Our strategy was first to identify soluble proteins whose metabolic stability could be controlled by a temperature shift, then attach these to a membrane anchor. Initially, we mutagenised a myc-tagged version of Ura3 and isolated temperature-sensitive mutants by their ability to grow at 37°C on 5-fluoro-orotic acid, which kills cells with active Ura3. These mutants were then screened for Ura3 degradation during a cycloheximide chase. Two mutants were obtained with non-conservative changes in distinct regions of the protein: Ura3-2 had Pro144 changed to Leu, and Ura3-3 had Asp243 changed to Gly, as well as the conservative mutation of Lys176 to Arg. The structural consequences of the two mutations are expected to be distinct [20]. Asp243 is at the start of a C-terminal helix of the Ura3 monomer, and its conversion to Gly might destabilise this helix. Pro144 on the other hand is in a turn on the opposite side of the molecule. Figure 1A shows that both these proteins persisted during a cycloheximide chase at 25°C, but after shifting to 37° they were rapidly degraded. In both cases this degradation was dramatically slowed in the temperature sensitive proteasome mutant cim3-1 (Figure 1B), suggesting that they were subject to proteasomal degradation, presumably ubiquitin-mediated.

Ura3 variants with random peptides appended to them have been shown to be substrates for the Doa10 ubiquitin ligase [21]. Our ts alleles had a longer lifetime in a doa10 deletion mutant, but the fact that degradation could still be observed implies that Doa10 is not the only ligase contributing to their fate (Figure 1A).

To test whether the metabolic instability of the Ura3 mutants could be extended to a fusion protein, we tagged them with GFP. The GFP fusions were degraded in a temperature-sensitive manner, and this was blocked by the ts proteasome mutant cim3-1 (shown for Ura3-3(GFP) in Figure 1D), although not significantly in a doa10 mutant (Figure 1C, D). However, we found that their degradation was substantially reduced by mutation of San1, a nuclear RING finger ubiquitin ligase that has been implicated in the turnover of ts proteins in the nucleus [16] (Figure 1C, D). Though Ura3 is primarily a cytoplasmic protein, it appears that there is sufficient contact between San1 and Ura3-GFP to allow ubiquitination and subsequent proteasomal degradation of the fusion protein. The enhanced dependence on San1 shown by the GFP fusions may relate to a weak affinity of GFP for the nucleus. Indeed, microscopic examination showed that Ura3-GFP was present throughout the cytoplasm and nucleus, though it was excluded from vacuoles (data not shown; see also below).

Doa10 is an ER membrane protein, and may share some functions with Hrd1, the other major ER ubiquitin ligase. However, Ura3-3(GFP) was efficiently degraded in a hrd1 deletion mutant, and as efficiently in a hrd1 doa10 double mutant as in the doa10 single mutant (Figure 1D). It appears therefore that this protein is not a substrate for Hrd1.

Having established that the ts alleles of Ura3 could induce degradation of attached proteins, we fused the mutant Ura3-GFP chimeras to the plasma membrane SNARE protein Sso1. Fluorescence microscopy confirmed that the bulk of the Ura3-2(GFP-Sso1) construct was indeed at the plasma membrane (Figure 2A). Cells expressing these constructs remained temperature-sensitive for growth in the absence of uracil (Figure 2B). Strikingly, however, temperature shift did not induce degradation of the resultant fusion proteins (Figure 2C).

We extended these studies by fusing Ura3-2(GFP) to three more SNARES: Snc1, also found mainly on the plasma membrane; Vam3, found on the vacuolar membrane; and Use1, an ER resident SNARE. The Snc1 and Vam3 fusions also avoided degradation after temperature shift (Figure 2D). In contrast, although GFP-Use1 itself was stable, addition of the Ura3-2 protein resulted in rapid degradation at 37°C (Figure 2E). Thus, a SNARE fusion can be degraded, but apparently only if it is in the ER where quality control of membrane proteins typically occurs. When located on the plasma membrane the fusion proteins escape proteasome-mediated degradation, even though they are exposed to the cytoplasm. Furthermore, though some protein could be seen on the vacuolar membrane, even after prolonged incubation at 37° there was no sign of GFP accumulation within the vacuole (Figure 2A). Nor could we detect free GFP by immunoblotting (not shown); this is a characteristic product of the vacuolar degradation of fusion proteins since GFP itself is protease resistant.

As an independent test of this phenomenon, we used a mutant (L57A, A66T) of the N-terminal 92 amino-acid region of the bacteriophage lambda repressor. The A66T mutation is present in the c1857 temperature-sensitive repressor [22], and the L57A mutation has been shown to induce physical unfolding at temperatures above 20°C [23]. Unlike the Ura3 mutants, this protein is monomeric and it should also unfold at normal yeast growth temperatures, thus avoiding any potential complications due to heat stress of the cells. A GFP fusion to the mutant repressor was stable at 16°, but was degraded at 30° in wild type cells. It was not degraded in the cim3-1 proteasome mutant, but was degraded in cells lacking Doa10, San1 or both (Figure 3). Thus, it is likely to be a substrate for one or more ubiquitin ligases that do not recognise the Ura3 mutants, or possibly for the proteasome itself.

Despite these differences between the properties of the lambda repressor and Ura3 mutants, fusion of the repressor-GFP construct to Sso1 inhibited its degradation, just as with the Ura3 mutants (Figure 3). However, equivalent fusions to the ER SNARE Use1 were degraded, indicating that they were capable of becoming substrates for the ER-associated degradation machinery (Figure 3).

It has been reported that some abnormal proteins, such as a temperature sensitive mutant of Ubc9, accumulate in discrete punctae prior to degradation [19]. If this is a necessary step, it could explain why membrane-tethered proteins, which presumably are restricted in their movement, suffer a different fate from soluble ones. We therefore investigated the distribution of our soluble GFP chimeras following temperature shift.

The Ura3-2(GFP) construct was quite uniformly distributed at 25°, with only an occasional bright spot (Figure 4A). However, within 30 minutes of shifting the temperature to 37° in at least 80–90% of the cells the majority of the protein was in large fluorescent spots, consistent with aggregation. Surprisingly, however, this aggregation was almost completely prevented by the addition of cycloheximide prior to the temperature shift (Figure 4A). This dramatic difference led us to investigate whether the degradation of the protein was affected by cycloheximide, which is used in our standard assay. We used the glucose-repressible GAL1 promoter to drive synthesis of Ura3-2(GFP) at low temperature, added glucose to repress transcription for four hours, then subjected the cells to high temperature. Figure 4B shows that the fusion protein was degraded just as it was in the presence of cycloheximide.

We also examined the GFP-tagged lambda repressor mutant, and found that it had a dispersed distribution with a few small bright dots, whether the cells were kept at 16° or had been shifted to 30° for one hour (Figure 4C). Taken together, these results indicate that the distribution of our mutant proteins does not correlate with their metabolic stability. There is thus little reason to believe that substantial aggregation or specific localisation is an obligatory or regulating step in their destruction. We cannot, of course, rule out the possibility that small or transient clusters are a necessary intermediate.

Yeast strains unless otherwise stated were from the Open Biosystems knockout collection in a BY4742 (MATα uraΔ0 leu2Δ0 his3Δ1 lys2Δ0) background. The doa10 san1 double mutant was made by homologous recombination with a replacement of the DOA10 open reading frame in the BY4742 san1 knockout strain by a PCR product encoding the Schizosaccharomyces pombe HIS5 gene. The doa10 hrd1 double mutant was made similarly. The cim3-1 and corresponding wild type strain [28] were obtained from Carl Mann (SBIGeM CEA/Saclay France).

PmUra414 was made by insertion of a c-myc tagged PCR product from the URA3 gene into a version of pRS414 bearing the TPI1 promoter. The ura3 mutants were made by error-prone PCR mutagenesis [29]. The mutated PCR products were co-transformed into SEY6210 (MATα ura3-52 leu2-3, -112 his3-Δ200 trp1-Δ901 lys2-801 suc2-Δ9) together with Nco1-Bpu101 cut pmUra414 and selected on plates lacking tryptophan. A pool of colonies was then plated onto 5-fluoroorotic acid plates at either 25°C or 37°C. Colonies from these plates were streaked out and further tested for temperature sensitive growth on plates lacking uracil. Rescued plasmids were sequenced.

In-frame GFP fusions were made between downstream EcoR1 and BamH1 sites. The SSO1 fusion was made by replacement of the GFP with a GFP-SSO1 fusion [30]. Fusions were similarly made to the N-termini of SNC1 and VAM3. USE1 was cloned by PCR into a TPI1-promoter-GFP containing version of pRS416 and the GFP-USE1 fusion used to replace GFP in the URA3 fusion constructs.

The region encoding the first 92 amino acids from the c1857 bacteriophage lambda repressor was cloned by PCR and subsequently mutated by PCR to introduce a L57A change [23]. This fragment was subsequently used to replace the URA3 moiety in the GFP fusion constructs. Galactose-inducible constructs were made by placing the fusion constructs in a version of YCpLac33 carrying the GAL1/10 promoter fragment. Sequences of all PCR generated constructs were verified by DNA sequencing.

Cells were grown into early log phase and after addition of cycloheximide (to 500 µg/ml) were shaken at the temperatures shown, and samples taken at the indicated times. These were spun down and stored on dry ice before being subjected to alkaline lysis and solublisation as described by [31]. Samples were run on PAGE and western blots probed with anti-GFP (Roche Diagnostics) and anti PGK (Molecular Probes). In experiments using the cim3-1 strain, cells were preincubated for 1 hour at 37°C prior to the addition of cycloheximide. For the galactose shutoff experiments cells were transferred from medium containing 2% galactose and raffinose into medium containing 2% glucose for 4 hours before temperature shift.

Live cell imaging was performed on concanavalin-A coated slides using a Zeiss LSM510. Images were adjusted for contrast and brightness, and in some cases, they were blurred to filter noise, by using Adobe Photoshop (Adobe Systems, Mountain View, CA).