Effects on vesicular transport pathways at the late endosome in cells with limited very long-chain fatty acids.

Very long-chain fatty acids (VLCFAs), fatty acids with chain-length greater than 20 carbons, possess a wide range of biological functions. However, their roles at the molecular level remain largely unknown. In the present study, we screened for multicopy suppressors that rescued temperature-sensitive growth of VLCFA-limited yeast cells, and we identified the VPS21 gene, encoding a Rab GTPase, as such a suppressor. When the vps21Δ mutation was introduced into a deletion mutant of the SUR4 gene, which encodes a VLCFA elongase, a synthetic growth defect was observed. Endosome-mediated vesicular trafficking pathways, including endocytosis and the carboxypeptidase Y (CPY) pathway, were severely impaired in sur4Δ vps21Δ double mutants, while the AP-3 pathway that bypasses the endosome was unaffected. In addition, the sur4Δ mutant also exhibited a synthetic growth defect when combined with the deletion of VPS3, which encodes a subunit of the class C core vacuole/endosome tethering (CORVET) complex that tethers transport vesicles to the late endosome/multivesicular body (MVB). These results suggest that, of all the intracellular trafficking pathways, requirement of VLCFAs is especially high in the endosomal pathways.

conditions used here, suggesting that the requirement for VLCFAs is especially high in the endosomal traffi cking pathways.

Yeast strains and media
The Saccharomyces cerevisiae strains used in this study are listed in Table 1 . Cells were grown in YPD medium (1% yeast extract, 2% peptone, and 2% D-glucose) or in synthetic dextrose (SD) medium (0.67% yeast nitrogen base without amino acids and 2% D-glucose) supplemented with 0.5% casamino acid and appropriate supplements. Single-deletion mutants, except MRY106 ( vps19 ⌬ ::LEU2 ), were purchased from Open Biosystems (Huntsville, AL). Gene disruption was performed by replacing the entirety or part of the coding region with a marker gene.

Screening of multicopy suppressors
MRY9 ( P tetO7 -PHS1 sur4 ⌬ ) cells were transfected with a yeast genomic library, in which genomic DNAs were cloned into the yeast expression vector pRS426 ( 2 , URA3 marker) ( 21 ). A pool of approximately 30,000 Ura + clones were plated on YPD plates and incubated for 4 days at 39°C. Plasmids were prepared from the 16 clones obtained and were reintroduced into MRY9. Of the 16 plasmids, 2 reproducibly restored growth of MRY9 at 39°C. Sequences of the DNA fragments inserted into the two plasmids were determined using a 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA). Sequencing revealed that one plasmid contained the SUR4 gene, and the other (pMR22) contained the VPS21 and YVC1 genes and the asparagine tRNA gene tN(GUU) O2 . The pMR31 ( VPS21 ) and pMR30 [ YVC1 and tN(GUU)O2 ] plasmids were created by subcloning the respective genes from the pMR22 plasmid into the pRS426 vector.

Plasmid construction
The pOK489 ( mRFP-CPS1 ) and pOK529 ( STE2-mRFP ) plasmids were constructed as follows. The CPS1 gene was amplifi ed by PCR from BY4741 genomic DNA to have Eco RI and Xho I sites at the 5 ′ and 3 ′ ends, respectively. The amplifi ed fragment was cloned into the Eco RI-Xho I site of the pOK107 plasmid ( 22 ), which contains the mRFP-2xFYVE domain, to replace the 2xFYVE domain with the CPS1 gene, thereby producing the pOK489 plasmid. The promoter and coding region, without the stop codon, of the STE2 gene was amplifi ed by PCR from BY4741 genomic DNA to have Sac I and Bam HI sites at the 5 ′ and 3 ′ ends, respectively. The mRFP sequence was amplifi ed by PCR from pOK107 so that Bam HI and Xho I sites were generated at the 5 ′ and 3 ′ ends, respectively. These fragments were cloned together into the Sac I-Xho I site of the p416TEF plasmid ( 23 ), creating the pOK529 plasmid.
The pMR8 ( FEN1 ) and pMR10 ( SUR4 ) plasmids were constructed as follows. The promoter, coding region, and 3 ′ -UTR of these genes were amplifi ed by PCR with creation of Bam HI site at the 5 ′ end. The amplifi ed fragments of FEN1 and SUR4 were fi rst cloned into the pGEM-T Easy vector (Promega, Madison, WI), then Bam HI-Not I fragments of the resulting plasmids were transferred to the yeast expression vector pRS316 ( CEN , URA3 marker) ( 24 ), producing the pMR8 and pMR10 plasmids.

FA elongation assays
In vitro FA elongation assays were performed essentially as described previously ( 25 ). Typical reaction mixtures contained total for both FEN1 and SUR4 are inviable, yet fen1 ⌬ and sur4 ⌬ single mutants are viable ( 12 ).
Involvement of VLCFAs in the yeast secretory pathway has been reported. For example, cells harboring deletions in both the SNC1 and SNC2 genes, which encode soluble NSF attachment protein receptors (SNARE) found on secretory vesicles, exhibit a temperature-sensitive growth phenotype and a reduction in invertase secretion ( 11 ). These phenotypes were suppressed by a loss of function mutation in SUR4 or FEN1 , implying that VLCFAs are potentially involved in the secretory pathway ( 11 ). In addition, in sur4 ⌬ cells, the plasma membrane-localized H + -ATPase Pma1 is mistargeted to the vacuole ( 8 ), and an interruption in the transcription of PHS1 , which encodes the 3-hydroxyacyl-CoA dehydratase that catalyzes the third step of the VLCFA elongation cycle, results in impaired accumulation of mature Gas1, a GPI-anchored protein localized in the plasma membrane ( 9 ). Thus, involvement of VLCFAs in secretory pathways is now recognized. However, no report has clearly demonstrated involvement of VLCFAs in the vesicular traffi cking pathway that functions in transport toward the vacuole. This is probably due to diffi culties in assessing their importance under VLCFAnull, lethal conditions, in which any indirect effects could not be excluded. Thus, investigation of VLCFA function under mild conditions is required for more understanding.
In the study presented here, we carried out a genetic screen to identify genes involved in VLCFA functions. We identifi ed the Rab GTPase VPS21 as a multicopy suppressor of a yeast mutant with limited VLCFA synthesis. Analyses of a sur4 ⌬ vps21 ⌬ double mutant suggested that VLCFAs are involved in Vps21-related vesicular transport processes. Furthermore, the gene SUR4 exhibited interactions with genes encoding the class C core vacuole/ endosome tethering (CORVET) subunits that function in tethering of vesicles to the late endosome/multivesicular body (MVB). Considering that any yeast mutant unable to synthesize VLCFA is nonviable, VLCFAs must be important for several essential cellular processes, such as general vesicular transport. However, transport defects at the endosome appeared even under the mild, VLCFA-limited

Pulse-chase experiment
Cells were grown at 30°C in SD medium containing appropriate supplements but lacking methionine and cysteine. Cells (2.0 A 600 units) were pulse-labeled with 100 Ci [ 35 S]methionine/cysteine (EasyTag EXPRE 35 S 35 S Protein Labeling Mix; PerkinElmer Life Sciences) at 30°C for 15 min. Chase was initiated upon addition of unlabeled methionine and cysteine to a fi nal concentration of 5 mM and 1 mM, respectively. The reaction was terminated by addition of sodium azide to a fi nal concentration of 20 mM, and cells were kept on ice. Cells were collected by centrifugation, suspended in lysis buffer [50 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, and 1% 2-mercaptoethanol], broken by vigorously mixing with glass beads, and boiled for 3 min. Then the resulting lysates (30 l) were diluted with 1 ml Triton buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.1 mM EDTA (pH 8.0), and 2% Triton X-100]. After removal of insoluble materials by centrifugation, samples were incubated with 750 ng of anti-CPY antibody (Molecular Probes) and 15 l of Protein G-Sepharose (GE Healthcare Bio-Sciences) for 16 h at 4°C with rotation. After beads were washed once with Triton buffer and once with Wash buffer [10 mM Tris-HCl (pH 8.0) and 150 mM NaCl], bound proteins were eluted with 2× SDS sample buffer [100 mM Tris-HCl, (pH 6.8), 4% SDS, 20% glycerol, 12% 2-mercaptoethanol, and a trace amount of bromophenol blue] and subjected to SDS-PAGE. The gel was fi xed with 10% acetic acid and 20% methanol, treated with Amplify fl uorographic reagent (GE Healthcare Bio-Sciences), and detected using a bioimaging analyzer BAS-2500 (Fuji Photo Film).

Microscopy
Cells were cultured to log phase in SD medium supplemented with 0.5% casamino acid and appropriate supplements and observed under a fl uorescence microscope (Axioimager M2, Carl Zeiss, Oberkochen, Germany). After the reaction, lipids were saponifi ed, acidifi ed, extracted, and dried as described previously ( 25 ). FAs were then converted to methyl esters by dissolving in 67 l of 3 M methanolic-HCl (fi nal 1 M), 123 l of chloroform, and 10 l of 2,2-dimethoxypropane (fi nal concentration 5%), and incubating at 100°C for 1 h. Samples were then treated with 233 l methanol, 177 l chloroform, and 300 l of 1% (w/v) KCl with vigorous mixing, and the organic phase was recovered and dried. Lipids were suspended in 20 l chloroform and separated by reverse-phase TLC on LKC18 Silica Gel 60 TLC plates (Whatman, Kent, UK) with chloroform/ methanol/water (5:15:1, v/v) as the solvent system. Labeled lipids were detected by autoradiography.

Immunoblotting
Total cell lysates were prepared by the alkaline/trichloroacetic acid method as described previously ( 22 ). Proteins were separated by SDS-PAGE and transferred to Immobilon polyvinylidene difl uoride membrane (Millipore, Billerica, MA). The membrane was incubated with an anti-DsRed (1:1000 dilution; a kind gift from Dr. S. Nishikawa), anti-CPY (0.25 g/ml; Molecular Probes, Eugene, OR), anti-Pep4 (1:5000 dilution; a kind gift from Dr. Y. Ohsumi), anti-ALP (1:5000 dilution; a kind gift from Dr. Y. Ohsumi), or anti-Pgk1 (0.25 g/ml; Molecular Probes) antibody, and then with HRP-conjugated anti-rabbit or anti-mouse IgG F(ab ′ ) 2 fragment (1:10,000 dilution; GE Healthcare Bio-Sciences, Piscataway, NJ). Signals were detected using an ECL prime (GE Healthcare Bio-Sciences) or Western Lightning Plus-ECL (Perkin-Elmer Life Sciences, Waltham, MA) kit and an ImageQuant LAS 4000 Biomolecular Imager (Fuji Photo Film, Tokyo, Japan). The Pep4 signal was enhanced using Can Get Signal Immunoreaction Enhancer Solution (TOYOBO, Osaka, Japan). To reveal any relationship between the chain-length of the VLCFAs and the growth requirement observed above, we examined the elongase activity of the P tetO7 -PHS1 sur4 ⌬ cells in vitro using palmitoyl-CoA as a substrate. The P tetO7 -PHS1 cells were able to elongate palmitoyl-CoA to C26-CoA, but in the P tetO7 -PHS1 sur4 ⌬ cells elongation beyond C22-CoA was largely impaired, with no C26-CoA and only small amounts of C24-CoA produced ( Fig. 1B ). These results suggest that VLCFAs with a chain-length of 26 and/or 24 carbons are required for cell growth at 39°C.

Overexpression of VPS21 suppresses the temperature sensitivity of the VLCFA-limited cells
To identify genes that exhibit a close relationship with VLCFA function, we employed a multicopy suppressor screening. P tetO7 -PHS1 sur4 ⌬ cells were transfected with a yeast genomic DNA library, and suppressors that restored growth at 39°C were isolated. Of two suppressors we obtained, one comprised a plasmid containing the SUR4 gene. The other suppressor harbored a plasmid containing a genomic region encompassing the VPS21 gene, which encodes a Rab-type small GTPase involved in vesicular traffi cking to the vacuole ( 27, 28 ) (supplementary Fig. I-A). Further subcloning and growth suppression assays determined that the VPS21 gene indeed functions as a multicopy suppressor in P tetO7 -PHS1 sur4 ⌬ cells at high temperatures (supplementary Fig. I-B). To clarify whether overexpression of Vps21 restored growth by facilitating synthesis of C24 and/or C26 VLCFA, we performed a VLCFA elongation assay. Expression of Sur4 restored the VLCFA elongation activity that produces C26-CoA and that was defective in the P tetO7 -PHS1 sur4 ⌬ cells, but no signifi cant difference was observed in the VLCFA elongation activity of P tetO7 -PHS1 sur4 ⌬ cells overexpressing Vps21 and those harboring an empty vector (supplementary Fig. I-C). Thus, Vps21 does not regulate VLCFA synthesis but may be related to the cellular functions in which VLCFAs are involved.

Deletion of the VPS21 gene causes a synthetic growth defect with deletion of the elongase genes
The growth defect in cells grown at 39°C was observed not only for P tetO7 -PHS1 sur4 ⌬ cells but also for sur4 ⌬ cells, albeit to a lesser extent ( Fig. 1A ). For simplifi cation, we used the sur4 ⌬ cells for further analyses. The sur4 ⌬ cells exhibited only slight growth retardation, both at 30°C and 36°C ( Fig. 2A ). Growth of the vps21 ⌬ cells was only slightly retarded at 36°C, although a signifi cant growth defect was observed at 39°C ( Fig. 2A ), as reported for 38°C ( 28 ). When a vps21 ⌬ mutation was introduced into the sur4 ⌬ cells, the resulting sur4 ⌬ vps21 ⌬ cells exhibited more severe growth retardation at 30°C than would be expected as an additive effect of the two single mutations ( Fig. 2A ), indicating a synthetic growth defect. Furthermore, we even observed a synthetic lethal phenotype at 36°C ( Fig. 2A ). A fen1 ⌬ vps21 ⌬ double mutant also displayed a synthetic growth defect both at 30°C and 36°C ( Fig. 2A ). To address whether the synthetic effects observed were the result of alterations in VLCFA synthesis, we performed FA elongation assays. The wild-type membrane elongated C16-CoA up to C26-CoA

Yeast cells' requirement of VLCFAs for growth is signifi cant at elevated temperature
To begin to understand the functions of VLCFA in yeast, we fi rst examined their roles in growth using a mutant with reduced VLCFA synthesis. Complete abrogation of VLCFA synthesis, such as that resulting from a double deletion of the redundant elongases FEN1 and SUR4 , is lethal ( 12 ), so we fi rst examined the growth of fen1 ⌬ and sur4 ⌬ single-mutant cells at high temperature. Both mutants exhibited normal growth at 30°C; however, at 39°C the growth of the sur4 ⌬ cells was retarded compared with the growth of wild-type cells ( Fig. 1A ). This phenotype was enhanced by decreasing the levels of the 3-hydroxyacyl-CoA dehydratase Phs1, which is involved in the third step of the VLCFA elongation cycle, using genetic manipulation. Replacement of the PHS1 promoter with the doxycyclinedependent tetO 7 promoter ( P tetO7 ) reduces the Phs1 levels to approximately 1/20 of normal even in the absence of doxycycline treatment ( 26 ). P tetO7 -PHS1 sur4 ⌬ cells could not grow at 39°C at all ( Fig. 1A ). P tetO7 -PHS1 fen1 ⌬ cells also exhibited a severe growth defect at 39°C. These results indicate that the cells' requirement for VLCFAs is particularly signifi cant at high temperatures.
, and MRY7 ( P tetO7 -PHS1 fen1 ⌬ ) cells were grown to stationary phase, serially diluted at 1:10, spotted on YPD plates, grown for the indicated times at 30°C or 39°C, and photographed against a dark background. (B) YOK2230 ( P tetO7 -PHS1 ) and MRY9 ( P tetO7 -PHS1 sur4 ⌬ ) cells were grown to log phase at 30°C, and total membrane fractions were prepared. Samples (20 g that substantial levels of CPY were present intracellularly as the mature, vacuole form in wild-type, fen1 ⌬ , and sur4 ⌬ cells ( Fig. 3A ). In vps21 ⌬ cells, the mature CPY was also apparent, but its levels were decreased compared with those in wild-type cells, and small amounts of the p2 (Golgi) form were observed intracellularly. The reduction in intracellular CPY was due to missorting of the p2 CPY to the extracellular space ( Fig. 3A ), as others have reported ( 27,29 ). In contrast, sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells accumulated signifi cant levels of the p1 (ER) and p2 (Golgi) forms of CPY, in addition to the mature form, indicating that the CPY pathway was affected synthetically by these double mutations ( Fig. 3A ). Secretion of the p2 form was also observed in both sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells, suggesting that the late secretory pathways are not impaired in these cells. Such synthetic effects were also confi rmed by a pulse-chase experiment ( Fig. 3B ). Nearly all newly synthesized CPY was converted to the mature ( Fig. 2B ). However, elongations beyond C18-CoA and C22-CoA were largely impaired in fen1 ⌬ and sur4 ⌬ cells, respectively. These results were consistent with the previous lipidomic analysis ( 6 ), in which C16/C18 and C22 sphingolipids were increased in fen1 ⌬ and sur4 ⌬ cells, respectively.
Sphingolipid synthesis is known to be affected by the sur4 ⌬ or fen1 ⌬ mutation ( 6,14 ). However, introducing the vps21 ⌬ mutation into the sur4 ⌬ or fen1 ⌬ mutant had no further effect on sphingolipid synthesis as judged by [ 14 C] serine or [ 3 H]dihydrosphingosine labeling (supplementary Fig. II-A, B). Thus, Vps21 is not involved in the regulation of VLCFA or sphingolipid synthesis, but it may be related to the functions of VLCFAs.
Pathways governing protein transport toward the vacuole via the late endosome/MVB are impaired in sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ mutants The genetic interactions between VPS21 and SUR4 / FEN1 observed above suggest that VLCFAs have roles in pathways involved in vesicular transport toward the vacuole via the endosome, in which Vps21 is involved. Thus, we monitored vesicular transport to the vacuole in sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells. To avoid potential secondary effects caused by the lethality of sur4 ⌬ vps21 ⌬ mutations at 36°C, we examined the vesicular transport at 30°C. The soluble vacuolar proteases carboxypeptidase Y (CPY) and Pep4 are synthesized in the ER and delivered to the vacuole via the Golgi apparatus and the late endosome/MVB, a pathway called the CPY pathway. Immunoblots revealed  into the cell by endocytosis and delivered to the vacuole via the late endosome/MVB ( 34,35 ). In wild-type cells, the fl uorescence of Ste2-mRFP was mainly detected in the vacuole lumen ( Fig. 5A and supplementary Fig. V). The endocytic pathway was also not signifi cantly affected in sur4 ⌬ or fen1 ⌬ cells, since Ste2-mRFP was observed in the lumen of the fragmented vacuoles ( Fig. 5A and supplementary Fig. V). In vps21 ⌬ cells, Ste2-mRFP was detected form within 20 min in wild-type, fen1 ⌬ , and sur4 ⌬ cells. In vps21 ⌬ cells, though, the mature CPY levels were decreased compared with those in wild-type cells, and signifi cant amounts of the p2 form were detected after a 20 min chase. Introducing the fen1 ⌬ or sur4 ⌬ mutation into the vps21 ⌬ cells enhanced the defect in the CPY transport. The mature CPY was nearly undetectable after a 20 min chase, and the p2 levels became more prominent in these double mutants ( Fig. 3B ). Similar results were obtained for Pep4. Nearly all Pep4 was detected as the mature form in wildtype, vps21 ⌬ , fen1 ⌬ , and sur4 ⌬ cells, whereas in vps21 ⌬ cells the Pep4 levels were decreased ( Fig. 3C ). In contrast, sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells signifi cantly accumulated the Pep4 proform. These results indicate that the deletions of VPS21 and SUR4 or FEN1 synthetically affect the CPY pathway.
We next examined the transport of the vacuolar carboxypeptidase Cps1. Cps1 is synthesized as a membrane protein in the ER, then transported to the vacuole lumen via the Golgi and the late endosome/MVB, where it is sorted into the intraluminal vesicles, then processed to its mature form. Fluorescence microscopy using mRFP-fused Cps1 (mRFP-Cps1) indicated that Cps1 was normally localized in the vacuole lumen in wild-type cells ( Fig. 4A ). In sur4 ⌬ and fen1 ⌬ cells, vacuoles were highly fragmented, as reported by others ( 30, 31 ) ( Fig. 4A and supplementary  Fig. III). Fluorescence of mRFP was detected in the fragmented vacuoles, indicating that mRFP-Cps1 did reach the vacuole lumen in these mutants ( Fig. 4A and supplementary Fig. IV). In vps21 ⌬ cells, mRFP-Cps1 was detected mainly in the vacuole lumen. In addition, some mRFP-Cps1 signals were detected at the vacuole membrane and as intracellular punctuates ( Fig. 4A ), indicating that mRFP-Cps1 transport was slightly retarded in the vps21 ⌬ cells. In the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ mutants, mRFP-Cps1 transport to the vacuole was severely impaired. Almost no mRFP-Cps1 was observed in the vacuole lumen, but mRFP-Cps1 accumulated on the vacuolar membranes, in a perivacuolar compartment, probably representing MVB, and in some punctate structures in the cytosol ( Fig. 4A and supplementary Fig. IV). Interestingly, fragmentation of the vacuole observed in the sur4 ⌬ and fen1 ⌬ mutants was reduced in these double mutants (supplementary Fig. III).
We also confi rmed the defect in Cps1 transport in sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells by immunoblot using an anti-DsRed antibody that detects mRFP. After delivery into the vacuole lumen, the linker portion of mRFP-Cps1 is cleaved by vacuolar proteases, releasing the mRFP moiety ( 32 ). The free mRFP was detected in wild-type, vps21 ⌬ , fen1 ⌬ , and sur4 ⌬ cells but not in sur4 ⌬ vps21 ⌬ or fen1 ⌬ vps21 ⌬ cells ( Fig. 4B ), indicating that mRFP-Cps1 was not delivered to the vacuole lumen in the sur4 ⌬ vps21 ⌬ or fen1 ⌬ vps21 ⌬ cells. These results indicate that the transport of membrane proteins via intraluminal vesicle formation is severely impaired in sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells.
Some vacuolar proteins are transported to the vacuole directly from the Golgi without passing through the late endosome/MVB. This pathway is called the AP-3 pathway. The alkaline phosphatase Pho8, a type II membrane protein, is the most characterized cargo of this pathway ( 37,38 ). The 74 kDa Pho8 proform is processed into a 72 kDa mature form upon delivery to the vacuoles, and some protein is further processed into a 66 kDa soluble mature form ( 39 ). In wild-type cells, most of the Pho8 detected was the mature form, and almost no proform was observed ( Fig. 3C ). In contrast, in vps21 ⌬ cells, a fraction of the Pho8 was present as the proform, as reported by others ( 40 ). In sur4 ⌬ and fen1 ⌬ cells, almost all of the Pho8 was the vacuole form, either the mature or soluble mature form ( Fig. 3C ), suggesting that the AP-3 pathway was not impaired. The soluble mature form was increased in the sur4 ⌬ and fen1 ⌬ cells compared with wild-type cells due to some unknown reason. Introduction of the vps21 ⌬ mutation into the sur4 ⌬ or fen1 ⌬ cells had no effect on the maturation of Pho8 ( Fig. 3C ), suggesting that the AP-3 pathway was intact in the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells. Overall, protein transport pathways utilizing the late endosome/MVB were impaired in the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells, but those pathways that bypass the organelle were not.

Sphingolipids containing VLCFAs function in Vps21related vesicular transport pathways
In yeast, the vast majority of VLCFAs are incorporated into sphingolipids as part of the hydrophobic backbone formed by ceramide, which contains FA such as a VLCFA ( 16,17 ). Thus, it is highly likely that VLCFAs acting in Vps21-involved traffi cking events function as components of sphingolipids. Each sphingolipid also comprises a polar head group on the ceramide. We considered the polar head groups of sphingolipids and investigated their roles in Vps21-related vesicular transport pathways. Although aur1 ⌬ cells, which lack all complex sphingolipids including IPC, MIPC, and M(IP) 2 C, are not able to grow in culture, yeast cells that lack only M(IP) 2 C ( ipt1 ⌬ cells) and those that lack MIPC and M(IP) 2 C ( csg1 ⌬ csh1 ⌬ cells) are able to grow without any retardation under normal vegetative conditions. We introduced the ipt1 ⌬ mutation or csg1 ⌬ csh1 ⌬ double-deletion mutation into vps21 ⌬ cells and examined the effect on cell growth and on the vesicular transports of the CPY and AP-3 pathways, as well as Cps1 transport and endocytosis.
Growth of the ipt1 ⌬ vps21 ⌬ cells at 30°C and 36°C was comparable to that of wild-type cells or those carrying either single mutation ( Fig. 6A ). Likewise, the csg1 ⌬ csh1 ⌬ at the plasma membrane and as intracellular punctuates, as well as in the vacuole lumen, indicating that endocytosis of Ste2-mRFP is severely affected in the vps21 ⌬ cells, consistent with reports that Vps21 is involved in endocytosis ( 36 ). In the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ double mutants, the Ste2-mRFP signal was mainly detected at the plasma membrane and in some punctate structures in the cytosol ( Fig. 5A and supplementary Fig. V). However, it was not clear whether the endocytotic pathways in the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ double mutants were more severely affected than those in the vps21 ⌬ cells from these fl uorescence microscopic results. Therefore, we also evaluated , and MRY91 ( sur4 ⌬ vps21 ⌬ ) cells, each harboring pOK529 (Ste2-mRFP), were grown at 30°C to log phase and subjected to fl uorescence microscopy. Left panels, DIC images; right panels, mRFP fl uorescence. Bar, 5 m. (B) Total lysates were prepared from cells indicated in (A) and separated by SDS-PAGE, followed by immunoblotting with anti-DsRed antibody or, to demonstrate uniform protein loading, anti-Pgk1 antibody. The lower panel represents the same immunoblot as the middle panel but a longer exposure to X-ray fi lm performed to enhance the image. transport of mRFP-Cps1 to the vacuole was impaired. Furthermore, impairments were observed in CPY and Pep4 transport ( Fig. 6D ) and in endocytosis of Ste2-mRFP ( Fig.  6C and supplementary Fig. V) in the csg1 ⌬ csh1 ⌬ vps21 ⌬ cells. Similar to results in sur4 ⌬ vps21 ⌬ cells, delivery of Pho8 to the vacuole via the AP-3 pathway was not affected in the csg1 ⌬ csh1 ⌬ vps21 ⌬ cells ( Fig. 6D ). These fi ndings suggest that sphingolipids are indeed involved in Vps21related vesicular transport at the late endosome/MVB. As the csg1 ⌬ csh1 ⌬ mutation and the sur4 ⌬ mutation affect mannose and VLCFA moieties of sphingolipids, respectively, both moieties are important in this transport.

DISCUSSION
VLCFAs have essential physiological functions in eukaryotes, and cells carrying double deletions of the redundant yeast FA elongases FEN1 and SUR4 , which are involved in VLCFA synthesis, cannot grow in any temperature ( 12 ). However, it still remains unclear in which cellular processes VLCFAs exert their essential functions. In the presented study using yeast cells that produce limited VLCFAs, we found that the VLCFA requirement is especially high at elevated temperatures. On the basis of this fi nding, we carried out a genetic screening to identify factors that are related to VLCFA functions at the molecular level, and we identifi ed the VPS21 gene as an important player in these processes. Subsequent genetic and cell biological analyses suggest that VLCFAs are involved in the vesicular transport were examined at the genetic level. The CORVET complex is composed of six subunits. Vps3 and Vps8 are specifi c to the CORVET complex, whereas four other subunits, Vps11, Vps16, Vps18, and Vps33, are shared with the homotypic fusion and vacuole protein sorting (HOPS) complex, which is involved in membrane tethering at the vacuole ( Fig. 7C ) ( 42 ). The HOPS complex also contains Vps39 and Vps41 as specifi c subunits in place of Vps3 and Vps8. The factors that are specifi c to the CORVET and HOPS complexes are interchangeable, and thus mixed complexes (i-CORVET containing Vps3 and Vps41, and i-HOPS containing Vps39 and Vps8) also exist ( 41 ). Severe synthetic growth defect at 36°C was observed in cells carrying a deletion of the SUR4 gene together with a deletion of VPS3 ( Fig. 7A ). Deletion of VPS8 , another gene encoding a CORVET-specifi c subunit, the Vps21 effector, conferred a synthetic growth defect at 36°C when combined with the SUR4 deletion ( Fig. 7A ), although the effect was weaker than that observed with the VPS3 deletion. This weak phenotype may be explained by the functional complementation of another Vps21 effector, Vps19 ( 43 ), as simultaneous deletion of VPS19 and VPS8 in sur4 ⌬ cells caused a severe synthetic growth defect at nearly the same levels as found in the sur4 ⌬ vps21 ⌬ and sur4 ⌬ vps3 ⌬ mutants ( Fig. 7B ). In addition, deletion of VPS41 or VPS39 , which encode , and MRY124 ( sur4 ⌬ vps19 ⌬ ) cells were grown to stationary phase, serially diluted at 1:10, spotted on YPD plates, grown for 48 h at 30°C or 36°C, and photographed against a dark background. (C) Schematic representation of the CORVET and HOPS complexes and interactions between their encoding genes and the SUR4 gene. Deletion of VPS21 or VPS3 causes severe synthetic growth defect when combined with the SUR4 deletion; these genes are illustrated as fi lled gray circles with thick lines. Simultaneous disruption of genes encoding the Vps21 effectors VPS8 and VPS19 also results in severe synthetic growth defect when combined with the SUR4 deletion; these genes are illustrated as white circles enclosed by thick lines. Two-headed arrows indicate protein-protein interactions between Rab GTPases and proteins involved in vesicle tethering. pathways active at the late endosome/MVB ( Fig. 6 ). This further indicates that not only the VLCFA moiety but also the polar head group of a sphingolipid is important for Vps21-related membrane fusion. Although the csg1 ⌬ csh1 ⌬ vps21 ⌬ cells possess IPCs, the simplest complex sphingolipids, they do not have mannose-containing MIPCs or M(IP) 2 Cs. We speculate that mannose facilitates clustering of sphingolipids through hydrogen bonds and stabilizes the membrane curvature generated by VLCFAs.
It is known that vacuoles in sur4 ⌬ and fen1 ⌬ mutants are fragmented ( 30,31 ), which we confi rmed in this study ( Figs. 4 and 5 and supplementary Fig. III). Interestingly, though, fragmentation of the vacuole was diminished in sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ cells. Cells carrying deletions for VPS21 possess a single enlarged vacuole and are known as having a class D vps phenotype ( 29 ). Hence, in regards to their vacuole morphology, sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ mutants exhibited an intermediate phenotype compared with each single mutant. This counteractive effect on vacuole morphology is in contrast to the synergetic effects observed here for cell growth at high temperature and vesicular transport. Thus, the mechanisms may differ for the effects of VLCFAs on vacuole morphology and on vesicular transport pathways active at the late endosome/MVB. Vacuole fragmentation is often accompanied by defects in vacuolar protein maturation and transport ( 44,45 ). However, neither sur4 ⌬ cells nor fen1 ⌬ cells exhibited signifi cant defects in transport of vacuolar proteins into their fragmented vacuoles ( Figs. 3 and 4 ). Interestingly, levels of the soluble mature form of Pho8, a highly processed form of the mature protein, were increased in sur4 ⌬ , fen1 ⌬ , sur4 ⌬ vps21 ⌬ , and fen1 ⌬ vps21 ⌬ cells ( Fig. 3C ), although the mechanism behind this enhanced processing is not readily understood. This phenomenon was not tightly coupled with vacuole fragmentation, as the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ double mutants, in which vacuole fragmentation was less prominent, also accumulated the soluble mature Pho8 at levels similar to those in the sur4 ⌬ and fen1 ⌬ single mutants. In addition, unlike the defects in the endosome-mediated transport pathways, this phenomenon was not enhanced synergetically by the combination of VPS21 deletion and deletion of SUR4 or FEN1 . Thus, this phenomenon was associated with the deletion of SUR4 or FEN1 but not with VPS21 deletion. Indeed, the soluble mature Pho8 was undetectable in the csg1 ⌬ csh1 ⌬ vps21 ⌬ triple mutant.
In the presented study, we revealed the function of VLCFAs in vesicular transport at the late endosome/MVB. However, considering that VLCFAs are essential for cell growth, it is highly likely that VLCFAs function in several membrane-involved processes, including other vesicular transport pathways, as reported for the secretory pathway. Indeed, we observed an accumulation of the p1 ER form of CPY in fen1 ⌬ vps21 ⌬ and sur4 ⌬ vps21 ⌬ cells, even under our mild conditions ( Fig. 3A ), suggesting that vesicular transport from the ER to the Golgi is also affected. Currently, the specifi c process(es) in which VLCFAs are actually involved, such as vesicle budding or fusion of the vesicle with the destination membrane, remains unclear.
pathways that act at the late endosome/MVB. Although VLCFAs have been implicated in the secretory pathway, our research provides the fi rst experimental evidence that VLCFAs are also involved in endosomal traffi cking.
The most severe synthetic growth defect was observed in sur4 ⌬ vps21 ⌬ or sur4 ⌬ vps3 ⌬ cells ( Figs. 2A and 7A ). Vps3 is a specifi c component of the CORVET complex, which captures the vesicle at the late endosome/MVB by physically interacting with Vps21 ( 41 ). Simultaneous deletion of the two Vps21 effectors ( VPS19 and VPS8 ) in sur4 ⌬ cells also caused a very severe synthetic growth defect ( Fig. 7B ). Thus, all the mutations that lead to loss of Vps21 function caused the severe synthetic growth defect when combined with sur4 ⌬ mutation. Since genetic interactions have been observed between genes functioning in closely related cellular processes in general, we speculate that Vps21 cooperates with VLCFAs in vesicle fusions with the late endosome/ MVB. Alternatively or in addition, both Vps21-mediated vesicle fusion with the late endosome/MVB and the function of VLCFAs in the late endosome/MVB may be highly important for the functional maintenance of this organelle.
Then what is the function of VLCFA in vesicular traffi cking, especially at the late endosome/MVB? During vesicle budding and fusion, local generation of highly curved membranes is essential, for example, at the neck of budding vesicles and at the contact sites of lipid mixing during vesicle fusion. The chain-length of VLCFAs allows them to span both leafl ets of the lipid bilayer, which would facilitate generation and stabilization of highly curved membranes. Such membrane-stabilizing functions of VLCFAs have been proposed for the formation of the nuclear pore complex, around which highly curved membranes also exist ( 10 ). Thus, it would seem likely that VLCFAs are involved in vesicular traffi cking by stabilizing highly curved membranes. Similar to conditions observed for vesicle budding and fusion, intraluminal vesicle formation at the late endosome/MVB requires generation of highly curved membrane at the neck of invagination. We detected mR-FP-Cps1 on the vacuole membrane in the sur4 ⌬ vps21 ⌬ and fen1 ⌬ vps21 ⌬ mutants ( Fig. 4A ), suggesting that VL-CFAs are also required for the formation of intraluminal vesicles. Such frequent generation of curved membranes at the late endosome/MVB may cause this organelle to be highly sensitive to VLCFA-limited conditions. In the present study, we found that the VLCFA requirement is significant at elevated temperatures ( Fig. 1A ). It is possible that curved membrane structures are destabilized at elevated temperatures due to enhanced membrane fl uidity, leading to a high VLCFA requirement.
In yeast, almost no free VLCFAs exist, and most VLCFAs are incorporated into sphingolipids, although small amounts are used for GPI-anchor generation. Since GPI-anchored proteins are localized in the plasma membrane and are not involved in vesicular transport at the late endosome/ MVB, the functions of VLCFAs revealed here must be elicited by those VLCFAs that are components of sphingolipids. Indeed, csg1 ⌬ csh1 ⌬ vps21 ⌬ triple-mutant cells, which are defective in MIPC synthesis, exhibited a growth defect at a high temperature and impairment in vesicular transport