Dictyostelium macroautophagy mutants vary in the severity of their developmental defects

atg6 because these two genes are the only representatives of the cognate that have been identified in D. discoideum . We chose atg8 because it is the only structural component identified in the Atg8 lipidation system. We show that atg1 , atg6 and atg8 genes are also required for macroautophagy in Dictyostelium . The three mutations produce defects in survival, recycling of cytoplasm and organelles during nitrogen starvation, and protein degradation during development. Additionally, development is aberrant in all three mutants. Mutation of atg1 produces the most severe macroautophagy and developmental defects, whereas mutations in atg6 and atg8 permit a further progression in development.


SUMMARY
Macroautophagy is the major mechanism that eukaryotes use to recycle cellular components during stressful conditions. We have previously shown that the Atg12-Atg5 conjugation system, required for autophagosome formation in yeast, is necessary for Dictyostelium development. A second conjugation reaction, Aut7/Atg8 lipidation with phosphatidylethanolamine, as well as a protein kinase complex and a phosphatidylinositol 3'-kinase complex, are also required for macroautophagy in yeast.
In this study, we characterize mutations in the putative Dictyostelium discoideum orthologues of budding yeast genes that are involved in one of each of these functions, ATG1, ATG6 and ATG8. All three genes are required for macroautophagy in Dictyostelium. Mutant amoebae display reduced survival during nitrogen starvation and reduced protein degradation during development. Mutations in the three genes produce aberrant development with defects of varying severity. As with other Dictyostelium macroautophagy mutants, development of atg1-1, atg6and atg8is more aberrant in plaques on bacterial lawns than on nitrocellulose filters. The most severe defect is observed in the atg1-1 mutant, which does not aggregate on bacterial lawns and arrests as loose mounds on nitrocellulose filters. The atg6and atg8mutants display almost normal development on nitrocellulose filters, producing multi-tipped aggregates that mature into small fruiting bodies. The distribution of a green fluorescent protein fusion of the autophagosome marker, Atg8, is aberrant in both atg1-1 and atg6mutants.

INTRODUCTION
In the social amoeba D. discoideum, starvation is a signal for the initiation of multicellular development. Starving amoebae aggregate in response to cAMP to form mounds. Within these cell aggregates, intercellular signals direct the formation of a multicellular slug that migrates to a suitable location for formation of a fruiting body.
The fruiting body is composed of a spore mass held aloft on a stalk composed of cells that vacuolate and die. Development is an energy-intensive process, and requires that amoebae cease production of growth-related proteins and lipids, and initiate a developmental program (reviewed in (1)). One mechanism employed by Dictyostelium and other eukaryotes to mobilize resources required for development is macroautophagy.
Macroautophagy is required for sporulation in Saccharomyces cerevisiae (2), differentiation in the yeast Podospora anserina (3), metamorphosis in Drosophila melanogaster (4), and dauer development in Caenorhabditis elegans (5). In this transport process, bulk cytoplasm and organelles are sequestered in double-membrane vesicles (autophagosomes/autophagic vacuoles) that fuse with and deliver their content to the lytic compartment of the cell, the lysosome or vacuole. Genetic studies in S. cerevisiae have identified 15 ATG genes that are required for the formation of these double membrane autophagosomes (2,6,7). A new unified nomenclature for autophagy-related genes was introduced recently (8), and we will use the atg designation henceforth. The genes can be grouped into several classes that function in the localization of autophagy components to the site(s) of autophagosome biogenesis, termed the preautophagosomal structure (PAS) in yeast, and in the generation of double-membrane vesicles (reviewed in (9)). Three signaling complexes are required for macroautophagy: a Tor-kinase complex, an Atg1phosphoprotein kinase complex, and a phosphatidylinositol-3' kinase (PI3K) complex.
Tor is a phosphotidylinositol kinase-related Ser/Thr kinase that functions in the control of protein metabolism in response to nutritional status (reviewed in (10)). In addition to increasing protein synthesis by multiple mechanisms, Tor also functions in macroautophagy by signaling to Atg1, a Ser/Thr kinase that forms a signaling complex with Atg13 and Atg17 (11,12). Tor influences the phosphorylation status of Atg13, which in turn alters the affinity of the Atg13-Atg1 association. During starvation, Tor is inactive, Atg13 is hypophosphorylated and binds tightly to and activates Atg1 kinase.
However, enhanced Atg1 kinase activity during starvation is apparently dispensable for induction of macroautophagy, suggesting that Atg1 function in autophagy may be structural (13).
Macroautophagy in yeast also requires a protein complex containing a class III phosphatidylinositol 3'-kinase (PI3K), Vps34, a membrane bound Ser/Thr kinase, Vps15, and a coiled-coil protein Vps30/Atg6 (14,15). This core complex functions both in macroautophagy and vacuolar protein sorting. Pathway specific components are required for the localization or targeting of the PI3K activity: Vps38 is sorting-specific, whereas the small coiled-coil protein Atg14 functions in macroautophagy (15), and is thought to target the PI3K complex to the PAS in yeast (16,17). The human homologue of ATG6, beclin 1, encodes a Bcl-2-interacting protein that plays a role in negatively regulating mammalian cell growth and tumorigenesis (18,19). Beclin 1 binds to the human Vps34 homologue, PtdIns 3-kinase, and both proteins localize to the trans-Golgi network, suggesting that the macroautophagy function of Beclin 1 may be related to protein sorting (20).
Two novel post-translational modifications are essential for efficient autophagosome formation. Atg12 is conjugated to Atg5 in an ubiquitination-like reaction requiring the E1-like enzyme Atg7 (21,22) and the E2-like enzyme Atg10 (23). In yeast, the Atg12-Atg5 conjugate binds a coiled-coil protein, Atg16, that mediates formation of a 350-kDa oligomeric complex (24), and localization of the complex to the PAS (17). In mammals and other eukaryotes (including Dictyostelium), the Atg16 orthologue, Atg16like protein (Atg16L), contains both a coiled-coil motif and WD repeats (25). Mouse Atg16L functions similarly to Atg16 in yeast, forming 800-kDa oligomers with the Atg12-Atg5 conjugate, and targeting the complex to the forming autophagosome, the cup-shaped isolation membrane.
Finally, a membrane protein complex composed of the binding partners Atg9 and Atg2 (31) is required for autophagosome formation. These two proteins localize to the PAS in yeast, but otherwise their function is poorly understood.
Suzuki and co-workers examined the localization of GFP-Atg8 and Atg5-GFP in different autophagy mutants in an attempt to order the function of the groups of autophagy proteins (17). The Atg6-containing PI3K complex and Atg9 are required for localization of Atg5 to the PAS. The Atg5-Atg12 conjugation system is required for recruitment of Atg8-PE to the PAS, while the Atg1-containing phosphoprotein kinase complex and Atg2 function after the recruitment of Atg8-PE to the PAS. Two distinct autophagy protein complexes exist at the PAS, one containing the Atg1 complex and Atg9, the other containing the conjugation proteins and Aut7/Atg8 (16).
D. discoideum provides a simple yet powerful model for studying the role of macroautophagy in multicellular development. We have used Dictyostelium to study the requirement for macroautophagy in the establishment of a replicative vacuole for the intracellular bacterial pathogen, Legionella pneumophila (32). In addition, mutations in two Dictyostelium macroautophagy genes representative of the Atg12 conjugation system, atg5 and atg7, produce severe developmental defects (33). The mutants produce aberrant fruiting bodies that are devoid of mature, detergent-resistant spores. Both mutations produce macroautophagy defects and developmental phenotypes of similar severity. We asked whether mutations in macroautophagy genes representative of the other functional classes produce an arrest at a similar stage in multicellular development.
We generated mutations in representatives of the phosphoprotein kinase complex (atg1), the PI3K-signaling complex (atg6), and the lipidation system (atg8). We chose atg1 and atg6 because these two genes are the only representatives of the cognate complexes that have been identified in D. discoideum. We chose atg8 because it is the only structural component identified in the Atg8 lipidation system. We show that atg1, atg6 and atg8 genes are also required for macroautophagy in Dictyostelium. The three mutations produce defects in survival, recycling of cytoplasm and organelles during nitrogen starvation, and protein degradation during development. Additionally, development is aberrant in all three mutants. Mutation of atg1 produces the most severe macroautophagy and developmental defects, whereas mutations in atg6 and atg8 permit a further progression in development.

Strains
All mutations were created in the strain DH1, which is a uracil auxotroph. Strains were grown in HL5 medium or on lawns of Klebsiella pneumoniae (34).

Development and spore production assays
Multicellular development was examined on nitrocellulose (NC) filters, on non-nutrient agar, or in plaques on bacterial lawns. For NC filters, axenically grown cells in mid-log phase (2-4 x 10 6 cells/ml) were washed twice in cold Sorensen C (SorC) buffer (16.7mM Na 2 H/KH 2 PO 4 , 50µM CaCl 2 , pH 6.0). The cells were resuspended in SorC buffer and plated on 25mm, 0.45µM nitrocellulose filters (Millipore Corp.), which rested on SorCsoaked Whatman Grade 17 filter pads, at a density of ~3.3x10 6 cells/cm 2 , or 1.6 x 10 7 cells/filter (34). For development on non-nutrient agar, cells were resuspended at 10 7 cells/ml, and 100 µl of these cells were plated on 35-mm SorC-1% Phytagel (Sigma) agar plates and allowed to dry. For development on bacterial lawns, 25-50 amoebae were mixed with 150 ul of an overnight culture of K. pneumoniae, and plated on SM plates.
Phenotype was examined after 5 days incubation at 22 °C. For spore production assays, the same protocol was followed as for development on NC filters. Filters were transferred to 5 ml SorC in 15 ml Falcon tubes 28-32 hours after initiating development, and all developing structures and cells dislodged by vigorous vortexing. The harvested structures were incubated for 5 minutes in 0.3% Triton X-100 to kill amoebae but leave mature spores intact. Spores were counted with a haemacytometer and appropriate dilutions were plated on K. pneumoniae lawns to determine viability. Transformants were selected with blasticidin (5 µg/ml) in HL5 medium for 1 week, harvested from Petri dishes, plated onto SM plates, and clones of mutant or wild-type phenotype were selected for further analysis. Homologous recombination of the targeting construct with the endogenous locus was confirmed by PCR or by Southern blot.

Fusion constructs
The GFP-Atg8 fusion was constructed as previously described (33). Briefly, the full-length coding sequence of atg8 was obtained by PCR from a genomic DNA template. The PCR product was cloned into pGEM-T Easy (Promega), cut from pGEM- SacI for cloning into SacI-digested, dephosphorylated pDXA-HC-CFP. Amoebae were transfected by electroporation (35), and transformants were selected with G418 (5 µg/ml) in HL5 medium for 1 week. Transformants were harvested from Petri dishes, plated onto SM + G418 (60 µg/ml) plates, and individual clones were selected for further analysis.
We confirmed that a protein of the correct size was produced in transformed strains by western blotting with rabbit polyclonal antiserum against GFP (Molecular Probes).

Northern blot analysis
Northern blot analysis was conducted as previously described (37). Cells were deposited on nitrocellulose filters for development, and a filter was harvested every 4 hours for RNA extraction. Five µg total RNA was glyoxylated, size fractionated on 1% agarose gels, transferred to nitrocellulose, and hybridized with random primer-labeled DNA probes. Probes were obtained by isolation of appropriate restriction fragments following separation on low-melting temperature agarose gels. The atg1 probe consisted of an ~700 bp 5' BglII-SacI fragment of the 1-1/1-4 PCR product in pGEM-T Easy. The atg8 probe consisted of the ~800 bp 3' ClaI-EcoRI fragment of the 8-5/8-4 PCR product in pGEM-T Easy.

Protein turnover assays
Protein turnover assays were based on those of White and Sussman (38).
Developing cells were recovered from Millipore filters in SorC buffer, pelleted by centrifugation, and resuspended in 50 mM Tris-HCl (pH 6.9) containing protease inhibitors (Complete, Mini-Roche). Cells were lysed by freeze-thaw before determining protein levels using the Pierce Coomassie® Plus Protein Assay Reagent, following the manufacturer's instructions.

Electron microscopy
Electron microscopy procedures were as described (39). Briefly, cells were fixed with 2% phosphate-buffered glutaraldehyde (pH 7.2), followed by 0.5-2% osmium tetroxide fixation in the same buffer, sedimented by centrifugation and enrobed in agar, dehydrated with an aqueous/acetone series, and embedded in TAAB epon resin (Energy Beam Sciences, Agawam, MA).

Fluorescence microscopy
To examine the localization of GFP fusion proteins, axenic cells were incubated overnight in HL5 medium supplemented with 10 µg/ml G418 in 35 mm glass bottom microwell dishes (MatTEK Corporation). The following day, the medium was replaced with SorC 1-6 hours prior to visualization (starving cells) or visualized directly (growing cells). GFP-expressing cells were viewed under a Nikon Eclipse TE300 microscope, images were captured with a cooled CCD camera, and processed using Metamorph 5 imaging software (Universal Imaging Corporation).
Chimaeric development was conducted as previously described (33).

Structural features of atg1, atg6 and atg8 genes in Dictyostelium discoideum
We previously identified putative Dictyostelium orthologues for Atg1 (AY191011), Atg6 (AY191013), and Atg8 (AY191015) (33). In addition, a putative open reading frame with homology to Atg6, which we call atg6B, was recently identified on chromosome 5 by the Dictyostelium Genome Sequencing Project at Baylor College of Atg1 is a serine-threonine kinase initially identified in budding yeast (40). Atg1p is composed of an N-terminal kinase domain (see narrow arrowheads in Figure 1A) and a C-terminal domain containing no recognizable motifs. Protein kinases contain twelve conserved subdomains that mediate ATP-binding, binding of the substrate, and phosphotransfer (41). All twelve subdomains are present in the Atg1 orthologues of S. cerevisiae, Arabidopsis thaliana, and D. discoideum. The putative Atg1 orthologues from mammalian cells (42,43) and Arabidopsis (44) have not been shown experimentally to function in autophagy, whereas the C. elegans Atg1 orthologue, unc-51, does function in autophagy (5). The homology between the Atg1 orthologues is highest in the Nterminal kinase domain, and lower in the poorly-conserved C-terminus. The predicted 668 residue Dictyostelium Atg1 has an asparagine-rich stretch immediately following the kinase domain. Poly-asparagine repeats are common in Dictyostelium proteins, but their function is unknown. Additionally, a large 29 residue glutamine-rich stretch is present in the C-terminal half of the protein. The Dictyostelium Atg1 kinase domain shares highest homology with its Arabidopsis counterpart (44% identity over 268 residues). The homology of the C-terminal domain is also highest with the Arabidopsis protein, (19% over 331 residues). The Dictyostelium protein lacks the predicted coiled-coil motif present at the C-terminus of S. cerevisiae Atg1 (13), as do the C. elegans and A. thaliana Atg1 orthologues when analyzed using the COILS server (http://www.ch.embnet.org/software/COILS_form.html (45)).
The predicted 122 residue Dictyostelium Atg8 orthologue shares a remarkably high degree of sequence identity to Atg8 proteins from Arabidopsis (72%) and budding yeast (63%), whereas identity to the human MAPLC3 is lower (35%). The glycine residue required for conjugation to PE is conserved, and corresponds to G119 in Dictyostelium Atg8, G117 in AtAtg8a, G120 in HsAtg8/MAP1LC3, and G116 in ScAut7/Atg8 (asterisk in Figure 1C). Presumably the 3 amino acids following G119 must be cleaved by the putative Dictyostelium Atg4 orthologues previously identified (33), a process that is conserved in other organisms.

Generation of mutations in atg1, atg6 and atg8
We generated mutations in atg1, atg6, and atg8 by insertional mutagenesis suggest that atg1-1 is a strong loss-of-function allele.
To generate an atg6 mutant, the blasticidin cassette was inserted into a BamHI site at position 435 in the 4104 bp atg6 open reading frame, leaving the possibility that a truncated 145 amino acid residue protein is produced. We have confirmed by PCR that the atg6 mutant phenotype is a result of homologous recombination (data not shown).
We could not detect an atg6 transcript by northern blot analysis or RT-PCR in the parental strain DH1, and so could not confirm that a truncated mRNA is present in the atg6 mutant. However, this insertion most likely creates a null mutation, for the following reasons: the phenotype of the atg6 mutant cells with this insertion is consistent with that of a macroautophagy mutant. Additionally, the insertion is at the 5' end of the gene, upstream of the presumed functional part of the protein, based upon homology of the Dictyostelium protein with other orthologues (see Figure 1B). Therefore, we refer to the resultant mutant strain as atg6 -.
To create the atg8 mutant, we inserted the blasticidin cassette into a HindIII site 15 base pairs into the atg8 coding sequence. We cannot detect an atg8 transcript in the atg8 mutant strain with a probe to the 3' 318 bp of the atg8 gene, whereas a developmentally regulated ~1.1-1.2 kb transcript (not 0.9 kb as previously described (33)) is detected in the parent DH1 (Figure 2, lower panel). The insertion creates a null mutation, and thus we refer to the atg8 mutant as atg8 -.

Autophagy mutants show developmental defects of varying severity
We tested the growth of autophagy mutants and the parental strain in both shaking axenic culture (by haemacytometer counts) and on bacterial lawns (by plaque size). The atg1-1 mutant grows slightly slower than the parental strain under both conditions (data not shown). The mutations in atg6 and atg8 produce no discernable growth defects. All three mutants show defective development when developing within plaques on bacterial lawns, and when developed on nitrocellulose filters (see Figure 3). Dictyostelium amoebae form plaques on lawns of K. pneumoniae, where the amoebae at the edge actively engulf bacteria, while those at the center starve and initiate development ( Figure   3D). The atg1-1 mutant does not aggregate when developing in plaques on K.
pneumoniae lawns, but forms loose mounds on nitrocellulose filters ( Figure 3B, E). The atg6and atg8mutants are the least severe of the autophagy mutants we have characterized thus far. Development is almost normal on nitrocellulose filters, where multi-tipped aggregates are formed that mature into small but otherwise normal fruiting bodies for atg8 - (Figure 3H), or small fruiting bodies with truncated stalks for atg6 -( Figure 3G). The developmental defect is more severe on bacterial lawns, where both strains aggregate mainly at the center of plaques, and not at the periphery as the plaque grows, and the multi-tipped aggregates that form produce very small fruiting bodies ( Figure 3J, K).
The atg1-1 mutant phenotype is significantly rescued on both nitrocellulose filters and bacterial lawns by complementation with CFP-Atg1 ( Figure 3C, F). The effects of complementing the atg8mutation by expressing GFP-Atg8 are not as obvious because the atg8phenotypic defect is relatively mild ( Figure 3I, L). We could not maintain the full-length atg6 gene stably in bacteria, and so could not perform complementation experiments.
Since the atg6and atg8mutants have the mildest developmental defect of the macroautophagy mutants we have examined to date, we determined whether the defective fruiting bodies produced by these mutants contained mature, detergent-resistant, viable spores. The atg1-1 mutant produces no spores, consistent with the early block in development prior to cell-type differentiation and morphogenesis (Table 1).
Complementation with CFP-Atg1 rescues spore production to ~10-25% of wild-type levels, consistent with the significant phenotypic improvement. The atg6and atg8mutants produce some detergent-resistant spores, although significantly fewer than wildtype. The majority of spores formed by atg8are round (83-92%), and not the elliptical shape of mature DH1 spores. The atg6mutant produces fewer spores than atg8 -, but most are elliptical. Expression of GFP-Atg8 in atg8yields an increase in spore production, and a greater proportion of the spores that are produced are elliptical (27-38%).

Autophagy mutations are cell autonomous
The atg6and atg8strains will produce spores when developed on nitrocellulose filters, but they produce fewer spores than wildtype, and many of the atg8spores appear to be immature. To test whether development in the presence of wild-type cells may rescue the spore production or maturation defect of these two mutants, we performed mosaic development experiments. Mutant and wild-type cells were mixed in a ratio of 1:3, and placed on nitrocellulose filters or non-nutrient agar for development. When GFPexpressing mutant cells mixed with unmarked parental DH1 are developed on nonnutrient agar, both atg6and atg8amoebae, but not atg1-1, are observed in the sorocarps of chimaeric fruiting bodies by fluorescence microscopy (Figure 4). The atg1-1 mutant cells, like atg5and atg7amoebae, are found in the basal disc, a structure that supports the fruiting body ( Figure 4D). We developed mixed populations on nitrocellulose filters and examined detergent-resistant spores produced by chimaeric fruiting bodies. atg6and atg8amoebae produce plaques that are clearly distinguishable from parental plaques (see Figure 3). We observed no plaques with the atg8phenotype of 223 examined. If sorocarps were harvested with a sterile metal loop into SorC, and the spores plated with bacteria on SM plates without detergent treatment, then 3% (17/587) of the resulting plaques had the atg8phenotype. With detergent treatment, chimaeras of atg6and DH1 yielded 6/217 (3%) plaques with the atg6phenotype, and this value did not change significantly if spores were plated without detergent treatment. Since the mutant input is 25%, it appears that the two mutants are predominantly excluded from the fruiting structures during mosaic development. In conclusion, all three mutations are cell autonomous, because production of mature, viable spores by mutant cells cannot be rescued by mixing with wild-type cells.

Autophagy mutants are hypersensitive to amino acid starvation
A hallmark of macroautophagy mutants in budding yeast (2) and in Dictyostelium (33) is an inability to survive amino acid starvation. We tested whether the mutations we created result in the reduction in viability expected of macroautophagy mutants, and tested atg5in the same experiment as a representative of the Atg conjugation system. We starved amoebae in amino acid-free FM medium, and plated aliquots with K. pneumoniae on SM/5 plates every 2 days to test for viability, as previously described (33). The parental strain DH1 survives well over the 8-day duration of the experiment ( Figure 5), whereas all three mutants show severely reduced viability after only 4 days.
The severity of the viability defect correlates with that of the developmental defect: the atg1-1 mutant is most severe, whereas atg6and atg8show a less significant reduction in viability, similar to that of the atg5mutant.

Bulk protein turnover is reduced in macroautophagy mutants
Macroautophagy is the major mechanism of bulk protein turnover in starving yeast cells (46), and Dictyostelium macroautophagy mutants degrade significantly less protein than wild-type cells during development (33). Therefore, we tested whether atg1-1, atg6and atg8had a similar defect in protein degradation. Total protein measured by the Bradford assay drops by 30-45% in wild-type cells developed for 24 hours on nitrocellulose filters ( Table 2). The three macroautophagy mutants contain only 5-15% less total protein than at the start of development, and the severity of the defect correlates with that of the developmental phenotype. The complemented atg1-1 strain, atg1-1 (act15/cfp-atg1), displays protein degradation comparable to the parent DH1, whereas rescue in atg8 -(act15/gfp-atg8) is poorer.

Transmission electron microscopy of amino acid-starved amoebae
A diagnostic feature of amino acid-starved Dictyostelium macroautophagy mutants is an absence of the significant cytoplasmic degradation observed in starved wild-type cells by transmission electron microscopy (TEM). We undertook TEM studies to confirm that our new mutations produce the ultrastructural changes expected for macroautophagy mutants. Growing mutant amoebae are indistinguishable from their growing wild-type counterparts (data not shown). We also examined amoebae starved of amino acids for 36 hours. As we have observed for the atg5and atg7mutants (33), atg1-1, atg6and atg8mutants all show little evidence of turnover of cytoplasmic constituents compared to wild-type DH1 cells ( Figure 6). Again, as we have observed for other measures of autophagic efficiency, the ultrastructural phenotype of atg1-1 is more severe than the atg6and atg8mutants. The cytoplasm of atg1-1 amoebae is invariably more dense than that of atg6or atg8amoebae (compare Figure 6C to B and E). The starved atg1-1 amoebae commonly show profiles of mitochondria encircled by rough endoplasmic reticulum (arrowheads in Figure 6C; see also Figure 7). We do not observe any structures that could be definitively identified as autophagosomes in any of the mutants examined in this study.
We commonly observe profiles of large vesicle clusters in the atg1-1 mutant that are absent in the parent DH1 (Figure 7). The clusters consistently contain three vesicle classes: larger vesicles with empty lumina (narrow arrow in Figure 7B), vesicles of intermediate size with electron-dense lumina (wide arrow in Figure 7B), and very small vesicles that are sometimes linearly arrayed (arrowheads in Figure 7B). We have observed similar profiles in atg5and atg7mutants, although the clusters are smaller than in atg1-1 (data not shown). The reduced size of these clusters in atg5and atg7may explain why we observe them less frequently than in atg1-1.
We also examined whether expression of GFP fusions could rescue the mutant phenotype of atg1-1 and atg8mutants. In both cases, we observed a gradation in the degree of complementation. A minority of starved cells are similar in appearance to mutant cells, and have dense cytoplasm and many organelles. A majority of cells had the phenotype expected for a complemented mutant, and were similar to starved wild-type cells in that the cytoplasm contained large electron-lucent regions and few organelles ( Figure 7D, F). We believe that the variable phenotype is a result of variable expression of the fusion protein, which has been documented for the pDXA vector backbone (47), and which we have observed by fluorescence microscopy for the pTX-GFP vector (data not shown).

The localization of GFP-Atg8 in macroautophagy mutants
Mutations in the genes of the Atg5-Atg12 conjugation system affect the localization of the GFP-Atg8 fusion protein in D. discoideum. We wanted to know whether mutations in the Dictyostelium orthologues of macroautophagy genes encoding components of signaling complexes, like Atg1 or Atg6, also affect the distribution of GFP-Atg8. We transformed atg1-1, atg6and atg8amoebae with an expression vector containing GFP-Atg8 under control of the constitutive actin15 promoter. In the parent DH1, we observe diffuse cytoplasmic fluorescence, small dots, and semi-circular and round structures ( Figure 8A, B) that are presumably the isolation membranes and autophagosomes, respectively, described in mammalian cells (48). In the atg1-1 mutant, GFP-Atg8 labels a similar structure to that previously observed for strains bearing mutations in the Atg5-Atg12 conjugation system. . However, the labeled structure is even larger, and appears diffuse with brighter areas in the atg1-1 mutant ( Figure 8C, D).
Often, in addition to the larger structure, a few small diffuse dots are observed in the cytoplasm of the cell ( Figure 8D). In contrast, in the atg6mutant we observe diffuse cytoplasmic localization and small dots similar to those seen in DH1. However, it appears that there are fewer of these dots than in wild-type ( Figure 8E), and often we observe a greater proportion of atg6amoebae containing semicircular fluorescent structures, which may be isolation membranes (arrowhead and inset in Figure 8F). In the atg8strain complemented by expression of the GFP-Atg8 fusion, we see structures that we interpret as isolation membranes (arrowheads in Figures 8G and H) and nearcomplete or completed autophagosomes (arrow in Figure 8G), just as in GFP-Atg8expressing DH1.

DISCUSSION
In the present work, we continue our characterization of macroautophagy in the social amoeba D. discoideum. We previously identified putative orthologues of budding yeast Atg1, Atg6 and Atg8, and here undertake more detailed analysis of them. The Dictyostelium proteins share significant sequence similarity and similar domain organization with their counterparts in S. cerevisiae, A. thaliana, and mammals ( Figure   1). To confirm the function of these genes, we generated mutations in atg1, atg6 and atg8 by insertional mutagenesis. All three mutants have developmental defects consistent with those of Dictyostelium macroautophagy mutants such as atg5and atg7 - (33). First, development is less robust in plaques on K. pneumoniae lawns than on nitrocellulose (NC) filters ( Figure 3). Second, atg6and atg8form multi-tipped aggregates on both NC filters and bacterial lawns, whereas atg5and atg7form multi-tipped aggregates on NC filters only. The atg1-1 mutant commonly arrests at the loose mound stage on NC filters ( Figure 3B), but we have observed rare progression beyond this arrest to form multitipped aggregates (data not shown). Spore production mirrors the severity of the developmental defect (Table 1). The three mutants also display other attributes of macroautophagy mutants: reduced protein degradation during development (Table 2) Figure 6). Therefore, we conclude that the Dictyostelium atg1, atg6 and atg8 genes are genuine orthologues of their budding yeast counterparts. All of the phenotypes described above are more severe in the atg1-1 mutant than in atg6or atg8 -. The atg1-1 amoebae die more rapidly when nitrogen-starved, they degrade even less protein during development than the other mutants, and by TEM they have a denser cytoplasm after 36 hours amino acid starvation. The mild phenotype of atg8compared to atg1-1 in Dictyostelium parallels the severity of the macroautophagy defect in S. cerevisiae mutants, where aut7∆ (atg8∆) cells survive for longer than atg1∆ mutants during nitrogen-starvation (29). Consistent with this result, loss of atg8 in yeast does not completely block autophagosome formation, but results in production of very small autophagosomes (29).
Two proteins with homology to S. cerevisiae Atg6 exist in Dictyostelium, which is unique among the organisms in which autophagy has been studied. Both Dictyostelium orthologues are larger than Atg6 in other eukaryotes, and have a unique, extended Nterminus. We show here that atg6 is required for autophagy, whereas the second gene, atg6B, was only identified recently and its function is uncharacterized. The mild autophagic and developmental defects in atg6may occur because of functional redundancy with atg6B. Interestingly, the pathway-specific components of the two yeast PI3K complexes, Atg14 and Vps38, have yet to be identified in the D. discoideum genome. Therefore, it is possible that the two functions ascribed to Atg6 in yeast and mammalian cells, namely autophagy and protein sorting, are performed by distinct polypeptides in Dictyostelium, obviating the need for specific targeting proteins. A test of this hypothesis would be to abrogate atg6B function and determine whether it is required for autophagy, and determine the requirement (if any) for either Atg6 homologue in protein sorting. Of the remaining components of the autophagy-specific PI3K complex, an orthologue of the yeast class III PI3K, Vps34, has been studied in Dictyostelium (49). The ddvps34 gene is duplicated in the genome, and is essential for growth. Reducing ddvps34 mRNA levels by antisense technology results in reduced growth and aberrant development on bacterial lawns, but little effect on growth in liquid medium or development on non-nutrient agar (49). Dictyostelium autophagy mutants also display a more severe developmental defect in plaques on bacterial lawns compared to non-nutrient agar or nitrocellulose filters. Whether either of the Atg6 proteins interacts with DdVps34 remains to be tested. A putative orthologue of the remaining PI3K complex component, Vps15/p150, is present in the Dictyostelium genome (unpublished observation).
The atg1-1 amoebae contain profiles indicative of large clusters of vesicles by TEM ( Figure 7). We have observed similar profiles in the atg5and atg7mutants, although in these mutants the clusters are smaller and they are observed less frequently (data not shown). We do not observe these clusters in the atg6mutant, suggesting that they are rare or absent in this genetic background. The occurrence and size of these clusters observed by TEM correlates with the frequency and size of the structure that is fluorescently labeled in the atg1-1, atg5and atg7mutants expressing GFP-Atg8 ( Figure   8). Therefore, we speculate that these vesicle clusters are the same structures that are labeled with GFP-Atg8. These large vesicle clusters are unique to Dictyostelium, and to our knowledge have not been described in budding yeast macroautophagy mutants.
We suggest that the vesicle clusters represent a stalled or unresolved site of autophagosome biogenesis and/or a membrane source for autophagosomes. In the absence of the Atg5-Atg12 conjugation system and Atg1 function, the target for membrane delivery may be absent and/or non-functional, resulting in an accumulation of vesicle traffic either at the membrane source or the membrane target. The absence of the vesicle clusters in atg6cells suggests that Atg6 may function in the generation of these vesicles. These speculations presuppose a spatially distinct source of membrane and cellular address for membrane delivery (the Dictyostelium equivalent of the PAS). Our future studies aim to address the subcellular localization of the vesicle clusters as a possible membrane source for autophagosomes, and to characterize the composition of this structure by biochemical isolation.         * Sporulation refers to the number of spores produced by a strain as a percentage of the number of spores produced by the parent DH1, expressed as the range of values observed for three independent experiments. For each experiment, the average number of spores produced by each strain on three filters was determined.