New Regulators of Clathrin-Mediated Endocytosis Identified in Saccharomyces cerevisiae by Systematic Quantitative Fluorescence Microscopy

A GFP-based screening of S. cerevisiae for novel endocytic proteins

We took advantage of the S. cerevisiae GFP library, containing the coding sequence of GFP immediately preceding the stop codon of each ORF in the yeast genome (Huh et al. 2003). Library proteins are therefore expressed from their endogenous promoter, with a GFP tag at the carboxy-terminal end. Creators of the library performed an initial classification of the subcellular localization for 4156 GFP-tagged proteins representing ∼75% of the proteome. We noted that well-established components of the endocytic machinery were found in three localization groups: the plasma membrane, actin, and punctate (Huh et al. 2003). Together the three groups contain 319 GFP-tagged proteins and include numerous proteins with unknown function. We reasoned that some of these unknown proteins may specifically localize to sites of endocytosis and constitute new components of the endocytic machinery. To test that possibility, MATa strains expressing the 319 GFP-tagged proteins in these categories were mated with MATα cells expressing Sla1–RFP from the endogenous locus and resulting diploid cells were selected with appropriate markers. Sla1 is a multifunctional clathrin adaptor and actin polymerization regulator present at all sites of CME and easily visible by fluorescence microscopy (Figure 1A) (Ayscough et al. 1999; Kaksonen et al. 2003; Kaksonen et al. 2005; Di Pietro et al. 2010; Feliciano and Di Pietro 2012; Feliciano et al. 2015). Each diploid strain expressing both the corresponding GFP-fusion protein and Sla1–RFP was subjected to live cell confocal fluorescence microscopy and colocalization analysis. To ensure an unbiased study, the operator did not know the identity of the strain subjected to imaging and random images were used to determine colocalization levels.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

GFP-based screening for endocytic proteins. (A) S. cerevisiae cells expressing Sla1–RFP and the indicated GFP-fusion proteins were analyzed by live-cell confocal fluorescence microscopy. Bar, 1 μm. (B) Pearson correlation coefficients (mean ± SD) between Sla1–RFP and each of the proteins demonstrated in A. (C) Distribution of PCC values for Sla1–RFP with the 319 GFP-tagged proteins tested in the screening. Endocytic proteins are indicated with red symbols and uncharacterized proteins with yellow symbols. The cutoff for colocalization (PCC = 0.2) is marked with a line. (D) Left, schematic of how the yeast genome was narrowed down to the 319 proteins selected for the screen. Right, functions of the 197 proteins showing colocalization with Sla1 at a level higher than PCC = 0.2.

Visual and quantitative data reveal candidate endocytic proteins

The Pearson correlation coefficient (PCC) was determined for each GFP-tagged protein by averaging at least three images, each containing multiple cells with several endocytic patches. The library proteins were then ranked from highest to lowest PCC with a range of 0.86 to −0.02 (Supporting Information, Table S1). Representative examples are shown in Figure 1A. The highest scoring protein, Pan1 (0.86 ± 0.01, mean ± SD), is known to have the same dynamics as Sla1 and therefore expected to display a high colocalization level (Kaksonen et al. 2003, 2005; Boettner et al. 2012; Weinberg and Drubin 2012). The capping protein β subunit, Cap2 (0.58 ± 0.11, mean ± SD), is a component of the actin network that assembles after the arrival of Sla1 (Kaksonen et al. 2005) and shows intermediate level of colocalization. The lysine permease Lyp1 (0.27 ± 0.05, mean ± SD) localizes to the plasma membrane in a relatively uniform manner and thus represents a low colocalization level. A nuclear pore protein classified as punctate localization, Kap95 (0.09 ± 0.04, mean ± SD) serves as an example of background PCC obtained with a noncolocalizing protein (Figure 1, A and B). Based on these observations, proteins with PCC >0.2 were considered to show colocalization above background, totaling 197 of the 319 strains (Figure 1C).

The functions of the 197 colocalizing proteins were obtained from the Saccharomyces Genome Database and the literature, showing representation from endocytic, cytoskeletal, trafficking, as well as other functions (Figure 1D, Table S1). Of the 45 known endocytic proteins included in the screening, only two (Sac6 and Arp2) fell below the 0.2 PCC cutoff for colocalization, indicating that the group of 197 proteins includes the vast majority of machinery components analyzed in the group of 319 strains (Table S1 and Table S2). Furthermore, 31 of the top 50 PCCs corresponded to well-studied endocytic proteins, such as Sla2, Ent2, and Bbc1. Thus, a majority of the known endocytic machinery proteins were located high in this ranking (Table S1 and Table S2). Interestingly, all four subunits of the AP-2 complex scored a colocalizing PCC, consistent with recent findings that AP-2 does in fact play a role in yeast CME (Carroll et al. 2009; Chapa-Y-Lazo et al. 2014). The lower range of the 197 colocalizing proteins was enriched in proteins spanning a variety of functions and containing predicted or known transmembrane domains. Such proteins correspond to known or likely CME cargo and typically showed a relatively even distribution throughout the plasma membrane similar to Lyp1 (Figure 1A and Table S1). To identify probable new machinery components, we next focused on proteins with unknown function and PCC similar to known endocytic proteins.

Importantly, 28 uncharacterized proteins colocalize with Sla1 and many of them are not predicted to have a transmembrane domain and thus may not be endocytic cargo (Figure 1, C and D and Table 1). Most notably, nine of such uncharacterized proteins were in the top 50 PCC scores, representing highly likely new components of the endocytic machinery. The highest scoring uncharacterized protein was Ubx3 with a PCC of 0.52 ± 0.20 (Figure 1C, Table S1, and Table 1). This protein was identified as having a punctate composite fluorescent pattern in the library, which also showed a higher cytosolic background compared to our images, probably due to their use of wide-field fluorescence microscopy. Ubx3 is defined by a ubiquitin-like UBX (ubiquitin regulatory X) domain in its C terminus (Figure 2A) (Dreveny et al. 2004; Schuberth et al. 2004; Schuberth and Buchberger 2008) and was subjected to further study to confirm its endocytic role.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Ubx3 is a novel component of the endocytic machinery. (A) Cartoon representation of Ubx3 domains predicted by the Phyre2 protein-folding recognition engine. UAS, domain of unknown function. (B) Ubx3–GFP shows strong colocalization with Sla1–RFP by live-cell confocal fluorescence microscopy. Solid arrows show an example of an endocytic patch demonstrating strong colocalization; open arrows show an example of Ubx3–GFP present at an endocytic patch without Sla1–RFP. Bar, 1 μm. (C) Dynamics of Ubx3–GFP and Sla1–RFP at endocytic sites were compared. Left, one frame of a movie indicating with a white box the endocytic site used for constructing a kymograph. Right, kymograph demonstrating dynamics of Ubx3–GFP and Sla1–RFP, and average patch lifetimes (22 patches from seven cells, mean ± SEM). Each frame represents 2 sec. (D) Patch lifetimes of Sla1–RFP and Ubx3–GFP were measured at room temperature and 37° in both wild-type (CHC1) and temperature-sensitive clathrin heavy-chain (chc1-ts) cells (15 patches from 5 cells per strain and temperature, mean ± SEM; **, P < 0.0001; *, P < 0.001). (E) GST–Ubx3 directly binds polyhistidine-tagged clathrin heavy-chain N-terminal domain as determined by GST-pulldown with purified proteins. Top: eluted proteins were analyzed by immunoblotting using an antibody to the polyhistidine tag (anti-His). Bottom: loading control Coomassie-stained gel of GST and GST–Ubx3 proteins bound to glutathione beads. (F) Alignment of S. cerevisiae Ubx3 amino acid sequence with that of other fungal species demonstrates high conservation of residues matching the W-box core clathrin-binding motif. Strictly conserved residues are shown in red. The fragment fused to GST for GST-pulldown assays in Figure 2G is indicated at the top (WXXW). (G) Ubx3 contains a W-box that binds the clathrin heavy-chain N-terminal domain. GST alone, GST–WXXW, and corresponding mutant GST–AXXA were bound to glutathione-Sepharose beads and incubated with polyhistidine-tagged N-terminal domain of clathrin heavy chain. The eluted proteins were analyzed by immunoblotting using an antibody to the polyhistidine tag (anti-His). Bottom: loading control Coomassie-stained gel of GST-fusion proteins bound to glutathione beads. (H) Ubx3–GFP is the only UBX domain-containing protein that localizes to endocytic patches. Yeast strains expressing each of the seven Ubx proteins tagged with GFP were analyzed by confocal fluorescence microscopy. Bar, 1 μm.

Ubx3 is the first UBX domain-containing protein in the endocytic machinery

By confocal fluorescence microscopy, Ubx3–GFP displayed similar patch dynamics to Sla1–RFP and other endocytic coat proteins arriving at intermediate stages. Ubx3–GFP had a patch lifetime of 45 ± 3 sec, appearing slightly before Sla1–RFP at the endocytic site and remaining slightly after the disappearance of Sla1–RFP (Figure 2, B and C and File S1). Ubx3–GFP is recruited after early coat proteins such as Ede1–RFP (Figure S1). Ubx3–GFP patches also showed movement away from the surface toward the center of the cell right before disappearing, a behavior typical of endocytic coat proteins (Figure 2C, last eight frames, and File S1). To assess whether Ubx3–GFP patch localization is affected by the endocytic coat, Ubx3–GFP dynamics was visualized in cells carrying a temperature-sensitive allele of the clathrin heavy-chain gene (chc1-ts) (Bensen et al. 2000). Upon incubation at 37°, the patch lifetime of Ubx3–GFP was significantly reduced (Figure 2D). In contrast, the patch lifetime of Ubx3–GFP in wild-type cells (CHC1) was not affected by incubation at 37° (Figure 2D). For comparison, the clathrin-binding adaptor protein Sla1–GFP (Di Pietro et al. 2010) was analyzed in parallel in cells carrying the chc1-ts allele and demonstrated a similar reduction in patch lifetime upon incubation at 37°, whereas wild-type cells did not (Figure 2D). This result indicates that Ubx3 associates with the endocytic coat in vivo. To investigate whether Ubx3 binds clathrin, we tested in vitro for physical interaction by GST pulldown. GST and GST–Ubx3 were immobilized on glutathione-Sepharose beads and incubated with polyhistidine-tagged clathrin heavy-chain N-terminal β-propeller domain (His–CHC N-term) (Kirchhausen et al. 2014). Immunoblotting analysis showed binding of His–CHC N-term to GST–Ubx3 but not to GST indicating a direct physical interaction (Figure 2E). Inspection of the Ubx3 amino acid sequence for clathrin-binding motifs revealed no obvious clathrin-box (LLDLD related sequence) (Dell’Angelica et al. 1998). However, a sequence conforming to the core W-box motif (WXXW), where X represents any amino acid (Ramjaun and Mcpherson 1998; Miele et al. 2004), was located upstream the UBX domain (Figure 2A). Sequence alignment revealed that these residues are extremely well conserved among the Ubx3 family, suggesting functional importance (Figure 2F). Consistent with a functional W-box capable of engaging the clathrin β-propeller domain, this stretch of the Ubx3 sequence is predicted to be disordered. The ability of this sequence to bind clathrin was determined by GST-pulldown. A GST-fusion protein containing the Ubx3 candidate W-box and flanking sequences (residues 337–350, GST–WXXW) bound His–CHC N-term significantly above background levels (GST) (Figure 2G). Mutation of the tryptophan residues to alanine (GST–AXXA) diminished His–CHC N-term binding to background levels, showing that binding was specific (Figure 2G). To our knowledge, this is the first example of a W-box type clathrin-binding motif in a nonmammalian protein. Together these results demonstrate Ubx3 binds clathrin and that it is a component of the endocytic coat. As there are seven UBX domain-containing proteins in yeast, we visualized each protein tagged with GFP to determine if any others localized to endocytic sites, but only Ubx3 displayed an endocytic punctate fluorescent pattern (Figure 2H).

To further test for a role of Ubx3 in endocytosis we performed two uptake assays. First we used Lucifer yellow, a fluid-phase fluorescent dye, to track bulk intake into cells. Wild-type cells and cells carrying a deletion of the UBX3 gene (ubx3∆) were incubated with media containing Lucifer yellow and the internalized dye was quantified by fluorescence microscopy. An internalization defect was observed for ubx3∆ cells compared to wild-type cells (Figure 3A). The extent of the defect was comparable to the one we observed for Las17-MP8-12, a mutant displaying altered Las17 recruitment to endocytic sites and misregulation of actin polymerization (Feliciano and Di Pietro 2012). We also developed a strain harboring a deletion of the UBX domain in the endogenous UBX3 gene (ubx3^ΔUBXd) (Figure 2A). This strain displayed a similar defect in Lucifer yellow internalization, suggesting that Ubx3 depends on its UBX domain for endocytic function (Figure 3A). GFP tagging of the ubx3^ΔUBXd allele and live cell imaging showed that the mutant Ubx3 protein is not degraded but localizes to fast-moving internal structures rather than endocytic sites (File S2). Thus, the UBX domain is required for normal Ubx3 localization and function at endocytic sites. We then used Mup1–GFP to track cargo endocytosis. Mup1 is a methionine transporter that is expressed at the plasma membrane when cells are starved for methionine, but quickly internalized via ubiquitin-dependent CME when cells are returned to methionine-rich media. Internalization of this cargo again portrayed a delay in ubx3∆ cells compared to wild-type cells (Figure 3B). Together, these data demonstrate that Ubx3 is a new component of the clathrin-mediated endocytic machinery needed for optimal endocytosis (see File S3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Ubx3 is needed for optimal endocytosis. (A) Wild-type cells (UBX3) and cells carrying a deletion of the UBX3 gene (ubx3∆) and a deletion of the UBX domain in the UBX3 gene (ubx3^ΔUBXd) were incubated with Lucifer yellow and imaged by confocal fluorescence microscopy. The fluorescence intensity inside the cell was measured, normalized by the intensity of the background, and expressed as the average relative fluorescence intensity (40 cells per strain, mean ± SEM; **, P < 0.001; *, P < 0.01). The experiment was performed two times with similar results. Bottom: representative images of the cells. Bar, 10 μm. (B) Endocytosis of the methionine transporter Mup1 tagged with GFP was analyzed in UBX3 and ubx3∆ cells as described in Materials and Methods. Fluorescence in the plasma membrane was measured using a mask drawn on the cell periphery and normalized to the background (15 cells per strain and time point, mean ± SEM; **, P < 0.001; *, P < 0.01). The experiment was performed three times with similar results. Bottom: representative images of cells at 0 and 20 min after change to media containing methionine. Bar, 10 μm.

Ubx3 regulates dynamics of the early arriving protein Ede1 but not later arriving coat proteins or actin assembly

In an effort to understand the mechanistic basis of Ubx3 function in CME, we examined the patch lifetimes of known endocytic proteins tagged with GFP in wild-type and ubx3∆ cells. We analyzed early (Ede1, Syp1), intermediate (End3, Ent2, Las17), and late (Myo5, Abp1, Rvs167) arriving components of the endocytic machinery (Kaksonen et al. 2005; Toshima et al. 2006; Boettner et al. 2009, 2012; Reider et al. 2009; Stimpson et al. 2009; Weinberg and Drubin 2012; Brach et al. 2014). Remarkably, Ede1–GFP had a significantly longer patch lifetime in ubx3∆ cells compared to wild-type cells while other endocytic proteins were unaffected (Figure 4, A–C). Ede1–GFP displayed a 34% increase in lifetime from 91 ± 4 sec to 123 ± 8 sec (mean ± SEM, P < 0.01) (Figure 4, A–C). Analysis of ubx3^ΔUBXd cells demonstrated a similar increase in Ede1–GFP lifetime (129 ± 11 sec, mean ± SEM, P < 0.01), again establishing that the UBX domain is required for Ubx3 function (Figure 4, B and C). The extension of the Ede1 patch lifetime occurs at the beginning of the endocytic process, as the time between arrival of Ede1–RFP and Sla1–GFP increases in ubx3∆ cells compared to wild-type cells (Figure 4, D–F). A similar delay was observed between the arrival of Ede1–RFP and Syp1–GFP in ubx3∆ cells compared to wild-type cells (Figure 4, D–F). The cause of this delay appeared to be a slower recruitment of Ede1–GFP to the endocytic site as we noted that in ubx3∆ cells and ubx3^ΔUBXd cells Ede1–GFP fluorescence intensity increased more slowly before reaching the peak (Figure 4C). Indeed, quantification of the Ede1–GFP relative recruitment rate demonstrated a significant defect in ubx3∆ cells and ubx3^ΔUBXd cells compared to wild-type cells (Figure 4, G and H). Slower Ede1 recruitment and prolonged endocytic patch lifetime in ubx3∆ cells and ubx3^ΔUBXd cells further link Ubx3 to the CME machinery and is consistent with the Lucifer yellow and Mup1–GFP endocytosis defects shown above.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Ede1 dynamics at sites of endocytosis depend on Ubx3. (A) Ede1–GFP has a large increase in endocytic patch lifetime in ubx3∆ cells, while other endocytic machinery proteins tested are unaffected. Strains expressing the indicated GFP-tagged proteins either in UBX3 or ubx3∆ background were analyzed by live-cell fluorescence microscopy (10–58 patches from four or more cells per strain, mean ± SEM; *, P < 0.01). (B) Frames from movies of cells expressing Ede1–GFP in UBX3, ubx3∆, or ubx3^ΔUBXd background. Boxed patches are tracked in the kymographs shown in C. Ede1–GFP patch lifetime (mean ± SEM) and number of patches analyzed for each strain are shown on the right (at least 10 cells per strain were analyzed). (C) Kymographs demonstrating the increase in Ede1–GFP lifetime in ubx3∆ cells and ubx3^ΔUBXd cells compared to UBX3 cells. Each frame represents 5 sec. D Kymographs displaying the patch lifetime of Ede1–RFP is extended at the beginning of the endocytic process, before Syp1–GFP or Sla1–GFP arrive. Each frame represents 5 sec. (E) Frames from the movies used to construct the kymographs shown in D, with the corresponding endocytic patches indicated by white boxes. (F) Cartoon representation of the Sla1, Syp1, and Ede1 relative timing of arrival to endocytic sites in UBX3 and ubx3∆ cells. (G and H) The recruitment rate of Ede1–GFP, measured from first appearance to peak patch intensity, is slower in ubx3∆ and ubx3^ΔUBXd cells compared to UBX3 cells (25 patches from 8 or more cells per strain, mean ± SEM; **, P < 0.0001). The peak fluorescence intensity between strains was unchanged.

Given the effect of UBX3 gene deletion on the dynamics and patch lifetime of Ede1–GFP, we examined the converse relationship. The Ubx3–GFP patch lifetime was determined in wild-type and ede1∆ cells. Ubx3–GFP displayed a significant decrease in patch lifetime from 43 ± 2 sec to 31 ± 3 sec (mean ± SEM, P < 0.01) suggesting a functional connection between Ubx3 and Ede1 (Figure S2).

Interestingly, UBX domain-containing proteins constitute a major family of cofactors for the ubiquitin-editing complex Cdc48 that determine its location and targets (Schuberth et al. 2004; Schuberth and Buchberger 2008; Stolz et al. 2011; Meyer et al. 2012; Buchberger 2013). A previous study reported that all seven S. cerevisiae UBX domain-containing proteins bind Cdc48 by yeast-two-hybrid analysis (Schuberth et al. 2004). We confirmed a physical interaction between Ubx3 and Cdc48 using a GST-pulldown assay (Figure 5A). To test for a function of Cdc48 in endocytosis we used the Lucifer yellow uptake assay with wild-type cells (CDC48) and cells carrying a temperature-sensitive allele (cdc48-3) (Ye et al. 2001). A defect in Lucifer yellow uptake was observed at 37°, indicating a function for Cdc48 in endocytosis (Figure 5B). Therefore Ubx3 function in endocytosis may be linked to the ubiquitin-editing complex Cdc48.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Ubx3 interacts physically with Cdc48 and endocytosis depends on Cdc48. (A) Ubx3 binds Cdc48. GST–Ubx3 and GST alone were bound to glutathione-Sepharose beads and incubated with total cell extract prepared with a strain expressing Cdc48–GFP from the endogenous locus. The eluted proteins were analyzed by immunoblotting using an antibody to the GFP tag (anti-GFP). Bottom: loading control Coomassie-stained gel of GST and GST–Ubx3 proteins bound to glutathione beads in the assay. (B) Lucifer yellow uptake was measured at room temperature and 37° in both wild-type cells (CDC48) and cells carrying a temperature-sensitive allele of the CDC48 gene (cdc48-3). The fluorescence intensity inside the cell was measured, normalized by the intensity of the background, and expressed as the average relative fluorescence intensity (30 cells per strain, mean ± SEM; **, P < 0.0001). Bottom: representative images of the cells.