Biochemical Journal

Research article

Analysis of the dual function of the ESCRT-III protein Snf7 in endocytic trafficking and in gene expression

Peter Weiss, Stefanie Huppert, Ralf Kölling


ESCRT (endosomal sorting complex required for transport)-III mediates the budding and scission of intralumenal vesicles into multivesicular endosomes in yeast. For the main ESCRT-III subunit Snf7, an additional role in activation of the transcription factor Rim101 (the ‘Rim pathway’) is now also firmly established. In the present study, we investigate how these two Snf7 functions are related to each other. By generating SNF7 mutations that severely affect endocytic trafficking, but leave the Rim pathway function intact, we show that the two functions of SNF7 can be separated genetically. We analysed in detail how the SNF7 mutations affect the interaction of Snf7 with its various binding partners. Although the interactions with proteins Rim13 and Rim20, necessary for the Rim-pathway-related functions, were not altered by the mutations, there was a strong effect on interactions with components of the ESCRT pathway. The interactions, as measured by co-immunoprecipitation, with the ESCRT-III subunits Vps20 and Vps24 were strongly increased by the mutations, whereas the interactions with proteins Vps4 and Bro1, acting downstream of ESCRT-III, were reduced. As Vps4 is required for disassembly of ESCRT-III these results suggest that ESCRT-III is more stable in our SNF7 mutants. In line with this notion, a higher fraction of mutant Snf7 protein was detected at the membrane. Upon a shift to alkaline pH, a stronger binding signal for virtually all interaction partners, except Vps4, was observed. This indicates that the ESCRT network at the endosomal membrane is more extensive under these conditions.

  • endocytosis
  • endosomal sorting complex required for transport (ESCRT)
  • multivesicular body
  • Rim101 pathway
  • Snf7/chromatin modifying protein 4 (CHMP4)
  • vacuolar protein sorting


During endocytosis, cell surface proteins are internalized and transported, via several endosomal intermediates, to the lysosome/vacuole for degradation and endocytic cargo proteins are marked for degradation by ubiquitination [1]. Ubiquitinated membrane proteins are then recognized by the endocytic machinery and are incorporated into vesicles that bud into the interior of late endosomal MVBs (multivesicular bodies). After fusion of the MVB with the lysosome/vacuole, the internal vesicles are released into the lumen of the hydrolytic compartment, where they are degraded.

Cargo recruitment and internal vesicle formation is mediated by ESCRT (endosomal sorting complex required for transport) proteins [2]. These proteins can be grouped into several complexes, called ESCRT-0, -I, -II and -III, which are thought to act sequentially in cargo recruitment and vesicle formation at the endosomal membrane to form an ‘ESCRT pathway’. The ESCRT complexes are released from the endosomal membrane by the AAA-ATPase Vps4 for further rounds of membrane budding [3].

ESCRT-III in conjunction with Vps4 appears to be the ‘machine’ that drives internal vesicle formation. The topology of membrane budding mediated by ESCRT-III is unique in that it is directed away from the cytoplasm. Other membrane budding and fusion processes, such as retrovirus budding [4] and cytokinesis [5,6], with the same topology also make use of the ESCRT-III machinery. The ESCRT-III/Vps4 machinery for membrane budding and fusion appears to be a very ancient system as it is also conserved in archaea, where it plays a role in cytokinesis [7]. Yeast ESCRT-III consists of the four core subunits, Snf7, Vps2/Did4, Vps20 and Vps24 [8], and two ESCRT-III-associated subunits, Did2 [9] and Mos10/Vps60 [10], and all these subunits are homologous to each other. Snf7 has the propensity to homo-oligomerize and is the most abundant subunit in the complex. Vps20 induces Snf7 oligomerization, whereas the other core subunits Vps2 and Vps24 seem to limit oligomerization by recruiting Vps4 and inducing disassembly [11].

Filament formation appears to be a general property of ESCRT-III proteins. Upon overexpression of either of the human Snf7 homologue, CHMP (chromatin modifying protein)-4A and CHMP-4B, in mammalian cells circular arrays of filaments on membranes were observed [12]. Budding and tubulation of membranes away from these circular scaffolds could be observed upon co-expression of CHMP4A in an ATP-hydrolysis deficient Vps4B mutant. Likewise, helical filament formation in vitro has been observed with purified yeast Vps24 [13] and human CHMP2A, the homologue of Vps2 [14]. Recently, it has been demonstrated that budding of intralumenal vesicles could be reconstituted in vitro with purified ESCRT-III components [15]. Vps20, Snf7 and Vps24 were sufficient to detach intralumenal vesicles into giant unilamellar vesicles, whereas Vps2 and Vps4 were required for ESCRT-III recycling and additional rounds of vesicle budding [15].

Although all ESCRT-III components function at the same step of MVB formation, individual ESCRT mutants also show distinct phenotypes [16], which points to a specific role of ESCRT proteins in gene regulation. It is now firmly established that Snf7 plays a role in activation of the transcription factor Rim101 [1720]. Under inducing conditions, such as a shift to alkaline pH, the inactive Rim101 precursor is processed to its active form by C-terminal cleavage [21]. This C-terminally processed transcription factor then translocates to the nucleus, where it regulates the expression of specific target genes [22]. We recently showed that the SUC2 gene, coding for the enzyme invertase, is also a target of the Rim pathway, offering an explanation for the sucrose non-fermenting phenotype of SNF7 mutants [16]. The current view is that Rim101 cleavage takes place on a platform consisting of Snf7, the Bro1-domain protein Rim20 and the calpain-like protease Rim13 [23]. In addition, Rim101 activation depends on a number of other Rim proteins thought to be involved in pH signalling [24] and on components acting upstream of Snf7 in the ESCRT pathway [1720], although at present it is unclear what precise role these proteins play.

In the present study, we investigate the connection between the two functions of Snf7, i.e. in the ESCRT pathway and in gene expression, and found that the two functions can be separated genetically. We were able to generate Snf7 mutants that are selectively blocked in endocytic trafficking, but that are normal with respect to gene expression. A detailed phenotypic analysis of the Snf7 mutants showed that Snf7 binding to downstream ESCRT pathway proteins, like Vps4 and Bro1, is reduced in the mutants. This suggests that disassembly of ESCRT-III is affected by the Snf7 mutations. We propose that the composition of the ESCRT-III complex may be important for Rim activation.


Yeast strains and plasmids

The yeast strains used are listed in Table 1. Yeast cells were grown in SDC (synthetic defined medium containing casamino acids; 0.67% yeast nitrogen base, 1% casamino acids and 2% glucose). In some experiments, 50 mM MOPS was added and the pH was adjusted to pH 3 or pH 7. Deletion strains and strains carrying Myc-tagged gene variants (13myc; Table 1) are derived from the wild-type strain JD52. They were constructed by one-step gene replacement with PCR-generated cassettes [25]. For N-terminal HA (haemagglutinin)-tagging of Rim101, the GAL1 promoter in pFA6a-His3MX6-PGAL1-3HA was replaced by a RIM101 promotor fragment. The deletions and insertions were verified by PCR. To construct pRK861, a 1843 bp PCR fragment, with attached BamHI and SalI sites, containing the SNF7 ORF (open reading frame) and including the SNF7 promoter, was cloned into YCplac33 [26].

View this table:
Table 1 Yeast strains

SNF7 mutagenesis

To generate SNF7 mutations an error-prone PCR was performed. The reaction mixture contained 2.5 mM MgCl2, 0.5 mM MnCl2, 0.5 mM dNTPs, 1 μM primers, 4 units of Taq polymerase and 2 ng of pRK861 template DNA, and 35 cycles were performed with 1 min annealing at 57.4°C, 4 min elongation at 68°C and 30 s denaturation at 94°C. The PCR fragments were co-transformed with YCplac33 linearized with BamHI and SalI into the Δsnf7 Δhis3 strain RKY1854, which carried a STE6HIS3 expression cassette integrated into the genome. In the yeast cells, an intact plasmid was regenerated by homologous recombination between the PCR-fragment and vector sequences. Ste6–His3 was expressed from the CUP1 promoter thus efficient expression required Cu2+ ions in the growth medium. His3-positive transformants were selected on SD (synthetic dropout) plates containining 0.5 mM CuSO4 and the auxotrophic requirements, but without histidine and uracil. Among 35000 transformants with intact plasmids, 128 transformants were obtained that were able to grow on plates lacking histidine. Among these transformants, 20 transformants were found that were wild-type with respect to temperature sensitivity and growth on raffinose.

Co-IP (co-immunoprecipitation) and cell fractionation

Cells were grown overnight to the mid-exponential phase (a D600<1.5, corresponding to approx. 5×107 cells/ml) in SDC. Approx. 108 cells were harvested, washed in cold 10 mM NaN3 and resupended in 100 μl of lysis buffer (50 mM Hepes, pH 7.5, containing 0.3 M sorbitol, 10 mM NaN3 and 1× protease inhibitors) and lysed by agitation with glass beads for 5 min. After addition of 650 μl of lysis buffer, samples were incubated on ice with 1% (v/v) Triton X-100 for 30 min. Then the cell extracts were centrifuged for 5 min at 500 g to remove cell debris. The supernatant was incubated for 1 h at 4°C with 5 μl undiluted anti-Myc antibody (Covance) and for another 1 h at 4°C with 50 μl of protein A–Sepharose beads [GE Heathcare; to form an approx. 20% (v/v) suspension]. The protein A–Sepharose beads were washed three times with lysis buffer in a table-top centrifuge (twice for 1 min at 150 g and once for 20 s at 13000 g), resuspended in 100 μl of SDS sample buffer and incubated for 15 min at 50°C before loading on to gels (7.5–12%). For cell fractionation, cell extracts were prepared in the same way, but the incubation with Triton X-100 was omitted. The cleared cell extract was centrifuged at 100000 g for 1 h at 4°C. The pellet (P100) was resuspended in the original volume of lysis buffer. For subsequent immunoprecipitation, 1% (v/v) Triton X-100 was added to the fractions (P100 and S100) and incubated on ice for 30 min. Immunoprecipitation was then further continued as described above.

Preparation of cell extracts for detection of Rim101 processing

Standard cell lysis triggers Rim101 processing (R. Kölling, unpublished work). Therefore, a fast-boil lysis protocol was used to assay Rim101 processing. Before harvesting of the cells, 1 mM PMSF was added to the culture. Cells were spun down (2 min at 13000 g), resuspended in 100 μl of lysis buffer and immediately heated to 95°C for 5 min. Then, cells were broken by glass bead lysis for 5 min.

Other methods

The pulse–chase, CPY (carboxypeptidase Y)-sorting and flotation experiments were performed as described previously [10,27,28]. Anti-Snf7 and anti-Ste6 antibodies were prepared by Seramum from antigens produced in-house.


Separation of snf7 phenotypes by mutagenesis

All of the Snf7 homologues appear to be involved in the formation of late endosomal MVBs, yet deletion mutants display distinct phenotypes [16]. These distinct phenotypes could reflect a particular function at endosomes, such as recruitment of specific cargo proteins for vacuolar degradation. Alternatively, they could point to additional functions for the Snf7 homologues completely unrelated to their endosomal function. If the observed phenotypes of the deletion mutants (sensitivity to high temperature, caffeine and Congo Red, and poor growth on raffinose) are unrelated to their endosomal function, it should be possible to separate these functions genetically from the endosomal function.

To test this prediction, we looked for SNF7 mutants that are blocked in the endocytic pathway but that are wild-type with respect to the other phenotypes. To be able to detect a block in the endocytic pathway, we made use of the ABC-transporter protein Ste6. We have shown previously that Ste6 is a short-lived protein that is transported via the endocytic pathway to the vacuole where it is degraded [28]. In Δsnf7 mutants, Ste6 is stabilized because its transport to the vacuole is blocked [10]. To facilitate detection of Ste6, it was fused to the marker protein His3, which is required for histidine synthesis. Under normal conditions, due to the high turnover of the Ste6–His3 fusion protein, the amount of Ste6–His3 is not high enough to promote growth of a his3 strain on medium lacking histidine. However, when Ste6–His3 is stabilized, as in a snf7 mutant, Ste6–His3 accumulates to higher levels and thus provides sufficient His3 activity to enable the his3 strain to grow.

SNF7 mutants were generated by error-prone PCR. Of the His3-positive transformants, 20 transformants were found that were wild-type with respect to temperature sensitivity and growth on raffinose (examples are shown in Figure 1A). This demonstrates that it is indeed possible to genetically separate the different Snf7 functions. Sequencing of 12 of these mutant SNF7 genes revealed that most of the snf7 genes carried multiple mutations (Table 2). Although many of the mutations are probably irrelevant to the phenotype the distribution of the mutations was quite uneven, with mutations clustering in the N-terminal half of Snf7 and at its very C-terminus (Figure 1B). Structural information is available for one ESCRT-III protein, human CHMP3, the homologue of yeast Vps24 [29]. However, it is likely that all ESCRT-III proteins have a similar structure (see our secondary structure prediction in [29a]), and hence these Snf7 mutations can be modelled on to the three-dimensional structure of CHMP3. The CHMP3 core structure consists of a four-helix bundle with two long helices (α1, α2) forming a 70 Å (1 Å=0.1 nm) hairpin structure and two additional short helices (α3, α4) across the α1–α2 hairpin. A fifth helix (α5) is connected to the core by a largely disordered linker and an additional sixth, C-terminal helix (α6) is not represented in the structure (Figure 1C). Hydrophobic amino acids show a heptad-repeat pattern in helices α2 and α6. It is therefore likely that these helices are engaged in coiled-coil interactions.

Figure 1 SNF7 functions can be separated genetically

(A) Growth phenotypes of SNF7 mutants. RKY1852 (left three panels, Δsnf7) or RKY1854 [right panel, Δsnf7 (CUP1p-STE6-HIS3)] were transformed with plasmids expressing: SNF7, wild-type Snf7; no Snf7 (vector control), snf7-M3; and snf7-M5. Samples of 10-fold serial dilutions of overnight cultures grown in SDC were spotted on to different plates and incubated for 3 days under different conditions: control, SDC at 25°C; 37°C, SDC at 37°C; raffinose, SDC with 2% raffinose, as the sole carbon source, and 2 μg/ml antimycin at 25°C; and −his, synthetic defined medium containing leucine, lysine, tryptophane, but not histidine, and 0.5 mM Cu2SO4 at 25°C. (B) Distribution of the Snf7 mutations obtained by PCR mutagenesis. The upper panel shows the positions of the mutations, indicated by arrows. Black boxes, predicted α-helical regions. The lower panel shows the coiled-coil forming probability determined with the program Macstripe 2.0a1 [36]. (C) Positions of M3 and M5 mutations modelled on to the CHMP3 structure.

View this table:
Table 2 SNF7 mutants

The Snf7 mutations generated in the present study cluster in the N-terminal half of Snf7, comprising the α1–α3 region, and in the C-terminal α6. Thus these regions appear to be important for the endosomal function of Snf7. As the region between α3 and α6 is largely devoid of mutations and given we selected for mutants that retain the gene expression function, we conclude that this region is crucial for the gene expression function of Snf7. Two Snf7 mutants with single mutations (snf7-M3, with a L26W substitution, and snf7-M5, with a S93P substitution; Figure 1C) were selected for further analysis.

To exclude the possibility that the higher His3 activity in the snf7 mutants was simply due to higher STE6HIS3 expression, the Ste6 half-life was determined by pulse–chase experiments in an Δsnf7 strain transformed with different SNF7 mutant plasmids (Figure 2A). Ste6 was rapidly turned over with wild-type SNF7, with a half-life of 21 min. The half-life increased by 5-fold, to 101 min, with the vector control (Δsnf7) as reported previously [10]. Half-life increases of 2- to 3-fold where observed with the mutant SNF7 variants. This demonstrates that the observed His3-positive phenotype of the SNF7 mutants is indeed due to stabilization of Ste6–His3.

Figure 2 Effect of SNF7 mutations on endocytic trafficking

(A) The turnover of Ste6 in a Δsnf7 strain transformed with different SNF7 plasmids was examined by pulse–chase analysis. Cells were labelled with [35S]-Trans label for 15 min and chased with an excess of cold methionine and cysteine for the time intervals indicated. Ste6 was precipitated from cell extracts with polyclonal antibodies directed against Ste6. Precipitated Ste6 was detected by autoradiography. The calculated Ste6 half-lives (τ) are indicated. (B) Δsnf7 cells were co-transformed with plasmids expressing Ste6–GFP and with plasmids expressing different Snf7 variants and the cells grown in SDC to the exponential phase. Left panels, GFP fluorescence; right panels, DIC (differential interference contrast) image of the yeast cells. (C) CPY-sorting was examined in the Δsnf7 strain transformed with different SNF7 plasmids. Cells were labelled for 10 min with [35S]-Trans label and chased with an excess of cold methionine and cysteine for 0 min or 40 min. CPY was immunoprecipitated from the spheroplast internal fraction (I) and external medium faction (E). The precipitated proteins were analysed by SDS/PAGE and autoradiography. The positions of the different CPY forms are indicated. SNF7, expressing wild-type SNF7; Δsnf7, vector control plasmid; snf7-M3, expressing snf7-M3; and snf7-M5, expressing snf7-M5.

To further demonstrate a Ste6-trafficking defect in these mutants, the intracellular distribution of a Ste6–GFP (green fluorescent protein) fusion protein was examined in the mutants (Figure 2B). Owing to high turnover of Ste6–GFP, only a faint staining of the vacuolar lumen is observed in a wild-type strain. The Δsnf7 mutant, in contrast, showed a typical ‘class E’ staining pattern [30], i.e. a brightly staining dot close to the vacuole with some faint staining of the vacuolar membrane. In line with a Ste6-trafficking defect, the mutants snf7-M3 and snf7-M5 showed a class E phenotype indistinguishable from Δsnf7.

To exclude the possibility that the SNF7 defects are restricted to Ste6 turnover, we also examined the trafficking of another protein, CPY, which passes through endosomes on its way to the vacuole. By passing from one compartment to the other, CPY receives different modifications, which can be distinguished by gel mobility. In the endoplasmic reticulum, it is core-glycosylated to the p1-form, in the Golgi it is converted into the slower migrating, outer-chain glycosylated p2-form and in the vacuole the mature m-form is finally generated by proteolytic cleavage of the precursor [30a]. In vacuolar protein sorting mutants, which are defective in endosomal transport, p2-CPY is missorted to the culture medium. To detect missorting, CPY was immunoprecipitated from cell extracts and from culture supernatants in a pulse–chase experiment (Figure 2C). Shortly after pulse labelling (0 min), all three CPY forms can be detected in the internal fraction. After a 40 min chase period in the wild-type strain, CPY is completely converted to the m-CPY form, which is exclusively found in the internal fraction. In contrast, a fraction of p2-CPY is detected in the culture medium of the Δsnf7 mutant. In this respect, the mutants snf7-M3 and snf7-M5 behaved like the Δsnf7 deletion mutant. This demonstrates that our SNF7 mutants have a general defect in endosomal trafficking.

As it is known that Snf7 is required for the activation of the transcription factor Rim101 (see above), we were specifically interested to determine whether our Snf7 mutants interfere with the activity of the Rim pathway. Therefore, several aspects of Rim pathway function were assayed in the mutants. Firstly, the induction of invertase after a shift from a high- to low-glucose medium was examined. We have recently shown that the invertase gene SUC2 is a target of the Rim pathway and that active Rim101 is required for high-level expression of invertase [16]. Expression of invertase from the SUC2 gene is subject to glucose repression [30b]. At 2h after the shift to low-glucose medium, the Δsnf7 mutant showed an invertase activity of approx. 30% of that of the wild-type, whereas the mutants snf7-M3 and snf7-M5 displayed wild-type activity (Figure 3A). This result is expected as the mutants were selected for wild-type growth on raffinose media (Figure 1A) and raffinose is an invertase substrate.

Figure 3 Test for rim phenotypes

(A) Invertase activity level. Cells were grown to exponential phase in SDC at 25°C with 5% glucose and transferred to medium with 0.1% glucose to induce invertase expression. Invertase activity was determined after 2 h in low-glucose medium as described previously [16]. The activity with wild-type SNF7 (WT) was set to 100%. The results are means± S.D. for three independent experiments. (B) LiCl sensitivity. Samples of 10-fold serial dilutions of cell cultures were spotted on to SDC plates with (+ LiCl, right panel) or without 0.3 M LiCl (left panel) and incubated for 3 d at 25°C. WT, wild-type strain; Δsnf7, Δsnf7 strain with vector control plasmid; Δrim13, Δrim13 strain with vector control plasmid; snf7-M3, Δsnf7 strain expressing snf7-M3; snf7-M5, Δsnf7 strain expressing snf7-M5. (C) The strains as used in (B) and (C) were co-transfected with a Rim101–3HA-expressing plasmid and were grown to exponential phase in SDC adjusted to pH 3. Half of the culture was adjusted to pH 7 and growth was continued for 30 min before extract preparation. Rim101–3HA was detected by Western blotting with anti-HA antibodies. no tag, wild-type strain with vector control plasmid. Even numbered lanes, pH 3; odd numbered lanes pH 7. (*), background band.

Another phenotype indicative of defects in the Rim pathway is sensitivity to LiCl [23]. To test for lithium sensitivity, serial dilutions of cultures were spotted on to plates containing 0.3 M LiCl. As can be seen in Figure 3(B), the Δsnf7 strain is as equally sensitive to LiCl as the Rim pathway mutant Δrim13. The mutants snf7-M3 and snf7-M5, however, are not affected by LiCl.

Upon a shift to alkaline pH Rim101 is activated by proteolytic cleavage [23], presumably mediated by the calpain-like protease Rim13, giving rise to a spectrum of bands, rather than to a single cleavage product. As expected, no cleavage was observed in the Δrim13 and Δsnf7 mutants upon a shift from pH 3 to pH 7 (Figure 3C). In contrast, cleavage indistinguishable from the wild-type was seen with the snf7-M3 and snf7-M5 mutants.

Taken together, our experiments clearly show that protein trafficking in the endocytic pathway is severely affected by our SNF7 mutations, whereas Rim pathway function is completely normal in the mutants. Thus our mutations selectively affect the endocytic function of Snf7 but leave the Rim pathway function intact.

Membrane association of Snf7

To further characterize the nature of the endosomal defect of the various SNF7 mutants, Snf7 membrane association was examined. It has been shown previously that Snf7 associates with endosomal membranes as part of the ESCRT-III complex [8]. As the membrane-associated fraction of Snf7 is higher in vps4 mutants [3], these experiments were performed in a Δvps4 Δsnf7 background. Membrane association was examined by flotation on Optiprep™ gradients. During centrifugation, membranes float to the top of the gradient, due to their low density, while proteins not attached to membranes remain behind in the lower, denser fractions of the gradient. Six fractions were collected from the gradients. As shown in Figure 4, the membrane proteins Pep12 [an endosomal SNARE (N-ethylmaleimide-sensitive factor-attachment protein receptor) protein] and Pma1 (the plasma membrane ATPase) were mostly found in fractions 1 and 2 of the gradient, whereas the soluble PFK (phosphofructokinase) protein was mostly found in fractions 4–6. About 60% of wild-type Snf7 protein were membrane-associated (found in fractions 1–3). For the mutants snf7-M3 and snf7-M5, a similar fraction of protein was observed at the membrane. This shows that membrane association is not disrupted by these mutations.

Figure 4 Membrane association of Snf7 mutants

Cell extracts of a Δsnf7 Δvps4 strain, transformed with different SNF7 plasmids, and grown in SDC were fractionated on Optiprep™ flotation gradients. Six fractions (lanes 1–6) were collected from the gradients (lanes 1–3 represent the float fraction, lanes 4–6 the non-float fraction). In the top three panels, fractions were analysed by Western blotting with anti-Snf7 antibodies. Snf7, expressing wild-type Snf7; Snf7-M3, expressing snf7-M3; Snf7-M5, expressing snf7-M3. The bottom three panels show a Δsnf7 strain expressing wild-type SNF7 subjected to the same experiment with antibodies directed against different marker proteins, as indicated.

Effect of mutations on Snf7 interactions

To understand the molecular basis of the selective trafficking defect of the Snf7 mutations, we analysed the effect of the mutations on Snf7 interactions. Previous studies presented evidence that Snf7 can exist in a filament formation [12,13]. To examine the effect of our mutations on such a possible multimer formation, cross-linking experiments were performed. Cell extracts were treated with different concentrations of the non-cleavable cross-linker DSS (disuccinimidyl suberate) and were then analysed by SDS/PAGE and Western blotting with anti-Snf7 antibodies. Upon addition of cross-linker, a regular evenly spaced pattern of anti-Snf7 reactive bands was observed (Figure 5). The apparent sizes of these protein species were roughly multiples of 38 kDa (38, 75, 110 and 160 kDa), which corresponds to the apparent size of monomeric Snf7 on SDS/PAGE gels. One cross-link band at 55 kDa did not fit this pattern and could correspond to a cross-link between Snf7 and another protein. In any case, as no difference in the cross-linking pattern was observed between wild-type and mutant Snf7 it is likely that multimer-formation is not affected by our mutations. In contrast with Snf7, only a single, monomeric band was observed for the ESCRT-III-associated protein Mos10/Vps60 (results not shown).

Figure 5 Detection of higher-order Snf7 complexes by cross-linking

Cell extracts were prepared from the Δsnf7 strain transformed with different SNF7 plasmids, expressing wild-type (WT) Snf7, snf7-M3 (M3) or snf7-M5 (M5) as indicated, were treated with increasing concentrations of DSS (0, 1, 2.5 and 5 mM). Snf7 reactive bands (marked by arrows with calculated sizes in kDa) were detected by Western blotting with anti-Snf7 antibody. M represents the protein standard.

We then examined the interactions of Snf7 with other binding partners by co-IP. The Myc-tagged binding partners were immunoprecipitated from cell extracts with anti-myc antibodies and the immunoprecipitates were then analysed by Western blotting with anti-Snf7 antibodies (Figure 6 and Table 3). The homomeric Snf7–Snf7 interaction was confirmed by the co-IP experiments and, as above, this interaction was unaffected by our Snf7 mutations. Likewise, no significant effect on co-IP efficiencies were observed for either Rim13 or Rim20. However, a clear effect was observed for the Vps20, Vps24, Vps4 and Bro1 co-IP efficiencies. The Vps20 and Vps24 co-IP signals were strongly enhanced in the mutants, whereas the Vps4 and Bro1 signals were significantly reduced. As Vps4 is required for disassembly of ESCRT-III and since interaction between the mutant Snf7 proteins and Vps4 is reduced, our results suggest that the Snf7 mutations interfere with the disassembly of ESCRT-III at the endosomal membrane, which would offer an explanation for the observed vacuolar protein sorting defect.

Figure 6 Effect of mutations on Snf7 interactions

The interaction of Snf7 with its interaction partners was examined by co-IP. Indicated Myc-tagged interaction partners, from the chromosomal copy of the corresponding genes, were co-transformed into the Δsnf7 strain background expressing wild-type (WT) Snf7, snf7-M3 (M3) or snf7-M5 (M5). Cells were grown in SDC adjusted to pH 3. The interaction partners were immunoprecipitated from cell extracts with anti-Myc antibodies. The immunoprecipitates were then analysed for the presensce Myc-tagged proteins with anti-Myc antibodies (anti-myc; lanes 1–3) and for Snf7 with anti-Snf7 antibodies (lanes 5–10) by Western blotting. I, input (lanes 5, 7 and 9); P, immunoprecipitates (lanes 6, 8 and 10).

View this table:
Table 3 Snf7 co-IP

The results are presented the means±S.D of the ratio of the Snf7 co-IP signal compared with the Snf7 input signal as. The wild-type ratio is set to 100.

If the above interpretation is correct, we would expect to observe an increased amount of mutant Snf7 protein at the membrane. To test this prediction, cell extracts were centrifuged at 100000 g to pellet the membranes, and the portion of Vps20 and Snf7 in the pellet fractions was determined (Figure 7 and Table 4). In wild-type cells, 33% of Vps20 and 14% of Snf7 was detected in the membrane fraction. In the snf7-M3 and snf7-M5 mutants, the fraction of Vps20 at the membrane was slightly increased compared with wild-type (by 30–40%). The fraction of mutant Snf7 at the membrane, however, was twice as high in the mutants as in wild-type cells. This is in line with the interpretation that there is more ESCRT-III at the endosomal membrane due to less efficient disassembly. We only observed a moderate increase in the membrane association of Vps20 in the mutants. The reason for this moderate effect could be that a large portion of Vps20 is already associated with the endosomal membrane, independently of ESCRT-III. In fact, the myristoylated Vps20 appears to have an intrinsic ability to associate with membranes and plays an important role in recruiting the other members of the ESCRT-III family [8].

Figure 7 Fractionation of Snf7 and Vps20 by sedimentation

Strain RKY2401 (Δsnf7 VPS2013myc) transformed with different plasmids was grown to exponential phase in SDC adjusted to pH 3. The strains expressed wild-type Snf7 (WT Snf7; lanes 1–7), snf7-M3 (lanes 8–12) or snf7-M5 (lanes 13–17). Cleared cell extracts (Total; lanes 3, 8 and 13) were centrifuged at 100000 g for 1 h and separated into P100 pellet (P100, lanes 4, 9 and 14) and S100 supernatant (S100 lanes 5, 10 and 15). Equal portions of the fractions were examined by Western blotting with anti-Myc antibodies to detect Vps20–13myc (Vps20; upper panel) and anti-Snf7 antibodies for detection of Snf7 (Snf7; lower panel). In addition, the interaction of Snf7 with Vps20–13myc was examined by co-IP. Vps20–13myc was immunoprecipitated from the P100 (IP-P100) and S100 fractions (IP-S100) with anti-Myc antibodies. The immunoprecipitates (lanes 1, 2, 6, 7, 11, 12, 16 and 17) were analysed for Vps20–13myc with anti-Myc antibodies (upper panel) and for Snf7 with anti-Snf7 antibodies (lower panel) by Western blotting. Lanes 1 and 2 show the control immunoprecipitation without anti-Myc antibodies.

View this table:
Table 4 Percentage of protein in the membrane fraction (P100) of cell extracts

In each case two experiments were performed with virtually identical results.

We have made the assumption that the observed interactions between Snf7 and its binding partners take place at the endosomal membrane. There is evidence that soluble, monomeric ESCRT-III proteins are in a closed conformation preventing homo- or hetero-typic ESCRT interactions [31]. Theoretically, our Snf7 mutations could alter the equilibrium between open and closed conformations. Under these conditions, Snf7 interactions may also occur in the soluble phase. To address this point, Vps20 was immunoprecipitated from the P100 (membrane) and S100 (soluble) fractions and assayed for Snf7 co-IP (Figure 7). For wild-type Snf7, as well as for the mutant Snf7 variants, co-immunoprecipitated signals could only be detected with Vps20 from the P100 fraction. This indicates that the interactions indeed occur at the endosomal membrane.

Effect of pH on Snf7 interactions

The Rim pathway is activated by alkaline pH [24]. We were therefore interested in determining whether the Snf7 interactions are affected by pH shift. Snf7 co-IP efficiencies were determined from cells grown at pH 3 and from cells shifted to pH 7 for 30 min. As above, Myc-tagged binding partners were immunoprecipitated from cell extracts with anti-Myc antibodies and the immunoprecipitates were analysed by Western blotting with anti-Snf7 antibodies. From these experiments (Figure 8), it was obvious that with most of the binding partners (Rim13, Rim20, Vps20 and Bro1), considerably more Snf7 could be co-precipitated at alkaline pH than at pH 3. The notable exception, however, was Vps4 where co-IP efficiencies were not altered by the pH shift. This suggests that the ESCRT-III network at the endosomal membrane is more extensive at alkaline pH, thereby offering more binding sites for Snf7 interactors.

Figure 8 Effect of pH on Snf7 interactions

The interaction of Snf7 with its interaction partners was examined by co-IP. Yeast strains expressing Myc-tagged interaction partners, as indictaed on the left, from the chromosomal copy of the corresponding genes were pre-grown in SDC adjusted to pH 3. Half of the culture was adjusted to pH 7, the other half was left untreated. Incubation was continued for 30 min. The interaction partners were immunoprecipitated from cell extracts with anti-Myc antibodies. The immunoprecipitates were then analysed for the 13 myc-tagged proteins with anti-Myc antibodies (anti-myc; left panels) and for Snf7 with Snf7 antibodies (anti-Snf7; right panels). pH 3, lanes 1–3; pH 7, lanes 4 and 5; I, input (lanes 1 and 4), P+, immunoprecipitate with antibodies (lanes 2 and 5); P−, mock immunoprecipitation, no antibodies (lane 3).


In the present paper, we explore the relationship between the function of Snf7 in the ESCRT pathway and in the activation of the transcription factor Rim101. The goal of the study was to determine whether these two types of Snf7 activity are different manifestations of a common function or whether they are separate from each other. In the latter case, it should be possible to obtain SNF7 mutants that are selectively disrupted in one of the functions and we succeeded in isolating SNF7 mutants that are defective in endocytic trafficking but that are wild-type for Rim101 activity. This indicates that activation of Rim101 is not a side effect of ongoing endocytic trafficking, but that it is a distinct activity of Snf7. Whether our mutations cleanly separate the two functions of Snf7 is less clear. The trafficking defect of our point mutants is less severe than in a SNF7 deletion. Thus, our Snf7 mutants appear to retain some residual activity. Although this small residual activity is not sufficient for efficient endocytic trafficking, it may still be adequate for activation of Rim101. However, our results do confirm that the requirements for endocytic trafficking and for Rim101 activation are different.

The Snf7 activities cannot be completely independent from each other, as the ESCRT and Rim pathways share a requirement for ESCRT functions acting upstream of Snf7 [1720]. However, if the ESCRT and Rim pathways are connected, then how can the selective defect of the Snf7 mutants on endocytic trafficking be explained? To gain mechanistic insight into the co-ordination of ESCRT and Rim pathways, a detailed phenotypic analysis of our Snf7 mutants was performed. To fulfil its function in MVB formation, Snf7 has to be recruited to the endosomal membrane. We found that membrane association was not affected in our mutants, thus all upstream events leading to membrane recruitment of Snf7 appear to be normal in the mutants. This would suggest that Snf7 membrane association is a prerequisite for both activities of the protein.

To determine mechanistically what distinguishes the Snf7 mutants proteins from the wild-type protein, we further examined the interactions with known Snf7 binding partners. We found that the interaction with Rim13 and Rim20 was not affected by the mutations, and likewise there was no effect on the Snf7 homotypic interaction. However, we observed a clear effect on the interactions with Vps20, Vps24, Vps4 and Bro1. The Snf7 co-IP signals with Vps20 and Vps24 were strongly increased in the mutants, whereas the signals with Vps4 and Bro1 were significantly reduced. As Vps4 is required for disassembly of ESCRT complexes at the endosomal membrane and since we see less binding of the mutant proteins to Vps4, we propose that the mutants are defective in disassembly of ESCRT-III complexes at the endosomal membrane. This appears to be a perfect solution to the demands of our genetic screen; through the block in ESCRT-III disassembly, endocytic trafficking is abrogated, whereas the ESCRT network at the endosomal membrane is preserved for activation of the Rim pathway.

The exact composition of ESCRT-III in our SNF7 mutants is unclear. It has been previously reported that the ESCRT-III core subunits do not appear to occur in an equimolar ratio in the complex; the stoichiometry of the complex in yeast has been estimated to be 1:10:5:3 (Vps20/Snf7/Vps24/Vps2) [32]. Thus Snf7 is the most important and most abundant subunit of the complex. It appears that the other ESCRT-III subunits serve a more regulatory role. For instance, binding of the Vps2–Vps24 subcomplex limits the size of the complex and promotes disassembly of ESCRT-III by recruiting Vps4, whereas Vps20 promotes Snf7 oligomerization [32]. In the present study, we found that the co-IP efficiencies between Snf7 and the ESCRT-III subunits Vps20 and Vps24 are much higher with the mutant Snf7 variants than with wild-type Snf7. This finding could be explained in two ways: either the stoichiometry of the ESCRT-III complex is altered in the SNF7 mutants or the ESCRT-III complexes are more stable leading to a higher steady-state concentration of complexes at the membrane. In line with both interpretations, we detected more mutant Snf7 protein in the membrane fraction.

Our mutations Snf7-M3 (L26W) and Snf7-M5 (S93P) are localized to helices α1 and α2. These regions have been implicated in homo- and hetero-typic ESCRT-III interactions [29]. The higher co-IP efficiencies of mutant Snf7 with Vps20 and Vps24 could thus be explained in part by tighter binding of these ESCRT-III subunits to helices α1 and α2. However, Vps4 and Bro1 are not likely to bind to the α1-α2 region. Binding sites for Vps4 or Alix, the mammalian Bro1 homologue, have been mapped to C-terminal regions of ESCRT-III proteins [3335]. However, the α1-α2 mutants do show reduced interaction with Vps4 and Bro1. One explanation for these findings is that conformational changes in the α1-α2 region are conveyed to the C-terminal region of Snf7 and alter the binding affinities for Vps4 and Bro1. Binding of ESCRT-III subunits to the α1-α2 region could thus modulate Vps4 and Bro1 binding to Snf7. This could be part of the normal functional cycle of ESCRT-III assembly and disassembly. An alternative view would be that the α1-α2 mutants used in the present study are ‘locked’ in a functional state that precludes Vps4 and Bro1 binding. For instance, owing to the special structure of the ESCRT-III complex, the binding sites for Vps4 and Bro1 may not be accessible.

The results in the present paper suggest that the Snf7 interactions are very sensitive to conformational changes in Snf7. The composition and stability of ESCRT-III could thus be a prime target for metabolic control. It appears that extensive ESCRT network formation at the endosomal membrane is also a prerequiste for Rim pathway activation. In line with this notion, we observed a stronger binding signal with virtually all interaction partners, both those that are Rim pathway specific and those that are ESCRT pathway specific, upon shift to alkaline pH, with the notable exception of Vps4, whose binding signal was not affected by pH. This model is further supported by the findings of Hayashi et al. [17] who showed that the Rim pathway is constitutively active when disassembly of the ESCRT complexes is blocked by mutating DID2, VPS24 or VPS4. A change in the stability or composition of ESCRT-III may thus be crucial for Rim pathway activation.


Stefanie Huppert performed the SNF7 mutant screen and the initial characterization of the mutants. Peter Weiss examined the rim phenotypes and the effects on the various binding partners. Ralf Kölling supervised the project and wrote the manuscript.


This work was supported by the Deutsche Forschungsgemeinschaft [grant number Ko 963/5-2].


We are grateful to Karin Krapka (Institut für Mikrobiologie, Universität Düsseldorf, Germany) for her assistance with some of the experiments.

Abbreviations: CHMP, chromatin modifying protein; co-IP, co-immunoprecipitation; CPY, carboxypeptidase Y; DSS, disuccinimidyl suberate; ESCRT, endosomal sorting complex required for transport; GFP, green fluorescent protein; HA, haemagglutinin; MVB, multivesicular body; SDC, synthetic defined medium containing casamino acids


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 29a.
  31. 30.
  32. 30a.
  33. 30b.
  34. 31.
  35. 32.
  36. 33.
  37. 34.
  38. 35.
  39. 36.
View Abstract