The MVB pathway delivers membrane proteins into the lumen of the vacuole for degradation. The resulting amino acids are pumped into the cytoplasm by vacuolar transporters where they are reused for protein synthesis ( 15 ). This amino acid recycling pathway might play an important role during starvation conditions. To test this hypothesis, wild-type (for ESCRT function) or vps4Δ strains, both leucine auxotrophs (leu2Δ), were transferred into growth medium lacking leucine, and the cellular concentrations of various free amino acids were determined at different time points. Overall, the starting amino acid levels and the response to leucine starvation was very similar between wild-type and the MVB mutant strain vps4Δ ( , Table S1 ). The levels of alanine, isoleucine, tyrosine and phenylalanine rapidly increased following leucine starvation, which is possibly caused by a broad translational shutdown combined with the continued uptake of these amino acids from the medium. The major difference between the two strains was the extent to which the leucine levels dropped. Wild-type cells maintained 3-5% of the initial leucine concentration, while that of vps4Δ cells decreased to non-detectable amounts of leucine ( ). This result was consistently observed in multiple independent experiments ( Fig. S2 A ). The detection limits of the amino acid analyses are at least 10-fold less than the values measured in starved wild-type cells (<0.3%). These data indicated that the MVB pathway plays a role in maintaining minimal levels of amino acids during starvation conditions. Interestingly, we observed a dramatic decrease in methionine levels in both strains, though methionine was not limited in the growth medium. Similar decreases in methionine levels were observed in cells starved for lysine ( ) or histidine (data not shown), suggesting that this drop in free methionine might be part of the starvation response and connected to a block in general translation initiation.

Deletion of two key vacuolar peptidases caused a loss in viability during starvation similar to that observed with ESCRT mutants (pra1Δ prb1Δ, ). This result is consistent with the idea that the diminished survival of ESCRT mutants during starvation was due to the lack of protein degradation via the MVB pathway. Finally, as a control we compared the survival rate of wild-type and vps4Δ cells in YNB complete medium after the cells reached stationary phase ( Fig. S1 B ). Under these naturally achieved starvation conditions, both strains showed survival rates similar to those observed following leucine depletion, indicating that the survival phenotype of the ESCRT mutants is not specific to leucine starvation.

During our work with ESCRT mutants, we observed that these strains rapidly lost viability when kept in stationary phase on agar plates. Furthermore, previous studies have shown that diploid strains carrying mutations in the ESCRT machinery exhibited sporulation defects ( 13 ). These observations suggested that a block in the MVB pathway might affect starvation- response pathways. To test this idea further, we examined the survival rate of wild-type and ESCRT-mutant strains under starvation conditions. Note that the yeast strain SEY6210, referred to as ‘wild type’, contains several mutations that render the strain auxotrophic for leucine, tryptophan, histidine, lysine and uracil ( ). For most of our experiments, we used leucine-free medium to induce starvation conditions. Leucine is the most common amino acid found in proteins, and leucine synthesis is dependent on LEU2, which is mutated in SEY6210. The data shown in demonstrated that lack of either Vps4 or Vps24, two proteins essential for a functional MVB pathway, dramatically decreased the survival rate during leucine-starvation conditions (see Fig. S1 A for a prolonged analysis of cell survival during starvation). This decreased survival rate was similar to that of the autophagy mutant atg29Δ. These data showed that half of the ESCRT mutant cells died within the first two days. In contrast, wild-type cells exhibited no loss of viability in the same time period, but instead showed a ~20 percent increase in cell number, possibly due to the completion of cytokinesis by cells in later stages of mitosis. This increase in cell number was not observed in cells deleted for IST1, a gene encoding a regulatory ESCRT factor not essential for MVB formation ( 9 , 14 ). These data demonstrated a strong correlation between MVB pathway activity and the survival of yeast cells under starvation conditions.

Ist1 protein levels regulate the MVB pathway

Ist1 has been shown to function in the recruitment of Vps4 to ESCRT-III, suggesting a supportive role for Ist1 in MVB trafficking (9, 12). However, overexpression of Ist1 results in reduced recruitment of Vps4 to ESCRT-III and a defect in MVB trafficking (9), suggesting that Ist1 can have both a positive and negative effect on Vps4 activity, depending on Ist1 protein levels. To test this idea, we preformed an in vitro ‘GST-pull-down’ experiment using purified recombinant proteins. Did2 is an important factor in the recruitment of Vps4 due to its interactions with Ist1, Vps4 and ESCRT-III (12). Therefore, we used the binding of Vps4 to Did2 as readout for recruitment efficiency. For the assay, the C-terminal half of Did2 (GST-Did2(CT), amino acids 113-204), which contains the Vps4 and Ist1 interaction regions, was immobilized on glutathione-sepharose beads. These beads were then incubated with an equal amount of the ATP-locked form of Vps4 (Vps4E233Q) in the presence of various concentrations of Ist1. The results indicated that, consistent with previously published data (12), the addition of an approximate equimolar amount of Ist1 increased the recruitment of Vps4 to Did2 ( ), possibly through the formation of a trimeric Vps4-Ist1-Did2 complex. In contrast, increasing amounts of Ist1 caused a reduction of Did2-associated Vps4 ( ) suggesting that at higher concentrations, Ist1 might bind to Did2 and Vps4 independently, resulting in the formation of Did2-Ist1 and Ist1-Vps4 complexes that inhibit the formation of the trimeric complex. This effect of Ist1 on Vps4 recruitment to GSTDid2 was consistently observed in other independent pulldown assays, although the Ist1 concentration necessary to affect the Vps4-Did2 interaction varied (Fig. S2 B). These variations are likely due to the tendency of Ist1 to oligomerize/aggregate (16), thereby changing the concentration of soluble Ist1 protein in the in vitro assay. These in vitro observations are consistent with the altered endosomal recruitment of Vps4 when Ist1 protein levels are elevated in vivo (9).

Additional support for this regulatory model comes from the observation that artificially high cellular levels of Ist1 induced by using a GAL1 promoter (pGAL1-Ist1-HA), resulted in loss of viability during leucine starvation ( ). This drop in viability was similar to the drop observed in cells expressing the dominant-negative mutant vps4(E233Q), which renders the MVB pathway nonfunctional (17). In comparison to the data shown in , this viability assay was performed using a different yeast strain (W303) and a different carbon source (galactose).

Analysis of cell-surface protein trafficking indicated that Ist1 protein levels in exponentially growing cells have a negative regulatory effect on the MVB pathway. GFP-tagged versions of the plasma membrane proteins Ftr1 (iron transporter), Fur4 (uracil transporter) and Can1 (arginine transporter) showed obvious vacuolar staining when expressed in cells deleted for IST1, whereas no vacuolar signal was observed in wild-type cells ( ). These observations suggested that loss of Ist1 caused heightened downregulation of plasma membrane proteins, likely due to increased MVB pathway activity in these mutant cells.

We analyzed Ist1 protein levels in growing cells by Western blot using a functional HA-tagged version of Ist1. The data showed that Ist1-HA levels change dramatically with growth conditions ( ). Ist1 levels increased at low cell densities, peaked during exponential growth and then fell to barely detectable levels as cells entered stationary phase. In contrast, the concentration of the control protein Snf7 (ESCRT-III subunit) showed only minor changes during the different growth phases of the cells.

The drop in Ist1 levels coincided with the end of the growth phase and an increase in phosphorylated translation-initiation factor eIF2 (time = 6h, ). Phosphorylation of eIF2 at the α-subunit has been shown to block general protein translation in response to amino acid starvation (1), suggesting that starvation might have caused the decrease in Ist1 levels. Consistent with this idea, free amino acid analysis of the same samples revealed a sharp drop in cellular lysine levels after six hours of growth, which occurred concurrently with a marked decline in Ist1 levels (red-marked time point in ). The yeast strain used for the experiment is a lysine-auxotroph, and after approximately six hours of growth, the lysine provided by the medium was depleted ( , bottom panel), causing the induction of starvation response pathways. The resulting changes in the levels of other tested amino acids were similar to those observed during leucine starvation ( ).

To identify the mechanism responsible for regulating Ist1 protein levels, the growth-dependent Ist1 expression analysis was repeated using a strain containing a genomically integrated HA-tag immediately downstream from the IST1 locus. This strain was grown in YNB Complete Synthetic Medium (CSM). Samples were taken every hour and analyzed by Western blot for the presence of Ist1-HA, phospho-eIF2α and Snf7. Furthermore, quantitative RT-PCR was performed to determine the amount of IST1-HA mRNA relative to the control mRNA of the actin gene ACT1 ( ). The results showed growth-dependent levels of Ist1-HA, eIF2α-P and Snf7 similar to the patterns observed in the experiment of , which used plasmid-encoded IST1-HA. In contrast to the Ist1 protein levels, the relative amount of IST1-HA mRNA exhibited only minor changes during growth of the yeast culture, indicating that Ist1 levels are not transcriptionally controlled; rather, they are likely regulated at the level of protein translation and/or degradation.

The rate of Ist1 degradation was determined by adding the translational inhibitor cycloheximide to cells and monitoring Ist1 levels over time by Western blot. The strains used for these experiments were deleted for the gene encoding the multidrug transporter Pdr5 in order to enhance the effect of the proteasomal inhibitor MG132. The results showed that in cells treated with cycloheximide, Ist1-HA was almost completely degraded within two hours; whereas, the control protein Snf7 (ESCRT-III subunit) remained stable during the same time period ( , lanes 1-3). The addition of the proteasomal inhibitor MG132 partially stabilized Ist1-HA ( , lanes 4-6). In contrast, no change in Ist1-HA turnover was observed in cells lacking Pra1 and Prb1, two proteases that play a key role in vacuolar protein degradation ( , lanes 13-15). These results indicated that Ist1-HA is rapidly degraded, and that the proteasome, rather than the MVB pathway, mediates the degradation.

We observed a decreased rate of Ist1-HA degradation in cells deleted for VPS4 ( , lanes 7-9), a mutation that causes the accumulation of Ist1 at the endosome with ESCRT-III (9). This stabilization effect was even more pronounced in vps4Δ cells that were treated with MG132 ( , lanes 10-12). These results suggested that predominantly the soluble pool of Ist1, not the membrane associated pool, is targeted for degradation by the proteasome.

The observed drop in Ist1 levels as cells enter stationary phase ( ) suggested that starvation conditions might result in the loss of Ist1 protein. To test this idea, we followed yeast Ist1-HA levels by Western blot during acute starvation. Starvation conditions were induced by either transferring cells to growth medium lacking leucine or by adding rapamycin, a drug that blocks the kinase activity of the TORC1 complex. In contrast to the non-treated control samples, both leucine starvation and rapamycin addition resulted in the rapid disappearance of Ist1 ( , set 1). This rapid drop in Ist1-HA levels was also observed when Ist1-HA was expressed from the constitutive SNF7 promoter (the promoter driving expression of the control protein Snf7), further support that Ist1 levels are not transcriptionally regulated ( , set 2). In contrast, a fusion of IST1-HA to the 5’ UTR region of CPS1 resulted in stable Ist1-HA protein levels during leucine-starvation ( , set 2). The expression of the vacuolar hydrolase Cps1 is upregulated during starvation conditions, indicating that Cps1 belongs to a set of proteins that is efficiently translated during these stress conditions (18). Together, the results suggested that the observed drop in Ist1 levels during starvation is mainly caused by general translational repression in combination with the intrinsically rapid turnover rate of the protein. We predicted that the starvation-induced drop in Ist1 levels increases the efficiency of the MVB pathway and thus increases the recycling of amino acids required for the stress-response pathways. Consistent with this idea, we observed diminished fitness in starving cells that maintain artificially high levels of Ist1 due to the CPS1 promoter fusion (P(CPS1)-IST1, ).

C-terminally GFP-tagged Ist1 (Ist1-GFP) is a non-functional fusion protein; it does, however, localize properly to MVBs (Dimaano et al., 2008). Ist1-GFP did not exhibit a starvation-induced drop in protein levels ( , set 3). This result suggested that the large, C-terminal GFP domain might interfere with the degradation of Ist1. Interestingly, the C-terminal region of Ist1 contains a cluster of lysine residues that could be targeted for ubiquitination. Rsp5 is the principle ubiquitin ligase acting in endocytosis and the MVB pathway. However, starvation-induced degradation of Ist1 was found to be independent of Rsp5 function (Fig. S3 A). Similarly, mutation of a reported Ist1 C-terminal phosphorylation site (19), showed no effect on the rapid degradation of the protein during starvation (Ist1(S244A), Fig. S3 A). Finally, microscopy of Ist1-GFP-expressing cells showed no obvious change in Ist1 localization during acute starvation (Fig. S3 B). Therefore we found that neither Rsp5, nor phosphorylation of S244 mediates the degradation of Ist1. Also, the degradation is not attributable to a change in Ist1 localization following starvation.

Because the lysine-rich C-terminus of Ist1 contains the Vps4-interaction domain, we tested whether high levels of Vps4 would also stabilize the Ist1 protein. The experiment demonstrated that overexpression of Vps4 (expressed from a 2μ high copy vector) was indeed able to inhibit the starvation-induced degradation of Ist1-HA. In contrast, overexpression of Did2, which binds to the N-terminal region of Ist1, had only a minor effect on the Ist1 degradation rate ( , set 3). In summary, our analysis suggested that soluble, non-Vps4-associated Ist1 was targeted for proteasomal degradation. This degradation, coupled with decreased translation, is the most likely cause for the rapid loss of Ist1 observed following the induction of starvation-response pathways.

To test if protein stability of Vps4 regulators other then Ist1 were affected by starvation, we constructed N-terminal HA-fusions of VPS60, VTA1, DID2 and IST1. All of these fusion proteins were expressed under the control of the SNF7 promoter. These constructs were expressed in wild-type cells and protein levels were followed by Western blot using anti-HA antibodies during one hour of leucine starvation (Fig. S4). The results demonstrated that, similar to the data obtained with the C-terminal HA fusion, HA-Ist1 was rapidly degraded upon induction of starvation. The protein levels of the other HA-tagged regulatory factors also declined, but to a lesser extent than observed for HA-Ist1. The MVB pathway was found to be functional during the initial phase of starvation and thus the observed decline in the levels of Vps60, Did2 and Vta1 were not sufficient to cause MVB sorting defects. Importantly, our data indicated that after prolonged starvation, cells seem to discontinue protein turnover via the MVB pathway and shift to the autophagy pathway. This is discussed in more detail later in this publication. This suggests that the loss of ESCRT regulatory factors might be inconsequential at later time points in starvation.