The yeast FIT2 homologs are necessary to maintain cellular proteostasis and membrane lipid homeostasis

Lipid droplets (LDs) have long been regarded as inert cytoplasmic organelles with the primary function of housing excess intracellular lipids. LDs arise from the endoplasmic reticulum (ER) and contain a core of the non-polar lipids triacylglycerol (TAG) and steryl ester (SE) surrounded by a phospholipid monolayer, with phosphatidylcholine (PC) as the major component (Grillitsch et al., 2011). More recently, LDs have been strongly implicated in conditions of lipid and protein dysregulation. These conditions are major contributors to the pathophysiology of metabolic diseases and concomitantly activate cellular stress response pathways, namely the unfolded protein response (UPR) and heat-shock response (HSR). Increased LD biogenesis has been extensively observed in cells under stress conditions. Introduction of oxidative stressors or the attenuation of antioxidant capacities of cells result in the formation of LDs (Lee et al., 2013; Liu et al., 2015; Nguyen et al., 2017). While the extent of UPR activation and the resulting transcriptional profile differs between proteotoxic and lipid stress, a similar increase in lipogenic markers and concomitant LD formation is observed (Fei et al., 2009; Fun and Thibault, 2019; Ho et al., 2020; Hou et al., 2014; Lee et al., 2012; Thibault et al., 2012). It remains to be determined whether LDs contribute to stress induction or if this is reflective of the adaptive role of LDs to mitigate the otherwise deleterious effects of stress. However, all these undoubtedly highlight the complex integration of LDs in stress response pathways. In addition to this, the UPR regulates metabolic pathways to a certain extent under normal physiological conditions (Lee et al., 2008), and similarly orchestrates the complex transcriptional metabolic reprogramming under ER stress induction (Thibault et al., 2012).

The fat storage-inducing transmembrane (FIT; also known as FITM) family of proteins constitutes a group of evolutionarily conserved proteins eponymously named for their role in lipid metabolism and LD formation (Kadereit et al., 2008). In mammals, FIT proteins exhibit differential expression patterns, with FIT1 (FITM1) being expressed primarily in cardiac and skeletal muscle tissue, and FIT2 (FITM2) being more ubiquitously expressed. FIT proteins (FIT1/FIT2) are ER-resident proteins with a total of six transmembrane domains and with both N- and C- termini facing the cytosol (Gross et al., 2010; Kadereit et al., 2008). The S. cerevisiae FIT2 homologs (ScFIT) SCS3 and YFT2 are predicted to share the same membrane topology. The pioneer study on the FIT proteins has identified their profound effect on the formation and accumulation of LDs both in vitro and in vivo (Kadereit et al., 2008). Transient overexpression of FIT2 was sufficient to drive the formation of LDs, a process that was later hypothesized to be mediated by the capacity of FIT2 to directly bind and partition TAG from the ER into storage in LDs in vitro (Gross et al., 2010). Transgenic expression of the mammalian FIT2 gene in the budding yeast S. cerevisiae (Moir et al., 2012) and in the plant models Arabidopsis thaliana and Nicotiana tabacum (Cai et al., 2017) induced the formation of cytoplasmic LDs. Similarly, transient expression of the ScFIT genes in mammalian cells in vitro led to the increased formation of LDs (Moir et al., 2012). A gain-of-function FIT2 mutant was identified with a 3-amino-acid mutation within its fourth transmembrane domain, which interestingly houses the most highly conserved amino acid residues from yeast to humans (Ramus et al., 2011). Conversely, it was found that the deletion of FIT2 greatly compromised LD formation in the model organisms Danio rerio and Caenorhabditis elegans, as well as in the pathogenic yeast Candida parapsilosis (Choudhary et al., 2015; Kadereit et al., 2008; Nguyen et al., 2011). All these reports further reinforced the initial hypothesis of the FIT proteins directly functioning in LD biogenesis.

In contrast, S. cerevisiae deletion mutants for either or both ScFIT genes retain their capacity to form LDs, with size and number comparable to that in wild type (WT) (Moir et al., 2012). Upon closer investigation, it was found that LDs fail to completely bud off from the ER in the absence of the ScFIT proteins (Choudhary et al., 2015), presumably through alterations in ER membrane lipid properties (Choudhary et al., 2018). These findings have since then led to the investigation of alternative functions of the FIT class of proteins. SCS3 was initially reported to have a putative role in the regulation of phospholipids (Hosaka et al., 1994), a function that was largely unexplored until a large-scale genetic screen reported on the strong interactions between the ScFITs and genes involved in phospholipid biosynthesis, including DGK1 and PSD1 (Moir et al., 2012). In-depth analysis of the ScFIT protein sequences have since then revealed the presence of the catalytic site of a lipid phosphatase (Hayes et al., 2017). In vitro analyses have identified the capacity of mammalian FIT2 to hydrolyze phosphates from phosphatidic acid (PA) and lyso-PA to yield diacylglycerol (DAG) and monoacylglycerol (MAG), respectively. On the other hand, substrates for the lipid phosphatase activity of the ScFIT proteins have yet to be determined. This conserved catalytic function has been associated with the aberrant ER whorling phenotype observed in cells devoid of the FIT proteins (Hayes et al., 2017).

Conjointly, several different perspectives now exist on the function of the ScFIT proteins. However, definitive evidence for the function of FIT proteins in either binding and partitioning neutral lipids (NLs) or influencing NL and phospholipid metabolism in the complex in vivo environment remains scarce. Moreover, how any of these functions impact ER homeostasis and the UPR are partially unexplored.

In this study, we investigated the role of SCS3, identified as one of the downstream UPR target genes. As the UPR transducer Ire1 is essential for viability in the absence of SCS3, we generated the temperature-sensitive allele scs3-1 to reveal its role without the masking effect of the UPR program. We demonstrated that dysfunctional SCS3 leads to the accumulation of TAG at the ER, a shift in phospholipid distribution and the biogenesis of aberrant LD morphology. Furthermore, we identified the interactome of ScFITs, with Scs3 being found to interact with components of the proteostatic machinery. Next, we demonstrated that ScFIT mutants impaired the clearance of ER-associated degradation (ERAD) client proteins, which was exacerbated by lipid imbalance. Together, our data support a model where ScFITs play an important role in lipid metabolism and proteostasis beyond their defined roles in LD biogenesis.

Previous synthetic genetic array (SGA) analyses have revealed the synthetic lethality between SCS3 and IRE1, which encodes for the sole UPR transducer in yeast (Moir et al., 2012). Additionally, the mutant scs3Δ strain has been reported to activate the UPR (Jonikas et al., 2009), while SCS3 has conversely been shown to be transcriptionally upregulated upon UPR activation resulting from either proteotoxic stress or lipid bilayer stress (LBS) (Ho et al., 2020; Thibault et al., 2012; Travers et al., 2000). As previously reported (Becuwe et al., 2020; Choudhary et al., 2015; Hayes et al., 2017), Scs3 and Yft2 proteins localize to the ER (Fig. S1A). Hence, we sought to further understand the role of SCS3 within the UPR program. First, we monitored UPR activation using the UPR element (UPRE)-LacZ reporter assay (Cox and Walter, 1996). Unexpectedly, no significant UPR activation was observed in scs3Δ mutants, contradicting a previous report (Jonikas et al., 2009) while being consistent with other findings (Moir et al., 2012) (Fig. 1A). Similarly, there was no significant UPR activation in yft2Δ nor in ScFITΔ cells. All mutant strains were able to mount an UPR response upon treatment with the ER stress-inducing agent tunicamycin (Tm) although the level of activation was significantly lower in scs3Δ and ScFITΔ. Similarly, the heat-shock response (HSR), a cytosolic proteotoxic stress compensatory pathway, was dampened in ScFITΔ upon heat stress (Fig. S1B,C). To complement the UPR assay, we asked whether SCS3 and YFT2 genes are upregulated in a UPR-dependent manner. SCS3 was significantly upregulated while YFT2 was mildly upregulated in WT cells treated with Tm (Fig. 1B). As depleting the medium of inositol induces the UPR through LBS (Cox et al., 1997; Halbleib et al., 2017; Ho et al., 2020), we measured the mRNA levels of both genes upon inositol depletion (− ino) and only SCS3 exhibited a mild, but significant, upregulation. To further assess the role of SCS3 during ER stress, we carried out a growth assay. The spotting assay revealed that the lack of SCS3 causes a growth defect in the presence of Tm which can be rescued with the overexpression (OE) of SCS3 (Fig. 1C). Together, these results demonstrate that SCS3 is essential during ER stress conditions.

As lipogenic pathways constitute one of the major effectors of the UPR (Cretenet et al., 2010) and LDs are associated with stress conditions, the putative function of Scs3 in LD formation might then provide a rationale for its UPR-dependent transcriptional upregulation. We further hypothesized that the inability of scs3Δ mutants to mount a maximal UPR under proteotoxic stress conditions may translate to the impairment of LD formation as part of the stress response. We made use of fluorescent BODIPY 493/503 to stain LDs in scs3Δ mutants challenged with Tm-induced stress to evaluate gross changes in LD formation. The mutant cells were still able to form LDs to the same extent as WT cells under unstressed and ER stressed conditions (Fig. 1D, Fig. S1D). Next, we sought to determine the extent by which the UPR compensates for the absence of SCS3, particularly in respect to LD formation. However, a scs3Δire1Δ mutant is synthetically lethal, thus rendering conventional deletion strategies unusable in titrating the phenotypic effects of the UPR upon the loss of SCS3. To address this, we generated a conditional temperature-sensitive scs3 allele (scs3-1) that is functional at a permissive temperature of 25°C but not at the restrictive temperature of 37°C (Fig. 1E, Fig. S2), using a screening strategy that we previously reported (Thibault et al., 2011, 2012; Thibault and Ng, 2011). As IRE1 is not essential in the absence of YFT2 (Fig. S2B), a temperature-sensitive allele of YFT2 was not needed. This provides strong support for the interdependence between IRE1 and SCS3, and that each is required for viability in the absence of the other, consistent with SGA results showing synthetic lethality between these two genes (Moir et al., 2012).

In scs3Δ and yft2Δ mutants, LDs remain irreversibly tethered to the ER and are wrapped by a membrane. Jacquier et al. suggest that, in yeast, LDs always remain connected to the ER (Jacquier et al., 2011). We hypothesized that while the knockout strains we used are fully capable of forming LDs (Fig. 1D), it is possible that these LDs remain tightly integrated in the ER membrane, resulting in the disruption of ER lipid homeostasis, as previously reported (Choudhary et al., 2018, 2015). Strains with temperature-sensitive alleles were grown to early log-phase at 25°C followed by a temperature shift of 2 h at 37°C. To assess the levels of neutral lipids in the ER, we extracted total lipids from microsomes of the cells, and TAGs were separated by thin layer chromatography (TLC) and quantified by gas chromatography with a flame ionization detector (GC-FID) (Fig. 2A; Fig. S3). A significant increase in TAGs within the microsomal fractions of scs3-1 strain was only observed in the absence of IRE1 at the restrictive temperature of 37°C, thereby suggesting that the presence of the UPR exerts a suppressive effect for this phenotype. As it has been previously shown that LDs are wrapped with ER membranes in scs3Δ and yft2Δ yeast mutants using the inducible TAG synthesis system (Choudhary et al., 2015), we asked whether the UPR plays a role in the budding and morphology of LDs. Transmission electron microscopy (TEM) was performed on scs3-1 and ire1Δscs3-1 cells following a temperature shift. Strikingly, we observed irregular LD morphology in scs3-1 and ire1Δscs3-1 cells (Fig. 2B; Fig. S4). The abnormally elongated LDs were embedded in ER. Additionally, LDs were smaller in ire1Δscs3-1 cells compared to those in scs3-1 cells at 37°C. Taken together, our findings suggest that Scs3 is necessary for LDs to form on the exoleaflet of the ER and that the UPR plays an important role in regulating TAG levels.

The transcription of a subset of genes encoding phospholipid biosynthesis enzymes are inhibited in the presence of phospholipid precursors inositol and choline (Carman and Henry, 1999; Carman and Kersting, 2004; Henry and Patton-Vogt, 1998). Additionally, the absence of inositol in yeast growth medium induces the UPR (Cox et al., 1997; Halbleib et al., 2017; Ho et al., 2020). Therefore, we asked whether Scs3 plays a role in modulating phospholipid homeostasis in the absence of inositol and choline. We performed lipid analysis of cells grown in medium containing inositol and choline to mid-logarithmic phase before being shifted to a 3 h incubation in medium lacking inositol and choline. Unexpectedly, scs3Δ mutant cells contained four times the phosphatidylcholine (PC) levels of WT cells (Fig. 2C). We also tested scs3Δ pct1Δ cells because Pct1 is a cholinephosphate cytidylyltransferase enzyme, so these cells are unable to synthesize PC from the precursor choline through the Kennedy pathway. PC levels of scs3Δ cells were significantly decreased in the pct1Δ background mutant, suggesting that PC is mostly synthesized through the Kennedy pathway in the absence of SCS3. Phosphatidylserine (PS) and phosphatidylethanolamine (PE) were also found to be two times more abundant in scs3Δ mutant compared to WT cells, while there was no significant difference in phosphatidylinositol (PI) levels. Next, we tested whether supplementing the medium with inositol and choline would restore phospholipid levels in the scs3Δ mutant. In contrast to what was found in the absence of both lipid precursors, we observed a decrease of about four times of PC in the scs3Δ mutant compared to WT cells (Fig. 2D). There was also a significant decrease of PI in the scs3Δ mutant. To further understand the role of Scs3 in modulating phospholipids, we measured the levels of PC and PI over the course of 3 h of cells grown in media depleted of inositol followed by a 0.5 h recovery (+ ino) period. In the presence of choline, scs3Δ cells failed to increase the synthesis of PI upon the re-introduction of inositol while PI level increased rapidly in WT (Fig. 2E). On the other hand, there was a constant decrease of PC levels in scs3Δ while PC levels continually increased in WT during the 3 h of inositol depletion. Together, these data reveal that Scs3 is essential to maintain phospholipid homeostasis. We speculate that the decrease of PI in scs3Δ cells might alter the composition of complex sphingolipids (Guan et al., 2009), which in turn would induce the UPR (Liu et al., 2012).

To gain further insight into the physiological relevance of ScFIT proteins within the cell, we employed the split-ubiquitin-based membrane yeast two hybrid (MYTH) screen (Snider et al., 2010). The reporter moiety was fused to the N-terminal or the C-terminal cytosolic domains of both Scs3 and Yft2. The four bait constructs were validated, and screening conditions for each were optimized with 3′-amino-1,2,4,-triazole (3′-AT) supplementation in the selection medium to reduce the occurrence of false positives. Following this, 1344 colonies were collectively screened for all reporter strains. From these, 664 colonies were positive for bait–prey interaction as manifested by blue colony growth on 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal)-supplemented selective medium and were designated as putative interactors (Fig. S5). These were further validated for specificity towards the bait protein of interest in comparison to the single-pass human cluster of differentiation 4 (CD4) receptor protein, which served as a negative control. From these, 189 showed specific interactions with the ScFIT proteins. Following sequence analysis, 88 unique protein interactors were identified. Considering the overlap in protein interactors, the MYTH screen identified a total of 73 genuine and unique interactors for the ScFIT proteins, which were further categorized according to cellular functions (Fig. 3). Moreover, our screen results show that more than half of the identified Yft2 protein interactors are shared with those of Scs3, thereby supporting a certain degree of functional redundancy between the two proteins.

Surprisingly, only few of the ScFIT protein interactors identified with the screen are directly involved in lipid metabolism, suggesting that ScFIT proteins may not function extensively in that cellular process. It should also be noted that, while the encoded proteins of genes that ScFIT had high degrees of genetic interactions with, such as ICE2, SEY1 and UBX2 (Moir et al., 2012; Tavassoli et al., 2013), were not identified with the MYTH screen, this does not exclude the possibility of a physical interaction. However, our results alternatively suggest that, despite having a high degree of genetic interaction, these proteins function in parallel but independently in the same cellular process, and that the loss of both is detrimental to cell viability.

Interestingly, several proteins that function in proteostasis and the ubiquitin-proteasome system (UPS) have been found to interact with Scs3 (Fig. 3, Fig. S5). Of note are the J-protein chaperone Zuo1 and the Hsp70 chaperone Ssb2, both of which have been reported to function in protein quality control (Allen et al., 2007; Chiabudini et al., 2012; Ohba, 1994). Doa10 is one of the key E3 ubiquitin ligases in yeast, and is involved in ER-associated degradation (ERAD) of proteins (Habeck et al., 2015; Ravid et al., 2006; Swanson et al., 2001). Taken together, these results suggest that Scs3 may function to a certain extent in protein quality control pathways, specifically in the UPS.

As the accumulation of misfolded proteins and the ensuing proteotoxicity is closely related to the ability of cells to efficiently process client proteins, we hypothesized that the sensitivity to Tm exhibited by scs3Δ and ScFITΔ mutants may be the result of impaired protein degradation pathways, as in the case of ubx2Δ mutants, which have impaired turnover of both misfolded ER and cytosolic substrates (Spear and Ng, 2003). Following the identification of UPS machinery components as protein interactors of Scs3, we asked whether ScFIT proteins play a role in protein ubiquitylation. We overexpressed Myc-tagged ubiquitin (Ub-Myc) through the inducible CUP1 promoter and quantified the extent of total protein ubiquitylation in cells by immunoblotting. The amount of ubiquitylated proteins in ScFITΔ mutants was significantly reduced, by ∼38%, relative to that of WT cells, with a less-pronounced reduction of 18% in scs3Δ single mutants (Fig. S6A). Since the stabilization of protein substrates was accompanied by a decrease, rather than an increase in high molecular mass ubiquitin antibody-reactive proteins, the inefficient turnover of the said substrates is likely due to a failure to mark them correctly for degradation and not because of efficient clearance in the proteasome.

To investigate whether the global decrease of ubiquitylated proteins correlates with protein stability, we measured the turnover of known ERAD substrates in ScFIT mutants by means of a cycloheximide chase assay. We expressed HA-tagged Sbh2, Yeh1 and Pgc1 in ScFITΔ mutants. These native proteins are dependent on Doa10-mediated ERAD for normal degradation (Habeck et al., 2015; Ruggiano et al., 2016). The turnover rates of Sbh2 and Yeh1 in ScFITΔ mutants was similar to that of WT (Fig. S6B,C). In contrast, the degradation of Pgc1 was significantly accelerated in ScFITΔ mutants (Fig. S6D). Along with the identification of Pgc1 as a Doa10-dependent ERAD substrate, its proper localization dynamics between the ER and LD membranes was found to be critical in determining its stability (Ruggiano et al., 2016). Doa10 reportedly recognizes ER-localised Pgc1 through its hairpin loop, which then serves as a degron that concentrates Pgc1 on the surface of LDs. As LDs fail to properly mature in the absence of the ScFIT proteins (Fig. 2B), the lateral diffusion of the pool of Pgc1 proteins to the ER may be increased in ScFITΔ mutants, resulting in continual degradation by Doa10 (Kory et al., 2016). We hypothesized that native proteins in their proper conformation, like Pgc1, may not illicit a proteotoxic effect on ScFITΔ cells, and that an otherwise compromised protein degradation pathway in this mutant could remain fully capable of clearing these endogenous proteins.

As misfolded model substrate, we monitored the protein levels of epitope-tagged versions of misfolded CPY (CPY*–HA) (Fig. 4A). A small but significant delay in the degradation of CPY* was only observed in ScFITΔ mutants and not in scs3Δ nor yft2Δ. Next, we measured the degradation rates of the engineered misfolded variant of the Pep4 vacuolar protease (ngPrA*Δ295-331–HA) (Kanehara et al., 2010) (Fig. 4B). In contrast to CPY*, ScFITΔ cells exhibited a strong defect in the degradation of ngPrA*Δ295-331–HA in comparisons to WT and single mutants. Both CPY*–HA and ngPrA*Δ295-331–HA are luminal soluble substrates, which are degraded in a Hrd1-dependent manner (Kanehara et al., 2010; Thibault et al., 2011). To further assess whether the global decrease of ubiquitylated proteins is associated with ScFIT, we monitored the degradation of San1-dependent cytosolic protein quality control (CytoQC) substrates Δ2GFP–HA and ssPrA–HA (Prasad et al., 2010). Consistent with our results using misfolded ERAD substrates, we found that ScFITΔ mutants are unable to efficiently clear away both cytosolic substrates compared to WT cells or either of the single mutants (Fig. S6E,F). As neither of the single deletion mutants resulted in a stabilization of the ERAD substrates, the two proteins may share a redundant yet poorly understood function. YFT2 is reported to have been the result of the segmental duplication of SCS3 (Moir et al., 2012). This is supported by the more-pronounced growth sensitivity to Tm in the double ScFITΔ mutant in comparison to a mild defect in scs3Δ cells (Fig. 1C). This, along with the broader range of Scs3 protein interactors (Fig. 3), also suggests an asymmetric redundancy wherein YFT2 only partially compensates for the absence of SCS3 functionality in the ERAD pathway, which ultimately results in less-apparent phenotypic defects in yft2Δ mutants.

To further investigate the role of lipid homeostasis and protein quality control, we monitored the degradation of CPY* in ScFITΔ mutant strains supplemented with inositol (+ino, −cho), choline (−ino, +cho), or both (+ino, +cho). There was a significant defect in the degradation of CPY* in scs3Δ and yft2Δ supplemented with choline compared to inositol (Fig. 4C,D). Similarly, the degradation of CPY* was slower in scs3Δ and ScFITΔ compared to WT in the presence of choline. Next, to validate the role of Scs3 in modulating ERAD, we monitored the degradation of ngPrA*Δ295-331–HA in ScFITΔ cells overexpressing (OE) SCS3. In ScFITΔ, ngPrA*Δ295-331–HA was degraded at a significantly slower rate than in WT cells (Fig. 5A). On the other hand, the degradation of ngPrA*Δ295-331–HA was similar in ScFITΔ OE SCS3 to WT and ScFITΔ strains, suggesting that SCS3 is sufficient to rescue the ERAD defect. In the presence of choline (− ino,+cho), the degradation of ngPrA*Δ295-331–HA was decelerated in the three strains (Fig. 5B). Interestingly, OE SCS3 in ScFITΔ cells supplemented with choline (− ino,+cho) significantly accelerated the degradation of CPY*-HA compared to WT and ScFITΔ (Fig. 5C). As lipid homeostasis correlates with ERAD fitness (Shyu et al., 2019; Thibault et al., 2012), these findings reinforce the notion that ScFIT is essential to regulate lipid levels at the ER and that these proteins contribute to ER proteostasis.

LDs have been increasingly implicated in disease pathophysiology. Despite this, our understanding of their involvement is obscure at best, as LD biology is still in its infancy, and more mechanistic insight into LD formation is warranted to grasp its relevance and importance in physiological processes. From the simple budding yeast, several proteins have been identified to influence LD generation (Adeyo et al., 2011; Cartwright et al., 2015; Szymanski et al., 2007). Among these, the FIT2 class of proteins has gained much interest in recent years, but its initial putative role in LD formation as a lipid-binding protein has recently been contested in favor of a broader function in membrane homeostasis. However seemingly disparate, the identification of lipid phosphatase activity in FIT2 may not be mutually exclusive with previous reports of its involvement in LD biogenesis. Given this, the molecular mechanism by which these two processes are linked is poorly understood, as well as the potential implication of FIT2 in the normal functioning of cells outside the context of LD formation. In this study, we report on the involvement of the yeast FIT homologs (ScFIT) not only in the maintenance of ER membrane homeostasis, but also in coordinating the cellular stress response pathway, namely the UPR, and the consequent impact on protein quality control (Fig. 6).

The complexity of lipid metabolic pathways is underscored by the highly interconnected conversion of intermediates as well as the various organelles and proteins that mediate these processes (Henry et al., 2012; Klug and Daum, 2014). In addition to this, perturbation of the lipid metabolic pathways results in the extensive reprogramming of the bioenergetic network (Natter and Kohlwein, 2013; Stordeur et al., 2014). Similarly, cellular insults also alter the lipidomic landscape of cells, suggestive of the buffering capacity of lipid pathways against stress conditions. Proteotoxic or LBS both activate the UPR and similarly culminate in the formation of LDs. Interestingly, none of the previously reported major protein effectors of LD biogenesis were identified as UPR targets. Moreover, apart from the SE biosynthetic ARE2, no other NL synthesis players are upregulated under conditions of ER stress (Thibault et al., 2012; Travers et al., 2000). Therefore, Scs3 upregulation under the UPR program could be part of the effort to orchestrate membrane remodeling.

It was reported that LDs in ScFITΔ cells remain embedded in the ER due to the enrichment of DAG, a lipid species with negative membrane curvature, as is the case with the accumulation of PE in a mutant of CHO2, a methyltransferase for PC synthesis (Choudhary et al., 2018). Interestingly, the addition of either of the positive curvature phospholipids, lyso-PC or lyso-PA, rescued the aberrant LD budding of ScFITΔ cells. The failure of the UPR to restore proper LD maturation in cho2Δ cells may result from the markedly reduced capacity to synthesize PC, which exhibits a neutral curvature and is an intermediate to lyso-PC (van Meer et al., 2008). While cho2Δ cells indeed accumulated high levels of PE, DAG levels are dramatically reduced (Thibault et al., 2012). This, taken together with the observation that the gain-of-function DAG-binding FIT mutant failed to rescue the aberrant ER membrane whorling in scs3Δ cells (Becuwe et al., 2020), strongly suggests that the defects in ER membrane properties that lead to impaired LD maturation are independent of DAG. In contrast, we observed an accumulation of TAG at the ER, together with irregularly shaped LDs that were detached to the ER in scs3-1 (Fig. 2B; Fig. S4), reinforcing the idea that Scs3 has a role in lipid homeostasis. While the catalytic activity of mammalian FIT2 on PA and lyso-PA was not identified in ScFIT proteins (Hayes et al., 2017), this does not exclude the possibility that the latter may instead act on other membrane lipid species in vivo. In a recent paper, mammalian FIT2 is proposed to be a lipid phosphate phosphatase enzyme based on in vitro evidence (Becuwe et al., 2020). As SCS3 is essential for viability in the absence of a functional UPR (Fig. 1), it can then be hypothesized that Scs3 regulates ER membrane lipid composition. Taken together, these findings highlight the important role of Scs3 in maintaining ER lipid homeostasis beyond LD biogenesis. A unifying model on the role of Scs3 in vivo with it acting as an enzyme that catalyzes lipid synthesis, regulates other enzymes or regulates lipid metabolism through other ways should emerge in future studies.

The loss of both FIT homologs in ScFITΔ mutants led to the unexpected stabilization of misfolded proteins in both within the ER and in the cytoplasm (Figs 4 and 5; Fig. S6E,F), which correlated with a global decrease in protein ubiquitylation (Fig. S6A). The maintenance of ER membrane integrity and lipid homeostasis are critical in supporting organellar function. Loss of ICE2 causes altered ER membrane dynamics, including defects in mother–daughter cell ER membrane inheritance and ER–plasma membrane tethering (Estrada de Martin et al., 2005; Loewen et al., 2007). This ER membrane perturbation further impaired cellular functions, such as phospholipid regulation and protein degradation (Markgraf et al., 2014; Quon et al., 2018; Schuldiner et al., 2005; Tavassoli et al., 2013). Similarly, mutants of the phospholipase Lpl1, which catalyzes the turnover of phospholipids, also exhibited ERAD defects (Selvaraju et al., 2014; Weisshaar et al., 2017). The general disruption of lipid metabolism by attenuating fatty acid synthesis in turn caused defects in processing of ERAD client proteins in mammalian systems (To et al., 2017), which may also be in part due to its indirect effects on membrane lipid composition. Conversely, defective protein turnover also exerted a direct effect on membrane composition. The deletion of the ERAD component UBX2 led to severe changes in ER membrane morphology due to dysregulation of Mga2 processing and the subsequent expression of its transcriptional target OLE1, a key regulator of membrane lipid saturation (Surma et al., 2013). Sterol content within membranes is also under tight control by protein quality control pathways, as the key enzymes Hmg2 and Erg1 are regulated in an ERAD-dependent mechanism (Foresti et al., 2013; Hampton et al., 1996). Intriguingly, these mutants have aberrant membranes and exhibit impaired LD formation in addition to curtailed protein turnover (Markgraf et al., 2014; Wang and Kaufman, 2012; Weisshaar et al., 2017).

In a previous study, we have shown that the singular UPR transducer Ire1 in yeast is strongly activated upon genetic alterations of ER membrane composition (Thibault et al., 2012). We also identified a LBS-sensing switch located at the interface of the amphipathic and transmembrane helices (Ho et al., 2020) while key residues within the amphipathic helix of Ire1 were reported to be important to sense LBS- and proteotoxic-induced ER stress (Halbleib et al., 2017). As ER membrane morphology is compromised in ScFIT mutants, it could be hypothesized that proper Ire1 and UPR activation may not proceed as efficiently, which in turn affects ERAD function. In line with this, several studies lend support for the modulation of LDs by protein quality control pathways including ERAD. The ubi4Δ mutant was reported to exhibit less LD accumulation compared to that of WT cells under stress (Ishii et al., 2018). While the rationale for this increase in LD remains enigmatic, it suggests that LD formation may in part be regulated by ubiquitylation processes. This agrees with previous studies that detailed on the dependence of the NL biosynthetic enzyme Dga1 and SE lipase Yeh1 on Doa10 for their endogenous turnover (Ruggiano et al., 2016), and that the recruitment of the mammalian ERAD factor UBXD8 onto the LD surface regulates LD growth by modulating lipolysis (Olzmann et al., 2013). Apart from LDs, ERAD pathways also regulate the ER membrane composition and phospholipid turnover. The Cdc48 ATPase mediates the processing of the ER membrane sensors Mga2 and Spt24, to yield the cognate transcription factor for OLE1 regulation (Kandasamy et al., 2004; Shcherbik and Haines, 2007; Surma et al., 2013), and the degradation of phosphorylation-inactive Pah1 is impaired in proteasome and ubiquitylation mutants (Hsieh et al., 2015; Pascual et al., 2014). Taken together, these greatly emphasize the interdependence of membrane homeostasis and protein quality control pathways.

In this study, we build on the current hypothesis on the role of ScFIT proteins in LD formation and membrane homeostasis, and further provide support for its functioning in cell stress response pathways to exert effects on these two processes (Fig. 6).

Saccharomyces cerevisiae strains used in this study are listed in Table S1. Strains were generated using standard cloning protocols. Anti-HA mouse monoclonal antibody HA.11 (1:2000; Covance MMS-101R-1000), anti-tubulin mouse monoclonal antibody 12G10 (1:10,000; DHSB), and anti-Myc mouse monoclonal antibody (1:5000; Invitrogen R950-25), were commercially purchased. Secondary antibodies goat anti-mouse IgG-DyLight 488 (Thermo Fisher 35503, Waltham, MA), goat anti-mouse IgG-IRDye 800 (LI-COR Biosciences 926-32210) and goat anti-rabbit IgG-IRDye 680 (LI-COR Biosciences 926-68021) were commercially purchased.

Plasmids and oligonucleotide primers used in this study are detailed in Tables S2 and S3, respectively. Plasmid constructs were generated through either conventional restriction enzyme cloning methods or Gibson assembly (New England Biolabs). The mutant scs3 library was generated by low-fidelity PCR using primers PS1 and PS2 to amplify the promoter, coding sequence and terminator regions of the SCS3 from the genomic DNA of wild-type (WT) cells. The PCR product was then digested with the enzymes EcoRI and XbaI before ligation into pGT0004. Plasmid pGT0364 was obtained through a colony sectoring screen detailed in the ‘Genetic screen for temperature sensitive alleles’ section below. Plasmid pGT0286, encoding for WT SCS3, was similarly generated using conventional PCR amplification. To generate reporter constructs for the membrane yeast two-hybrid screen, the coding sequences of SCS3 and YFT2 were amplified from WT yeast DNA using primer pairs PS39- and PS40, and PS159 and PS160, respectively. These were then inserted via Gibson assembly into vector backbones generated through PCR from pGT0317 using primer pairs PS39 and PS40, and PS157 and PS158, respectively, to generate pGT0374 and pGT0427. Plasmids pGT0426 and pGT0428 were generated through Gibson assembly by amplifying the coding sequence of SCS3 and YFT2, terminating immediately before the stop codon using WT yeast DNA with primer pairs PS107 and PS108, and PS101 and PS102, respectively. These were then cloned into PCR-amplified vector backbones using pGT0318 as template with primer pairs PS105 and PS106, and PS36 and PS99, respectively.

Strains were grown to saturation in appropriate selective medium overnight at 30°C (or at 25°C for the temperature sensitive strains). Cultures were diluted to 0.2 OD₆₀₀/ml and serially diluted five-fold for a total of four dilutions. The cell suspensions were then spotted onto appropriate agar plates and incubated at indicated temperatures until the appearance of colonies.

where t is time, and V_A and V_R represent the actual volume assayed and the volume used to measure OD₆₀₀, respectively.

Cells were grown to an early log phase overnight at 30°C. Tunicamycin was added to a final concentration of 2.5 µg/ml and incubated 1 h at 30°C or depleted of inositol for 2 h, when indicated. Total RNA was extracted using an RNeasy Mini Kit (Qiagen) following the manufacturer’s protocol. DNase treatment in columns was carried out with RNase-free DNase (Qiagen, Venlo, Netherlands) following the manufacturer's protocol. cDNA was synthesized from 2 μg of total RNA using RevertAid reverse transcriptase (Thermo Fisher, Waltham, MA) following the manufacturer's protocol. SYBR Green qPCR experiments were performed following the manufacturer's protocol using a QuantStudio 6 Flex Real-time PCR system (Applied Biosystems, Waltham, MA). cDNA (30 ng) and 50 nM of paired primer mix was used for each reaction. Relative mRNA was determined with the comparative Ct method (ΔΔCt) normalized to housekeeping gene ACT1. Oligonucleotide primers used are listed in Table S3.

Cells were grown to early log phase, and 500 µl of the suspension was transferred on a coated slide with 10 mg/ml concanavalin A (Sigma-Aldrich, St Louis, MO) mounted onto an Attofluor cell chamber (Thermo Fisher, Waltham, MA) and imaged at room temperature. Tunicamycin was added to a final concentration of 2.5 µg/ml and incubated at 30°C for 1 h, when indicated. To stain LDs, cells were incubated with 0.05 µg/ml BODIPY 493/503 (Invitrogen) in phosphate-buffered saline (pH 7.4) for 10 min at room temperature, washed and resuspended in liquid medium before transferring into the Attofluor cell chamber for viewing. Samples were imaged with a Leica DMi8 system (HCX PL APO 100×/1.4–0.70 NA oil immersion objective) under the control of Metamorph ver. 7.8.10.0, or a Zeiss LSM710 microscope (100×1.4 NA Plan-Apochromat oil-immersion objective) under the control of Zen software (Carl Zeiss MicroImaging).

The genetic screen was performed as previously reported (Thibault et al., 2011). A mutant library of the SCS3 open reading frame flanked by 500 bp of its endogenous promoter and 300 bp of its terminator was generated by low-fidelity PCR using Taq DNA polymerase in the presence of 0.05 and 0.1 mM MnCl₂. DNA fragments were digested with EcoRI and XbaI and ligated into digested pRS316 to produce a plasmid library scs3* with random point mutations. The strain YGT0492 was transformed with the mutant library pRS316-scs3* and transformants were spread on selecting synthetic complete medium lacking uracil (SC −Ura) plates with limiting adenine (low Ade) at 6 µg/ml. Plates were incubated at 25°C until colonies developed fully red pigmentation due to low Ade. Colonies with a sectoring phenotype were streaked in duplicate on SC −Ura, low Ade, and incubated at 25°C and 37°C. For the primary screen, 123 colonies were screened for positive clones, which sectored at 25°C but remained red at 37°C. From the positive clones, the cells with a sectoring phenotype at 25°C were re-streaked in duplicate on SC-Ura, low Ade, and incubated at 25°C and 37°C to eliminate false positives. Positive clones were isolated without the plasmid pDN388. Plasmids were extracted from the clones and subjected to DNA sequencing analysis to identify the mutation present in the scs3 temperature-sensitive (ts) alleles. The plasmid pGT0364, containing scs3 ts allele (scs3-1) encodes for Scs3 with the following mutations D277G and I328V. The plasmid pGT0364 was transformed in strain YGT0492.

Cells were grown to early log phase at 25°C followed by a 2 h incubation at 37°C. For whole-cell lipid extraction, 10 OD₆₀₀ of cells was washed, pelleted in a glass vial and lyophilised using a Virtis freezer dryer under vacuum. All subsequent steps were carried out at 4°C. For lipid extraction of microsomes, 50 OD₆₀₀ of cells was pelleted and resuspended in lysis buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 1 mM PMSF and 1:200 dilution protease inhibitor cocktail (PIC, Sigma P8215)] and lysed mechanically by 15 times of 30 s interval using 0.5 mm zirconium beads at maximum speed of a vortex mixer. The supernatant was collected by spinning down the lysate 5 min at 800 g. The clarified lysate was spun down 1 h at 100,000 g. The pellet was resuspended in 100 µl ddH₂O and sonicated for 30 min before quantifying total protein using the bicinchoninic acid (BCA) protein quantification assay (Sigma-Aldrich). A volume corresponding to 0.5 mg of total protein (Klug and Daum, 2014) was transferred into glass vials and lyophilized using a Virtis freeze dryer under vacuum to record the dry weight of each sample. For lipid extraction from whole cells, samples were resuspended in 100 µl ddH₂O. Afterwards, 300 µl of 0.5 mm zirconium beads and 900 µl of chloroform (CHCl₃):methanol (2:1) were added before rigorous agitation of 2 h at 4°C. From here, 300 µl each of CHCl₃ and ddH₂O were added to the mixture and vortexed 15 s twice. The vials were centrifuged 6 min at 4250 g, and the lower organic phase was transferred to a new glass vial. The extraction step was repeated by the addition of 500 µl of CHCl₃ and further agitation for 2 h. Lipid extraction from microsomes was performed similarly with scaled-down reagent volumes. Combined extracts were concentrated, resuspended in 100 µl CHCl₃:methanol (2:1), and 30 µl was spotted on HPTLC Silica gel 60 plates (Merck Millipore) using Linomat 5 (CAMAG). Triacylglycerol (TAG) was separated with hexane:diethyl ether:acetic acid (75:25:2) and visualized under long-wave ultraviolet light (320 nm) by spraying 0.05 mg/ml of primuline dye in acetone:water (80:20) onto the dried plates.

Spots corresponding to TAG were scraped off the silica plates and transferred into glass vials. A total of 100 µl 1 mM pentadecanoic acid (C15:0) was added to the tubes as internal standard. TAGs were hydrolyzed and esterified to fatty acid methyl esters (FAME) with 300 µl of 1.25 M HCl-methanol for 1 h at 80°C. FAMEs were extracted three times with 1 ml of hexane. Combined extracts were dried under nitrogen, and resuspended in 20 µl of hexane. FAMEs were separated by gas chromatography with a flame ionisation detector (GC-FID; GC-2014, Shimadzu, Kyoto, Japan) using an ULBON HR-SS-10 50 m×0.25 mm column (Shinwa, Tokyo, Japan). Supelco 37 component FAME mix was used to identify corresponding FAs (Sigma-Aldrich, St Louis, MO). Data were normalized using the internal standard C15:0.

Samples for transmission electron microscopy (TEM) were prepared as previously described (Wright, 2000). One OD₆₀₀ unit of early log-phase cells grown at 25°C or 37°C was collected and pre-fixed with glutaraldehyde overnight at 4°C. Post-fixation was performed in the presence of 2% potassium permanganate for 1 h at room temperature. After dehydration in ethanol, cells were infiltrated with Spurr's resin and incubated for 24 h at 60°C to allow polymerization. Silver-gray sections were prepared using Ultracut UCT (Leica) equipped with a diamond knife and stained with lead citrate. Micrographs were taken using a transmission electron microscope (Joel JEM-1230).

Cells were grown to mid-log phase overnight in medium containing 75 µM inositol with (+cho) or without (−cho) 1 mM choline in the presence of 10 µCi/ml [³²P]orthophosphate. Cells collected by filtration were resuspended in medium with or without inositol or choline, as indicated, and in the presence of 10 µCi/ml [³²P]orthophosphate. Cells were harvested after 3 h following the shift or at indicated time point. Labeled lipids were extracted as previously described (Gaspar et al., 2006). The individual phospholipid species were resolved by two-dimensional thin layer chromatography. Phospholipids were separated using the solvent system chloroform:ethanol:water:triethylamine (30:35:7:35) for at least 2 h. Phospholipid identity was based on the mobility of known standards and quantified on a STORM 860 PhosphorImager (Amersham Biosciences).

The bait plasmid for the expression of Scs3 and Yft2 fused with the Cub-LexA-VP16 reporter tag at either N- or C-terminus under the control of a TEF1 promoter was generated by cloning the respective coding sequences into the pTLB-1 or pTMBVα plasmid (Snider et al., 2010). Expression of the ScFIT bait proteins in the NMY51 reporter strain was verified with immunoblotting using anti-LexA antibodies (1:5000; Abcam ab50953) and anti-rabbit IgG IRDye 680 (1:15,000 dilution). Following verification of bait construct expression in the NMY51 reporter strain, the functionality of the reporter system was validated by co-expression of the bait construct with either a positive or negative control prey protein. Growth medium for bait constructs that activate the reporter with the negative control prey protein was adjusted for stringency using His3 competitor 3′-amino-1,2,4,-triazole (3-AT). The final concentrations of 3-AT supplemented to the growth medium, which yielded minimal self-activation, were 50 mM and 10 mM 3-AT for N-terminally tagged Scs3 and Yft2, respectively. The reporter strains bearing the bait constructs were then transformed with the NubG-X prey cDNA plasmid library (DualSystems), plated and incubated at 30°C until colonies appeared. Colonies were randomly selected and plated on selective plates supplemented with X-Gal for each screen. Plasmids from blue-colored colonies, indicative of positive bait–prey interaction, were recovered and amplified in DH5α competent bacteria cells. This was followed by plasmid extraction and sequencing. To reduce false-positive interactions, prey constructs were screened once more for specific interaction by retransformation back into yeast strains bearing the original bait protein. This was performed in comparison to a yeast strain expressing an unrelated negative control bait construct encoding for the human CD4 T-cell surface glycoprotein. Only prey constructs that exclusively activate the reporter with the bait construct of interest are included in the final list of interactors.

The cycloheximide chase assay was performed as previously described (Prasad et al., 2010). In brief, 6 OD₆₀₀ units of early log phase cells were grown in appropriate selective medium. To induce lipid perturbation, 1 mM of choline chloride was added a day prior to harvesting, whereas inositol depletion was performed 2 h beforehand. Protein synthesis was inhibited by adding 200 μg/ml cycloheximide. Samples were taken at designated time points and trichloroacetic acid (TCA) was added to a final 10% volume. Cells were mechanically disrupted with 300 µl of 0.5 mm zirconium beads at 6500 rpm for 2×30 s using a tissue homogeniser (Precellys 24, Bertin Instruments). Precipitated proteins were pelleted 10 min at 21,000 g, 4°C, and resuspended in 40 µl of TCA resuspension buffer (100 mM Tris-HCl pH 11, 3% SDS, 1 mM PMSF and PIC). Solubilized proteins were separated by SDS-PAGE, transferred on nitrocellulose membranes. Immunoblotting was performed with appropriate primary antibodies and IRDye-conjugated secondary antibodies. Proteins were visualized using the NIR fluorescence system (Odyssey CLx Imaging System). Values for each time point were normalized using anti-Tub1 as loading controls. Tonal quality was adjusted for representative images through ImageStudio Lite Version 5.2 (LI-COR Biosciences) where appropriate and was followed by quantification. All comparative analyses were performed on immunoblots performed in parallel using samples derived from the same experiment.

Strains expressing Myc-tagged ubiquitin (Ub–Myc) under the control of the inducible CUP1 promoter were grown to early log phase. The culture medium was supplemented with a final concentration of 100 µM Cu₂SO₄, and cells were incubated for 3 h to allow Ub–Myc expression. Proteins were extracted from 2 OD₆₀₀ units of cells and separated on SDS-acrylamide gels. Following transfer on nitrocellulose membranes, total protein on each lane was stained with REVERT Total Protein Stain (LI-COR Biosciences) followed by visualization on the Licor Odyssey CLx system. This was then followed by membrane blocking and incubation with antibodies against the Myc epitope tag (1:1000 dilution), and anti-mouse IgG IRDye 800 secondary antibodies (1:15,000). Total Myc-tagged protein signal was normalized against the total protein signal present as quantified through ImageStudio Lite Ver 5.2 (LI-COR Bioscienes).

Error bars indicate the s.e.m., calculated from at least three biological replicates, unless otherwise indicated. P values were calculated using one-way ANOVA with Tukey's post hoc test, unless otherwise indicated and reported as P values with four significant digits in the figures. All statistical tests were performed using GraphPad Prism 7 software.

We are grateful to Dr Davis Ng for providing reagents. We thank members of Thibault lab for critical reading of the manuscript. Results and discussion segments in this paper are reproduced from the PhD thesis of Peter Shyu, Jr (School of Biological Sciences, Nanyang Technological University, 2019).

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.248526.reviewer-comments.pdf