ESCRT machinery plays a role in microautophagy in yeast

Background Microautophagy, which degrades cargos by direct lysosomal/vacuolar engulfment of cytoplasmic cargos, is promoted after nutrient starvation and the inactivation of target of rapamycin complex 1 (TORC1) protein kinase. In budding yeast, microautophagy has been commonly assessed using processing assays with green fluorescent protein (GFP)-tagged vacuolar membrane proteins, such as Vph1 and Pho8. The endosomal sorting complex required for transport (ESCRT) system is proposed to be required for microautophagy, because degradation of vacuolar membrane protein Vph1 was compromised in ESCRT-defective mutants. However, ESCRT is also critical for the vacuolar sorting of most vacuolar proteins, and hence reexamination of the involvement of ESCRT in microautophagic processes is required. Results Here, we show that the Vph1-GFP processing assay is unsuitable for estimating the involvement of ESCRT in microautophagy, because Vph1-GFP accumulated highly in the prevacuolar class E compartment in ESCRT mutants. In contrast, GFP-Pho8 and Sna4-GFP destined for vacuolar membranes via an alternative adaptor protein-3 (AP-3) pathway, were properly localized on vacuolar membranes in ESCRT-deficient cells. Nevertheless, microautophagic degradation of GFP-Pho8 and Sna4-GFP after TORC1 inactivation was hindered in ESCRT mutants, indicating that ESCRT is indeed required for microautophagy after nutrient starvation and TORC1 inactivation. Conclusions These findings provide evidence for the direct role of ESCRT in microautophagy induction.


Background
In microautophagy, cytoplasmic cargos are directly engulfed by lysosomal/vacuolar membranes, sorted into the vacuolar lumen, and degraded [1][2][3]. Because vacuolar membrane proteins together with vacuolar membranes are degraded in the vacuole in the course of microautophagy, overall microautophagic ux is estimated using green uorescent protein (GFP)-tagged vacuolar transmembrane proteins, Vph1 (a subunit of the vacuolar-ATPase V0 domain) and Pho8 (vacuolar alkaline phosphatase) in the budding yeast Saccharomyces cerevisiae [4]. When Vph1-GFP and GFP-Pho8 are incorporated into the vacuolar lumen by microautophagy, Vph1 and Pho8, but not the stable GFP moiety, are degraded by vacuolar proteases, generating free GFP, which is detectable by immunoblotting. Nutrient starvation or inactivation of target of rapamycin complex 1 (TORC1) protein kinase evokes microautophagy [4][5][6].
The endosomal sorting complex required for transport (ESCRT) system was originally identi ed as being involved in the formation of intraluminal vesicles within multivesicular bodies [7][8][9]. The ESCRT-0 complex binds to ubiquitinated proteins on endosomal membranes and recruits the ESCRT-I and -II complexes, eliciting the assembly of ESCRT-III, which promotes the invagination, constriction, and abscission of endosomal membranes. In microautophagy, carbon starvation after diauxic shift or TORC1 inactivation promotes recruitment of ESCRT-0 onto vacuolar membranes [4,10]. Furthermore, free GFP production from Vph1-GFP after diauxic shift was signi cantly impeded in cells lacking ESCRT-0, -I, -II or -III, inferring a model where ESCRT is required for the deformation of the vacuolar membrane and, thus, also microautophagy [4]. However, most proteins destined for vacuoles pass through late endosomes, named the vacuolar protein sorting (VPS) pathway (or the carboxypeptidase Y pathway), and hence they are trapped in the class E compartment in ESCRT mutants [11,12]. This suggested a possibility that Vph1-GFP is not located properly on vacuolar membranes during microautophagy induction, thereby causing a reduction in the autophagic degradation of Vph1-GFP. Thus, it remains unclear whether or not ESCRT is directly necessary for microautophagic processes on vacuolar membranes.
We found that Vph1 accumulated in the perivacuolar class E compartment in vps27∆ cells defective in ESCRT-0, but we did not further investigate the involvement of ESCRT in sorting of Vph1-GFP and GFP-Pho8 [5]. In addition, we also found that loss of ESCRT compromised microautophagic degradation of GFP-Pho8 after rapamycin treatment, but we did not check whether Pho8 was properly localized on vacuolar membranes in ESCRT mutant cells [10]. Here, we show that Pho8 destined for vacuolar membranes via an alternative adaptor protein-3 (AP-3) pathway, but not Vph1, is correctly targeted to vacuolar membranes even in ESCRT mutants and after TORC1 inactivation. Nevertheless, free GFP production from GFP-Pho8 after TORC1 inactivation was hindered in ESCRT mutants. These ndings demonstrate that the ESCRT machinery is genuinely necessary for microautophagic processes. Another AP-3 pathway-dependent vacuolar membrane protein Sna4 con rmed this idea.

Results
Vph1-GFP is an unsuitable tool to assess the involvement of ESCRT in microautophagy To assess microautophagic ux/activity using a processing assay with a GFP-tagged vacuolar membrane protein, its proper localization in vacuolar membranes is prerequisite. First, we assessed whether Vph1 is correctly localized in cells defective in each ESCRT complex, ESCRT-0 to -III, both in normal (nutrient-rich and TORC1 active) conditions and during microautophagy induction after TORC1 inactivation. We found that ESCRT-lacking mutant cells, with mutations in Vps27 (ESCRT-0), Vps28 (ESCRT-I), Vps36 (ESCRT-II) or Vps24 (ESCRT-III), showed massive accumulation of Vph1-GFP in the perivacuolar class E compartment in normal (nutrient-rich and TORC1 active) conditions ( Fig. 1, control) as described previously [5,[13][14][15][16]. This clearly con rmed that Vph1 is delivered to the vacuolar membrane via the VPS pathway. The aberrant accumulation of Vph1-GFP still remained during microautophagy induction after rapamycin treatment ( Fig. 1, +Rap). Namely, Vph1-GFP did not properly reach the vacuolar surface in ESCRT-de cient cells regardless of TORC1 activity. This demonstrated that the Vph1-GFP processing assay is not suitable to evaluate whether ESCRT is directly implicated in microautophagic processes, although Oku et al. proposed it based on results obtained using this assay [4].
GFP-Pho8 is a suitable marker to evaluate the involvement of ESCRT in microautophagy Next, we similarly assessed GFP-Pho8. In contrast to most vacuolar proteins tra cked via the VPS pathway, a small number of proteins including the vacuolar membrane alkaline phosphatase (ALP) Pho8 are transported directly from the Golgi to the vacuolar surface, even in class E mutants, via the adaptor protein-3 (AP-3) pathway (or the ALP pathway) [14,[17][18][19]. We also observed that GFP-Pho8 was appropriately located on vacuolar membranes in ESCRT mutants in normal conditions (Fig. 2, control). Furthermore, its vacuolar localization was not lost during microautophagy induction after cells were treated with rapamycin ( Fig. 2, +Rap). Thus, GFP-Pho8 is properly localized on the vacuolar membrane in ESCRT mutants regardless of TORC1 activity. We concluded that GFP-Pho8 is a suitable marker to address the involvement of ESCRT in microautophagy ESCRT is required for microautophagy after TORC1 inactivation In a parallel study, we recently reported that free GFP generation from GFP-Pho8 after treatment with rapamycin was massively compromised in cells defective in ESCRT, vps27∆, vps28∆, vps36∆, and vps24∆ cells [10]. Given the proper localization of GFP-Pho8 on vacuolar membranes in these ESCRT mutants, these observations indicated that ESCRT is required for microautophagy itself. To further con rm this we subjected cells to nitrogen starvation, which is a natural condition where TORC1 is inactivated. GFP-Pho8 still showed proper distribution on the vacuolar surface after nitrogen starvation even in ESCRT mutants (Fig. 3a). Autophagic degradation of GFP-Pho8 after nitrogen starvation was mildly reduced in these ESCRT mutants (Fig. 3b) (see "Discussion"). These ndings con rmed the idea that ESCRT is indeed required for proper microautophagic induction after TORC1 inactivation.
Search for other appropriate markers of microautophagy Thus, proteins destined for vacuolar membranes via the AP-3 pathway (the ALP pathway) have potential as appropriate markers to investigate ESCRT-mediated microautophagy. We assessed this using the following vacuolar membrane proteins targeted via the AP-3 pathway: Nyv1 (vacuolar v-SNARE), Yck3 (casein kinase) and Sna4 (protein of unknown function) [20][21][22]. We found that Nyv1-GFP was not clearly distributed on vacuolar membranes even in wild-type cells (Additional le 1: Fig. 1a). By contrast, clear localization of Yck3-GFP on vacuolar membranes was found in wild-type cells in the absence or presence of rapamycin (Additional le 1: Figs. S1a, b). However, unlike Vph1-GFP and GFP-Pho8, Yck3-GFP did not generate free GFP after rapamycin treatment, although Yck3-GFP was almost lost (Additional le 1: Fig.  S1c). This suggested that Yck3 is degraded outside vacuoles (see "Discussion"). Thus, Nyv1-GFP and Yck3-GFP were not suitable for the assessment microautophagic ux. Finally, we found that Sna4-GFP was located on vacuolar membranes in wild-type cells in the absence and presence of rapamycin and that similar vacuolar localization was maintained in ESCRT mutants, although Sna4-GFP signals on the vacuolar membranes were largely lost after rapamycin treatment (Fig. 4, Additional le 1: Fig. S2). Sna4-GFP produced free GFP in wild-type cells after rapamycin treatment (Fig. 5, see also Additional le 1: Fig.  S3), which was repressed in vps27∆ cells, similar to GFP-Pho8 [10]. In addition, free GFP generation from Sna4-GFP after rapamycin treatment was also reduced in vps28∆, vps36∆, and vps24∆ cells, although uncharacterized protein bands accumulated additionally in wild-type and the mutant cells. These ndings con rmed the hypothesis that ESCRT is required for proper microautophagy induction.

Discussion
In this study, we showed that GFP-Pho8 and Sna4-GFP are suitable tools to evaluate the involvement of ESCRT in microautophagic processes on vacuolar membranes and that ESCRT is genuinely required for this process after TORC1 inactivation. In contrast, Vph1-GFP is not suitable for use in this assay, because Vph1-GFP was not correctly localized on the vacuolar membranes in ESCRT mutants.
We tested whether other vacuolar membrane proteins sorted via the AP-3 pathway (Nyv1 and Yck3) are suitable as markers of microautophagic ux. However, they were found not to be useable markers for microautophagy assays. In the course of this study, we noticed that the vacuolar membrane protein Yck3 was not degraded in the vacuole after TORC1 inactivation, although microautophagy is induced: there was no accumulation of free GFP (a marker of vacuolar degradation) from Yck3-GFP after rapamycin treatment, although Yck3-GFP was lost. It is plausible that Yck3 is degraded by proteasomes in the cytoplasm after TORC1 inactivation. The armadillo repeat protein Vac8 is localized on the vacuolar membrane, and has multiple functions including vacuole inheritance [23,24]. No free GFP production from Vac8-GFP appeared after rapamycin treatment (Morshed et al., unpublished data). These ndings indicated that some, but not all, vacuolar membrane proteins are degraded by microautophagy after TORC1 inactivation. It is an interesting open question how some vacuolar membrane proteins escape from microautophagic degradation after TORC1 inactivation. Furthermore, Sna4-GFP was lost with no clear free GFP accumulation after rapamycin treatment in vps27∆ cells (Fig. 5), indicating that Sna4 was degraded outside vacuoles in speci c situations. These ndings imply that some vacuolar membrane proteins are degraded by multiple pathways after TORC1 inactivation.
In the case of nitrogen starvation, loss of ESCRT mildly reduced nitrogen starvation-induced microautophagy ( Figure 3B). This suggested a possibility that ESCRT-independent microautophagy is promoted after nitrogen starvation. Alternatively, a microautophagy-independent pathway might mediate degradation of vacuolar membrane proteins under these conditions. It has been reported that ESCRTindependent homotypic vacuole-vacuole fusion mediates degradation of vacuolar membrane proteins (the intraluminal fragment pathway) [25].
This study concluded that ESCRT is indeed involved in microautophagic processes in budding yeast.
However, the molecular mechanism of ESCRT-mediated microautophagy is still largely unknown. This study provides a caution for the selection of markers to assess microautophagic ux. To evaluate microautophagic activity in mutants using GFP-tagged vacuolar membrane proteins, con rmation of their proper localization on vacuolar membranes is prerequisite. It is suggested that microautophagy is involved in protein homeostasis in mammalian cells [26]. However, the molecular mode and physiological roles of mammalian microautophagy are still largely unknown [27,28]. This study is helpful for dissection of the mechanism and physiological meaning of microautophagy not only in budding yeast but also other organisms including humans.

Conclusions
Microautophagic degradation of GFP-Pho8 and Sna4-GFP after TORC1 inactivation was hindered in ESCRT-de cient cells, although they were properly sorted onto vacuolar membranes. This fact indicates that ESCRT is indeed required for microautophagy after nutrient starvation and TORC1 inactivation.

Strains and media
The S. cerevisiae strains and plasmids are listed in Additional les: Tables S1 and S2, respectively. Glucose-containing YPAD (YPD containing 0.01% adenine) and synthetic minimal medium (SD) complemented with the appropriate nutrients for plasmid maintenance were prepared using standard methods. For assessment of microautophagy after rapamycin treatment, when cells harbored plasmids, cells were precultured in SD medium with the appropriate nutrients, and then cultured in YPAD, followed by rapamycin treatment. For nitrogen-starvation experiments, cells were transferred into SD-N medium without ammonium sulfate.

Western blotting analysis
Exponentially growing cells were used and proteins were extracted using a post-alkaline extraction method in accordance with a previous report [29]. We used the following antibodies: anti-GFP mouse monoclonal antibody (Santa Cruz, Cat#sc-9996; RRID: AB_627695) and anti-Pgk1 mouse monoclonal antibody (Abcam, Cat# ab113687; RRID: AB_10861977). All western blotting experiments were performed independently at least three times to con rm the reproducibility of the results. Relative protein amounts were measured using ImageJ software. To assess microautophagic ux, free GFP processed from of GFP-Pho8 after rapamycin treatment or nitrogen starvation was measured and quanti ed by calculating the ratio of cleaved free GFP versus the sum of uncleaved GFP-fused proteins and free GFP, and the values relative to that in wild-type cells are shown, as described previously [30]. In the case of Sna4-GFP, uncleaved Sna4-GFP was almost lost after rapamycin treatment. Therefore, to assess microautophagic ux using Sna4-GFP, the ratios of cleaved free GFP versus Pgk1 were calculated, and the values relative to that in wild-type cells are shown. The average and standard deviation were determined for each sample from three independent experiments, and relative values normalized against the value in control cells are shown. For group comparison, data analysis was performed with the software GraphPad Prism (v. 8.1.2) followed by two-way ANOVA (or mixed model) with Bonferroni correction for multiple comparisons in statistical hypothesis testing.

Microscope observations
Exponentially growing cells were used for experiments. Cells were stained with the vacuolar membrane dye FM4-64 (10 ng/ml) for 1 h, washed to remove free dye and further incubated at least for 1.5 h prior to treatment with rapamycin or nitrogen starvation. Cell, GFP, and FM4-64 images were captured using a Carl Zeiss Axio Imager M1 microscope with a cooled CCD camera (Carl Zeiss AxioCam MRm) [10]. All microscope observations were performed independently at least three times to con rm the reproducibility of the results. For quanti cation of intensities of Sna4-GFP signals on the vacuolar membrane, we used a microscopy-equipped software (Carl Zeiss AxioVision) according to a previous study [31].

Statistical analysis
Statistical analysis was performed using Microsoft Excel and GraphPad Prism 8 software.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
Data and materials used in the current study are available from the corresponding authors on request.

Acknowledgments
We thank Muhammad Arifur Rahman for providing preliminary ndings and laboratory members of TU for helpful discussions.

Funding
This work was supported in part by the Japan Society for the Promotion of Science (JSPS) (grant No. 18K06212 and 20H05317) and Institute for Fermentation, Osaka (IFO) (grant No. 2020-2-055). The funding bodies played no role in the design of the study, collection, analysis, interpretation of data, or writing of the manuscript.
Authors' contributions TU designed the project and experiments. SM mainly conducted the experiments. MNT conducted some experiments. SM and MNT analyzed the data. TU mainly wrote the paper and SM helped to write the paper. All authors read and approved the nal manuscript.