Endosome Sorting and Autophagy Are Essential for Differentiation and Virulence of Leishmania major*

Cellular remodeling during differentiation is essential for lifecycle progression of many unicellular eukaryotic pathogens such as Leishmania, but the mechanisms involved are largely uncharacterized. The role of endosomal sorting in differentiation was analyzed in Leishmania major by overexpression of a dominant-negative ATPase, VPS4. VPS4E235Q accumulated in vesicles from the endocytic pathway, and the mutant L. major was deficient in endosome sorting. Mutant parasites failed to differentiate to the obligate infective metacyclic promastigote form. Furthermore, the autophagy pathway, monitored via the expression of autophagosome marker GFP-ATG8, and shown to normally peak during initiation of metacyclogenesis, was disrupted in the mutants. The defect in late endosome-autophagosome function in the VPS4E235Q parasites made them less able to withstand starvation than wild-type L. major. In addition, a L. major ATG4-deficient mutant was found also to be defective in the ability to differentiate. This finding, that transformation to the infective metacyclic form is dependent on late endosome function and, more directly, autophagy, makes L. major a good model for studying the roles of these processes in differentiation.

Cargo destined for the lysosome/vacuole compartment in eukaryotic cells requires sorting from the biosynthetic, secretory, and endocytic pathways. Transport of newly synthesized lysosomal proteins from the trans-Golgi network to the lysosomal compartment involves the delivery of cargo into the endosomal system, which serves as a sorting compartment as well as a collection site for endocytosed materials. The endocytic process, which is important for internalization of portions of the plasma membrane as well as extracellular fluids, involves a number of mechanisms (many of them receptor-mediated), including phagocytosis, macropinocytosis, caveolae, and clathrin-dependent or clathrin-independent endocytosis (1). Once internalized, receptor-bound ligands, solutes, and lipids are subject to complex intracellular trafficking pathways and can be recycled back to the plasma membrane or degraded in the lysosomal compartment. The endosomal system itself is complex, and several types of vacuoles have been identified, including late endosomes having a multivesicular aspect and called multivesicular bodies (MVB 2 ; see Refs. 2 and references therein for a review).
A screen for vacuolar protein sorting (Vps) mutants in yeast led to the discovery of the molecular machinery responsible for MVB formation and, in particular, the class E group of Vps mutants, which contained a large multilamellar cisternal compartment thought to represent an endosome unable to form intraluminal vesicles (3,4). Vps class E proteins are organized in several sub-complexes named "endosomal sorting complexes required for transport" (ESCRT) I, II, and III (5)(6)(7)(8). The ESCRT proteins are recruited from the cytoplasm to the endosomal membrane where they function sequentially in the formation of MVB vesicles and the sorting of proteins into the MVB pathway. It has been proposed that a multimeric AAA-type ATPase, Vps4, binds to ESCRT III and catalyzes the disassembly of this complex in an ATP-dependent manner (9). This Vps4-dependent dissociation of the ESCRT machinery is the final step of protein sorting into the MVB and is a prerequisite for vesicle formation (6). The Vps4 homologue in mammals, SKD1, is involved in membrane transport through endosomes and overexpression of a dominant negative mutant, SKD1 E235Q , resulted in the production of aberrant endosomes defective in membrane transport between late endosomes and lysosomes (10).
In the protozoan parasite Leishmania, as in related kinetoplastid flagellates such as Trypanosoma, there is a complex membrane network highly polarized around the flagellar pocket, an invagination of the plasma membrane where the flagellum emerges from the cell body (11). This is an important zone of interaction between the parasite and its environment, because it is the only site in the cell for endocytosis and exocytosis and so is responsible for crucial exchanges such as uptake of nutrients via receptor-mediated endocytosis (12) and secretion of virulence factors that can interact with the host. Another characteristic of Leishmania parasites is that their promastigote (insect stage) form possesses a rather unusual lysosomal compartment, named the multivesicular tubule (MVT)-lysosome (13)(14)(15). This compartment has a low lytic capacity and a relatively high luminal pH in multiplicative procyclic promastigotes but acquires the properties of mature lysosomes as the parasite differentiates into its infective, but non-replicative, metacyclic form (13). So far, only a few components of the Leishmania endosomal machinery have been characterized at the molecular level such as earlyendosomal Rab5 (16) and late-endosomal Rab7 (17). Indeed, although it has been shown that the MVT-lysosome is downstream of a MVB-like network of vesicular endosomes that surrounds the flagellar pocket (13), very little is known about them in Leishmania.
Among the pathways involving vesicular traffic in eukaryotic cells is autophagy, a process that is important for protein and organelle degradation during cellular differentiation and also as a defense against starvation conditions (18). The autophagic pathways of yeast and mammals have been characterized, and, although they have many common features, they appear to differ in some ways (19,20). Central to the process in both is a vesicular compartment called the autophagosome, which is formed by a membranous structure that engulfs the cytoplasm/organelles that are to be degraded. The genesis of autophagosomal struc-tures requires the activity of two protein conjugation systems, one involving ubiquitin-like protein Atg8 proteolytically processed by Atg4, and a second involving ubiquitin-like protein Atg12, covalently attached to Atg5 (18,20,21). The autophagosome delivers the internalized material to the lysosomal compartment for degradation. It seems that, at least in mammals, but perhaps not in yeast, autophagosomes first fuse with endosomal vesicles (22)(23)(24). Thus in this case, the autophagosome requires strong interactions with the endosomal compartments to reach its full degradative potential. Among the cellular functions proposed for autophagy is a role in the architectural changes occurring during development and differentiation, although experimental evidence supporting this hypothesis is not extensive (18). Recent reports, however, have shown that stress-induced differentiation is impaired in autophagy mutants of Dictyostelium discoideum (25), Caenorhabditis elegans (26), and even yeast, where autophagy genes seem to be important for sporulation (27).
In this study, we investigated the function of MVBs in Leishmania major by characterizing the leishmanial Vps4 homologue. A dominantnegative VPS4 mutant (VPS4 E235Q ) accumulated the mutated protein around vesicular structures of the endocytic system and showed a defect in transport to the MVT-lysosome. This is similar to what has been observed in yeast and mammalian Vps4 mutants, suggesting a conservation of role for this protein in MVB architecture from the early branching kinetoplastid flagellate lineage to mammals. Moreover, L. major-overexpressing VPS4 E235Q were impaired in their differentiation in culture and their resistance to starvation, suggesting a crucial role for VPS4 and the MVB compartment in these processes. In addition, by demonstrating the presence of autophagosomes in Leishmania using ATG8 as a marker and showing that these structures accumulated and were not functional in the VPS4 E235Q mutant, we have revealed that the autophagic pathway is intimately involved with the endosomal system. Production of an L. major ATG4.2 null mutant, which could not undergo metacyclogenesis, further confirms that the lack of a functional autophagic pathway correlates with, and probably mediates, the differentiation defect of the VPS4 mutant parasite that impairs its virulence.

EXPERIMENTAL PROCEDURES
Parasites-L. major (MHOM/JL/80/Friedlin) promastigotes were grown in modified Eagle's medium with 10% (v/v) heat-inactivated fetal calf serum (designated complete HOMEM medium) at 25°C. Metacyclic promastigotes were isolated by the peanut agglutinin method (28) from cultures that had been in stationary growth phase for 2-5 days. When referring to the stage of growth of the different cell lines throughout this report, early log phase corresponds to ϳ5 ϫ 10 5 parasites/ml and mid-log phase to ϳ5 ϫ 10 6 parasites/ml. Early stationary phase corresponds to ϳ9 ϫ 10 6 parasites/ml for pXG-VPS4 E235Q and to ϳ2 ϫ 10 7 parasites/ml for pXG-VPS4 or wild type L. major. Unless otherwise stated, the pXG-VPS4 and pXG-VPS4 E235Q cell lines were maintained in culture with 12.5 g/ml G418.
The GFP fusion constructs were obtained as follows. LmjVPS4 was obtained by PCR from plasmid pXG-VPS4 with primers OL1543 and OL1544 and cloned into BglII/KpnI-digested GFP-containing pNUS-GFPnH vector (30). The resulting plasmid was named GFP-VPS4 and contained LmjVPS4 in-frame with the 3Ј-end of GFP. A plasmid (GFP-VPS4 E235Q ) containing LmjVPS4 E235Q was obtained by site-directed mutagenesis on plasmid GFP-VPS4, as described above. L. major ATG8 was amplified by PCR from genomic DNA with primers OL1506 and OL1507 and cloned into BglII/XhoI-digested pNUS-GFPnH vector to give GFP-ATG8.
Generation of a L. major ATG4.2 Null Mutant and Re-expressing Cell Lines-The 1005-bp 5Ј-flank fragment of LmATG4.2 (LmjF30.0270) was generated by PCR from L. major genomic DNA with primers NT168 and NT169 (Table 1), digested with HindIII and SalI, and inserted into HindIII/SalI-digested pGL345-HYG (31) to give pGL345ATG4.2-HYG5Ј. The 3Ј-fragment was generated by PCR using primers NT170 and NT171. The resulting 1104-bp fragment was digested by SmaI and BglII and cloned into pGL345ATG4.2-HYG5Ј to give pGLATG4.2-HYG5Ј3Ј. The cassette used for transfection was released by HindIII/ BglII digestion. pGLATG4.2-BSD5Ј3Ј plasmid, used for the replacement of the second ATG4.2 allele, was generated from plasmid pGLATG4.2-HYG5Ј3Ј by replacing the SpeI/BamHI cassette containing the hygromycin resistance gene by a SpeI/BamHI cassette containing the blasticidin S deaminase gene. A population of parasites resistant to hygromycin was generated after transfection with pGLATG4.2-HYG5Ј3Ј. This population was used for a second round of transfection with the pGLATG4.2-BSD5Ј3Ј construct. Two blasticidin-resistant clones from independent transfection events, designated ⌬atg4.2A1 and ⌬atg4.2B4, were selected for analysis.
For the re-expression experiments, the ATG4.2 gene, modified with a poly-histidine tag at its C-terminal end, was inserted into the pNUS episomal vector. PCR using the primers NT158 and NT159 produced the poly-histidine-tagged version of ATG4.2. The resulting 1185-bp fragment was digested by NdeI/XhoI and ligated into pNUS-HnN plasmid (30), previously digested by the same enzymes, to give the pN-ATG4.2 plasmid. L. major wild-type promastigotes were electroporated with 15 g of the episome, and transfectants were selected with the appropriate antibiotic (Geneticin, G148, Invitrogen).
Complement-mediated Lysis and Macrophage Infections-The sensitivity of procyclic and metacyclic promastigotes to complement lysis was assessed with a protocol modified from (32). Briefly, 10 7 parasites were incubated at 37°C in PBS mixed with an appropriate volume of fresh normal human serum. After 20 min, the samples were diluted 2-fold in PBS. The percentage of lysis, assessed by loss of promastigote motility and morphological changes, was calculated relative to control samples incubated in heat-inactivated serum, in which all cells remained viable.
Peritoneal macrophages from CD1 mice were adhered overnight in RPMI medium (Sigma) at 37°C in 5% CO 2 /95% air onto eight-chamber tissue culture plastic slides (Labtech) and then infected using late stationary phase promastigotes or peanut lectin-purified metacyclics at a ratio of 10 promastigotes per macrophage. After incubation for 3 h at 35°C in 5% CO 2 /95% air, non-phagocytosed promastigotes were removed by washing gently with RPMI four times, and the cultures were then incubated at 35°C in 5% CO 2 /95% air. The initial uptake of promastigotes by macrophages and their subsequent intracellular survival and growth as amastigotes was determined by counting the number of infected macrophages and the number of intracellular parasites in stained slides 3, 24, and 120 h after infection. Slides were fixed in methanol and stained with Giemsa to identify the parasites. The number of L. major per 100 macrophages was determined by examination of ϳ200 macrophages per assay.
Monitoring Autophagy-Live promastigotes of wild-type or VPS4expressing L. major expressing GFP-ATG8 were observed daily by fluorescence microscopy, and the proportion of autophagosome-bearing cells as well as the number of these structures per cell were assessed. At least four series of 200 cells were counted per experiment for each point. To inhibit the formation of autophagosomes, promastigotes were grown with either 10 M wortmannin (Sigma, 1 mM stock solution in Me 2 SO) or 10 mM 3-methyladenine (Sigma, 30 mg/ml stock solution in water). To investigate responses to starvation conditions, promastigotes from early log phase cultures were sedimented by centrifugation and washed twice in PBS before being resuspended in pre-warmed PBS and incubated at 25°C for up to 2 h.
Parasite Viability Assay-Viability of L. major during starvation was measured by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT, Sigma) assay as adapted for Leishmania species (33). Briefly, 5 ϫ 10 6 promastigotes were starved in PBS as described above and then incubated for 45 min at 37°C with MTT to a final concentration of 1 mg/ml, and the absorbance was measured at 620 nm using a microtiter plate reader. Control cells were grown in complete medium during the same period of time and similarly assessed with MTT. The results were expressed as a fraction of the values obtained for the starved cells relative to the values of the control cells (in percentages).
Fluorescent Staining of Cells-10 7 L. major promastigotes were harvested by centrifugation and washed once in serum-free HOMEM. For FM4-64 labeling, the cells were incubated with 40 M FM4-64 (from a 12 mM stock solution in Me 2 SO; Invitrogen) for 15 min at 4°C and then washed in fresh medium and incubated for various times at 25°C. Cells were then washed in cold PBS and processed for microscopy. For dextran endocytosis, cells were incubated with Alexa Fluor 594-conjugated dextran at 500 g/ml (M r 10,000, Invitrogen) in complete HOMEM medium for various times at 25°C. Cells were then washed five times in cold PBS and processed for microscopy. For tomato lectin labeling, cells were resuspended in 100 l of cold serum-free HOMEM, and 1 l of fluorescein isothiocyanate-conjugated tomato lectin (1.3 mg/ml solution, Sigma) was added for 10 min at 4°C. Cells were then washed and fixed for fluorescence after a 30-min further incubation in the absence of tomato lectin at 25°C. Cells were fixed with 2% (v/v) formaldehyde in PBS for 30 min at 4°C, resuspended in PBS, and allowed to dry onto glass slides. For concanavalin A lectin labeling, cells were resuspended in 100 l of cold serum-free HOMEM, 1 l of Texas red-conjugated concanavalin A (5 mg/ml solution, Invitrogen) was added, and cells were incubated for 2 h at 25°C. They were then washed in cold PBS and processed as described above for microscopic observation.
Statistical Analyses-Values were expressed as mean Ϯ S.D. Data from macrophage infections, MTT assay, GFP-ATG8 puncta, and motile cell analyses during starvation were pooled (n ϭ 15) for comparison using unpaired t-tests. A p value of Ͻ0.05 was used as the level of significance.

RESULTS
Expression Profile of L. major VPS4 and Generation of a Dominantnegative Mutant-The L. major VPS4 (LmjF29.2500 gene) was identified by searching the L. major-predicted proteins data base (www. genedb.org/) with the yeast Vps4 (P52917) as a query. Sequence comparisons of LmjVPS4 with yeast (Vps4) and mouse (SKD1) orthologues revealed 53.7% and 52.9% amino acid identity, respectively. SKD1/VPS4 proteins belong to the AAA ATPase family (supplementary Figs. S1 and S2). An antiserum raised to a peptide located in the C-terminal region of LmjVPS4 detected the 50-kDa VPS4 protein in lysates of L. major promastigotes from various growth stages ( Fig. 1A, insets, and Fig. 2A, arrowed), as well as a cross-reacting protein of ϳ25 kDa. The only difference in VPS4 levels during in vitro growth was an apparent downregulation of the protein in the non-dividing metacyclic form of the parasite. To investigate the function of the VPS4 protein in L. major, a line overexpressing a LmjVPS4 E235Q mutant was generated. This Leishmania mutant is equivalent to the Vps4 E233Q yeast mutant (37) and the SKD1 E235Q mouse mutant (10), and the amino acid change introduced was expected to cause a defect in ATP hydrolysis (9). L. major promastigotes were transfected with pXG plasmid expressing VPS4 and VPS4 E235Q . The wild-type parasite and cell line containing pXG-VPS4 grew normally as promastigotes in vitro, but the pXG-VPS4 E235Q mutant showed a premature exit from exponential growth phase. Indeed, the promastigotes overexpressing pXG-VPS4 E235Q grew similarly to wild-type parasites throughout most of the logarithmic phase of growth in vitro but went into stationary phase at a lower density than wild-type or pXG-VPS4 promastigotes (Fig. 1B). During this growth phase in culture, an increasing number of promastigotes expressing VPS4 E235Q showed morphological peculiarities and became broader and shorter (for example see GFP-fused VPS4 E235Q in Fig. 2B, right). The use of an episomal vector for the expression of the VPS4 gene allowed us to increase expression levels with higher amounts of selective drug. This way we have shown that the growth defect was even more pronounced when the pXG-VPS4 E235Q mutant was grown in higher concentrations of G418 (Fig. 1A, lower panel), with growth being hindered even in early logarithmic phase, and this correlated with increased amounts of VPS4 E235Q protein at the higher G418 concentrations (Fig. 1A, lower panel inset). In contrast, cells transfected with the empty vector (pXG cell line) could withstand concentrations of G418 up to 50 g/ml with little or no effect on growth (Fig. 1A, upper panel), confirming that neither the vector nor the resistance marker per se were toxic to the cells. Moreover, overexpression of VPS4 was shown not to be toxic to the cells as the pXG-VPS4 cell line grew normally in the presence of increased amounts of G418 and only showed a slight growth defect at 50 g/ml (Fig. 1A, middle panel).
Localization of VPS4 and VPS4 E235Q -L. major promastigotes were transfected with N-terminal GFP fusions of VPS4 and VPS4 E235Q and expression of the proteins confirmed with anti-VPS4 antibody ( Fig. 2A, right panel, arrowhead) and anti-GFP antibody (data not shown). GFP-VPS4 E235Q was apparently expressed at a lower level than GFP-VPS4, but, because the expression of GFP-VPS4 E235Q was detrimental to Leishmania promastigotes (Fig. 1A), it is possible that the cells downregulated the protein to compensate for this effect. Live cells from these populations were observed by fluorescence microscopy (Fig. 2B). GFP-FIGURE 1. The VPS4 E235Q mutant prematurely exits exponential phase of growth. A, effect of the increased overexpression of VPS4 E235Q on growth. Parasites maintained in 12.5 g/ml G418 were resuspended at 5 ϫ 10 5 promastigotes/ml in various concentrations of antibiotic (as indicated above the inset panels), and growth was monitored. Protein extracts were taken at 120 h, and Western blot analyses were performed using anti-VPS4 antibodies (inset), except for VPS4 E235Q cells grown with the two highest concentrations of drug, where the cells were dead. VPS4 is indicated by an arrow, and molecular masses are indicated in kDa. B, cell lines were inoculated at 5 ϫ 10 5 promastigotes/ ml, and cell densities were measured over time. Cell lines expressing pXG-VPS4 or pXG-VPS4 E235Q were maintained with 12.5 g/ml G418. Values shown are the mean Ϯ S.D. of triplicate experiments. major and late-log phase GFP fusion-expressing cell lines. The arrow and arrowhead indicate VPS4 protein and GFP-fused VPS4 proteins, respectively. Molecular mass is shown on the right (in kDa). B, localization of GFP-fused VPS4 (left) and GFP-fused VPS4 E235Q (right) in log phase promastigotes. The GFP fluorescence image (green) has been merged with the equivalent DAPI image (blue; n, nucleus; k, kinetoplast). Differential interference contrast image is shown in the inset panel. Scale bar ϭ 10 m. C, colocalization of GFP-fused VPS4 or GFP-fused VPS4 E235Q with endocytosed tomato lectin (left) and concanavalin A (right). GFP signal is shown in green, and the fluorescent-lectin signal is red; the merged images are magnified. Scale bar ϭ 10 m. Fp, flagellar pocket; L, late endosomes.
VPS4 showed a diffuse pattern throughout the cytosol, a similar localization to the GFP-fused SKD1 of mammals (10,38) and HA-tagged Vps4 in yeast (9). However, the labeling obtained with GFP-VPS4 E235Q was concentrated in structures (Figs. 2B and 3) with shape and size similarities to the class E compartment of yeast, which is composed of exaggerated endosome-like organelles (3). The majority of these GFP-VPS4 E235Q -containing structures in Leishmania promastigotes were present in the anterior part of the cell body, the location of the endocytic pathways in the parasite (16,17).
To further characterize the compartment in which GFP-VPS4 E235Q accumulated in Leishmania, co-localization experiments were performed with markers of endosomal compartments. Tomato lectin, which recognizes poly-N-acetyllactosamine of glycan-containing proteins localized in early endosomes of Trypanosoma brucei (39), was visualized close to GFP-VPS4 E235Q but was clearly distinct (Fig. 2C, left panels). Concanavalin A, which had previously been shown to co-localize with Rab7-endosomes in Leishmania (17), accumulated in the flagellar pocket before being internalized to reach vesicles (presumably Rab7-containing late endosomes) next to the nucleus. When left for 2 h to be internalized, concanavalin A showed a partial co-localization with GFP-VPS4 E235Q -positive compartments present in the anterior part of the cell (Fig. 2C, right panels). This suggests that the compartment in which GFP-VPS4 E235Q accumulates is either part of the endosomal system or closely associated with it. This is fully consistent with VPS4 acting in a late endosomal localization as described for other organisms (9,10,38,40).
Transport of Cargo to the Leishmania MVT-lysosome Is Defective in the VPS4 E235Q Cell Line-The lipophilic dye FM4-64 was used to assess the integrity of transport to the lysosomal compartment in the pXG-VPS4 E235Q promastigote line. This fluorescent endocytic marker is delivered to the vacuolar membrane in yeast (41) and has been used to identify class E compartments (4). Similarly, FM4-64 has been used in Leishmania as a marker for the MVT-lysosome in the stationaryphase promastigote form of the parasite (13). When used in a timecourse experiment, the dye was found initially to be primarily labeling the flagellar pocket and the surface of the parasite (Fig. 3A, 0Ј). After a chase of 15-90 min, FM4-64 in the GFP-VPS4 line was apparent in the MVT-lysosome as well as the flagellar pocket (Fig. 3A, 15Ј-90Ј). In contrast, FM4-64 remained largely in the flagellar pocket in the GFP-VPS4 E235Q line but also after the 90-min chase was observed in compartments associated with GFP-VPS4 E235Q (Fig. 3A, 90Ј, merged  image). Additionally, we used fluorescent dextran as a fluid-phase endocytic tracer, because dextrans are known to be trafficked through the endosomal system to the MVT-lysosome of Leishmania (15). Dextran-red was trafficked normally to the MVT-lysosome in the pXG-VPS4-expressing cell line, whereas it accumulated in the VPS4 E235Q -positive compartments in the mutant cells (Fig. 3B). These results indicate that there is possibly a defect in transport to the MVT-lysosome, or that there could be an alteration of that compartment, in the VPS4 E235Q mutant. Furthermore, the fact that the FM4-64 signal did not co-localize with GFP-VPS4 E235Q until 90-min incubation is in contrast with the lysosome already being labeled after 15 min in the control cell line. This suggests that the kinetics of internal vesicular transport are altered in VPS4 E235Q -expressing parasites.
VPS4 E235Q Mutants Fail to Differentiate into the Infective Metacyclic Form-Metacyclogenesis is crucial for the survival and pathogenesis of Leishmania in the mammalian host (42). Because the pXG-VPS4 E235Q cell line was observed to prematurely enter stationary phase of growth during in vitro cultivation (Fig. 1B), we assessed if metacyclogenesis was affected in this cell line. Several properties distinguish the procyclic and metacyclic promastigote forms of L. major, including morphology (43), agglutination with peanut lectin (28), expression of stage-specific proteins HASPB (35) and SHERP (36), sensitivity to human serum (32), and infectivity to macrophages (28). By all of these criteria, wild-type and pXG-VPS4 cells in late stationary phase of in vitro growth had differentiated into the metacyclic form. In contrast, the VPS4 E235Q cell line failed to undergo metacyclogenesis. pXG-VPS4 E235Q promastigotes did not have metacyclic form morphology (i.e. thin and short cell body with extended flagellum (Fig. 4A)), did not express metacyclic-specific proteins HASPB or SHERP (Fig. 4B), and had greater sensitivity to lysis by human serum relative to the controls (Fig. 4C). Indeed, the only criterion consistent with the pXG-VPS4 E235Q line having undergone metacyclogenesis was that a proportion of the cells were peanut agglutininnegative (Fig. 4D, PNA Ϫ ). The pXG-VPS4 E235Q mutant cell line was found to be less efficient at infecting and surviving within macrophages in comparison with wild-type parasites transfected with the empty vector (pXG line) or the pXG-VPS4 line (Fig. 5A). Importantly, the number of amastigotes inside the infected macrophages after 5 days incubation was found to be greatly reduced compared with the controls, showing that there had been no parasite replication (Fig. 5B). Similar results were found when the experiment was performed using peanut agglutininnegative cells (data not shown), indicating that the presence of metacyclic-specific lipophosphoglycan (28) was not sufficient to confer an infective phenotype and that the pXG-VPS4 E235Q cells had indeed failed to differentiate fully. VPS4 E235Q Mutant Parasites Have a Defect in Autophagy-Endosomal membrane trafficking mediated by Vps4/SKD1 has been proposed to be involved in the late stages of the autophagic pathway in mammals (23) and also in regulating autophagy in yeast (24). Several orthologues of yeast proteins that have been implicated in the Atg8 lipidation pathway involved in autophagosome expansion and completion can be iden- tified in L. major, including Atg3, two putative Atg4 cysteine peptidases, and several families of Atg8 proteins (44). 3 The L. major ATG8 orthologue (LmjF19.1630), with the highest sequence identity to yeast Atg8 proteins was used in this study. ATG8 has been a useful marker for visualizing autophagosomes in yeast and mammalian cells (19,45).
Thus, to track formation of autophagosomes in L. major we expressed ATG8 fused with GFP at its N terminus.
In log phase L. major promastigotes, GFP-ATG8 was distributed throughout the cytoplasm in most cells (Fig. 6A, top). However, in some early stationary phase promastigotes, autophagosomes could be identified as punctate structures clearly observable in the cytoplasm (Fig. 6A,  center). To evaluate the extent of the variation in the occurrence of  . VPS4 E235Q mutants fail to differentiate into metacyclic promastigotes. A, non-agglutinated promastigotes were purified and their morphology observed by differential interference contrast light microscopy. Scale bar ϭ 10 m. B, extracts of 10 7 early log phase and mid stationary phase promastigotes were assessed by Western blot analysis. Elongation factor 1␣ (EF1␣) serves as an internal loading control. Molecular masses (kDa) of the detected proteins are shown on the right. C, complement-mediated lysis assay. 10 7 log phase or peanut agglutinin-negative promastigotes were subjected to lysis by various percentages of fresh human serum for 20 min at 37°C. Intact promastigotes were counted, and results are expressed as the percentage of living cells (as compared with control reactions using heat-inactivated serum). Represented data are the mean from three measurements from two independent experiments Ϯ S.D. *, data are significantly different (p Ͻ 0.01). D, promastigote lines were grown to stationary phase, and the percentage of non-agglutinated cells was determined by the peanut agglutinin method. Values shown are the mean Ϯ S.D. of three independent experiments. autophagosomes during growth phases, the number of autophagosomes was determined during the growth cycle (Fig. 6B). This analysis showed that the number of cells containing autophagosomes, as well as the average number of autophagosomes per positive cell, increased from early log to early stationary phase cultures in wild-type promastigotes (Fig. 6B, top and bottom). As the parasite progressed into stationary phase, concomitant with differentiation into the metacyclic form, the number of autophagosomes per cell and the percentage of cells with autophagosomes rapidly decreased. Concomitantly, a larger proportion of cells were observed with GFP-ATG8 labeling in a tubular compartment, which was identified as the MVT-lysosome by co-localization studies with FM4-64 (Fig. 6A, lower panel). The appearance of GFP-ATG8 positive autophagosomes in L. major promastigotes could be prevented by the addition of 3-methyladenine or wortmannin, inhibitors of autophagosomes formation in yeast and mammalian cells (19), to the culture medium (Fig. 6C).
To investigate the involvement of VPS4 in autophagy in Leishmania, GFP-ATG8 was transfected into pXG-VPS4 and pXG-VPS4 E235Q promastigotes. Log phase pXG-VPS4 promastigotes were similar to wildtype parasites in that, in most cells, GFP-ATG8 was found evenly dis-tributed throughout the cytoplasm (Fig. 7A, left), whereas a few cells contained autophagosomes in addition to the evenly stained cytosol (Fig. 7A, center). Promastigotes expressing pXG-VPS4 E235Q more frequently contained autophagosomes (Fig. 7A, right). Indeed, when analyzed for their production of autophagosomes, the pXG-VPS4 promastigotes showed a profile similar to that of wild type parasites (Fig. 7B, compare with Fig. 6B). In contrast, autophagosomes were more evident in the pXG-VPS4 E235Q mutants even in early-log phase promastigotes, and the number of autophagosomes per positive cell was also found to be higher than in the pXG-VPS4 cell line (Fig. 7B).
Autophagy is induced in yeast and mammalian cells as a survival response under starvation conditions (18,21). We found that the occurrence of autophagosomes in the pXG-VPS4 cells of L. major could be induced by starving early log phase promastigotes in PBS for 2 h (Fig.  7B). In contrast, the pXG-VPS4 E235Q mutants, which initially displayed a higher number of autophagosomes than the pXG-VPS4 cell line, showed no increase in the number of autophagosomes under the same stress conditions (Fig. 7B). ATG8 undergoes several post-translational processing events resulting in conjugation to phosphatidylethanolamine (PE) and recruitment to the autophagosomal membrane (46). We used Western blotting analysis with anti-GFP antibody on cell extracts from pXG-VPS4 and pXG-VPS4 E235Q cells expressing GFP-ATG8 to detect the membrane-associated form GFP-ATG8-PE (Fig.  7C). For the pXG-VPS4 cell line, the proportion of GFP-ATG8-PE increased during progression from log phase to early stationary phase of growth, or when cells were starved for 2 h in PBS. In contrast, the pXG-VPS4 E235Q mutant displayed a high level of the conjugated form both in log and early stationary phase. This is consistent with the data obtained by fluorescence observation of the GFP-ATG8 puncta (autophagosomes) in the same cell lines (Fig. 7B).
During stress by starvation, pXG-VPS4 E235Q promastigotes died significantly quicker than did the pXG-VPS4 promastigotes (Fig. 7D), showing that the mutant is more sensitive to nutrient deprivation than the control cell line. Consistent with this, the presence of autophagosomes steadily increased during starvation of pXG-VPS4 promastigotes, whereas the number of pXG-VPS4 E235Q promastigotes with GFP-ATG8-containing autophagosomes, which was initially higher in the controls, did not increase (Fig. 7E). These data suggest that the pXG-VPS4 E235Q promastigotes have an autophagy defect, in which autophagosomes are not processed, and so accumulate, which makes the cells more sensitive to nutrient deprivation.
An ATG4 Mutant Has a Defect in Autophagy-To further assess if there is a direct involvement of autophagy in the cellular differentiation process, we produced a mutant of autophagy-related gene ATG4.2 (LmjF30.0270), which is one of two ATG4 homologues present in the genome of L. major (44). 3 ATG4 is a cysteine peptidase involved in ATG8 processing and is crucial for autophagosome function (47,48). We produced L. major ATG4.2 null mutant (⌬atg4.2) cell lines by targeted gene disruption (Fig. 8A). Two independent clones (A1, B4) were obtained in which the absence of the ATG4.2 gene and the correct integration of the two replacement cassettes were assessed by Southern blot (Fig. 8A) and PCR (data not shown). Hybridization of NdeI/BglIIdigested genomic DNA with a 5Ј-flank probe revealed a 7.3-kb product corresponding to the locus of the wild-type parasites, which was absent in the ⌬atg4.2 cell lines (Fig. 8A). In these, a DNA fragment of ϳ8.5 kb was visible showing correct integration of the pGL345ATG4.2HYG and pGL842ATG4.2BSD cassettes (co-migrating 8.9-and 8.3-kb DNA fragments, respectively). All subsequent phenotype analyses were performed with both clones, but as they behaved similarly, data are shown for clone B4 only. The ability of L. major ⌬atg4.2 null mutants to form autophagosomes was assessed by transfecting them so that they expressed the GFP-ATG8 marker. The proportion of ⌬atg4.2 promastigotes with GFP-ATG8-containing autophagosomes and the number of autophagosomes in each cell was higher than observed for wild-type cells at a similar stage of growth ( Fig. 8B and data not shown). In addition, the number of autophagosomes did not decrease during stationary phase of growth, as observed with wild-type cells. Western blot analysis using anti-GFP antibody on cell extracts of early stationary phase wild-type and ⌬atg4.2 promastigotes expressing GFP-ATG8 showed that the mutant cell line had an increased proportion of the lipidated form of GFP-ATG8 relative to the cleaved form (Fig. 8B). This result is consistent with an accumulation of the membrane-bound conjugated form of GFP-ATG8 in ⌬atg4.2 promastigotes.

The ATG4 Mutant Fails to Differentiate into the Infective Metacyclic
Form-We assessed the ability of ⌬atg4.2 promastigotes to differentiate into metacyclic forms using the same criteria as pXG-VPS4 E235Q cells. Stationary phase ⌬atg4.2 were found to be defective in expression of stage-specific metacyclic proteins HASPB and SHERP (Fig. 9A) and were more sensitive to lysis by human serum (Fig. 9B). One difference between the pXG-VPS4 E235Q cells and ⌬atg4.2 was their agglutination by peanut lectin (Fig. 9C). Whereas ϳ30% of wild-type and pXG-VPS4 E235Q cells were agglutinated by peanut lectin, only 5-10% of ⌬atg4.2 promastigotes were agglutinated, showing some phenotypic variance between the two different mutant cell lines. All these phenotypes were similar to wild type in the ⌬atg4.2[pN-ATG4.2] mutant (Fig.  9). These data show that ⌬atg4.2 promastigotes are defective in metacyclogenesis.

DISCUSSION
In this study we have investigated the properties of L. major MVB by creating dominant-negative mutants of VPS4, which disrupt MVB sorting function. In yeast, Vps4 is involved in the disassembly of the ESCRT III complex from the endosomal membrane, and the ATP hydrolysis mutant cannot dissociate from the late endosomal membrane (9). In L. major, GFP-VPS4 E235Q was localized to a vesicular compartment, which partially co-localized with the late endosomal marker concanavalin A at the anterior end of the cell, between the flagellar pocket and the nucleus, whereas GFP-VPS4 was found throughout the cytosol (Fig. 2B). This is consistent with VPS4 having a similar role in L. major as in yeast and mammalian cells, and this is further supported by the finding of genes encoding other components of the ESCRT III complex in the FIGURE 9. Autophagy mutant ⌬atg4. 2 fails to differentiate into metacyclic promastigotes. A, extracts of 10 7 mid stationary phase promastigotes were assessed by Western blot analysis. EF1␣ serves as an internal loading control. Molecular masses (kDa) of the detected proteins are shown on the right. B, complement-mediated lysis assay. 10 7 log phase or mid stationary phase cells were subjected to lysis by various percentages of fresh human serum for 20 min at 37°C. The remaining intact promastigotes were counted, and results are expressed as the percentage of living cells (as compared with control reactions using heat-inactivated serum). Data are the means Ϯ S.D. from three measurements from three independent experiments. *, data are significantly different from wild type (p Ͻ 0.01). C, promastigote lines were grown to early stationary phase and the percentage of nonagglutinated cells was determined by the peanut lectin method. Values shown are the means Ϯ S.D. from three independent experiments. *, data for Datg4.2 cell line are significantly different from wild type (p Ͻ 0.01). genome of L. major (supplementary Table SI). It has been proposed for the mammalian SKD1 mutants that these abnormal endosomes might represent a hybrid organelle between late endosomes and lysosomes (49). Our data are consistent with a close association of VPS4 E235Q with endosomal structures. Additional experiments with FM4-64 and the fluid phase marker dextran-red confirmed that in the VPS4 E235Q mutant endocytosed cargo could not reach the MVT-lysosome. However, the cellular distribution and expression levels of glycosylphosphatidylinositol-anchored surface peptidase GP63 remained unchanged in the mutant (data not shown), showing that there were no major defects in exocytosis in the mutants. The impairment in lysosomal transport is similar to the situation observed for yeast and mammalian Vps4/SKD1 mutants (9,49). This impairment in L. major, however, caused a severe growth phenotype resulting in cells that lacked many established markers of metacyclic promastigotes, showing that they had a defect in metacyclogenesis. The one exception was that the VPS4 E235Q cells expressed metacyclic-specific lipophosphoglycan on their cell surface. Lipophosphoglycan assembly is complex and necessitates biosynthetic steps in the Golgi, before being transported to the surface (50), so it is possible that this biosynthetic pathway is induced independently of metacyclicspecific proteins such as SHERP and HASP and earlier in the process of metacyclogenesis. One of the consequences of this inability to fully differentiate into the metacyclic form was reduced infectivity and survival of the mutant parasites in macrophages.
In yeast and mammalian cells, late endosomal membrane trafficking mediated by Vps4/SKD1 have been shown to be important for the function of the autophagic pathway (23,24). In Leishmania, autophagy, monitored by the presence of GFP-ATG8-positive autophagosomes, was most active during the differentiation of Leishmania procyclic promastigotes to the infective metacyclic promastigote form (Fig. 6). It also could be induced in procyclic promastigotes by starving the cells (Fig. 7,  B and D), a stimulus that induces autophagy in yeast and mammals. Moreover, the appearance of the structures could be prevented by the use of known autophagy inhibitors wortmannin and 3-methyladenine (19), thus further confirming their autophagosomal nature.
In VPS4 E235Q cells, the number of GFP-ATG8-positive cells and autophagosomal structures within them was higher than in the VPS4 control, both in log phase and stationary phase cells. However, they did not increase over a certain limit, suggesting that there is some feedback signal that prevents further production of autophagosomes. Furthermore, the lipidated form of ATG8 (ATG8-PE) was found to be consistently present at higher levels in VPS4 E235Q cells than in wild-type cells. This suggests that the VPS4 E235Q cells accumulated autophagosomes, due to a defect in the ability of the autophagosomes to fuse with the endosomal/MVT-lysosomal compartment. In wild-type parasites, the GFP-ATG8 signal localized to the MVT-lysosome in metacyclic parasites (Fig. 6A). We propose that the differentiation between dividing procyclic promastigotes and the infective metacyclic form, which differ by many aspects of cell shape, metabolism, and surface proteins, is dependent on endosome function and autophagy. The functioning of the autophagic pathway in mammalian cells requires a functional MVB compartment as illustrated by the defect in autophagy in the Vps4 mutant (23). However, the extent of the involvement of the endosomal compartment in the autophagic pathway in yeast is less clear (20, 51).
We have now clearly shown that in Leishmania, the VPS4 E235Q cells had a defect in autophagy, and we observed that VPS4-positive structures were located close to autophagosomal structures in the cell (data not shown), suggesting that in this lower eukaryote the autophagic pathway could require interaction with the endosomal network in order to function.
More direct evidence for the involvement of autophagy in differentiation of the parasite was provided by the phenotype of the ATG4.2 null mutants. ATG4 is essential for autophagy in both yeast and mammals (47,48); it is a cysteine peptidase that is involved in the regulation of the autophagosomal membrane association of ATG8 by the conjugation/deconjugation of the protein to PE. In yeast, where there is only one isoform of ATG4, the ATG4 mutant loses the ability to both conjugate and deconjugate ATG8, leading to a defect in autophagy (46). There are two ATG4 isoforms in mammalian cells, but their respective substrates and specific roles have not yet been clearly elucidated (52,53). We report here that the null mutant of one of the two L. major ATG4 genes (⌬atg4.2) accumulates autophagosomes and has a defect in autophagy, as assessed by the reduced ability of the mutants to withstand starvation. It leads us to speculate that ATG4.2 might be involved in ATG8 deconjugation prior to fusion with late endosomes/lysosomes, a process that is necessary for efficient progression of autophagy in yeast (46). The ⌬atg4.2 mutant is also impaired in its ability to differentiate into the metacyclic form, thus there is a clear link between the autophagic machinery and the differentiation process. It is notable that for most markers of metacyclogenesis VPS4 E235Q and ⌬atg4.2 were similar, however, the latter did not express metacyclic-specific lipophosphoglycan on its cell surface. This could highlight a different role between endosome sorting and autophagy in the completion of differentiation.
The VPS4 E235Q cells were more sensitive to starvation than the control promastigotes and, unlike the control cells, did not increase the number of autophagosomes in this condition (Fig. 7). It is interesting to note that previous work done on csc1/Vps4 in Saccharomyces cerevisiae had shown that starvation-induced autophagy was impaired in a null mutant of this gene (24). Thus, VPS4 mutants do not seem to be affected in their ability to initiate the formation of ATG8-bearing autophagosomes but rather accumulate these structures.
The increased sensitivity of the VPS4 E235Q cells to starvation could be a combination of several defects. Firstly, the VPS4 E235Q line appeared defective in the transport of endocytosed cargo to the MVT-lysosome, and thus its ability to acquire and transfer nutrients from the medium would be compromised. However, the MVT-lysosome itself does not appear to be significantly altered, because the CPB lysosomal cysteine peptidase is correctly trafficked, processed, and activated in the MVTlysosome in VPS4 E235Q cells (data not shown). Secondly, defects in the final stages of the autophagy pathway in VPS4 E235Q cells make the parasites more sensitive to starvation conditions due to a reduced ability of the parasites to recycle, as reserve nutrients, cytoplasmic or organellar material. This latter defect is likely to be important, because similar results were found when the ATG4.2 mutant was put in similar conditions (Fig. 8), directly highlighting the role of autophagy in the survival of Leishmania during starvation.
Starvation conditions and high population density are strong inducers of autophagy in both yeast and mammalian cells (18,21). In those conditions, autophagy has been proposed to be associated with stress-induced cell differentiation (18), where it could help remodel cell shape. This notion was until recently not supported by consistent experimental evidence, but a few examples of such a role of autophagic processes in differentiation have now emerged (25)(26)(27). Leishmania promastigotes reach an in vitro stationary phase of growth where their cell density is at a peak, and it is at this stage that they differentiate from dividing promastigote to the infective metacyclic form. We observed an increased number of autophagosomal structures in correlation with the start of differentiation process (Fig. 6). Thus, autophagy normally occurs during differentiation and, as this study has shown, apparently is necessary for it.
While most autophagy studies made so far have focused on the role of autophagy during stress-induced cellular remodeling, we have shown for the first time that autophagy has an important role in the differen-tiation and acquisition of virulence of a protozoan parasite. Proposed roles for autophagy extend from nutrient acquisition and cellular remodeling to growth control, but its role in differentiation is essentially unexplored. Full characterization of the molecular mechanisms of autophagy, which are largely conserved from unicellular eukaryotes to metazoa, but probably differ in detail through evolutionary adaptation, can greatly benefit from the study of Leishmania, which has distinct morphological forms and can be genetically manipulated.