S6A-B′). In contrast, both Rst and Lva were detectable upon permeabilization (Fig. S6C-D′). These results demonstrate that Rst is present at the PM in Sac1^ts following fixation at RT.

To determine whether maintenance of Rst distribution following ice fixation requires Sac1 catalytic activity, we performed rescue experiments. Rst localization in Sac1^ts tissue fixed on ice was rescued by expression of mCh-Sac1(WT), but not mCh-Sac1(PR) (Fig. 6A-C′). In contrast, Kirre localized to the 1°pc:IOC border regardless of which transgene was expressed. The average ratio of Rst:Kirre in Sac1^ts expressing mCh-Sac1(WT) was 103% of WT (n=30, P<0.3), whereas the ratio in Sac1^ts expressing mCh-Sac1(PR) was 51% of WT (Fig. 6D; n=30, P<1×10⁻⁶). Thus, Rst IOC surface localization during incubation in the cold shows a specific sensitivity to loss of Sac1 function not seen with its paralog Kirre.

Mammalian IRM proteins are continuously turned over and replenished at the cell surface by de novo synthesis and transport (Satoh et al., 2014). We hypothesized that if turnover occurs in a similar manner in Drosophila, loss of Rst at the PM in Sac1^ts mutants fixed on ice could result from failed replacement of internalized protein in a cold-sensitive manner. MTs, which are important for vesicle trafficking, are known to be labile under cold conditions (Behnke and Forer, 1967; Roth, 1967; Tilney and Porter, 1967). We stained WT and Sac1^ts pupal retinas for acetylated α-tubulin and F-actin at 24 h APF. Under both RT (Fig. 7A-B′) and cold (Fig. 7E-F′) fixation conditions, acetylated (stable) MTs were less abundant and appeared disorganized in Sac1^ts compared with WT. These defects were more pronounced under cold fixation conditions. Line scans across individual IOCs (Fig. 7A′,B′, yellow lines) revealed a significant decrease in the peak intensity of acetylated α-tubulin in Sac1^ts mutants, whereas the levels and distribution of F-actin were largely unchanged (Fig. 7G; n=30). Similarly, examination of a β-tubulin-GFP transgene revealed a decrease in MT abundance in Sac1^ts compared with WT following RT fixation (Fig. 7C-D′).

To determine whether Sac1 phosphatase activity is needed for MT stability, pupal retinas from flies expressing mCh-Sac1(WT) or mCh-Sac1(PR) were fixed at RT and stained for acetylated α-tubulin. mCh-Sac1(WT), but not mCh-Sac1(PR), rescued the MT organization defect in Sac1^ts mutant flies (Fig. 8A-D), confirming that MT stability, abundance and organization in the pupal retina require Sac1 phosphatase activity.

To test whether Sac1^ts mutants fixed on ice have defects in continued delivery of Rst to the PM, we examined localization of the exocyst complex, which is dependent on microtubules in mammalian cells (Vega and Hsu, 2001). WT and Sac1^ts pupal retinas at 24 h APF were fixed at RT or on ice, and co-stained for Rst and the exocyst subunit Sec8. When prepared at RT, WT and Sac1^ts pupal retinas exhibited similar staining for Sec8 (Fig. 9A-B′). In subapical regions of the IOCs, we observed small Sec8-positive puncta that showed partial colocalization with Rst. Thus, exocyst-associated trafficking is unperturbed in Sac1^ts mutants following RT fixation. In contrast, cold fixation resulted in severe defects in Sec8 distribution in Sac1^ts (Fig. 9D,D′). Sac1^ts IOCs contained enlarged Sec8-positive structures (>0.3 µm²) in addition to the small puncta seen in WT (Fig. 9C-D′). Quantification of these structures revealed that WT tissue contained an average of 1.5 enlarged Sec8-positive puncta per ommatidium, whereas Sac1^ts mutants contained an average of 7.0 (Fig. 9G; n=14, P<1×10⁻⁵). In Sac1^ts, 87% of these enlarged Sec8 structures were positive for Rst (Fig. 9H; n=14). In addition, the percentage of Rst-positive puncta that were positive for Sec8 increased from 3.5% to 24.8% (Fig. 9I; n=14).

To determine whether Rst accumulates in the endocytic pathway, we co-stained for Rst and the late endosome marker YFP-Rab7 in WT and Sac1^ts retinas fixed on ice. Rst was present in YFP-Rab7-positive endosomes in both WT and Sac1^ts retinas and no significant difference in colocalization was detected (Fig. 9E-F′). Quantification revealed that although the average number of YFP-Rab7-positive puncta was similar in WT and Sac1^ts retinas, Sac1^ts retinas exhibited a significant decrease in the average number of Rst-positive puncta per ommatidium (Fig. 9G).

Taken together, these findings suggest a specific cold-sensitive step in exocyst-dependent trafficking of Rst to the plasma membrane in Sac1^ts.

To confirm that Rst is internalized upon ice fixation, we performed anti-Rst antibody uptake experiments at 24h APF. In the control experiment, where Sac1^ts and WT retinas were fixed and permeabilized prior to anti-Lva (as an indicator of permeabilization) and anti-Rst staining, numerous Rst-positive and Lva-positive puncta were observed (Fig. 10A-B′). For antibody uptake, retinas were briefly incubated with anti-Rst and anti-Lva at RT prior to fixation, then washed and fixed on ice (see Materials and Methods). Following this protocol, we observed subapical Rst-positive puncta in IOCs of both Sac1^ts and WT retinas, with a somewhat higher number in WT (Fig. 10C-D′,I; n=60, P<0.01). Lva staining was not observed, confirming lack of permeabilization during incubation with primary antibodies. When the retinas were incubated for an additional 20 min chase at RT prior to fixation, the number of puncta per ommatidium increased significantly in both WT (Fig. 10E,E′; n=60, P<1×10⁻⁶) and Sac1^ts (Fig. 10F,F′; n=60, P<1×10⁻¹²), to roughly similar levels (Fig. 10I). Together, these results indicate that Rst is internalized from the plasma membrane in retinas fixed on ice, and that Rst turnover occurs ex vivo prior to fixation. In addition, when retinas were co-stained with anti-Sec8 following anti-Rst uptake and cold fixation (see Materials and Methods), instances of colocalization were observed in both Sac1^ts and WT, suggesting endocytosed Rst is partially recycled to the plasma membrane via an exocyst-mediated pathway (Fig. 10G-H′).

Conditional alleles are powerful tools for studying the functions of essential genes. Here, using a temperature-sensitive allele of Sac1, we show that proper phospholipid regulation by Sac1 is required for normal development of the Drosophila retinal epithelium.

Our results indicate that retinal patterning is highly sensitive to elevated levels of PI4P. Loss of Sac1 leads to increased Golgi PI4P and IOC patterning defects that are rescued by expression of catalytically active Sac1. Deletion of PI4KII suppresses Sac1^ts, whereas expression of a WT PI4KII transgene exacerbates the observed defects. Thus, Sac1 and PI4KII regulate a pool of PI4P that is required for retinal patterning. Loss of PI4KII itself results in mild patterning defects (i.e. occasional extra bristles and fused ommatidia), indicating that reducing PI4P levels also has consequences for retinal development. Although these data point to a clear role for PI4KII in signaling required for 2°/3°pc adhesion and patterning, our experiments do not rule out possible contributions of other PI4Ks to this process. Moreover, we anticipate that, in addition to the Golgi, other pools of PI4P at the PM (Cheong et al., 2010; Faulhammer et al., 2005; Manford et al., 2012; Roy and Levine, 2004; Tahirovic et al., 2005) and elsewhere (Hammond et al., 2014) may be regulated by Sac1.

Because PI4P is phosphorylated by PIP 5-kinases to produce PI(4,5)P₂, Sac1^ts mutants might be expected to show increased levels of PI(4,5)P₂ as well as PI4P. Strikingly, however, we found that Sac1^ts mutants exhibit reduced levels of the PI(4,5)P₂ marker PLCδ-PH at the PM. Although this result is counterintuitive, similar observations have been reported in budding yeast (Foti et al., 2001; Guo et al., 1999; Hughes et al., 2000), and we previously observed the same phenomenon in Drosophila embryos homozygous for a lethal allele of Sac1 (unpublished data). We note that this observation is in contrast to other Drosophila tissues, where Sac1 loss was reported to have no effect on PI(4,5)P₂ levels (Forrest et al., 2013; Yavari et al., 2010). It is possible that PIP 5-kinases lack access to the increased PI4P in Sac1 mutants, thereby preventing the expected increase in PI(4,5)P₂, or perhaps increased PI4P triggers a feedback mechanism that activates PIP phosphatases or phospholipases, leading to PI(4,5)P₂ breakdown. Thus, in addition to elevated PI4P levels, reduced levels of PI(4,5)P₂ might contribute to the IOC patterning defects in Sac1^ts. Indeed, in preliminary results we observed a genetic interaction between Sac1 and the PI4P 5-kinase skittles (not shown), which could stem from reduced PI(4,5)P₂ or increased PI4P levels, or a combination of the two.

Drosophila Sac1^ts mutants exhibit rough eyes, ommatidial mispatterning and abnormal 2°/3°pc morphology, phenotypes that resemble mutants for the IRM protein Rst. Maintenance of Rst at the 1°pc:IOC border is sensitive to ice fixation in Sac1^ts mutant retinas. Because MTs are labile under cold conditions, this suggested a MT defect in Sac1^ts. MTs are disorganized in Sac1^ts mutant IOCs fixed at RT, and even more dramatically affected in mutant IOCs fixed on ice. Moreover, Sac1^ts IOCs fixed on ice display enlarged Sec8-positive vesicles containing Rst, suggesting that Rst travels to the cell surface via an exocyst- and MT-associated pathway that is sensitive to cold fixation in Sac1^ts.

The simplest model to explain our data is that Rst is constantly turned over at the PM and requires continuous replenishment from internal stores, as has been reported for mammalian IRM proteins (Satoh et al., 2014). This turnover appears to occur in both WT and Sac1^ts, as we observed similar uptake of anti-Rst antibody in each. It remains possible that subtle defects in the activity or turnover of Rst (or other proteins) at the PM contribute to the observed defects in Sac1^ts. Indeed, we observed compromised expression of the differentiation marker BA12-lacZ in Sac1^ts IOCs. Defects in BA12-lacZ expression, as well as IOC number and specification, also occur in Rst^D mutants, which show a delay in Rst accumulation at the 1°pc:IOC border (Araujo et al., 2003). Relevant to this point, we note that when quantification of Rst distribution after RT or cold fixation was independently validated at a later date, we observed a slight but significant reduction in Rst accumulation in Sac1^ts following RT fixation (84% of WT; n=90). However, as this difference was not observed in our earlier experiments (Fig. 5G,H,O), we are unable to conclude that there is a reproducibly significant decrease in Rst accumulation in Sac1^ts.

Sac1 was originally identified as a suppressor of yeast actin mutants (Cleves et al., 1989; Novick et al., 1989). Thus, much of the literature has focused on how Sac1 loss affects the actin cytoskeleton. Here, we observed that retinal cells in Drosophila Sac1^ts mutants exhibit normal actin organization during early stages of pupal eye development (24-30 h APF), yet show defects in MT stability and organization. Interestingly, both increased levels of PI4P and reduced levels of PI(4,5)P₂ have previously been associated with defects in MT organization (Forrest et al., 2013; Gervais et al., 2008; Liu et al., 2009; Wei et al., 2008). The molecular mechanisms involved remain unknown, and further investigation will be required to determine the link between Sac1 and MT stability. However, our observations suggest that regulation of the MT cytoskeleton by Sac1 may be more important when moving from fungal to animal models.

In summary, we have characterized a requirement for Sac1 in maintaining PI4P and PI(4,5)P₂ levels and promoting normal development of the Drosophila retina. Additionally, we have demonstrated that Rst distribution in Sac1^ts is sensitive to cold fixation, and that Sac1^ts exhibits a microtubule defect at 24 h APF. Our results indicate that Rst is delivered to the PM via exocyst- and microtubule-based trafficking (which is cold sensitive in Sac1^ts) and is turned over at the PM, similar to its mammalian homologs. Further investigation will be required to unravel precisely how Sac1 regulates microtubules, and how this contributes to normal retinal development.

Flies were raised on standard cornmeal molasses agar (Ashburner, 1990). Crosses and staging were performed at 25°C unless otherwise indicated. Stocks were Oregon R (wild type, WT), w⁺; Sco/CyO; FRT80B, Sac1^ts (Sac1^ts, previously known as l(3)2107 or sac1^L4H) (Wei et al., 2003b), w⁺; Sco/CyO; P{w⁺, CG14671}, P{neoFRT}82B, Df(3R)730 (PI4KII) (Burgess et al., 2012) and w⁺; Sco/CyO; Sac1^ts, P{w⁺, CG14671}, P{neoFRT}82, Df(3R)730 (Sac1^ts PI4KII) (this work). The following transgenes on chromosome II were crossed into the Sac1^ts mutant background: P{w⁺, α₁-tubulin::mCherry-PI4KII} (Burgess et al., 2012), P{w⁺, α₁-tubulin::mCherry-Sac1(WT)} (WT Sac1) and P{w⁺, α₁-tubulin::mCherry-Sac1(PR)} [phosphatase reduced (PR) Sac1] (this study). Additional stocks were w, BA12-lacZ (gift of R. Ramos, University of São Paulo, Brazil) (Araujo et al., 2003), w; P{w⁺, UAS-mRFP-PH-FAPP}/CyO (this study), w; P{w⁺, UAS-mCherry-2xFYVE} (gift of A. Kiger, UCSD, CA, USA) (Velichkova et al., 2010), w; P{w⁺, UAS-PLCδ-PH-GFP}/CyO (gift of L. Cooley, Yale University, CT, USA) (von Stein et al., 2005), P{w⁺, YFP-Rab7} (gift of S. Eaton, Dresden, Germany) (Marois et al., 2006), w; P{w⁺, β-tub-GFP}/CyO (gift of Y. Akiyama-Oda and H. Oda, JT Biohistory Research Hall, Takatsuki City, Japan) (Inoue et al., 2004) and y¹ w*; P{Act5C-GAL4}25FO1/CyO (to drive expression of UAS-mRFP1-PH-FAPP, UAS-mCherry-2xFYVE and UAS-PLCδ-PH-GFP; Bloomington Drosophila Stock Center, IN, USA).

To generate Drosophila Sac1 constructs for expression in cultured mammalian cells, WT Sac1 was amplified from cDNA (GH08349, BDGP, Berkeley, CA, USA) and a Flag tag was added to the 5′ end using standard molecular techniques; Flag-Sac1 was amplified using primers 5′-GCGGTACCATGGACTACAAGGACGACGATGACAAGATGGACAGCAGGGAGGAGAACGCC-3′ and 5′-GCTCTAGATCATGGGTCTCGAAAGGAGATGGG-3′ (Invitrogen) and cloned into the mammalian expression vector pCDNA3.1 with Asp718 and XhoI. Phosphatase reduced (PR, C391S) and temperature sensitive (TS, G382R) mutations were introduced using QuikChange XL (Stratagene). Flag-Sac1[PR] and Flag-Sac1[TS] were generated using the mutagenesis primers 5′-GAACGAATTGTATCGACTCTCTCGATAGGACGAACGTCGCGACGTTCGTCCTATCGAGAGAGTCGATACAATTCGTTC-3′ and 5′-GTCCACGC AGACTCGTGTCTTCCGAACGAATTGCAATTCGTTCGGAAGACACGAGTCTGCGTGGAC-3′, respectively (Invitrogen). Note that although the mutated cysteine in Drosophila Sac1(PR) is conserved and present in the catalytic motif (C391), it is not the same residue that is mutated in phosphatase dead (PD, C389S) human SAC1 (Rohde et al., 2003), likely explaining the residual phosphatase activity.

To generate transgenic flies, Sac1(WT) and Sac1(PR) were subcloned from pCDNA3.1 into a modified pCaSpeR4 vector containing the αTub84B promoter (α₁-tubulin) fused to mCherry (Goldbach et al., 2010) using KpnI and XbaI. mRFP1-PH-FAPP was amplified from tv3::mRFP-PH-FAPP (Wei et al., 2008) using standard molecular techniques and subcloned into pUAST (Brand and Perrimon, 1993) with NotI and Asp718. Transgenic lines carrying pUAST::mRFP-PH-FAPP were generated in the lab. All other transgenic lines were made by BestGene using standard P-element transformation techniques.

For SEM, heads from 3-day-old flies were bisected and fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer for 2 h. Samples were rinsed with 0.1 M sodium cacodylate buffer in 0.2 M sucrose for 20 min and dehydrated in an ethanol series of 20 min washes. After dehydration, samples were subjected to critical-point drying (Bal-Tec CPD 030 Critical Point Dryer), attached to an aluminum stub, and sputter-coated with gold (Denton Desk II sputter coater) (Advanced Bioimaging Center, Mount Sinai Hospital, Toronto, Canada). Samples were imaged using a Phillips/FEI XL30 scanning electron microscope. Images were uniformly edited with Adobe Photoshop CS6.

Fly eyes were prepared for TEM as described (Pellikka et al., 2002). In brief, heads from 3-day-old flies were immobilized in PBS, bisected and fixed in 1% OsO₄ in 0.1 M cacodylate buffer. Samples were left in fixative for 3 days on a nutator at 4°C. Samples were washed with 0.1 M cacodylate buffer, placed in fixative in the dark for 1 h, washed again with 0.1 M cacodylate buffer, dehydrated in an ethanol series (50%, 70%, 80%, 90% and 100% for 5 min each), embedded in fresh Spurr's resin and polymerized in rubber molds at 65°C for 8 h. Embedded samples were sectioned (Reichert Ultracut E ultramicrotome) and imaged at 5000× or 19,000× using a FEI Tecnai 20 transmission electron microscope and AMT digital camera (Advanced Bioimaging Center, Mount Sinai Hospital, Toronto, Canada). Images were uniformly edited with Adobe Photoshop CS6.

HeLa cells (ATCC HeLa 229 cell line, Manassas), which were authenticated and tested for contamination, were grown in DMEM (Sigma-Aldrich) supplemented with 10% fetal bovine serum. Cells were maintained at 37°C in a humidified atmosphere with 5% CO₂. Cells were seeded on culture plates 24 h before transfection. Cells were transfected with plasmids for transient expression of Flag-tagged Drosophila (this study) or human (Rohde et al., 2003) Sac1 constructs using Lipofectamine 2000 (Invitrogen). Transfected cells were kept for 4 h at 37°C and then further incubated for 18 h at either 37°C or 25°C. Cells were lysed in mRIPA buffer (1% NP-40, 1% sodium deoxycholate, 150 mM NaCl, 50 mM Tris-HCl, pH 8.0) supplemented with protease inhibitor cocktail (Roche). Flag-tagged Sac1 proteins were analyzed by SDS page and immunoblotting (1:1000) using mouse monoclonal anti-Flag M2 antiserum (Sigma-Aldrich F3165). Anti-GAPDH antibody (1:1000; Abcam, ab9485) was used as the loading control.

HeLa cells were transfected as above with plasmids for transient expression of Flag-tagged Sac1. Two days after infection, cells were washed once with PBS and lysed in mRIPA buffer supplemented with protease inhibitor cocktail (Roche). Following centrifugation at 13,000 g for 15 min, Flag-tagged Sac1 proteins were collected on M2 agarose beads. A modified version of a published protocol (Blagoveshchenskaya et al., 2008) was used to measure phosphatase activity: 24 µl of reaction buffer (200 mM sodium acetate, 100 mM Bis-Tris, 100 mM Tris, pH 6.0, 0.002% porcine gelatin and 4 mM DTT), 1.4 µl of 1 M dioctanoyl-PI4P in water (Sigma-Aldrich) and M2 agarose beads containing 4 µg of recombinant Sac1 protein were mixed and incubated at 37°C for various times. Reactions were stopped by addition of an equal volume of 100 mM NEM and 25 µl of each supernatant was pipetted into 96-well plates. 50 µl of Malachite Green solution (1 volume 4.2% ammonium molybdate in 4 M HCl, 3 volumes of 0.045% Malachite Green and 0.01% Tween 20) was added to each well. After incubation for 20 min at RT, OD₆₂₀ was measured.

Pupal eyes were dissected in PBS as described (Walther and Pichaud, 2006) and fixed in 4% paraformaldehyde (PFA) for 30 min. Fixation was performed on ice, unless otherwise stated. After fixation, samples were washed in PBS containing 0.3% saponin (PBSS), blocked in PBSS with 5% normal goat serum (NGS) for 1 h at RT, and incubated with primary antibodies in PBSS with 5% NGS overnight at 4°C. Samples were washed and incubated with secondary antibodies in PBSS with 10% NGS. DE-Cad staining was as described above, except phosphate buffer was used instead of PBS. Rhodamine-conjugated phalloidin was used at 4 U/ml to visualize F-actin (Invitrogen). DAPI (Thermo Fisher Scientific, D1306) was used at 1:1000 to stain DNA. Samples were mounted in Dako fluorescence mounting medium (Agilent Technologies).

Primary antibodies were mouse anti-acetylated α-tubulin 6-11B (Sigma-Aldrich, T7451, 1:1000), mouse anti-Arm N2 7A1 [Developmental Studies Hybridoma Bank (DSHB), AB_528089, 1:150] (Riggleman et al., 1990), mouse anti-β-gal (Promega, Z3781, 1:250), rat anti-DE-Cadherin DCAD2 (DSHB, AB_528120, 1:50) (Oda et al., 1994), mouse anti-Dlg 4F3 (DSHB, AB_528203, 1:500) (Parnas et al., 2001), rat anti-dsRed (ChromoTek, 5f8-100, 1:500), mouse anti-GFP 3E6 (Life Technologies, A-11120, 1:500), chicken anti-GFP (Abcam, 13970, 13970, 1:500), rabbit anti-Hibris AS14 (from Karl-Friedrich Fischbach, 1:300) (Bao et al., 2010), rabbit anti-Kirre (from Karl-Friedrich Fischbach, 1:300), rabbit anti-Lava lamp (gift of J. Sisson and O. Papoulas, 1:1000) (Sisson et al., 2000), mouse anti-Notch^ICD C17.9C6 (DSHB, AB_528410, 1:500) (Fehon et al., 1990), mouse anti-Roughest mAb24A5.1 (from Karl-Friedrich Fischbach, 1:50) (Schneider et al., 1995), guinea pig anti-Sec8 (gift of U. Tepass, 1:1000) (Beronja et al., 2005), Secondary antibodies conjugated to Alexa 488, 568 or 633 were used at 1:500 (Molecular Probes, Life Technologies).

Pupal eyes were dissected in PBS as described (Walther and Pichaud, 2006), kept at RT, and stained according to one of two methods. For no permeabilization, tissue was incubated with anti-Rst and anti-Lva in PBS containing 5% NGS for 1 h prior to fixation in 4% PFA for 30 min. For permeabilization, tissue was fixed in 2% PFA for 30 min and then incubated with anti-Rst and anti-Lva in PBS containing 5% NGS for 1 h. In either case, samples were then washed, incubated with secondary antibodies in PBS containing 5% NGS, washed again, and mounted for imaging.

Pupal eyes were dissected in Schneider's media (1% fetal bovine serum) as described (Walther and Pichaud, 2006), and stained according to one of three methods. For permeabilization control, tissue was fixed on ice in 4% PFA for 30 min, washed in PBSS for permeabilization, and incubated with anti-Rst and anti-Lva in PBS containing 5% NGS for 10 min. For antibody uptake (no chase), tissue was incubated in anti-Rst and anti-Lva in Schneider's media (1% fetal bovine serum) containing 5% NGS for 10 min, washed in PBS, and then fixed on ice in 4% PFA for 30 min and washed in PBS. For antibody uptake (with chase), tissue was incubated in anti-Rst and anti-Lva in Schneider's media (1% fetal bovine serum) containing 5% NGS for 10 min, washed in Schneider's media, incubated for 20 min at RT, washed in PBS, and then fixed on ice in 4% PFA for 30 min, and washed in PBS. In all cases, samples were then washed in PBSS for permeabilization, blocked in PBSS with 5% NGS for 1 h, incubated with secondary antibodies in PBSS containing 10% NGS, washed again, and mounted for imaging in ProLong Diamond Antifade Mountant (Thermo Fisher).

Images were acquired on a Quorum spinning disk confocal microscope featuring the following components: Yokogawa CSU-X1 scanhead, Olympus IX81 microscope, 60×/1.35 oil-immersion objective, Spectral Applied Research laser merge module (excitation wavelengths: 405 nm, 491 nm, 561 nm, 640 nm) and Hamamatsu C9100-13 EM-CCD camera (SickKids Imaging Facility, Toronto, Canada). Serial optical sections were acquired at 0.3 µm and deconvolved using the Iterative Restoration function of Perkin Elmer Volocity 6.3 software (SickKids Imaging Facility, Toronto, Canada). A single focal plane (xy plane) was selected at a comparable depth in the tissue and an image was obtained using the Capture Snapshot function of Volocity. Images were then exported from Volocity and uniformly edited for brightness and contrast using Adobe Photoshop CS6.

Ommatidial mispatterning scores (OMS) were determined as previously described (Johnson and Cagan, 2009). An average OMS was calculated for each genotype and Student's t-test was used for statistical analysis and to determine P-values. Fluorescence intensity values across the 1°pc:IOC border were determined using the line tool in ImageJ. Puncta size and number were measured using the Volocity Measurements tool. Statistical analysis of puncta size, puncta number, average intensity and intensity ratios between WT and Sac1^ts was performed with Student's t-test using values normalized to WT. For fluorescence intensity and puncta measurements, non-neighboring IOCs or ommatidia in equivalent focal planes were selected for analysis. All experiments were replicated at least three times; quantification was performed using combined results from three independent experiments.

We thank J. Ashkenas, S. Egan, H. McNeill, R. Fehon and members of the Brill lab for helpful discussions; C.-I. J. Ma, J. Tan, L. Fabian and especially M. Pellikka for advice and assistance with experiments; L. Cooley, A. Kiger, U. Tepass, K.-F. Fischbach, J. Sisson and O. Papoulas, the Developmental Studies Hybridoma Bank, and the Bloomington Drosophila Stock Center for generously providing flies and reagents; M. Woodside and P. Paroutis (SickKids Imaging Facility) for assistance with confocal imaging and analysis; H. Hong and A. Darabie (Imaging Facility, Department of Cell & Systems Biology, University of Toronto) for assistance with TEM; D. Holmyard (Advanced Bioimaging Center, Mount Sinai Hospital) for assistance with SEM.