LDO proteins and Vac8 form a vacuole-lipid droplet contact site to enable starvation-induced lipophagy in yeast

,


INTRODUCTION
Cellular adaptation to changing metabolic demands requires efficient communication between organelles and remodeling of subcellular structures.2][3] Membrane contact sites are established by an array of tethering proteins that bridge virtually all pairs of organelles, facilitating the bidirectional exchange of biochemical information in form of metabolites, ions, and lipids. 1,2,4,5These organelle contacts are key to intracellular signaling and are emerging as important sites for metabolic adaptation. 3,6,7n yeast, a contact site that dynamically changes size and architecture in response to nutrient availability is the nucleus-vacuole junction (NVJ).2][13] LDs are dynamic fat storage organelles with critical roles in lipid and energy metabolism that allow cells to adapt to changing nutritional cues. 14,15They consist of a neutral lipid core, particularly triacylglycerols (TAGs) and sterol esters, delimited by a phospholipid monolayer that originates from the outer leaflet of the ER membrane and contains integral and peripheral proteins. 143][24][25] Lipophagy is based on en bloc import into the lysosome/vacuole and plays a key role in LD consumption.In mammalian cells, uptake into the lysosome is mediated by macroautophagy, in which autophagosomes sequester LDs for delivery to the lysosome. 26,27Notably, an alternative lipophagic route for LD consumption has recently been described in hepatocytes. 280][31][32] During microlipophagy, LDs redistribute around the yeast vacuole and associate with liquid-ordered (L o ) microdomains on the vacuolar membrane before being engulfed. 33,34Though contact formation via molecules that bridge the LDs and the vacuole/ lysosome is anticipated to support LD engulfment, no such molecules have been identified.Still, proximity-based approaches employing LD-localized and vacuolar resident proteins suggest the existence of such a vacuole-lipid droplet contact site (vCLIP). 20ere, we uncover the molecular architecture of the contact site that anchors LDs to the vacuole to facilitate lipophagy in response to nutrient exhaustion (also see Diep et al. 35 submitted in parallel to this study).We demonstrate that the LD organization (LDO) proteins Ldo16 and Ldo45, encoded by overlapping genes and products of a splicing event, 36,37 attach LDs to the vacuolar membrane via Vac8, forming the vCLIP.Ldo16 was transcriptionally upregulated specifically upon nutrient depletion and anchored LDs to the vacuolar membrane via direct interaction of its C-terminal intrinsically disordered region with Vac8.Experimentally redirecting Vac8 to the nuclear membrane was sufficient to re-route LDs to juxtanuclear locations.In sum, we identify the molecular bridges that form vCLIP and show that this contact is critical for lipophagy and full lifespan extension via caloric restriction.

LDO proteins accumulate at the vacuole-LD interface upon nutrient exhaustion
The shift from rapid proliferation in nutrient-rich conditions to stationary phase upon nutrient exhaustion is associated with a metabolic switch that involves the transition from LD biogenesis to storage and gradual LD consumption.To assess how this transition affects LD-localized proteins, we followed the subcellular distribution of endogenous GFP fusions of the fatty acyl-coenzyme A (CoA) synthetase Faa4, the ergosterol biosynthesis enzyme Erg1, the phosphatidylinositol transfer protein Pdr16 and the LD organization proteins Ldo16/45. 36,37We analyzed cells in glucose-rich conditions (8 h) and after prolonged incubation (48 h), resulting in gradual glucose exhaustion, a switch to respiratory metabolism and entry into stationary phase. 9Under nutrient-rich conditions, all GFP fusions decorated the LDs, visualized using the neutral lipid stain monodansylpentane (MDH) (Figure 1A).As reported previously, 38 Erg1 also localized to the ER.In glucose-exhausted cells, Faa4 and Erg1 remained evenly distributed on the LD surface, while Pdr16 and in particular the LDO proteins formed distinct foci on LDs, mostly limited to one structure per LD.Simultaneous visualization of the vacuole using Vph1 mCherry revealed that these foci mark the vacuole-LD interface, suggesting that Pdr16 and the LDO proteins are redirected to contact sites between these organelles (Figure 1B).The targeting of the LDO proteins to the interface was not affected by genetic ablation of Pdr16 (Figure 1C).In contrast, the lack of both LDO proteins (DDldo) prevented the accumulation of Pdr16 at these sites and resulted in its cytosolic distribution (Figure 1C), suggesting that the LDO proteins are required to recruit Pdr16 to these contacts.The LDO proteins are encoded by overlapping open reading frames (Figure 1D), and C-terminal tagging with GFP (referred to as LDO GFP ), thus allows the simultaneous visualization of both proteins (Figures 1A-1C).Imaging over time revealed that LDO GFP rarely co-localized with the vacuole in exponentially growing cells but progressively formed foci at the vacuole after the diauxic shift (Figures S1A and S1B).Immunoblotting demonstrated that the expression of Ldo45 and Ldo16 was differentially affected by nutrient depletion (Figures 1E and 1F).Ldo16 protein levels progressively increased with time spent in stationary phase, while the levels of Ldo45 slightly decreased.Likewise, LDO16 mRNA levels (but not LDO45 levels) increased over time, indicating transcriptional upregulation of specifically Ldo16 (Figure S1C).Simultaneous visualization of Faa4 mCherry and LDO GFP in glucose-exhausted cells demonstrated that the LDO proteins were targeted selectively to the vacuole-LD interface, while Faa4 remained evenly distributed at the LD surface, independent of which fluorescent tags were used (Figure 1G).Next, we ruled out that the starvation-induced targeting of LDO GFP to the vacuole-LD interface was a simple consequence of increased LD size by expanding the LDs in exponentially growing cells on glucoserich media using oleate supplementation. 39Although oleate prominently enlarged the LDs, the LDO proteins remained distributed over the LD surface and only accumulated at the vacuole-LD interface in few, small foci when glucose was still available (Figure 1H).Again, also under oleate-supplemented conditions, entry into stationary phase redirected the LDO proteins to the expanded contact between the enlarged LDs and the vacuole.Furthermore, subjecting cells to phosphate restriction or acute nitrogen depletion resulted in the partitioning of the LDO proteins to the vacuole-LD interface (Figure 1I) and a specific increase of Ldo16 (Figures S1D-S1F).Collectively, this shows that limitation of the macronutrients glucose, nitrogen or phosphate induces Ldo16 expression and results in the specific targeting of the LDO proteins to the vacuole-LD interface, likely reflecting vCLIP that emerge in response to starvation.

LDO proteins target LDs to the vacuole to facilitate lipophagy
We directly tested whether LDO proteins are required for vCLIP formation.In wild-type (WT) cells, LDs were targeted to the vacuolar membrane (visualized via Vph1 mCherry ) shortly after glucose exhaustion.Upon prolonged starvation, a large part of the LD population was engulfed into the vacuole via lipophagy  (legend continued on next page) (Figure 2A).In contrast, such starvation-induced lipophagy was absent in cells lacking both LDO proteins (DDldo), which have already been proposed to contribute to lipophagy by yet unclear mechanisms. 37In DDldo cells, LDs remained clustered at one side of the vacuole, likely reflecting the nER and its contact to the vacuole at NVJs (Figure 2A).Indeed, visualization of the NVJs employing Nvj1 GFP demonstrated that in glucose-exhausted DDldo cells, a few LDs remained associated with the NVJs, but most LDs clustered at the nER (Figures 2B and  S2A).Transmission electron microscopy showed that in glucose-exhausted WT cells, more than 80% of LDs were closely attached to the vacuolar membrane at contact sites that reflect vCLIP, often already engulfed by the vacuole (Figures 2C and  2D).In contrast, in DDldo cells, only 10% of the LDs contacted the vacuole, while 80% remained associated with the ER, mostly nER (Figures 2C-2E).This suggests a critical function of the LDO proteins in vCLIP formation and supports the notion that contact formation between LDs and the vacuole is a prerequisite for microlipophagy.Pdr16 as additional factor that is targeted to vCLIP was not required for lipophagy (Figure S2B).In line with a reduction of neutral lipid consumption via lipophagy, the lack of LDO proteins resulted in increased levels of TAGs in glucose-exhausted cells (Figure 2F).We reasoned that the impaired consumption of neutral lipids may reduce viability during caloric restriction, a regime known to induce longevity across species. 40imilar to what was observed on standard media, the loss of vCLIP formation precluded lipophagy also when cells were grown into stationary phase on caloric restriction media containing 0.4% glucose (Figure S2C).Interestingly, inactivation of vCLIP compromised full lifespan extension induced by caloric restriction but did not affect cellular survival during chronological aging on control media with 2% glucose (Figure 2G).
To discriminate between Ldo16 and Ldo45 functions, we introduced selective deletion mutations to individually express either Ldo16 or Ldo45 from the chromosomal locus.Confocal microscopy and automated image quantification revealed that the individual loss of Ldo16 (but not of Ldo45) caused a slight reduction of vacuolar-LD engulfment upon glucose exhaustion, but not to the same extent as the simultaneous lack of both proteins (Figures 2H and 2I).Moreover, in DDldo cells but not the single deletion mutants LDs displayed a different morphology, being fewer in number and prominently enlarged, suggesting that the presence of either LDO protein is sufficient to adjust the number and size of LDs to cellular needs (Figures 2J and 2K).In line with this, both Ldo16 GFP and Ldo45 GFP still efficiently accumulated at vCLIP to establish contact when expressed individually (Figure 2L).Furthermore, overexpression of either Ldo16 or Ldo45 triggered massive accumulation of LDs inside the vacuole, suggesting induction of lipophagy (Figures 2M and S2D).As the complete Ldo16 sequence is also present in Ldo45, we created a truncated Ldo45 variant that lacks the part shared with Ldo16.This Ldo45 DC148 mutant still targeted the LDs but failed to attach LDs to the vacuole (Figures S2E and S2F).Taken together, both LDO proteins establish vCLIP by tethering the LDs to the vacuole via the shared Ldo16 domains, yet Ldo16 expression is selectively upregulated upon nutrient exhaustion.
The C-terminal disordered region of Ldo16 anchors LDs to the vacuole Next, we generated a set of Ldo16 mutants to assign function to distinct protein regions.The N-terminus of Ldo16 likely corresponds to hydrophobic transmembrane a helixes (Figure 3A) and includes a proline-leucine-leucine-glycine (PLLG) motif as typical helix breaker.This might facilitate hairpin-like membrane insertion, characteristic for LD proteins that are targeted to LDs from the ER membrane. 41,42In addition, Ldo16 is predicted to contain a cationic amphipathic helix (CAH; Figure 3B), again a common motif to target proteins to LDs. 43,44 The remainder of Ldo16 in its C-terminus is predicted to be an intrinsically disordered region, ending with a short a helix.We created a series of C-and N-terminally truncated Ldo16 variants as well as point mutants within the CAH, all ectopically expressed as GFP fusions in DDldo cells (Figures 3C-3E).Immunoblotting confirmed the expression of all Ldo16 mutants, though modification of the predicted CAH impaired protein stability (Figure 3D).Assessing their subcellular localization demonstrated that the N-terminal hydrophobic region served as ER targeting signal, as its deletion prevented ER import (Ldo16 DN49 and Ldo16 DN72 ).Consistently, the Ldo16 variant containing only the N-terminal hydrophobic helixes (Ldo16 DC98 ) was still imported into the ER.However, this mutant failed to associate with LDs, while a slightly longer variant that still contained the CAH (Ldo16 DC54 ) was fully redirected to LDs (Figure 3C), demonstrating that the CAH is critical to target Ldo16 from the ER to LDs in the so-called ERTOLD pathway. 42Exchanging 5 hydrophobic residues in the CAH with alanine did not prevent LD targeting, though accumulation of this Ldo16 5xA variant at the contact site was reduced (Figures 3C-3E).However, a stronger mutation of the CAH by inserting 2 glutamates resulted in severely compromised protein stability (Figure 3D) and ER retention (Figure 3C), supporting the notion that amphipathicity of the CAH contributes to Ldo16 targeting to LDs.Importantly, Ldo16 DC54 , which harbors a functional CAH but lacks the C-terminal disordered region, was efficiently redirected to LDs but decorated the complete LD surface  instead of accumulating at vCLIP.In addition, the complete LD population remained attached to one side of the vacuole, likely reflecting the nER (Figure 3C), demonstrating that the C-terminus of Ldo16 is critical for contact formation.Removing only the short a helix (24 C-terminal residues) did not prevent accumulation at vCLIP, indicating that it is the intrinsically disordered region that is essential for Ldo16 function in contact formation.Collectively, this suggests that Ldo16 is targeted to the ER via its N-terminal hydrophobic region, is redirected from the ER membrane to the LD surface via its CAH and attaches to the vacuolar membrane via the intrinsically disordered region in its C terminus to establish vCLIP (Figures 3F and 3G).

Vac8 is required for vCLIP formation
To identify cellular processes and molecular determinants involved in Ldo16-mediated LD-vacuole tethering, we examined LDO GFP localization in a set of mutants with established functions in LD biosynthesis, autophagy, the endosomal sorting complex required for transport (ESCRT), NVJ formation and vacuolar membrane lipid composition.All mutants, expressing Vph1 mCherry to visualize the vacuole and stained with MDH, were grown to glucose exhaustion and assessed microscopically for contact formation (Figures 4A and S3A), scored as co-localization between LDO GFP and Vph1 mCherry (Figure 4B).The LDO proteins have previously been shown to interact with seipin (Sei1) at the ER membrane to support LD biogenesis in growing cells. 36,37However, the lack of seipin did not compromise LDO GFP targeting to the contact sites.Likewise, neither genetic inactivation of autophagy nor impairment of endosomal protein sorting via deletion of genes coding for ESCRT components, which have been suggested to contribute to lipophagy, 30,31,33,45 altered LDO-mediated tethering of LDs to the vacuole.The formation of L o microdomains on the vacuolar membrane has also been associated with stationary phase lipophagy, 33,34,46 but defective L o domain formation did not prevent vCLIP formation.Similarly, changes in phosphatidylinositol metabolism had no effect on contact formation despite altered vacuolar morphology (Figures 4A, 4B, and S3A).Finally, we tested whether NVJs are required for LD tethering to the vacuole, as the LDO proteins have been shown to decorate a LD subpopulation associated with the NVJs. 36,37LDO-mediated tethering was unaffected in cells lacking either Nvj1 or Mdm1, bridging proteins required for NVJ formation, 10,47 or lacking the NVJ regulator Snd3. 9Importantly, one strain showed severely impaired vCLIP formation: the loss of the vacuolar protein Vac8 prevented LDO GFP targeting to the vacuolar-LD interface, resulting instead in its spreading over the complete LD surface (Fig- ure 4C).Quantification of LDO GFP -Vph1 mCherry co-localization supported the absence of vCLIP in Dvac8 cells (Figures 4B and  4D).In line with a critical function of Vac8 in recruiting LDs to the vacuole, cells lacking the palmitoyltransferase Pfa3, which lipidates Vac8 to attach it to the vacuolar membrane, 48 resembled Dvac8 cells (Figures 4C and 4D).The armadillo (ARM) repeat protein Vac8 has previously been linked to lipophagy 31 and contributes to multiple cellular processes via direct interaction with distinct proteins, including Nvj1 to tether the vacuole to the nER at NVJ, 10 Vac17 to facilitate vacuole inheritance, 49 as well as Atg11 and Atg13 to support phagophore assembly during selective and bulk macroautophagy, respectively. 50,51The lack of any of these Vac8 interactors did not preclude vCLIP formation (Figure 4F), suggesting that it is the absence of Vac8 per se rather than the impairment of associated processes that prevents vCLIP formation.Overall, these data demonstrate that Vac8 is required to recruit LDs to the vacuole.vCLIP formation is critical but not sufficient for lipophagy To assess whether the formation of vCLIP is sufficient to trigger vacuolar uptake of LDs, we quantified lipophagy in the deletion mutants analyzed for changes in vCLIP formation.Cells lacking ESCRT components were omitted, as a clear discrimination between LDs inside or outside the highly fragmented vacuole was not possible.Plotting lipophagy as a function of LDO GFP colocalization with the vacuolar membrane revealed that uniquely the lack of Vac8 and its palmitoyltransferase Pfa3 precluded vCLIP formation as well as lipophagy (Figure 4G).Despite efficient vCLIP formation, cells lacking regulators of macroautophagy such as Atg1, Atg6, and Atg8, all previously linked to lipophagy, 30 displayed reduced vacuolar LD uptake (Figure 4G).Similarly, the loss of Atg14, which has been suggested to mediate lipophagy through re-distribution to the vacuolar surface, 30 reduced LD uptake to some extent without affecting vCLIP formation (Figure 4G).Microscopic analysis demonstrated that Atg14 GFP targeted the rim of some, but not all, vCLIP established by the LDO proteins when cells were grown into glucose exhaustion (Figure S3B).Collectively, this indicates that key regulators of autophagy are not required for vCLIP formation but contribute to vacuolar LD uptake.
Interestingly, we observed that NVJ inactivation impacted on lipophagy but not vCLIP formation.Compromising NVJ formation by genetic ablation of Nvj1 or Mdm1 increased the vacuolar uptake of LDs (Figure 4G).Mdm1 has been suggested to mediate a three-way connection between ER, vacuole and LDs to support localized LD biogenesis at NVJ in proximity to the vacuole. 12,47,52Thus, we tested for co-localization of Mdm1 and the  S6 for statistical analyses.
LDO proteins and indeed detected Mdm1 GFP in close proximity to a subpopulation of vCLIP (Figure S3B).Still, Mdm1 was dispensable for vCLIP formation (Figures 4A and 4B), suggesting a potential regulatory function at the ER-vacuole-LD interface.
Finally, we assessed whether bringing the LDs to the surface of the vacuole was sufficient to induce lipophagy even in absence of the LDO proteins.To this end, we fused Faa4 and Vph1 to splitVenus fragments and established the reconstitution of splitVenus via bimolecular fluorescence complementation (Figure 4H).The enforced tethering of LDs to the vacuolar surface did not restore the lipophagy defect of DDldo cells, suggesting that the LDO proteins provide lipophagic function beyond mere tethering (Figure 4I).

Modification of the C terminus of Vac8 disrupts vCLIP formation
Next, we tested for co-localization of endogenously expressed LDO GFP and Vac8 mScarlet (Figure 5A).C-terminal tagging of Vac8 has extensively been used in the field, mostly without impairing Vac8 functions, 53 and we confirmed that mScarlettagging resulted in a functional protein based on growth (Figures S4A and S4B).Interestingly, the C-terminal tagging of Vac8 triggered the re-distribution of LDO GFP to cover the complete LD surface and precluded vCLIP formation, a phenotype reminiscent of its distribution in Dvac8 cells (Figures 4C and  5A).A similar disruption of contact formation was observed when using HA as alternative epitope (Figure S4C), while N-terminal tagging caused the loss of Vac8, most likely due to interference with its palmitoylation and thus defective membrane anchoring (Figure 5A).The interactions of Vac8 with Nvj1 and Atg13 have previously been studied at the structural level, showing that the last ARM repeat at its C terminus (ARM12) is not involved in the interaction. 54,55As C-terminal tagging selectively disrupted vCLIP formation while not affecting other Vac8 functions, we created a series of genomically encoded C-terminal truncations of Vac8, successively deleting the a helixes of ARM12 (Figure 5B).We monitored LDO GFP localization in combination with the vacuole (using FM4-64) and LDs (using MDH) (Figure 5C) and scored for LDO GFP -vacuole co-localization (Figure 5D) as well as for accumulation of LDO GFP at vCLIP versus the entire LD surface (Figure 5E).The truncation of the last a helix (a3) only slightly compromised contact formation (Figures 5D and 5E), although vacuolar morphology was already affected (Figure 5C).Further truncation of ARM12 disrupted contact formation, resulting in the spreading of LDO GFP across the complete LD surface, reminiscent of Dvac8 cells, and prominent vacuolar fragmentation (Figures 5C-5E).To assess whether these C-terminal truncations of Vac8 would also interfere with well-established interactions, we used Nvj1 GFP to monitor NVJ formation in cells harboring the Vac8 variants at the endogenous locus.This revealed a similar pattern of disruption of Vac8 binding to its partner, here Nvj1 (Figures 5F and S4D).The progressive loss of Vac8 interactions upon truncation of the a helixes of ARM12 correlated well with the extent of growth arrest (Figures 5G and S4E).As ARM12 does not directly contribute to the interaction interface of the Vac8-Nvj1 complex, determined by crystallography, 54,55 it is likely that these C-terminal truncations impair a general aspect of protein structure.The regions of both Nvj1 and Atg13 that interact with Vac8 are disordered loops near their C-termini. 54,55Despite these sharing little sequence homology, both loops run as extended peptides along the minor groove of the ARM repeat superhelix.We used ColabFold 56 to examine if an interaction surface in Vac8 could be predicted for Ldo16.Of the five best models obtained from full-length sequences of both proteins, all showed a portion of the disordered C terminus of Ldo16 in contact with the minor groove of the Vac8 superhelix (Figure 5H), with the top ranked model having the majority of the C terminus of Ldo16 making multiple contacts with Vac8 (22 of 34 residues between 104 and 137 had R8 contacts defined as atom centers %4 A ˚apart).Notably, the predicted position of Ldo16 is the same as that previously found for Nvj1 (Figure 5I).Thus, Ldo16 and Nvj1 might bind to the same Vac8 groove, potentially competing for Vac8 interaction when Ldo16-decorated LDs enrich at the NVJs shortly after glucose exhaustion.The elongated interaction surface along the minor groove of Vac8 is well conserved despite being spread across multiple ARM repeats (Figure 5J).The five ColabFold models aligned Ldo16 in the same direction, which was parallel to Vac8 (i.e., opposite to the direction of Nvj1 and Atg13).According to this prediction, the C-termini of Vac8 and Ldo16 would be in close proximity, and insertion of C-terminal tags simultaneously on both proteins might interfere with their interaction.To test this, we monitored LD localization in a strain equipped with Vac8 mScarlet and untagged LDO proteins.Indeed, Vac8 mScarlet accumulated at sites where LDs made contact with the vacuole when the LDO proteins were untagged (Figures 5K  S6 for statistical analyses.and 5L).Overall, these results suggest that (1) Vac8 is recruited to vCLIP and (2) C-terminal tagging of both proteins at the same time interferes with vCLIP formation.Clustering of Vac8 at LDs was absent in DDldo cells, while clear Vac8 patches at NVJs were visible (Figure 5K), demonstrating that the LDO proteins are critical for Vac8 recruitment to vCLIP.

Vac8 interacts with Ldo16 and recruits LDs to cellular membranes
We used electron microscopy and immunogold labeling to test for a specific targeting of native Vac8 to vCLIP in glucoseexhausted WT cells.As expected, we detected Vac8 at the entire vacuolar membrane, with an enrichment at the NVJs (Figures 6A  and 6B).Of note, Vac8 in addition clustered at vacuolar membrane regions that made contact with LDs and that often were in the process of engulfing LDs, suggesting that Vac8 indeed accumulates at vCLIP.Next, we probed for a direct interaction between Vac8 and Ldo16.We co-expressed glutathione-S-transferase-tagged Vac8 (GST-Vac8) with the soluble form of Ldo16 (Ldo16 DN49 lacking the hydrophobic membrane domain; Figure 3C), fused to the solubility tag Sumo, in E. coli.GST pulldown experiments revealed a direct interaction between Vac8 and Ldo16, with approximately 40% of the total expressed Ldo16 protein co-purifying with GST-Vac8 (Figures 6C and  6D).Thus, the C-terminal part of Ldo16 is sufficient to form a complex with Vac8.
To test whether Vac8 serves as a tether protein that recruits LDs to cellular membranes, we reconstituted this process at the nER, employing a previously established tethering system. 50usion of the N-terminal half of Nvj1 to Vac8DN, a variant that lacks the domains necessary to attach it to the vacuolar membrane, drives the attachment of Vac8DN to the nuclear envelope. 50Importantly, ectopic expression of this nER-attached Vac8 in Dvac8 cells recruited LDs to the nER (Figure 6E).The LDO proteins and Vac8 accumulated at concave nER patches in which attached LDs were buried, suggesting that Vac8 is the single necessary and sufficient component of the vacuolar membrane that recruits LDs (Figures 6E and 6F).Notably, targeting LDs to the nER resulted in a strong deformation of the nER at sites enriched in LDO and Vac8, raising the possibility that the interactions of these proteins at the interface between LDs and cellular membranes support membrane bending to push LDs inward.As a positive control to validate chimera functionality, Vac8DN was fused to Vph1 to retain it at the vacuolar membrane, which predicably resulted in co-localization of LDO GFP and Vac8 at the vacuole and LD attachment.Collectively, these data suggest that Vac8 interacts with LDO proteins to establish vCLIP and is sufficient to recruit LDs to an organelle, thus fulfilling the formal criteria of a membrane contact site tether. 1

Proteolysis associated with lipophagy depends on vCLIP
We quantitatively assessed the impact of defective vCLIP formation on lipophagic activity associated with proteolytic breakdown of cargo.To this end, GFP liberation from Faa4 GFP by vacuolar proteases was followed by immunoblotting.Faa4 GFP levels progressively decreased with time spent in glucose exhaustion, which was accompanied by the simultaneous increase of free GFP (Figures 7A-7C).Microscopic analysis confirmed vacuolar uptake and breakdown of Faa4 GFP -deco-rated LDs (Figure S5A).Genetic ablation of the LDO proteins reduced the level of free GFP but not Faa4 GFP , suggesting a drop in lipophagic activity (Figures 7A-7D).Notably, no GFP liberation was detectable in cells devoid of Vac8 (Figure 7D).Consistently, Dvac8 cells did not sequester LDs into the vacuole and displayed an increase in LD size comparable to DDldo cells (Figures 7E-7G).While loss of Vac8 completely blocked lipophagy, DDldo cells still maintained residual lipophagic activity.Here, some LDs were still engulfed into the vacuole in particular upon prolonged glucose exhaustion (72 h) (Figure 7F), and lipophagy-associated proteolytic GFP liberation from Faa4 GFP was reduced but not completely blocked (Figures 7A-7D).This suggests that LDO-independent mechanisms for lipophagy exist that still require Vac8.

Disruption of vCLIP induces NVJ expansion and PMN
Vac8 is a key resident of both NVJs and vCLIP, raising the possibility of cross-talk via this shared and perhaps limiting component.Indeed, inhibition of NVJ formation by removing Nvj1 or both Nvj1 and Mdm1 resulted in increased uptake of LDs into the vacuole (Figures 4G, 7E, and 7F).Additional loss of the LDO proteins blocked this uptake.Vice versa, the inactivation of vCLIP in DDldo cells affected NVJ size and shape, resulting in expanded NVJs that frequently appeared fragmented due to ongoing piecemeal microautophagy of the nucleus (PMN) (Figures 7H and 7I).This special form of microautophagy occurs only at the NVJs and is characterized by vacuolar membrane invagination that leads to the pinching-off of vesicles carrying portions of nuclear cargo for subsequent vacuolar degradation. 57Electron microscopy demonstrated a 3-fold increase of sections with PMN vesicles upon inactivation of vCLIP, and often multiple PMN vesicles formed simultaneously along the expanded NVJs (Figures 7J and 7K).Confocal microscopy using Nvj1 GFP confirmed a prominent increase of PMN in DDldo cells (Figures 7L and 7M).Likewise, immunoblotting demonstrated increased vacuolar GFP liberation from Nvj1 GFP in these cells, suggesting increased cargo turnover via PMN upon vCLIP inactivation (Figures S5B-S5D).Conceptually, the mutually negative regulation of NVJs and vCLIP is likely caused by a tug-of-war for limiting Vac8 but may also involve compensatory cellular responses to maintain needed lipid flux to the vacuole.These results point to a functional link between the NVJ and the vCLIP, both of which are established by Vac8 at the vacuolar membrane and metabolically regulated.

DISCUSSION
The mobilization of fat stored in LDs plays a central role in both physiological lipid metabolism and human pathology. 16,58,59Lipophagy, one of two routes for LD consumption, enables eukaryotic cells to use lipids as an energy source under nutrient deprivation.This route requires docking of LDs to the lysosome/ vacuole before uptake by so far unknown means.Here, we have identified the molecular machinery that tethers LDs to the vacuole to enable LD consumption via lipophagy in yeast.The LD-localized LDO proteins attach to the vacuolar ARM repeat protein Vac8 to form vCLIP, the vacuole-LD contact site.We demonstrate that vCLIP emerges specifically upon nutrient  S6 for statistical analyses.exhaustion and is critical for efficient LD consumption via lipophagy.
The LDO proteins Ldo45 and Ldo16 were initially identified as accessory factors of yeast seipin, contributing to spatial organization of LD biogenesis at the ER. 36,37Similar to the LDO proteins in yeast, also the lipid droplet assembly factor 1 (LDAF1 alias promethin), the putative mammalian LDO homolog, supports LD biogenesis by interacting with seipin and subsequently dissociates from seipin to target the surface of mature LDs. 60,61e demonstrate that at a later step, when nutrients are exhausted, the LDO proteins establish vCLIP by redistributing to defined Vac8-positive foci of adjacent vacuoles.Similarly, also Pdr16, a cytosolic protein that is recruited to LDO-positive LDs, 36 redistributed to vCLIP in nutrient-exhausted cells.However, Pdr16 was not needed for either vCLIP formation or lipophagy.Interestingly, vCLIP formation occurs independently of several cellular processes and factors previously linked to lipophagy, including vacuolar microdomain formation, the ESCRT machinery and autophagy regulators. 31,33,34,45,46,62Still, despite efficiently forming vCLIP, several of the analyzed mutants, in particular those defective in autophagy, displayed reduced vacuolar LD uptake, demonstrating that vCLIP formation is a prerequisite but not sufficient for successful lipophagy.Artificial docking of LDs to the vacuole in absence of LDO proteins did not restore lipophagy.Hence, mere contact formation does not trigger vacuolar uptake of LDs, suggesting that LDO proteins not only tether these organelles but in addition participate in the sequestration step.
Our data demonstrate that both LDO proteins are able to form vCLIP and to induce lipophagy, which contributes to long-term survival under caloric restriction.Their differential expression related to the cellular metabolic state likely reflects different roles in LD biology, and only Ldo16 is transcriptionally upregulated upon nutrient depletion.Our series of Ldo16 mutants revealed that its N-terminal hydrophobic domains insert it into the ER, its CAH redirects it to the LD surface, and its C-terminal intrinsically disordered peptide region directly binds Vac8 at the vacuolar surface.Formation of vCLIP and subsequent lipophagy was prevented by deleting VAC8, and phenocopied by loss of Pfa3, which palmitoylates Vac8 to anchor it to the vacuolar membrane.Notably, re-targeting Vac8 to the nuclear envelope was sufficient to re-route LDs to these sites, demonstrating that Vac8 is a vCLIP tether.
Multiple aspects of vacuolar homeostasis require the ARM repeat protein Vac8, which interacts with Atg13 and Atg11 to recruit the phagophore assembly site (PAS) to the vacuole to support bulk and selective autophagy, respectively, and interacts with Nvj1 to form the NVJ. 50,51,57,63Crystallographic studies demonstrated that both Atg13 and Nvj1 interact with Vac8 via disordered loops that associate with the minor groove of the ARM repeat superhelix.According to ColabFold, the disordered loop of Ldo16 is predicted to occupy the same groove.Thus, Atg13, Nvj1 and Ldo16 might compete for binding, and Vac8 may switch from one binding partner to another to fine-tune multiple autophagic processes.Related to this, recruitment of the selective PAS to the vacuolar membrane has been proposed to derive from avidity-mediated Vac8-Atg11 interactions with low affinity that are stabilized by a high local concentration and limited diffusion of the interaction partners. 50,64The same concept of a body to be autophagocytosed acting as a platform to concentrate Vac8 has been applied to the NVJ 55 and now can be seen in the context of LDs.Once an Ldo16-decorated LD comes into proximity with the rim of the NVJ, it can access a pool of highly concentrated Vac8, which exchanges its partner from Nvj1 to Ldo16 to form vCLIP.This concept places Vac8 as a key regulator of autophagic processes during starvation, with all of macroautophagy, PMN and lipophagy being routed through the same final common pathway.In support of this, we find that the loss of vCLIP formation upon genetic ablation of the LDO proteins results in NVJ expansion and induction of PMN.Vice versa, inactivation of NVJ formation triggers lipophagy.Though a functional link between PMN and lipophagy remains to be explored, our findings establish Vac8 as critical regulator of the contact sites that enable delivery of diverse cargo to the vacuole.

Limitations of the study
This study identifies the molecular tethering machinery critical to attach LDs to the vacuole for subsequent en bloc uptake, but knowledge gaps remain.How is the invagination of the vacuolar membrane at vCLIP achieved to push LDs inward and drive their uptake, and how do autophagy-related proteins contribute to this engulfment?In addition, the process of LD consumption inside the vacuolar lumen remains elusive.While our data suggest a slow and progressive consumption of LDs that supports longterm viability in stationary phase, how such gradual neutral lipid mobilization within the vacuole is mechanistically accomplished remains to be explored.Then, the sequence corresponding to spliced LDO45 was amplified from pRS313-spLdo45 and recombined to replace the cassette.For plasmid construction for the Vac8-Ldo16 interaction assay in E. coli, a 2.9 kb DNA fragment was synthesized, encompassing codon-optimized Ldo16 50-148 -RBS-GST-Vac8 with sequences homologous to pSUMO-YHRC.Following yeast homologous recombination with pSUMO-YHRC as previously described, 70 pCA1057 (coding for Sumo-Ldo16 50-148 -RBS-GST-Vac8) was isolated as a kanamycin-resistant clone.The derivative pCA1056 (coding for Sumo-Ldo16 50-148 -RBS-GST) was constructed by HindIII restriction of pCA1057 to release VAC8 sequences followed by religation.

Yeast culturing conditions
All strains were grown in baffled Erlenmeyer flasks at 28 C and shaking at 145 rpm in synthetic complete medium (SC), containing 0.17% yeast nitrogen base (BD Difco), 0.5% (NH 4 ) 2 SO 4 (Carl Roth) and 30 mg/l of all amino acids (except 80 mg/l histidine and 200 mg/l leucine and, for the BY4742 strain, 120 mg/l lysine), 30 mg/l adenine and 320 mg/l uracil, with 2% glucose (SCD) or 0.4% glucose (SCD 0.4%) for caloric restriction.Plasmid-containing strains were grown in SCD without histidine, uracil or leucine.Overnight cultures were incubated for 16-20 h in SCD and used to inoculate cultures to OD 600 0.1 in SCD, followed by culturing into glucose exhaustion and stationary phase.For caloric restriction, overnight cultures in SCD were used to inoculate cells to OD 600 0.1 in SCD 0.4%, followed by culturing and rapid exhaustion of the limited glucose.For nitrogen starvation conditions, cells were pre-inoculated at 0.01 OD 600 and grown in SCD for 12 h, washed with water and diluted 1:10 into SD without (NH 4 ) 2 SO 4 and amino acids (SD-N).For phosphate restriction, cells were inoculated in SCD prepared using YNB without phosphate (FORMEDIUM).Phosphate was supplemented as NaH 2 PO 4 to a final concentration of 0.2 mM (instead of 7 mM as in standard SCD), allowing regular growth but leading to early entry into stationary phase due to phosphate exhaustion as described before. 71,72For oleate supplementation, cultures were inoculated to OD 600 0.1 in SCD and grown for 5 h before addition of 0.5% sodium oleate (final concentration; dissolved in 0.1% Tween-20) (Sigma-Aldrich O3880) or of the respective solvent control.For deletion and tagging of genes, yeast cells were grown in rich medium (YPD) containing 20 g/l peptone (Gibco Bacto BD Biosciences), 10 g/l yeast extract (Bacto BD Biosciences) and 4% glucose.For subsequent selection of mutants, YPD plates containing hygromycin B (FORMEDIUM, HYG5000), nourseothricin sulphate (Jena Biosciences, AB102XL) or G418 (Sigma-Aldrich, A1720-5G) or SCD plates with all amino acids except for histidine, uracil or leucine, were used.

Analysis of cell growth
Cells from overnight cultures were used to inoculate 250 ml SCD to OD 600 0.1 in 96-well microplates with clear, flat bottom (Greiner Bio-ONE).Plates were shaking at 999 rpm and 28 C and growth was measured by monitoring OD 600 every 2 h for 10 h using a plate reader (2300 EnSpire, Perkin Elmer).

Flow cytometric analysis of cellular survival during aging
Cellular survival was determined using propidium iodide (PI) staining, indicative of loss of plasma membrane integrity and thus cell death. 73Cultures were inoculated in standard glucose media with 2% glucose (SCD) or caloric-restricted media (SCD 0.4%) as described above, and aliquots were collected at indicated days during chronological aging.Cells were transferred into 96-well plates, pelleted by centrifugation at 3500 rpm for 1 min and resuspended in 250 ml PBS containing PI (Sigma-Aldrich, 81845) with a final concentration of 0.02 mg/ml (early days) or 0.1 mg/ml (late days), followed by incubation in the dark for 10 min.Cells were pelleted, resuspended in PBS and evaluated via flow cytometry using a Guava easyCyte HT with guavaSoft 3.3 software (Luminex Corporate).5000 cells were analyzed per sample, and PI negative cells were scored as alive.

Quantitative Real-Time PCR
To assess gene expression via qRT-PCR, approximately 30 OD 600 of cells were collected and total RNA was purified using the Ribopure-Yeast kit (Thermo Fisher, AM1926).Genomic DNA was digested using TURBO DNase (Invitrogen AM2238) according to the supplier's protocol.2 mg of total RNA was reverse transcribed with SuperScript transcriptase (Thermo Fisher, 18064014) following the manufacturer's instructions.qRT-PCR was performed in triplicates with the KAPA SYBR Fast qPCR Master mix (Sigma-Aldrich, KK4600) using a Rotor-Gene Q (Qiagen) PCR cycler.Data are presented as fold changes using the comparative Ct method (DDCT) 74 and UBC6 as housekeeping gene.All primers used for qRT-PCR are listed in Table S5.
Quantification of total cellular triacylglycerol levels 10 OD 600 of cells, cultured for 48 h in SCD, were harvested, washed with cold PBS and resuspended in 5% IGEPAL CA-630 (Sigma-Aldrich, I8896).Cells were mixed with glass beads and lysed with a Bioprep-24 homogenizer (Allsheng), using 3 cycles of 30 s. Similar volumes of the resulting lysates, corresponding to similar amounts of cells, were used to measure triacylglycerol levels using the Triglyceride Assay Kit (Abcam, ab65336) according to the manufacturer's instructions.To account for background signal due to the presence of diacylglycerol and in particular glycerol in the samples, each sample was corrected by a background control of its own without addition of lipase.Thus, the values obtained without lipase addition were subtracted from the values obtained for the same sample upon addition of lipase.Cells lacking the two acyltransferases Dga1 and Lro1 and hence almost completely devoid of triacylglycerols were used as negative controls, resulting in values resembling background.Measurements were obtained by recording fluorescence intensity (excitation at 535 nm; emission at 570 nm) with a plate reader (2300 EnSpire, Perkin Elmer).
Confocal Fluorescence Microscopy 1 OD 600 of cells were harvested at indicated time points, stained using indicated dyes as described below and seeded on 3% agar/ PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8mM KH 2 PO 4 ; pH 7.4).The samples were imaged using a ZEISS LSM780 microscope equipped with an 63x/1.40Oil M27 objective using the ZEN black software.For Figures 1G, 4H, and S3B, a ZEISS LSM800 Airyscan microscope equipped with an 63x/1.40Oil M27 objective and the ZEN blue software control was used.Appropriate laser and detector settings were applied to visualize endogenously tagged proteins (GFP, mCherry or mScarlet) or dyes.For LD stainings, wells were incubated for 10 min in the dark with BODIPY493/503 (0.12 nM; Fischer Scientific, 11540326) or monodansylpentane (MDH) (100 mM; AUTODOT Abcepta, #SM1000b) and then washed with PBS prior to seeding on agar-coated slides and microscopic analysis.For vacuolar staining, cells were either incubated for 30 min in the dark with CellTracker Blue 7-amino-4-chloromethylcoumarin (CMAC) Dye (1 mM; Thermo Fisher, C2111) and washed with PBS before imaging, or supplemented with FM 4-64FX (Invitrogen, 11574816) to a final concentration of 15 mM directly in the culture media until collecting the cells for microscopic analysis.

Image analysis and quantification
Images obtained with the ZEISS LSM800 Airyscan were first processed using the ''Airyscan processing'' algorithm in the ZEN blue software.The open-source software Fiji 66 was used to further process and quantify all confocal micrographs.To process the images, Gaussian filtering (s = 0.8-1.5)was applied, followed by background subtraction (rolling ball radius = 25-50 pixels) and Unsharp mask settings when required.Images from the same experiment were processed with similar settings.Brightness and contrast were adjusted for each channel equally in individual experiments for analysis.The ratio of 'LDs in/out' of the vacuole using BODIPY was calculated by automatically measuring Integrated Density (IntDen) 'inside of vacuoles' segmented with the Huang algorithm and divided by the IntDen of 'outside of vacuoles' (calculated by subtracting the IntDen 'inside the vacuoles' from the IntDen of the whole cell, segmented with YeastMate or Cellpose 2 75,76 ).'Relative lipophagy' was obtained by calculating the 'LDs in/out' ratio of MDH-stained cells and normalizing by the control averaged ratio.All intensity measurements were performed in the original, unprocessed images.LD size was quantified by segmenting LDs with the Yen algorithm and measurement of the area.Number of LDs per cell was quantified by automated counting of the segmented LDs against the mask of segmented cells with Find Maxima.In both cases, the results were analyzed with a frequency distribution and were fitted to a gaussian curve.The colocalization analysis to obtain the Pearson's and Manders M2 coefficients was performed with the JaCoP plugin, thresholding images with only Gaussian blur, subtract background and Unsharp Mask adjusted.Quantification of cells according to LDO GFP distribution (Figure 5E) as well as according to NVJ formation using Nvj1 GFP (Figure 5F) in the Vac8 truncation mutants was performed manually using the Cell counter plugin.

Transmission electron microscopy
For transmission electron microscopy, samples were prepared as described. 9Briefly, a Wohlwend Compact 03 (M.Wohlwend GmbH, Sennwald, Switzerland) was used for high-pressure freezing of samples, and freeze-substitution was performed for 1 h in a Leica EM AFS2 (Leica Microsystems, Vienna, Austria) using the Leica reagent bath with a flow-through ring of 2% uranyl acetate dissolved in 90% acetone and 10% methanol at -90 C. 77 Samples were washed twice in acetone while the temperature was gradually raised (2.9 C per h to -50 C).Samples were infiltrated with increasing amounts of Lowicryl HM20 (Polysciences, Warrington, PA, 15924-1) mixed with acetone (1:4, 2:3, 1:1, 4:1 and 100% 3x) at -50 C and a duration of 2 h per step.UV light was used to induce polymerization for 72 h at -50 C, followed by 24 h at room temperature.A Reichert-Jung Ultracut E Ultramicrotome (C.Reichert, Vienna, Austria) with an ultra 45 diamond knife (Diatome, Biel, Switzerland) was used to cut sections of 70 nm, which were collected on copper slot grids coated with 1% Formvar (TAAB).2% uranyl acetate in dH 2 O and Reynold's lead citrate were applied for on-section contrast staining. 78A Tecnai T12 electron microscope equipped with a Ceta CMOS 16M camera (FEI Co., Eindhoven, the Netherlands) was used for imaging of samples at 120 kV.Quantification of LDs in direct physical contact with the vacuole or the ER as well as PMN events per section was performed manually.

Immuno-electron microscopy
For immunogold labelling, the sample blocks were prepared and embedded in HM20 resin as described above, as this sample preparation has been shown to be amenable to immune-labelling. 79Grids with 70 nm-thick sections were fixed in 1% paraformaldehyde in PBS for 10 min and blocked with 0.1% fish skin gelatin and 0.8% BSA in PBS for 1 h.For primary antibody labeling, samples were incubated overnight with an antibody against Vac8 (dilution 1:30, rabbit, gift from Christian Ungermann), followed by 3x wash in PBS for 20 min and 1 h incubation with a secondary gold-conjugated antibody (anti-rabbit IgG 10 nm gold; EMS Electron Microscopy Sciences).After washing with PBS, glutaraldehyde (2.5%) was applied to the sections for postfixing for 1 h, followed by contrast staining using uranyl acetate and Reynold's lead citrate as described above.The detection of gold particles was done automatically using the IMODfindbeads program 80 inside of an area including 30 nm (the size of the antibody sandwich plus the gold particle) on either side of the membranes.Special drawing tools were used to efficiently model such areas in the IMOD suite of programs, followed by automated extraction of the quantification of the gold beads per area.Vac8 labeling density was quantified at vacuolar membrane regions in contact with LDs (vCLIP), in contact with the nER (NVJs), not in contact with other organelles (vacuolar membrane) and at the nER excluding the NVJs.In total, 57 sections were quantified, with varying numbers of data points for the 4 membrane categories, depending on their presence in respective section.

Structural modelling
To model the conservation of the Vac8 surface, a multiple sequence alignment was made that focussed on Vac8 only using 2 rounds of HHblits, which searches into a nr30 database (non-redundant above 30% identity).After the first iteration 106 hits were included with e-value <10 -20 .After the second iteration, 294 proteins were included (e-value < 10 -42 ).This excluded other Armadillo repeat proteins.Aligned sequences were submitted along with the solved structure of Vac8 to the ConSurf server 81 to obtain conservation scores for residues on the surface scaled between 0-10.To model the interaction of Vac8 with Ldo16, ColabFold 56 was seeded with Vac8 residues (560 residues, 19-578, missing the disordered N-terminus) and full length Ldo16 (148 residues).The structure shown in Figure 5H is the rank 1 model, for which a version was obtained with side-chains positioned in relaxed conformations.
Vac8-Ldo16 interaction assay E. coli BL21(DE3) cells were transformed with pCA1056 and pCA1057 that co-express Sumo-Ldo16 DN49 with GST or GST-Vac8, respectively.Cells were grown in salt free 2xYT (1.6% tryptone and 1% yeast extract) supplemented with 50 mg/L kanamycin and 2 mM MgSO 4 at 30 C until OD 600 reached 0.8-1.Protein expression was induced by the addition of 0.5 mM IPTG and 300 mM NaCl.The cells were harvested after 4 h and were resuspended in LWB150 lysis buffer (40 mM Hepes-KOH pH 7.4, 150 mM KCl, 5 mM MgCl 2 , 5% (v/v) glycerol, 1 mM PMSF, 10 mM b-mercaptoethanol) and lysed by three passages through a Emulsiflex-B15 homogenizer (Avestin).The lysates were centrifuged at 19 000 g for 30 min at 4 C and the supernatants were incubated with Protino Glutathione Agarose 4B beads (MACHEREY-NAGEL GmbH & Co. KG) for 2 h at 4 C.The beads were washed thrice with LWB150 buffer and eluted with the same buffer containing 10 mM reduced glutathione.The proteins levels were analyzed by SDS-PAGE and immunoblotting.

Statistical analysis
Data are presented either as dot plots, showing individual data points, mean (line) and error bars representing standard error of mean (SEM), or as line graphs with symbols depicting mean and SEM (survival as well as protein and mRNA level) or as line graphs depicting a gaussian non-linear regression fit histogram (frequency distribution).Sample size, referring to independent biological replicates, is indicated in the respective figure legends.Statistical analysis was performed using GraphPad Prism (v8.0).Shapiro-Wilk's test and visual inspection of Q-Q-plots was used to check for normal distribution of data, and analysis of variance (ANOVA) with Tukey's post hoc test was used for comparisons between multiple groups.The equality of group variances was checked with the Brown-Forsythe-Test.Due to unequal variances of data shown in Figures 2F and 6B, a Welch's t Test or a Welch's ANOVA with Dunnett T3 post hoc test was used.Where appropriate, a two-way ANOVA, corrected with Tukey's or Bonferroni's multiple comparisons test, was applied.Significances are presented as ***p<0.001,**p<0.01,and *p<0.05.Details for all statistical analyses performed are listed in Table S6.

Figure 4 .
Figure 4. Vac8 is required for vCLIP formation (A-D) Micrographs of WT and indicated mutants expressing LDO GFP and Vph1 mCherry , stained with MDH and analyzed at 48 h (A and C) and corresponding co-localization analysis of Vph1 mCherry with LDO GFP (B and D).Mean ± SEM; n = 4-10, at least 30 cells per n.Scale bars, 2 mm.(E) Schematics of Vac8 functions and interactors.(F) Micrographs of cells lacking Vac8 interactors and expressing LDO GFP and Vph1 mCherry , stained with MDH at 48 h.Scale bars, 2 mm.(G) Relative lipophagy, determined by quantifying the ratio of MDH intensity inside/outside the vacuole, plotted against the Pearson's coefficient for Vph1 mCherry and LDO GFP co-localization as in (B).Mean ± SEM; n = 4-10, at least 30 cells per n.(H) Micrographs of WT and DDldo cells expressing one or both splitVenus fragments fused to Vph1 (Vph1 VN ) or Faa4 (Faa4 VC ), stained with FM4-64 and MDH at 48 h.Scale bars, 2 mm.(I) Micrographs in (H) were used to quantify the ratio of MDH intensity inside/outside the vacuole.Mean ± SEM; n = 5-6, with 40-50 cells per n.*p < 0.05, ***p < 0.001, ns, not significant.See related Figure S3 and TableS6for statistical analyses.
Detailed methods are provided in the online version of this paper and include the following: Developmental Cell 59, 1-17, March 25, 2024 15 Please cite this article in press as: A ´lvarez-Guerra et al., LDO proteins and Vac8 form a vacuole-lipid droplet contact site to enable starvation-induced lipophagy in yeast, Developmental Cell (2024), https://doi.org/10.1016/j.devcel.2024.01.014