Roles for L _o microdomains and ESCRT in ER stress-induced lipid droplet microautophagy in budding yeast

Abstract

Microlipophagy (µLP), degradation of lipid droplets (LDs) by microautophagy, occurs by autophagosome-independent direct uptake of LDs at lysosomes/vacuoles in response to nutrient limitations and ER stressors in Saccharomyces cerevisiae. In nutrient-limited yeast, liquid-ordered (L_o) microdomains, sterol-rich raftlike regions in vacuolar membranes, are sites of membrane invagination during LD uptake. The endosome sorting complex required for transport (ESCRT) is required for sterol transport during L_o formation under these conditions. However, ESCRT has been implicated in mediating membrane invagination during µLP induced by ER stressors or the diauxic shift from glycolysis- to respiration-driven growth. Here we report that ER stress induced by lipid imbalance and other stressors induces L_o microdomain formation. This process is ESCRT independent and dependent on Niemann-Pick type C sterol transfer proteins. Inhibition of ESCRT or L_o microdomain formation partially inhibits lipid imbalance-induced µLP, while inhibition of both blocks this µLP. Finally, although the ER stressors dithiothreitol or tunicamycin induce L_o microdomains, µLP in response to these stressors is ESCRT dependent and L_o microdomain independent. Our findings reveal that L_o microdomain formation is a yeast stress response, and stress-induced L_o microdomain formation occurs by stressor-specific mechanisms. Moreover, ESCRT and L_o microdomains play functionally distinct roles in LD uptake during stress-induced µLP.

INTRODUCTION

Lipid droplets (LDs), organelles that form at and bud from ER membranes, consist of a phospholipid monolayer with associated proteins surrounding a core of neutral lipids including triacylglycerol and sterol esters. LDs have well-established functions in storage of lipids that are used for energy production or for synthesis of membrane components, signaling molecules, and other macromolecules. Storage of lipids within LDs also protects cells from exposure to excess free fatty acids or sterols (Kounakis et al., 2019). Finally, recent studies revealed a role for LDs in protein quality control (Welte and Gould, 2017; Garcia et al., 2018). Indeed, studies from our laboratory indicate that LDs function in removing unfolded proteins that accumulate in ER during ER stress and targeting those proteins for autophagic degradation by the lysosome (vacuole in yeast) (Vevea et al., 2015; Garcia et al., 2020). The central role of LD function in cellular fitness is evident from diseases such as obesity, nonalcohol fatty liver disease, cardiovascular disease, neutral lipid storage disease, lipodystrophy, and hereditary spastic paraplegia that are associated with LD dysregulation (Olzmann and Carvalho, 2019).

LDs are taken up into vacuoles for degradation by multiple autophagic pathways, referred to collectively as lipophagy. During macroautophagy-driven lipophagy, LDs are engulfed within autophagosomes, which fuse with and are degraded by lysosomes/vacuoles (Singh et al., 2009). In contrast, during microautophagy-driven lipophagy (microlipophagy, µLP), LDs undergo autophagosome-independent direct uptake into the lysosomes/vacuoles. In yeast, µLP is critical for mobilizing lipids for ATP production under nutrient-limited conditions, degrading excess lipids under conditions of lipid imbalance, degrading misfolded or aggregated proteins that are sequestered to LDs and targeting lipids and sterols to the vacuolar membranes (van Zutphen et al., 2014; Wang et al., 2014; Vevea et al., 2015; Oku et al., 2017; Seo et al., 2017; Tsuji et al., 2017; Garcia et al., 2020). Recent studies revealed that µLP occurs in hepatocytes in response to nutrient limitation. As in yeast, hepatocyte µLP involves autophagosome-independent interaction of LDs with lysosomes and uptake of LDs into lysosomes at sites of invagination in the lysosomal membrane (Schulze et al., 2020).

The mechanism of LD uptake into lysosomes in mammalian systems is not well understood. However, studies in yeast revealed two possible mechanisms for vacuolar membrane invagination during µLP. During transition into stationary phase or nitrogen starvation, LD uptake occurs at liquid-ordered (L_o) microdomains in the vacuolar membrane (Wang et al., 2014; Tsuji et al., 2017). L_o microdomains are specialized raftlike regions that are enriched in sterols and saturated lipids and specific proteins and coexist with liquid-disordered (L_d) domains. They are sites for membrane bending and invagination (Simons and Ikonen, 1997; Tsuji and Fujimoto, 2018).

L_o microdomains of the yeast vacuolar membrane are larger and more stable than other rafts (Toulmay and Prinz, 2013). Ultrastructural studies in yeast reveal that entry into stationary phase or nutrient limitation induces vacuolar L_o microdomain formation and roles for multivesicular bodies (MVBs, late endosomes that contain sterol-rich intraluminal vesicles) and Niemann-Pick type C (NPC) sterol transporter proteins, Ncr1 and Npc2, in this process. Specifically, MVBs mediate transport of sterols to the vacuole while NPC proteins transfer sterols from the lumen to membrane of vacuoles during L_o microdomain formation and expansion. These studies also find that L_o microdomains are a site for LD uptake into the vacuole during µLP, and that MVBs and NPC proteins are also required for this LD uptake (Tsuji et al., 2017).

Studies from our laboratory and others revealed an alternative mechanism for LD uptake at the vacuole during µLP in yeast. This form of µLP depends on endosome sorting complex required for transport (ESCRT) and occurs in response to ER stressors or the diauxic shift from glycolysis- to respiration-driven growth (Vevea et al., 2015; Oku et al., 2017; Garcia et al., 2020). Deletion of ESCRT components blocks µLP induced by these conditions. Moreover, ESCRT components are recruited to the vacuolar membrane and localize to sites of vacuolar membrane invagination and scission during LD uptake into the organelle (Vevea et al., 2015; Oku et al., 2017; Garcia et al., 2020). These findings support the model that ESCRT plays a direct role in µLP in response to ER stressors or during the diauxic shift by mediating membrane invagination and scission during LD uptake into the vacuole. However, whether L_o microdomain-dependent and ESCRT-dependent mechanisms interact is still not clear.

Here we obtained additional evidence that L_o microdomain biogenesis is induced by multiple stressors in yeast and may be a general stress response. Moreover, we find that L_o microdomain formation occurs by different mechanisms in growing or stressed yeast compared with nutrient-limited yeast. Finally, our studies indicate that L_o microdomains and ESCRT have independent functions in LD uptake at the vacuolar membrane and that both mechanisms are active in µLP in yeast exposed to lipid imbalance.

RESULTS AND DISCUSSION

Lipid imbalance results in alteration of the redox state of ER

The ER houses key enzymes in all lipid biosynthetic pathways and is the site for trafficking and folding of 30% of the proteins in eukaryotic cells. Studies from our laboratory and others revealed that lipid imbalance, and more specifically, alterations in the ratio of phosphatidylcholine (PC) to phosphatidylethanolamine (PE), induces ER stress (Thibault et al., 2012; Surma et al., 2013; Vevea et al., 2015). Since PC is a cylindrical lipid that stabilizes lipid bilayers and PE is a conical, bilayer-destabilizing lipid, PC:PE imbalance can induce ER stress through effects on lipid bilayer stress (Cullis et al., 1996). On the other hand, we found that similar proteins are removed from the ER by LDs and targeted for degradation in response to lipid imbalance and treatment with dithiothreitol (DTT), a reducing agent that induces ER stress (Garcia et al., 2020). This finding raises the possibility that lipid imbalance may also induce ER stress through effects on oxidative protein folding in the ER. We tested this hypothesis using an ER-targeted, redox-sensing GFP (eroGFP) (Merksamer et al., 2008).

Here, lipid imbalance was induced using a model system described previously (Vevea et al., 2015). CHO2 encodes a phosphatidyl methyltransferase that catalyzes the first step in the conversion of PE to PC during de novo PC biosynthesis. In cho2∆ cells, PC can also be produced from exogenous choline (cho2∆^+C1). However, withdrawal of choline from the medium for 1 d (cho2∆^–C1) results in defects in PC biosynthesis, severe imbalance in the PC:PE ratio, and activation of the unfolded protein response (UPR), LD biogenesis and µLP (Vevea et al., 2015).

We confirmed that the ER lumen is an oxidizing environment. Indeed, treatment with the oxidizing agent H₂O₂ did not cause a detectable change in the redox state of the ER lumen using the eroGFP indicator (Figure 1, A and B). In contrast, the ER lumen became 3.6-fold more reduced on treatment with DTT. We also found that lipid imbalance rendered the ER lumen 1.4-fold more reduced compared with that observed in nonstressed control cells (Figure 1). These findings support the model that the accumulation of unfolded proteins and UPR activation in response to lipid stress may be due in part to defects in oxidative folding in the ER.

FIGURE 1:

Lipid imbalance results in alteration of the redox state of ER. (A) Representative images of redox state of ER, visualized with eroGFP, in WT cells grown in SC and treated for 30 min with 5 mM DTT or 5 mM H₂O₂. Reduced, oxidized, and reduced:oxidized eroGFP ratio images are shown. Color scale in the bottom panel shows the dynamic range of ratios, with warmer colors indicating a more reducing environment. Bar, 2 μm. (B) Quantification of reduced:oxidized eroGFP ratios as shown in A. The box indicates the middle quartiles with the horizontal representing the median; whiskers show the 10th and 90th percentile, and red dots represent values in the top and bottom 10th percentiles. Representative trial from three independent experiments (n = 29–49 for each condition, ****p < 0.0001, by one-way ANOVA with Bonferroni’s post-hoc test for multiple comparisons). (C) Representative images of redox state of ER in cho2Δ^+C1 (normal lipid levels) and cho2Δ^–C1 (lipid-stressed) cells visualized with eroGFP. Bar, 2 μm. (D) Quantification of reduced to oxidized eroGFP ratios in cho2Δ^+C1 and cho2Δ^–C1 cells as shown in C. Representative trial from three independent experiments (n = 35–38 for each condition, ****p < 0.0001, by unpaired two-tailed Student’s t test).

Lipid imbalance and chemically induced ER stress induce vacuolar L _o microdomain formation

Our previous studies revealed that µLP is induced by all ER stressors studied (Vevea et al., 2015; Garcia et al., 2020); µLP in stationary-phase and nitrogen-starved yeast cells requires vacuolar L_o microdomains (Wang et al., 2014; Tsuji et al., 2017). However, whether a similar mechanism occurs during µLP in response to ER stress is still unknown. Here we tested the effect of ER stressors, including lipid imbalance, DTT, and tunicamycin (TM)-mediated inhibition of glycosylation of protein in the ER, on vacuolar L_o microdomains.

We confirmed that Vph1p and Ivy1p are effective for visualization of different types of L_o microdomains in yeast entering stationary phase (Martinez-Munoz and Kane, 2008; Dawaliby and Mayer, 2010; Toulmay and Prinz, 2013). Vph1p, a subunit of the vacuolar ATPase, localizes to vacuolar L_d domains and is excluded from L_o microdomains. In contrast, the phospholipid-binding inverse BAR protein Ivy1p localizes to punctate structures in L_o microdomains, where it mediates vacuolar membrane invagination during µLP (Toulmay and Prinz, 2013; Numrich et al., 2015). Ivy1p also associates with components of the TORC1-regulating EGO complex and itself regulates the TORC1 complex (Numrich et al., 2015). Since Ivy1p has functions that may be L_o independent, we scored L_o microdomains as sites in the vacuolar membrane where Vph1p is excluded and Ivy1p is enriched.

L_o microdomains are classified into three groups defined by the shape and size of areas where Vph1 is excluded: type I L_o microdomains constitute large regions in the vacuolar membrane where Vph1p is excluded, type II L_o microdomains contain multiple smaller semisymmetrical Vph1p-free structures, and type III L_o microdomains contain small circular Vph1p-free regions that are similar in diameter and regularly spaced within the vacuolar membrane (Supplemental Figure S1A). Ivy1p localizes to all of these microdomains (Supplemental Figure S1B). As expected, all three types of L_o microdomains were evident in all cells within 1 to 2 d after entry into stationary phase and type II and III L_o microdomains in ∼60% of cells during entry into stationary phase (Supplemental Figure S1, C and D). We also confirmed a significant increase in the number of Ivy1p puncta in yeast 2 d after entry into stationary phase (Supplemental Figure S1E) (Toulmay and Prinz, 2013; Wang et al., 2014).

Next, we tested whether L_o microdomains form in midlog phase yeast that are exposed to lipid imbalance-, DTT-, or TM-induced ER stress. We detected regions in the vacuolar membrane that were Vph1p-free in some midlog phase cells (type I L_o microdomains). These regions are possible sites of contact of vacuoles with organelles including nuclei (at nuclear–vacuolar junctions, NVJs) and mitochondria (at vCLAMPs) (Kane, 2006; Martinez-Munoz and Kane, 2008; Dawaliby and Mayer, 2010; Takatori et al., 2016). Using Vph1p exclusion as a marker, we detected an increase in the formation of type I and II L_o microdomains in yeast exposed to lipid imbalance- or DTT-induced ER stress (Figure 2, A, B, D, and E). However, L_o microdomain formation occurred to a lesser extent in ER-stressed yeast compared with that observed during entry into stationary phase (Figure 2, B and E, and Supplemental Figure S1D). Quantification of Vph1p exclusion in TM-treated yeast is difficult because vacuoles are highly fragmented, and the vacuolar membrane is not well resolved at sites of contact or overlap of fragmented vacuoles. Nonetheless, we observed a mild but insignificant increase in type I L_o microdomains in response to TM-induced ER stress (Supplemental Figure S1, F and G). We also detected a statistically significant increase in the number of Ivy1p puncta in vacuoles on exposure to any of the 3 ER stressors (Figure 2, C and F, and Supplemental Figure S1H). Thus we obtained evidence that lipid imbalance- and chemically induced ER stress both stimulate L_o microdomain formation.

FIGURE 2:

Lipid imbalance- and DTT-induced ER stress increase L_o microdomain formation. (A, D) Representative images of vacuoles with L_o domains in cho2Δ^+C1, cho2Δ^–C1cells (A), or WT cells grown in SC or SC + 5 mM DTT for 8 h (D). Cells express Ivy1p-GFPEnvy (red) and Vph1p-mCherry (gray). Mid, single optical section through the middle of the cell; Top, single optical section across the top of the vacuole; HC, high contrast. Bar, 2 μm. (B, E) Percentage of cells with vacuoles showing no L_o microdomains (ND) or L_o microdomain type I, II, or III in cho2Δ^+C1 and cho2Δ^–C1cells (B); WT cells grown in SC or SC + 5 mM DTT for 8 h (E). Bars represent average + SEM from three independent trials (n > 60 cells for each condition per trial. *p < 0.05; **p < 0.01; ***p < 0.001, by unpaired two-tailed Student’s t test). (C, F) Quantification of number of Ivy1p puncta per cell in cho2Δ^+C1, cho2Δ^–C1(C) and WT cells in SC or SC + 5 mM DTT for 8 h (F). Representative trial from three independent trials (n > 40 cells per conditions per trial. ****p < 0.0001 by unpaired two-tailed Mann–Whitney test).

L_o microdomains were originally identified in yeast exposed to short-term glucose starvation or treatment with cycloheximide or weak acids (Toulmay and Prinz, 2013). Other studies reveal that nitrogen limitation or transition into stationary phase induces L_o microdomain formation (van Zutphen et al., 2014; Wang et al., 2014; Tsuji et al., 2017). Interestingly, hypertonic stress induces L_o microdomain formation at contact sites between vacuoles and nuclei, ER, mitochondria, or other vacuoles (Takatori et al., 2016). Our finding that lipid imbalance and other ER stressors can induce L_o microdomain formation in yeast (Figure 2) provides additional evidence that vacuolar L_o microdomain formation is a general stress response.

Interestingly, L_o microdomain formation in response to ER stress and during transition to stationary phase may occur by distinct mechanisms. We detected an increase in all types of L_o microdomains in yeast transitioning into stationary phase (Supplemental Figure S1, C and D). In contrast, ER stress induced type I and II L_o microdomains in vacuoles, but no obvious type III microdomains (Figure 2, A, B, D, and E). In addition, the level of vacuolar L_o microdomains observed in response to ER stress is lower than that observed in stationary phase cells. It is possible that different types of L_o microdomains in vacuoles reflect maturation of those structures. However, it is also possible that there are functionally distinct L_o microdomain populations, reflected in part by microdomain type. These models are not mutually exclusive.

Impact of NPC and ESCRT proteins on ER stress-induced vacuolar L _o microdomain formation

During nutrient limitation-induced L_o microdomain formation, sterols are transferred to the vacuolar lumen by fusion of particles including MVBs, AP-3 vesicles, autophagosomes, or LDs with vacuoles or by lipid transfer from other organelles at contact sites including NVJs or vCLAMPs (Tsuji and Fujimoto, 2018). Sterols within vacuoles are transferred to the vacuolar membrane for L_o microdomain formation by NPC proteins in stationary phase and nitrogen-limited yeast (Tsuji et al., 2017). Here we studied whether NPC proteins and ESCRT, which mediates MVB formation, play roles in ER stress-induced L_o microdomain formation.

We found that deletion of NCR1 and NPC2 had no obvious effect on type I L_o microdomains in midlog phase yeast. In contrast, L_o microdomain formation in response to the ER stressors studied was partially inhibited by deletion of NPC proteins (Figure 3, A–D). Moreover, we detected a subtle reduction in type I L_o microdomains in TM-treated ncr1∆ npc2∆ yeast compared with wild-type (WT) yeast (Supplemental Figure S2, A and B). Finally, the increase in the number of Ivy1p puncta in vacuoles induced by all three ER stressors was also partially inhibited by deletion of NPC proteins (Supplemental Figure S2, C–E). Thus deletion of NPC proteins inhibits but does not block L_o microdomain formation in response to the ER stressors studied. This observation is consistent with previous findings that deletion of NPC proteins results in partial inhibition of L_o formation during transition into stationary phase or nitrogen starvation (Tsuji et al., 2017).

FIGURE 3:

L_o microdomain formation under lipid imbalance- and DTT-induced ER stress is dependent on NPC proteins. (A, C) Representative images of vacuoles with L_o domains in cho2Δ^+C1, cho2Δ^–C1cells in the presence or absence of NPC2 and NCR1 (A), or WT cells in the presence or absence of NPC2 and NCR1 grown in SC or SC + 5 mM DTT for 8 h (C). Cells express Ivy1p-GFPEnvy (red) and Vph1p-mCherry (gray). Mid, single optical section through the middle of the cell; Top, single optical section across the top of the vacuole; HC, high contrast. Bar, 2 μm. (B, D) Percentage of cells with vacuoles showing no L_o microdomains or L_o microdomain type I, II, or III as shown in A and C, respectively. Bars represent average + SEM from three independent trials (n > 60 cells for each condition per trial. *p < 0.05; ***p < 0.001; ****p < 0.0001, by one-way ANOVA with Sidak’s multiple comparisons test).

We used fenpropimorph, an antifungal agent that inhibits enzymes in the ergosterol biosynthesis (Marcireau et al., 1990), to assess the role of sterols in ER stress-induced L_o microdomain formation. We confirmed previous findings that fenpropimorph treatment effectively inhibits L_o microdomain formation in yeast during transition into stationary phase (Toulmay and Prinz, 2013). We also found that fenpropimorph compromised ER stress-induced L_o microdomain formation (Supplemental Figure S2, F and G). However, the inhibition of L_o microdomain formation observed during ER stress was subtle and less severe than that observed on entry into stationary phase or in ncr1∆ npc2∆ cells during ER stress or stationary phase. Thus, while sterol transfer is critical for L_o microdomain formation under all conditions studied, sterol biogenesis is a minor contributor to ER stress-induced L_o microdomain formation.

Next, we tested whether ESCRT is required for L_o microdomain formation in midlog phase cells and in cells exposed to ER stressors. We found that deletion of the ESCRT III component SNF7 did not inhibit L_o microdomain formation in midlog phase yeast (Figure 4, A–D and Supplemental S3, A and B). Indeed, deletion of SNF7 results in a subtle but statistically significant increase in exclusion of Vph1p at type I vacuolar L_o microdomains in midlog phase (Figure 4D and Supplemental Figure S3B). Moreover, deletion of SNF7 or the ESCRT protein Vps4p did not affect L_o microdomain formation in cells challenged with DTT-, TM-, or lipid imbalance-induced ER stress (Figure 4, A–D and Supplemental Figure S3, A–D). Consistent with this, deletion of VPS4 or SNF7 had no effect on the localization of Npc2p to the vacuolar lumen or of Ncr1p to the vacuolar membrane under stress conditions (Figure 4, E–H and Supplemental Figure S3, E–H). Thus we found that ESCRT is not required for L_o microdomain formation in midlog phase cells or in yeast exposed to ER stressors.

FIGURE 4:

L_o microdomain formation under lipid imbalance- and DTT-induced ER stress is ESCRT-independent. (A, C) Representative images of vacuoles with L_o domains in cho2Δ^+C1, cho2Δ^–C1cells in the presence or absence of SNF7 (A), or WT cells in the presence or absence of SNF7 grown in SC or SC + 5 mM DTT for 8 h (C). Cells express Ivy1p-GFPEnvy (red) and Vph1p-mCherry (gray). Mid, single optical section through the middle of the cell; Top, single optical section across the top of the vacuole; HC, high contrast. Bar, 2 μm. (B, D) Percentage of cells with vacuoles showing no L_o microdomains (ND) or L_o microdomain type I, II, or III as shown in A and C, respectively. Bars represent average + SEM from three independent trials (n > 60 cells for each condition per trial. *p < 0.05; ***p < 0.001; ****p < 0.0001; ns, no significance; by unpaired one-way ANOVA with Sidak’s multiple comparisons test). (E, G) Representative images of cho2Δ and vps4Δcho2Δ cells tagged with Npc2p-GFPEnvy (E) or Ncr1p-GFPEnvy (G) and Vph1p-mCherry grown with (cho2Δ^+C1) or without (cho2Δ^–C1) 1 mM choline for 24 h. Images are single optical sections through the middle of the cell. Arrows show no lumen structures in (E) or no rim structures in (G). Bar, 2 μm. (F, H) Quantification of Npc2p (F) and Ncr1 (H) localization in vacuoles from images in (E) and (G), respectively. Graph shows average + SEM of the total percentage of vacuoles that show Npc2p (F) or Ncr1p (H) inside the vacuole (lumen), at the rim, in puncta, or absent from each vacuole. (n = 3, **p < 0.01; ***p < 0.001; ns, no significance by one-way ANOVA with Bonferroni’s post-hoc test for multiple comparisons).

Our studies revealed that mechanisms underlying L_o microdomain formation during stationary phase and midlog phase/ER stress are distinct. NPC proteins, presumably through their function in sterol transfer within the vacuole, are required for L_o microdomain formation in all cases. However, the source for sterols appears to be distinct during stationary phase compared with the other conditions studied. Previous studies indicate that ergosterol biosynthesis and ESCRT function in MVB-mediated transport of sterols to vacuoles are critical for L_o microdomain formation in stationary-phase yeast (Toulmay and Prinz, 2013; Tsuji et al., 2017). In contrast, we find that inhibition of ergosterol biosynthesis has only minor effects and inhibition of ESCRT has no obvious effect on L_o microdomain formation in midlog phase or under ER stress. This finding is also supported by previous findings that treatment of vacuoles with methyl-β-cyclodextrin, an agent that extracts sterols from membranes, results in an increase in L_o microdomains in vacuoles in midlog phase yeast but reduces L_o microdomains in vacuoles in stationary-phase yeast (Toulmay and Prinz, 2013).

Differential roles for L _o microdomains and ESCRT in stress-induced µLP

Finally, we studied the relative contribution of L_o microdomains and ESCRT in vacuolar uptake of LDs during µLP. We observed an increase in the percentage of cells containing LDs within vacuoles in response to lipid or ER stress, which provides additional support for the notion that lipid or ER stress induces lipophagy (Supplemental Figure S4, A–F). We also used an established Western blot-based assay to measure lipophagy (Klionsky et al., 2012; Vevea et al., 2015). We tagged the LD marker protein Erg6p at its chromosomal locus with mCherry and carried out quantitative analysis of the degradation of Erg6p-mCherry to free mCherry. Using this assay, we found that lipid imbalance-, DTT-, or TM-induced ER stress induces Erg6p degradation and that this process is dependent on vacuolar proteases and does not require ATG genes (e.g., Figure 5, A–D and Supplemental Figure S4, G and H) (Vevea et al., 2015; Garcia et al., 2020). Thus, Erg6p degradation under the conditions studied is a sound readout for µLP.

FIGURE 5:

ESCRT and L_o microdomains have differential roles in µLP in response to lipid imbalance or DTT-induced ER stress. (A, C) Representative Western blots of Erg6p-mCherry in cho2Δ, npc2Δncr1Δcho2Δ, snf7Δcho2Δ, and npc2Δncr1Δsnf7Δcho2Δ cells grown with or without 1 mM choline (Cho) for 24 h (A), or in WT, npc2Δncr1Δ, and snf7Δ cells grown in SC, SC + 5 mM DTT for 8 h (C). (B, D) Quantification of vacuolar degradation of Erg6p-mCherry from Western blots in A and C. Bar graph shows average + SEM of total intensity of free mCherry bands normalized to TCE for each lane and to cho2Δ^+C1or WT. (n = 15 and n = 9 independent trials in F and H, respectively. *p < 0.05, **p <0.01, ***p < 0.001, and ****p < 0.0001 by one-way ANOVA with Bonferroni’s post-hoc test for multiple comparisons). (E) Schematic of the role of L_o microdomains and ESCRT in µLP under DTT-, TM-, or lipid imbalance-induced ER stress.

We found that deletion of NPC proteins had no detectable effect on DTT- or TM-induced Erg6p degradation. In contrast, inhibition of ESCRT (by deletion of SNF7) blocked DTT- or TM-induced Erg6p degradation (Figure 5, A and B and Supplemental Figure S4, G and H ). Since L_o microdomain formation requires NPC proteins during ER stress, our findings indicate that µLP in response to DTT- or TM-induced ER stress is not dependent on L_o microdomains. Moreover, we find that L_o microdomain formation during ER stress does not require ESCRT (Figure 4, A–D and Supplemental Figure S3, A–D). Since ESCRT localizes to sites of vacuolar membrane invagination and scission during ER stress-induced µLP (Garcia et al., 2020), our findings support the model that ESCRT functions in DTT- or TM-induced µLP through effects on vacuolar membrane invagination and not through effects on L_o microdomain formation. Finally, we detected a statistically significant decrease in lipid imbalance-induced µLP in ncr1∆ npc2∆ cells (Figure 5, C and D) or in snf7∆ cells and a complete block in this process in snf7∆ ncr1∆ npc2∆ cells (Figure 5, C and D). These data support the model that vacuolar membrane invagination during µLP occurs by both L_o microdomain- and ESCRT-dependent mechanisms in response to lipid imbalance.

Overall, we find that ER stress induced by lipid imbalance and other stressors induces L_o microdomain formation, which is ESCRT independent and dependent on NPC proteins. Although all stressors can induce L_o microdomains, the contribution of L_o microdomains is different in response to different stressors. During DTT- or TM-induced ER stress, µLP is fully dependent on ESCRT and independent of NPC proteins. In contrast, LD uptake during lipid stress-induced µLP is both ESCRT and L_o microdomain dependent (Figure 5E). Our findings support the model that L_o microdomain formation is a general stress response that occurs by distinct stressor-specific mechanisms. They also indicate that ESCRT and L_o microdomains play functionally distinct roles in LD uptake during stress-induced µLP.

Previous studies raise the possibility that there are functionally distinct types of L_o microdomains. In midlog phase cells, Vph1p is excluded from regions in the vacuolar membrane (type I L_o microdomains) at NVJs and vCLAMPs (Kane, 2006; Martinez-Munoz and Kane, 2008; Dawaliby and Mayer, 2010; Takatori et al., 2016). On the other hand, type III L_o microdomains are the primary sites for vacuolar membrane invagination during LD uptake in µLP induced by nutrient limitation (Tsuji et al., 2017). We find that L_o microdomains contribute to lipid imbalance-induced LD uptake. Since type I and II L_o microdomains are the only microdomains that are detected under these conditions, our findings imply that LD uptake occurs at I or II L_o microdomains and provide additional support for the idea that the types of L_o microdomains are functionally distinct. Further studies may reveal the function of and markers for different populations of L_o microdomains.

It is not clear why LD uptake during µLP occurs under lipid imbalance-induced ER stress by two independent mechanisms but is entirely ESCRT dependent under DTT- or TM-induced ER stress. It is possible that lipid imbalance results in changes in the lipid composition of the vacuolar or LD membrane, which affect L_o microdomain formation or function, or LD interactions with vacuolar membranes. Alternatively, previous studies indicate that ESCRT function in lipophagy differs in cells exposed to acute glucose restriction (GR) compared with gradient, less severe GR (Zhang et al., 2020). Since ER redox state is perturbed to a greater extent by DTT treatment compared with lipid imbalance (Figure 1), the differential effects of these two stressors on LD uptake at the vacuole may be due to differences in the severity or nature of ER stress. Finally, it is possible that L_o microdomains induced by chemically induced ER stress are functionally distinct from those produced by phospholipid imbalance.

It is also not clear why DTT- and TM-induced ER stress induces formation of L_o microdomains that either do not function in or are not required for µLP induced by those stressors. What, then, is the function(s) of L_o microdomains under these conditions? Do stress-induced L_o microdomains promote formation or stabilization of contact sites between vacuoles and other organelles, and if so, how do those contacts contribute to the cellular stress response? What is the mechanism for stress-induced L_o microdomain formation? Ongoing and future studies are needed to address these questions.

MATERIALS AND METHODS

Request a protocol through Bio-protocol.

Yeast strains and growth conditions

All strains were derived from WT BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) from Open Biosystems (GE Dharmacon, Lafayette, CO) and are listed in Table 1. All strains were grown at 30°C with shaking at 200 rpm (Sherman, 2002); cho2Δ yeast strains were grown on liquid rich-glucose medium (yeast-peptone-dextrose, YPD) or synthetic complete (SC) medium supplemented with 1 mM choline chloride (Sigma-Aldrich, St. Louis, MO) or on solid YPD supplemented with 4 mM choline chloride. To induce acute phospholipid imbalance, cho2Δ strains first were grown for 6 h on SC + 1 mM choline chloride until midlog phase (OD₆₀₀ 0.10–0.35). Next, cho2Δ cells were washed once with choline-free SC medium and grown for 24 h on choline-free SC to induce acute phospholipid imbalance or on SC + 1 mM choline chloride to maintain normal phospholipid levels. Cells grown on SC + choline are noted as cho2Δ^+C1while cells undergoing acute phospholipid imbalance are noted as cho2Δ^–C1 (Vevea et al., 2015). Midlog phase cells were used in all experiments unless otherwise noted.

TABLE 1:

Yeast strains in this study.

Strain	Genotype	Source
BY4741	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0	Open Biosystems (Huntsville, AL)
EGS065	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 [pPM28-eroGFP:URA3]	This study
EGS091	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 cho2Δ::LEU2 [pPM28-eroGFP:URA3]	This study
EGS163	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4	This study
EGS203	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 snf7Δ::LEU2	This study
EGS208	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mcherry::hphMX4 cho2Δ::LEU2	This study
EGS222	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 VPH1-GFPEnvy::HIS3	This study
EGS230	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 cho2Δ::LEU2 vps4Δ::KamMX6	This study
EGS237	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 VPH1-GFPEnvy::HIS3 cho2Δ::LEU2	This study
EGS270	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NCR1-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2	This study
EGS271	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NPC2-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2	This study
EGS295	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 cho2Δ::LEU2 snf7Δ::KanMX6	This study
EGS316	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NPC2-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2 vps4Δ::KanMX6	This study
EGS317	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mCherry::hphMX4 IVY1-GFPEnvy::HIS3	This study
EGS353	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mcherry::hphMX4 IVY1-GFPEnvy::HIS3 cho2Δ::LEU2	This study
EGS377	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 npc2Δ::loxP ncr1Δ::URA3	This study
EGS389	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 cho2Δ::LEU2 snf7Δ::URA3	This study
EGS405	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ERG6-mCherry::hphMX4 npc2Δ::loxP ncr1Δ::URA3 cho2Δ:: KamMX6	This study
EGS458	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 ncr1∆::URA3 npc2∆::loxP snf7∆::LEU2 cho2∆:: KanMX6 ERG6-mCherry::hphMX4	This study
EGS515	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NCR1-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2 vps4Δ::KanMX6	This study
CTY174	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mCherry-hphMX4 IVY1-ENVY-HIS3 snf7∆::KanMX6	This study
CTY175	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mCherry-hphMX4 IVY1-ENVY-HIS3 cho2∆::LEU2 snf7∆::KanMX6	This study
CTY178	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mCherry-hphMX4 IVY1-ENVY-HIS3 npc2∆::ura3 ncr1∆::KanMX6	This study
CTY179	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 VPH1-mCherry-hphMX4 IVY1-ENVY-HIS3 npc2∆::ura3 ncr1∆::KanMX6 cho2∆::LEU2	This study
CTY222	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NCR1-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2 snf7Δ::KanMX6	This study
CTY225	MAT a his3∆1 leu2∆0 met15∆0 ura3∆0 NPC2-GFPEnvy::HIS3 VPH1-mCherry::hphMX4 cho2Δ::LEU2 snf7Δ::KanMX6	This study

ER stress was induced by treating midlog phase cells for 8 h with 5 mM DL-DTT (Sigma-Aldrich, St. Louis, MO) or 2 µg/ml TM (Sigma-Aldrich, St. Louis, MO). Specifically, cells were grown overnight on SC to reach midlog phase. The following morning, cells were diluted and grown for 8 h on SC, SC + 5 mM DTT, or SC + 2 µg/ml TM until midlog phase unless otherwise noted. To reduce sterols, cells were treated with 5 µM fenpropimorph (Santa Cruz, Dallas, TX). For analysis at 2-d (2D) stationary phase, cells were grown as previously described (Wang et al., 2014). Briefly, cells were grown overnight on SC to late midlog phase. The next day, cells were diluted to OD₆₀₀ = 0.15 and allowed to continue growing. When the OD₆₀₀ reached ∼1.7, cells were left to grow for another 48 h before performing imaging experiments.

Yeast strain construction

To delete genes of interest, the loci of interest were replaced with an auxotrophy selection marker amplified from pOM12/13 (P30387/P30388 Euroscarf) or a kanMX6 cassette from pFA6a-kanMX6 (Longtine et al., 1998; Gauss et al., 2005). pFA6a-kanMX6 was a gift from Jurg Bahler and John Pringle (Addgene plasmid #39296; RRID: Addgene_39296). These knockout strains were generated in corresponding amino acid dropout SC media or on YPD media with the required antibiotics (200 μg/ml G418 and 300 μg/ml hygromycin B; Sigma-Aldrich, St. Louis, MO) for selection.

To generate GFPEnvy or mCherry fusion proteins, GFPEnvy or mCherry was inserted in the endogenous locus at the C terminus of the coding sequence using modules amplified from pFA6a-link-GFPEnvy-SpHis5 (Slubowski et al., 2015) or pCY3090-02 (Young et al., 2012), respectively. pFA6a-link-GFPEnvy-SpHis5 was a gift from Linda Huang (Addgene plasmid # 60782; RRID: Addgene_60782) while pCY 3090-02 was a gift from Anne Robinson (Addgene plasmid # 36231; RRID: Addgene_36231).

eroGFP strains were generated by transforming either WT or cho2Δ with pPM28 (eroGFP CEN/ARS URA3), a plasmid that expresses roGFP2-HDEL C-terminally fused to the signal sequence of Kar2p (Merksamer et al., 2008). pPM28 was a gift from Feroz Papa (Addgene plasmid # 20131; RRID: Addgene_20131).

Analysis of ER redox potential

Strains transformed with pPM28 undergoing acute lipid imbalance or ER stress were imaged with an AxioObserver.Z1 microscope equipped with a 100×/1.3 oil EC Plan-Neofluar objective (Zeiss, Thornwood, NY) and an Orca-ER cooled CCD camera (Hamamatsu, Bridgewater, NJ) controlled by Zen Blue Edition (Zeiss). For validation of the probe, WT cells transformed with pPM28 were treated with 5 mM DTT or 5 mM H₂O₂ for 30 min and imaged. Oxidized and reduced channels were excited using a 405-nm LED and a 470-nm LED, respectively. Emission was acquired with a modified GFP filter (Zeiss filter 46 HE without excitation filter, dichroic FT 515, emission 535/30). Z-series were acquired through the entire cell with a z step of 0.3 μm and 1 × 1 binning. Images were deconvolved using a constrained iterative restoration algorithm assuming 507 nm excitation wavelength, 100% confidence level, and 60 iterations using Volocity 6.3 (Quorum Technologies, Puslinch, Ontario, Canada). The reduced:oxidized ratio channel was calculated by dividing the intensity of the reduced channel (λ_ex = 470 nm, λ_em = 525 nm) by the intensity of the oxidized channel (λ_ex = 405 nm, λ_em = 525 nm) after background subtraction and thresholding for each channel individually.

Western Blotting

Western blot analysis was performed as previously described (Vevea et al., 2015). Briefly, the same amount (0.5 OD₆₀₀ • ml) of cells was collected for each condition and resuspended in 150 μl of lysis buffer (50 mM imidazole, pH 7.4, 10 mM EDTA, 1% triton X-100, 2 mM phenylmethylsulfonyl fluoride, and protease inhibitor cocktail: pepstatin A, chymostatin, antipain, leupeptin, aprotinin, benzamidine, and phenanthroline). Samples were vortexed with 100 μl of glass beads for 5 min. After vortexing, 50 μl of 4× SDS sample buffer was added and samples were boiled at 100°C for 10 min; 35 μl of protein lysate was loaded for each condition onto a 10% SDS–PAGE gel with 0.5% 2,2,2-trichloroethanol (TCE, Sigma-Aldrich, St. Louis, MO). Before transferring, TCE was activated by exposing the gel to UV light (300 nm) for 2.5 min to detect total loading proteins as loading control (Ladner et al., 2004). Proteins were transferred to a PVDF membrane (Immobilon-FL; EMD Millipore, Billerica, MA). After transfer, the PVDF membrane was rinsed with H₂O and dried for 1 h prior to blocking with 3% skim milk in TBST for 1 h. Primary antibodies used include mouse monoclonal anti-mCherry (Abcam, Cambridge, MA; #ab125096; 1/2000 dilution) and mouse monoclonal anti-GFP (Proteintech, Rosemont, IL; 66002-1-Ig, 1/1000 dilution). Western blots were imaged using Luminata Forte Western HRP substrate (EMD Millipore, Billerica, MA) and the Chemidoc MP imaging system (Bio-Rad, Hercules, CA).

Western blot images were analyzed with Image Lab v5.2.1 (Bio-Rad, Hercules, CA) as follows. First, individual free mCherry bands were selected and a rolling disk background subtraction was applied with disk size = 10.0 mm. Later, the complete lane for each corresponding sample in the TCE image was selected and was analyzed with a rolling disk background subtraction with disk size = 70.0 mm. Total integrated intensity of free mCherry band was normalized to the integrated intensity of total protein loading control for each corresponding lane. Finally, for each independent experiment, all the samples were normalized to their experimental control, either cho2Δ^+C1 cells or cells grown on SC.

Fluorescence microscopy

Yeast cells were collected by centrifuging for 30 s at 3,800 × g at RT, and 1.6 μl of cells were placed on a glass slide and covered with a #1.5 coverslip. Images were acquired with an Axioskop 2 microscope equipped with a 100×/1.4 Plan-Apochromat objective (Zeiss) and an Orca-ER cooled CCD camera (Hamamatsu) and a pE-4000 LED illumination system (coolLED, Andover, UK) controlled by NIS Elements 4.60 Lambda software (Nikon, Melville, NY). GFP and mCherry were excited using a 470-nm LED with a ET470/40× filter and a 561-nm LED with a ET572/35× filter, respectively (Chroma, Bellows Falls, VT). Emission was collected through a dual eGFP/mCherry cube (59222, Chroma, Bellows Falls, VT). GFP and mCherry images were deconvolved using a constrained iterative restoration algorithm assuming 507 nm and 610 nm excitation wavelength, respectively, with 100% confidence limit and 60 iterations using Volocity 6.3.

All image analysis and processing were performed with Volocity 6.3 or Fiji (Schindelin et al., 2012). For visualization, all images were contrast-enhanced with similar parameters in each channel. All the analysis was performed on deconvolved unenhanced images. To measure the colocalization levels, images were thresholded and colocalization between ER proteins and LDs was quantified by measuring Manders’ overlap coefficient (R) for each cell in Volocity 6.3 (Dunn et al., 2011). To measure integrated intensity, objects of interests (ROI) were first identified after thresholding and appropriate size exclusion on the deconvolved images. The total integrated intensity for each cell was determined by the sum of voxel values of all identified objects in the cell.

Statistical Analysis

GraphPad Prism7 (GraphPad Software) was used for statistical analysis. All data were analyzed for normality with the D’Agostino and Pearson normality test. For comparison of two groups, p values were determined with an unpaired two-tailed Student’s t test for parametric distributions and an unpaired Mann–Whitney test for nonparametric data. For multiple group comparisons, p values were determined with a one-way ANOVA with a Bonferroni post-hoc test or a Sidak’s multiple comparisons test for parametric distributions and a Kruskal–Wallis test with Dunn’s post-hoc test for nonparametric distributions. Bar graphs show the mean and SEM, while boxes indicate the middle quartiles with the midline representing the median. For all tests, p values are classified as follows: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Abbreviations used:

DTT

dithiothreitol

ESCRT

endosome sorting complex required for transport

GR

glucose restriction

LD

liquid droplet

MVB

multivesicular body

NPC

Niemann-Pick type C

NVJ

nuclear–vacuolar junction

PC

phosphatidylcholine

PE

phosphatidylethanolamine

SC

synthetic complete

TCE

2,2,2-trichloroethanol

TM

tunicamycin

UPR

unfolded protein response

YPD

yeast-peptone-dextrose

WT

wild type.