Srv2/CAP is required for polarized actin cable assembly and patch internalization during clathrin-mediated endocytosis

The coordinated regulation of the assembly and disassembly of actin filaments is required for various cellular functions, including the formation and internalization of endocytic vesicles (Goode et al., 2015; Kaksonen et al., 2006; Perrais and Merrifield, 2005). In yeast, the actin cytoskeleton exists in three distinct types of structure: cortical patches, cables and the contractile ring (Moseley and Goode, 2006). The actin cortical patches are a dense dendritic network of actin filaments that are assembled at endocytic sites (Goode et al., 2015; Kaksonen et al., 2006; Winter et al., 1997). Actin cables are bundles of actin filaments that generally align with the long axis of the cell, and serve as tracks for vesicle transport and organelle segregation (Bretscher, 2003; Yang and Pon, 2002). A previous study that used three-dimensional imaging has demonstrated that actin cables attach both at their ends and laterally to cortical patches (Amberg, 1998). We and other groups have also shown previously that actin patches (clathrin-coated vesicles, CCVs) associate with and move toward early endosomes along actin cables after being internalized into the cytosol (Huckaba et al., 2004; Toshima et al., 2006), suggesting that actin cables might function as tracks to transport CCVs to early endosomes. Although evidence increasingly indicates a functional connection between actin patches and cables, the timing of their interaction and the molecular machinery tethering CCVs to actin cables are still poorly understood.

Actin disassembly during endocytosis in yeast is driven by the coordinated activities of multiple actin-binding proteins, such as Cof1p (yeast cofilin), Crn1p (yeast coronin), Aip1p (actin-interacting protein 1), and Srv2/cyclase-associated protein (CAP) (Goode et al., 2015). All of these proteins are components of the actin cytoskeleton that are widely conserved among eukaryotes. Cof1p binds to the sides of actin filaments and severs the filaments with the help of Aip1p or Crn1p (Jansen et al., 2015). Activities of Srv2p with Cof1p also promote actin filament disassembly in vitro (Balcer et al., 2003; Moriyama and Yahara, 2002). The timing of the localization of Srv2p to actin patches during endocytosis has not been investigated. Cof1p, Crn1p and Aip1p appear at actin patches about 5 s later than Abp1p, and are lost from the patches later than Abp1p (Lin et al., 2010; Okreglak and Drubin, 2007). A study by Lin et al. has revealed that Cof1p assembles at a time point when the initial movement of the cortical patches had already begun and that the actin patches keep moving as they accumulate Cof1p (Lin et al., 2010). Thus, Cof1p and its interacting proteins seem to function in the later stages of endocytic internalization.

Srv2p, also called cyclase-associated protein (CAP), is a highly conserved actin-binding protein that is required for normal actin organization in organisms ranging from yeast to mammals (Ono, 2013). Srv2p has several motifs and domains, including an N-terminal oligomerization domain, a dimeric helical folded domain (HFD) and a Wasp homology 2 (WH2) domain, which are sandwiched between two poly-proline (P1 and P2) motifs, and a dimeric folded domain termed the CARP domain, entirely comprising β-sheets, at the C-terminus (Ono, 2013). The function of Srv2p in actin turnover has been established in several excellent biochemical studies. The C-terminal region of Srv2p has a role in G-actin nucleotide exchange, and also in the recycling of cofilin, a ubiquitously expressed actin depolymerization factor. These functions are accomplished through a high-affinity association of the Srv2p C-terminus with ADP-G-actin, which releases the actin from cofilin (Balcer et al., 2003; Balderhaar et al., 2013; Chaudhry et al., 2010; Mattila et al., 2004; Moriyama and Yahara, 2002). By contrast, the N-terminal oligomerization and the HFD domains can catalyze cofilin-mediated severing of actin filaments (Chaudhry et al., 2013; Jansen et al., 2014). In addition, several proteins have been reported to selectively bind to either the P1 or P2 regions of Srv2p; profilin, a nucleotide exchange factor for actin, binds to P1 (Bertling et al., 2007), and Abp1p, an actin-binding protein, binds to P2 (Freeman et al., 1996; Lila and Drubin, 1997). Although these studies provide a potential framework for understanding how Srv2p and its binding proteins regulate the assembly and disassembly of actin filaments in vitro, the physiological role of Srv2p in the formation of actin patches and/or cables has still remained unclear.

In the present study, we screened yeast strains carrying single-gene deletion mutations for defects in actin cable formation and dynamics, and identified SRV2 as such a gene. We also demonstrated that the srv2Δ mutant exhibits severely defective actin patch disassembly and endocytic internalization. Furthermore, we analyzed several domain-deletion and point mutants of Srv2p, and show the functional regions of Srv2p that are required for the proper assembly and disassembly of actin patches and cables. These data indicate that there are multiple in vivo roles for Srv2p in the regulation of the actin cytoskeleton that are required for normal endocytosis.

Previous studies have demonstrated that internalization of actin patches is mediated by polarized actin cables (Huckaba et al., 2004; Kaksonen et al., 2003; Toshima et al., 2006). However, how actin patches and cables cooperate to drive the formation and internalization of endocytic vesicles has not been clarified. To help us study these processes, we characterized Abp140p that had been fused with three tandem copies of GFP (3GFP) to confirm that Abp140–3GFP can serve as a marker of both actin patches and actin cables. Consistent with a previous observation, most Apb140p-labeled actin patches (>98%) were found to colocalize with patches that were labeled by Abp1–mCherry, a widely used actin patch marker (Fig. S1A and Movie 1) (Toshima et al., 2006). We also found that the timing of Abp140p recruitment to cortical patches, and the lifetimes and dynamics of Abp140p patches, were almost the same as those of Abp1p patches (Fig. S1B–E). From these results, we concluded that Abp140–3GFP can serve as a marker for both actin cables and patches. We then sought to identify the proteins required for actin cable assembly at endocytic sites. We selected 19 genes that have been reported to be related to actin patch assembly in yeast (Engqvist-Goldstein and Drubin, 2003; Moseley and Goode, 2006) and expressed Abp140–3GFP as a marker for actin cables (Yang and Pon, 2002) in cells that lacked each of the individual genes (Fig. 1A). Among the 19 mutants examined, five mutants – bni1-12 bnr1Δ, cap1Δ, cap2Δ, sac6Δ and srv2Δ – showed prominent localization of Abp140–3GFP at actin patches (Fig. 1A; Movie 2). Quantitative analysis revealed that the levels of actin patch fluorescence in these five mutants were remarkably increased (>1.5 fold), relative to that in wild-type cells, in all of these mutants (Fig. 1B). In contrast, the fluorescence intensity of actin cables was decreased in the bni1-12 bnr1Δ, cap1Δ, cap2Δ and sac6Δ mutants (Fig. 1A,C). Similar phenotypes have been reported previously for cap1Δ, cap2Δ and sac6Δ mutants (Adams et al., 1991; Amatruda and Cooper, 1992; Kim et al., 2004). The yeast formin mutant (bni1-12 bnr1Δ) displayed a decreased intensity of actin cable fluorescence, as described previously (Evangelista et al., 2002; Sagot et al., 2002). Aberrant actin structures in the srv2Δ mutant have been reported in previous studies (Chaudhry et al., 2013,, 2010). Consistent with previous studies, the srv2Δ mutant showed highly depolarized actin patches, and disorganized and partially diminished actin cables (Fig. 1A). In the srv2Δ mutant, the fluorescence intensity of actin patches was markedly increased (Fig. 1B), suggesting that excessive actin polymerization might occur at cortical patches. In contrast, the fluorescence intensity of actin cables was not significantly altered, although many of the actin cables were fragmented (Fig. 1A,C). Interestingly, the dynamics of the fragmented short actin cables were quite unique and never observed in other mutants (Fig. 1A; Movie 2).

As shown previously, over 85% of actin patches, even in single focal plane images, were found to associate with actin cables during internalization (see Fig. 2I). However, as the precise timing of actin cable recruitment to cortical patches has not been determined, we next investigated the spatio-temporal dynamics of actin patches and actin cables at endocytic sites. Live-cell imaging using cells expressing Abp140–3GFP showed that some actin cables were recruited to cortical patches before actin patch formation, whereas others were recruited thereafter (Fig. S2A,B; Movie 3). Quantitative analysis to categorize actin cable recruitment as occurring before, after or simultaneously with actin patch formation revealed that approximately 39% of actin cables were recruited after actin patch formation, whereas around 48% were recruited beforehand in wild-type cells (Fig. S2C). Similar analysis was performed for the deletion mutants, except for sla2Δ in which severely defective actin patch formation has been observed (Kaksonen et al., 2003). The proportion of actin cables recruited after actin patch formation was markedly increased (>60%) in the bni1-12 bnr1Δ, cap1Δ, cap2Δ and srv2Δ mutants (Fig. S2C). As shown in Fig. 1B, all of these mutants exhibited increased intensity of actin patch fluorescence. Among these mutants, srv2Δ exhibited the most intriguing phenotypic change in actin cable dynamics, and therefore became the mutant that we decided to focus our analysis on.

Srv2p has been demonstrated to be part of a multimeric complex that recycles actin monomers from ADF and cofilin, making them available for new rounds of filament assembly (Goode et al., 2015; Ono, 2013). In wild-type cells, actin cables showed a polarized alignment, extending from the bud neck towards the opposite tip, parallel to the longitudinal axis of the cell (Fig. 2A). In contrast, in the srv2Δ mutant, these polarized actin cables were observed much less frequently than in wild-type cells (7.2% and 92.5%, respectively) (Fig. 2B), and fragmented short actin cables oriented in random directions were more evident (Fig. 2A, arrowheads). Quantitative analysis of actin cable lengths revealed that the number of short actin cables (<2 µm) was significantly increased in srv2Δ cells relative to wild-type cells (Fig. 2C,D). Short actin cables move faster than long ones, yet the velocities of actin cables were almost the same between wild-type and srv2Δ cells when cables of the same length were compared to one another (Fig. 2D). We next examined the dynamics of actin cables at cortical patches in the srv2Δ mutant. In wild-type cells, we observed that actin patches invariably associate with actin cables and are then lost (Fig. 2E), as reported previously (Huckaba et al., 2004; Toshima et al., 2006). In the srv2Δ mutant, however, we found that actin cables were directed to pre-existing actin patches, became associated with them and then moved away from them (Fig. 2F, red arrowheads; Movie 4). In this mutant, we also found that multiple actin cables moved past a single actin patch (Fig. 2G). Such events were observed frequently only in srv2Δ cells (∼85.5%) and were rare in wild-type cells (<5%) (Fig. 2G). Additionally, we detected one actin cable that sequentially moved past two different actin patches located close to each other (Fig. 2F, yellow arrowheads, 2H, upper panels; Movie 4). Interestingly, actin cables occasionally exhibited a curved structure when simultaneously moving past two patches, indicating that, in the srv2Δ, mutant actin cables can attach to multiple actin patches at the same time and that one attachment is fixed and the other isn't, resulting in the curve. Although internalization events occurred less frequently in the srv2Δ mutant, over 80% of actin patches were internalized along actin cables at the internalization step of endocytosis, similar to the situation in wild-type cells (Fig. 2I). These observations suggest that, in the srv2Δ mutant, short actin cables are regularly recruited to sites of actin patch assembly but that tethering of endocytic vesicles on actin cables is defective.

Previous studies have indicated that deletion of the SRV2 gene does not result in changes in endocytosis-related parameters, such as the internalization of radiolabeled α-factor and the lifetimes of endocytic proteins (Kaksonen et al., 2005; Wesp et al., 1997). To confirm this, we next examined the effect of the deletion of SRV2 on endocytosis by assessing the internalization and transport of Alexa-Fluor-594-labeled α-factor (A594-α-factor) (Toshima et al., 2014). We examined the localization of A594-α-factor at several time points after its addition to wild-type and srv2Δ cells. In wild-type cells, the majority of A594-α-factor had been transported to the vacuole 15 min after addition (Fig. 3A). In contrast, an obvious delay of A594-α-factor internalization was observed in the srv2Δ mutant (Fig. 3A). Quantitative analyses revealed that the fluorescence intensity of A594-α-factor decreased to ∼20% in wild-type cells at 5 min after α-factor addition, and the fluorescence was mostly lost after 15 min, whereas approximately 61% and 26% of A594-α-factor remained at the plasma membrane in srv2Δ cells after 15 and 30 min, respectively (Fig. 3B). We further examined the effect on endocytosis by assessing the internalization of ³⁵S-labeled α-factor. In contrast to a previous observation (Wesp et al., 1997), we found that srv2Δ cells exhibited a marked defect in ³⁵S-labeled α-factor internalization (Fig. 3C). We also examined the effect of SRV2 deletion on recycling of the plasma membrane vesicle (v-)SNARE Snc1p, a plasma-membrane-recycling marker in yeast (Gurunathan et al., 2000). Consistent with A594-α-factor internalization, we found that recycling of Snc1p was impaired at the stage of endocytic internalization, whereas transport of Snc1p to the plasma membrane seemed to be normal (Fig. 3D). Furthermore, we found that binding of ³⁵S-labeled α-factor to the cell surface was almost the same between wild-type and srv2Δ cells (Fig. 3E). These results suggest that the expression level of the α-factor receptor on the cell surface does not differ between wild-type and srv2Δ cells, and that srv2Δ cells have a defect in the internalization step of endocytosis.

We next examined whether the dynamics of the late coat module and actin patch were affected in the srv2Δ mutant. In contrast to previous studies, the lifetime of Sla1–GFP, a marker of the late clathrin coat, was significantly increased in the srv2Δ mutant (141±66 s; mean±s.d.), compared to that of wild-type cells (29±6 s) (Fig. 3F,G). Two-color simultaneous imaging revealed that Sla1–GFP was internalized at the same time as Abp1–mCherry (Fig. 3F; Movie 5). Similar to Sla1p, the lifetime of Abp1p was considerably increased in the srv2Δ mutant, relative to that in wild-type cells (43±11 and 16±5 s, respectively) (Fig. 3F,H). Quantification of the fluorescence intensity for individual Abp1p patches showed that the time required to reach the maximum fluorescence intensity and the time required to decrease to the minimum intensity were both markedly increased in the mutant (Fig. 3I). Furthermore, the tracking of individual patches showed that the distance and time of the migration that patches engaged in after being released from the plasma membrane were increased in the srv2Δ mutant (Fig. 3J,K). These results suggest that the srv2Δ mutants have a defect in both the assembly and disassembly of actin patches, resulting in a severe defect in endocytic internalization.

To investigate in more detail the role of Srv2p on actin-mediated endocytosis, we created Srv2 mutants in which different portions of the protein were deleted, either the HFD domain (ΔHFD), the WH2 and two proline-rich motifs (ΔPWP), or the C-terminal CARP domain (ΔCARP). Each mutation was integrated into the endogenous SRV2 locus (Fig. 4A), replacing the original gene. We also did this with the full-length SRV2 gene, putting it back into the endogenous locus in the srv2Δ mutant to confirm the functionality of this approach. These three mutant cell lines showed fewer polarized actin cables and increased Abp1p lifetimes, compared to wild-type cells (Fig. 4B–D). In contrast, cells into which the full-length SRV2 gene had been integrated exhibited normal actin structures, similar to wild-type cells (Fig. 4B–D). Quantitative analyses revealed that srv2ΔCARP cells mostly, and srv2ΔHFD and srv2ΔPWP cells partially lacked polarized actin patches and cables, corresponding to the growth phenotype (Fig. 4C; Fig. S3A). The lifetime of Abp1p patches was considerably increased in srv2ΔCARP cells (∼35±12 s; mean±s.d.), whereas srv2ΔHFD and srv2ΔPWP cells exhibited a moderate increase (∼28±7 and 24±5 s, respectively) (Fig. 4D). These results suggest that the CARP domain has a more important role in endocytic vesicle formation. Particle-tracking analysis revealed that in srv2ΔHFD and srv2ΔPWP cells, relative to wild-type cells, the time taken for actin patches to reach maximum fluorescence intensity was not significantly altered, but the time required for the fluorescence intensity to decrease was longer (Fig. 4E). This suggests that deletion of the HFD and PWP domains primarily affects disassembly of actin patches. In contrast, in srv2ΔCARP cells, the times required for the fluorescence intensity to both increase and decrease were longer (Fig. 4E), indicating a requirement of the CARP domain for both assembly and disassembly of actin patches.

We wished to determine whether any of the changes in actin behavior that we observed in the Srv2p domain mutants were due to changes in the localization of the protein. A previous indirect immunofluorescence study has demonstrated that Srv2p colocalizes with cortical actin patches and that binding of poly-proline motifs to Abp1p are required for this localization (Freeman et al., 1996; Lila and Drubin, 1997; Yu et al., 1999). To perform an in-depth characterization of the localization and dynamics of Srv2p, we tagged Srv2p and its domain-deletion mutants with GFP, and examined their localization in living cells. Western blotting analysis using an antibody against GFP confirmed that each gene product was expressed with the correct molecular mass in each mutant cell (Fig. 5A). Live-cell imaging revealed that Srv2–GFP patches formed at the cell cortex with a lifetime of 11±4 s (mean±s.d.), culminating in inward movement (Fig. 5B,C; Movie 6). Srv2p patches had similar dynamics to actin patches, initially appearing as a stationary patch and then undergoing rapid inward movement away from the original position before being lost in the cytoplasm (Fig. 5D). We also examined the localization of Srv2p domain-deletion mutants and found that deletion of the PWP domain significantly decreased Srv2p localization at cortical patches (Fig. 5B,F), consistent with previous observations (Lila and Drubin, 1997; Yu et al., 1999). The localization of Srv2ΔHFD and Srv2ΔCARP at cortical patches slightly decreased compared to that of full-length Srv2p, whereas the mean lifetimes were increased (24±1 s and 27±2 s, respectively) (Fig. 5B,C,F). Unexpectedly, the fluorescence intensity of the cortical patches of GFP-tagged Srv2ΔHFD protein was increased about 1.6-fold compared to that of full-length Srv2p (Fig. 5B,G). Interestingly, the fluorescence intensity of Abp1p patches was similarly increased in srv2ΔHFD cells (Fig. 5G), suggesting that the loss of the F-actin-disassembly-promoting activity of HFD probably causes an increase of F-actin in the patches. Because it has been reported that Abp1p has a role in the localization of Srv2p (Freeman et al., 1996), we next examined whether the localization of each deletion mutant depends on the presence of Abp1p. The Srv2p cortical localization was affected by deleting the ABP1 gene, consistent with the findings of an earlier report, but the average number of Srv2ΔPWP patches in abp1Δ cells was almost the same as that of Srv2p patches in the srv2ΔPWP cells (Fig. 5E,F). Intriguingly, we found that deletion of the HFD or CARP domain of Srv2p further decreased the cortical localization of Srv2p when expressed in abp1Δ cells (Fig. 5E,F). These observations suggest that Srv2p localization at cortical patches is dependent on other interactions that are mediated by the HFD and CARP domains, in addition to interactions with Abp1p through the PWP region.

To determine the timing of Srv2p recruitment to actin patches, we performed simultaneous two-color imaging of Srv2–GFP and Abp1–mCherry (Movie 6). In wild-type cells, Srv2–GFP was recruited to cortical patches several seconds after the initial assembly of the actin patch (Fig. 5H,I, blue arrowheads) and reached its maximum concentration with the same timing as Abp1p (Fig. 5H,I, red arrowheads). Deletion of the CARP domain barely affected the timing of Srv2p recruitment (Fig. 5I, blue arrowheads), suggesting that the CARP domain is unnecessary for Srv2p recruitment. Interestingly, Srv2p that lacked the HFD domain was recruited with almost the same timing as Abp1p, whereas deletion of the PWP region significantly delayed the recruitment of Srv2p to cortical patches (Fig. 5I, blue arrowheads). The fluorescence intensity of Srv2ΔHFD–GFP and Srv2ΔCARP–GFP peaked at almost the same time point as Abp1–mCherry, whereas Srv2ΔPWP–GFP showed a marked delay (Fig. 5I, red arrowheads).

We next verified the phenotypes of the domain-deletion mutants, by comparing them to the phenotypes of several point mutants (Fig. 6A) that have been previously generated and biochemically characterized (Bertling et al., 2007; Chaudhry et al., 2010; Mattila et al., 2004; Quintero-Monzon et al., 2009). Srv2-91 contains four alanine residue substitutions at conserved solvent-exposed residues (RILKE mutated to AIAAA) in the HFD domain and has defects in binding to the cofilin–actin complex, and thus defects in cofilin-mediated actin turnover (Quintero-Monzon et al., 2009). Srv2-104 contains replacements of two charged residues with two alanine residues (ENYE mutated to ANYA) in the CARP domain and shows reduced affinity for ADP-actin in vitro (Mattila et al., 2004). Each mutant was integrated into the endogenous SRV2 locus, replacing the original gene. These point mutants displayed no apparent defects in growth at 37°C (Fig. 6B). Similar to the srv2ΔHFD mutant (Figs 4D and 5C), the srv2-91 mutant showed increased lifetimes of Abp1p and Srv2p (∼30±11, and 28±12 s; mean±s.d.) (Fig. 6C). GFP-fused Srv2-91 also exhibited a higher fluorescence intensity and was also recruited to cortical patches with almost the same timing as Abp1p (Fig. 6D,F,G), indicating that the phenotypes in these assays caused by deletion of the HFD domain (Fig. 5F,I) are the result of defective binding to the cofilin–actin complex. The srv2-104 mutant also showed similar defects to the srv2ΔCARP mutant (compare Fig. 6C,G to Figs 4D and 5C,I), suggesting that the phenotypes observed in the srv2ΔCARP mutant are caused by defective binding to actin monomers.

srv2-201, srv2-98 and srv2-203 contain mutations in the P1 (PPP mutated to AAA), WH2 (LKKV mutated to AAAA) and P2 regions (PPP mutated to AAA), respectively (Fig. 6A), and show major defects in binding to profilin, ATP- and ADP-actin, and Abp1p, respectively (Bertling et al., 2007; Chaudhry et al., 2010). Among these, the srv2-98 mutant exhibited increased lifetimes of Abp1p and Srv2p (∼29±11 and 25±10 s, respectively), whereas srv2-201 and srv2-203 mutants had no obvious defect in the lifetime of Abp1p, compared to that in wild-type cells (Fig. 6C). In contrast, the srv2-203 mutant exhibited significantly decreased Srv2p localization at cortical patches, whereas srv2-201 and srv2-98 exhibited only modest effects on the localization (Fig. 6E,F). Additionally, similar to Srv2ΔPWP, the timing of Srv2p recruitment to actin patches was markedly delayed in the srv2-203 mutant (Fig. 6G). These results suggest that the defective localization of Srv2p observed in the srv2ΔPWP mutant is caused by defective binding to Abp1p through the P2 region.

To further analyze the mechanism of Srv2p recruitment to actin patches, we examined Srv2p localization in sla2Δ cells, in which endocytic internalization is defective and actin comet tails are stably assembled from endocytic sites in the cell cortex (Kaksonen et al., 2003). Interestingly, we found that Srv2–GFP exhibited a localization that was distinct from the Abp1–mCherry-labeled actin structures in sla2Δ cells (Fig. 7A,B, upper panels; Movie 7). The localization of Srv2p and Abp1p was quantified as a function of fluorescence intensity along the actin comet tail. As shown in Fig. 7C, the peak of Srv2–GFP intensity was located at the cytoplasmic edges of actin comet tails. Because it has been demonstrated previously that actin subunits actively treadmill through the actin comet tail from the cortex toward the center of the cell (Kaksonen et al., 2003), this observation suggests that Srv2p might bind to and function around the pointed ends of actin filaments, which have abundant ADP-actin. We also examined the effect of each domain-deletion mutant on Srv2p localization in sla2Δ cells. A previous study has shown that cells in which both the SLA2 and SRV2 genes have been deleted were inviable (Lila and Drubin, 1997). We found that deletion of the CARP domain of Srv2p in the absence of the SLA2 gene was also lethal, whereas deletion of the HFD domain or PWP region was non-lethal (Fig. S3B,C). When the PWP region was deleted in sla2Δ cells, Srv2ΔPWP–GFP exhibited a localization similar to that of full-length Srv2–GFP, namely at the edges of actin comet tails, the lengths of which were slightly increased (Fig. 7A–C, middle panels and graph; Movie 7). In contrast, deletion of the HFD domain dramatically changed the localization of Srv2p so that it became associated with the actin structures that were labeled entirely with Abp1–mCherry (Fig. 7A–C, lower panels and graph; Movie 7). This observation is consistent with the reported interactions of the HFD domain with F-actin (Jansen et al., 2014). These results suggest that localization of Srv2p to cortical patches through the HFD domain might depend on the nucleotide-bound state of actin, or a protein that has a preferential affinity for ADP-actin.

Srv2p localization at the edges of actin comet tails in sla2Δ cells is quite similar to that of Cof1p, yeast cofilin (Okreglak and Drubin, 2007), which exhibits ADP-actin-dependent localization to cortical patches, prompting us to examine the effect of Srv2p domain deletions on Cof1p localization. As expected, Cof1–GFP colocalized with Srv2–mCherry at cortical patches in wild-type cells (Fig. 7D). We also found that Cof1p localized to cortical patches in srv2Δ cells as well as in wild-type cells (Fig. 7E), indicating that Cof1p can be recruited to cortical patches independently of Srv2p. Consistent with this idea, deletion of the HFD domain of Srv2p did not affect the localization of Cof1p at the cortical patches (Fig. 7E). Simultaneous two-color imaging revealed that Cof1p is recruited to cortical patches in a manner similar to Srv2p after actin patch assembly in wild-type cells (Fig. 7F). In srv2Δ cells, Cof1p was also recruited to cortical patches after actin patch assembly, although its appearance there was substantially delayed (Fig. 7F). Interestingly, in srv2ΔHFD cells, Cof1p and Srv2p were recruited to cortical patches with distinct timings – Cof1p was recruited after Abp1p with a little delay (Fig. 7G), whereas Srv2p was recruited simultaneously with Abp1p (Fig. 5I). This observation suggests that localization of Srv2p is not dependent on Cof1p in srv2ΔHFD cells.

Although the function of Srv2p in actin turnover and the physiological importance of Srv2p in maintaining proper actin organization has been established by several previous studies (Balcer et al., 2003; Chaudhry et al., 2013,, 2010; Dodatko et al., 2004; Johnston et al., 2015; Ksiazek et al., 2003; Mattila et al., 2004; Quintero-Monzon et al., 2009), the physiological role of Srv2p in endocytic internalization has remained ill defined. We investigated the phenotypes of several srv2 mutants in living yeast cells and obtained in vivo evidence for a requirement of Srv2p in actin cable assembly and endocytic internalization. On the basis of the data presented above and in previous studies, we propose a mechanism of actin patch and cable assembly that is regulated by Srv2p (Fig. 8). In wild-type cells, Srv2p is recruited to cortical patches, through an association with Abp1p (Freeman et al., 1996; Lila and Drubin, 1997; Yu et al., 1999) and Cof1p through the PWP and HFD domains, respectively. The C-terminal CARP domain of Srv2p has a role in recycling ADP-actin to ATP-actin (Balcer et al., 2003; Balderhaar et al., 2013; Chaudhry et al., 2010; Mattila et al., 2004; Moriyama and Yahara, 2002) and, thereby, accelerates the assembly of both actin patches and cables. By contrast, the N-terminal HFD domain catalyzes cofilin-mediated severing of actin filaments (Chaudhry et al., 2013) and could promote actin patch disassembly. Recent studies have suggested a possible role for Cof1p and Aip1p at actin cables, based on observations that Cof1p localizes to actin cables in aip1Δ cells (Rodal et al., 1999), and that Cof1p and Aip1p cooperatively promote rapid cable turnover (Okada et al., 2006). Therefore, Cof1p might recruit Srv2p onto actin cables to promote rapid cable turnover (Fig. 8, wild type) (Chaudhry et al., 2013). In contrast, in srv2Δ cells, actin turnover through the CARP domain and activation of cofilin by the HFD domain are impaired, and these cause a delay in both the assembly and disassembly of actin patches. The decrease in ATP-actin levels also leads to defective formation of actin cables.

Here, we have demonstrated that deletion of the HFD and PWP domains primarily affects disassembly of actin patches, whereas the CARP domain is required for both assembly and disassembly of actin patches. Deletion of the HFD domain changed the localization of Srv2p in wild-type and sla2Δ cells, and in both cases, Srv2p displayed a different localization from that of Cof1p. The srv2-91 mutant, which contains a crucial mutation in the HFD, also exhibited phenotypes of altered Srv2p localization and Abp1p dynamics, similar to those seen in the srv2ΔHFD mutant. This suggests that Srv2ΔHFD loses the ability to bind to Cof1p, and therefore Cof1p activation by Srv2p is also impaired, causing a delay in actin patch disassembly (Fig. 8, ΔHFD). The P2 region of Srv2p functions as a binding site for the SH3 domain of Abp1p during Srv2p localization at cortical actin patches (Freeman et al., 1996; Lila and Drubin, 1997; Yu et al., 1999). We have also demonstrated that deletion of the PWP domain and that the srv2-203 mutation significantly decreases and/or delays Srv2p recruitment to cortical patches. This causes a delay in Cof1p activation by Srv2p at the cortical patches and results in a delay in actin patch disassembly (Fig. 8, ΔPWP). The C-terminal CARP domain of Srv2p is vitally important for the proper assembly of actin patches and cables. In the srv2ΔCARP and srv2-104 mutants, Srv2p is recruited to the cortical patches normally, probably through the HFD and PWP domains, and Cof1p activation is induced normally but is unable to catalyze nucleotide exchange on actin monomers. This probably causes a decreased supply of ATP-actin monomers to be available to the growing end of actin filaments, resulting in a delay in actin patch assembly mediated by the Arp2/3 complex and in actin cable formation mediated by formins (Fig. 8, ΔCARP).

Several of our findings in the present study indicate that localization of Srv2p at cortical actin patches is regulated by cofilin as well as Abp1p. First, deletion of the PWP region, even in abp1Δ cells, did not completely impair the cortical localization of Srv2p, but deletion of the HFD domain of Srv2p further decreased the cortical localization of Srv2p when expressed in abp1Δ cells. As the HFD domain binds directly to F-actin, in addition to its actin-dependent interaction with cofilin (Jansen et al., 2014), the HFD interaction with F-actin also probably contributes to Srv2p localization. Second, Srv2p and Abp1p exhibited apparently different localizations in sla2Δ cells, with Abp1p stably associating with the actin comet tail and actively treadmilling through the elongated actin structures (Kaksonen et al., 2003; Okreglak and Drubin, 2007), whereas Srv2p localized only at the cytosolic edges of the actin tail. Interestingly, Cof1p displayed a similar localization to Srv2p in both wild-type and sla2Δ cells (Okreglak and Drubin, 2007), suggesting that the in vivo localization of Srv2p is dependent on Cof1p, as well as on Abp1p. The reduced binding affinity of Srv2ΔHFD to Cof1p seems to increase the dependency of Srv2p localization on Abp1p. Both Srv2p and cofilin are the most highly conserved proteins that regulate the actin cytoskeleton in eukaryotic cells. There is much evidence for the importance of the interaction between these proteins in actin turnover, both in yeast and mammals (Balcer et al., 2003; Chaudhry et al., 2013; Moriyama and Yahara, 2002; Quintero-Monzon et al., 2009). Additionally, the interaction of Srv2p with cofilin through the HFD domain has also been reported in mammals (Jansen et al., 2014; Moriyama and Yahara, 2002; Quintero-Monzon et al., 2009), suggesting conservation of the mechanisms of Srv2p regulation between yeast and mammals.

srv2Δ cells exhibited several unexpected actin cable phenotypes, suggesting that there are, as yet, undiscovered mechanisms regulating the interaction between endocytic vesicles and actin cables. In srv2Δ cells, fragmented short actin cables frequently moved toward endocytic sites and moved past them. Actin cables occasionally exhibited a curved structure while moving between two patches in srv2Δ cells, implying that the cables are anchored to endocytic sites and that Srv2 is involved in their proper anchoring. The molecular machinery that attaches actin cables to endocytic sites has not yet been elucidated, although a likely candidate would be an endocytic protein that binds to F-actin. Among such proteins, Sla2p, the yeast homolog of HIP1R, is a potential candidate because it can bind to both the plasma membrane and F-actin through its N-terminal ANTH domain and C-terminal talin-like domain (Sun et al., 2005; Yang et al., 1999). A recent study has shown that Ent1p, the yeast homolog of epsin, is also capable of binding to the membrane and F-actin through its ENTH domain and actin cytoskeleton-binding (ACB) domain (Skruzny et al., 2012). Interestingly, Sla2p and Ent1p can interact redundantly with F-actin, and strains carrying deletions of the actin-binding regions of both proteins exhibit severely defective internalization of endocytic vesicles (Skruzny et al., 2012). The Eps15-like protein Pan1p is another possible candidate because it forms multiple complexes with several endocytic proteins and actin (Goode et al., 2015). Whichever protein is required, our data argues that Srv2p serves to increase its affinity for binding to vesicles or actin cables. Clarifying the mechanism that mediates these interactions is an important area of future study.

The yeast strains used in this study are listed in Table S1. All strains were grown in standard rich medium (YPD) or synthetic medium supplemented with 2% glucose and appropriate amino acids. C-terminal GFP or mCherry tagging of proteins was performed as described previously (Longtine et al., 1998). Integration plasmids for full-length SRV2 and for the ΔHFD, ΔPWP and ΔCARP SRV2 domain mutants were created as follows: 329 bp of the 5′ UTR, the full length of the SRV2 open reading frame (nucleotides 1–1581) and 274 bp of the 3′ UTR from the SRV2 gene were amplified by using PCR and cloned into the BamHI site of pBluescript II SK (pBS-SRV2). Then, the NotI site was destroyed by treating the constructs with the Klenow fragment to create pBS-SRV2 NotIΔ. pBS-SRV2 NotIΔ was digested by BstZ171, and the SmaI fragment of the LEU2 gene was inserted to make pBS-SRV2 LEU2. The ΔHFD, ΔPWP and ΔCARP plasmids were generated by ligating the NotI-digested PCR products that had been amplified using primers listed in Table S2, using pBS-SRV2 NotIΔ as a template. To remove the mutations that might have been introduced into the plasmids, the resultant plasmids were digested with Aor51HI and MluI, and the Aor51HI-MluI-digested fragment (containing the ΔHFD, ΔPWP or ΔCARP fragments) was inserted into the Aor51HI- and MluI-digested pBS-SRV2 LEU2 construct to replace the full-length SRV2 gene with each mutant. Specific point mutations were introduced through site-directed mutagenesis using the pairs of mutagenic oligonucleotides listed in Table S2. To integrate srv2 mutants at the endogenous locus of the SRV2 gene, the plasmids were digested with BamHI and transformed into srv2Δ cells.

Fluorescence microscopy was performed using an Olympus IX81 microscope equipped with a ×100/NA 1.40 (Olympus) objective and an 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 that passed the red component through a 630/50 nm filter and the green component through a 530/30 nm filter. FM4-64 staining was performed as described previously (Toshima et al., 2005).

Fluorescence labeling of α-factor was performed as described previously (Toshima et al., 2006). For endocytosis assays, cells were grown to an OD₆₀₀ of ∼0.5 in 0.5 ml YPD, briefly centrifuged, and resuspended in 20 µl of synthetic medium with 5 µM of Alexa-Fluor-594-labeled α-factor. After incubation on ice for 2 h, the cells were washed with ice-cold synthetic medium. Internalization was initiated by addition of synthetic medium that contained 4% glucose and amino acids at 25°C.

Preparation and internalization of ³⁵S-labeled α-factor was performed as described previously (Toshima et al., 2005). For the binding assay, cells were grown to an OD₆₀₀ of ∼0.3 in 1 ml YPD at 25°C, briefly centrifuged, and resuspended in 50 µl of synthetic medium with 1% (w/v) bovine serum albumin (BSA) and ³⁵S-labeled α-factor on ice. After incubation on ice for 2 h, cells were washed with ice-cold synthetic medium, and the radioactivity was then measured.

Immunoblot analysis was performed as described previously (Toshima et al., 2005). The antibody against GFP (Life Technologies, A-11122) was used at a dilution of 1:500, and the rabbit anti-mouse IgG conjugated to horseradish peroxidase (Amersham Biosciences, Piscataway, NJ) at 1:1000 dilution was used as the secondary antibody. Immunoreactive protein bands were visualized using ELC plus (GE Healthcare).

The fluorescence intensity of cortical patches in wild-type cells was calculated by averaging the fluorescence intensities of the center area (1×1 pixel area) of individual patches. To calculate the relative fluorescence intensity in mutants, the average fluorescence intensity of the center area of the patch in mutant cells was divided by the average fluorescence intensity of this area in wild-type cells. The fluorescence intensity of actin cables in wild-type cells was calculated by averaging the fluorescence intensities of a randomly selected area (1×1 pixel area) from at least 20 actin cables. For the relative fluorescence intensity of actin cables in mutants, the fluorescence intensity of a randomly selected area (1×1 pixel area) in 50 actin cables was divided by the average fluorescence intensity of actin cables in wild-type cells. The fluorescence intensity was analyzed by using the program ImageJ v1.44.

We are grateful to Anthony Bretscher (Cornell University, NY) for providing the bni1-12 bnr1Δ (Y4135) strain.