Eps15/Pan1p is a master regulator of the late stages of the endocytic pathway

Clathrin-mediated endocytosis (CME) is a fundamental cellular function involved in various physiological processes, such as nutrient uptake, receptor internalization, synaptic vesicle recycling, and defense against pathogens. The molecular mechanism regulating CME is highly conserved in eukaryote cells, and it can be roughly divided into an early phase and a late phase. The early-phase proteins, including early coat proteins, function in determining the endocytic-site location and recruiting cargo, whereas the late-phase proteins, including mid/late coat proteins, actin and actin regulators, and amphiphysin, induce deeply invaginated membranes by triggering transient actin assembly and facilitating scission (Kaksonen and Roux, 2018). The molecular mechanisms underlying each phase have been investigated intensively, but how these phases link to each other remains unclear. Thus, clarifying the molecular mechanism regulating transitions between the phases is important for understanding how the progression of endocytosis is regulated. In previous studies, it has been shown that endocytic vesicle formation in the late phase can still occur even when genes encoding several early-phase proteins are simultaneously deleted, suggesting there is a degree of autonomy in the execution of the late phase in endocytosis (Brach et al., 2014).

One promising candidate protein for coupling the late endocytic phase to early phase is the yeast Eps15-like protein Pan1p, which serves as a key scaffolding protein in the late stages of the yeast CME pathway. Previous studies have demonstrated that Pan1p interacts with numerous endocytic proteins, including clathrin adaptors and coat proteins (Tang et al., 2000; Wendland and Emr, 1998; Wendland et al., 1999), and that these interactions are negatively regulated by the Prk1 family protein kinases, Prk1p and Ark1p (mammalian GAK/AAK1 homologs; Smythe and Ayscough, 2003). These kinases are recruited to endocytic sites just after the initiation of actin assembly, and they phosphorylate several endocytic proteins, including Pan1p, to disassemble the CCV (Cope et al., 1999; Toret et al., 2008; Zeng et al., 2001). The EH domain protein End3p and Pan1p form a stable complex, which is not regulated by Ark1p/Prk1p phosphorylation (Sun et al., 2015b; Toshima et al., 2007), and the complex, together with Sla1p, seems to function as a core scaffold that resembles mammalian intersectins (Goode et al., 2015; Yamada et al., 1998). Recent studies using the auxin-inducible degron system for rapid deletion of Pan1p alone or Pan1p and End3p simultaneously from the cell have suggested that endocytic site initiation and actin assembly are separable processes linked by a Pan1p complex (Bradford et al., 2015; Sun et al., 2015a). These observations clearly indicate the importance of Pan1p in the late stages of endocytosis, although the mechanisms whereby Pan1p drives the progression of endocytosis from the early to late stage remain unknown.

In addition to its role as a scaffolding protein, Pan1p functions as an actin regulator during CCV internalization. Previous studies have revealed that the actin cytoskeleton plays a critical role in the formation and internalization of CCVs and the subsequent transport of endocytic vesicles to the early endosome (Goode et al., 2015; Kaksonen and Roux, 2018; Weinberg and Drubin, 2012). In yeast endocytosis, several actin nucleation promoting factors (NPFs), such as the yeast WASP homolog Las17p (Winter et al., 1999), the type I myosins Myo3p and Myo5p (Evangelista et al., 2000), the actin-binding protein Abp1p (Goode et al., 2001), and Pan1p (Duncan et al., 2001), promote actin polymerization to generate the force required for CCV formation and internalization. Among these NPFs, Las17p and Myo3p/5p, which form a complex with yeast WIP (Vrp1p), have been shown to have high activity and play an important role in the initiation of actin assembly at sites of endocytosis (Sun et al., 2006). In contrast, Pan1p has a lower NPF activity compared to Las17p and Myo3/5p, and mutation in the Arp2/3 complex binding region of Pan1p causes only a minor defect in actin polymerization at the endocytic sites (Sun et al., 2006; Toshima et al., 2005). However, the dephosphorylation-mimicking mutant of Pan1p binds to F-actin with high affinity, and its expression causes aberrant actin polymerization (Toshima et al., 2005; Toshima et al., 2016). Interestingly, this Pan1 mutant also leads to stable interaction between endocytic vesicles and actin cables (Toshima et al., 2016). Thus, although Pan1p seems to play somewhat important roles in actin-mediated vesicle internalization and transport, details of its contribution to actin assembly during endocytosis remain ambiguous.

In the present study, we sought to clarify the function of Pan1p during the progression of the late steps of endocytosis using cells in which Pan1p had been relocated to the peroxisome using a rapamycin-induced dimer formation system. We found that anchoring Pan1p to the peroxisome triggered most of the events occurring during the late stage of endocytosis, including recruitment of the mid- and late-coat proteins and adaptors, initiation of actin polymerization, and disassembly of the endocytic machinery. Pan1p anchored to the peroxisome also recruited actin cables to the peroxisome simultaneously with the initiation of actin polymerization. These findings suggest that the recruitment of Pan1p to endocytic sites is sufficient to initiate and complete the later processes of clathrin-mediated endocytosis.

In comparison to Las17p and Myo3/5p, Pan1p has lower NPF activity, and mutation in the Arp2/3 complex binding region of Pan1p causes only a minor defect of actin polymerization at endocytic sites (Sun et al., 2006; Toshima et al., 2005). Thus, it has been unclear whether Pan1p is required for the initiation of actin polymerization at endocytic sites. To investigate this issue, we ectopically anchored Pan1p to peroxisomes using the FRB and FKBP domains, which form stable dimers in the presence of rapamycin (Banaszynski et al., 2005). We first employed Pex-FRB-mCherry, a fusion protein containing the peroxisomal membrane-targeting signal of Pex3p coupled to mCherry fluorescent protein and FRB (Fig. 1 A) and confirmed that it became localized at Pex3-GFP-labeled peroxisomes (Fig. S1 A). We also fused Pan1p to GFP and FKBP (Pan1-GFP-FKBP; Fig. 1 A) and tested whether the anchoring of Pex-FRB-mCherry to peroxisomes upon the addition of rapamycin resulted in the recruitment of Pan1p to peroxisomes. Before rapamycin addition, little colocalization between Pan1-GFP-FKBP and Pex-FRB-mCherry was observed, but after rapamycin addition, ∼94 and ∼98% colocalization was observed at 15 and 30 min, respectively (Fig. 1, B and C). To assess actin polymerization, we examined the localization of a marker for this process, Abp1-GFP, in a strain expressing Pex-FRB-mCherry and Pan1-FKBP (Fig. 1 D). When Pan1-FKBP was anchored to peroxisomes, around ∼62.0% of Abp1-GFP was locally recruited, but the recruitment was remarkably inhibited in the presence of the actin-sequestering drug, LatrunculinA (LatA), suggesting that Abp1p localization at peroxisomes requires F-actin (Fig. 1, E and F). To directly confirm that actin polymerization occurs at the peroxisome, we used GFP-fused actin (GFP-Act1) instead of Abp1-GFP and investigated whether GFP-Act1 was recruited to peroxisomes upon treatment with rapamycin. Similar to Abp1-GFP, localization of GFP-Act1 at the peroxisomes was clearly observed (Fig. 1, G and H). To exclude the possibility that the localization of Pan1p and actin to the peroxisome are caused by the tethering of endocytic vesicles to the peroxisome, we examined the localization of the fluorescent endocytic tracer FM4-64 in the presence of rapamycin. When added to wild-type cells, FM4-64 is immediately incorporated into the plasma membrane and internalized via bulk-phase endocytosis. FM4-64 is then transported to the endosomes at 5–10 min, but FM4-64-labeled puncta were rarely localized at peroxisomes, although Abp1-mCherry is efficiently recruited there (Fig. S1, B and C). LatA treatment caused a decrease of Abp1p recruitment to the peroxisome to ∼4.4% at 20 min after treatment, but the recruitment recovered to ∼69.5% at 20 min after washing away LatA (Fig. 1, E and F; and Video 1). These observations suggest that Pan1p has the ability to initiate actin polymerization in vivo by itself or through its interacting proteins.

We next utilized the Arp2/3 complex inhibitor CK-666 that specifically disassembles actin assembly at the endocytic sites in yeast (Nolen et al., 2009). The effect of CK666 was fully reversible, and Abp1p disappeared from peroxisomes rapidly 1 min after the treatment but the localization recovered within 2 min after washing away the drug (Fig. 2, A and B; and Video 2). To examine the dynamics of actin polymerization at the peroxisome upon Pan1p anchoring, we expressed GFP-fused Pan1-FKBP and CFP-fused Abp1p in a strain expressing Pex-FRB-mCherry (Fig. 2 C). Triple-color live cell analysis showed that Pan1p and Abp1p behaved differently at the peroxisome in cells treated with rapamycin. Pan1p persisted at the peroxisome labeled by Pex-FRB-mCherry, whereas Abp1p regularly appeared at the peroxisome and then disappeared (Fig. 2, D and E; and Video 3). Quantification of the fluorescence intensities of these proteins confirmed that Pan1p remained localized to the peroxisome constantly and revealed that Abp1p showed dynamic localization that increased gradually and then decreased after reaching a maximum (Fig. 2 F). We noticed that the behavior of Abp1p at the peroxisome was similar to that at endocytic sites (Kaksonen et al., 2003).

We sought to assess if Pan1p recruits interacting proteins to modulate actin assembly. At the endocytic site, actin polymerization is regulated by several actin nucleation promoting factors (NPFs), including Myo3/5p, Las17p, Abp1p, and Pan1p (Goode et al., 2015; Weinberg and Drubin, 2012). As it has been reported that Las17p and Myo3p/5p have higher NPF activity relative to Pan1p and Abp1p and that their NPF activities are most important for endocytic vesicle formation, we next examined if these proteins are recruited to the peroxisome, dependent on Pan1p (Fig. 2 G). Before the addition of rapamycin, Las17p and Myo3/5p were rarely localized at the peroxisome (Fig. 2 I and Fig. S1 D), but after Pan1-FKBP had become anchored to the peroxisome upon treatment with rapamycin, around ∼49.3%, ∼19.3, and ∼42.0% of Las17p, Myo3p, and Myo5p were recruited to the peroxisome, respectively (Fig. 2, H and I). We also examined whether Las17-, Myo3-, or Myo5-FKBP was conversely capable of recruiting Pan1p to the peroxisome; however none of these proteins could (Fig. 3, A and B). Surprisingly, direct anchoring of Las17-, Myo3-, or Myo5-FKBP to the peroxisome did not induce actin polymerization, as assessed by Abp1p recruitment (Fig. 3, C–E and Fig. S1 E). A previous study has shown that accumulation of Las17p and Myo3p above a certain threshold at endocytic sites triggers actin polymerization (Sun et al., 2017). Thus, we next examined if increasing amounts of the NPFs can induce actin polymerization at the peroxisome. Consistent with the previous model, simultaneous anchoring of Las17p, Myo3p, and Myo5p, or Myo3p and Myo5p alone, could substantially cause actin polymerization, as assessed by Abp1p recruitment (∼44.7 and ∼26.7%, respectively; Fig. 4, A and B; and Fig. S1 F). Intriguingly, Abp1p regularly appeared and disappeared when Pan1p was anchored (Fig. 2 E), whereas when Las17p and/or Myo3p/5p were anchored, Abp1p was stably localized at peroxisome (Fig. 4, C and D). Anchoring of overexpressed Las17p also induced similar actin polymerization (Fig. 4, E–H). To further clarify the contribution of these NPF activities for actin polymerization at the peroxisome, we utilized inactive or deletion mutants. When a Pan1 mutant that lacks both actin binding and Arp2/3 binding regions (Pan1ΔWA) is anchored to the peroxisome, actin polymerization occurred at similar levels to that seen upon the anchoring of wild-type Pan1p (Fig. 4, I and J). This suggests that Pan1p’s NPF activity is not essential for the initiation of actin polymerization. In contrast, the presence of the las17ΔWCA mutant, which lacks the C-terminal Arp2/3 complex activation region, reduced the frequency of actin polymerization at the peroxisome (Fig. 4, I and J), suggesting that Las17p’s NPF activity contributes to the initiation of actin polymerization. Deletion of either Myo3p or Myo5p did not affect actin polymerization at the peroxisome (Fig. 4 J and Fig. S1 G), probably because of the redundant function of these proteins (Anderson et al., 1998). Thus, FKBP-tagged Pan1p was able to recruit all other NPFs to the peroxisome when anchored there, and among them Las17p and Myo3p/5p seemed to be most important for actin polymerization at the peroxisome.

The timing of the appearance and disappearance of many endocytic proteins at endocytic sites has been clarified, and these proteins can be categorized into at least five modules: early proteins, early/mid/late coat, WASp/myosin, actin, and scission modules (Lu et al., 2016). We sought to determine which endocytic proteins are recruited to the peroxisome dependent on Pan1p. To this end, we employed Syp1p and Ede1p as markers of early proteins, Yap1801p as a marker of early coat protein, Sla2p, Ent1p, and Ent2p as markers of mid-coat proteins, and End3p and Sla1p as markers of late-coat proteins (Fig. 5 A; Weinberg and Drubin, 2012). We fused GFP to these proteins and expressed them in a strain expressing Pex-FRB-mCherry and Pan1-FKBP (Fig. 5 B). None of these proteins localized at the mCherry-labeled peroxisome before the addition of rapamycin (Fig. 5 E and Fig. S2 A). At 30 min after rapamycin addition, however, the mid- and late-coat proteins, End3p, Sla2p, Ent1p, and Ent2p, changed their localization to mCherry-labeled peroxisomes, whereas the localization of the early proteins, Syp1p and Ede1p, and the early coat protein, Yap1801p, was not affected (Fig. 5 C). The recruitment of mid- and late-coat proteins by Pan1-FKBP was not affected in the presence of LatA (Fig. 5 D), indicating that these interactions are not mediated by actin filaments. Among the proteins recruited to mCherry-labeled peroxisomes, ∼94.7% of End3p was localized to the peroxisome, while the localization of the other proteins was partial (Ent1p, ∼64.0%; Ent2p, ∼66.7%; Sla1p, ∼63.3%; Sla2p, ∼72.0%; Fig. 5 E). Thus, it is likely that End3p and the other proteins interact with Pan1p by different mechanisms.

Since previous studies have reported that Pan1p and End3p form a stable complex with 1:1 stoichiometry (Sun et al., 2015a; Toshima et al., 2007), we speculated that End3p stably interacts with Pan1p at the peroxisome as well. Quantification of the End3-GFP’s fluorescence intensity confirmed that End3p was constantly localized to the peroxisome (Fig. 6, A and B; and Fig. S2 B and Video 4). In contrast, the interaction between Pan1p and Sla1p has been reported to be negatively regulated by Ark1p/Prk1p (Zeng et al., 2001). Two-color analysis using GFP-tagged endocytic proteins and Pex-FRB-mCherry showed that localization of Sla1p and Ent1p at the peroxisome regularly cycles, in a similar manner to Abp1p, with repeated appearance and disappearance at the peroxisome (Fig. 6, C and E; and Fig. S2, C and D; and Videos 5 and 6). Quantification of the fluorescence intensities of these proteins revealed that Sla1p and Ent1p showed a dynamic localization that increased gradually, and then decreased after reaching a maximum, similar to Abp1p (Fig. 6, D and F). Localization of Las17p or Myo3p was also found to change regularly, like Sla1p and Ent1p (Fig. S3, A–D), whereas the change in localization of Sla2p appeared to occur less regularly (Fig. S3, E and F), suggesting a distinct mechanism of localization for Sla2p. We also examined whether Sla2p, Ent1p, or Ent2p is conversely capable of recruiting Pan1p to the peroxisome when anchored to the latter, as these mid-coat proteins are reported to appear at endocytic sites before Pan1p (Weinberg and Drubin, 2012). To this end, we fused Sla2p, Ent1p and Ent2p to GFP and FKBP, and confirmed that these proteins were targeted to the peroxisome in the presence of rapamycin (Fig. S4, A–C). In contrast to the situation when Pan1p was anchored to the peroxisome, no Pan1p recruitment was observed when Sla2-, Ent1-, or Ent2-FKBP was thus anchored (Fig. S4, D–F). Furthermore, rapamycin treatment elicited no recruitment of Abp1p to the peroxisome (Fig. S4, G–I). These results suggest that neither Sla2p, Ent1p, or Ent2p is able to recruit Pan1p and initiate actin polymerization by itself.

Finally, we examined the timing of recruitment of Sla1p and Abp1p, markers of late coat and actin polymerization, respectively, to the peroxisome. At the endocytic sites, Sla1p appeared ∼10–15 s earlier than Abp1p, the intensity of Sla1p reached the maximum when Abp1p appeared at the site, and then Sla1p disappeared from the site before Abp1p (Kaksonen et al., 2005). Interestingly, triple-color live cell analysis showed that Sla1p and Abp1p behaved in a similar manner as at the endocytic site, where Sla1p appeared at the peroxisome first, and then Abp1p appeared when the fluorescence intensity of Sla1p approached the maximum (Fig. 6, G and H; and Fig. S2 E and Video 7). These data suggest that Pan1p regulates the assembly of the endocytic machinery required for the mid-to-late stage of endocytosis and also regulates its disassembly.

We have shown that localization of mid- and late-coat proteins depend on Pan1p by ectopically localizing Pan1p to the peroxisomes. We next wished to determine the extent to which the localization of each endocytic protein to the actual endogenous endocytic site is dependent on Pan1p. We first targeted Pan1-FKBP to peroxisomes and examined the effect on Pan1p localization at cortical patches. While most Pan1p changed its localization to the peroxisome, ∼24% still remained localized at the cortical patches (Fig. S5 A). To remove Pan1p from endocytic sites more efficiently, we next utilized Mito-mCherry-FRB, a fusion protein containing the mitochondrial outer membrane protein OM45 coupled to mCherry and FRB, to anchor Pan1p to mitochondria (Fig. 7 A). At 30 min after the addition of rapamycin, in cells expressing Mito-FRB-mCherry, Pan1-GFP-FKBP changed its localization to the mitochondria (Fig. 7 B), with only ∼0.8% of the Pan1 present at cortical patches before addition (Fig. 7 C). Since Pan1p is essential for endocytosis, its removal from the endocytic site should severely impair endocytosis. To confirm this, we labeled cells expressing Pan1-FKBP and Mito-mCherry-FRB with Alexa Fluor 594-labeled α-factor (A594-α-factor; Toshima et al., 2006) in the presence or absence of rapamycin and followed the localization at several time points after α-factor internalization. As expected, in rapamycin-untreated cells the majority of A594-α-factor was transported to the vacuole after 20 min, whereas an obvious delay of α-factor internalization was observed in rapamycin-treated cells (Fig. 7, D and E). Similar to the situation when Pan1p was targeted to the peroxisome and when Pan1p was targeted to mitochondria, the mid- and late-coat proteins exhibited localization at mitochondria, but localization of early or early coat proteins was not observed (Fig. 7 F; and Fig. S5, B and C). To test if the recruitment of Pan1-FKBP causes tethering of endocytic vesicles to mitochondria, we again examined the localization of FM4-64 and found that FM4-64-labeled puncta were rarely observed at mitochondria in the presence of rapamycin although Abp1p is recruited there (Fig. S5 D). Consistent with this localization at the mitochondria, in cells treated with rapamycin, the fluorescence intensities of the mid or late coat proteins at cortical patches were significantly decreased, whereas those of early or early coat proteins were not affected (Fig. 7, F and G). The fluorescence intensity of End3-GFP at cortical patches decreased to a similar level (∼1.3%) as that of Pan1-GFP (Fig. 7, C, F and G), indicating that most End3p was targeted to mitochondria by forming a stable complex with Pan1p. The marked decrease of Sla1-GFP localization at cortical patches (∼33.9%) also suggested that Sla1p is a major binding partner of Pan1p (Fig. 7, F and G), consistent with previous findings (Tang et al., 2000). As Abp1p localizes to cortical patches dependent on actin filaments, more Abp1p may reroute to mitochondria where Pan1p-dependent ectopic actin polymerization occurs (Fig. 7, F and G). Intriguingly, contrary to previous observations (Bradford et al., 2015; Sun et al., 2015a), localization of the mid-coat proteins Ent1p, Ent2p, and Sla2p was also partially affected by ectopic targeting of Pan1p to mitochondria (Fig. 7, F and G).

We demonstrated that at Pan1p-anchoring peroxisomes, the mid/late coat proteins, and Abp1p appear periodically and then disappear in a manner similar to the dynamics observed at sites of endocytosis (Fig. 6, G and H), suggesting that endocytic coat disassembly and subsequent actin depolymerization occur regularly at the peroxisome. This motivated us to further examine the localization of proteins such as Prk1p, cofilin (Cof1p), and Srv2p, which regulate endocytic coat disassembly and actin depolymerization (Fig. 8 A). To this end, we fused GFP to these proteins and expressed them in a strain expressing Pex-FRB-mCherry and Pan1-FKBP (Fig. S5 E). As expected, we found that ∼40% of Prk1p became localized at peroxisomes after the addition of rapamycin (Fig. 8, B and C; and Video 8), and that this localization was remarkably inhibited by LatA treatment (Fig. S5 F). To determine the timing of Prk1p recruitment to the peroxisome, we compared the dynamics of Abp1-mTurquise2 (mTur2) and Prk1-GFP in cells anchoring Pan1p to the peroxisome. Triple-color imaging demonstrated that Prk1p recruitment to the peroxisome was slightly delayed in relation to Abp1p (Fig. 8 D), similar to Prk1p dynamics at the endocytic site (Toret et al., 2008). This suggested that Pan1p might be phosphorylated by Prk1p at the peroxisome, as well as at the endocytic site, resulting in disassembly of the mid/late coat proteins and actin. We also found that the actin depolymerization factor Cof1p, and Srv2p, which cooperatively accelerates actin depolymerization with Cof1p, were recruited to the peroxisome (∼51.0 or ∼52.6%, respectively; Fig. 8, E, G, H, and J; and Videos 9 and 10). Triple-color imaging using Abp1-mTur2 and Cof1-GFP or Srv2-GFP revealed that these proteins were recruited to the peroxisome several seconds after Abp1p recruitment and reached their maximum fluorescence intensities after Abp1p (Fig. 8, E, F, H, and I). Thus, Cof1p and Srv2p are likely to regulate the disassembly of the actin cytoskeleton assembled by the Pan1p complex at peroxisomes.

In addition to the role of Pan1p in CCV formation, Pan1p is also reported to regulate the interaction between endocytic vesicles and actin cables (Toshima et al., 2016; Toshima et al., 2006). Accordingly, we next examined whether actin cables interact with peroxisomes when Pan1p is anchored there. To this end, we labeled actin cables with triple GFP (3GFP)-tagged Abp140p and examined the spatiotemporal relationship between Pan1p-anchoring peroxisomes and actin cables. In the absence of rapamycin, little interaction between peroxisomes and actin cables was observed (∼4.3%), but the interaction increased markedly after rapamycin addition (∼47.0%; Fig. 9, A–C; and Video 11). Interestingly, we observed Pan1p-anchored peroxisomes transiently interacting with and moving along actin cables, labeled by Abp140-3GFP, when actin assembly was initiated at the peroxisome (Fig. 9 A). Then, the peroxisome dissociated from the actin cable at almost the same time point as actin disassembly (Fig. 9 A, lower panels). In contrast, such peroxisome dynamics along actin cables were rarely observed in the absence of rapamycin (Fig. 9 B). Quantification of peroxisome velocity revealed that in the presence of rapamycin, peroxisomes move on actin cables with an average speed of 217.6 ± 83.7 μm/s, whereas peroxisomes in the absence of rapamycin move little (36.9 ± 18.6 μm/s; Fig. 9 D). This movement of peroxisomes closely resembles that of endocytic vesicles, which move on actin cables when being internalized into the cell (Huckaba et al., 2004; Toshima et al., 2006). These findings suggest that the Pan1p complex directly regulates the interaction between endocytic vesicles and actin cables.

Clathrin-mediated endocytosis is driven by sequential assembly and disassembly of numerous endocytic proteins at the plasma membrane. The timing of the appearance and behavior of these proteins at endocytic sites has been well characterized (Goode et al., 2015; Kaksonen and Roux, 2018; Weinberg and Drubin, 2012), but the molecular mechanisms whereby endocytic proteins are recruited have remained unclear. One particularly important issue remaining to be clarified has been the manner in which mid/late coat proteins are recruited to endocytic sites because in cells lacking most of the genes encoding early-phase proteins, the mid/late coat proteins are regularly recruited to endocytic sites and the dynamics of vesicle budding are largely normal (Brach et al., 2014). In the present study, we showed that Pan1p anchored to the peroxisome was capable of initiating all of the events occurring at the late stage of endocytosis, including mid/late coat protein recruitment, NPF-mediated actin polymerization, coat and actin disassembly, and actin cable recruitment at the peroxisome (Fig. 10). These findings suggest that Pan1p serves as a master regulator of the late stage of endocytosis.

A transient burst of actin polymerization and subsequent depolymerization at the endocytic site is required for the formation and internalization of a CCV (Kaksonen et al., 2003; Merrifield et al., 2002). Previous studies have reported that mutant Las17p, which lacks the ability to activate the Arp2/3 complex, causes a severe defect of actin polymerization at endocytic sites. In contrast, it has been demonstrated that mutant forms of Pan1p and Myo3p/5p which lack Arp2/3 activation capability do not significantly affect actin polymerization (Sun et al., 2006). These observations suggested that the NPF activity of Las17p plays a central role in the initiation of actin polymerization at endocytic sites, while Pan1p has only an auxiliary role. In the present study, however, we showed that anchoring of Las17p to the peroxisome elicited little actin polymerization in situ, whereas anchoring of Pan1p was sufficient to trigger substantial actin polymerization. Consistent with this observation, previous studies reported that in cells lacking full-length Pan1p by the auxin-inducible degron system, actin polymerization at endocytic sites was severely defective, and the number of cortical patches was significantly decreased (Bradford et al., 2015; Sun et al., 2015a). These suggested that the function of Pan1p as a scaffold protein, in addition to the NPF activity, is important for accumulating all NPFs in a particular region and initiating explosive actin polymerization at the endocytic site.

A recent study has shown that the accumulation of Las17p and Myo3p above a certain threshold at endocytic sites triggers actin polymerization and that the Pan1p complex is implicated in the regulation of that process (Sun et al., 2017). Pan1p anchored to the peroxisome likely recruits Las17p and Myo3p/5p in a similar manner as at the endocytic site, causing accumulation of these NPFs to the peroxisome. On the other hand, anchoring of Las17p or Myo3p/5p alone might not recruit sufficient NPFs to initiate actin polymerization. Taken together, these findings suggest that Pan1p likely accumulates the actin assembly machinery at endocytic sites and generates the force for CCV formation and invagination through intensive polymerization of actin.

Previous studies demonstrated that Pan1p possesses multiple functional regions, such as two EH domains in the N-terminal region, a central coiled-coil region, and actin and Arp2/3 binding regions in the C-terminal region (Toshima et al., 2005; Wendland and Emr, 1998), and forms complexes with various endocytic proteins. However, it has not been definitively identified which proteins, especially early and mid-coat proteins, Pan1p interacts directly with in vivo. The EH domains are reported to interact with NPF motifs in Yap1801p/2p and Ent1p/2p, which are early and mid-coat proteins (Suzuki et al., 2012; Wendland and Emr, 1998; Wendland et al., 1999). However, our analysis showed that Pan1p was not capable of recruiting Yap1801p/2p to the peroxisome, and that localization of Yap1801p/2p was not affected even when most of the Pan1p was targeted to mitochondria. Similarly, localization of Yap1801p/2p was not affected in cells depleted of Pan1p (Bradford et al., 2015), suggesting that Yap1801p/2p, rather than Pan1p, might preferentially interact with components in the early clathrin coat (Huang et al., 1999). In contrast, Pan1p was able to recruit the mid-coat proteins Ent1p/2p and Sla2p, which contain the ANTH/ENTH domains that interact specifically with phosphoinositides (Legendre-Guillemin et al., 2004). These proteins are reported to be normally localized at cortical patches in cells depleted of Pan1p (Bradford et al., 2015; Sun et al., 2015a); however in our analysis, their localization at cortical patches is decreased when Pan1p was targeted to mitochondria, as it partially reroutes these proteins to mitochondria. This observation suggests that Pan1p or the Pan1p complex directly interacts with these mid coat proteins, thereby being recruited to the endocytic site and progressing to later stages of the endocytic pathway.

Interaction between Pan1p and several endocytic proteins is regulated by the Pan1p phosphorylation cycle (Smythe and Ayscough, 2003), but previous studies have failed to demonstrate which interactions are phosphorylation-dependent in vivo because the Pan1p complex disappears from the cortical patch when the endocytic vesicle is internalized (Kaksonen et al., 2003). Our analysis clearly demonstrated which proteins stably or transiently interacted with Pan1p in vivo. End3p has been shown to interact stably with Pan1p (Sun et al., 2015b; Toshima et al., 2007), and this was also confirmed in the present analysis. Interaction of Sla1p and Ent1p with Pan1p is reported to be regulated by phosphorylation (Watson et al., 2001; Zeng et al., 2001), and our present analysis also showed that it is periodically regulated. Interestingly, the interaction between Pan1p and Las17p or Myo3p/5p appeared to be regulated by the Pan1p phosphorylation cycle. Since these interactions are mediated by Sla1p (Feliciano and Di Pietro, 2012; Rodal et al., 2003), phosphorylation-dependent dissociation of Sla1p from Pan1p could also cause dissociation of these proteins.

Important regulators of actin disassembly are the Ark1p/Prk1p kinases (Smythe and Ayscough, 2003). Among the many substrates for Ark1p/Prk1p kinases, Pan1p is the major in vivo substrate, as pan1, mutated to prevent phosphorylation at all 18 threonines (pan1-18TA), displays almost the same endocytic defect as ark1Δ prk1Δ cells (Toshima et al., 2016). Recruitment of these kinases to sites of endocytosis depends on Abp1p (Cope et al., 1999), and is delayed relative to Abp1p recruitment by 1–2 s (Toret et al., 2008). This timing of Prk1p recruitment is almost coincident with that at the Pan1p-anchoring peroxisome and the timing of Pan1p complex disassembly there. Thus, one potential model that can be considered is that at both peroxisomes and endocytic sites, Pan1p triggers actin assembly, which in turn recruits Abp1p, and Ark1/Prk1 kinases are then recruited in an Abp1p-dependent manner to phosphorylate Pan1p, resulting in the dissociation of the actin filament from Pan1p.

Cof1p and Srv2p enhance actin disassembly by binding around the pointed ends of actin filaments, which have abundant ADP-actin (Balcer et al., 2003). Abp1p also mediates the recruitment of Srv2p to cortical patches (Freeman et al., 1996; Lila and Drubin, 1997). Our present analysis showed that these proteins were also recruited to the Pan1p-anchoring peroxisome, suggesting that Cof1p and Srv2p function cooperatively with Ark1/Prk1 kinases to facilitate rapid actin disassembly. Clathrin coat disassembly is regulated by the lipid phosphatase Sjl2p and recruitment of Sjl2p to endocytic sites is dependent on Abp1p (Stefan et al., 2005). Thus, Sjl2p could be similarly recruited to the Pan1p-anchoring peroxisome and regulates the disassembly of the endocytic machinery.

The yeast strains used in this study are listed in Table S1. All strains were grown at 25°C in standard rich medium (YPD) or synthetic medium (SM) supplemented with 2% glucose and appropriate amino acids. For rapamycin-induced heterodimerization of FRB and FKBP 10 μM, rapamycin was added to cultures 15 or 30 min before imaging. The GFP-FKBP- or FKBP-tagged proteins were expressed as follows: the NotI-SacII fragment, which contains the S. cerevisiae ADH1 terminator (TADH1) and the URA3MX6 module, was amplified by PCR using pFA6a-mCherry-URA3MX6 (Kawada et al., 2015) as a template and inserted into NotI- and SacII-digested pBS-GFP (pBS-TADH1-URA3). The NotI-SacII fragment containing GFP-FKBP or FKBP was amplified by PCR using GFP-FKBP-LCa plasmid (#59353; Addgene) as a template and inserted into BamHI- and NotI-digested pBS-TADH1-URA3 (pBS-GFP-FKBP-TADH1-URA3 or pBS- FKBP-TADH1-URA3). To create integration vectors, the C-terminal fragment of Pan1 (nt 3608-4440), Sla2 (nt 2231-2904), Ent1 (nt 634-1362), Ent2 (nt 634-1362), Las17 (nt 1021-1888), Myo3 (nt 3481-3816), or Myo5 (nt 3661-3957) was amplified by each primer set listed in Table S2, digested with BamHI or BglII, and inserted into BamHI-digested pBS-GFP-FKBP-TADH1-URA3 or pBS- FKBP-TADH1-URA3. To integrate the C-terminal GFP-FKBP or FKBP tag at the endogenous of the target gene, the integration plasmid for Pan1, Sla2, Ent1, Ent2, Las17, Myo3, or Myo5 was linearized by EcoNI, MfeI, MfeI, AflII, EcoRV, SpeI, or EcoRI, respectively, and transformed into JTY9191. Pex-mCherry-FRB or Mito-mCherry-FRB was expressed as follows: pRS305 containing the S. cerevisiae TPI1 promoter (PTPI1) and STE2 terminator (TSTE2) was created by inserting the SacI-XhoI fragment of PTPI1-(BamHI)-TSTE2 extracted from pRS316-PTPI1-(BamHI)-TSTE2 plasmid (Abe et al., 2019) into BamHI site of pRS305 (pRS305-PTPI1-(BamHI)-TSTE2). The N-terminal fragment of Pex3p (nt1-135; Pex) was amplified by PCR with JT3556 and JT3674 using yeast genome DNA as a template, and the mCherry-FRB fragment (mCherry-FRB) was amplified by PCR with JT3675 and JT3291 using pMito-mCherry-FRB plasmid (#59352; Addgene) as a template. Next, the Pex-mCherry-FRB fragment was amplified by PCR with JT3556 and JT3291 using Pex and mCherry-FRB fragments as the template and inserted into the BamHI site of pRS305-PTPI1-TSTE2 (pRS305-PTPI1-Pex-FRB-mCherry-TSTE2). To create Mito-mCherry-FRB integration plasmid, the BglII fragment containing Mito-mCherry-FRB was amplified by PCR using pMito-mCherry-FRB plasmid (#59352; Addgene) as a template and inserted into BamHI site of pRS305-PTPI1-TSTE2 (pRS305-PTPI1-Mito-mCherry-FRB-TSTE2). To integrate these plasmids into the endogenous locus of the HIS3 gene, the plasmids were linearized by EcoRV and transformed into yeast.

3-Thiopropionyl-G₃ was appended to the free ε-amine of K₇ in otherwise fully-protected α-factor by standard DCC/HOBT FMOC solid phase chemistry, and Alexa-594 maleimide (Thermo Fisher Scientific) was coupled to the purified peptide in NMM-HOAc buffer, pH 8.0. Peptides were purified by reverse-phase HPLC; structure and purity (>95%) were assessed by ESI-FTICR mass spectrometry. For endocytosis assays, cells were grown to an OD600 of ∼0.5 in 0.5 ml YPD, briefly centrifuged, and resuspended in 20 μl SM with 5 μM Alexa Fluor-labeled α-factor. After incubation on ice for 2 h, the cells were washed with ice-cold SM. Internalization was initiated by the addition of SM containing 4% glucose and amino acids at 25°C.

Fluorescence microscopy was performed using an Olympus IX83 microscope equipped with a ×100/NA 1.40 (Olympus) or a ×100/NA 1.49 (Olympus) objective and Orca-AG cooled CCD camera (Hamamatsu) using Metamorph software (Universal Imaging). Triple color imaging was performed using an Olympus IX81 microscope equipped with a high-speed filter changer (Lambda 10-3; Shutter Instruments) that can change filter sets within 40 ms. All cells were imaged during the early- to mid-logarithmic phase.

Statistical analysis was conducted using Prism 7 (GraphPad Software). Significance was tested by a two-sided Student’s t test for two-group comparisons or ANOVA with Turkey’s post-hoc test as specified in the figure legends for comparisons of three or more groups.

Fig. S1 shows targeting of Pan1p to the peroxisome leads to actin polymerization by recruiting other NPFs. Fig. S2 shows localization and fluorescence intensity of endocytic proteins at Pan1p-anchoring peroxisomes. Fig. S3 shows dynamic localization of endocytic proteins at Pan1p-anchoring peroxisomes. Fig. S4 shows ectopically localized Ent1p, Ent2p, and Sla2p cannot recruit Pan1p. Fig. S5 shows localization of endocytic proteins in cells expressing Pan1-FKBP and Mito-FRB-mCherry, and actin-dependent recruitment of Prk1-GFP to the Pan1p-anchoring peroxisome. Table S1 shows yeast strains used in this study. Table S2 lists primers used in this study. Data S1 contains source data for LatA treatment and recovery in Fig. 1 F. Video 1 shows de novo synthesis of F-actin at Pan1p-anchoring peroxisome. Video 2 shows Arp2/3-dependent actin assembly at Pan1p-anchoring peroxisome. Video 3 shows dynamics of Pan1-GFP and Abp1-CFP at Pan1p-anchoring peroxisome. Video 4 shows dynamics of End3-GFP at Pan1p-anchoring peroxisome. Video 5 shows dynamics of Sla1-GFP at Pan1p-anchoring peroxisome. Video 6 shows dynamics of Ent1-GFP at Pan1p-anchoring peroxisome. Video 7 shows dynamics of Sla1-GFP and Abp1-mTq2 at Pan1p-anchoring peroxisome. Video 8 shows dynamics of Prk1-GFP and Abp1-mTq2 at Pan1p-anchoring peroxisome. Video 9 shows dynamics of GFP-Cof1 and Abp1-mTq2 at Pan1p-anchoring peroxisome. Video 10 shows dynamics of Srv2-GFP and Abp1-mTq2 at Pan1p-anchoring peroxisome. Video 11 shows recruitment of actin cable to Pan1p-anchoring peroxisome.

This work was supported by JSPS KAKENHI GRANT #18K062291, and the Takeda Science Foundation to J.Y. Toshima, as well as JSPS KAKENHI GRANT #19K065710, the Uehara Memorial Foundation, and Life Science Foundation of JAPAN to J. Toshima.

The authors declare no competing financial interest.

Author contributions: M. Enshoji, Y. Miyano, N. Yoshida, M. Watanabe, and M. Kunihiro: Investigation, Data curation. M. Nagano: Investigation, Data curation, Project administration, D.E. Siekhaus: Writing-Review and Editing. J.Y. Toshima: Investigation, Data curation, Funding acquisition, Supervision. J. Toshima: Supervision, Conceptualization, Funding acquisition, Writing-original draft, Writing-Review and Editing.