Suppression of respiratory growth defect of mitochondrial phosphatidylserine decarboxylase deficient mutant by overproduction of Sfh1, a Sec14 homolog, in yeast

Interorganelle phospholipid transfer is critical for eukaryotic membrane biogenesis. In the yeast Saccharomyces cerevisiae, phosphatidylserine (PS) synthesized by PS synthase, Pss1, in the endoplasmic reticulum (ER) is decarboxylated to phosphatidylethanolamine (PE) by PS decarboxylase, Psd1, in the ER and mitochondria or by Psd2 in the endosome, Golgi, and/or vacuole, but the mechanism of interorganelle PS transport remains to be elucidated. Here we report that Sfh1, a member of Sec14 family proteins of S. cerevisiae, possesses the ability to enhance PE production by Psd2. Overexpression of SFH1 in the strain defective in Psd1 restored its growth on non-fermentable carbon sources and increased the intracellular and mitochondrial PE levels. Sfh1 was found to bind various phospholipids, including PS, in vivo. Bacterially expressed and purified Sfh1 was suggested to have the ability to transport fluorescently labeled PS between liposomes by fluorescence dequenching assay in vitro. Biochemical subcellular fractionation suggested that a fraction of Sfh1 localizes to the endosome, Golgi, and/or vacuole. We propose a model that Sfh1 promotes PE production by Psd2 by transferring phospholipids between the ER and endosome.

Plasmid YCp111-PSD1 to express PSD1 under the control of its native promoter was constructed as follows: A DNA fragment containing ORF with 5'-and 3'-flanking regions of PSD1 was amplified by PCR using primers PSD1-KpnI-f and SFH1-BamHI-r with total DNA of W303-1A as a template. This fragment was digested with KpnI and BamHI and cloned into the KpnI-BamHI site of YCplac111. Plasmids YEp181-PSD2, YEp181-DPL1, and YEp181-PSS1 to overexpress PSD2, DPL1, or PSS1 were constructed as follows: DNA fragment containing ORF with 5'-and 3'-flanking regions of PSD2, DPL1, or PSS1 was amplified by PCR using primers PSD2-HindIII-f and PSD2-SalI-r, PstI-DPL1-f and DPL1-SacI-r, or PstI-PSS1-f, and PSS1-SacI-r, respectively. These fragments were digested by HindIII and SalI, or PstI and SacI and cloned into the HindIII-SalI site or PstI-SacI site of YEplac181. Plasmid YCp33-PSS1-EGFP to express Pss1-EGFP fusion protein under the control of its native promoter was constructed as follows: DNA fragments were amplified by PCR using primers BamHI-PSS1-f and PSS1ctag-KpnI-r, or KpnI-EGFP-f and EGFP-SalI-r with total DNA of W303-1A or plasmid pEGFP as a template, respectively. These fragments were digested by BamHI and KpnI or KpnI and SalI. YCpTGAP111 was digested with SalI and HindIII to obtain DNA fragment of TDH1 terminator. These three digested fragments were cloned into the BamHI-HindIII site of YCplac33. YCp22-PSD2-FLAG to express Psd2-FLAG fusion protein was constructed as follows: DNA fragment was amplified by PCR using primers PSD2-HindIII-f and PSD2ctag-SalIr, SalI-FLAG-f and FLAG-GAPDHt-r, or FLAG-GAPDHt-f and GAPDHt-SpeI-r with total DNA of W303-1A, p3xFLAG-myc-CMVTM-26, or YCpTGAP111 as a template, respectively. The DNA fragments containing FLAG and TDH1 terminator were ligated by PCR using primers SalI-FLAG-f and GAPDHt-SpeI-r. This fragment and PSD2 fragment were digested by SalI and SpeI or HindIII and SalI, and cloned into the HindIII-SpeI site of YCplac22. Plasmids YEp195-MDM34 and YCp111-MDM34 to express MDM34 under the control of its native promoter were constructed as follows: A DNA fragment was amplified by PCR using primers SalI-MDM34-f and HindIII-MDM34-r with total DNA of BY4741 as template. This fragment was digested with SalI and HindIII, and cloned into the SalI-HindIII site of YEplac195 or YCplac111. Yeast cells were grown on minimal medium (0.17% yeast nitrogen base without amino acid and ammonium sulfate, 0.5% ammonium sulfate) containing 2% glucose (SD medium) or 4.4% lactate (pH 5.5) (SLac medium), or semi-synthetic lactate medium (4.4% lactate, 1.6% NaOH, 0.17% yeast nitrogen base without amino acid and ammonium sulfate, 0.5% ammonium sulfate, 0.3% yeast extract, 0.05% glucose, 0.05% CaCl 2 �2H 2 O, 0.05% NaCl, 0.06% MgCl 2 �6H 2 O, 0.1% NH 4 Cl, 0.1% KH 2 PO 4 , pH 5.5), with required nutrients at 30˚C. For growth curve analyses, cells precultured in SD medium were seeded at a starting OD 600 = 0.005 to the medium of interest. Cells were cultured at 30˚C and the growth curve was obtained with an automatically recording incubator TN1506 (Advantec).

Lipid analysis
For analyzing phospholipid composition, cells were grown in SD medium or semi-synthetic lactate medium to a final OD 600 between 1.0 and 2.0, harvested by centrifugation, and washed with ice-cold 0.15 M KCl. Cells were broken with glass beads in chloroform/methanol/water (2:4:1). For analysis of mitochondrial phospholipids, yeast spheroplasts were lysed gently by French pressure cell press (SLM instruments), and mitochondria were purified by sucrose density gradient centrifugation by the method of Zinser and Daum [1]. Total lipids were extracted from organic layer by the method of Bligh and Dyer [43]. The extracted lipids were separated by two-dimensional thin layer chromatography (TLC) as previously described [44]. The spots corresponding to phospholipids were scraped from TLC plates. Phosphorus assay was done for each of phospholipid species according to the method of Bartlett [45].
For the analysis of Sfh1-binding lipids, W303-1A expressing Sfh1-ZZ, Sfh1 S175I,T177I -ZZ, or non-tagged Sfh1 were grown in SD medium to a final OD 600 between 1.5 and 2.0, and harvested by centrifugation. Cells were suspended in Lysis buffer (25 mM HEPES (pH 7.5), 100 mM KCl, 10%(w/v) Glycerol, 1 mM DTT, Protease Inhibitor Cocktail for Fungal and Yeast cells (Sigma)), and disrupted with glass beads using Multi-Beads Shocker (Yasui Kikai). Cytosolic fractions were obtained by stepwise centrifugation at 13,000 g for 10 min at 4˚C, and at 100,000 g for 60 min at 4˚C twice (S100 fraction). The S100 fractions (4 mg protein) were incubated with IgG-Sepharose TM 6 Fast Flow (GE Healthcare) beads overnight at 4˚C. Beads were washed with ice-cold lysis buffer and lipids were extracted from the beads by the method of Bligh and Dyer with short-chain phospholipids, diC10 PC, diC8 PE, diC8 PI, or diC12 PS (Avanti Polar Lipids) (0.35 nmol each) for internal standard. Phospholipid species were analyzed by ESI-MS/MS according to the method previously described [44,46,47]. For ionization efficiency correction, mixture of equal molar of short-chain phospholipids used for internal standards with DOPC, DOPE, SoyPI, or DOPS was subjected to ESI-MS/MS analysis. Ionization rates of DOPC, DOPE, SoyPI, and DOPS against short-chain phospholipids were 92 ± 5.5%, 22 ± 1.0%, 23 ± 0.7%, and 26 ± 5.6%, respectively, and employed as correction factors.

Protein expression and purification
Recombinant His 8 -Sfh1 and His 8 -Sfh1 S175I,T177I mutant were expressed in Escherichia coli BL21 (DE3) strain and purified as described previously [48] with several modifications. E. coli cells were lysed by French pressure cell press in lysis buffer containing 25 mM HEPES (pH7.5), 300 mM KCl, 10% (w/v) Glycerol, Protease Inhibitor Cocktail for use in Histidinetagged proteins purification (Sigma), and 2 mM β-mercaptoethanol. Cell extracts were incubated with Talon Metal Affinity Resin (clontech) beads overnight at 4˚C. Beads were washed with lysis buffer and lysis buffer containing 5 mM imidazole. Proteins were eluted by lysis buffer containing 200 mM imidazole. Buffer exchange against lysis buffer was performed using PD-10 desalting column (Amersham Biosciences).

Fluorescence microscopic observation
Cells were cultured in the SD medium to logarithmic phase and fluorescence was detected by Olympus BX-52 microscope equipped with a digital CCD camera ORCA-ER 95-4742-ER (Hamamatsu Photonics) and an imaging system AQUACOSMOS (Hamamatsu photonics).

Suppression of Etn auxotrophy of psd1Δ on non-fermentable carbon sources by SFH1 overexpression
The PSD1 deletion mutant showed a significant defect in growth on non-fermentable carbon sources, but supplementation of Etn suppressed this defect [52], suggesting that PE synthesized in other organelle(s) is transported to the mitochondria. To identify the genes involved in PE supply to the mitochondria in S. cerevisiae, we carried out genetic screening for multi-copy suppressors of Etn auxotrophy of psd1Δ on lactate. A multi-copy genomic library was introduced into psd1Δ and clones that restored the growth on synthetic lactate (SLac) medium in the absence of Etn supplementation were isolated. These clones are expected to increase the amount of mitochondrial PE by enhancing PE synthesis or PE import to the mitochondria. Indeed, genes involved in PE synthesis were obtained from this screening; overexpression of PSD2 and DPL1, which encodes a dihydrosphingosine phosphate lyase involved in the synthesis of PE via the Kennedy pathway, partially recovered the growth of psd1Δ on SLac (Fig 2A). After screening approximately 12,500 transformants, we isolated a clone containing SFH1/ YKL091c encoding a member of the Sec14 family proteins. Overexpression of SFH1 by a multi-copy vector suppressed the growth defect of psd1Δ on lactate more efficiently than PSD2 or DPL1 (Fig 2A). SFH1 overexpression also restored the growth of psd1Δ on the medium containing other non-fermentable carbon sources, including glycerol and ethanol (S1A Fig).
Overexpression of SEC14 weakly suppressed the growth defect of psd1Δ on lactate, but overexpression of other Sec14-family protein genes, including SFH4, which is critical for Psd2 activity in vivo [10], did not ( Fig 2B). Previous studies have indicated that the endogenous Sec14 level was more than 20-fold higher than that of Sfh1 in wild-type S. cerevisiae cells cultured in the medium containing glucose [53,54]. Therefore, it is conceivable that the suppressor function of Sec14 is much weaker than that of Sfh1 and that the suppression of the growth defect of psd1Δ on non-fermentable carbon sources is a unique function of Sfh1 amongst the yeast Sec14-family proteins. The Y113C substitution confers Sec14-like functions in vesicle transport to Sfh1 [34], but did not abolish the activity of Sfh1 in suppressing the growth defect of psd1Δ on lactate (Fig 2C).
Crystal structure analysis revealed that recombinant Sfh1 binds PI, PC, and PE in its hydrophobic pocket [35]. Sec14 is reported to possess PITP and PCTP activities in vitro [31], but both the PITP and PCTP activities of Sfh1 are approximately fivefold lower than those of Sec14 [34]. Sfh1 S175I,T177I , Sfh1 R61A,T238D , and Sfh1 L179W,I196W mutants are reported to be defective in binding to PC/PE, PI, and PC/PE/PI, respectively [35]. Overexpression of Sfh1 S175I,T177I and sfh1 L179W,I196W mutant genes did not suppress the growth defect of psd1Δ on lactate while overexpression of sfh R61AT238D partially restored the growth of psd1Δ (Fig 2C). We analyzed the levels of Sfh1 mutant proteins tagged with EGFP, whose overproduction showed similar effects with Sfh1 mutants without EGFP tag on the growth of psd1Δ, and expression levels of these Sfh1 mutants tagged with EGFP were found to be not less than that of the wild-type Sfh1 tagged with EGFP (S2 Fig), suggesting that these mutants were overexpressed. These results imply that PC/PE binding (and probably transport) is a prerequisite for Sfh1 function whereas PI binding (and probably transport) is partially dispensable.
In the psd1Δ cells, there remain two routes to synthesize PE: the Kennedy pathway in the ER and PS decarboxylation by Psd2 in the endosome, Golgi, and/or vacuole [52]. To clarify the contribution of these routes to the Sfh1-mediated growth recovery on lactate, we validated the growth of the double deletion mutant of PSD1 and ECT1, which encodes ethanolaminephosphate cytidylyltransferase, a critical enzyme in the PE synthesis through the Kennedy pathway, and that of PSD1 and PSD2 or SFH4 in the presence of SFH1 overexpression. Overexpression of SFH1 restored the growth of psd1Δect1Δ on lactate, suggesting that the Kennedy Role of Sfh1, a Sec14 homolog, in yeast pathway is dispensable for the suppressor ability of SFH1 (Fig 2D). In contrast, SFH1 overexpression did not recover the growth defect of psd1Δpsd2Δ or psd1Δsfh4Δ on lactate (Fig 2E and  S1B Fig), revealing that PE synthesized by Psd2 is critical for the suppression of the growth defect of psd1Δ on lactate by the overexpression of SFH1. In accordance with this result, simultaneous overexpression of SFH1 and PSD2 improved the growth of psd1Δ on lactate compared with overexpression of each gene alone (S1C Fig).

Restoration of mitochondrial function of psd1Δ by SFH1 overexpression
To determine whether overexpression of SFH1 increased the mitochondrial PE level of psd1Δ, we purified the mitochondrial fraction from psd1Δ in the presence or absence of SFH1 overexpression and analyzed the phospholipid composition. Deletion of PSD1 caused a three-fold reduction in the mitochondrial PE level, and overexpression of SFH1 in psd1Δ partially restored it (Fig 3A).
We next analyzed the cellular phospholipid composition of psd1Δ with and without SFH1 overexpression. The psd1Δ cellular PE level increased following SFH1 overexpression (Fig 3B), while the cellular PE level of psd1Δpsd2Δ did not (S3 Fig), suggesting that overexpression of SFH1 leads to an increase in PE synthesized by Psd2. Moreover, the fact that overexpression of the Sfh1 S175I,T177I mutant did not increase the cellular PE level ( Fig 3B) agreed with the inability of this mutant to restore the growth of psd1Δ on lactate (Fig 2C).
A possible reason for the elevated cellular PE level by SFH1 overexpression is that Sfh1 directly activates the enzyme involved in the synthesis of PE or its precursor PS. To test this possibility, we analyzed the activity of Pss1 and Psd2 in vitro. However, the addition of the recombinant His 8 -Sfh1 did not increase the activity of Pss1 in the isolated ER fraction ( Fig  4A). In addition, overexpression of SFH1 had no significant effect on Psd2 activity in the psd1Δ cell extract (Fig 4B). Furthermore, SFH1 overexpression did not significantly increase the amount of epitope-tagged Pss1 and Psd2 proteins, which were expressed from the low copy plasmids under their native promoters (Fig 4C and 4D and S4 Fig).

Binding and transfer of phospholipids by Sfh1
Another possibility for the elevated PE level by SFH1 overexpression is that Sfh1 mediates interorganellar transfer of PS and/or PE and its overproduction enhanced PE synthesis by Psd2 through increased PS supply and/or efficient removal of PE. To determine whether Sfh1 binds to PS and/or PE in vivo, Sfh1 was purified from the cytosolic fraction of yeast cell extracts and the lipid species recovered with Sfh1 were analyzed. To accomplish this, Sfh1 was fused with an IgG binding domain (ZZ tag) at its C-terminus (Sfh1-ZZ) and this protein was overproduced in psd1Δ. This fusion protein was confirmed to be functional, because its overproduction suppressed the growth defect of psd1Δ on lactate (S5A Fig). Sfh1-ZZ was then affinitypurified from the cytosolic fraction (S5B Fig), and the phospholipids bound to Sfh1-ZZ were analyzed by electrospray ionization tandem mass spectrometry (ESI-MS/MS). Significant amounts of PC, PE, PI, and PS were recovered in the extract prepared from Sfh1-ZZ, whereas much smaller amounts of these phospholipids were detected in the fraction similarly prepared from the cells overexpressing the non-tagged Sfh1 (Fig 5A), excluding the possibility of nonspecific binding of lipids to the IgG sepharose beads used for protein purification. Compared with phospholipids detected from Sfh1-ZZ, significantly lower amounts of PE and PC were recovered from Sfh1 S175I,T177I -ZZ, in agreement with the defect in PE/PC binding of recombinant Sfh1 S175I,T177I . Furthermore, the amount of PS recovered from Sfh1 S175I,T177I -ZZ appeared to be lower than that from Sfh1-ZZ (C32:1 and C34:1, p = 0.06), raising the possibility that Ser 175 and Thr 177 are also important for PS binding. Together, the ability of Sfh1 to bind aminophospholipids was suggested to be critical for its function. Maeda et al. have analyzed the lipids bound to Sfh proteins and showed that Sfh1 binds PI and PC, but not PS and PE, in vivo [55]. The reason for this discrepancy remains unclear, but it might be due to the difference in the amounts of phospholipids that bind to Sfh1. Alternatively, it may be explained by the difference in the experimental conditions. We next investigated whether Sfh1 possesses the ability to transfer aminophospholipids between liposomes in vitro by a fluorescent dequenching assay. Donor liposome containing 7-nitro-2-1,3-benzoxadiazol-4-yl (NBD)-phospholipid and rhodamine-PE were incubated with recombinant proteins and acceptor liposome mimicking the phospholipid composition of psd1Δ cells. While the NBD fluorescence is quenched by rhodamine when NBD- The psd1Δ strains harboring YCp111-PSD1, YEplac181, or YEp181-SFH1 were cultured in semi-synthetic lactate medium to late logarithmic phase, and mitochondria were purified by sucrose density gradient. Lipids were extracted by the Bligh and Dyer's method, separated by thin-layer chromatography, and analyzed as described in the Material and Methods. Data are the means of three independent assays. Error bars represent S.E. �� , p < 0.005 (two-tailed Student's t-test) (B) Cellular phospholipid composition of psd1Δ overexpressing SFH1. The psd1Δ strains harboring YCp111-PSD1, YEplac181, YEp181-SFH1, or YEp181-SFH1 S175I,T177I were cultured in semi-synthetic lactate medium to late logarithmic phase. Lipids were extracted and analyzed as described above. Data are the means of three independent assays. Error bars represent S.E. �� , p < 0.005 (two-tailed Student's t-test). https://doi.org/10.1371/journal.pone.0215009.g003 Role of Sfh1, a Sec14 homolog, in yeast phospholipid and rhodamine-PE are on the same liposome, it is detectable when NBD-phospholipid is transferred to the rhodamine-free liposome (Fig 5B). The addition of recombinant Sfh1 dequenched the fluorescence of NBD-PS in the transfer assay (Fig 5C). No dequenching was observed when the acceptor liposome was absent in the reaction mixture ( S5C Fig). Importantly, dequenching of NBD-PS by Sfh1 S175I,T177I mutant was less efficient than the wild-type Sfh1. Addition of non-LTP protein (LDH, lactate dehydrogenase) showed no increase in NBD fluorescence, indicating that liposome fusion by non-specific proteins did not occur in this condition [56]. Dequenching of the NBD-PS fluorescence by Sfh1 was observed in a dose-dependent manner (Fig 5D). These results suggest that Sfh1 has the ability to transfer NBD-PS between liposomes in vitro. The fluorescence of NBD-PC and NBD-PE declined in a time-dependent manner for an unknown reason, but the decrease in the fluorescence was significantly alleviated by the addition of the wild-type Sfh1 (S5D and S5E Fig). The addition of the Sfh1 S175I,T177I mutant showed the same effects. It is also possible that Sfh1 transferred non-labeled phospholipids from the acceptor liposome to the donor liposome, resulting in dilution of the fluorescent phospholipids, which culminates in dequenching of NBD fluorescence. However, because significant increase in the dequenching signal by Sfh1 was detected

Fig 5. Sfh1 binds phospholipids in vivo and transfers PS between liposomes in vitro.
(A) Phospholipids bound to Sfh1-ZZ and Sfh1 S175I,T177I -ZZ. The wildtype strains harboring YEp181-SFH1-ZZ, YEp181-SFH1 S175I,T177I -ZZ or YE181-SFH1 were cultured in SD medium to logarithmic phase. Proteins were purified from S100 fractions (4 mg protein) prepared from these cells.  concentration (0, 400, 800, and 1200 nM). (E) Simultaneous SFH1 overexpression recovers the growth of psd1Δ overexpressing PSS1 on lactate. Cells were cultured in SD medium to logarithmic phase, and were spotted on SD or SLac medium in ten-fold serial dilutions and were incubated on SD medium for 2 days or on SLac medium for 7 days. https://doi.org/10.1371/journal.pone.0215009.g005 Role of Sfh1, a Sec14 homolog, in yeast when NBD-PS, but not NBD-PE or NBD-PC, was used, it is plausible that NBD-PS was transported from the donor liposome to the acceptor liposome.
While overexpression of PSD2 recovered the growth of psd1Δ on lactate (Fig 2A), overexpression of PSS1 impaired it (Fig 5E), implying that an excess of PS in the psd1Δ mitochondria may impair mitochondrial function. Simultaneous overexpression of SFH1, however, recovered the growth of psd1Δ overexpressing PSS1 (Fig 5E). Enhanced transport of PS from the ER to endosome, Golgi, and/or vacuole by SFH1 overexpression and the following conversion of PS to PE by Psd2 may have reduced the mitochondrial PS level. We analyzed the cellular phospholipid levels in psd1Δ, psd1Δ overexpressing PSS1, and psd1Δ overexpressing PSS1 and SFH1 (S6 Fig). However, PSS1 overexpression did not significantly increase the cellular PS level, and simultaneous overexpression of SFH1 did not decrease it. It is possible that PSS1 overexpression resulted in increased the local PS levels of mitochondria or other organelles and simultaneous overexpression of SFH1 decreased it. Overexpression of SFH4 did not recover the growth of psd1Δ overexpressing PSS1 (S5F Fig).

Localization of Sfh1 to the cytosol and organellar membranes
The subcellular localization of the Sec14 family proteins has been studied by fluorescent microscopy and it was suggested that the Sfh1 fused with enhanced green fluorescent protein (Sfh1-EGFP) localizes to the cytosol and nucleus [57] and we obtained similar results (Fig 6A). However, fractionation by differential centrifugation suggested that Sfh1-EGFP localizes to the 30,000, 40,000, and 100,000 × g microsomal fractions, in addition to the cytosolic fraction [57]. To further assess the organelle to which Sfh1 localizes, we investigated the subcellular localization of Sfh1 tagged with the 3xHA epitope at its C-terminus (Sfh1-HA) by sucrose density gradient centrifugation. Analysis of the distribution of Sfh1-HA and organelle marker proteins suggested that, in addition to the cytosolic fraction, Sfh1-HA was enriched in the fractions containing the marker protein of the endosome, Golgi, and vacuole when overexpressed in the psd1Δ cells (Fig 6B and S7 Fig). Less Sfh1-HA was detected in the fractions, in which the ER and mitochondrial markers were enriched. Not only when overexpressed using a multi-copy vector, Sfh1-HA was also recovered in the fractions containing marker proteins of the endosome, Golgi, and vacuole when expressed under its native promoter using a low copy vector (S8 Fig). These results suggest that a fraction of Sfh1 localized to the endosome, Golgi, vacuole, and cytosol. As Psd2 has been reported to localize to the endosome, Golgi, and vacuole [9,10], the localization of Sfh1-HA to those organelles could be important for enhancing Psd2-mediated PE synthesis.

Deletion of SFH1 impairs the growth of psd1Δ
Expression of SFH1 using a low-copy vector also recovered the growth of psd1Δ, albeit to a lesser extent than overexpression of SFH1 using a multi-copy vector (Fig 7A). SFH1 has little effect on the growth of the wild-type S. cerevisiae strain [58], and its physiological role has remained obscure. To determine the physiological function of SFH1, we examined the growth of psd1Δsfh1Δ. The growth of psd1Δ was slightly but significantly impaired by the deletion of SFH1 both on solid and liquid SLac media (Fig 7B and S9 Fig).
To determine whether SFH1 deletion affects phospholipid synthesis in psd1Δ, we analyzed the cellular phospholipid composition of psd1Δsfh1Δ (Fig 7C). Compared to the psd1Δ cells, the proportion of PI tended to increase (p = 0.07) and the sum of PS, PE, and PC tended to decrease in the psd1Δsfh1Δ cells (p = 0.06). These results imply that deletion of SFH1 results in a slight depression in phospholipid synthesis through the PS branch of the CDP-DAG pathway in the psd1Δ cells.

Involvement of MCSs in the suppression of psd1Δ by SFH1 overexpression
Overexpression of SFH1 increased the cellular and mitochondrial PE levels of psd1Δ in a Psd2-dependent manner, indicating that PE synthesized by Psd2 was supplied to the mitochondria. Contribution of MCSs in lipid transfer in cooperation with LTPs have been argued [19][20][21]. Mitochondria form MCSs with the ER by ERMES complex composed of Mmm1, Mdm10, Mdm12, and Mdm34 or the EMC composed of Emc1-Emc6. Mitochondria also form a MCS with vacuole (vCLAMP) marked by Vps39 and Ypt7, and a MCS with endosome by Vps13 in S. cerevisiae [22][23][24][25][26]. To determine whether these MCSs are required for the supplementation of PE synthesized by Psd2 to the mitochondria, SFH1 was overexpressed in the psd1Δ strains, in which genes involved in these MCSs are simultaneously deleted, and their growth on lactate was observed.
Double deletion of PSD1 and ERMES component genes exhibited much more severe growth defects on glucose or lactate in the presence or absence of Etn than the single deletion mutants, and the growth defect of psd1Δmdm34Δ was not rescued by introduction of PSD1 or MDM34, suggesting that the mitochondria were irreversibly damaged by the deletion of these genes (S10A Fig). Therefore, to examine the importance of the ERMES complex in the growth of psd1Δ supported by Sfh1 overproduction, we performed a plasmid shuffling assay for Role of Sfh1, a Sec14 homolog, in yeast MDM34 (Fig 8A). In this experiment, MDM34 on the chromosome was deleted in psd1Δ harboring the plasmid containing MDM34 with URA3 as a marker (YCp33-MDM34). Then the low-copy plasmid carrying MDM34 or PSD1 or multi-copy plasmid carrying SFH1 was introduced into the strain. These strains were cultured on SLac medium containing 5-FOA, which allows selection of cells that lost URA3, and growth of the cells that lost YCp33-MDM34 was observed. Overexpression of SFH1 did not significantly affect the growth of psd1Δmdm34Δ without the MDM34-harboring plasmid on SLac medium (Fig 8A, SLac + 5-FOA), suggesting that ERMES is critical for the growth of psd1Δ on lactate supported by Sfh1 overproduction. On the other hand, disruption of genes encoding the components of the EMC did not show a significant effect on growth of psd1Δ on lactate (Fig 8B and S10B Fig), implying that the EMC does not contribute to the increased supplementation of PE synthesized by Psd2 in the presence of SFH1 overexpression to the mitochondria.
Deletion of VPS39 impaired the growth of psd1Δ both on glucose and lactate (S10C Fig). Deletion of YPT7 also slightly impaired the growth of psd1Δ (S10C Fig). The products of these two genes are also components of the HOPS complex involved in vacuole-vacuole or late endosome-vacuole tethering. Deletion of VPS41, a gene encoding a specific component of the HOPS complex, impaired the growth of psd1Δ to a similar degree to VPS39 deletion (S10C Fig), and therefore the growth defects caused by VPS39 or YPT7 deletion may not be due to a defect in vCLAMP. Overexpression of SFH1 restored the growth of both psd1Δvps39Δ and psd1Δypt7Δ (Fig 8C), showing that vCLAMP is dispensable for the increased supply of PE synthesized by Psd2 in the presence of SFH1 overexpression to the mitochondria. Deletion of VPS13 did not exhibit a significant effect on the growth of psd1Δ (S10D Fig), and the growth restoration of psd1Δvps13Δ on lactate by SFH1 overexpression was comparable to that of psd1Δ (Fig 8D). These results suggested that VPS13 is not required for the increased supplementation of PE synthesized by Psd2 in the presence of SFH1 overexpression to the mitochondria in psd1Δ.

Discussion
In this study, we identified SFH1 as a multi-copy suppressor of the growth defect of psd1Δ on non-fermentable carbon sources and characterized its physiological function.

Enhanced Psd2-mediated PE synthesis by Sfh1 overproduction
Overexpression of SFH1 partially restored cellular and mitochondrial PE levels (Fig 3). One possible reason for this increase in the cellular PE level is the upregulation of the amounts or activities of Psd2 and Pss1, but we did not observe any significant change in protein abundance or activities of Psd2 and Pss1 following SFH1 overexpression (Fig 4). Alternatively, Sfh1 could potentially activate Psd2 by regulating PIP signaling, as Sec14-family proteins are proposed to facilitate PI4P production by presenting PI to PI kinases [36,37]. However, this possibility appears unlikely for the following reasons: i) Sfh1 does not activate the PI kinase in vivo [34], and ii) overexpression of the Sfh1 mutant deficient in PI binding partially but significantly suppressed the growth defect of psd1Δ on lactate (Fig 2C). In addition, activation of Pss1 or Psd2 by Sfh1 was not observed in vitro (Fig 4A and 4B). Activation of Pss1 appears to be harmful for psd1Δ mitochondria as overexpression of PSS1 impaired the growth of psd1Δ particularly on lactate (Fig 5E), whereas the simultaneous overexpression of SFH1 rescued the growth of psd1Δ (Fig 5E), implying that Sfh1 does not activate Pss1 solely in vivo.
The binding of Sfh1 to PS in vivo (Fig 5A), the possibility of the transport of NBD-labeled PS by Sfh1 in vitro (Fig 5C and 5D), and the localization of Sfh1 to the endosome, Golgi, and vacuole ( Fig 6B) suggest that Sfh1 transports PS from the ER to those organelles and that PE containing YCp33-MDM34 with YCp11-MDM34, YCp111-PSD1, YEp181-SFH1, or YEplac181 was cultured in SD medium to logarithmic phase and were spotted on SD or SLac medium with or without 0.5 mM 5-FOA in five-fold serial dilutions. Strains were incubated on SD or SLac medium without 5-FOA for 2 or 7 days, and on SD or SLac medium with 5-FOA for 7 or 10 days, respectively. Slight growth of psd1Δmdm34Δ containing YCp11-MDM34 and YEplac181 on SLac medium in the presence of 5-FOA was probably due to longer incubation than that on SLac medium in the absence of 5-FOA. (B)-(D) Cells were cultured in SD medium to logarithmic phase and spotted on SD or SLac medium in ten-fold serial dilutions and were incubated on SD medium for 2 days or on SLac medium for 7 days. synthesis is enhanced by the increased PS supply to Psd2 following Sfh1 overproduction (Fig  9). PS could be transported from the ER to endosome, Golgi, and vacuole through vesicle transport, but it might be insufficient. In addition, given that Sfh1 also bound to PE in vivo ( Fig 5A) and in vitro [48], it is possible that PE synthesis increases because of the efficient export of PE from the endosome, Golgi, and vacuole to the ER or other organelles (Fig 9). Friedman et al. have reported that Psd1 is present in mitochondria and the ER, and indicate the importance of the optimal PE level to the ER function [11]. It may be possible that Sfh1 transfers PE synthesized by Psd2 to the ER to maintain an appropriate level of PE in the ER.

Contribution of MCSs in the growth of psd1Δ overproducing Sfh1
A variety of MCSs have been discovered and their involvement in lipid transport has been discussed. One of the most well studied MCS components is the ERMES complex, which tethers the ER and mitochondria. While it has been reported that the ERMES components are dispensable for the transport of PS from the ER to mitochondria [59,60], there is evidence that ERMES is involved in phospholipid transfer between the ER and mitochondria [22,[61][62][63]. In addition, ERMES has been shown to influence the mitochondrial phospholipid composition in concert with vCLAMP components or EMC [23,24]. One possible reason for the requirement of ERMES for the growth suppression of psd1Δ on lactate by SFH1 overexpression (Fig  8) is that PE synthesized by Psd2 is transported to the ER and then transported to mitochondria by an ERMES-dependent mechanism (Fig 9). Alternatively, the simultaneous deletion of the ERMES component genes with PSD1 could have a deleterious effect on mitochondrial function, which cannot be recovered by Sfh1-dependent PE supplementation.
vCLAMP and the ER-endosome MCS formed by Vps13 were reported to dissociate from the mitochondria when cultured in medium containing the non-fermentable carbon source, glycerol or acetate [25,26]. These MCSs may not participate in phospholipid transport during respiratory growth. As overexpression of SFH1 restored the growth of vps39Δpsd1Δ or vps13Δpsd1Δ on lactate (Fig 8), it is conceivable that PE synthesized by Psd2 was transported to the mitochondria through different routes. Disruption of EMC had little effect on the growth of psd1Δ. Lahiri et al. reported that 5x-emc, in which EMC1, EMC2, EMC3, EMC5, and EMC6 are deleted, did not grow on the rich medium containing glycerol (YPGly) and had a growth defect on the rich medium containing glucose (YPD) [23]. However, the strain we constructed (5x-emcΔ) grew similarly to the wild-type strain on the minimal medium containing glucose or lactate (SD or SLac) (S10B Fig). This discrepancy may be due to the difference in medium and/or the genetic backgrounds. Although VPS13 and EMC genes were dispensable for the growth recovery of psd1Δ by SFH1 overexpression, we cannot exclude the possibility that the MCSs formed by these proteins partially contribute to the import of PE synthesized by Psd2 to the mitochondria. In addition, it is possible that PerMit contact is involved in phospholipid transport to mitochondria [29].

Physiological role of Sfh1
Although Sfh1 exhibits the highest similarity to Sec14 in the S. cerevisiae Sec14 family of proteins, deletion of SFH1 in the wild-type strain does not confer a significant effect on growth. Other Sec14 family members have been reported to be involved in various cellular processes [33,[64][65][66][67], yet the physiological role of Sfh1 has remained unclear. In this study, it was suggested that Sfh1 has a physiological function distinct from other Sfh proteins in S. cerevisiae. Deletion of SFH1 decreased the sum of the levels of PS, PE, and PC and increased PI level in psd1Δ cells (Fig 7C). This could be explained by our suggestion that Sfh1 is involved in the phospholipid transport between the ER and endosome, Golgi, and/or vacuole and less effective phospholipid transport between these organelles by SFH1 deletion partially abrogated the PS branch of the CDP-DAG pathway, resulting in increased conversion of the excess of CDP-DAG to PI.
SFH1 deletion aggravated the growth of psd1Δ on SLac albeit to a small extent (Fig 7B and  S9 Fig). SEC14, and perhaps other SFH genes, may function redundantly with SFH1, reducing the impact of SFH1 disruption. SFH1 and SEC14 are supposed to be generated by whole genome duplication [68], and these two genes may partially share functions. If this is the case, then where does the functional difference arise from? Sec14 was reported to localize to the cytosol and Golgi, while Sfh1 was suggested to localize to the cytosol, nucleus, endosome, Golgi, and vacuole [57,69] (Fig 6). This difference in the subcellular localization may affect their functions. Moreover, Sec14 has high PITP/PCTP activities whereas those of Sfh1 are very low. High PITP/PCTP activities of Sec14 may be derived from strong binding preferences for PI and PC over PS and PE.
SFH4 has been reported to play a critical role in PE synthesis by Psd2 [70]. Nonetheless, overexpression of SFH4 did not rescue the growth defect of psd1Δ on lactate ( Fig 2B). As Sfh4 forms a complex with Psd2 and Pbi1, overexpression of SFH4 alone may not influence the rate of PE production by Psd2. Overexpression of SFH4 partially rescued the growth defect of sec14 ts at a restrictive temperature, indicating that there is some functional redundancy between SEC14 and SFH4 [57]. In contrast, SFH1 appears to have a distinct physiological function from SFH4 (Fig 2B). Indeed, while Sfh1 was suggested to transfer NBD-PS in vitro (Fig 5C  and 5D), Sfh4 has been reported not to possess in vitro PS transfer activity [67].
Large-scale surveys have revealed that the expression of SFH1 is induced under various stress conditions, including high osmolality, high salt concentrations, temperature shifts, rapamycin treatment, and the diauxic shift [71][72][73]. Vacuole is involved in the responses to high osmolality, high salt concentrations, and rapamycin treatment [74]. PE synthesized by Psd2 was suggested to be important for vacuolar function [10], and PE synthesis by Psd2 could be upregulated in response to these stresses by transcriptional induction of SFH1. During the diauxic shift, yeast cells adapt themselves to respiring growth by the induction of genes involved in the assimilation of non-fermentable carbon sources and proliferation of the mitochondria. Enhanced expression of SFH1 during the diauxic shift implies that Sfh1 participates in mitochondrial function. Further study is required to clarify the physiological role of Sfh1 under these stress conditions. Supporting information S1 Fig. Overexpression of SFH1 restores the growth of the psd1Δ strain on non-fermentable carbon sources. (A) Overexpression of SFH1 restores the growth of psd1Δ on various non-fermentable carbon sources. Strains were spotted on synthetic medium with glucose, glycerol, ethanol, and lactate as sole carbon sources in ten-fold serial dilutions and were incubated on glucose for 2 days or on glycerol, ethanol, and lactate for 7 days. (B) SFH4 is critical to recover the growth of psd1Δ on lactate by SFH1 overexpression. Strains were spotted on SD or SLac medium in the presence or absence of 1 mM Etn in ten-fold serial dilutions and were incubated for 2 or 7 days, respectively. (C) Simultaneous overexpression of SFH1 and PSD2 leads to improved growth of psd1Δ on lactate compared to single overexpression of one of these genes. Strains were spotted on SD or SLac medium in five-fold serial dilutions and were incubated for 2 or 7 days, respectively. (TIF)

S2 Fig. Expression of Sfh1 mutants. (A)
The psd1Δ strains overexpressing SFH1, SFH1-EGFP, sfh1 S175I,T177I , sfh1 S175I,T177I -EGFP, sfh1 R61A,T238D , sfh1 R61A,T238D -EGFP, sfh1 L179W,I196W , and sfh1 L179W,I196W -EGFP were cultured in SD medium to logarithmic phase and were spotted on SD or SLac media in ten-fold serial dilutions and were incubated on SD medium for 2 days or on SLac medium for 7 days. (B) The psd1Δ strains overexpressing SFH1, SFH1-EGFP, sfh1 S175I,T177I , sfh1 S175I,T177I -EGFP, sfh1 R61A,T238D , sfh1 R61A,T238D -EGFP, sfh1 L179W,I196W , and sfh1 L179W,I196W -EGFP were cultured in SD medium to logarithmic phase and the levels Sfh1 proteins tagged with EGFP were evaluated by immunoblot using anti-EGFP antibody. (A) Sfh1 fused with ZZ tag is functional in S. cerevisiae. Strains were spotted on SD or SLac medium in ten-fold serial dilutions and were incubated for 2 or 7 days, respectively. (B) Affinity purification of Sfh1-ZZ and Sfh1 S175I,T177I -ZZ. Affinity purified proteins were eluted from IgG sepharose beads and concentrated by TCA precipitation. The purity was checked by SDS-PAGE. Arrow head represents Sfh1-ZZ and Sfh1 S175I,T177I -ZZ. � represents contaminated proteins. (C) NBD-PS dequenching assay was performed in the absence of acceptor liposomes. (D) and (E) NBD-PC (D) and NBD-PE (E) transfer activities of Sfh1 and Sfh1 S175I,T177I mutant were measured at room temperature. NBD fluorescence intensities were set to 0 at 0 s. Data are the means of three independent assays. Error bars represent S.E. Proteins were added to final concentration of 800 nM. (F) Overexpression of SFH4 does not rescue the growth of psd1Δ overexpressing PSS1. Strains were spotted on SD or SLac medium in the presence or absence of 1 mM Etn in ten-fold serial dilutions and were incubated for 2 or 7 days, respectively. (TIF)