Localization of Ykt6 in yeast cells
Despite its localization to the cytoplasm and membranes, the function of Ykt6 seems to be restricted to membranes because a transmembrane-anchored Ykt6 can partially complement the wild-type function (9) (our unpublished observations). In contrast, mutants of Ykt6 that can no longer stably associate with membranes cannot complement loss of the wild-type protein (17,20). However, attempts to determine the localization of membrane-associated wild-type Ykt6 failed because of the high cytosolic background. Therefore, we decided to place the YKT6 open reading frame (ORF) under the control of the inducible GAL1 promoter. If placed into glucose-containing medium, the GAL1-YKT6 strain grew like wild type for 13 h and then slowed down significantly (Figure 1A). We then compared the subcellular distribution of Ykt6 at three time-points after depletion. In wild-type cells, Ykt6 was found in the vacuole-containing membrane fraction (P13) and in the soluble S100 fraction throughout the experiment (Figure 1B). In GAL1-YKT6 cells, the depletion was most obvious for the cytosolic Ykt6 S100 fraction, which was successively lost after 9 and 13 h. The membrane-bound fraction remained fairly constant until the 9-h time point. We reason that cytosolic Ykt6 is recruited to membranes rather than being selectively degraded because cells still expanded their total membranes because of growth and division.
Figure 1. Depletion of Ykt6 from yeast cells. A) Growth curves. Wild-type (wt) yeast cells (CUY1598, see Table S1) or cells with YKT6 under the control of a GAL1 promoter (CUY1870) were grown in YP + galactose before being transferred into glucose-containing YPD medium for 17 h at 30°C. After 10 h, cells were diluted and grown again at 30°C. B) Subcellular fractionation. Yeast strains BY4741 and CUY1870 were subjected to subcellular fractionation at the indicated time-points. Cells were osmotically lysed and cleared by low-speed centrifugation (300 × g, 4°C). Lysates were then fractionated by centrifugation in two subsequent steps to yield P10 (10 000 × g, 4°C, 15 min) and P100 (100 000 × g, 4°C, 60 min) pellets and the S100 supernatant. These were analysed by SDS–PAGE and western blotting using antibodies against Vti1 and Ykt6. C) Localization of Ykt6-eGFP. Cells expressing eGFP-tagged Ykt6 (CUY2367) were grown in YPG or shifted to YPD for 9 h and then analysed by fluorescence microscopy. DIC, differential interference contrast.
Download figure to PowerPoint
To visualize Ykt6, we generated a strain that expressed green fluorescent protein (GFP)-tagged Ykt6 from a GAL1 promoter and depleted cells for 9 h. The GFP-tagged Ykt6 was functional and behaved like the wild-type protein (data not shown). The cytosolic portion of Ykt6–GFP masked any membrane staining if expressed from its natural (data not shown) or the GAL1 promoter (Figure 1C). Upon depletion, Ykt6 accumulated on membranes (Figure 1B) and became visible on circular structures and dots scattered throughout the cell (Figure 1C).
To identify the compartments, to which Ykt6 localized under these conditions, we added C-terminal red fluorescent protein (RFP) tags to selected markers of the endoplasmic reticulum (ER) (Sec63), the Golgi (Mnn9), the endosome (Snx41) and the vacuole (Vac8). We observed a clear colocalization of GFP–Ykt6 with RFP-tagged Vac8 and Mnn9 (Figure 2A,C). Likewise, some Ykt6–GFP dots colocalized with the endosomal protein Snx41 (Figure 2D), but we did not observe Ykt6 on peripheral or nuclear ER structures stained by Sec63 (Figure 2B). It is possible that Ykt6 was present on the ER or plasma membrane prior to the depletion and that the quantities remaining after a 9-h depletion are below the detection limit of our system. Our localization data are consistent with the function of Ykt6 in Golgi, endosomal and vacuolar SNARE complexes.
Figure 2. Localization of GFP–Ykt6 in live cells. Yeast strains CUY325, CUY331, CUY851 and CUY327 expressing Ykt6-eGFP and RFP-tagged Vac8 (A), Sec63 (B), Mnn9 (C) or Snx41 (D) were shifted from an overnight culture grown in galactose-containing medium to glucose-containing medium for 9 h before being analysed by fluorescence microscopy. Scale bar: 10 μm.
Download figure to PowerPoint
Intramolecular interactions and localization of Ykt6
We screened mutants of conserved residues within the longin domain of Ykt6 to identify mutants that are affected in their dynamic localization. We introduced plasmids carrying the selected mutants into ykt6 deletion strains that were kept viable by the simultaneous presence of a plasmid coding for the Ykt6 wild-type protein. Upon loss of the wild-type plasmid, a ykt6 deletion mutant died, as did the F42S mutant (Figure 3A). These data are consistent with previous findings that mutations of F42 to a hydrophilic residue inside the hydrophobic pocket of the longin domain render Ykt6 unable to complement loss of the wild-type protein by preventing the interaction between longin and SNARE domains (21). Among the other mutants tested, only the introduction of a polar serine residue at position 17 (L17S) produced a non-functional protein. To determine the localization of these mutants, we generated GFP-tagged proteins and expressed them in the wild-type background. As shown in Figure 3B, the L17S and H67A mutants displayed a strong cytosolic background comparable to the wild type. However, the F42S mutant accumulated on several organelle membranes and inside the vacuole (Figure 3B), consistent with previous observations in mammalian cells (19). This suggests that the F42S mutant lost the ability to cycle between the membranes and the cytosol. We therefore decided to use this mutant as a tool to explore the sorting and palmitoylation of Ykt6 in more detail.
Figure 3. Analysis of Ykt6 longin domain mutants. A) ykt6Δ cells containing a URA3 plasmid coding for wild-type (wt) Ykt6 and plasmids expressing the indicated Ykt6 mutants (CUY2336, CUY1613, CUY2345, CUY2337, CUY2341, CUY2560, CUY2561 and CUY2559) were spotted as serial dilutions on synthetic complete dextrose without histidine (SDC-HIS) to select for the mutant plasmid or 5-Fluoroorotic acid (5-FOA) to displace the plasmid coding for wild-type Ykt6. B) Localization of GFP-tagged Ykt6 wild-type and mutant proteins by fluorescence microscopy in yeast strains CUY2484, CUY2485, CUY2486 and CUY2487.
Download figure to PowerPoint
First, we introduced Ykt6 F42S as an additional GFP-tagged copy into selected temperature-sensitive mutants of the secretory pathway. If Ykt6 F42S associates with membranes initially at the ER, we would expect that a block in vesicular transport between the ER and the Golgi (sec12), within the Golgi (sec7) or to the plasma membrane (sec6) would lead to an accumulation of the protein. However, our data do not give a clear picture (Figure 4A). Under non-restrictive conditions, wild-type and mutant cells show the same distribution of Ykt6 F42S. If shifted to 37°C, some Ykt6 F42S accumulated in the vacuole in wild-type cells (discussed below). The same redistribution is also seen in the sec12 and sec6 mutants. Only in the sec7 mutant, some additional dot-like structures appear that are distinct from the vacuolar staining and could reflect a Golgi accumulation. Attempts to address this question using wild-type Ykt6 and our depletion assay described above failed because of technical reasons (data not shown).
Figure 4. Sorting and release of Ykt6 F42S. A) Localization of Ykt6 F42S in selected temperature-sensitive (ts) mutant strains CUY2727, CUY2728, CUY2730 and CUY2729. GFP-tagged Ykt6 F42S was expressed in wild-type (wt), sec6, sec7 or sec12 ts strains and grown in liquid medium at 23°C or 37°C for 1 h before being analysed by fluorescence microscopy. Scale bar: 10 μm. B) Analysis of Ykt6 release from membranes in yeast strains BJ3505 and CUY3346. Vacuoles were obtained after osmotic lysis of cells and flotation on a discontinuous Ficoll density gradient from the indicated strains. Wild-type vacuoles or vacuoles expressing myc-tagged Ykt6 F42S were incubated for 10 min at 26°C in reaction buffer together with 10 μm CoA and with or without ATP-regenerating system and centrifuged for 10 min at 12 000 × g. Pellets and TCA-precipitated supernatants were analysed by SDS–PAGE and western blotting using an antibody to Ykt6. C) Release of Ykt6 in a mutant lacking the acyltransferase Pfa3 in strains CUY1465 and CUY1494. The experiment was carried out as described for (B) except that membranes were probed for Vti1 and Sec17 in addition to Ykt6. P, pellet; S, supernatant.
Download figure to PowerPoint
In a second approach to address Ykt6 dynamics, we isolated vacuoles carrying myc-tagged F42S Ykt6 and followed ATP-dependent release of the protein (20). As seen in Figure 4B, wild-type Ykt6 is released from membranes into the supernatant in a reaction dependent on addition of ATP and Sec18. In contrast, the myc-tagged Ykt6 F42S mutant protein was not released from membranes regardless of the absence or presence of ATP. This observation is consistent with the F42S mutant accumulating on membranes in vivo.
To address the question whether the vacuolar acyltransferase Pfa3 plays a role in the dynamics of Ykt6, we analysed release of the protein from pfa3Δ vacuoles but observed no change compared with wild-type membranes (Figure 4C). The well-described release of Sec17 was also unaltered under these conditions (23).
It was previously shown for wild-type Ykt6 that membrane localization is determined by its prenyl and palmitoyl anchors (20). To test whether this also applies to the F42S mutant, point mutations in the cysteine residues of the C-terminal CCIIM sequence were generated. We then localized the GFP-tagged Ykt6 wild-type (CC) and mutant (CA, AC and AA) proteins by fluorescence microscopy and subcellular fractionation (Figure 5A). The mutant lacking the palmitoyl anchor (AC) was still found on membranes, whereas loss of the prenyl anchor (CA) shifted a significant portion of Ykt6 to the cytosolic S100 fraction. However, complete cytosolic localization was only observed if both anchors (AA) were missing, indicating that the dual lipid anchor promotes membrane localization.
Figure 5. Analysis of Ykt6 membrane anchoring. A) GFP-tagged Ykt6 F42R proteins with mutations in the C-terminal CCIIM sequence (CC, palmitoylated and farnesylated; CA, non-farnesylated and AC, non-palmitoylated) were analysed by fluorescence microscopy in strains CUY2684, CUY2555, CUY2685 and CUY2556; Scale bar: 10 μm (left) and or subcellular fractionation (right). The analysis was performed as before. B) Localization of fusion proteins anchored to membranes through the Ykt6 membrane targeting CCIIM motif by fluorescence microscopy in strains CUY2332, CUY2397, CUY2334, CUY2398, CUY2573 and CUY2574. CaaX (yEGFP fused to CCIIM), Ykt6ΔC (Ykt6 longin domain fused to CCIIM), Ykt6ΔN (Ykt6 coiled-coil domain fused to CCIIM), Ykt6N–Vam3CC (fusion of Ykt6 longin domain with the Vam3 coiled-coil motif and CCIIM), Ykt6N–Nyv1CC (fusion of Ykt6 longin domain with the Nyv1 coiled-coil motif and CCIIM) and Vam3ΔTMD (Vam3 with CCIIM replacing the transmembrane domain). Scale bar: 10 μm. C) Scheme of constructs used for localization in (B).
Download figure to PowerPoint
In previous studies, we showed that the CCIIM sequence can be transplanted to the vacuolar SNAREs Vam3 and Nyv1, which we then co-purified with vacuoles (20,24). Both proteins did not get released from membranes, in contrast to the Ykt6 protein. To analyse whether the C-terminal anchor determines the subcellular localization of Ykt6, we fused CCIIM sequence of Ykt6 to the longin domain (Ykt6ΔC) and the coiled-coil domain (Ykt6ΔN). The Ykt6ΔC mutant localized to the plasma membrane and mostly punctate internal structures, whereas the Ykt6ΔN construct localized mainly to the plasma membrane (Figure 5B). Next, we linked the Ykt6 longin domain to the Vam3 and Nyv1 coiled-coil domains and replaced their respective transmembrane domains with CCIIM anchor of Ykt6 (Ykt6N–Vam3CC and Ykt6N–Nyv1CC, respectively) and compared it with Vam3 with its transmembrane domain replaced by CCIIM (Vam3ΔTMD; Figure 5B). None of the two chimaeras showed any remarkable alteration in localization when compared with Vam3ΔTMD. Apparently, the longin domain of Ykt6 is not sufficient to impart localization to internal membranes onto either the Vam3 or the Nyv1 coiled-coil domains. These observations support the notion that its intramolecular interaction and not either domain alone determines the sorting and dynamics of Ykt6.
Dynamics and palmitoylation of Ykt6
Dynamics of Ykt6 must be governed by a consecutive palmitoylation and depalmitoylation event, which seems to be linked to its intramolecular interactions (17). Several studies indicate that a family of acyltransferases with a common cytosolic DHHC (Asp-His-His-Cys) motif mediates palmitoylation. In yeast, seven isoforms exist. Specific targets have been reported for some DHHC proteins, but they also seem to have a substantial overlap in substrate recognition (25) (our unpublished observations). For instance, loss of the vacuolar DHHC Pfa3 leads to a mislocalization of the fusion factor Vac8 to the cytosol (26,27).
To address the question how Ykt6 localizes in the absence of single or multiple DHHCs, we employed our previously described depletion assay (Figure 1) in strains lacking selected acyltransferases (25). These depletions – contrary to those presented in Figures 1 and 2– were carried out with a strain still containing the genomic copy of wild-type YKT6 because we were not able to delete it in our DHHCΔ background strains. When we depleted the Ykt6 protein to observe potential changes in its membrane localization, only a deletion in five DHHC proteins, which included Pfa3, led to a dot-like accumulation of Ykt6 proximal to the vacuole (Figure S2A). This was not observed in all cells, and some accumulation was already seen in wild-type cells. We speculate that the presence of the untagged version of Ykt6 obstructed the accumulation of GFP-tagged Ykt6 on membranes.
We then tried to address acyltransferase-mediated palmitoylation of Ykt6 by expressing the GFP-tagged F42S mutant under the control of the GAL1 promoter in various DHHC deletion backgrounds. However, we were unable to detect a strong effect on the localization of GFP–Ykt6 F42S because of the loss of any one or multiple DHHC proteins (Figure S2B), suggesting that multiple DHHCs can palmitoylate Ykt6. In a further attempt to test the effect of deletions in DHHCs, we expressed the Ykt6ΔN construct (Figure 5C) in the same background strains. However, removal of the longin domain and expression of this construct in the different DHHCΔ backgrounds did not show any changes in its localization compared with the wild-type strain (Figure S2C).
Therefore, we decided to take advantage of a DHHC overexpression approach in yeast lacking five of the seven DHHC proteins to address the question of acyltransferase-mediated palmitoylation of Ykt6 F42S. If the vacuolar DHHC protein Pfa3 is overexpressed, it localizes to the vacuole and a prevacuolar compartment and can redirect proteins to the vacuolar surface (H. Hou, C. Meiringer, K. Subramanian and C. Ungermann, submitted). We expressed the GFP-tagged F42S mutant under the control of the constitutive NOP1 promoter in wild-type cells and the 5×DHHC deletion background overexpressing Pfa3 from the GAL1 promoter (Figure 6A). The latter strain grown in glucose is depleted of Pfa3, whereas in galactose-containing media, Pfa3 is strongly overexpressed compared with wild-type levels of the protein. Ykt6 F42 was found on the plasma membrane and on internal membranes including the vacuolar rim in the 5×DHHCΔ background under Pfa3 depletion conditions (Figure 6A). In contrast, some Ykt6 F42S appears to be sorted to the vacuolar lumen in wild-type cells under these conditions (compare also with Figure 5). Strikingly, overexpression of Pfa3 in the DHHC deletion background led to an almost complete shift of Ykt6 F42S into the vacuolar lumen (Figure 6A). In the wild-type background, some increase in vacuolar accumulation was also observed; however, the plasma membrane and most of the endomembranes still retained considerably more Ykt6 F42S compared with the Pfa3 overexpressor (Figure 6A). When we tested wild-type Ykt6 in the same strain backgrounds, localization of the protein was always cytosolic (Figure S1A).
Figure 6. Depalmitoylation of Ykt6 is necessary for its recycling. A) Localization of eGFP-Ykt6 F42S in the wild-type (wt) yeast strain CUY3180 and the 5× DHHCΔ background (CUY2929) with Pfa3 under the control of the GAL1 promoter. Cells were grown either in glucose- – depletion of Pfa3 – or in galactose-containing medium – overexpression of Pfa3. In (B), an eGFP-Ykt6 F42S version lacking the palmitoylatable cysteine (CCAC) is expressed in the same strain backgrounds as above (CUY3181 and CUY2968, respectively). C) Localization of GFP-tagged Vam3–CaaX in the absence or presence of overproduced Pfa3. The analysis was performed as in (A) with strains CUY3242 andCUY3065. Scale bar: 10 μm. D) Ykt6 F42S and Vam3–CaaX localization in strains overexpressing either Pfa4 or Pfa5 (CUY3079 and CUY3081 as well as CUY3082 and CUY3084, respectively).
Download figure to PowerPoint
To control that relocalization of the F42S mutant was not because of differences in the growth conditions, but attributable to palmitoylation, we tested how a non-palmitoylable mutant (AC) would localize. Indeed, overexpression of Pfa3 had no effect on the localization of this mutant (Figure 6B). This is in agreement with our finding of intravacuolar accumulation of Ykt6 F42S carrying the wild-type membrane anchor, whereas the AC mutant was only found on vacuolar surfaces in the wild-type strain background (compare with Figure 5A).
The vacuolar t-SNARE Vam3 travels to the vacuole by the adaptor protein (AP)3 pathway (28). It thereby bypasses the endosome and also has no reported function on this organelle. To address the question whether any vacuolar protein carrying lipid anchor of Ykt6 (CCIIM) is sorted into the vacuolar lumen upon Pfa3 overexpression, we investigated the localization of GFP–Vam3 with its transmembrane domain replaced by CCIIM (Vam3–CaaX) in the same strains as the Ykt6 F42S mutant. Akin to Ykt6 F42S, Vam3–CaaX is found on vacuoles and the plasma membrane and does not get released during fusion (20), (Figure 6C). However, in the 5×DHHCΔ background, Vam3–CaaX localized exclusively to the plasma membrane when cells were depleted of Pfa3. Upon overexpression of Pfa3, Vam3–CaaX accumulated on the vacuolar membrane. However, Vam3–CaaX was not sorted into the vacuolar lumen. To control that interference of the AP3 pathway does not play a role in the sorting of Vam3–CaaX, we deleted the AP3 coat subunit gene APL5 in the 5×DHHCΔ background overexpressing Pfa3, thus forcing Vam3–CaaX into the MVB pathway. Neither Vam3–CaaX nor Ykt6 F42S was affected in its localization because of the apl5 deletion (Figure S1B). Therefore, both proteins take the same route to the vacuole by the endocytic pathway but encounter a different fate when MVBs are formed.
To address the question whether any other acyltransferase affects sorting of Ykt6 F42S and Vam3–CaaX, we overproduced ER-localized Pfa4 and plasma membrane-localized Pfa5. Ykt6 F42S was sorted into the vacuolar lumen upon overproduction of Pfa4 but not Pfa5 (Figure 6D). In contrast, both Pfa4 and Pfa5 imparted vacuolar localization upon Vam3–CaaX, although the effect was stronger for Pfa4 overproduction. We can only speculate that Vam3–CaaX may be more readily palmitoylated by Pfa5 than Ykt6 F42S.
Next, we directly investigated the palmitoylation state of Ykt6 F42S by the biotin switch method (29). Upon overproduction of Pfa3, GFP–Ykt6 F42S becomes strongly palmitoylated (Figure 7A), whereas a mere shift to galactose of the 5×DHHC deletion strain not overexpressing Pfa3, which showed some sorting of Ykt6 F42S into the vacuole, did not yield any palmitoylated Ykt6 F42S. Similarly, relocalization of Vam3–CaaX to the vacuole upon Pfa3 overproduction also coincided with an increase in palmitoylation (Figure 7B). Sorting into the vacuole through MVBs therefore coincides with an acyltransferase-mediated palmitoylation of the Ykt6 F42S mutant. This increase in palmitoylation is distinct from the Ykt6 autoacylation, which can be observed in vitro for wild-type protein but only poorly for the Ykt6 F42S mutant (Figure S3) (30). Thus, Ykt6 F42S is directly palmitoylated by Pfa3 and as a consequence gets sorted into the vacuole where it is then degraded.
Figure 7. Sorting of Ykt6 F42S depends on palmitoylation and ESCRT. A) Palmitoylation of Ykt6 F42S in the 5×DHHCΔ strain with Pfa3 either under the control of the GAL1 promoter (CUY2929) or knocked out (CUY2966) were analysed by the biotin switch method after growth in galactose-containing medium. Free cysteines in lysates were quenched by incubation with 25 mm NEM followed by treatment with hydroxylamine (H3NO), pH 7.4, in the presence of 300 μm biotin-BMCC to specifically label palmitoylated cysteines. Afterwards, biotinylated proteins were precipitated using neutravidin–agarose. The 5% load corresponds to 5% of the input before the precipitation by neutravidin–agarose. Proteins were separated by SDS–PAGE, blotted and blots decorated with an antibody against Ykt6. B) Palmitoylation of Vam3–CaaX. Strain CUY3065 was grown in parallel in glucose- and galactose-containing medium and analysed by the biotin switch method as described above. C) ESCRT dependence for Ykt6 F42S sorting into MVBs. Yeast strains CUY3008 and CUY3086 were analysed as described in Figure 6.
Download figure to PowerPoint
Finally, we asked whether sorting of Ykt6 F42S into the lumen of the vacuole depends on the ESCRT machinery. Indeed, luminal sorting of Ykt6 F42S under conditions were Pfa3 is overexpressed depends on palmitoylation and ESCRT function: deletion of the class E gene VPS4 relocalized Ykt6 F42S to the endosome and the vacuolar surface (Figure 7C) and prevented its entry into the vacuolar lumen. Deletion of VPS4 also resulted in an accumulation of Vam3–CaaX in distinct dots adjacent to the vacuole when Pfa3 was overexpressed (Figure 7C), indicating that it indeed travels by the endosomal pathway towards the vacuole.
Both Ykt6 F42S and Vam3–CaaX are anchored identically to membranes and follow the same transport pathways. However, whereas Vam3–CaaX ends up on the vacuolar surface, Ykt6 F42S is internalized into the MVB. Internalization is dependent on the ESCRT machinery and stably membrane-anchored Ykt6. Ykt6 can be palmitoylated by the DHHC acyltransferase network and appears to depend on depalmitoylation to avoid sorting into MVBs and ultimately the vacuolar lumen.