Cooperative regulation of endocytic vesicle transport by yeast Eps15-like protein Pan1p and epsins

Dynamic actin filaments are required for the formation and internalization of endocytic vesicles. Yeast actin cables serve as a track for the translocation of endocytic vesicles to early endosomes, but the molecular mechanisms regulating the interaction between vesicles and the actin cables remain ambiguous. Previous studies have demonstrated that the yeast Eps15-like protein Pan1p plays an important role in this interaction, and that interaction is not completely lost even after deletion of the Pan1p actin-binding domain, suggesting that additional proteins mediate association of the vesicle with the actin cable. Other candidates for mediating the interaction are endocytic coat proteins Sla2p (yeast Hip1R) and Ent1p/2p (yeast epsins), as these proteins can bind to both the plasma membrane and the actin filament. Here, we investigated the degree of redundancy in the actin-binding activities of Pan1p, Sla2p, and Ent1p/2p involved in the internalization and transport of endocytic vesicles. Expression of the nonphosphorylatable form of Pan1p, Pan1-18TA, caused abnormal accumulation of both actin cables and endocytic vesicles, and this accumulation was additively suppressed by deletion of the actin-binding domains of both Pan1p and Ent1p. Interestingly, deletion of the actin-binding domains of Pan1p and Ent1p in cells lacking the ENT2 gene resulted in severely defective internalization of endocytic vesicles and recruitment of actin cables to the site of endocytosis. These results suggest that Pan1p and Ent1p/2p cooperatively regulate the interaction between the endocytic vesicle and the actin cable.


Edited by Enrique De La Cruz
Dynamic actin filaments are required for the formation and internalization of endocytic vesicles. Yeast actin cables serve as a track for the translocation of endocytic vesicles to early endosomes, but the molecular mechanisms regulating the interaction between vesicles and the actin cables remain ambiguous. Previous studies have demonstrated that the yeast Eps15-like protein Pan1p plays an important role in this interaction, and that interaction is not completely lost even after deletion of the Pan1p actin-binding domain, suggesting that additional proteins mediate association of the vesicle with the actin cable. Other candidates for mediating the interaction are endocytic coat proteins Sla2p (yeast Hip1R) and Ent1p/2p (yeast epsins), as these proteins can bind to both the plasma membrane and the actin filament. Here, we investigated the degree of redundancy in the actin-binding activities of Pan1p, Sla2p, and Ent1p/2p involved in the internalization and transport of endocytic vesicles. Expression of the nonphosphorylatable form of Pan1p, Pan1-18TA, caused abnormal accumulation of both actin cables and endocytic vesicles, and this accumulation was additively suppressed by deletion of the actin-binding domains of both Pan1p and Ent1p. Interestingly, deletion of the actin-binding domains of Pan1p and Ent1p in cells lacking the ENT2 gene resulted in severely defective internalization of endocytic vesicles and recruitment of actin cables to the site of endocytosis. These results suggest that Pan1p and Ent1p/2p cooperatively regulate the interaction between the endocytic vesicle and the actin cable.
Endocytosis is a critical process in which a vesicle containing various cargos, such as extracellular molecules, membrane proteins, or certain types of viruses, buds off from the plasma membrane to internalize its content into the cell. Among the many forms of endocytosis, the best characterized is clathrinmediated endocytosis (1,2). In this process, over 50 proteins have been shown to be required for formation and transport of clathrin-coated vesicles (CCVs) (3,4). Recent live-cell imaging of various eukaryotic cells has revealed that the actin cytoskeleton plays an essential role in the formation and internalization of CCVs (5)(6)(7), and the molecular mechanisms involved have been well characterized. Endocytic vesicles also require actin cables, which are polarized linear bundles of actin filaments (8,9), to mediate their transport to the early endosome (10,11), but it still remains unclear how endocytic vesicles interact with such actin cables. Previous studies have demonstrated that endocytic vesicles move on the actin cable, independently of type V myosins (Myo2p and Myo4p), and passively with the cable itself (7,10). This suggests the existence of endocytic protein(s) that fix the endocytic vesicle to the actin cable through an ability to bind to the vesicle, the cable, or both.
In yeast, after internalization, each CCV is uncoated through the combined actions of phosphatidylinositol 4,5-bisphosphate phosphatases Sjl1p and Sjl2p, Arf3p GTPase, and Ark1p/Prk1p kinases (4,12). The Ark1p/Prk1p kinases, which are related to the mammalian proteins GAK and AAK1 (13), have been shown to regulate the disassembly of endocytic coat proteins and actin by phosphorylating several target proteins, such as Sla1p, Ent1/ 2p, Yap1801/2p, Scd5p, and Pan1p (14)(15)(16). A study using a prk1 analog-sensitive mutant has revealed that transient inactivation of Prk1p causes stable association of endocytic vesicles with actin cables (17,18), and inhibits dissociation of the vesicle from the cable, suggesting that substrates of Ark1p/Prk1p might be key regulators of endocytic vesicle transport along the actin cable. The Eps15-like protein Pan1p is the major in vivo target of Ark1p/Prk1p kinases, and its ability to both bind actin filament and promote actin polymerization is negatively regulated by Prk1p phosphorylation (19). Pan1p contains 18 Ark1p/Prk1p phosphorylation consensus sequences ([L/I/V/M]xx[Q/N/T/S] xTG), and mutation of the 18 threonine to alanine (Pan1p-18TA) causes abnormal cytoplasmic actin aggregates, also called actin clumps, containing endocytic vesicles and endosomes with several endocytic and endosomal proteins (18). The pan1-18TA mutant exhibits less-polarized and more-aggregated actin cable structures in the actin clumps (18), indicating that the actin cable might stably associate with the unphosphorylated form of Pan1p, which resides on the endocytic vesicle.
Other candidates for mediating the interaction between the endocytic vesicle and the actin cable are the yeast epsin homologs Ent1p and Ent2p, as these proteins can bind to both the plasma membrane and the actin filament via the N-terminal epsin N-terminal homology domain and C-terminal actinbinding domain, respectively (20)(21)(22)(23). Ent1p and Ent2p also have Ark1p/Prk1p phosphorylation consensus sequences. The threonine residues in the consensus sequences are phosphorylated by Prk1p in vitro, and the overexpression of the unphosphorylated form of Ent1p partially suppresses the formation of the aberrant actin aggregates observed in ark1Δ prk1Δ cells (16). In addition, Prk1p phosphorylation sites are located within or in close proximity to the actin cytoskeleton-binding (ACB) domain, and the phospho-mimicking mutation of the Ent1 ACB domain reduces its F-actin-binding activity (22). Considering that Ent1p and Ent2p interact directly with Pan1p (24), the Pan1p-Ent1p complex might play an important role in regulating the interaction between the endocytic vesicle and the actin cable. A previous study has shown that Ent1p and the yeast HIP1R homolog Sla2p interact redundantly with actin filament and that deletion of the actin-binding domains of these proteins causes a severe defect in endocytosis (22). As Sla2p also binds to both the plasma membrane and actin filament via the N-terminal ANTH domain and the C-terminal THATCH/talinlike domain (22,25,26), Sla2p might also be involved in transport of the endocytic vesicle along the actin cable.
Here, we demonstrate a degree of redundancy in the actinbinding activities of Pan1p, Sla2p, and Ent1p/2p required for the formation and transport of endocytic vesicles. The pan1 mutant lacking the actin-binding domain was shown to exhibit slightly defective endocytic vesicle internalization when combined with the sla2 mutant, but markedly defective vesicle internalization as well as recruitment of the actin cable to the endocytic site when combined with the ent1 ent2 mutant. These results suggest that Pan1p and Ent1p/2p cooperatively regulate the interaction between the endocytic vesicle and the actin cable.

Results
Actin-binding domains of Pan1p, Sla2p, and Ent1p/2p are required for the formation of actin cable aggregates in the pan1-18TA mutant We previously demonstrated that Pan1p directly mediates the interaction between endocytic vesicles and actin cables via its actin-binding activity (18). Expression of the Pan1p-18TA mutant causes severe defective endocytosis and abnormal clumping of Pan1p, which contains endocytic vesicles and actin cables (18). Deletion of the actin-binding domain of Pan1p-18TA (pan1-18TAΔABD) decreased the actin cable aggregates, but they were not completely lost in the mutant (18), implying the existence of additional actin-binding coat protein(s) that stabilize vesicle association with actin cables in the clump. A previous study had shown that Sla2p and Ent1p, both of which are mid-coat endocytic proteins, interact redundantly with actin filaments (4,22). We therefore speculated that Sla2p and/or Ent1p/2p, in addition to Pan1p, might play a role in the interaction between endocytic vesicles and actin cables. To confirm this, we first examined whether Sla2p, Ent1p, and Ent2p are localized to clumps in the pan1-18TAΔABD mutant. We used Abp1-GFP as a marker for actin patch, assumed to be an endocytic vesicle (27), and found that Abp1-GFP is observed in most Pan1p clumps (Fig. 1, A and B). As expected, the localization of Sla2p, Ent1p, and Ent2p clearly corresponded to that of Pan1p accumulation (Fig. 1A). Quantification analysis revealed that Sla2p, Ent1p, or Ent2p was localized at 71.3%, 86.7%, or 98.7% of Pan1-mCherrylabeled clumps, respectively (Fig. 1B). To examine the contribution of the actin-binding activities of these proteins to the formation of actin cable aggregates in the pan1-18TA mutant, we constructed the pan1-18TA or pan1-18TAΔABD mutant carrying an individual deletion of the THATCH/talinlike domain of Sla2p (Sla2ΔTHATCH) or deletion of the ACB domain of Ent1p (Ent1ΔACB) ( Fig. 2A). In WT cells, actin cables, labeled by Abp140-3GFP, are highly dynamic polarized structures, whereas those in the pan1-18TA mutant were less polarized and had a more-aggregated structure, and 58.7% of the mutant contains such actin aggregates (Fig. 2, B and C). The actin cable aggregates were significantly suppressed in cells expressing the Pan1p-18TA mutant lacking the Figure 1. The localization of Sla2-GFP, Ent1-GFP, and Ent2-GFP in WT and pan1-18TA cells. A, cells expressing each GFP-tagged protein and Pan1-mCherry were grown to the early logarithmic phase to mid-logarithmic phase in the YPD medium at 25 C and observed by fluorescence microscopy. Merged images of GFP and mCherry channels are shown in the right panels. The scale bar represents 2.5 μm. B, quantification of actin clumps including GFPtagged proteins. The percentages were calculated as the ratio of actin clumps (n = 50) including each protein in each experiment. Error bars indicate the SD from at least three independent experiments. Different letters indicate significant difference at p < 0.05, one-way ANOVA with Tukey's post hoc test. YPD, yeast extract peptone dextrose.
C-terminal actin-binding domain (pan1p-18TAΔABD), but still remained in 30.1% of the cells (Fig. 2, B and C) (18). The sla2ΔTHATCH mutation little affected the actin cable aggregation in the pan1-18TA mutant (55.3%) (Fig. 2, B and C), and a combination of the pan1-18TAΔABD and sla2Δ-THATCH mutations did not significantly suppress the aggregation of actin cables (17.8%), compared with pan1-18TAΔABD mutation (Fig. 2, D and E). In contrast, the ent1ΔACB mutation significantly decreased the actin cable aggregation in the pan1-18TA mutant (41.4%) (Fig. 2, B and C), and deletion of the ENT1 or ENT2 gene in the pan1-18TAΔABD mutant resulted in a significant decrease of actin cable aggregation (9.5% or 8.3%, respectively) ( Fig. 2, D and E). Because the functions of Ent1p and En2p partially overlap, and disruption of both genes is lethal (23), we combined ent1ΔACB ent2Δ mutation with pan1-18TAΔABD mutation to examine the extent to which Ent1p/2p contributes to the actin-binding activity. Intriguingly, in the pan1-18TAΔABD ent1ΔABD ent2Δ triple mutant, accumulation of Pan1p and actin cables was completely abolished (Fig. 2, D and E). These results suggest that the actin-binding function of Pan1p and Ent1p/2p might contribute to the interaction between endocytic vesicles and actin cables.

Pan1p and Ent1p/2p cooperatively regulate actin-dependent endocytic vesicle internalization
We next examined whether lack of the actin-binding activity of these proteins affected the dynamics of endocytic vesicles. The pan1ΔABD, sla2ΔTHATCH, or ent1ΔACB ent2Δ mutant exhibited a slightly increased Abp1p patch lifetime (17.8 s, 20.0 s or 16.5 s, respectively) relative to WT cells (13.5 s) (Fig. 3, A and B). The pan1ΔABD ent1Δ or pan1ΔABD ent2Δ double mutant also exhibited a slightly increased Abp1p lifetime (18.8 s or 19.9 s). These effects were clearly enhanced in the pan1ΔABD sla2ΔTHATCH (22.7 s) or pan1ΔABD ent1ΔACB ent2Δ mutant (21.9 s) (Fig. 3, A and B). A previous study had reported that deletion of the actin-binding domains of Sla2p and Ent1p arrested endocytic vesicle budding and that actin was polymerized continuously at sites of endocytosis. We also found that combined deletion of the ENT2 gene and the actin-binding Regulation of actin-mediated endocytic vesicle transport domains of Sla2p and Ent1p (sla2ΔTHATCH ent1ΔACB ent2Δ mutation) had a lethal outcome (Fig. 3C), suggesting an essential role of these actin-binding domains in endocytic vesicle formation. Kymographs and particle tracking analysis showed that the majority of Abp1p patches in the pan1-ΔABD, sla2ΔTHATCH, or ent1ΔACB ent2Δ mutant were internalized normally (81.3%, 90.0%, or 95.3%, respectively), like WT cells (93.3%) (Fig. 3, A, D and E). The pan1ΔABD ent1Δ or pan1ΔABD ent2Δ double mutant also exhibited normal Abp1p patch internalization (82.0% or 86.7%), but the pan1ΔABD sla2ΔTHATCH and pan1-ΔABD ent1ΔACB ent2Δ mutants showed Abp1p patch behaviors distinct from those of the WT cells (Fig. 3, A, D and E). In the pan1ΔABD sla2ΔTHATCH mutant, the efficiency of Abp1p patch internalization was slightly reduced (53.3%) and inward movement from the plasma membrane was irregular, in comparison with the WT cells (Fig. 3, A, D and E). In the pan1ΔABD ent1ΔACB ent2Δ mutant, Abp1p patch dynamics were affected much more severely; the majority of Abp1p patches were not internalized (17.3%), and although changes in their fluorescence intensity were similar to those in WT cells, a longer time was required (Fig. 3, A, D and E). When the Abp1p patch increased and reached maximum fluorescence intensity, the pan1ΔABD ent1ΔACB ent2Δ mutant exhibited transient formation of actin tail-like structures that arose from the Pan1-GFP patch, thereafter gradually shortening and disappearing after Pan1p had done so ( Fig. 3A and Movie S1). These findings suggested that actin polymerization and depolymerization probably occur regularly at endocytic sites but that actin-dependent vesicle internalization is impaired in the pan1ΔABD ent1ΔACB ent2Δ mutant. The scale bars represent 2.5 μm. B, average lifetimes of Abp1-GFP patches ± SD in WT and mutant cells. Data were taken from 1-min movies with a 1-s frame interval. n = 50 patches for each strain. C, synthetic lethality between sla2ΔTHATCH and ent1ΔACB ent2Δ mutant. Tetrad analysis after sporulation of an sla2ΔTHATCH::kan r /SLA2 ent1ΔACB::kan r /ENT1 ent2Δ::kan r /ENT2 diploid. The tetrads were dissected on the YPD plate, and the plate was photographed after 3 days of growth at 25 C. The right panel represents the genotype of each spore. D, the bar graphs represent the percentage of patches internalized into the cytosol in WT or mutant cells. Data show the mean ± SD from at least three experiments, with 50 patches counted for each strain per experiment. Different letters indicate significant difference at p < 0.001, one-way ANOVA with Tukey's post hoc test (B and D). E, quantification of fluorescence intensity (red) and distance from the site of patch formation (blue) as a function of time for patches of Abp1-GFP. Data from ten patches from each strain were averaged using single-color movies of Abp1-GFP. Fluorescent intensity over time was corrected for photobleaching. Tracking of individual cortical Abp1p patches is indicated in right. Abp1-GFP was visualized every 1 s, and patch movement traces were obtained for the entire life of the patches. Green and red dots indicate the first and the last positions, respectively. YPD, yeast extract peptone dextrose.

Distinct requirement of actin-binding domains for endocytic internalization
To clarify the step of endocytic transport that requires the actin-binding activities of Pan1p, Sla2p, and Ent1p, we next examined the effect of deleting each of the actin-binding domains on the endocytic pathway. We first carried out an α-factor internalization assay to examine the effect on endocytic cargo internalization. pan1ΔABD mutants exhibited only a slight defect of 35 S-labeled α-factor internalization (Fig. 4A, red line), whereas the effect in ent1ΔACB ent2Δ or sla2Δ-THATCH mutants was negligible (Fig. 4, A and B, yellow  lines). Combination of the pan1ΔABD and ent1ΔACB ent2Δ mutations resulted in a marked defect (Fig. 4A, purple line), but a combination of the pan1ΔABD mutant with the sla2Δ-THATCH mutant resulted in a moderate defect (Fig. 4B, green line), consistent with the observation that internalization of Abp1p patches was mostly inhibited in the mutant (Fig. 3E).
We next labeled WT and mutant cells with Alexa Fluor 647-α-factor (Alexa-α-factor), a marker of the receptormediated endocytic pathway, and followed the localization at 15 min after α-factor internalization to examine the effect on transport of endocytic cargo from the plasma membrane to the vacuole. The pan1ΔABD mutant showed a slight delay in αfactor transport to the vacuole (PM only, 0.0%; PM + End + Figure 4. Effect of deleting each of the actin-binding domains of Pan1p, Sla2p, and Ent1p on the endocytic pathway. A and B, radiolabeled α-factor internalization assays performed on the indicated strains at 25 C. Each curve represents the average of three independent experiments, and error bars indicate the SD at each time point. Different letters indicate significant difference at p < 0.05, one-way ANOVA with Tukey's post hoc test. C, transport of Alexa Fluor 647-α-factor in pan1ΔABD, pan1ΔABD sla2ΔTHATCH, or pan1ΔABD ent1ΔACB ent2Δ mutant. WT and mutant cells were treated with Alexa-αfactor, and the images were acquired 0 and 15 min after washing out unbound Alexa-α-factor and warming the cell to 25 C. The scale bars represent 2.5 μm. D, quantification of the localization of Alexa-α-factor in WT and mutant cells at 15 min after internalization. The bar graphs represent the percentages of cells exhibiting Alexa-α-factor localized at the endosome and vacuole (red), the plasma membrane, endosome, and vacuole (yellow), or the plasma membrane (blue). Data show the mean ± SD from three experiments, with 50 cells counted for each strain per experiment. Alexa-α-factor, Alexa Fluor 647-α-factor.

Regulation of actin-mediated endocytic vesicle transport
Vac, 25.7%; End + Vac, 74.3%), but the delay was additive when combined with the sla2ΔTHATCH mutant (PM only, 10.0%; PM + End + Vac, 70.0%; End + Vac, 20.0%) (Fig. 4, C and D). Consistent with the results of analysis using 35 Slabeled α-factor, the pan1ΔABD ent1ΔACB ent2Δ mutant showed a much severer defect in transport of Alexa-α-factor from the plasma membrane to the cytosol (PM only, 97.3%; PM + End + Vac, 2.7%; End + Vac, 0.0%) relative to the pan1ΔABD sla2ΔTHATCH mutant (Fig. 4, C and D). These results clearly indicated that the pan1ΔABD ent1ΔACB ent2Δ mutant has a defect at the internalization step in the endocytic pathway.
The actin-binding domains of Pan1p and Ent1/2p are required for recruitment of actin cables to endocytic sites We previously showed that over 80% of endocytic vesicles were internalized along actin cables at the internalization step of endocytosis (28). Therefore, we next examined actin cable dynamics in the pan1ΔABD sla2ΔTHATCH or pan1ΔABD ent1ΔACB ent2Δ mutant. While the pan1ΔABD sla2Δ-THATCH mutant exhibited a near-normal polarized actin cable structure, the pan1ΔABD ent1ΔACB ent2Δ mutant showed aberrant actin cable dynamics ( Fig. 5A and Movie S2). The pan1ΔABD sla2ΔTHATCH mutant exhibited a slightly increased Pan1p patch lifetime (35.7 s) relative to WT cells (29.7 s) (Fig. 5B). In contrast, the pan1ΔABD ent1ΔACB ent2Δ mutant exhibited remarkably increased Pan1p lifetime (80.6 s). In the pan1ΔABD ent1ΔACB ent2Δ mutant, particularly, the fluorescence intensity of actin cables, labeled by Abp140-3GFP, was decreased, and instead actin tail-like structures, which were also labeled by Abp1-GFP (Fig. 3A), were more clearly evident (Fig. 5, A and C and Movie S2). Because Abp140p is reported to localize to both actin patches and cables (29), this observation suggests that localization of Abp140-3GFP at actin patches is increased in the mutant because of decrease of actin cables. Most of the actin tail-like structures attached to Pan1-mCherry were transiently assembled and disassembled on the plasma membrane, as described above, but recruitment of actin cables to Pan1-mCherrylabeled endocytic sites and the following Pan1p patch internalization were significantly decreased (13.3%) (Fig. 5, D and  E). In the pan1ΔABD sla2ΔTHATCH mutant, Pan1p patches internalized linearly along actin cables were slightly decreased (59.3%), whereas in WT cells, 84.7% of them, even in a single focal plane image, were internalized along actin cables (Fig. 5, D and E and Movie S3). These observations suggested that the actin-binding activities of these three proteins, especially Pan1p and Ent1p/2p, are necessary for the recruitment of actin cables to endocytic vesicles.

Discussion
In a previous study, we investigated the timing of actin cable recruitment to sites of endocytosis and demonstrated that 48% of actin cables appear at such sites before actin patch formation (28), implying CCV budding, whereas 39% of the cables appear thereafter. This observation suggested that actin cables are recruited to endocytic sites independently of CCV budding and that an endocytic protein, which is recruited to the site before vesicle formation and has the ability to bind to the actin filament, mediates the interaction. In yeast endocytosis, early coat proteins, including clathrin and the AP-2 complex, are recruited to the endocytic site first, and then mid-coat proteins, such as Sla2p and Ent1/2p, are recruited around 10 to 20 s before CCV formation (Fig. 6). The Pan1p complex, including Sla1p and End3p, is classified as late coat protein and recruited to the endocytic site slightly after Sla2p and Ent1/2p (4). Considering that the interaction between the endocytic vesicle and the actin cable should be transient and controllable, Pan1p and Ent1/2p are ideal candidates for mediating this interaction. Pan1p can bind to actin filaments directly, and the binding is regulated by phosphorylation through the Ark1p/Prk1p kinases (19). Ent1p and Ent2p are also targets of Ark1/Prk1 kinases (16), and a phosphomimicking mutant form of the Ent1p ACB domain severely reduces the actin-binding activity (22). As Prk1p and Ark1p are recruited to the endocytic site at 12 s after Abp1p arrival (12), it seems reasonable to assume that-before endocytic vesicle budding-Pan1p and Ent1p/2p recruit actin cables to the endocytic site and are then phosphorylated by Ark1p/ Prk1p kinases after vesicle internalization, thereafter dissociating from the actin cable and resulting in separation of the latter from the endocytic vesicle. Endocytic vesicle movement occurs at the same velocity and direction as that of actin cables Regulation of actin-mediated endocytic vesicle transport (11), and this suggests that endocytic vesicles remain fixed on the actin cables and move as a result of actin cable flow. The endocytic vesicle eventually fuses with the early/recycling endosome after these events (Fig. 6).
It has been reported that Pan1p, Sla2p, and Ent1/2p are able to bind actin filaments, but that strains carrying individual deletions of these actin-binding domains exhibit near-normal endocytic dynamics, although the Pan1ΔABD mutation suppresses the formation of actin cable aggregates in the pant1-18TA mutant (18,19,22,26). Previous studies have demonstrated that the C-terminal actin-binding domains of Sla2p and Ent1p/2p interact redundantly with actin filaments and that deletion of both actin-binding domains (sla2Δ-THATCH ent1ΔACB) causes a specific endocytic block referred to as the "uncoupling phenotype," which is also observed in the sla2Δ mutant and suggests that the actin cytoskeleton is not coupled to the membrane efficiently (22,27). The N-terminal lipid-binding domains of Sla2p and Ent1p were shown to coassemble in a PI(4,5)P 2 -dependent manner and form an organized lattice structure, which is required for CCV formation (30). These observations, taken together with our finding that the sla2ΔTHATCH ent1ΔACB ent2Δ triple mutant is nonviable, suggest that the actin-binding domains of Sla2p and Ent1/2p are essential for CCV formation.
In contrast to the sla2ΔTHATCH ent1ΔACB ent2Δ mutant, the pan1ΔABD sla2ΔTHATCH or pan1ΔABD ent1ΔABD ent2Δ mutant was viable, and both exhibited distinct actin and endocytic phenotypes. While the pan1ΔABD ent1ΔACB ent2Δ mutant exhibited severely defective endocytosis, the pan1-ΔABD sla2ΔTHATCH mutant was only mildly defective in terms of both endocytosis and the actin cytoskeleton, suggesting a distinct role of the actin-binding domains of Pan1p and Sla2p. This idea is also supported by the previous observation that Sla2p specifically inhibits the ability of Pan1p to bind to the actin filament (31). Sla2p binds directly to Pan1p via its coiled-coil domain, and a mutant with partial loss of function of pan1 shows suppression of the endocytic and actin phenotypes observed in the sla2 mutant lacking the coiled-coil domain (31). Thus, Pan1p might bind to the actin filament after dissociating from Sla2p.
The actin structure observed in the pan1ΔABD ent1ΔACB ent2Δ mutant probably does not correspond to the "uncoupling phenotype" observed in the sla2Δ or sla2ΔTHATCH ent1ΔACB mutant because endocytic coat assembly and disassembly accompanied by actin polymerization and depolymerization occur regularly. Activation of the Arp2/3 complex by nucleation-promoting factors such as Las17p (yeast WASP) and Myo3p/5p (yeast type I myosins) plays an important role in actin patch formation, and these activities are regulated by the endocytic coat protein complex including Pan1p, Sla1p, and End3p (32,33). In contrast, disassembly of the actin patch is regulated by phosphorylation of coat proteins by Ark1p/Prk1p kinases (13). In the pan1ΔABD ent1ΔACB ent2Δ mutant, actin patch internalization was severely blocked, whereas the lifetime of actin patch was increased only slightly, suggesting that the vesicle budding step mediated by the interaction between endocytic coat proteins and actin regulators is normal; however, the vesicle scission or internalization step is impaired. The interaction between the endocytic vesicle and the actin cable may play a role in these steps, as well as transport to the endosome.
Our results suggest the important role of Pan1p and Ent1/ 2p in the interaction between the endocytic vesicle and the actin cable, but the role of clathrin in the interaction remains unclear. Previous studies have demonstrated that deletion of the CHC1 or CLC1 gene, encoding yeast clathrin heavy chain or light chain, dramatically reduces the inward movement of coat proteins that implies the formation of endocytic vesicle, whereas internalization of actin patches is not inhibited (34,35). It has been also reported that the timing of actin cable recruitment to actin patches in both chc1Δ and clc1Δ mutants is similar to that in WT cells (28). These observations suggest that clathrin is required for the link between the coat proteins and actin for endocytic vesicle formation, rather than for the fixing of endocytic vesicle to the actin cable.

Experimental procedures
Yeast strains, growth conditions, and plasmids The yeast strains used in this study are listed in the strain list (Table S1). All strains were grown in yeast extract peptone dextrose (YPD) or synthetic medium (SM) supplemented with 2% glucose and appropriate amino acids. C-terminal GFP or mCherry tagging of proteins was performed by PCR-based homologous recombination using pFA6a-GFP(S65T)-HIS3 (36) or pFA6a-mCherry-URA3 (37), respectively, as a template. The pan1-18TA and sla2ΔTHATCH mutants were generated as described previously (18,26). The pan1ΔABD or ent1ΔACB mutant was generated by PCR-based homologous recombination. To generate the pan1ΔABD mutant, the mCherry-URA3 fragment containing 50-bp flanking homologous regions (nt 2513-2562 and the 3 0 noncoding region of PAN1 gene) was amplified using primers (JT2676 and JK2) and pFA6a-mCherry-URA3 as a template and transformed into WT cells. The pan1ΔABD mutants were selected on the synthetic complete plate lacking uracil and confirmed by PCR using JT3 and JT2769. To generate the ent1ΔACB mutant, the 3HA-HIS3 fragment containing 50-bp flanking homologous regions (nt 965-1014 and the 3 0 noncoding region of ENT1 gene) was amplified using primers (JT2874 and JT2875) and pFA6a-3HA-HIS3 as a template and transformed into WT cells. The ent1ΔACB mutants were selected on the synthetic complete plate lacking histidine and confirmed by PCR using JT2876 and JT763.

Fluorescence microscopy
Fluorescence microscopy was performed using an Olympus IX81 microscope equipped with a 100×/NA 1.40 (Olympus) objective and Orca-AG cooled CCD camera (Hamamatsu), using MetaMorph software (Universal Imaging). Simultaneous imaging of red and green fluorescence was performed using an Olympus IX81 microscope, described above, and an image splitter (Dual-View; Optical Insights) that divided the red and green components of the images with a 565-nm dichroic mirror and passed the red component through a 630/50 nm filter and the green component through a 530/30 nm filter. These split signals were taken simultaneously with one CCD camera, described above.
Fluorescent labeling of the α-factor and endocytosis assays Fluorescent labeling of the α-factor was performed as described previously (10). For endocytosis assays, cells were grown to an A 600 of 0.5 in 0.5-ml YPD, briefly centrifuged, and resuspended in 20-μl SM with 5 μM Alexa-α-factor. After incubation on ice for 2 h, the cells were washed with ice-cold SM. Internalization was initiated by addition of SM containing 4% glucose and amino acids at 25 C. 35 S-labeled α-factor internalization assay Preparation and internalization of 35 S-labeled α-factor was performed as described previously (19). Briefly, cells were grown to an OD600 of 0.3 in 50-ml YPD, briefly centrifuged, and resuspended in 4-ml YPD containing 1% (w/v) bovine serum albumin, 50 mM KH 2 PO 4 , pH 6.0, and 20 μg/ml uracil, adenine, and histidine. After adding 35 S-labeled α-factor, cell aliquots were withdrawn at various time points and subjected to a wash in pH 1 buffer to remove surface-bound α-factors so that internal α-factors could be measured, or in pH 6 buffer to determine the total (internal and bound) α-factors. The amount of cell-associated radioactivity after each wash was determined by scintillation counting. Each experiment was performed at least three times.

Data availability
All data generated and analyzed during this study are contained within the article and can be shared upon request (jtosiscb@rs.tus.ac.jp).  Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.