shibire mutations reveal distinct dynamin-independent and -dependent endocytic pathways in primary cultures of Drosophila hemocytes

We have developed a primary cell culture system derived from embryonic and larval stages of Drosophila. This allows for high-resolution imaging and genetic analyses of endocytic processes. Here, we have investigated endocytic pathways of three types of molecules: an endogenous receptor that binds anionic ligands (ALs), glycosylphosphatidylinositol (GPI)-anchored protein (GPI-AP), and markers of the fluid phase in primary hemocytes. We find that the endogenous AL-binding receptor (ALBR) is internalized into Rab5-positive endosomes, whereas the major portion of the fluid phase is taken up into Rab5-negative endosomes; GPI-APs are endocytosed into both classes of endosomes. ALBR and fluid-phase-containing early endosomes subsequently fuse to yield a population of Rab7-positive late endosomes. In primary culture, the endocytic phenotype of ALBR internalization in cells carrying mutations in Drosophila Dynamin (dDyn) at the shibire locus (shits) parallels the temperature-sensitive behavior of shits animals. At the restrictive temperature in shits cells, receptor-bound ALs remain completely surface accessible, localized to clathrin and α-adaptin-positive structures. On lowering the temperature, ALs are rapidly sequestered, suggesting a reversible block at a late step in dDyn-dependent endocytosis. By contrast, GPI-AP and fluid-phase endocytosis are quantitatively unaffected at the restrictive temperature in shits hemocytes, demonstrating a constitutive dDyn and Rab5-independent endocytic pathway in Drosophila.


Introduction
Metazoan systems like Caenorhabditis and Drosophila provide a unique opportunity to extract molecular details about the mechanisms of endocytosis as well as a broader idea about the role of endocytic trafficking in the development and physiology of multicellular animals (Narayanan and Ramaswami, 2001;Nurrish, 2002). For example, numerous genetic screens in Drosophila for mutations that result in reversible temperature-sensitive paralysis have led to the discovery of molecules affecting various aspects of neural transmission. Such behavioral screens, originally conceived as a means of isolating conditional mutations affecting muscle physiology, have largely identified molecules that affect axonal conduction (para) (Grigliatti et al., 1973) and synaptic vesicle recycling (shibire, stoned, comatose, syntaxin) (Grigliatti et al., 1973;Littleton et al., 1998;Siddiqi and Benzer, 1976). Complementary reverse-genetic approaches have also identified other molecules affecting nervous system function (cysteine string protein, α-adaptin) (Gonzalez-Gaitan and Jackle, 1997; Zinsmaier et al., 1994). Whereas specialized modes of trafficking such as synaptic vesicle recycling have been extensively investigated in Caenorhabditis and Drosophila, the diversity of endocytic pathways in these systems has yet to be characterized.
Not surprisingly, essential mediators of synaptic endocytosis like α-adaptin (Gonzalez-Gaitan and Jackle, 1997) and the shibire (shi) gene product Drosophila Dynamin (dDyn), which is the only reported Drosophila homolog of vertebrate Dyn (Kosaka and Ikeda, 1983a;van der Bliek and Meyerowitz, 1991), are not restricted to the nervous system (Chen et al., 1992;Dornan et al., 1997). dDyn is known to regulate endocytosis in a variety of fly tissues (Kosaka and Ikeda, 1983a;Kosaka and Ikeda, 1983b;Tsuruhara et al., 1990). Zygotic null mutations at the shi locus are lethal and the mutants show neural hyperplasia (Poodry, 1990). Electronmicroscopy (EM) studies that addressed the phenotypes of temperature-sensitive alleles at the shibire locus (shi ts ) in larval garland cells showed an accumulation of 'coated pits' at the plasma membrane and reduced uptake of fluid-phase tracers like horse radish peroxidase (HRP) (Kosaka and Ikeda, 1983b). These studies suggested that dDyn is required for all pathways of endocytosis in the cell types examined.
By contrast, over-expression of an analogous temperature-We have developed a primary cell culture system derived from embryonic and larval stages of Drosophila. This allows for high-resolution imaging and genetic analyses of endocytic processes. Here, we have investigated endocytic pathways of three types of molecules: an endogenous receptor that binds anionic ligands (ALs), glycosylphosphatidylinositol (GPI)-anchored protein (GPI-AP), and markers of the fluid phase in primary hemocytes. We find that the endogenous AL-binding receptor (ALBR) is internalized into Rab5-positive endosomes, whereas the major portion of the fluid phase is taken up into Rab5-negative endosomes; GPI-APs are endocytosed into both classes of endosomes. ALBR and fluid-phase-containing early endosomes subsequently fuse to yield a population of Rab7-positive late endosomes. In primary culture, the endocytic phenotype of ALBR internalization in cells carrying mutations in Drosophila Dynamin (dDyn) at the shibire locus (shi ts ) parallels the temperature-sensitive behavior of shi ts animals. At the restrictive temperature in shi ts cells, receptor-bound ALs remain completely surface accessible, localized to clathrin and α-adaptin-positive structures. On lowering the temperature, ALs are rapidly sequestered, suggesting a reversible block at a late step in dDyn-dependent endocytosis. By contrast, GPI-AP and fluid-phase endocytosis are quantitatively unaffected at the restrictive temperature in shi ts hemocytes, demonstrating a constitutive dDyn and Rab5-independent endocytic pathway in Drosophila.
sensitive mutant of Dyn (Dyn ts ) in HeLa cells led to an initial reduction of fluid-phase uptake at 39°C that subsequently recovered (Damke et al., 1995); the internalization of human transferrin was completely inhibited at this temperature. These results argued for pathways of fluid-phase uptake in mammalian cells that are induced or upregulated when Dyn function is perturbed. Glycosylphosphatidylinositol-anchored proteins (GPI-APs) (reviewed by Chatterjee and Mayor, 2001) and the D2 dopamine receptor (Vickery and von Zastrow, 1999) continue to be internalized into mammalian cells upon expression of dominant-negative Dyn isoforms. GPI-APs are endocytosed in a Dyn-independent manner into distinct endocytic compartments that contain a majority of the internalized fluid phase . These studies suggest that mammalian cells exhibit constitutive Dynindependent pathways of endocytosis for both membrane and fluid-phase markers. However, mammals have three distinct genes encoding Dyn and as many as 25 splice variants (Cao et al., 1998), leaving open the possibility that alternate forms of Dyn might be involved in these endocytic events (McNiven et al., 2000). By contrast, in Drosophila dDyn, a multi-domain protein (see later) is encoded by a single locus that has six splice variants (Staples and Ramaswami, 1999;van der Bliek and Meyerowitz, 1991). The availability of shi ts alleles that map to domains conserved across all the splice variants makes Drosophila an attractive system to dissect the involvement of dDyn in different endocytic processes. Drosophila macrophages and hemocytes are a part of the innate immune system of the animal (Lanot et al., 2001;Tepass et al., 1994) and are capable of internalizing a variety of ligands by a scavenger receptor-mediated (dSR) pathway (Abrams et al., 1992). These cells are also phagocytic and have been shown to engulf both apoptotic cells and microbes (Franc et al., 1999); cells of this lineage in other metazoa are known to have multiple pathways of endocytosis (Gold et al., 1999;Racoosin and Swanson, 1992).
Here, we establish a methodology to reproducibly obtain primary cultures of macrophages and hemocytes from wildtype and mutant Drosophila embryos and larvae, respectively. The larval hemocytes have an endogenous anionic-ligand binding receptor (ALBR) with a similar ligand-binding specificity as dSR. In this system, we probed the existence of multiple endocytic pathways. Specifically, we have asked whether cells from embryonic and larval stages of wild-type and temperature-sensitive shibire animals are capable of dDyn-dependent and -independent endocytosis. We find that, first, the endocytic phenotype of ALBR-mediated endocytosis in cells from temperature-sensitive shibire mutants parallels the behavior of mutant flies; ALBRmediated internalization is reversibly blocked by raising the temperature. Second, at the restrictive temperature, cellsurface, ALBR-bound ligands remain completely accessible to relatively large molecules (70 kDa proteins); upon shifting to low temperatures, the arrested structures are internalized in a surge, restoring 'endocytic competence'. This is suggestive of a block in a late step of endocytosis in shi ts mutants. By contrast, fluid-phase and GPI-AP endocytosis, which occurs via distinct endosomal structures, remains unaffected at the restrictive temperature in the shi ts mutants, providing evidence for a constitutive dDyn-independent pathway.

Cell culture
Stage 11 and 12 embryos were collected from flies freshly transferred to sucrose-agar bottles. Cell cultures were derived by homogenizing 40-80 embryos in a Potter-Elvehjem Tissue Grinder (2 ml; Wheaton, Millville, NJ) using a loose-fitting pestle in complete medium [Schneider's Incomplete Medium (Gibco-BRL, Gaithersberg, MD) supplemented with 10% non-heat-inactivated FBS (Gibco-BRL), 1 µg/ml bovine pancreatic insulin, penicillin, streptomycin and Lglutamine]. The resulting cell suspension was transferred to 35 mm cover-slip bottom dishes , and maintained in an incubator at 21°C. Macrophages were identified by their characteristic morphology (Eschalier, 1997) in culture after a period of 2-4 days. Complete medium was 'aged' for a period of 24-36 hours at 4°C prior to use.
Hemocytes from the third larval instar were obtained as described previously (Lanot et al., 2001) with modifications (Sriram et al., 2003). Male flies carrying UAS-GFP-protein transgenes were crossed with virgin shi ts ;CollagenGal4 flies and the progeny male larvae expressing GFP-tagged proteins were used to obtain hemocytes expressing the appropriate transgene in the mutant background. The cells were used for experiments 2 hours after dissection, unless otherwise mentioned.
Probes for endocytosis and immunodetection mBSA was prepared as described earlier (Haberland and Fogelman, 1985). Fluorescent conjugates were made according to the instructions provided by the manufacturer with minor modifications. Biotinylated-Cy3-mBSA (B-Cy3mBSA) was made by conjugation of BSA with BiotinXXSSE at a molar ratio of 1:5 according to the instructions provided by the manufacturer, prior to conjugation to Cy3 and subsequent maleylation. FITC or Lissamine Rhodamine conjugated to Dextran (10 kDa; F-Dex, LR-Dex, respectively) was used as a fluid-phase tracer.
Immunofluorescence detection of antibodies was carried out as described (Sriram et al., 2003). To enhance contrast in detecting membrane-bound forms of ectopically expressed GFP-Rab5 or GFP-Clc, cells were fixed for shorter time periods and permeabilized to remove cytosolic fluorescence.
For uptake assays at 31°C, the dishes were secured onto an aluminium block immersed in a water bath. The dishes were then covered and probes were added through a hole in the dish lid. The water bath and dishes inside were kept covered except when the probe was added or the cells were washed. In all uptake experiments at 31°C, cells were first incubated at this temperature for a period of 5 minutes prior to being labeled.
To visualize the specific endocytic uptake of GFP-GPI, cells were pre-incubated with Fl-anti-GFP (2-4 µg/ml, 0°C) for 20 minutes followed by incubation of cells at room temperature for 45 seconds or 5 minutes. PI-PLC treatment was carried out (0.3 mg/ml; 30 seconds, room temperature) to quantitatively remove cell-surface GFP-GPI and bound Fl-anti-GFP prior to fixation and imaging.
Assays for surface accessibility of mBSA Qualitative assay: after endocytosis of B-Cy3mBSA by ALBRs, to detect surface-accessible receptors, cells were washed and cooled rapidly on ice and further incubated with Cy5-SA for an additional 15 minutes to probe for surface accessibility of B-Cy3-mBSA, and then fixed and imaged on a wide-field microscope. Cy5-SA-labeled cells were imaged and the Cy5-fluorescence was completely photobleached prior to imaging Cy3 to eliminate the possibility of quenching B-Cy3mBSA fluorescence by bound Cy5-SA.
Quantitative assay: to quantify the total accessible pool of receptors, cells were labeled with B-Cy3mBSA on ice for 15 minutes, washed and then labeled with Cy5-SA for 15 minutes, and taken for imaging. These assays used medium 1 as incubation buffer since Schnieder's medium contains free biotin that competed for the binding of Cy5-SA to B-Cy3mBSA. To measure the amount accessible at any other temperature, cells were labeled with B-Cy3mBSA for a period of 5 minutes at the given temperature, washed rapidly and labeled with Cy5-SA for an additional 15 minutes at the same temperature. In experiments designed to assay the reversibility of the temperature-sensitive endocytic defect, after incubation with B-Cy3mBSA at high temperatures, cells were rapidly transferred to ice prior to the addition of Cy5-SA at the same temperature. The ratio of Cy5 to Cy3 fluorescence represents the fraction of the B-Cy3mBSA probe accessible to Cy5-SA. The ratios of Cy5 to Cy3 fluorescence at a given temperature were normalized to the ratio obtained for total accessible pool. To estimate the extent of nonspecific binding of Cy5-SA, in each experiment cells were labeled with B-Cy3mBSA for 15 minutes on ice and subsequently incubated with biocytin-treated (2 µM) Cy5-SA. The nonspecific value never exceeded 10% of the fraction of the total accessible pool in a given experiment.
Fluorescence imaging, quantification and processing Quantitative digital imaging and confocal microscopy was carried out as described previously . Fluorescence images were processed using MetaMorph software. Images were pseudo-coloured using Adobe Photoshop and composites were assembled using the same software.

Results
A system for the study of endocytosis Primary cultures derived from Canton S (CS; wild-type strain) embryos of stage 11/12 (Materials and Methods) yield neurons, different types of muscle, and fat cells, as identified by both morphology and specific markers (e.g. antibodies and GAL4mediated cell-type-specific expression of reporters (Brand and Perrimon, 1993;Eschalier, 1997;Seecof and Unanue, 1968) (A. Guha and S. Mayor, unpublished). In these cultures, macrophages are morphologically distinct and can be visually discriminated from other cells after 2-4 days ( Fig. 1A; A. Guha and S. Mayor, unpublished). Embryonic macrophages in vivo internalize polyanionic ligands (ALs) such as acetylated LDL and maleylated BSA (mBSA) via a unique class of dSRs (Abrams et al., 1992). Embryonic cultures incubated with Cy3labeled mBSA (Cy3mBSA) revealed a population of cells that specifically internalized this anionic ligand into endosomes via ALBRs (Fig. 1B). None of the other cell types in culture internalized Cy3mBSA at detectable levels at the concentrations used (data not shown). The hemolymph of third instar larva is also a rich source for cells of the hemocyte lineage (Lanot et al., 2001) that are endocytically active and show ALBR-mediated Cy3mBSA uptake (Fig. 1C,D). Moreover, the binding and internalization of Cy3mBSA by cells in both types of cultures could be competed by the addition of excess unlabeled mBSA (Fig. 1D, inset; and 1E). The binding of Cy3-mBSA could also be competed by the anionic ligands, fucoidan or LPS (Fig. 1E), but not by BSA, poly IC or Dextran (data not shown). This is similar to the binding specificity of type C1 dSRs (Pearson et al., 1995;Ramet et al., 2001). These results show that internalization of Cy3mBSA via ALBRs is likely to be mediated by an endogenous dSR.
In this study, we have mainly used larval hemocytes since they are better suited for the study of endocytic processes because large numbers of ALBR-expressing cells can be obtained more reproducibly. They also exhibit considerably less auto-fluorescence than their culture-derived embryonic counterparts. Nevertheless, qualitatively similar results have been obtained in both types of cell cultures from different genetic backgrounds (Supplementary figure S2).

Pathways of receptor-mediated and fluid-phase endocytosis in hemocytes
To study the pathways of receptor-mediated and fluid-phase endocytosis in larval hemocytes from wild-type (CS) animals, Cy3mBSA was used to label the ALBR-mediated pathway whereas FITC or Lissamine Rhodamine conjugated to Dextran (10 kDa; F-Dex, LR-Dex, respectively) were used as fluidphase tracers. When cells are pulsed with Cy3mBSA and F-Dex for 5 minutes and chased for 2 minutes both probes extensively colocalize in endosomes ( Fig. 2A-C). These endosomes are positive for the endosomal marker Rab7 ( Fig.  2D-F) and are multivesicular (Sriram et al., 2003), indicating that F-Dex and Cy3-mBSA are rapidly delivered to late endosomes in these hemocytes. These rates are similar to those observed in mammalian macrophages in culture wherein multivesicular endosomes containing fluid-phase cargo are visualized as early as 5-8 minutes after a pulse of BSA-gold in mammalian macrophages (Rabinowitz et al., 1992). Consistent with the passage through an early endosomal compartment, endosomal intermediates formed in a short pulse of 45 seconds of Cy3mBSA and F-Dex do not colocalize with Rab7 (Fig.  2G,H).
To confirm that the punctate structures formed by Cy3mBSA in short internalization times are endosomal, we incubated cells with biotinylated Cy3mBSA (B-Cy3mBSA) for 45 seconds and assayed for the surface accessibility of the biotin-tag by incubation with Cy5-labeled Streptavidin (Cy5-SA) on ice. As a control, we ascertained that, if cells were labeled at 0°C, almost all punctate Cy3mBSA structures are colocalized with Cy5-SA (data not shown). After 45 seconds, a fraction of peripherally distributed B-Cy3mBSA remains accessible to Cy5-SA, indicating cell-surface localization (compare Fig. 3A and B, small arrowheads). At the same time, endocytosed B-Cy3mBSA (defined as Cy5-SA inaccessible structures) is distributed in numerous small endosomal structures (Fig. 3A, small arrows) distinct from endosomes formed by F-Dex present in the incubation medium (Fig. 3C, arrowheads; merge in Fig. 3D). After the 45 seconds pulse, when cells were chased for 5 minutes in the absence of fluorescent probes, both Cy3mBSA and F-Dex colocalize in late endosomes (Supplementary figure S1A-C). Following a 2 minutes pulse, the majority of Cy3mBSA and F-Dex were found to colocalize in late endosomal structures, but compartments containing only F-Dex or Cy3-mBSA are also seen (Supplementary figure S1D-F).
The difference in distribution of B-Cy3mBSA-and F-Dexcontaining endosomes after a 45 seconds pulse suggested two possibilities. First, that F-Dex and B-Cy3mBSA are internalized via independent pathways. Second, that F-Dex and Cy3mBSA are internalized via a common intermediate wherein the observed segregation reflected differences in their relative rates of internalization. To discriminate between these alternatives, we incubated cells with B-Cy3mBSA alone for 3 minutes to label all early endosomal structures and then pulsed in F-Dex and B-Cy3mBSA together for an additional 45 seconds. Under these conditions, we find that the majority of F-Dex-containing structures remain devoid of B-Cy3mBSA ( Fig. 3E-H). After a 5 minutes chase, F-Dex and B-Cy3mBSA are once again extensively colocalized (data not shown).
These experiments suggest that the difference in distribution of Cy3mBSA and F-Dex after brief pulses could be due to the involvement of different endocytic pathways, and that the ALBR-mediated pathway has a poor capacity for fluid-phase uptake. Consistent with this, if cells were pulsed with tenfold higher concentrations of F-Dex (5-8 mg/ml), the fluid-phase marker was detected in the distinct endosomes as well as in Cy3mBSA-labeled endosomes under the same conditions of imaging (data not shown). These results confirm that Cy3mBSA (or B-Cy3mBSA)-labeled early endosomes do not account for a major fraction of the fluid-phase uptake in hemocytes.
The Cy3mBSA structures accessed in 45 seconds are positive for Rab5 ( Fig. 4A-C, small arrows) consistent with their designation as early endosomes (Zerial and McBride, 2001). At the same time, early (~45 seconds) endosomes of the fluid-phase pathway do not stain for Rab5 ( Fig. 4D-F, arrowheads), suggesting that these endosomal compartments are part of a Rab5-independent pathway.
In Chinese Hamster Ovary (CHO) cells, fluid-phase uptake takes place predominantly via a pathway that selectively incorporates GPI-APs. This pathway is also Rab5 independent, and independent of clathrin, dynamin and caveolin . To examine if fluid-phase uptake occurs via a similar pathway in Drosophila hemocytes, we have followed the endocytosis of exogenously added fluorescently labeled antibodies (Fl-anti-GFP) against GFP-Journal of Cell Science 116 (16) Fig. 1. Embryonic macrophages and larval hemocytes in primary culture bind and internalize Cy3mBSA via endogenous scavenger receptors. Embryonic macrophages (A, arrows) and larval hemocytes (C, arrows) in culture are the only population of cells from wild-type animals to internalize Cy3mBSA. The fluorescence image of cells (outlined in A and C) indicates that Cy3mBSA uptake is specifically mediated by ALBR endogenously expressed in embryonic macrophages (B) and larval hemocytes (D). Inset in panel D shows that binding and uptake of Cy3mBSA (D, arrows) can be competed by excess unlabeled mBSA. Histogram in panel E shows the relative amount of Cy3mBSA internalized via ALBR in hemocytes incubated in the absence (Control) or presence of excess unlabeled mBSA, fucoidan or LPS, expressed as a fraction of total endocytosed Cy3mBSA internalized in the absence of any other competing ligand. Bar, 5 µm.
GPI transgenically expressed in hemocytes using the UAS-Gal4 system (Brand and Perrimon, 1993). Visualization of endocytosed GFP-GPI was facilitated by the quantitative removal (data not shown) of cell-surface GFP-GPI (and Flanti-GFP) by treatment with PI-PLC (Materials and Methods). Observation at 45 seconds post-internalization shows that the peripheral small early endosomes containing Cy5mBSA as well as the large fluid-filled early endosomes are both labeled with PI-PLC-resistant Fl-anti-GFP specifically internalized by GFP-GPI (Fig. 5A,B); cells that do not express GFP-GPI do not show uptake of detectable amounts of Fl-anti-GFP (data not shown). These results suggest that, similar to CHO cells, the fluid-phase uptake is also mediated by a Rab5-negative, GPI-AP-enriched endosomal pathway. At later times, Fl-anti-GFP (endocytosed via GFP-GPI) is delivered to late endosomes (Fig. 5C,D). The delivery of GPI-APs to late endosomes in Drosophila hemocytes is in contrast to the eventual delivery of GPI-APs to endosomal recycling compartments in CHO cells. However, it is similar to the route followed by GPI-APs in Baby Hamster Kidney cells, consistent with the differential sorting and fate of endocytosed GPI-APs observed in multiple cell types (Fivaz et al., 2002;Sabharanjak et al., 2002;Sharma et al., 2002).
Together, these results suggest that hemocytes from Drosophila internalize ALBR-bound ligands and fluid-phase markers into two distinct classes of early endosomes, Rab5positive and negative, respectively. These later fuse to form Rab7-positive late endosomes.

Role of dDyn in endocytosis
We next asked if dDyn plays any role in endocytosis of the two types of probes. For this purpose we obtained cells from shi ts alleles, shi ts1 and shi ts2 . Both these alleles contain missense mutations (van der Bliek and Meyerowitz, 1991) conserved across all splice variants (Staples and Ramaswami, 1999). The shi ts2 mutation (G141S) lies in the GTP binding and hydrolysis domain (G domain) and the shi ts1 mutation (G267D) lies at the interface of the G domain and the 'middle' domain ( Fig. 6A). shi ts1 and shi ts2 flies paralyze within 3 minutes at 27°C and 27.5°C (the restrictive temperature), respectively, and recover just as rapidly when returned to 21°C (the permissive temperature) (Krishnan et al., 1996).
Hemocytes from CS, shi ts1 and shi ts2 animals were pulsed with Cy3mBSA and F-Dex for 5 minutes at 21°C. At this temperature, the uptake of both probes in shi ts1 and shi ts2 cells is indistinguishable from that of CS cells; both Cy3mBSA and F-Dex colocalize in large centrally located endosomes (Fig.  6B-D). In CS cells, following a 5 minutes pulse at 31°C, Cy3mBSA and F-Dex are seen in endosomes indistinguishable from those formed at 21°C (compare Fig. 6B with 6E). However, the distribution of Cy3mBSA is dramatically altered in both mutant shi alleles; Cy3mBSA has a punctate peripheral distribution (Fig. 6F,G; small arrowheads). The distribution of Cy3mBSA in shi ts1 and shi ts2 cells at 31°C is comparable to cells labeled on ice (Fig. 6H-J), a condition in which all endocytic activity is inhibited, suggesting an arrest in the internalization of ALBR-ligands. In contrast to Cy3mBSA, fluid-phase endocytosis appears unaffected; F-Dex is internalized into similar endosomes in cells from the mutant shi animals (Fig. 6F,G; green) when compared with the endocytic structures observed in cells from CS animals (Fig. 6E, green) at the non-permissive temperature (see below for a detailed analysis of fluid-phase and GPI-AP endocytosis in shi ts mutants). These observations confirm that the inhibition of ALBR internalization at the restrictive temperature is not result of any nonspecific cellular toxicity.
Nature of endocytic defect in shi ts mutants at restrictive temperature The process of formation of the endocytic vesicle minimally comprises three distinct stages: invagination, closure and fission (Sever et al., 2000a). At the synapse, shi ts mutants exhibit a block at a late stage in this process (Ramaswami et al., 1994). To understand the nature of the endocytic defect in hemocytes, and to quantify the extent of Cy3mBSA internalization at 31°C, we devised a quantitative surface-accessibility assay (Fig. 7A). In this assay, we determined the accessibility of Biotinylated-Cy3mBSA (B-Cy3mBSA) to exogenously added Cy5-labeled Streptavidin (Cy5-SA). Thus, the ratio of Cy5-SA to B-Cy3mBSA fluorescence is a read out of the extent of the accessibility of B-Cy3mBSA to Cy5-SA added at a given temperature (Materials and Methods). Maximum accessibility was determined from cells incubated with B-Cy3mBSA on ice followed by Cy5-SA incubation at the same temperature (Fig. 7B, shi ts1 ice). In shi ts1 cells at 31°C, most of the B-Cy3mBSA was surface accessible; Cy5 to Cy3 fluorescence ratio in a population of shi ts1 cells labeled on ice is indistinguishable from that observed when the cells are labeled at 31°C (Fig. 7B, shi ts1 31°C). In sharp contrast, only a small fraction of the ALBR ligand is accessible to Cy5-SA at the permissive temperature of 21°C; Cy5 to Cy3 fluorescence ratio in a population of shi ts1 cells at 21°C (Fig. 7B, shi ts1 21°C) resembles that obtained for nonspecific binding of the Cy5-SA to cells (Fig. 7B, shi ts1 Noise). Greater than 90% of total ligand-bound ALBR remained arrested at the cell surface in both shi ts1 and shi ts2 cells at 31°C. At the same temperature, 50% of ligand-bound ALBR remains at the cell surface in CS cells (Fig. 7C), compared with the ~25% detected at 21°C. These experiments show that shi ts1 and shi ts2 cells quantitatively arrested ALBR sequestration at 31°C but not at 21°C, confirming that ligand-bound ALBR is internalized by a dDyndependent mechanism. These results are consistent with studies conducted in mammalian cells overexpressing Dyn ts (analogous to shi ts1 ) on the internalization of biotinylated transferrin endocytosed via the transferrin receptor wherein the coated pits with wide openings also appear to accumulate at the restrictive temperature (Damke et al., 1995).
To test whether the reversible nature of the temperature-sensitive shi mutation (Koenig et al., 1983;Kosaka and Ikeda, 1983b) is reproduced in the primary culture cells, we asked if Cy3mBSA arrested at the cell surface at 31°C in shi ts cells (Fig. 7C) is capable of Journal of Cell Science 116 (16) Fig. 3. B-Cy3mBSA and F-Dex are internalized into distinct endosomal compartments at very early times. Hemocytes from CS animals were incubated at room temperature with B-Cy3mBSA (A,E; red in D,H) and F-Dex (C,G; green in D,H) for 45 seconds either without (A-D) or with (E-H) a pre-pulse of B-Cy3mBSA for 3 minutes. Cells were then immediately chilled on ice and Cy5-SA (B,F; blue in D,H) was added on ice to detect surface-accessible receptors, prior to fixation and imaging on wide-field microscope. Note that after 45 seconds a significant fraction of the cell-associated B-Cy3mBSA is distributed in endosomes inaccessible to Cy5-SA (A,D and E,H; arrows). Most of the brightest F-Dex-labeled structures (C,D and G,H; arrowheads) do not colocalize with B-Cy3mBSA. A large fraction of the peripherally distributed B-Cy3mBSA is colocalized with Cy5-SA, remaining surface accessible at both times (A,B,D, and E,F,H; small arrowheads). Insets show magnified view of the area marked by an asterisk. Bar, 5 µm; inset, 1 µm. shibire-independent pinocytosis in Drosophila being internalized into subsequently formed Cy5mBSAcontaining endosomes at the permissible temperature (see scheme in Fig. 8A). The extensive colocalization of Cy3mBSA in Cy5mBSA-containing endosomes formed at the permissive temperature in hemocytes from shi ts animals provides evidence for the reversible nature of the arrest of dDyn-mediated internalization in these cells (Fig. 8B).
To gain further insight into the nature of structures containing surface-accessible receptors that accumulate at the restrictive temperature we followed the scheme outlined in Fig.  8C. This protocol addresses the rapidity with which surfaceaccessible structures that accumulate at the restrictive temperature in shi mutant cells become inaccessible to largemolecular-weight molecules (Cy5-SA). After rapid transfer to ice (<5 seconds to equilibrate to 0°C), we find that a surprisingly significant fraction (~65%) of surface-accessible B-Cy3mBSA pre-bound at 31°C (Fig. 7C) becomes inaccessible to Cy5-SA (Fig. 8D). In comparison with the quantitative accessibility of B-Cy3mBSA at 32°C, transfer to ice results in a significant reduction in accessible receptors. This suggests a rapid sequestration of the B-Cy3mBSA in this experimental paradigm. As observed in a representative figure (right panel in Fig. 8C), although the majority (87%) of the B-Cy3mBSA fluorescence remains inaccessible to Cy5-SA, the B-Cy3mBSA is predominantly peripherally distributed (Fig.  8C, 32°C, arrowheads), which is indistinguishable from cells labeled on ice (Fig. 6H-J), or cells from shi ts animals held at the high temperature (Fig. 6F,G). The distribution of Cy5-SAinaccessible ligands in this paradigm is quite different from the distribution of internalized B-Cy3mBSA in cells incubated at the permissive temperature (21°C) post to transfer to ice (Fig.  8C, 21°C). These results suggest that the temperature-sensitive step in dDyn mutants is at a late step in the sequestration process mediated by dDyn.

Surface-accessible ligands accumulate in clathrin-and adaptin-positive structures
Since various Dyn-dependent pathways described in mammalian cells (McNiven et al., 2000) are clathrin dependent or independent, we ascertained whether internalization of ALBR occurs by a clathrin-mediated pathway. For this purpose, we examined the colocalization of ectopically Fig. 4. Cy3mBSA is internalized into GFP-Rab5-positive early endosomes distinct from LR-Dex early endosomes that are devoid of GFP-Rab5. Hemocytes from Collagen-Gal4;UAS-GFPdRab5 larvae were incubated for 45 seconds with Cy3mBSA (A-C) or LR-Dex (D-F) prior to fixation, permeabilization and imaging on wide-field microscope. At 45 seconds, a fraction of Cy3mBSA-labeled structures (A, red in C; small arrows) are also positive for GFP-dRab5 (B, green in C; small arrows). Several peripherally distributed Cy3mBSA-labeled structures (A, red in C; small arrowheads) are negative for GFP-dRab5 (B, green in C) and are likely to be surface accessible (see Fig. 3). LR-Dex-containing endosomes (D, red in F; arrowheads) do not colocalize with GFP-dRab5-positive structures (E; green in F; small arrows). Insets show magnified view of the area marked by an asterisk. Bar, 5 µm; inset, 1 µm.

Fig. 5. GFP-GPI is internalized into LR-Dex early endosomes and
Cy5mBSA early endosomes. Hemocytes from Collagen-Gal4;UAS-GFP-GPI (A-D) were incubated with Fl-anti-GFP (pseudo-colored green; Cy5-anti-GFP in A,C and Alexa568-anti-GFP in B,D) at room temperature for the indicated times in the presence of LR-Dex (A,C; red) or Cy5mBSA (B,D, red). Note that Fl-anti-GFP is partially colocalized with LR-Dex (A, arrowheads; C, arrows; insets) and with Cy5mBSA (B, small arrows; D, arrows; insets) at these times. However, at late times, all three probes colocalize in late endosomes (arrows, C,D). Insets (red, top; green, middle; merge, bottom) show the area marked by the asterisk. Bar, 5 µm; inset, 1 µm. expressed GFP-clathrin light chain (GFP-Clc) (Chang et al., 2002) with Cy3mBSA-labeled ALBR incubated at 0°C. We find that most of the surface-localized (and quantitatively surface-accessible; Fig. 7B) receptors are present in GFP-Clcpositive punctate structures ( Fig. 9A-C, small arrowheads). At the restrictive temperature in shi ts2 cells, the surface-accessible receptors are also localized in GFP-Clc-positive structures ( Fig. 9D-F, small arrowheads). Furthermore α-adaptin, a component of the clathrin-recruiting AP-2 complex (Chang et al., 1993), also colocalizes with the peripheral, surfaceaccessible Cy3mBSA puncta in both shi ts1 and shi ts2 cells (Supplementary figure S3). These data support the conclusion that ALBR ligands are internalized into primary larval hemocytes by a clathrin-and dDyn-mediated pathway.
Fluid-phase and GPI-AP endocytosis in shi ts cells Unlike the internalization of Cy3mBSA, uptake of F-Dex into shi ts1 and shi ts2 cells at the restrictive temperature does not appear to be perturbed (Fig. 6E-G). A quantification of the amount of F-Dex internalized (Fig. 10A) shows that fluid-phase uptake in shi ts1 and shi ts2 cells at 21°C, 31°C and 33°C is comparable with CS cells at the same temperatures. A common pathway of endocytosis for both Cy3mBSA and F-Dex would predict that a decrease in the rate of mBSA internalization should be accompanied by a decrease in the rate of F-Dex uptake. This does not take place, thus ruling out the possibility of a common pathway. It is unlikely that the conditions of primary culture medium components upregulate alternative endocytic pathways in hemocytes, since cells directly extracted into buffered saline also show dDynindependent F-Dex internalization (Supplementary figure  S2A-F).
To determine whether F-Dex-containing endosomes formed at the higher temperature in shi ts cells are bona fide endocytic structures, we examined if these endosomes are capable of merging with ALBR-ligand-containing endosomes formed during the recovery of shi ts cells at 21°C. For this purpose, we first incubated cells with F-Dex (and Cy3mBSA) at 31°C and then shifted the temperature to 21°C in the presence of Cy5mBSA (see schematic, Fig. 8A). After shifting to 21°C, Cy3mBSA, Cy5mBSA and F-Dex colocalize in both shi ts1 and Journal of Cell Science 116 (16) (red) and F-Dex (green) into endosomes following a pulse for 5 minutes at 21 o C (B, arrows) and 31°C (E, arrows). Cells from shi ts1 (C,F,I) and shi ts2 (D,G,J) animals internalize Cy3mBSA and F-Dex at 21°C (C,D, arrows). At 31°C in cells from shi ts animals (F,G), Cy3mBSA remains peripherally distributed in punctate (small arrowheads) and hazy distribution, whereas F-Dex continues to be internalized into large endosomes (arrows). Cells labeled with Cy3mBSA for 20 minutes on ice (H,I,J) show a distribution similar to shi ts cells at 31°C (small arrowheads). Bar, 5 µm. shibire-independent pinocytosis in Drosophila shi ts2 cells (Fig. 8B). These results show that the fluid-filled structures formed at restrictive temperatures are capable of merging with ligand-bound ALBR subsequently internalized from the cell surface in shi ts cells at the permissible temperature.
The endocytosis of GPI-APs via the fluid-filled endosomes in mammalian cells has been shown to be Dyn independent . To ascertain the requirment for dDyn in GPI-AP internalization, we examined GPI-AP endocytosis in cells from shi ts animals. At the restrictive temperature, Fl-anti-GFP is taken up into cells from shi ts2 in fluid-filled endosomes similar in morphology (Fig. 10B) and extent as in cells from wild-type animals at the same temperature (data not shown). Furthermore, in the same cells, Cy5-mBSA is excluded from the endocytic pathway at the restrictive temperature (Fig. 10C). In cells expressing comparable levels of GFP-GPI, the ratio of internalized Fl-anti-GFP to the total amount of GFP fluorescence is 0.57±0.37 (15 cells) in wild type and 0.79±0.27 (20 cells) in the shi ts2 background. Thus, the shi ts2 mutation does not prevent the internalization of GPI-APs. This further reiterates the existence of a distinct dDynindependent pathway involved in GPI-AP endocytosis in Drosophila hemocytes. Taken together, these data show that mBSA internalization and F-Dex (and GPI-AP) uptake occur by distinct mechanisms; ALBR-bound mBSA internalization is dDyndependent. The cartoons indicate the expected surfaceaccessible or -inaccessible status of ALBR ligand (B-Cy3mBSA) as probed using Cy5-SA. (C) Histogram shows the average extent of B-Cy3mBSA bound to the cells that remains accessible to Cy5-SA (Cy5/Cy3 ratio) after a pulse for 5 minutes at 21°C and 31°C in the indicated alleles, relative to the total accessible pool of receptors (Cy5/Cy3 ratio determined for Ice condition; arrow). Bars represent mean±s.e. obtained from three experiments, each with at least 75 cells from two dishes in each condition. Note that both shi ts alleles quantitatively arrest Cy3mBSA internalization at this temperature.

Discussion
The results presented in this report provide evidence that primary cultures of Drosophila hemocytes are capable of at least two distinct constitutive pathways of endocytosis (see model in Fig. 10D shibire-independent pinocytosis in Drosophila figure S2). In terms of the molecular identity of the ALBR, the competition studies carried out in larval hemocytes suggest that the endogenous ALBR has a binding specificity similar to that of type C1 dSRs (Pearson et al., 1995;Ramet et al., 2001). Furthermore, our results suggest that ALBR is internalized via a clathrin-and adaptin-mediated pathway and that type C1 dSRs have an adaptin-binding motif (YXXφ) at 549-552. However, these results do not unequivocally establish that the ALBR is identical to a specific dSR. At the restrictive temperature, the receptors are trapped at the cell surface in clathrin-and α-adaptin-decorated sites, accessible to large molecules (Cy5-SA). This accessibility is quantitatively similar to that observed when cells from both wild type and mutants are directly labeled on ice, indicating a complete blockage of uptake and sequestration. These results are analogous to observations in mammalian cells overexpressing Dyn ts , a homologue of shi ts1 , wherein coated pits with wide openings accumulated at the restrictive temperature (Damke et al., 1995). In contrast to studies where, in mammalian cells overexpressing Dyn ts , the receptors retained avidin accessibility even after transfer to ice (Damke et al., 1995), the receptors in shi ts cells pre-treated at restrictive temperature become completely sequestered upon transfer to ice. This difference is likely to reflect the intrinsic nature of mutations in dDyn in the absence of interference from copolymerization with endogenous wild-type Dyn molecules. The rapidity with which receptor sequestration takes place (within 5 seconds), suggests that the block in dDyn function in the shi ts mutants is at a stage where commitment to endocytic sequestration has already occurred. The sharp transition in the extent of sequestration induced by shifting to ice after incubation at the restrictive temperature is suggestive of a structural role for dDyn in sequestration. Consistent with these results is a previous report where synaptic endocytosis is blocked in shi mutants at a late stage (Ramaswami et al., 1994).
Current models for the role of dynamin in receptor-mediated endocytosis via clathrin-coated pits fall into two broad classes (reviewed in Hinshaw, 2000;Sever et al., 2000a), one in which the GTPase activity of the enzyme is directly involved in a mechano-chemical scission of the invaginated pit (Marks et al., 2001), and the other where Dyn is a regulatory GTPase involved in recruiting molecular players responsible for the fission process (Sever et al., 2000b). Although our results do not distinguish between these two models, they support a structural role for dDyn at late stages in the sequestration process without precluding a regulatory role in the earlier stages of the formation of the endocytic invagination. Understanding of the stage in the GTP-GDP cycle at which the shi mutants are blocked will further discriminate between the two models (Damke et al., 2001).
These results reported here also demonstrate that the process of dDyn-dependent internalization is intrinsically temperature sensitive; >85% of the receptors are sequestered at permissive temperatures compared with ~50-60% at the restrictive Fig. 9. ALBR is localized to GFP-clathrin light chain marked sites at the plasma membrane. Hemocytes from Collagen-Gal4;UAS-GFP-Clc (A-C) or shi ts2 ;Collagen-Gal4;UAS-GFP-Clc animals (D-F) were incubated with Cy3mBSA (A,D; red in C,F) on ice (0°C; A-C) or at a restrictive temperature (31°C; D-F) for 5 minutes prior to fixation, permeabilization and imaging on wide-field microscope. On ice, the peripherally distributed Cy3mBSA-positive structures (A, red in C; small arrowheads) are positive for GFP-Clc (B, green in C; small arrowheads). At 31°C in cells from shi ts2 animals, Cy3mBSA remains peripherally distributed in punctate (D, small arrowheads) structures also positive for GFP-Clc (E, small arrowheads). This distribution in shi ts2 cells at 31°C is similar to cells labeled with Cy3mBSA on ice (small arrowheads). Insets show magnified view of the area marked by an asterisk. Bar, 5 µm; inset, 1 µm. temperature in the same time interval in wild-type flies. Interestingly, this also has a correlate in the temperature sensitivity of the paralysis of wild-type (CS) flies in our hands; CS flies paralyze in a manner very similar to shi ts flies but at temperatures >40°C. Mutations at the shi locus, shi ts1 and shi ts2 , can thus be seen as enhancers of the intrinsic temperature sensitivity of this process. By contrast, fluid-phase uptake does not exhibit any temperature sensitivity in wild-type and shi ts animals, arguing strongly for independent regulation. Cells completely lacking any shibire gene product are likely to be autonomously cell-lethal (Grant et al., 1998), precluding an analysis of fluid-phase uptake in cells without any dDyn expression. Therefore, these data do not completely rule out a role for dDyn in fluidphase uptake but they certainly rule out a role similar to the function of dDyn in receptor-mediated uptake in this endocytic process. Uptake of fluid occurs by various means (Lamaze and Schmid, 1995); however, except for macropinocytosis and the pathway for the internalization of GPI-APs , all other pathways require Dyn function in mammalian cells. Live imaging with cultured larval hemocytes expressing either GFP-actin or other actinbinding proteins fused to GFP have revealed that these cells undergo extensive membrane ruffling (A. Guha and S. Mayor, unpublished). Whereas macropinocytosis correlates with membrane ruffling (Racoosin and Swanson, 1992), the earliest F-Dex-filled compartments are unlike typical macropinosomes. Macropinocytosis is usually a triggered process resulting in large, phase-lucent organelles at the time of formation (Racoosin and Swanson, 1992). Although endosomes labeled in a 5 minutes pulse in larval hemocytes develop into phasebright structures, the majority of the earliest F-Dexfilled endosomes formed in a 45 seconds pulse are not, arguing against the dDyn-independent pathway resembling macropinocytosis.
Another marker that has been extensively used to study fluid-phase endocytosis in Drosophila and other cell culture systems is HRP. The reduced uptake of HRP seen in garland cells from shi ts1 animals at the restrictive temperature (Kosaka and Ikeda, 1983b) would, therefore, appear contrary to the results presented here. However, at comparable molar concentrations, it is likely that HRP may be taken up via multiple means, including receptor-mediated pathways. We find that HRP is internalized via a yeast-mannan-competable pathway in hemocytes (Sriram et al., 2003), consistent with the possibility that HRP uptake is not restricted to the fluid phase. Furthermore, similar to the distinct cdc42-regulated, dynamin-independent pinocytic pathway recently described in mammalian cells , our results show that the Drosophila hemocytes are also capable of dDyn-independent GPI-AP endocytosis via the fluid-phase pathway. Another point of similarity with the fluid-phase pathway recently described in mammalian cells is the absence of Rab5 on fluid-phase-containing endosomes . This suggests that Rab5 defines only a subset of early endosomal Journal of Cell Science 116 (16) Fig. 10. Fluid-phase and GPI-AP uptake is unperturbed at restrictive temperatures in cells from mutant alleles of shi. Histogram shows the average amount of F-Dex (FDx; ±s.d.) internalized per cell in the indicated alleles and temperatures in an incubation period of 5 minutes. The similarity in uptake across all alleles, even at restrictive temperatures, demonstrates that mutations at the shi locus do not perturb the uptake of the fluid phase. Data shown are from three experiments each with at least 75 cells. (B,C) Hemocytes from shi ts2 ;Collagen-Gal4;UAS-GFP-GPI animals were incubated with Fl-anti-GFP (green) at 31°C for 15 minutes in the presence of LR-Dex (B, red) or Cy5mBSA (C, red) prior to fixation and imaging on a wide-field microscope. Endocytosis of Fl-anti-GFP into LR-Dex (B, arrows)-containing endosomes is unaffected, whereas Cy5mBSA is blocked at the cell surface (C, small arrowheads). Insets (red, top; green, middle; merge, bottom) show the area marked by the asterisk. Bar, 5 µm; inset, 1 µm. (D) Model depicts the existence of dDyn-independent endocytic pathways for the fluid phase (yellow) and GPI-APs (green) in larval hemocytes. ALBR ligands (red) are endocytosed via clathrin and dDyndependent pathways into Rab5-positive early endosomes (EE), whereas the fluid phase (yellow) marks a separate Rab5-negative EE, before finally coming together in Rab7-positive late endosomes (LE). compartments, strengthening the notion that there may be multiple early endosomal systems that are connected to each other via cell-type-dependent trafficking pathways (Fivaz et al., 2002;Sabharanjak et al., 2002). This primary cell culture system will provide an opportunity for analysis of the molecular players involved in this pathway in a genetically amenable system, and will allow an exploration of the trafficking consequences of alterations in this pathway in the whole animal.
In conclusion, we show that Drosophila cells exhibit Dyn (dDyn)-dependent and -independent pathways of endocytosis. These pathways converge in a population of Rab7-positive endosomes and subsequently target endocytosed cargo for degradation in lysosomes (Sriram et al., 2003). It is conceivable that several GPI-AP-containing membranes in Drosophila, including 'argosomes' (Greco et al., 2001) and modulators of signaling such as heparan sulfate proteoglycans (Selleck, 2000) are endocytosed via the dDyn-independent pathway, distinct from dDyn-dependent pathways utilized to regulate signaling receptors (Ceresa and Schmid, 2000;Pierce and Lefkowitz, 2001). This could provide a mechanism by which cells can independently modulate their responses to the extracellular environment. At least in the case of Notch and Delta, the clearance of inactive Notch at the cell surface appears to be facilitated by a dDyn-independent pathway (Parks et al., 2000). The primary culture system described in this report makes it possible to examine, at high resolution, endocytic trafficking of molecules internalized by various pathways in Drosophila and use available genetic tools to dissect these pathways.
We thank N. Gay for anti-α-adaptin antibodies; M. Zerial for anti-Rab7 antisera; V. Verkhusha for GFP-actin; D. Kiehart for GFPmoesin; M. Gonzalez-Gaitan, I. Mellman, S. Eaton and C. Dearolf for fly strains; Rahul Chadda for PI-PLC and critically reading the manuscript. A.G. was supported by a PhD fellowship from TIFR. V.S. is supported by a Kanwal Rekhi Fellowship from the TIFR Endowment Fund, and Appam. This work was supported by a grant from the Department of Biotechnology, India, and from intra-mural funds from NCBS. S.M. is a Senior Research Fellow of the Wellcome Trust (UK), and thanks K. Belur and F. F. Bosphorus for inspiration.