Loss of the starvation-induced gene Rack1 leads to glycogen deficiency and impaired autophagic responses in Drosophila.

Autophagy delivers cytoplasmic material for lysosomal degradation in eukaryotic cells. Starvation induces high levels of autophagy to promote survival in the lack of nutrients. We compared genome-wide transcriptional profiles of fed and starved control, autophagy-deficient Atg7 and Atg1 null mutant Drosophila larvae to search for novel regulators of autophagy. Genes involved in catabolic processes including autophagy were transcriptionally upregulated in all cases. We also detected repression of genes involved in DNA replication in autophagy mutants compared with control animals. The expression of Rack1 (receptor of activated protein kinase C 1) increased 4.1- to 5.5-fold during nutrient deprivation in all three genotypes. The scaffold protein Rack1 plays a role in a wide range of processes including translation, cell adhesion and migration, cell survival and cancer. Loss of Rack1 led to attenuated autophagic response to starvation, and glycogen stores were decreased 11.8-fold in Rack1 mutant cells. Endogenous Rack1 partially colocalized with GFP-Atg8a and early autophagic structures on the ultrastructural level, suggesting its involvement in autophagosome formation. Endogenous Rack1 also showed a high degree of colocalization with glycogen particles in the larval fat body, and with Shaggy, the Drosophila homolog of glycogen synthase kinase 3B (GSK-3B). Our results, for the first time, demonstrated the fundamental role of Rack1 in autophagy and glycogen synthesis.

Introduction

The proper balance of anabolic and catabolic processes ensures cellular and organism-wide homeostasis under physiological conditions. Major intracellular catabolic pathways include the ubiquitin-proteasome system, responsible for the regulated degradation of selected proteins and autophagy, capable of degrading all intracellular macromolecules and even whole organelles. The major pathway of autophagy utilizes double-membrane vesicles called autophagosomes to deliver cytoplasmic cargo for degradation in lysosomes. Basal levels of autophagy are required to prevent the accumulation of abnormal protein aggregates that are detrimental for cell function and survival, especially in long-lived cells such as neurons.- Autophagic activity is tightly controlled in all cells: factors increasing cell growth rate usually decrease, while nutrient or growth factor deprivation dramatically increases the levels of autophagy.

A set of about 20 evolutionarily conserved genes required for autophagy has been identified in yeast. These Atg (autophagy-related) gene products orchestrate the formation of early autophagic structures where autophagosome generation takes place. The Atg1 protein kinase complex is directly controlled by TOR (target of rapamycin), a kinase that coordinates growth-promoting stimuli with the availability of nutrients, ATP and oxygen. Inactivation of TOR rapidly activates the Atg1 complex, resulting in autophagy induction. Autophagy also requires a specific phosphatidylinositol-3-kinase (PtdIns3K) complex, the transmembrane protein Atg9 and factors involved in its cycling from cellular reservoirs to phagophore assembly sites, and two ubiquitin-like protein conjugation systems that ensure lipidation and membrane association of the ubiquitin-like protein Atg8. Atg8 is associated with the surface of autophagosomes similar to a coat protein. Atg8 needs to be cleaved off from the outer membrane to allow the fusion of autophagosomes with lysosomes and endosomes, while the Atg8 pool associated with the inner membrane is degraded along with sequestered cargo in autolysosomes., This way, half of these Atg8 protein molecules are broken down in each autophagosomal cycle.

Yeast studies found that the levels of Atg8 can become rate-limiting during autophagy, and that upregulation of this gene is normally observed during induction. In higher eukaryotes, increased transcription from Atg8 homologs and other Atg genes has been documented in certain settings, for example in murine muscles, or in Drosophila fat body and salivary gland cells at the onset of metamorphosis, but not in all cases of autophagy induction.- Basal levels of Atg proteins and the extent of autophagy induction must together determine if transcriptional upregulation of these genes is necessary during the response. The presence of multiple homologs of a single yeast gene may also supply enough proteins to support autophagy induction. Interestingly, the Atg8 gene family has expanded in higher eukaryotes, for example there are two members in Drosophila (Atg8a and Atg8b), and five in mammals with specialized functions during autophagosome formation (GABARAP, GABARAPL1, GABARAPL2, GABARAPL3 and MAP1LC3—this latter gene also produces four different protein isoforms).,

Drosophila is an excellent model for autophagy studies, as polyploid tissues like the fat body grow rapidly during the larval stages, and store nutrients to support metamorphosis later on. Therefore, low levels of autophagy are observed during the feeding period in larval stages L1 (approximately 24–48 h after egg laying—AEL), L2 (48–72 h AEL) and most of L3 (72–120 h AEL). Larvae leave the food at around 108 h AEL (the onset of the wandering period) to find a proper, dry place for pupariation, which is observed around 120 h AEL. Robust induction of autophagy is seen in the larval fat body and midgut in wandering larvae, or in response to starvation during earlier stages.,

We have previously used transcription profiling to search for gene expression changes in dissected fat bodies of feeding and wandering animals. By following the expression levels of 3,200 genes, we found that Atg8 homologs were upregulated, and we have successfully shown that the downregulated gene FKBP39 (FK506-binding protein of 39 kD) is an inhibitor of autophagy. FKBP39 likely acts by inhibiting the transcription factor Foxo, and we found for the first time that Foxo is an important positive regulator of autophagy. Since then, Foxo-dependent transcriptional upregulation of Atg8 homologs during autophagy has been described in murine muscles.,

We reasoned that a similar strategy could be applied by looking at gene expression changes during starvation. An earlier study reported a whole-genome analysis of starvation responses in 48 h AEL (L2 stage) larvae, and found the transcriptional induction of only Atg8a (annotated as CG1534 at that time) and Atg9 of all the autophagy genes. We decided to use a more sensitive newer generation microarray platform to test 86 h AEL (L3 stage) larvae as these are used in most cases for studies on starvation-induced autophagy. In addition, this time point is after the developmental milestone known as the ‘70 h change.’ This simply refers to the observation that all larvae before 70 h AEL die in a few days upon complete starvation, while a fraction of the larvae past 70 h AEL (near the L2- to L3-molt) that have presumably accumulated appropriate nutrient stores to sustain metamorphosis are able to pupariate and give rise to small adults.

Results

Microarray analysis of starvation-induced autophagy in Drosophila larvae

We compared genome-wide expression changes of 86 h AEL control larvae subjected to a 4-h complete starvation to their well-fed siblings. We have also analyzed previously described autophagy-deficient null mutants for Atg7 (on a similar genetic background with control larvae) and Atg1 (from a distinct genetic background), with the purpose of finding gene networks that are specifically regulated in starved autophagy mutants., Genes that showed at least a two-fold induction or repression with a p value < 0.05 were considered regulated. We found that the expression of 2,819 vs. 3,690 and 3,971 of 13,613 genes (represented by unique FBgn identifiers) was starvation-regulated in controls, Atg7 and Atg1 mutants, respectively. Please see for detailed results.

Genes encoding antimicrobial peptides such as Drosomycin and IM10 (immune-induced molecule 10) were among the most highly upregulated genes in all three genotypes (). Growth/insulin signaling is inhibited in response to starvation, which activates the transcription factor Foxo (forkhead box, sub-group O). Induction of these immune effector molecules is directly controlled by Foxo. The FoxA/Forkhead transcription factor target gene CG6770 which is transcriptionally induced upon inhibition of TOR was also induced 37- to 95-fold., Another highly upregulated gene was CG7224, encoding a small, 118-amino acid protein. As it was transcriptionally induced 19.5- to 30.5-fold in the different genotypes, we named this gene Sirup (starvation-upregulated protein). Interestingly, expression of Sirup and its paralog CG15283 was also induced during developmental autophagy of the fat body.

Of the 20 genes showing the strongest downregulation, only two were found in all three genotypes: CG15615 encoding an oxidoreductase (36- to 61-fold repression) and CG15155 encoding a transferase that transfers acyl groups other than amino-acyl groups (30- to 60-fold repression; ). The biological processes in which these gene products are involved are not known, and no phenotypic data are available for these genes.

We performed gene ontology (GO) analysis of the significantly regulated genes in the three different genotypes (Fig. 1A and B; for details). The analysis revealed that genes belonging to 10 enriched GO categories were specifically induced in autophagy mutants (please see for a list of induced genes in these 10 Atg mutant-specific enriched GO categories). Some of these GO categories that are high in the hierarchy of GO terms (such as GO0009056: catabolic process) included genes that were also present in other, more specific GO groups enriched in all three genotypes, which is a clear limitation of GO analyses. Nevertheless, these groups include genes involved in the metabolism of purine nucleotides and in various catabolic processes, such as CG9363 encoding an enzyme involved in the breakdown of tyrosine and l-phenylalanine. Numerous genes encoding members of the ubiquitin-dependent protein catabolic process were also found here, including CHIP, CG1950, CG4661, CG18341 and Ubc84D. Only 3 enriched GO terms were specific for autophagy mutants among repressed genes (please see for a list of repressed genes in these 3 Atg mutant-specific enriched GO categories). Genes in these categories encode proteins involved in DNA metabolism, amplification and replication, such as DNA polymerases CG8142, DNA polymerase α 73 kD and 180 kD, DNA polymerase delta, and essential DNA replication factors Mcm5 (Minichromosome maintenance 5), Mcm10, disc proliferation abnormal, replication factor C subunits RfC3 and RfC38, double parked, DNA ligase I, SuUR (suppressor of under-replication), mus209/PCNA (proliferating cell nuclear antigen) and origin recognition complex subunit 2.

Figure 1. Transcriptional analysis of starvation responses. The number of gene ontology (GO) terms enriched (p < 0.05) among the significantly upregulated (A) and downregulated (B) genes (minimum two-fold change in expression level during the 4-h complete starvation, p < 0.05) in control, Atg7 and Atg1 mutant larvae. See text and for details. (C) lists starvation-induced transcriptional changes (gene expression levels in starved animals relative to the expression levels detected in fed animals) for Drosophila homologs of yeast and human Atg genes, involved in the core mechanism of autophagy. Atg8a encoding a ubiquitin-like protein showed the highest induction during starvation. Interestingly, the highest average upregulation was observed for genes encoding members of the Atg9 cycling complex. (D) Quantitative real-time (QT) PCR analysis of fat bodies dissected from fed and starved control larvae. Numbers represent fold change values, and nonsignificant changes are crossed out in (C and D).

Interestingly, we detected many more genes significantly regulated upon a 4-h complete starvation when compared with an earlier study. Zinke et al. reported 369 and 279 genes significantly induced or repressed, respectively, while we found that 1,584 and 1,235 genes were significantly up- or downregulated in control larvae, respectively. Comparing the list of hits revealed that 170 genes were found induced and 134 repressed in both studies (please see for a list of these genes), which is almost half of all genes identified by Zinke et al. While some of these differences are likely due to the different genetic backgrounds and larval stages tested, we are convinced that our approximately 4.3-fold higher discovery rate is mostly due to the use of a different, more sensitive microarray. This is further supported by a previous study that compared different microarray technologies and found that the long oligonucleotide platform (such as in the case of our study utilizing arrays printed with 60-mer oligos) is more sensitive and produces lower levels of technical variability when compared with cDNA-based systems (used by Zinke et al.).

Most autophagy genes are transcriptionally induced during starvation

Several GO terms that include Atg genes (such as GO0048102: autophagic cell death) were found enriched among the upregulated genes in all three genotypes. To focus more on autophagy, we looked at the expression of Atg gene homologs. Many of these genes showed increased induction in autophagy mutants compared with controls (Fig. 1C). Expression from the Atg7 and Atg1 loci was still observed in the mutants; since these alleles had been previously characterized as genetic nulls, presumably truncated versions of wild-type transcripts were detected.,

Of the genes encoding members of the Atg1 kinase complex, Atg1, FIP200 and Atg101 showed a moderate induction, while the levels of Atg13 did not change. This is in agreement with recent findings that Atg13 is continuously degraded by the ubiquitin-proteasome system, and it is stabilized by the binding of Atg101 during autophagy induction., Therefore, Atg13 protein levels appear to be mainly regulated on the post-translational level. It is important to note that Atg101 showed the highest upregulation in this group (on average 3.5-fold in the 3 different genotypes), but it was mostly due to its stronger induction in autophagy mutants. In controls, FIP200 shows the highest induction in this group (4.6-fold).

Genes encoding members of the autophagy-specific PtdIns3K complex (Vps34, Vps15, Atg6, Atg14) showed an even milder response: only Vps34 that codes for the lipid kinase itself was upregulated in all three genotypes. To our surprise, EDTP (Egg-derived tyrosine phosphatase) showed the highest average induction in this lipid signaling group: 2.9-fold. This gene encodes the Drosophila homolog of the lipid phosphatase Jumpy, a potent antagonist of the autophagy-promoting activity of Vps34 in mammals. The upregulation of a potential inhibitor of autophagy may represent a negative feedback loop that acts to prevent hyperactivation. Elucidation of this hypothetical mechanism needs further studies.

Interestingly, genes encoding members of the pre-autophagosomal cycling complex showed the strongest average induction, between 3.2- and 6.2-fold. Atg9 is the only transmembrane protein among Atg gene products, and it has been shown to cycle between early autophagic structures and potential intracellular membrane sources to ensure membrane transport to forming autophagosomes. Atg2 and Atg18 are required for the cycling of Atg9 in yeast.

Half of the genes (5/10) encoding members of the ubiquitin-like protein conjugation systems did not even show an upregulation. Atg4a, Atg7, Atg8b and Atg16 were moderately induced, while the highest upregulation of all the Atg genes was detected for Atg8a, 8.4-fold on average.

Atg8a itself is a specific cargo selectively degraded during autophagy. Additional substrates have also been described, such as p62/Ref(2)P, which is an intracellular receptor of ubiquitinated proteins for selective autophagy, in addition to its scaffolding roles in multiple signaling pathways. p62 binds to Atg8 homologs, so it is continuously degraded during autophagy. We found that p62 is transcriptionally induced 6-fold during starvation, potentially in part to compensate for its accelerated rate of degradation during increased autophagy. A more moderate upregulation was detected for Blue cheese/Alfy, whose protein product is involved in the selective degradation of cytosolic protein aggregates.

Next we sought to expand on our findings obtained by microarray analysis of starved whole larvae by focusing on the fat body. We followed the expression of selected genes by quantitative real-time PCRs, comparing fat bodies dissected from starved vs. fed control larvae (Fig. 1D). These QT-PCR experiments further supported our most important findings: among the genes encoding members of the Atg1 kinase and lipid kinase complexes EDTP showed the highest upregulation (6.5-fold), which is still moderate when compared with genes that showed the strongest induction. Atg8a was by far the most highly upregulated gene (169-fold), and all members of the membrane transport group were strongly induced (11- to 35-fold). The expression of Sirup also increased 60.4-fold. There were also important differences, such as in the case of Atg5 that showed a 7.6-fold upregulation in fat bodies while no significant change was detected in whole larvae, as it was probably masked by other tissues.

Rack1 is required for a full autophagic response to starvation

We selected a set of genes for further studies that were starvation-regulated in all three genotypes, and no autophagic function had been assigned to these in spite of having been already characterized to some extent. For example, Rack1 transcription increased 4.1-, 5.5- and 4.1-fold in controls, Atg7 and Atg1 mutants, respectively. QT-PCR analysis also showed that Rack1 was upregulated 21.6-fold in the fat body upon starvation.

Silencing of Rack1 expression reduced autophagy based on mCherry-Atg8a (Fig. 2A and N), Lysotracker Red staining and Lamp-GFP, a reporter for lysosomes. Numerous small, presumably inactive Lamp-GFP positive lysosomes were seen in fat body cells of well-fed control larvae, which were not strongly acidic as revealed by lack of Lysotracker Red labeling (Fig. 2B). In contrast, large Lamp-GFP and Lysotracker-positive autolysosomes were formed during starvation (Fig. 2C). Silencing of Rack1 reduced the number and size of Lysotracker and Lamp-GFP positive autolysosomes formed during starvation (Fig. 2D, see also Fig. 2O–Q for statistics).

Figure 2. Rack1 is required for the full autophagic response to starvation, and for efficient glycogen synthesis. (A) Knockdown of Rack1 results in reduction of autophagy in response to starvation. Please compare the number of mCherry-positive dots in RNAi cells vs. control cells in each panel (red; the red channel is also shown in greyscale in insets). RNAi expressing cells are marked by coexpression of GFP (green). Cell nuclei are labeled with DAPI (blue). (B) Numerous small Lamp1-GFP dots are seen in fat body cells of well-fed control larvae, which are not co-labeled by Lysotracker Red. (C) In contrast, large Lamp1-GFP and Lysotracker-positive autolysosomes are formed during starvation. (D) Silencing of Rack1 strongly reduces the number of Lysotracker particles that are formed during starvation; note that the size of Lamp1-GFP dots is also more similar to fed control cells. (E and F) The number of starvation-induced mCherry-Atg8a dots is reduced in Rack1 null mutant fat bodies compared with controls. (G and H) Similarly, the number of Lysotracker-positive autolysosomes is reduced in Rack1 mutants compared with controls. (I) Starvation-induced autophagy is rescued by heat shock-mediated expression of a Rack1 transgene in Rack1 null mutant animals. (J) Transmission electron microscopy shows the presence of large glycogen stores (g), numerous double-membrane autophagosomes (arrowheads) and autolysosomes (arrows) in fat body cells of control animals starved for 4 h in 20% sucrose. (K) A portion of (J) is shown enlarged. (L) Fewer autophagic structures and glycogen particles are detected in Rack1 mutants. (M) Heat-shock mediated expression of a Rack1 transgene in Rack1 null mutant animals partially rescues the reduced autophagy and glycogen accumulation phenotypes of the mutants. (N–S) Statistical analyses of mutant and RNAi phenotypes on autophagy and glycogen stores. Graphs represent dots/cell (N–P), average dot size (Q) and normalized cytoplasmic area (R and S), **p < 0.01 (two-tailed, two-sample unequal variance Student’s t-test). (T) Knockdown of Rack1 expression cell-autonomously decreases glycogen stores. (U and V) No specific Rack1 labeling is observed in Rack1 mutants, and glycogen stores are strongly decreased compared with heterozygous controls. Scale bars: 30 μm for all fluorescent images (indicated in A, E and T–V), 10 μm for (J, L and M, indicated in J) and 2 μm for (K).

To expand on these RNAi studies we have also analyzed a previously described null mutant Rack1[1.8]. Using the same mCherry-Atg8a and Lysotracker assays, we observed again a strong reduction of autophagy in Rack1 mutants compared with heterozygous controls (Fig. 2E–H; see also Fig. 2N and O for statistics). As a final proof for the role of Rack1 in autophagy, we have generated transgenic animals expressing Rack1 with a C-terminal FLAG-HA tag by using a DNA construct from the ongoing Drosophila proteomics project. Heat shock-mediated expression of Rack1 in mutant larvae partially rescued their autophagy defect (Fig. 2I and O). To avoid a possible effect of the heat shock on autophagy, all genotypes were heat shocked in these Lysotracker and electron microscopy experiments (see below).

Similar to the above findings, ultrastructural analysis showed that while numerous autophagosomes and large autolysosomes were observed in heterozygous controls, the cumulative area of these structures was decreased in cross-sections of Rack1 mutant fat body cells, and heat shock-mediated expression of Rack1 in mutant larvae again showed a partial rescue (Fig. 2J–M; see Fig. 2R for statistics). The partial rescue effect may reflect insufficient expression levels, or more likely it may be caused by the C-terminal tag interfering with full protein activity. Still, the reduction of autophagosome area was completely rescued in these experiments (p = 0.02 for mutant and rescue, and p = 0.84 for control and rescue sample pairs, respectively), suggesting that Rack1 is required for efficient autophagosome formation (Fig. 2L).

Interestingly, we found that Rack1 was dispensable for developmental autophagy induced in fat body and midgut cells prior to metamorphosis in late L3 stage wandering animals (). In contrast, p62 aggregates accumulated in Rack1 RNAi cells in well-fed early L3 stage animals (). This finding suggests that Rack1 is also involved in basal autophagy, as the ubiquitin- and Atg8-binding protein p62 is continuously degraded by autophagy along with aggregated ubiquitinated proteins.

Rack1 is involved in glycogen synthesis

Glycogen stores are normally present in fat body, midgut and muscle cells in Drosophila larvae. A striking phenotype of these Rack1 mutant fat body cells was the 11.8-fold reduction of glycogen particle area compared with heterozygous controls in electron micrographs, and this effect was also partially rescued by Rack1 expression (Fig. 2J–M; see Fig. 2S for statistics). Reduced glycogen stores were most likely caused by defects in glycogen synthesis induced by the 4 h starvation in 20% sucrose rather than accelerated degradation, as no glycogen particles could be observed in fat bodies of well-fed control or Rack1 mutant larvae (). In these latter cases, the lack of glycogen was probably due to the high-protein diet that we routinely feed larvae before starving them, by supplementing the medium with live yeast paste. Taking advantage of a monoclonal antibody specific for glycogen, we were also able to show that silencing of Rack1 decreased glycogen stores in a cell-autonomous manner in larvae fed a high sugar-high protein diet, that is, medium supplemented with live yeast paste prepared in 20% sugar; similar phenotypes were observed in mutants (Fig. 2T–V).

Endogenous Rack1 localizes to early autophagic structures and glycogen

Immunofluorescence analysis revealed that Rack1 is present in numerous cytoplasmic particles in most tissues. Knockdown of Rack1 expression reduced staining both in fat body and midgut cells (Fig. 3A and B). To assess the potential role of Rack1 in autophagy, we performed colocalization experiments. Atg8a fusion proteins are the best characterized fluorescent reporters for autophagosome generation. We found that endogenous Rack1 colocalized with GFP-Atg8a in 5.8% of the cases (n = 10/171 dots) (Fig. 3C). Vice versa, 6.1% of GFP-Atg8a particles colocalized with Rack1 (n = 11/181 dots). In contrast, not a single case of colocalization was observed between endogenous Rack1 and the lysosome reporter Lamp1-GFP (Fig. 3D). We next used immunogold labeling of endogenous Rack1 to localize it on the ultrastructural level. We found that Rack1 was associated with 7.7% of pre-autophagosomal phagophores and autophagosomes (n = 17/221) (Fig. 3E; see also Fig. 3G). Again, no labeling of lysosomes was seen.

Figure 3. Localization of endogenous Rack1. (A and B) Knockdown of Rack1 reduces anti-Rack1 immunostaining in both midgut and fat body cells in a cell-autonomous manner. (C) Rack1 shows a partial colocalization with the (pre-) autophagosome reporter GFP-Atg8a. (D) In contrast, no overlap was found between Rack1 and the lysosome reporter Lamp1-GFP. Boxed areas in (C and D) are shown enlarged. (E) Immunogold labeling of Rack1 is associated with a pre-autophagosomal phagophore near the growing edges (asterisk) and an autophagosome (red arrowhead). Please note the characteristic cleft (empty-looking space) between the two membrane sheets of phagophores and autophagosomes, which facilitates their recognition in glutaraldehyde-fixed samples. In contrast, no labeling of a lysosome is observed (arrow in E). Some of the gold particles are highlighted by yellow arrowheads. (F) Gold particles representing endogenous Rack1 are frequently associated with the peripheral parts of glycogen stores (g) in fat body cells. (G) Rack1 is also present in cytoplasmic clusters (c); an autophagosome (red arrowhead) is also labeled (yellow arrowhead). (H and I) Rack1 is closely associated with glycogen granules in the fat body, and it colocalizes with Shaggy, the Drosophila homolog of GSK-3B. Boxed areas of (A and B) are shown enlarged. Scale bars: 10 μm for (A–D, H and I), 5 μm for (C’–D” and H’–I”) and 1 μm for (E–G). All colocalization experiments were performed using confocal laser scanning microscopy.

Immuno-electron microscopy analysis also identified that endogenous Rack1 localized in the periphery of, or immediately next to glycogen particles in fat body cells, in addition to being present in unidentified cytoplasmic clusters (Fig. 3F and G). In indirect immunofluorescence experiments most Rack1-positive dots localized next to or in the periphery of glycogen particles in laser scanning confocal microscopy as well (Fig. 3H). 77% of glycogen particles had associated Rack1 dots (n = 426/553), while 48% of Rack1 dots were associated with glycogen (n = 394/822).

The activity of glycogen synthase is a major determinant of glycogen levels. Phosphorylation of glycogen synthase by Shaggy, the Drosophila homolog of glycogen synthase kinase 3B (GSK-3B) promotes glycogen synthesis. A physical interaction between the human homologs of Rack1 and GSK-3B was already reported. In line with that, we found that 73.9% of endogenous Rack1 colocalized with endogenous Shaggy, and 39.4% of Shaggy positive dots were also positive for Rack1 (n = 82/111 and n = 82/208 dots, respectively) (Fig. 3I). In contrast, no colocalization was found between Shaggy and GFP-Atg8a (not shown).

Discussion

We have used whole-genome microarrays to identify genes regulated during the starvation response in Drosophila, and also compared gene expression changes between control, Atg7 and Atg1 mutant larvae. Based on a gene ontology analysis, genes involved in catabolic processes such as proteasomal degradation and amino acid catabolism were more significantly upregulated in starved autophagy mutants, which is likely a compensatory reaction. On the other hand, genes required for DNA replication were specifically downregulated. Mitotic tissues continue to grow and divide even in Drosophila larvae deprived of all nutrients. Repression of genes required for DNA replication in autophagy mutants suggests that energy-consuming DNA replication processes are strongly inhibited in completely starved autophagy mutants, presumably because polyploid cells are not able to supply nutrients to support diploid cell divisions in the absence of autophagy.

We detected increased transcription of most Atg genes upon starvation, and an even higher upregulation was seen for most genes in fat bodies dissected from starved vs. fed animals. The most highly induced gene was Atg8a in all cases, in agreement with previous studies that showed the importance of Atg8 induction during autophagy in various models.,, It is important to note that Atg8b is mostly expressed in adult testis, and its very low level larval expression is restricted to the fat body. In line with this, mutation of Atg8a completely blocks starvation-induced autophagy in the fat body confirming that Atg8b expression is unable to sustain autophagy in this setting, although Atg8b was also upregulated 118-fold during starvation in the fat body. Induction of Atg8a is likely necessary to make up for the degradation of half of its protein products involved in each autophagosomal cycle. Members of the Atg9 cycling complex showed the strongest average upregulation of the functional groups of Atg proteins. We speculate that the sudden induction of autophagy in polyploid larval Drosophila tissues requires a lot of membrane to sustain the high level of autophagosome generation, explaining the increased transcription of genes whose products are required for membrane transport.

In this work we chose Rack1 (receptor of activated protein kinase C 1) for further analysis. Rack1 was induced in whole animals and in dissected fat bodies during starvation, and loss of Rack1 impaired both starvation-induced and basal autophagy, while it had no effect on developmentally programmed autophagy of the fat body. These findings indicate that Rack1 is involved in multiple but not all types of autophagy.

Autophagosomes generally have a short half-life of 5–10 min, which is the most likely explanation why basal levels of autophagy are very difficult to visualize in most tissues. In contrast, autophagosomes appear in high numbers during starvation. The autophagosomal compartment was reduced in Rack1 mutants which was rescued completely by transgenic expression of Rack1, and Rack1 was at least transiently associated with phagophore assembly sites and autophagosomes, altogether suggesting that Rack1 is required for efficient autophagosome formation. As we observed no defects in cell size/cell growth, DNA polyploidization or lipid droplet accumulation in Rack1 loss of function cells, the autophagy defect seems to be a highly specific phenotype and not just a consequence of general problems with normal cellular functions.

Rack1 is an evolutionarily conserved guanine nucleotide-binding scaffold protein with a WD40-repeat: 77% (243/315) of the amino acid residues are identical and 87% (275/315) are similar between Drosophila Rack1 and human GNB2L1. A proteomic study of Atg complexes found that GNB2L1 interacts with human homologs of Atg1, Atg4, Atg14 and Atg18. Although GNB2L1 was not classified as a high-confidence interacting partner, these data support our hypothesis that Rack1 may act as a scaffold, transiently binding multiple Atg proteins at phagophore assembly sites to promote maximal activity. In line with that, it is interesting to note that the potential interacting partners of human Rack1 include a member from all four Atg protein complexes. Further biochemical studies such as co-immunoprecipitation experiments are necessary to verify this hypothetical mechanism.

In addition, half of the Rack1-positive dots localized to glycogen particles, and loss of Rack1 prevented the proper formation of glycogen stores in larval fat body cells. The high degree of colocalization with GSK-3B strongly suggests that Rack1 is associated with a pool of GSK-3B that promotes glycogen synthesis. The known interacting partners of Rack1 include three subunits of AMPK (AMP-activated protein kinase): PRKAA1, PRKAA2 and PRKAB2. The β-subunit of AMPK has a glycogen-binding domain that targets a pool of this kinase to bind the surface of the glycogen particle. Activated AMPK turns on catabolic processes to generate ATP, and it also inhibits glycogen synthesis through direct phosphorylation of glycogen synthase. We hypothesize that a signaling complex containing Rack1, GSK-3B and AMPK assembles on glycogen particles to regulate the rate of synthesis through phosphorylating different amino acid residues of glycogen synthase, with AMPK being responsible for targeting this complex to glycogen. In this scenario, the activity of multiple kinases including AMPK and GSK-3B would determine the rate of glycogen synthesis. These interactions may be less relevant for Rack1’s role in autophagy for a number of reasons. First, GSK-3B never colocalized with Atg8a, and overexpression of GSK-3B actually inhibits starvation-induced autophagy (our unpublished observation). Second, the exact role of AMPK in starvation-induced autophagy of the fat body is not clear: mutation of SNF4Aγ encoding a regulatory subunit of AMPK has been shown to lead to impaired autophagic responses in mutants, but a more recent report suggested that AMPK mutants show a persistent starvation phenotype that may interfere with further induction of autophagy by additional experimental starvation., In line with that, none of the more than 10 different transgenic RNAi and dominant-negative AMPK lines we tested showed any suppression of starvation-induced autophagy in clonal analysis (our unpublished observations), potentially supporting the hypothesis that the autophagy phenotype of AMPK mutants may not be cell autonomous in the Drosophila fat body.

Rack1 was originally described as a cytoplasmic receptor for activated protein kinase C (PKC). The structure of Rack1 resembles that of the β-subunit of heterotrimeric G proteins. The individual WD40 repeats can simultaneously bind to different proteins, making Rack1 a candidate platform for integrating several signaling pathways. Rack1 was shown to physically interact with β-integrin, various kinases including PKC, AMPK, GSK-3B and Src, protein phosphatase 2A, focal adhesion components, and even ribosomes. Indeed, Rack1 plays a role in a wide range of processes including cell adhesion and migration, cell survival and translation., Recent findings showed that Rack1 promotes and is required for progression of several types of cancers, and its increased expression predicts poor clinical outcome for breast cancer patients.-

Here we have shown that Drosophila Rack1 is also involved in the autophagic response to starvation, potentially acting again as a scaffold protein during the formation of autophagosomes. In addition, Rack1 is necessary for the proper generation of glycogen particles in the larval fat body, likely through recruiting Shaggy/GSK-3B to promote glycogen synthesis. Taken together, we have demonstrated that novel roles in autophagy and glycogen synthesis must be added to the already diverse list of functions for Rack1.

Materials and Methods

Drosophila stocks and culture

Flies were raised at 25°C on standard cornmeal/agar media, at 60% humidity and a 12-h light/12-h dark daily cycle. The following stocks were used: CG5335[d30], Atg7[d77], Atg7[d14], (d14 deletes the promoters and the first parts of both Atg7 and Sec6, d77 deletes both CG5335 and Atg7, and d30 deletes CG5335 only; therefore, d77/d14 animals are mutant for Atg7 only and heterozygous for both CG5335 and Sec6, while d30/d14 are heterozygous for all three genes), Atg1[d3D], UAS-Lamp1-GFP, UAS-GFP-Atg8a, hs-Flp; UAS-Dcr2; R4-mCherry-Atg8a, Act > CD2 > Gal4, UAS-GFPnls (gift of Tom Neufeld), Df(2L)rack1-mts, Rack1[1.8], (the 1.8 allele carries a premature stop codon instead of the sixth amino acid; we used hemizygous animals in trans with the deficiency Df(2L)rack1-mts that deletes both Rack1 and mts in all experiments), hs-Gal4 (Bloomington Drosophila Stock Center, BDSC), and transgenic RNAi lines for Rack1 KK104470 (designated as Rack1 RNAi throughout the text and figures), HMS01173 (designated as Rack1 RNAi_2 in ) (provided by the Vienna Drosophila RNAi Center, VDRC and BDSC, respectively). Starvations were routinely performed by floating well-fed 72–96 h AEL (after egg laying) larvae on top of a 20% sucrose solution in an Eppendorf tube for 4 h. For complete starvations (no calorie), we placed well-fed animals onto filter papers soaked in PBS. Twenty to 30 larvae were transferred into a new vial amply supplemented with live yeast culture (high protein diet) 12–24 h before starvation to reduce crowding and make sure that all animals were well-fed at the onset of the starvation experiments, which is necessary for a proper autophagic response. For diet high in both protein and sugar, yeast paste was prepared in a 20% sucrose solution.

Microarray analysis

Our microarray experiments conformed to MIAME guidelines, and full data sets are available at www.ebi.ac.uk/arrayexpress under accession number E-MEXP-3352.

Four microarrays were used for analyzing each genotype: two independent fed and starved samples were collected (biological replica), and a dye swap was performed for each fed/starved sample pair (technical replica—please see below). Eighty-six ± 2 h AEL L3 stage larvae were allowed to feed on standard fly food supplemented with yeast (fed) or starved in PBS for 4 h (starved). Approximately 30 mg of larvae were collected and frozen immediately in liquid nitrogen in RNAlater (Sigma, R0901). Total RNA was purified using a NucleoSpin RNA II RNA isolation kit (Macherey-Nagel, 740955). RNA samples were stored at –80°C in the presence of 30 U Prime RNase inhibitor (Fermentas, EO0382). For probe preparation, 1 μg of total RNA were first reverse transcribed in 10 μl volume using Oligo(dT) Primer and ArrayScript enzyme, then the second cDNA strand was synthesized in 50 μl final volume using DNA polymerase and RNase H. Amino allyl modified aRNA were then synthesized by in vitro transcription using aaUTP and T7 Enzyme mix. All these steps were done using AminoAllyl MessageAmpTM II aRNA Amplification Kit (Ambion, AM1753). Six micrograms of amino allyl modified amplified RNA was labeled with either Cy5 or Cy3 dyes in 10 μl volume (GE Healthcare, PA23001, PA25001) and purified using RNA purification columns (Macherey Nagel, 740955). The specific activity and concentration of the labeled aRNA samples was checked by spectrophotometry (NanoDrop 3.1.0, Rockland). 4 × 44 k format Drosophila Gene Expression Microarrays (Agilent, G2519F-021791) were used to determine gene expression changes. 300 ng of Cy5 and Cy3 labeled aRNA in 19 μl volume, 5 μl 10 × blocking agent and 1 μl 25 × fragmentation buffer were mixed together and incubated at 60°C for 30 min. Twenty-five microliters 2× GEx hybridization fuffer was added to each sample to stop the fragmentation reaction. All these steps were done using gene expression hybridization kit (Agilent, 5188-5242). Forty-eight microliters of these mixes were used for the hybridization in microarray hybridization chambers (Agilent, G2534A). The chambers were then loaded into a hybridization rotator rack (~5 rpm) and incubated at 65°C for 17 h. After hybridization the slides were washed in wash buffer 1 (Agilent, 5188-5327) at room temperature for 1 min then in wash buffer 2 at 37°C for another 1 min before scanning. Each array was scanned at 543 nm (for Cy3 labeling) or at 633 nm (for Cy5 labeling) in Agilent Scanner using the built-in XDR (Extended Dynamic Range) function with 5 μm resolution. Output images were analyzed with the Feature Extraction software from Agilent (ver.9.5.1.1.) using the two-color gene expression protocol (GE2_v5_95_Feb07) to get the raw data. All ratios were normalized using the Lowess normalization method. Two-tailed two-sample unequal variance Student’s t-test with a two-fold change cut off was applied to determine which genes showed significantly altered expression levels (p < 0.05). Additionally, we applied multiple testing correction (Benjamini-Hochberg method; p value 2 in ) using Genspring to estimate false discovery rate (corrected p value in supplemental tables) associated with such a large number of comparisons.

QT-PCR experiments

RNA samples were prepared from fat bodies dissected from fed and starved control larvae as described above. QT-PCR reactions were performed on 4 independently collected fed and 4 independently collected starved fat body samples. The reactions were performed on a RotorGene 3000 instrument (Corbett Research) with gene-specific primers and SybrGreen protocol to monitor gene expression. Two micrograms of total RNA was reverse transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems, 4322171) according to the manufacturer’s instructions in a final volume of 30 μL. The temperature profile of the reverse transcription was the following: 10 min at room temperature, 2 h at 37°C, 5 min on ice and finally 10 min at 75°C for enzyme inactivation. After dilution with 30 μL of water, 1 μL of the diluted reaction mix was used as template in QT-PCR reactions, using FastStart SYBR Green Master mix (Roche, 04673484001) at a final primer concentration of 250 nM under the following conditions: 15 min at 95°C, 45 cycles of 95°C for 15 sec, 60°C for 25 sec and 72°C for 25 sec. Fluorescence intensity of SYBR Green was detected after each amplification step. Melting temperature analysis was done after each reaction to check the quality of the products. Relative expression ratios were calculated as normalized ratios to GAPDH. As an additional control, no significant change was observed in expression from the Actin5C locus. Nontemplate control sample was used for each PCR run to check the primer-dimer formation. The final relative gene expression ratios were calculated as ΔΔCt values. Primer sequences are available upon request. Two-tailed two-sample unequal variance Student’s t-test was used to calculate p values.

Bioinformatic analysis

GO analysis was performed using DAVID Bioinformatics Resources 6.7 at http://david.niaid.nih.gov. We selected unique FBgn identifiers of genes showing an at least two-fold expression change in response to starvation with p < 0.05 from all three genotypes. Separate lists of induced and repressed FBgn identifiers were uploaded, and the GO terms of biological processes functional annotation tool was used for functional annotation clustering with default options. We then compared the significantly enriched (p < 0.05) GO categories among the three genotypes. As most genes are represented by multiple 60-mer oligos on the whole-genome microarray, all p < 0.05 fold change values were collected for each gene of interest, averaged, and then visualized using JColorGrid for the expression analysis of individual Atg genes. BLASTP searches were used to find homologous gene products in yeast, Drosophila and human, and the top-scoring hit is shown for each gene in Figure 1C.

Construction of UAS-Rack1 transgenic lines and genetic rescue experiments

The pUAST-Rack1-FLAG-HA clone UFO04269 was obtained from DGRC (Drosophila Genomics Resource Center). Following sequencing, standard PhiC31-mediated transformation procedures were used to obtain transgenic Drosophila lines carrying UAS-Rack1 on a 65B2 platform chromosome (Bestgene). Expression of UAS-Rack1 was induced with hs-Gal4 in a Rack1 null mutant background by a 1 h heat shock at 37°C, followed by a 4 h starvation.

Histology and imaging

Clonal analysis in polyploid larval Drosophila tissues and Lysotracker stainings were performed as described previously., For statistics, Lysotracker, mCherry-Atg8a and Lamp1-GFP puncta were manually counted in Photoshop. Two-tailed two-sample unequal variance Student’s t-test was used to calculate p values. The following antibodies were used for immunostainings: rabbit polyclonal anti-Rack1 (1:500, gift from Julie Kadrmas), mouse monoclonal anti-glycogen (1:20, gift from Otto Baba), mouse monoclonal anti-Shaggy (4G1E11, 1:100, gift from Marc Bourouis), mouse monoclonal anti-GFP (1:1,000, Invitrogen, A11120), Alexa488-conjugated goat anti-mouse (Invitrogen, A11001, 1:1,000), Alexa568-conjugated goat anti-rabbit (Invitrogen, A21069, 1:1,000). A Zeiss Axioimager M2 microscope equipped with an Apotome2 unit and a Plan-NeoFluar 40× 0.75 NA objective was used to capture images with Axiovision software using MinMax setting for automatically adjusting image levels. Confocal images were acquired on an Olympus FV500 laser scanning microscope with a 63× 1.45 NA oil-immersion objective in sequential scanning mode. Primary images were processed in Adobe Photoshop to produce final figures.

Transmission electron microscopy (TEM)

Tissues were processed for TEM analysis as described previously. For immuno-electron microscopy, rabbit polyclonal anti-Rack1 (1:50) was used as described previously. Images were captured on a JEOL JEM-1011 transmission electron microscope using an Olympus Morada 11 megapixel camera and iTEM software (Olympus). For statistical analysis, four 5,000 × magnification images were taken randomly from four larvae in each genotype (16 images per genotype), and the area of autophagic structures and glycogen has been manually encircled in Photoshop to obtain area values. Relative cytoplasmic area values and two-tailed two-sample unequal variance Student’s t-tests were calculated using Excel.