The Na⁺ pump Ena1 is a yeast epsin-specific cargo requiring its ubiquitylation and phosphorylation sites for internalization

Endocytosis is a vital cellular process required for multiple cellular functions including nutrient uptake and control of cell membrane composition. This process involves the selective retention of cargo proteins into nascent endocytic sites by the endocytic machinery through the recognition of sequence-encoded motifs or post-translationally added tags.

In addition to its classical role in proteosome-mediated protein degradation, ubiquitin (Ub) plays an important role as a post-translational tag for protein sorting into different subcellular compartments. Indeed, the Ub tag mediates membrane protein internalization and targeting to the lysosomal/vacuolar and recycling compartments (reviewed in Mukhopadhyay and Riezman, 2007; Traub and Lukacs, 2007).

Since Ub constitutes the main trafficking signal in Saccharomyces cerevisiae (reviewed in Lauwers et al., 2010), studies in yeast have been instrumental in advancing our understanding of the mechanisms operating in Ub-mediated sorting. For example, it has been clearly established that budding yeast utilizes the E3 ubiquitin ligase Rsp5 to promote cargo ubiquitylation relevant to endocytosis and other vesicle trafficking events (Belgareh-Touze et al., 2008; Lauwers et al., 2010). Indeed, Rsp5-mediated ubiquitylation is required for the internalization of cargoes such as the pheromone receptor Ste2 (Dunn and Hicke, 2001; Hicke and Riezman, 1996), the uracil transporter Fur4 (Galan et al., 1996), the general amino acid permease Gap1 (Soetens et al., 2001) and the zinc transporter Zrt1 (Gitan and Eide, 2000) among others. Although Rsp5 is capable of directly interacting with PPxY motif-containing substrates via its WW domains (McNatt et al., 2007; Oestreich et al., 2007; Stawiecka-Mirota et al., 2007; Sullivan et al., 2007a), this consensus sequence is not present in all cargoes that undergo Ub-dependent trafficking (Belgareh-Touze et al., 2008; Polo and Di Fiore, 2008). In those cases, the PPxY motif-containing, arrestin-related trafficking (ART) adaptor family specifically bridges cargo proteins to Rsp5 (Alvaro et al., 2014; Becuwe et al., 2012; Lin et al., 2008; Nikko and Pelham, 2009; Nikko et al., 2008; Prosser et al., 2015).

Following ubiquitylation, plasma membrane cargoes are recognized by endocytic adaptors, such as epsin (Ent1 and Ent2 in yeast) and Eps15 (or Ede1 in S. cerevisiae), which contain ubiquitin-binding elements like the Ub-interacting motifs (UIMs) or Ub-associated (UBA) domains. As a consequence of these interactions, cargo is incorporated into nascent endocytic sites and trafficked towards degradative vacuolar/lysosomal compartments or recycling routes.

Specifically, Eps15 is known to be responsible for the internalization of the AMPA receptor (Lin and Man, 2014) and Cx43 (Girão et al., 2009) among other ubiquitylated cargoes; while epsin is involved in the uptake of vascular endothelial growth factor receptor 2 (VEGFR2) (Pasula et al., 2012), notch ligands (Overstreet et al., 2004; Tian et al., 2004; Wang and Struhl, 2004, 2005) and the epithelial sodium channel ENaC (Staruschenko et al., 2005; Wang et al., 2006).

The different relevance of these adaptors for the internalization of various ubiquitylated cargoes highlights a crucial knowledge gap in the field: how do identical ubiquitin units attached to different cargoes mediate specific, but differential recognition by Ub-binding proteins? (which in turn is crucial to promote the needed trafficking outcomes).

Here, we report the identification of the Na⁺ pump Ena1 (Exitus Natru 1) as the first described yeast epsin-specific cargo. Further, we established that the internalization of this cargo required K¹⁰⁹⁰ and integrity of UIMs in epsin.

Importantly, our results suggest that epsin-mediated internalization of Ena1 requires two independent post-translational events – phosphorylation of an upstream Ser/Thr-rich motif and ubiquitylation in the cytoplasmic domain of the transporter. Epsin co-immunoprecipitated with an Ena1 mutant that emulated constitutive phosphorylation and ubiquitylation.

In summary, our results indicate that the yeast endocytic adaptor epsin likely acts upon its cargo based on the presence of a phosphorylated Ser/Thr motif upstream of a ubiquitylated lysine residue. We speculate that other Ub-binding elements of the sorting system (such as Ede1) also use coincidence detection of determinants (likely different from the ones recognized by epsins) to identify their targets among multiple ubiquitylated cargoes.

S. cerevisiae is known to use Ub as its main sorting tag and has been instrumental to study the function of endocytic adaptors (Lauwers et al., 2010). Therefore, we relied on this organism to investigate the mechanism by which epsin recognizes its specific targets among multiple ubiquitinated-cargoes. However, it was first necessary to identify a yeast membrane protein that would substantially rely on this adaptor for its internalization.

In contrast to mammals, yeast does not have homologs for established epsin-specific cargoes such as Notch-ligands and VEGFR2-like receptor tyrosine-kinases. Although ENaC-sodium channel homologs are not present either, S. cerevisiae has sodium pumps, such as those of the Ena protein family. Pumps and channels are evolutionarily/mechanistically related (Artigas and Gadsby, 2002; Gadsby, 2009) and their plasma membrane levels have been shown to be controlled by Ub-mediated endocytosis (Helenius et al., 2010; Lecuona et al., 2007). Therefore, we speculated that members of the Ena family (Ena1–Ena5) were suitable candidates to be internalized in an epsin-dependent manner.

Given its relevance among family members (Garciadeblas et al., 1993; Haro et al., 1991; Wieland et al., 1995), we focused our studies on the Ena1 paralog (also known as Pmr2) using an ENA1::GFP construct under the control of the MET25 methionine-repressible promoter (Logg et al., 2008; Wadskog et al., 2006). The stability of wild-type (WT) and mutated versions of Ena1–GFP was verified by western blotting, while also monitoring their cellular distribution pattern in comparison to free GFP and ER-retained transmembrane proteins (Fig. S1).

In agreement with previous observations (Logg et al., 2008; Wadskog et al., 2006), we observed that, in WT cells, Ena1 partitioned between the plasma membrane and intracellular compartments (Fig. 1A). Furthermore, more than 95% of Ena1–GFP intracellular structures were stained by the lipophilic dye FM4-64 (following a 30 min chase), indicating that they are connected to the endocytic pathway (Fig. S2).

We quantified the internal (I) GFP fluorescence intensity as a fraction of the total (T) fluorescence intensity (denoted I/T) and defined an ‘intracellular localization’ (‘IL’) index as the cumulative fraction of cells with more than 50% of Ena1 in intracellular compartments (see Materials and Methods). WT cells exhibited an ‘IL’ of 0.89±0.04 (mean±s.e.m., Fig. 1A) indicating that ∼89% of this cell population contained Ena1 enriched (>50%) in internal compartments. In addition, individual WT cells showed Ena1–GFP localized at the plasma membrane (P), corresponding to only 13% of their total fluorescence intensity [to give a fraction (P/T) of median value 0.13 P/T; Fig. 1B]. The evaluation of cell-based observations (P/T) along with the quantitative, population-based IL index is important to avoid result misinterpretation due to variability of individual cell images. See Materials and Methods and Fig. S3 for a more detailed description of IL and P/T value estimation and their complementary meaning.

As expected for proteins undergoing endocytosis, Ena1 accumulated at the plasma membrane of the general endocytosis-deficient sla2Δ strain and therefore cells had a median amount of Ena1–GFP localized at the plasma membrane double that of WT cells (Fig. 1B) while exhibiting an IL=0.08±0.04 (Fig. 1A). As described previously (Roth and Davis, 1996), internalization of the constitutively endocytosed pheromone receptor Ste3 was also substantially inhibited in sla2Δ as compared to WT cells (Fig. 1A,B).

If as hypothesized, Ena1 internalization was dependent on epsin function, then the transporter should be largely localized at the plasma membrane (i.e. should display low IL and high P/T values) in ent1Δent2Δ double epsin deleted cells (hereafter referred to as ΔΔ). Indeed, as opposed to the WT strain, only ∼11% of the ΔΔ cells had Ena1 enriched in internal compartments (IL=0.11±0.05; Fig. 1A). Similar to sla2Δ, ΔΔ cells showed a substantial accumulation of Ena1–GFP at the plasma membrane (Fig. 1B). In contrast, as expected for a cargo that does not depend on the epsins for internalization (Maldonado-Báez et al., 2008), Ste3 was properly localized to internal structures in 100% of ΔΔ cells (Fig. 1A). Importantly, re-introduction of either yeast epsin Ent1 or Ent2 could rescue the Ena1 localization defect in ΔΔ cells (Fig. 1C). Along the same lines, Ena1 localization to intracellular compartments was not affected in either single epsin knockout (ent1Δ or ent2Δ), also suggesting that the epsins are redundant for promoting Ena1 uptake (Fig. 1C).

Further supporting the epsin specificity of Ena1 internalization, the transporter distribution was not affected by deletion of the genes corresponding to the Yap180 family of endocytic adaptors (Fig. S4A).

Although the Ena1–GFP distribution phenotype in ΔΔ cells was identical to the one seen in the classical endocytosis-defective strain sla2Δ (Fig. 1), a direct observation of Ena1 internalization defects in epsin-deficient cells was lacking. Therefore, based on previous knowledge, we devised a strategy to directly test the presence of Ena1 endocytosis abnormalities in ΔΔ cells. It is known that under salt stress conditions Ena1 expression is induced and the protein is translocated to the plasma membrane to promote Na⁺ efflux (Ruiz and Ariño, 2007; Serrano et al., 1999; Yenush, 2016). Furthermore, Ena1 has been proposed to undergo internalization (Logg et al., 2008) and vacuolar degradation (Wadskog et al., 2006). Therefore, we speculated that upon salt stress relief, Ena1 will be subjected to internalization mediated by epsins followed by vacuolar degradation, and that this process will be affected in ΔΔ cells.

To test the potential contribution of epsins to low salt-induced downregulation of Ena1, we designed a salt stress relief assay (see Materials and Methods). Briefly, WT and ΔΔ cells were subjected to salt stress (0.5 M NaCl) following a short induction of Ena1–GFP expression. This was done to ensure that the expressed Ena1–GFP was recruited to and maintained at the plasma membrane to enable the efflux of excess Na⁺. Under Ena1 expression-suppressed conditions, and after lowering the salt concentration in the medium, cells were imaged over time and the peripherally localized fluorescence intensity of Ena1–GFP was quantified at each time-point. Fig. 2A shows a representative experiment where at least 50 cells at each time point were analyzed. In WT cells, the peripheral fluorescence intensity of Ena1–GFP underwent a sharp decrease 45 min after salt excess removal, indicating transporter uptake (Fig. 2A, upper panel). At later time points, the vacuolar fluorescence also dramatically diminished, indicating Ena1–GFP degradation (Fig. 2A, upper panel). In striking contrast, the fluorescence intensity of GFP–Ena1 at the periphery of ΔΔ cells remained at high values over time (Fig. 2A, lower panel), indicating that epsins are required for efficient Ena1–GFP removal from the plasma membrane.

The role of epsins in Ena1–GFP retrieval from the plasma membrane was confirmed in complementary experiments using flow cytometry. Specifically, WT and ΔΔ cells were allowed to express Ena1–GFP overnight, and samples were analyzed by flow cytometry at 0 h and 5 h under Ena1–GFP expression-suppressed conditions. Fig. 2B shows that 5 h after turning off the MET25 promoter (with 2 mM methionine) fluorescence intensity associated with WT cells was low due to internalization and vacuolar degradation of GFP–Ena1, whereas at 5 h, ΔΔ cells displayed high fluorescence intensity closer to that seen at the 0 h time point. It should be noted that, despite other endocytic adaptors (e.g. Ede1) being present, Ena1 was severely impaired for uptake in ΔΔ cells.

These results indicate that in the absence of epsins, internalization of Ena1 triggered by salt stress relief is substantially delayed. Also, as a whole, these experiments support the conclusion that changes in Ena1 cellular distribution (e.g. Fig. 1, and all throughout this study) upon deletion of the PM-localized epsin endocytic adaptors, largely reflects impairment in the Na⁺-transporter uptake.

Since ubiquitylation has been implicated in the trafficking of Ena1 (Wadskog et al., 2006) and epsin is a known Ub-binding protein, we hypothesized that the UIMs of epsin would be required for Ena1 internalization. Indeed, we observed that Ent1/2 variants bearing mutations that abolish the functionality of the two epsin UIMs [Ent1/2^Um: Ent1^{S177D, S201D} and Ent2^{S187D, S218D} (Aguilar et al., 2003)] were unable to rescue the Ena1 internalization defect seen in ΔΔ cells (Fig. 3).

In agreement with these findings, the UIM-containing C-termini of Ent1 or Ent2 (ΔENTH fragments), but not their lipid-binding ENTH domains, were also sufficient to suppress Ena1 uptake deficiency in ΔΔ cells (Fig. 3). It is worth reiterating that although ΔΔ cells express other Ub-binding endocytic adaptors such as Ede1, Ena1 uptake showed a strong dependence on epsin UIM integrity.

Since epsin UIMs are required for Ena1 internalization, we hypothesized that this cargo will undergo lysine-ubiquitylation as a requisite for endocytosis. It should be noted that Ena1 is a relatively large (1091 residues) protein with ten predicted transmembrane regions and a large number of cytosol-facing lysine residues as potential targets for ubiquitylation. In order to tackle the daunting task of identifying putative internalization-relevant ubiquitylation sites, we took into consideration pioneering findings from other authors working with yeast multi-spanning endocytic cargoes. Specifically, it has been determined that the internalization of the a-factor receptor Ste3, and the α-factor receptor Ste2, required ubiquitylation of their cytosolic C-termini (Hicke and Riezman, 1996; Rohrer et al., 1993; Roth and Davis, 2000; Roth et al., 1998). Based on these findings, we hypothesized that the Ena1 C-terminal tail (residues 1044–1091) is important for its uptake. In agreement with this, C-terminal truncations of Ena1 displayed intracellular distribution defects (Fig. 4A,B). Furthermore, an Ena1 truncation that lacked only the last four amino acids GIKQ (Ena1^Δ1088-1091) failed to internalize in WT cells (Fig. 4A,B). Since lysine residues are sites for ubiquitylation, we speculated that K¹⁰⁹⁰ was the residue within the GIKQ sequence responsible for Ena1 internalization. In fact, the Ena1^K1090R mutant was severely impaired for uptake, thereby indicating a requirement of K1090 for Ena1 internalization (Fig. 4C). Point mutations of the adjacent residues (Ena1^G1088A, Ena1^I1089A and Ena1^Q1091A) did not grossly affect the localization of the transporter (although the Q1091A mutant showed a slight accumulation of Ena1 at the PM) (Fig. 4C).

Importantly, an in-frame C-terminal fusion of ubiquitin with the Ena1^K1090R mutant was able to re-establish its intracellular localization (Fig. 4C), suggesting that K¹⁰⁹⁰ was a target for ubiquitylation.

Furthermore, and despite challenges for detecting notoriously difficult targets such as mono-ubiquitylated products and for pulling down multi-pass transmembrane proteins like Ena1, we found that Ena1 WT, but not a K1090R mutant (lacking the proposed ubiquitylation site) could be affinity-captured using Ub-binding beads (Fig. 4D). The data also suggest that, from all possible ubiquitylation sites, K¹⁰⁹⁰ is an important target, as mutation of this residue led to no detectable signal (Fig. 4D). Among other causes, this could be because a putative Ub unit introduced at K¹⁰⁹⁰ is target for sequential di-ubiquitylation (and/or higher-order modification).

Ub fusion in-frame to Ena1^K1090R was used as a positive control (100% of Ena1 molecules were ‘ubiquitylated’). This Ena1^K1090R–Ub construction was efficiently affinity captured; however, differences in band intensity between mono- and di/multi-Ub product suggest that in-frame fusion of Ub was perhaps less efficiently targeted for Ub-linkage than Ena1 WT. In summary, Ena1 undergoes ubiquitylation and K1090 (a residue required for its endocytosis) plays an important role in such process.

Experiments performed with rcy1Δ (recycling-deficient) strain showed that none of these mutant proteins accumulated in the characteristic/aberrant endosomal compartment of this strain suggesting the absence of mistrafficking to recycling routes (Fig. S4B). To test whether in-frame fusion of Ub to Ena1^K1090R was steering this mutated protein from the Golgi complex directly to the vacuole rather than restoring its uptake, we compared the intracellular fluorescent signal of Ena1^K1090R variants fused to GFP with the signal from the same proteins tagged with pHluorin (Fig. S4C). pHluorin is a GFP variant unable to emit fluorescence in low pH environments; therefore, the fluorescence signal produced by any Ena1–pHluorin fusions are expected to be more intense at the plasma membrane (if the protein indeed reaches the cell surface) where the pH is out of the quenching range. On the one hand, Fig. S4C shows that Ena1^K1090R accumulates at the plasma membrane and therefore no difference is seen when this protein is labeled with GFP or pHluorin. On the other hand, Ena1^K1090R–Ub fused to GFP shows a signal in the vacuole that is substantially more intense than in any other compartment, while the pHluorin fusion (which is dimmed in more acidic organelles like the vacuole) showed a weak signal that was mostly restricted to the plasma membrane (as expected for an efficiently internalized cargo; Fig. S4C). Furthermore, absence of the epsins (ent1/2; i.e. the ΔΔ strain) caused the pHluorin–Ena1^K1090R–Ub fusion to strongly accumulate at the cell cortex (Fig. S4C). These results strongly support the idea that Ena1^K1090R–Ub reaches the plasma membrane from where it is internalized towards the vacuole in an epsin-dependent manner.

Rsp5 is an E3 ubiquitin ligase in yeast involved in internalization of cargo, which can directly interact with its substrates via its WW domains (Lauwers et al., 2010). However, many cargoes do not contain the PPxY motif required for interacting with the WW domains of Rsp5 (Belgareh-Touze et al., 2008). Nevertheless, it is very well established that the arrestin-related trafficking (ART) family of adaptors specifically recognizes cargoes and they recruit Rsp5 to mediate selective cargo ubiquitylation via their own PPxY motif (Becuwe et al., 2012; Lauwers et al., 2010). Since Ena1 does not contain a PPxY motif, we tested whether its uptake was ART dependent. We quantified Ena1 localization within a panel of single ART-deleted cells (from art1Δ through art10Δ; Fig. S5). While Ena1 uptake was normal in most of these strains, it was significantly affected in art3Δ (Fig. 5A,C) and partially in art6Δ cells (Fig. S5). It should be noted that Art3 and Art6 are paralogs, and while the double Art3/6 deletion showed significant abnormalities in Ena1 distribution, this was not substantially worse than art3Δ (Fig. S5). This result suggests that although Art6 contributes to Ena1 internalization, Art3 is the major player in this process. Importantly, and as expected, this defect could be bypassed by an in frame fusion with Ub (Fig. 5A,C). We also monitored the uptake of transferrin receptor-like protein 1 (Tre1), which is a PPxY motif-containing protein that undergoes Rsp5-dependent ubiquitylation in an Art-independent manner (Stimpson et al., 2006; Sullivan et al., 2007). As predicted, Tre1 internalization was not affected in art3Δ cells (Fig. 5A). Importantly, expressing Art3 from a plasmid in art3Δ cells re-established Ena1–GFP internalization (Fig. 5B,C).

Interestingly, Ena1 uptake was also partially affected by deletion of the Art3 paralog Art6, but not by knoukout of Art9 (Fig. S5). Since Art9 has been shown to be involved in Ena1 functional regulation (Marqués et al., 2015), but our data showed that is dispensable for internalization, we speculate that this Art protein might play a role in a different Ena1 trafficking event or other regulatory step.

In summary our results indicate that Ena1 is ubiquitylated (Fig. 4D) and its internalization is impaired by: (1) epsin deletion or epsin UIM mutation (rescued by re-expression of either Ent1 or Ent2 or their corresponding UIM-containing C-terminal fragments); (2) mutation of the C-terminal Ena1 K1090R (rescued by a C-terminal Ub fusion-in-frame); or (3) deletion of the Ub-ligase Rsp5 adaptor Art3 (rescued by a C-terminal Ub fusion-in-frame). Therefore, we speculate that yeast epsins (via their UIMs) recognize Ena1 ubiquitylated (by an Art3–Rsp5 complex) at K¹⁰⁹⁰.

Although our results strongly suggest that Ena1 ubiquitylation is necessary for UIM-mediated recognition by epsin, other ubiquitylated yeast cargoes (e.g. Ste3) are internalized independently of Ent1 and Ent2. A corollary observation (further emphasizing Ub-binding adaptor specificity) is that, even in the presence of other ubiquitin-binding endocytic proteins (e.g. Ede1), Ena1 uptake is severely affected by the absence of epsins. Therefore, this led us to a very important unanswered question in the field: given that all Ub units are identical, what is the basis for cargo (e.g. Ena1) specificity for different adaptors (e.g. epsin)?

We speculated that a detection of an ubiquitylated lysine residue at the same time as another putative Ena1 determinant leads to epsin specificity. Since the C-terminus of Ena1 holds one of these elements (K¹⁰⁹⁰), we started our search for a hypothetical Ub-independent determinant within this region. Specifically, we performed a scanning mutational analysis of the Ena1 C-terminal tail by simultaneous mutation of four consecutive amino acids to glycine (denoted Gly ×4 mutants) (Fig. 6). This approach was preferred to alanine scanning to avoid the introduction of a potentially hydrophobic stretch of four alanine residues. It should also be highlighted that point mutations or truncations introduced between residues 1050–1070 in Ena1 C-terminus led to ER retention (see example in Fig. S1).

As expected, mutation of the G¹⁰⁸⁸IKQ¹⁰⁹¹ sequence led to loss of Ena1 internalization by elimination of the critical K¹⁰⁹⁰ (Fig. 6A,B). While most Gly ×4 Ena1 mutants displayed normal endocytosis, mutation of the S¹⁰⁷⁶TST¹⁰⁷⁹ residues caused a substantial impairment of cargo uptake (Fig. 6A,B). Interestingly, although 90% of cells expressing the T¹⁰⁸⁰IHT¹⁰⁸³-to-GGGG Ena1 mutant showed enrichment of the transporter in intracellular structures (Fig. 6A), they exhibited a PM-associated GFP fluorescence fraction slightly above normal levels (Fig. 6B). Therefore, we speculate that the threonine residues included in this region are functionally linked to the S¹⁰⁷⁶TST¹⁰⁷⁹ patch. The identified S/T-rich (including T¹⁰⁸⁰) sequence constituted an obvious candidate for phosphorylation. Indeed, mutational analysis and detection of the corresponding phosphorylated peptides by mass spectrometry indicated this to be the case (see below).

Mutation of S¹⁰⁷⁶TST¹⁰⁷⁹ to TSTS did not affect Ena1 uptake (Fig. 6C,D), suggesting that, rather than the exact sequence, the biochemical nature of S/T residues (e.g. as targets for phosphorylation) is important for its role in Ena1 internalization. Next, we changed the glycine residues in the uptake-defective G¹⁰⁷⁶GGG¹⁰⁷⁹ (Ena1^G1076GGG1079=Ena1^GGGG) quadruple mutant to aspartate (one at a time or in pairs) to mimic constitutive phosphorylation and to assess the relative importance of the residues within this ST patch. While single mutants failed, a double aspartate mutation (Ena1^{G1076D, G1077D}; denoted Ena1^DDGG) rescued the uptake of Ena1 (Fig. 6C). All other double aspartate mutations showed some level of improvement (Fig. 6C); that is, a slight increase of the number of cells with the corresponding Ena1–GFP mutant enriched in intracellular compartments and a decrease in levels of the transporter at the PM as compared to Ena1^GGGG (except for the GDDG mutant, Fig. 6C). However, only the Ena1–GFP DDGG and GGDD mutants also showed median P/T values within the normal range for the transporter (Fig. 6C). Since only Ena1^DDGG exhibited all signs of substantial rescue (i.e. significant increase in IL values and significant decrease in PM-associated mutant Ena1–GFP levels compatible with the normal range for the transporter), we concluded that the first two positions within the ST patch substantially contribute to Ena1 uptake.

To complement these studies, we generated S/T-to-alanine mutations of the ST motif in Ena1^WT to mimic constitutive dephosphorylation. While all Ena1 S/T mutants displayed defective uptake at certain extent, Ena1^{S1076A T1077A} (Ena1^AAST) was the most affected (Fig. 6D). In addition, the sole presence of S¹⁰⁷⁶ and T¹⁰⁷⁷ was enough to maintain proper Ena1–GFP uptake (Ena1^STAA–GFP; Fig. 6D).

As a whole, these results strongly suggest that the ST motif is a target for phosphorylation, and also suggest that in terms of function, S¹⁰⁷⁶ and T¹⁰⁷⁷ are the most relevant residues.

In order to gain insight as to what kinases could target the ST patch for phosphorylation, we monitored Ena1 distribution in seven strains bearing deletions in genes encoding for different S/T kinases (yck1-Δ1::ura3, yck2-2ts; ypk1Δ; ypk2Δ; pkh1Δ; pkh2Δ; ptk2Δ and kkq8Δ). Specifically, we selected strains deficient in S/T kinases implicated in salt homeostasis and/or endocytosis. While most of them did not lead to Ena1 uptake abnormalities, a yeast casein kinase-1 (Yck1) knockout and Yck2 temperature-sensitive double mutant strain (yck1-Δ1::ura3, yck2-2ts; hereafter referred to as yck^ts) was significantly affected in Ena1 internalization at the restrictive temperature of 37°C (Fig. 7A). Although this result suggests that Yck function is required for the transporter uptake, it should be noted that given the widespread function of Yck1/2 in endocytosis (Feng and Davis, 2000; Hicke et al., 1998; Marchal et al., 2002, 2000; Panek et al., 1997), yck^ts cells show general endocytosis defects and an overall decreased fitness (even at permissive temperatures). Importantly, we reasoned that if the Yck proteins are directly involved in Ena1 phosphorylation at the ST patch, and this is required for uptake, then the Ena1^{S1076D, T1077D} (Ena1^DDST) mutant that emulates constitutive phosphorylation of part of the ST patch, should be able to bypass the requirement for Yck1/2-mediated phosphorylation. Indeed, in contrast to Ena1^WT, Ena1^DDST was able to significantly internalize in yck^ts cells even at the restrictive temperature (Fig. 7B). Furthermore, in vitro phosphorylation assays showed that His₆-tagged Yck1/2 proteins (purified from yeast lysates) were capable of phosphorylating the bacterially produced purified Ena1^K1090R–Ub CT (where CT refers to C-terminal fragment, i.e. Ena1^1076-1091; Fig. 7C).

It should be noted that as a whole, the ST patch of Ena1 is flanked by the putative casein kinase phosphorylation sites E¹⁰⁷³AYS¹⁰⁷⁶ and T¹⁰⁸³E. Indeed, both S¹⁰⁷⁶ and T¹⁰⁸³ were found to be phosphorylated in whole-cell lysates of ΔΔ cells expressing Ena1–GFP and Yck1/2 by mass spectrometry (Table S1). These phosphorylated S/T residues are expected to yield additional casein kinase phosphorylation sites, for example: pS¹⁰⁷⁶TS¹⁰⁷⁸ and T¹⁰⁸⁰IHpT¹⁰⁸³ (where pS or pT represent phosphorylated serine or threonine, respectively), which could further propagate the post-translational modification within the ST patch. These sites were also found to be phosphorylated using mass spectrometry (Table S1). This propagation mechanism is expected to continue further and to yield a highly phosphorylated ST patch.

To further confirm the relevance of casein kinases for the internalization of Ena1, we also tested the effect of Yck1/2 overexpression on the uptake of an endocytosis-defective mutant (Fig. 7D). We reasoned that if Yck enzymes were involved in inducing Ena1 internalization, then their overexpression should enhance the transporter uptake. Since endocytosis of WT Ena1–GFP is highly efficient (IL≥0.8; P/T of ∼0.13) with a small window for improvement, we selected a transporter mutant with IL<0.5 and P/T>0.13 to improve the experiment sensitivity. Specifically, we chose the partial phosphorylation S¹⁰⁷⁶A Ena1 mutant (Ena1^ATST), which should be less efficient at initiating Yck-sequential phosphorylation of the ST patch. We predicted that Yck overexpression would ameliorate such phosphorylation impairment, improving the Ena1 mutant uptake. Indeed, overexpression of either Yck1 or Yck2 led to an increase in the proportion of cells with substantial intracellular Ena1^ATST–GFP and decreased the median accumulation of the mutant at the plasma membrane (Fig. 7D).

In summary, this section shows that Ena1 internalization is impaired by mutation of the S¹⁰⁷⁶TST¹⁰⁷⁹ patch to glycine or alanine (but this defect was reduced by introduction of asparate residues in positions 1076 and 1077). This ST patch is flanked by putative casein kinase phosphorylation sites and phosphorylation of this region was verified by mass spectrometry of yeast lysates. Interestingly, overexpression of Yck1/2 improved the uptake of an Ena1 ST patch partial mutant (S1076A). Ena1 internalization is also impaired upon Yck1/2 lack-of-function (but this defect could be bypassed by emulation of constitutive phosphorylation using the Ena1^{S1076D, T1077D} mutant). Furthermore, Yck1 or Yck2 purified from yeast lysates is able to phosphorylate the Ena1 C-terminus. Therefore, we propose that Yck-mediated phosphorylation of the S¹⁰⁷⁶TST¹⁰⁷⁹ patch is required for Ena1 internalization in addition to its ubiquitylation.

S/T phosphorylation has been shown to be a prerequisite for ubiquitylation during internalization of epsin-independent cargoes such as Ste6 (Kelm et al., 2004). We reasoned that if the ST motif was required for Ena1 ubiquitylation, then Ena1^AAST mutant with an Ub fused in-frame should bypass this requirement and undergo normal internalization. However, the resulting Ena1^AAST–Ub construction exhibited a substantial defect in uptake as compared to WT (Fig. 8A). A complementary experiment showed that Ena1^{DDST, K1090R} was also not internalized efficiently (Fig. 8A).

Although we cannot completely rule out the possibility that the ST patch is required for ubiquitylation and for a Ub-independent activity, taken together, these results strongly suggest that the ST motif and the K¹⁰⁹⁰ residue are independently required for epsin-mediated Ena1 internalization. It should be noted that these experiments do not rule out the possibility that phosphorylation of other determinants different from the S¹⁰⁷⁶TST¹⁰⁷⁹ play a role in promoting ubiquitylation of Ena1. An important corollary conclusion is that ubiquitylation would be necessary but not sufficient for Ena1 internalization.

To further define characteristics and requirements of this emerging ‘STK’ epsin-specific combined motif, we investigated whether the spacing between its constitutive elements impacts its function. Therefore, we modified the residue-distance between the transmembrane domain (TM) and the ST patch, as well as the residue separation between the ST patch and K¹⁰⁹⁰ by deleting or duplicating 5-residue long segments of the Ena1 C-terminal sequence. We adopted this approach instead of using artificial spacers to minimize the possibility of introducing changes in flexibility or hydrophobicity that could independently affect Ena1 behavior. However, we cannot completely rule out the possibility that besides spacing, our results may also reflect sequence-specific effects. In fact, our data indicate that eliminating residues between the TM and the ST patch led to ER retention, further supporting the idea that the integrity of the 1050–1070 region is crucial for Ena1 proper folding. In contrast, duplication of the 1066–1070 sequence did not affect the ability of Ena1 to exit the ER and to be internalized (Fig. S6A). When the ST–K residue distance was shortened from 10 to 5 (Δ1083–1087) there were deficiencies in Ena1 uptake, whereas duplication of a 5-residue sequence (1083–1087) did not affect internalization (Fig. S6A). These results suggest that a TM–ST spacing of ≥29 and ST-K of ≥10 sustains proper function of the STK sequence as a combined epsin-specific endocytic motif.

Interestingly, known epsin mammalian cargoes such as VEGFR2, selta-like 1 and ENaC bear C-terminal cytoplasmic sequences with similar composition (rich in serine, threonine or acidic residues) at the appropriate distance from a lysine residue, constituting potentially analogous STK motifs to the one identified in Ena1 (Fig. S6B). In fact, the ‘ST’-like motifs from these mammalian proteins were able to functionally replace Ena1 ST patch sustaining epsin-dependent Ena1 internalization (Fig. S6B).

Finally, and to evaluate whether epsin was able to associate with a transporter mutant emulating the double post-translationally modified STK Ena1 motif, we performed pull-down and co-immunoprecipitation (co-IP) experiments.

Glutathione beads bearing immobilized GST or GST–Ub–Ena1^{S1076D, S1077D}CT as described above and incubated with whole-cell lysates (WCL) from cells expressing HA–Ent1^149-454 (this truncation was preferred to the more aggregation-prone, ENTH-domain containing, full-length Ent1; see Materials and Methods). After incubation and washes, the beads were boiled in sample buffer, resolved by SDS-PAGE and the presence of bound HA–Ent1^149-454 investigated by western blotting with anti-HA antibodies. Results shown in Fig. 8B indicate that recombinant ‘constitutively phosphorylated/ubiquitylated’ cytoplasmic C-terminal tail of Ena1 was able to interact with in vivo-produced epsin protein from cell lysates.

Although co-IP of intrinsically low affinity sorting machinery–cargo complexes is challenging (and further aggravated by the presence of ten transmembrane domains in Ena1) we were able to co-immunoprecipitate full-length Ena1^{S1076D, S1077D}–Ub–GFP with HA-tagged Ent1^149-454 using an anti-HA antibody (a representative experiment is shown in Fig. 8C). This result indicated that epsin is capable of in vivo recognition of constructions that emulate constitutively phosphorylated/ubiquitylated STK from Ena1.

Epsin is an endocytic protein essential for cell viability in yeast and required for embryo development in metazoans (Sen et al., 2012). The function of this protein family as endocytic adaptors is necessary for activation of the Notch signaling pathway (Tian et al., 2004; Wang and Struhl, 2004, 2005) and for the temporal-spatial coordination of endocytosis and RhoGTPase signaling (Aguilar et al., 2006; Coon et al., 2010; Mukherjee et al., 2009). Epsin endocytic function is required for VEGFR2 (Pasula et al., 2012), VEGFR3 (Liu et al., 2014) and ENaC internalization (Staruschenko et al., 2005; Wang et al., 2006) and (although not essential) is known to contribute to the uptake and regulation of EGFR, insulin receptor, mu opioid receptors, dopamine transporter, protease activated receptor 1, Ste2 and others (Bertelsen et al., 2011; Chen et al., 2011; Dores et al., 2010; Hawryluk et al., 2006; Henry et al., 2012; Kazazic et al., 2009; Shih et al., 2002; Sigismund et al., 2005; Sorkina et al., 2006; Sugiyama et al., 2005).

However, how epsin (or any other Ub-binding endocytic adaptors) recognize their specific cargoes among multiple ubiquitylated proteins is unknown. Here, we speculate that there is a coincidence detection mechanism that assures specific recognition of ubiquitylated cargo by yeast epsins. We also suggest that other adaptors, such as the Eps15-homolog Ede1, would require post-translational modification of other determinants (in addition to ubiquitylation) for recognition of their corresponding specific cargoes.

In fact, this study unveils several important aspects of yeast epsin-mediated internalization of cargo. We have established the Na⁺ pump Ena1 as the first epsin-specific cargo identified in S. cerevisiae and showed that epsins sustain fast internalization of Ena1–GFP upon salt-stress relief. Specific residues within the Ena1 C-terminus (S¹⁰⁷⁶TST¹⁰⁷⁹ and K¹⁰⁹⁰) critical for its epsin-mediated internalization were also mapped. Furthermore, our data suggest that the K¹⁰⁹⁰ residue is a target for Art3/Rsp5-mediated ubiquitylation, while the ST patch is a target for yeast casein kinase (Yck1/2)-mediated phosphorylation. Our results also support the idea that the two regulatory determinants – Ena1 S¹⁰⁷⁶-T¹⁰⁷⁹ and K¹⁰⁹⁰ – are independently required for epsin-mediated Ena1 internalization. Therefore, and very importantly, our data also suggest that Ub is not sufficient for Ena1 uptake and suggest the existence of a coincidence detection mechanism for recognition of ubiquitylated cargoes by epsin.

Our results indicate that internalization of Ena1 by epsins does not rely on direct recognition of sequence-encoded sorting signals [such as Y- or L-motifs (Traub and Bonifacino, 2013)], on a double post-translational modification (ubiquitylation and phosphorylation) of specific cargo regions.

On the one hand, Ena1 internalization was dependent on the presence of the arrestin-related adaptor Art3 linking the yeast ubiquitin ligase Rsp5 to this process. This is in agreement with the well-known role of Rsp5 in endocytosis (Rotin et al., 2000). Nevertheless, it is possible that other Ub-ligases might be involved. Although Art9 has been shown to be important for the regulation of Ena1 (Marqués et al., 2015), our data indicates that is not crucial for internalization of the transporter. We speculate that Art9 is involved in the ubiquitylation of K residues relevant to other steps of intracellular traffic while Ena1 K¹⁰⁹⁰-Ub is required for removal from the plasma membrane. On the other hand, our results show that the Ena1 S¹⁰⁷⁶TST¹⁰⁷⁹ patch can be indeed phosphorylated in vivo and strongly suggest that yeast casein kinases are involved in the transporter uptake. Although we cannot completely rule out involvement of other S/T kinases in Ena1 internalization, our findings are in agreement with the well-known role of Yck proteins in endocytosis (Feng and Davis, 2000; Hicke et al., 1998; Marchal et al., 2002, 2000; Panek et al., 1997).

It has been clearly shown that Yck activity is a pre-requisite for ubiquitylation of the epsin-independent cargoes Ste2 and Ste3 (Ballon et al., 2006; Feng and Davis, 2000; Horák, 2003; Panek et al., 1997). Along the same lines, it has been reported that, following Yck-mediated phosphorylation, several specific ART-family proteins recruit Rsp5, resulting in the ubiquitylation of these cargoes (Alvaro et al., 2016, 2014; Prosser et al., 2015). In contrast, our data indicated that K¹⁰⁹⁰ and the ST patch are independently needed for cargo uptake. Nevertheless, it is possible that phosphorylation of other determinants in Ena1 cytoplasmatic domains could support Ena1 ubiquitylation. Therefore, it is tempting to speculate that independence between phosphorylation of this specific ST patch and ubiquitylation in Ena1 is a hallmark of epsin-specific internalization.

Our results also suggest the existence of an epsin-recognized combined STK motif with specific required spacing between its components and from the last transmembrane domain. We speculate that such requirements may reflect steric/spatial constraints for binding the transmembrane STK-bearing cargo by the membrane-bound epsins.

Searching the S. cerevisiae database (http://www.yeastgenome.org/), we found several transmembrane proteins displaying putative STK-like sequences with similar spacing and position to that in Ena1 (including the whole Ena protein family). However, the functionality of putative STK motifs for epsin-mediated internalization must be experimentally tested (and it will be the focus of future investigations) as unknown context factors and additional determinants might also play an important role in the control of PM cargo levels.

Indeed, the complexity of the trafficking information packed in the cytosolic (and possibly transmembrane) portions of membrane proteins should not be underestimated. Important work done in the Chen lab clearly indicates that, in mammalian epsins, specificity determinants for the recognition of the receptor tyrosine kinase VEGFR2 are embedded within the UIM sequence environment (Dong et al., 2017, 2015; Rahman et al., 2016). Furthermore, it should be kept in mind that multi-spanning proteins like Ena1 usually contain additional/critical determinants in their intracellular loops; these regions may play a role in the recruitment of post-translational modifying enzymes (e.g. kinases, ubiquitin ligases/ARTs or proteins that may assist in cargo recognition). A binding site for the fly Ub-ligase Mind bomb has been found within the cytosolic domain of the epsin cargo protein Delta (Palardy and Chitnis, 2015); very interestingly, this region is flanked by STK-like elements. Furthermore, the simple transplantation of the Ena1 STK motif was not enough to convert Ste3 into an epsin-specific cargo but rendered a plasma membrane-localized hybrid protein. A similar result was observed upon replacement of Ena1 C-terminus for the analogous Ste3 sequence (Fig. S4D). These observations further suggest a complex organization of partially overlapping motifs required for cargo internalization.

Based on our results, we speculate about the specific regulation of the plasma membrane levels of the Na⁺ pump Ena1, and propose the following working model (Fig. 8D). Upon salt stress, Ena1 is expressed and translocated to the cell cortex to contribute to the efflux of Na⁺ (Yenush, 2016) (Fig. 8D). When salt levels decrease, Ena1 is post-translationally modified by (1) ubiquitylation at K¹⁰⁹⁰ (by an Art3–Rsp5 complex) and (2) phosphorylation at the S¹⁰⁷⁶TSTT¹⁰⁸⁰ patch (by casein kinases) (Fig. 8D). The double modified C-terminus of Ena1 then becomes competent for specific recognition of yeast epsins and recruitment to nascent endocytic sites by these Ub-binding adaptors. Epsins, in turn, contribute to the binding of accessory factors and the clathrin coat (Sen et al., 2012), which eventually leads to the pinching-off of endocytic vesicles, thereby removing Ena1 from the plasma membrane (Fig. 8D).

We believe that this model, supporting findings and reagents will be the basis of future investigations that will further our understanding of endocytic function of epsin in yeast. Importantly, one could speculate that alternative coincidence detection mechanisms confer recognition specificity by other ubiquitylated-cargo/ubiquitin-binding adaptor combinations (e.g. involving Ede1). Finally, since mammalian STK motifs sustain Ena1 internalization in yeast, we also speculate that this strategy might be conserved across evolution. Nevertheless, these emerging hypotheses will need to be tested taking into account the complexity (see above) and potential redundancy of endocytic signals and endocytic adaptors.

Materials were purchased from Thermo Fisher Scientific (Fairlawn, NJ) or Sigma (St Louis, MO) unless stated otherwise. Plasmids and strains used in this study are listed in Tables S2 and S3. DNA constructs were prepared using standard techniques. Site directed mutagenesis was done using a QuikChange^® Lightning site directed mutagenesis kit (Agilent Technologies, Inc., Santa Clara, CA).

Yeast SEY6210-derived (MATα leu2-3,112 ura3-52 his3-Δ200 trp1-Δ901 suc2-Δ9 lys2-801; GAL) and other strains were grown overnight at 30°C (or room temperature for temperature-sensitive strains) with shaking at 250 rpm in standard yeast extract–peptone–dextrose (YPD) or synthetic selective (amino acid dropout) medium supplemented with dextrose and lacking appropriate amino acids for plasmid maintenance. Yeast were transformed by the Li-acetate method following standard techniques.

Images were acquired using a Zeiss Axiovert 200M microscope equipped with a Zeiss Axiocam MRm monochrome digital camera and Carl Zeiss Axiovision image acquisition software (version 4.4). For imaging MET25 promoter-driven Ena1–GFP-derived constructs, cells were grown overnight in selective medium in the presence of 2 mM methionine. The cultures were diluted, washed with sterile H₂O, and Ena1 expression was induced by growing the cells in selective medium without methionine for 5 h. A cell culture volume containing the equivalent of 2–5 units of optical density at 600 nm (OD_600nm) of cells expressing fluorescently tagged proteins were pelleted and resuspended in 50–100 μl selective medium, and 10 μl of cell suspension was spotted on a pre-cleaned slide, covered with a 22×22 mm coverslip and imaged with appropriate filters as ∼0.3 μm-spaced Z-stacks.

To visualize the endocytic pathway and vacuolar membrane, staining using FM4-64 was performed according to Baggett et al. (2003). Briefly, cells were grown overnight to 0.3–0.8 OD₆₀₀/ml in selective medium. 1 ml of cells were pelleted and chilled on ice, followed by the addition of cold FM4-64 (1:40 dilution from a stock of 1 mg/ml in DMSO) in a final volume of 50 µl of selective medium. The cells were incubated for 15–20 min on ice and subsequently imaged (t=0 min). Cells were supplemented with 900 µl of fresh pre-warmed selective medium and allowed to incubate at 30°C for 30 min. At this time point, cells were quickly chilled on ice prior to imaging at 100× using GFP and Rhodamine filters.

To quantify the localization of fluorescently tagged Ena1, we used an approach based on the analysis of fluorescence signal distribution within cells in randomly acquired images. First, we manually outlined each cell and measured the total GFP fluorescence intensity (T) using ImageJ software. T represents the sum of fluorescence contributions associated with the internal compartments (I), plasma membrane-localized signal (P) and cytosolic background (C). The fluorescence intensity inside of the cell (I_C=I+C) was measured by outlining the cell of interest excluding the peripheral region, while the P component value was estimated subtracting I_C from T. The average contribution of cytosolic fluorescence was quantified by measuring and averaging the fluorescence intensity of three rectangular regions within the cytosol of each analyzed cell. This value multiplied by the cell area yielded an integrated cytosolic background intensity (C). Internal fluorescence intensity was calculated as I=I_C−C. The I component arises from endocytic uptake of Ena1–GFP as shown by >95% colocalization with the hydrophobic dye FM4-64 (Fig. S2). All measurements were performed on a single plane with the best focus of cell periphery. This analysis was performed on at least 30 cells and each experiment was performed at three 3 times.

The results were processed and represented in two complementary ways. First, histograms (fraction of cells vs I/T values) provided a measurement population distribution. In addition to direct histogram comparisons, an ‘intracellular localization’ (IL) index was defined as the cumulative fraction of cells with intracellular fluorescence greater than 50% of the total fluorescence (i.e. the fraction of cells present at the right-hand side of a line positioned at I/T=0.5 over at least three histograms). This value integration procedure yielded a normally distributed, very convenient descriptor of population skewness with respect to Ena1–GFP localization. The higher the IL, the higher the proportion of cells with cargo enrichment in intracellular compartments. Second, we analysed P/T median values to provide a measure of Ena1–GFP plasma membrane accumulation in cells. Since P=T−I_C (see above), the calculation of this quantity does not require estimation of cytosolic background, thereby decreasing error propagation. Although showing typical cell-to-cell biological variation (and following a non-normal distribution), P/T values were found to be fairly independent of Ena1–GFP levels of expression (Fig. S3C). Nevertheless, highly overexpressing cells were excluded from all analyses. A P/T median value ≤0.2 (Fig. S3C) was considered within the limits for typical (normal) Ena1-GFP PM fraction range.

It should be noted that populations with similar ILs can show substantially different P/T median values.

WT and ent1Δent2Δ cells expressing Ena1–GFP were grown overnight in selective medium supplemented with 0.2 mM methionine to maintain low levels of Ena1 expression (controlled by MET25 methionine-repressible promoter). Cells were grown to early log phase, washed and transferred to medium lacking methionine for 2 h, thereby inducing high Ena1 expression for a short time. Next, 0.5 M NaCl was added to the cells to simulate salt stress and simultaneously supplemented with 2 mM methionine to suppress the MET25 promoter (stopping Ena1 production), and cells were incubated under these conditions for 3 h. This was done to ensure that the Ena1–GFP produced by the cells would be recruited and maintained at the cell membrane to pump out excess sodium. The cells were imaged (time 0), and then washed to remove NaCl excess (‘salt stress relief’ – under these conditions we expected Ena1 to be internalized) and resuspended in medium containing 2 mM methionine. Cells were imaged over time and the amount of peripherally localized Ena1 was quantified.

WT or ent1Δent2Δ cells transformed with Ena1–GFP plasmid were cultured in synthetic complete medium lacking methionine at 30°C overnight. The cells were diluted to 0.2 OD_600nm/ml with fresh medium and grown for 2 h. Cells were supplemented with methionine to a 2 mM final concentration, and a sample was taken immediately after (‘0 h’), and another after 5 h incubation at 30°C (‘5 h’). After staining with propidium iodide (BD Biosciences) for 10 min, 5×10⁴ cells were analyzed using a FC500 flow cytometer (Beckman-Coulter) with FL1 and FL3 detectors. The results were processed using FlowJo v7.6 software.

Cells were grown in selective medium and spheroplasted following standard techniques (Katzmann and Wendland, 2005). Briefly, cells at ∼50 OD₆₀₀ were harvested and incubated for 10 min at room temperature in softening solution [0.1 M Tris-HCl pH 9.4, 10 mM DTT and 5 mM N-ethylmaleimide (NEM)]. Cells were then resuspended in spheroplasting solution composed of YPD medium containing 1 M sorbitol, 1 unit of zymolyase per OD₆₀₀, protease inhibitor cocktail (Pierce) and 0.1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride (AEBSF) for 30 min at 30°C. Spheroplasts were washed once with PBS containing 1 M sorbitol, pelleted by centrifugation at 300 g for 5 min and dounce-homogenized with in HEPES/KOAc lysis buffer (0.2 M sorbitol, 50 mM KOAc, 2 mM EDTA, 20 mM HEPES pH 6.8, 0.2% Triton X-100, 50 mM NEM and protease inhibitor cocktail).

For Ena1–GFP detection by western blotting, cells were induced to express Ena1–GFP (WT or mutant) by maintaining them in medium without methionine for 5 h. The equivalent to 10 OD₆₀₀ of cells were collected and resuspended in 200 μl of 0.2 M NaOH for 15 min at room temperature. Cells were spun down at 1000 g for 2 min and resuspended in 50 μl of 2× Laemmli buffer with 8% SDS (100 mM Tris-HCl pH 6.8, 8% SDS, 40% glycerol, 2% mercaptoethanol, 0.005% Bromophenol Blue) and incubated at 55°C for 15 min. The equivalent to 50 μl of 0.15-0.25 mm acid washed beads were added to the sample and vortexed for 1 min at room temperature. The samples were chilled on ice for 2 min and spun down at 300 g for 5 min. The supernatants were collected and resolved on 8% SDS-PAGE gels, transferred onto nitrocellulose followed by immunoblotting using an anti-GFP antibody (Thermo Fisher Scientific, A-11122) at a 1:1000 dilution. The loading control, VDAC1 (porin) was detected by using a specific mouse monoclonal antibody (clone 16G9E6BC4, Thermo Fisher Scientific) at 1:1000 dilution. Blots were developed using the enhanced chemiluminescence detection system.

Cells expressing Ena1–GFP were grown overnight in the presence of 0.2 mM methionine. Cells were diluted in medium without methionine to a final 0.5 OD₆₀₀/ml and to induce Ena1–GFP expression for 5 h. Cells at 200 OD₆₀₀ were collected by centrifugation at 1000 g for 5 min, resuspended in ice-cold water and aliquoted to give 10 OD₆₀₀/tube. The cells were then pelleted down at 2000 g for 2 min, and 0.15–0.25 mm acid-washed glass beads (equivalent to a 100 μl volume) and 1 ml of ice-cold lysis buffer [30 mM Tris-HCl pH 7.5, 5 mM EDTA, 25 mM DTT, 250 mM NaF, 2× protease inhibitors and 2× phosphatase inhibitor cocktails (Pierce)] were added to each tube. The cells were broken by vortexing for 3 min, then chilled on ice for 2 min, repeating the cycle twice. The crude membranes were collected by pelleting the lysates at 300 g for 5 min, the supernatants were pooled and centrifuged at 20,000 g 30 min at 4°C. The pellet was resuspended in 1 ml of lysis buffer containing 0.1% SDS.

GFP-tagged Ena1 was purified by incubating extracted crude membranes with GFP-Trap agarose beads (ChromoTek, NY) for 1 h at 4°C. The GFP-Trap beads were washed with RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40, 0.5 mM EDTA, 0.1% SDS, 0.5% sodium deoxycholate, and 50 mM NaF) three times, and enriched Ena1–GFP proteins were denatured in 8 M urea and reduced with 5 mM dithiothreitol for 30 min at 37°C. The proteins were further alkylated in 15 mM iodoacetamide for 1 h in the dark at room temperature and then digested with proteomics grade trypsin at a 1:100 (w/w) ratio overnight at 37°C. The digested peptides were desalted by SDB-XC reverse phase-Stage Tips (Rappsilber et al., 2007). Phosphopeptides from Ena1–GFP were enriched by polyMAC (Iliuk et al., 2010). The eluted phosphopeptides were dried in a SpeedVac and stored at −20°C.

The phosphopeptides were dissolved in 4 μl of 0.1% formic acid and injected into an Easy-nLC 1000 (Thermo Fisher Scientific). The mobile phase buffer consisted of 0.1% formic acid in ultra-pure water (buffer A) with an eluting buffer of 0.1% formic acid in 80% ACN (buffer B) run over a linear 90 min gradient with a flow rate of 250 nl/min. The Easy-nLC 1000 was coupled online with a Velos LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific). The mass spectrometer was operated in the data-dependent mode in which a full-scan MS (from m/z 300–1500 with the resolution of 60,000 at m/z 400) was followed by top 15 MS/MS scans of the most abundant ions. The raw files were searched directly against the S. cerevisiae database (UniprotKB) with no redundant entries using both the SEQUEST HT and BYONIC algorithm on Proteome Discoverer^TM version 2.0 (Thermo Fisher Scientific). Peptide precursor mass tolerance was set at 10 ppm, and MS/MS tolerance was set at 0.6 Da. Search criteria included a static carbamidomethylation of cysteine residues (+57.0214 Da) and variable modifications of (1) oxidation (+15.9949 Da) on methionine residues, (2) Gln to pyro-Glu (−17.027 Da) at the N-terminus of peptides, (3) acetylation (+42.011 Da) at N-terminus of proteins, and (4) phosphorylation (+79.996 Da) on serine, threonine or tyrosine residues were searched. Searching was performed with full tryptic digestion and allowed a maximum of two missed cleavages on the peptides analyzed from the sequence database. Relaxed and strict false discovery rates (FDR) were set for 0.05 and 0.01, respectively.

Bacterially produced recombinant proteins were purified based on the procedure described in Aguilar et al. (2003). Rosetta cells (Novagen) were transformed with GST- or His₆-tagged constructs (listed in Table S1) and protein production was induced by incubation with 0.2 mM (final concentration) isopropyl-β-D-thiogalactopyranoside (IPTG) for 5 h at 30°C. The cells were harvested and resuspended in PBS containing 0.1% Tween (PBST) plus 5% glycerol, protease inhibitor cocktail and 0.1 mM AEBSF. Cell lysis was induced by incubating the cell suspension with 1 mg/ml lysozyme (4°C, 15 min) followed by sonication. Clarified lysates were incubated either with Ni-NTA beads in the presence of 5 mM imidazole (for purification of His₆-tagged proteins) or with glutathione-coupled beads (for purification of GST-tagged proteins), for 2 h at 4°C. The beads were washed four times with PBST with imidazole (in increasing concentrations of 5 mM, 10 mM, 15 mM and 20 mM) for His₆-tagged proteins, and four times with PBST alone for GST-tagged proteins. Elution was performed with either 500 mM imidazole (His₆-tagged proteins) or 50 mM glutathione (GST-tagged proteins). The supernatants containing the purified tagged proteins were collected, and the imidazole from His₆-tagged and glutathione from GST tagged protein fractions were removed using Zeba™ spin desalting columns (Thermo Scientific, Waltham, MA).

C-terminal His₆-tagged Yck1/2 proteins were expressed in yeast and purified from whole-cell lysates (equivalent to ∼100 OD₆₀₀ of cells) using Ni²⁺-NTA beads and eluted with 500 μl of 500 mM imidazole and desalted into 500 μl of 2× kinase buffer (100 mM Tris-HCl pH 7.5, 20 mM MgCl₂, 2 mM DTT). Purified Yck1/2 proteins were incubated with GST–Ena1^K1090RCT–Ub immobilized on 50 μl of glutathione beads at 30°C for 45 min in 1× kinase buffer with 5 mM ATP. The beads were then washed three times with 1× kinase buffer and boiled with 25 μl of 2× protein sample buffer. Samples were resolved with SDS-PAGE and phosphorylated proteins were detected using the Pro-Q^® Diamond phosphoprotein gel stain kit (ThermoScientific, Waltham, MA) following the manufacturer's instructions.

Overexpression of HA–Ent1^149-454 under the regulation of a MET25 promoter was induced by growing the cells overnight in selective medium in the absence of methionine. For these and for co-immunoprecipitation experiments, the Ent1 truncation indicated above was preferred to the more aggregation-prone, ENTH-domain containing, full-length Ent1. Cell lysates were prepared as described above. The samples were centrifuged at 5000 g for 5 min, and clarified lysates were incubated with GST or GST fusion proteins immobilized on glutathione-coupled beads for 2 h at room temperature. The beads were washed five times with PBST, resuspended in Laemmli's protein sample buffer, followed by SDS-PAGE resolution and immunoblotting. A mouse monoclonal anti-HA antibody was used (Cat# MMS-101R, Covance, Princeton, NJ) at 1:2500 dilution.

Yeast cells expressing MET25 promoter-driven Ena1–GFP transformed with empty vector or a plasmid encoding for HA-tagged Ent1^149-454 (under the endogenous promoter), were grown overnight in selective medium in the absence of methionine. Cells were lysed by spheroplasting and dounce homogenization as described above and subjected to cross-linking using 2 mM DSP [Dithiobis (succinimidyl propionate)] for 30 min at room temperature with constant rotation. The crosslinking reaction was quenched by incubating with 50 mM Tris-HCl pH 7.5 for 15 min. For protein extraction, mild alkali treatment was used (Kushnirov, 2000). NaOH was added to the mix to a final concentration of 0.2 M and incubated for 5 min at RT. Samples were centrifuged at 5000 g for 5 min and the supernatant was discarded. The pellet was resuspended in extraction buffer (0.06 M Tris-HCl pH 7.5, 1% SDS, 5% glycerol, 2% β-mercaptoethanol, supplemented with NEM, protease and phosphatase inhibitors) and boiled for 3 min. Following centrifugation at 5000 g for 5 min, the clarified supernatant was diluted at least 20 times in 1 M Tris-HCl pH 7.5. Anti-HA coupled beads (EZview red anti-HA affinity gel, Sigma) were used to immunoprecipitate HA-tagged proteins (at 4°C for 2 h). Following three washes, the beads were boiled in Laemmli buffer for 5 min. Samples were resolved by SDS-PAGE and immunoblotted. A rabbit monoclonal anti-GFP antibody (Cat# G10362, Invitrogen, Carlsbad, CA) at 1:10 dilution was used to detect GFP-tagged proteins from the immunoprecipitates. HA tagged proteins were detected using a mouse monoclonal α-HA antibody (Cat# MMS-101R, Covance, Princeton, NJ) at 1:2500 dilution.

Cells were induced to express Ena1–GFP (WT or mutant) by maintaining them in medium without methionine for 5 h. Cells at ∼100 OD₆₀₀ were spheroplasted as described in the previous section (see Cell lysate preparation). Total ubiquitylated proteins were affinity-captured using Ubiquitin Affinity beads (UBA01; which uses a proprietary formulation of ubiquitin-binding domain proteins able to detect difficult targets such as mono-ubiquitylated products) included in the Signal-Seeker™ ubiquitylation detection kit (Cytoskeleton Cat. #BK161) following the manufacturer's instructions. The Ena1-GFP from samples and 5% input lysates were resolved in 8% SDS-PAGE gels, transferred onto nitrocellulose followed by immuno-blotting using an anti-GFP antibody (Thermo Fisher A-11122) at a 1:1000 dilution.

The analysis was performed as described below and in Taylor (1997). When appropriate, the magnitude of errors associated with values derived from algebraic operations using experimentally measured quantities were calculated following standard rules of error propagation. Statistical significance of differences between fluorescence distribution histograms were analyzed using the Kolmogorov–Smirnov test. The Student's t-test was used to evaluate differences among normally distributed ILs, while the Wilcoxon's test was used to evaluate the significance between non-normal P/T of value samples. Bonferroni's correction for multiple comparisons was performed whenever applicable [α_C=p/n; n being the number of comparisons].

We thank Drs Chris Staiger, Donald Ready, Henry Chang and Tony Hazbun (Purdue University) for stimulating discussions and/or critical reading of the manuscript. We thank Dr Jeremy Thorner (UC Berkeley) for strains and for graciously allowing the use of his lab space and equipment. We also thank Drs Lucy Robinson (Louisiana State University), Ingrid Wadskog (Göteborg University) and Martha Cyert (Stanford) for reagents.