The S. pombe adaptor protein Bbc1 regulates localization of Wsp1 and Vrp1 during endocytic actin patch assembly

Clathrin-mediated endocytosis (CME) is an essential process conserved across eukaryotes, which allows for the internalization of external materials into a cell. This process is highly regulated with >40 proteins known to assemble into patch structures at sites of endocytosis in a consistent spatial and temporal pattern (Goode et al., 2015; Kovar et al., 2011; McMahon and Boucrot, 2011; Weinberg and Drubin, 2012). How these proteins coordinate their activities to ensure reproducible endocytic patch assembly and efficient endocytosis is an important open question in the field. Although many of the endocytic patch proteins are conserved among species, some are specific to animals or yeast. Bbc1 (also known as Mti1) is a yeast-specific protein that localizes to sites of endocytosis and is thought to function in regulating actin assembly (Kaksonen et al., 2005; Mochida et al., 2002; Picco et al., 2018; Rodal et al., 2003; Sun et al., 2006).

To help overcome membrane tension in yeast and animal cells, endocytosis relies upon the assembly of actin into structures called actin patches, which provide the scaffold and force needed for efficient membrane invagination and scission (Aghamohammadzadeh and Ayscough, 2009; Basu et al., 2014; Boulant et al., 2011; Engqvist-Goldstein and Drubin, 2003; Goode et al., 2015; Kaksonen and Roux, 2018; Kaksonen et al., 2006). The branched actin network at endocytic sites (Collins et al., 2011; Rodal et al., 2005; Young et al., 2004) is assembled by the Arp2/3 complex upon its activation by nucleation-promoting factors (NPFs) (Higgs and Pollard, 2001). Multiple NPFs are present at endocytic sites in fission yeast Schizosaccharomyces pombe (Basu and Chang, 2011; Carnahan and Gould, 2003; Kovar et al., 2011; Sirotkin et al., 2005, 2010) including the Wiskott–Aldrich Syndrome protein (WASp) homolog Wsp1, myosin-1 (Myo1), the EPS15 homolog Pan1, and the SPIN90 homolog Dip1. Homologs of these proteins are also present in budding yeast Saccharomyces cerevisiae, including the WASp Las17 and the myosin-1 proteins Myo3 and Myo5 (Myo3/5) (Boettner et al., 2012; Goode et al., 2015; Weinberg and Drubin, 2012). NPFs share the ability to stimulate the Arp2/3 complex but differ in the strength and mechanism of Arp2/3 complex activation and have additional distinct functions (Galletta et al., 2008; Sirotkin et al., 2005; Sun et al., 2006; Wagner et al., 2013). Specifically, in both budding and fission yeast, WASp is the strongest Arp2/3 complex activator and is considered the primary NPF driving actin assembly for membrane deformation, while myosin-1 is a weak activator that is thought to contribute force via its actin-dependent motor activity (Basu et al., 2014; Berro et al., 2010; Sirotkin et al., 2005; Sun et al., 2006). These diverse NPFs collaborate during endocytosis. In budding yeast, Las17 and Myo3/5 form a WASp–myosin-1 module (Kaksonen et al., 2005; Weinberg and Drubin, 2012) that also includes the F-BAR protein Bzz1 and the WASp-interacting protein (WIP) homolog verprolin (Vrp1), both of which regulate NPF activity in budding and fission yeast (Arasada and Pollard, 2011; Sirotkin et al., 2005; Soulard et al., 2002; Sun et al., 2006). A similar collaboration exists in fission yeast where Myo1 and Wsp1 interact directly (Carnahan and Gould, 2003) or indirectly via Vrp1, which is recruited to patches by Wsp1 but can also bind Myo1 and stimulate its NPF activity (Sirotkin et al., 2005).

The localization and activity of NPFs are regulated to ensure proper actin polymerization. One proposed regulatory component of the budding yeast WASp–myosin-1 module of endocytosis is Bbc1 (Kaksonen et al., 2005; Weinberg and Drubin, 2012). Bbc1 was originally identified in S. cerevisiae as a binding partner of the myosin-1 Myo5 (Mochida et al., 2002) and was shown to mildly inhibit Myo5 NPF activity in vitro (Sun et al., 2006). However, Bbc1 also interacts with Las17 (Tong et al., 2002). Consistent with this interaction, Bbc1, along with Myo5, is recruited to Las17-coated beads but not to Myo5 in comet tails in in vitro reconstitution assays (Michelot et al., 2010). Moreover, a pioneering study by Rodal et al. (2003) demonstrated that Bbc1, in cooperation with adaptor protein Sla1, inhibits Las17 NPF activity in vitro. This work led to a widely accepted model for Bbc1 and Sla1 as dual Las17 inhibitors (Weinberg and Drubin, 2012). Supporting an inhibitory function of Bbc1, bbc1Δ sla1Δ and, to a lesser extent, bbc1Δ cells demonstrate enhanced actin patch assembly (Kaksonen et al., 2005; Picco et al., 2018). The bbc1Δ cells are also characterized by deeper internalizations of endocytic structures (Kaksonen et al., 2005; Picco et al., 2018). However, this depends on the NPF activity of Myo5 but not Las17 (Sun et al., 2006), consistent with identification of Bbc1 as a Myo5 ligand. The behavior and precise localization of Bbc1 at endocytic sites further suggest that Bbc1 interaction with Myo5 may be important for additional, yet to be discovered, functions of Bbc1 in cells. Specifically, S. cerevisiae Bbc1 is recruited to actin patches with timing similar to Myo3/5 (Kaksonen et al., 2005) and, while immuno-electron microscopy (EM) showed that Bbc1 was localized near both Las17 and Myo3/5 in vivo (Idrissi et al., 2008), a recent super-resolution microscopy study placed Bbc1 closer to Myo3/5 than Las17 (Mund et al., 2018). However, the significance of the Bbc1 interaction with myosin-1, aside from the ability of S. cerevisiae Bbc1 to mildly inhibit Myo5 NPF activity in vitro (Sun et al., 2006), remains largely unknown.

Interestingly, the localization pattern of key NPFs differs in budding and fission yeasts. In budding yeast, Myo3/5 and Las17 remain at the base of the invagination and do not internalize with the endocytic vesicle (Kaksonen et al., 2005), except for a small portion detected by immuno-EM (Idrissi et al., 2008) or unless Las17 is overexpressed (Galletta et al., 2008). In contrast, S. pombe Myo1 and Wsp1 show distinct localization and behavior. Myo1 does not internalize with the actin patch and remains at the base of the endocytic invagination, while Wsp1 internalizes with the patch and separates from Myo1 (Sirotkin et al., 2005). This spatial separation of Myo1 and Wsp1 in S. pombe makes it possible to determine whether regulators of actin patch assembly function via association with Myo1 or Wsp1 by following protein dynamics in live cells.

We detected a protein homologous to budding yeast Bbc1/Mti1 in pulldowns of the contractile ring (CR) proteins Cyk3 (Roberts-Galbraith et al., 2010) and Fic1, piquing our interest in this thus far uncharacterized S. pombe protein. Here, we take advantage of distinct behaviors of Myo1 and Wsp1 in S. pombe to reveal a novel mechanism for how Bbc1 regulates actin assembly based on a competition between Bbc1 and Vrp1–Wsp1 for the Myo1 tail.

The protein SPAC23A1.17/Mti1/Bbc1 was identified by mass spectrometry (MS) in affinity purifications of the CR proteins Cyk3 (Roberts-Galbraith et al., 2010) and Fic1 (Table S1), prompting us to investigate localization of this uncharacterized S. pombe protein in dividing cells (Fig. 1). Bbc1–mNeonGreen (mNG) was absent from the CR labeled with mCherry–Cdc15 but was present at the division site in two types of structures: first, in dynamic cortical puncta and, upon completion of CR constriction, also in stable puncta (Fig. 1A–D). Imaging cells expressing Bbc1–mNG, the spindle pole body (SPB) marker Sid4–mNG and the CR marker mCherry–Cdc15 revealed that Bbc1–mNG first appeared in dynamic puncta at the division site 20.9±2.7 min (mean±s.d.) after the separation of SPBs (Fig. 1C). This timing is similar to the time of appearance of endocytic actin patches at the division site (Wu et al., 2006). Imaging of cells expressing Bbc1–mGFP and endocytic actin patch marker Fim1–mCherry confirmed that the dynamic cortical puncta of Bbc1 at the division site are indeed actin patches (Fig. 1E) and stable puncta are distinct from actin patches (Fig. 1F). Bbc1 in stable puncta at new cell tips colocalized with Cdc15 and Cyk3 (Fig. 1G,H), thus explaining isolation of Bbc1 with these proteins in affinity purifications.

To gain insight into the physiological role of Bbc1, we deleted bbc1⁺ and examined genetic interactions of bbc1Δ and sensitivity of bbc1Δ cells to a variety of stresses. The bbc1Δ cells were sensitive to benomyl and latrunculin A (Fig. S1A) and showed negative interactions with fim1Δ, myo1Δ, zds1Δ and mid1Δ (Fig. S1B), implicating Bbc1 in endocytosis, cell wall integrity and cytokinesis (Chang et al., 1996; Sirotkin et al., 2005; Skau and Kovar, 2010; Sohrmann et al., 1996; Yakura et al., 2006). However, bbc1Δ cells were morphologically normal and had no apparent cytokinesis defects, and Bbc1-overexpressing cells had a normal septation index (Fig. S1C). Given that S. pombe Bbc1 localizes to actin patches and co-purifies with actin patch components Myo1, Pan1 and Cdc15 in affinity purifications (Table S1), we then investigated the role of S. pombe Bbc1 in endocytic actin patches.

We confirmed, by colocalization analysis with actin patch components Fim1 and Pan1, that in interphase cells, as during cell division, Bbc1 localizes to both actin patches and actin-free puncta (Fig. 2A,B; Fig. S2A). During interphase, Bbc1 colocalizes with Cdc15 in actin patches and also in a subpopulation of stable puncta (Fig. 2C; Fig. S2B,C). During cell division, Cdc15 localizes to the CR and is absent from actin patches (Arasada and Pollard, 2011; Carnahan and Gould, 2003; Wu et al., 2006).

Quantitative imaging of strains combining Fim1–mCherry with mGFP-tagged Bbc1 or other patch proteins provided detailed accounting of Bbc1 dynamics in patches relative to Myo1, Wsp1 and Cdc15, patch markers with known dynamics (Arasada and Pollard, 2011; Sirotkin et al., 2010). We determined the time courses of the numbers of molecules of these proteins in actin patches (Fig. 2D; Fig. S2D–H) based on previously measured numbers of actin patch proteins (Sirotkin et al., 2010). Time courses of the numbers of molecules were aligned to the time when patches marked by Fim1–mCherry first moved (time zero), which corresponds to the initiation of actin patch internalization (Berro and Pollard, 2014). S. pombe Bbc1 appeared in actin patches 4–6 s before the start of actin patch internalization, reached peak number of 48 molecules at −2 s, and quickly dissipated 2–6 s after time zero. Cdc15 showed similar behavior, appearing at −4 to −2 s, peaking at 74 molecules at time zero, and disappearing 2–4 s later, comparable to previous measurements (Arasada and Pollard, 2011). Both Myo1 and Wsp1 are present in excess over Bbc1 and Cdc15, reaching peaks of 359 and 155 molecules, respectively, at −2 s. In contrast to Wsp1, which in S. pombe internalizes with the actin patch, neither Bbc1 nor, as previously reported (Arasada and Pollard, 2011), Cdc15 internalized with actin patches. This behavior of Bbc1 and Cdc15 resembles that of Myo1, suggesting Bbc1 and Cdc15 function with Myo1 at the base of the endocytic invagination.

To determine whether Myo1 or Wsp1 play a role in Bbc1 localization in patches, we examined localization of Bbc1–mGFP in myo1Δ and wsp1Δ cells using Cdc15, which is known to depend on Myo1 for localization to actin patches (Arasada and Pollard, 2011), as a control (Fig. 2E,F). We found that both Bbc1–mGFP and Cdc15–mGFP failed to localize to dynamic cortical patches in myo1Δ cells (Fig. 2E) and instead localized exclusively in stable cortical puncta (Fig. S2A,B). In contrast, in the absence of Wsp1, both Bbc1–mGFP and Cdc15–mGFP continued to localize in dynamic patches (Fig. 2E), albeit with slower dynamics (Fig. 2F) typical for endocytic patches in wsp1Δ cells (Basu et al., 2014). Interestingly, while Cdc15 accumulated in patches to the same level in wsp1Δ as in wild-type cells, the peak level of Bbc1 in patches in wsp1Δ cells was increased by ∼25% (Fig. 2F). Thus, Bbc1 is recruited to patches by Myo1 and may compete with other ligands for binding Myo1.

To determine which part of Myo1 is responsible for recruiting Bbc1 and Cdc15 to patches, we examined the ability of Bbc1 and Cdc15 to localize to patches marked by Fim1–mCherry in strains featuring a series of Myo1 tail truncations using two-color imaging (Fig. 3A; Fig. S3A). We uncovered that the Myo1 TH2 and SH3 domains are both required for Bbc1 patch recruitment (Fig. 3A). In contrast, the Myo1 TH2 and SH3 domains contribute but are not required for Cdc15 recruitment. When the Myo1 TH2 domain, SH3 domain or both were deleted (Fig. S3A), Cdc15 continued to localize to some of the patches, albeit at decreased levels. This is consistent with previous mapping of Myo1–Cdc15 interaction to the TH1 domain in addition to a portion of the TH2 domain (Carnahan and Gould, 2003).

To determine whether S. pombe Bbc1 can associate with the Myo1 tail, we developed an in vivo binding assay since Bbc1 is a 170 kDa proline-rich protein, which hinders its purification and characterization in vitro. We fused the C-terminal domain of the SPB protein Ppc89 (Rosenberg et al., 2006) to mCherry and to Myo1 tail fragments consisting of the TH2, SH3 and LCA domains (Ppc89–mCherry–23A), the SH3 and LCA domains (Ppc89–mCherry–3A), or the LCA domain alone (Ppc89–mCherry–LCA). These constructs were introduced into cells expressing mGFP-tagged Bbc1, other known Myo1 tail ligands Vrp1, Wsp1 and Cdc15 (Carnahan and Gould, 2003; Sirotkin et al., 2005), or Sla1, which is not known to interact with Myo1. Rather than localizing to SPBs as might have been expected, each construct formed aggregates within the cytoplasm. These were still useful, however, to determine whether mGFP-tagged proteins were recruited by mCherry-marked Myo1 fragments (Fig. 3B). With the exception of Sla1, all examined proteins relocalized to the Ppc89–mCherry–23A puncta (Fig. 3C). However, only Bbc1, Vrp1 and Wsp1 relocalized to the Ppc89–mCherry–3A puncta (Fig. 3D). Neither Bbc1 nor Vrp1 relocalized to the Ppc89–mCherry–LCA aggregates, indicating that the Myo1 SH3 domain alone is sufficient to relocalize both Bbc1 and Vrp1 (Fig. S3B). The binding of Cdc15 to the TH2–SH3–LCA but not SH3–LCA Myo1 construct is consistent with Cdc15 interacting with the TH1 and TH2 domains (Carnahan and Gould, 2003) and suggests that SH3 domain helps recruit Cdc15 to patches via indirect interactions. In contrast, the Myo1 SH3 domain is sufficient for interaction with Bbc1 and together with the TH2 domain is necessary for Bbc1 localization to patches.

Since the Myo1 SH3 domain is necessary for recruitment of Bbc1 to patches but also interacts with Vrp1 in vivo (Fig. 3D) and in vitro (Sirotkin et al., 2005), we tested whether Vrp1 and Bbc1 compete for localization to patches in cells. We measured accumulation of Bbc1–mGFP in patches in vrp1Δ cells and Vrp1–mGFP in bbc1Δ cells versus wild-type cells, expecting increased accumulation of Bbc1 or Vrp1 in patches in the absence of its potential competitor. Indeed, while whole-cell intensities were unchanged (Fig. S4A,B), Bbc1 patch accumulation was increased by 40% in the absence of Vrp1 (Fig. 4A), and Vrp1 patch accumulation was increased by 51% in the absence of Bbc1 (Fig. 4B). This competition mechanism may also explain the modest 25% increase in the Bbc1 level in patches observed in wsp1Δ cells (Fig. 2F) since Wsp1 recruits Vrp1 to patches (Sirotkin et al., 2005).

Next, we examined the effects of competition between Bbc1 and Vrp1 on recruitment of Myo1 to patches by measuring Myo1 levels in patches in the absence of Bbc1, Vrp1 or both proteins (Fig. 4C). Myo1 accumulation in patches was reduced by 20–30% in vrp1Δ and bbc1Δ vrp1Δ cells, suggesting that Vrp1 helps retain Myo1 in patches even though these proteins differ in their patch dynamics (Sirotkin et al., 2005). In contrast, Myo1 accumulation was mildly increased in bbc1Δ cells. This slight increase, 15% on average, was reproducible across multiple experiments (Fig. 4D) and can be explained if Myo1 association with Vrp1 is favored in the absence of Bbc1 (Fig. 4E), thus prolonging Myo1 retention in patches.

Verprolins bind long-tailed myosin-1 and WASp homologs through two distinct domains, a proline-rich region and the C-terminal WASp-binding domain, respectively (Anderson et al., 1998; Carnahan and Gould, 2003; Naqvi et al., 1998; Ramesh et al., 1997; Sirotkin et al., 2005). Therefore, S. pombe Vrp1 can bridge Myo1–Wsp1 association and may help retain both Myo1 and Wsp1 in patches (Fig. 4E). We hypothesized that increased interaction of Bbc1 with Myo1 upon Bbc1 overexpression will disrupt this bridging interaction and decrease accumulation of Vrp1, Myo1 and Wsp1 in patches. Indeed, the levels of Vrp1, Myo1 and Wsp1 were significantly reduced in Bbc1 overexpressing cells (Fig. 4F–H, Fig. S4C,D). Collectively, these data support the competitive binding of Bbc1 and Vrp1 to Myo1 in cells, with Vrp1 promoting and Bbc1 reducing accumulation of Myo1 in patches.

Actin patch assembly is enhanced in S. cerevisiae cells lacking Bbc1 (Picco et al., 2018) and to an even greater degree in cells lacking both Bbc1 and the adaptor protein Sla1 (Kaksonen et al., 2005). To test whether inhibition of actin assembly by Sla1 and Bbc1 is conserved in S. pombe, we examined the effects of bbc1Δ alone, sla1Δ alone or a combination of bbc1Δ and sla1Δ on accumulation in patches of F-actin labeled with Fim1–mGFP, Crn1–GFP or LifeAct–mCherry (Huang et al., 2012; Pelham and Chang, 2001; Sirotkin et al., 2010) and NPFs Myo1, Wsp1, Vrp1 and Pan1 (Fig. 5; Fig. S5).

In agreement with the previous observations (Fig. 4), the loss of Bbc1 alone resulted in a mild increase in the accumulation of Myo1 in patches and a notable increase in accumulation of Vrp1 in patches but had no effect on accumulation of Pan1–GFP in patches (Fig. 5A–C; Fig. S5A,B). In contrast to the increased accumulation of WASp Las17 and actin in patches in S. cerevisiae bbc1Δ cells (Kaksonen et al., 2005; Picco et al., 2018), the peak levels and lifetimes in patches for mGFP–Wsp1 and F-actin markers Fim1–mGFP and Crn1-GFP were unaltered in S. pombe bbc1Δ cells (Fig. 5D,E; Fig. S5C,D). Likewise, even though F-actin levels observed with LifeAct–mCherry were inconsistent between experiments, the mean intensities across multiple experiments indicated no change in actin levels in the absence of Bbc1 (Fig. S5E,F). This observation also suggests that Bbc1 may play a role in suppressing variability of F-actin levels in patches.

Examination of sla1Δ cells revealed only modest increases in accumulation of all markers except Pan1. Compared to wild-type cells, peak patch intensities of Myo1–GFP, Vrp1–mGFP, mGFP–Wsp1, Fim1–mGFP and LifeAct–mCherry in sla1Δ cells increased by 28%, 9%, 21%, 30% and 11%, respectively (Fig. 5A,B,D,E; Fig. S5E), while the peak intensity of Pan1–GFP increased by 125% (Fig. 5C). In addition, Pan1 patch lifetime was significantly increased from 35 s to 64 s. The reasons for these dramatic effects of sla1Δ on Pan1 are not clear but a similar increase of Pan1 patch lifetime in sla1Δ cells has been reported in S. cerevisiae (Kaksonen et al., 2005).

Combining bbc1Δ with sla1Δ resulted in greatly increased accumulation of NPFs and F-actin in patches. Compared to wild-type cells, in sla1Δ bbc1Δ cells, Myo1–GFP, Vrp1–mGFP, Pan1–GFP, mGFP–Wsp1, Fim1–mGFP and LifeAct–mCherry peak patch intensities increased by 91%, 146%, 242%, 51–73%, 57% and 120%, respectively (Fig. 5; Fig. S5E). The same trends were observed with N-terminally tagged Myo1, but to a lesser extent (Fig. S5A). Combining bbc1Δ sla1Δ with vrp1Δ, but not with bzz1Δ, reduced Wsp1 accumulation in patches, indicating that Vrp1 helps retain Wsp1 in patches (Fig. 5F). The synergistic rather than additive effects of sla1Δ and bbc1Δ on assembly of NPFs and actin into patches suggest that S. pombe Sla1 and Bbc1 play redundant roles in constraining actin patch assembly.

In addition to increased accumulation of Myo1 in patches, sla1Δ and bbc1Δ sla1Δ cells displayed an unusual behavior of Myo1 at the endocytic sites. In wild-type and bbc1Δ cells, Myo1 always remained at the base of endocytic invaginations and did not internalize with vesicles (Sirotkin et al., 2005) (Fig. 6A). In contrast, in 6% of Myo1–mGFP patches in sla1Δ cells, a small portion of Myo1 moved into the cytoplasm, presumably in association with the tip of the endocytic invagination or endocytic vesicle, while a larger portion of Myo1 remained at the base (Fig. 6A,B). This partial internalization of Myo1, observed as splitting of the Myo1 signal, was significantly more pronounced in bbc1Δ sla1Δ cells where the percentage of Myo1 patches exhibiting this behavior increased from 6% to 64% (Fig. 6A,B). We suspect that the Myo1 splitting events were more evident in bbc1Δ sla1Δ cells because, compared to sla1Δ alone, Myo1 patches were brighter and internalized two times farther (Fig. 6C).

We also observed infrequently in sla1Δ and, more often, in sla1Δ bbc1Δ cells elongated actin structures that we called actin plumes (Fig. 6D,D′; Movie 1). These are similar to enlarged actin structures found in S. cerevisiae sla1Δ bbc1Δ cells (Kaksonen et al., 2005). In addition to localizing to the base of these plumes, Myo1 associated with both the tip of the actin plumes and at the middle where the plume appeared to kink (Fig. 6D, inset). In single-color imaging of mGFP–Myo1 in sla1Δ and bbc1Δ sla1Δ cells, these plumes appeared as stable cytoplasmic dots of Myo1 swinging like a pendulum from stable Myo1 platforms on the membrane (Fig. 6E–F′; Movie 2). Thus, besides suppressing actin patch assembly, Sla1 appears to play a role in regulating Myo1 positioning at the endocytic site.

Close inspection of mGFP–Wsp1 dynamics revealed that this NPF also exhibits an unusual behavior but in bbc1Δ and bbc1Δ sla1Δ rather than in sla1Δ cells. Unlike S. cerevisiae Las17 and Vrp1 (Kaksonen et al., 2003, 2005), S. pombe Wsp1 and Vrp1 normally internalize with an endocytic vesicle for a few seconds before dissipating into the cytoplasm (Sirotkin et al., 2005). However, in bbc1Δ and bbc1Δ sla1Δ cells, Wsp1 and Vrp1 patches frequently split in two, with a fraction internalizing normally and another fraction remaining at the base of invagination (Fig. 6G,H; Fig. S6A, Movie 3). 84% and 95% of mGFP–Wsp1 patches split in bbc1Δ and bbc1Δ sla1Δ cells, respectively, while mGFP–Wsp1 patches rarely split in sla1Δ and wild-type cells (Fig. 6I). Vrp1–mGFP patches split at a similar frequency; 78% in bbc1Δ, 84% in bbc1Δ sla1Δ cells, but fewer than 15% of Vrp1–mGFP patches split in sla1Δ and wild-type cells (Fig. 6J).

Since Vrp1 can bind both Myo1 and Wsp1 and thereby bridge Myo1–Wsp1 interaction (Carnahan and Gould, 2003; Sirotkin et al., 2005), and Vrp1 and Bbc1 appear to compete for Myo1 binding (Fig. 4), splitting of Wsp1 patches in bbc1Δ cells could be due to enhanced Vrp1-mediated association of Wsp1 with Myo1 at the base of the endocytic invagination (Fig. 6K). To determine whether Vrp1 mediates Wsp1 patch splitting, we examined the effects of Vrp1 loss on Wsp1 patch splitting in bbc1Δ sla1Δ cells where Wsp1 patches are brighter and splitting is easier to observe (Figs 5D and 6G). Indeed, the frequency of Wsp1 patch splitting decreased from 92% in bbc1Δ sla1Δ cells to 21% in bbc1Δ sla1Δ vrp1Δ cells (Fig. 6L). In contrast, the loss of Bzz1, which is another Wsp1 binding partner that localizes to the base of endocytic invaginations (Arasada and Pollard, 2011; Idrissi et al., 2012; Kishimoto et al., 2011; Soulard et al., 2002; Sun et al., 2006), had no effect on Wsp1 patch splitting (Fig. 6L) or Wsp1 accumulation in patches in bbc1Δ sla1Δ cells (Fig. 5F). A reduction of Wsp1 patch splitting was also observed when vrp1Δ, but not bzz1Δ, was combined with just bbc1Δ, while in vrp1Δ alone cells, as in wild-type cells, Wsp1 patches rarely split (Fig. S6B).

To test whether the residual Wsp1 patch splitting observed in bbc1Δ sla1Δ vrp1Δ cells is due to direct binding of Wsp1 to the Myo1 SH3 domain that was observed by two-hybrid analysis (Carnahan and Gould, 2003), we combined bbc1Δ sla1Δ with the deletion of the Myo1 SH3 and LCA domains to disrupt the Myo1 interaction with both Vrp1 and Wsp1. As predicted, Wsp1 patch splitting completely ceased in bbc1Δ sla1Δ myo1ΔSH3-LCA cells (Fig. 6M), suggesting that both indirect Vrp1-mediated and direct binding of Wsp1 to Myo1 are responsible for Wsp1 patch splitting in the absence of Bbc1. In contrast, deletion of the CA domain resulted in only a minor effect on splitting Wsp1 patches, providing further evidence that Wsp1 patch splitting is mediated by Myo1 SH3 domain (Fig. S6C).

The observation that mGFP–Myo1 patches internalized nearly twice as far in bbc1Δ sla1Δ cells compared to sla1Δ cells (Fig. 6C) suggested that, similar to S. cerevisiae Bbc1 (Buser and Drubin, 2013; Kaksonen et al., 2005; Picco et al., 2018), S. pombe Bbc1 might have a role in regulating the length of the endocytic invagination. To determine the effect of bbc1Δ on the internalization of endocytic structures in S. pombe (Fig. 7A), we measured the accumulation of amphiphysin Hob1 and distances traveled by endocytic adaptors End4, Ent1 and Sla1 (Fig. 7; Fig. S7). Since Rvs167, an S. cerevisiae homolog of Hob1, binds around the tubular invagination (Idrissi et al., 2012, 2008; Kishimoto et al., 2011; Kukulski et al., 2012; Picco et al., 2015), accumulation of Hob1 provides a measure of the length of endocytic invagination. The three adaptors associate with the tip of invaginations (Boettner et al., 2011; Idrissi et al., 2012; Picco et al., 2015; Skruzny et al., 2012); therefore, the distance traveled by these markers represents the depth of endocytic internalizations. Compared to wild-type cells, Hob1–mGFP patch accumulation increased by 88% in bbc1Δ cells and 77% in bbc1Δ sla1Δ cells but it did not increase in sla1Δ cells (Fig. 7B,C). Additionally, the average distance traveled by Hob1–mGFP patches doubled in both bbc1Δ and bbc1Δ sla1Δ mutants compared to patches in wild-type and sla1Δ cells (Fig. 7B). Similarly, End4–mGFP and Ent1–mGFP patches internalized ∼2 times farther in bbc1Δ and bbc1Δ sla1Δ cells than in wild-type and sla1Δ cells: ∼1 µm in bbc1Δ and bbc1Δ sla1Δ cells versus ∼0.5 µm in wild-type and sla1Δ cells (Fig. 7D,E). The increased distance measured for End4–mGFP patches is not a result of the ability to track brighter patches longer since End4–mGFP patches had similar intensities and lifetimes in wild-type and bbc1Δ cells (Fig. S7A). Likewise, another invagination tip marker, Sla1, moved twice as far in bbc1Δ cells compared to wild-type cells (Fig. S7B). Collectively, these measurements are consistent with a 2-fold increase in the length of endocytic invaginations resulting in a 2-fold deeper internalizations of the endocytic structures in the absence of Bbc1. We hypothesized that this increase may be due to a prolonged association of Myo1 with Vrp1 and Wsp1, the main NPF at the endocytic site (Berro et al., 2010), at the base of endocytic invaginations. In agreement with this hypothesis, the distance traveled by the tip marker End4–mGFP returned to the normal 0.5 µm when bbc1Δ or bbc1Δ sla1Δ were combined with myo1ΔSH3-LCA (Fig. 7F), a mutant that disrupts Wsp1 retention at the base of invaginations (Fig. 6M). In contrast, End4–mGFP continued to internalize twice as far in bbc1Δ myo1ΔCA and bbc1Δ sla1Δ myo1ΔCA strains (Fig. 7G). Therefore, we conclude that the deeper internalizations observed in the absence of Bbc1 are mediated by the Myo1 SH3 domain, likely through the retention of Wsp1 and Vrp1 at the base of endocytic invaginations.

We initially identified the previously uncharacterized S. pombe protein Bbc1 in pulldowns of CR proteins. However, subsequent analysis revealed that Bbc1 does not localize to the CR, instead localizing to dynamic actin patches. While Bbc1 was originally discovered in S. cerevisiae as a myosin-1-tail-binding protein (Mochida et al., 2002), most of the subsequent work (Kaksonen et al., 2005; Picco et al., 2018) focused on the ability of Bbc1 to bind (Tong et al., 2002) and inhibit NPF activity of budding yeast WASp Las17 (Rodal et al., 2003). Thus, aside from mild inhibition of myosin-1 NPF activity (Sun et al., 2006), the significance of Bbc1 interaction with myosin-1 remained largely unknown. By taking advantage of the distinct behavior of Wsp1–Vrp1 and Myo1 in S. pombe, we provide evidence that the interaction of Bbc1 with myosin-1 is important for the role of Bbc1 as a negative regulator of endocytosis. Specifically, we propose a model (Fig. 8) where association of Bbc1 with Myo1 tail limits localization of Wsp1 and Vrp1 at the base of endocytic invagination, thereby preventing excessive tubulation of endocytic invaginations.

In S. pombe, Myo1 and Wsp1 are recruited to endocytic sites independently of one another but can interact with each other directly or indirectly via Vrp1. Vrp1 is recruited to patches by Wsp1 but also binds Myo1 (Carnahan and Gould, 2003; Sirotkin et al., 2005) and therefore can bridge Myo1–Wsp1 interaction. We hypothesize that these interactions of Myo1 with Wsp1 and Vrp1 take place before the initiation of endocytic internalization and help retain both Myo1 and Wsp1–Vrp1 in patches. Indeed, accumulation of Myo1 is reduced in the absence of Vrp1 and slightly increased in the absence of Bbc1, which also increases Vrp1 accumulation in patches. This interaction is likely transient since upon initiation of patch internalization, which occurs 2 s after Myo1, Wsp1 and Vrp1 reach peak assembly, Wsp1 and Vrp1 internalize with the patch, while Myo1 remains at the base of the invagination (Sirotkin et al., 2005, 2010).

Collectively, our results support a simple model based on competition of Bbc1 with Wsp1 and Vrp1 for the Myo1 tail (Fig. 8). Bbc1 completely depends on Myo1 for localization to patches, and specifically requires the Myo1 TH2 and SH3 domains. Bbc1, Vrp1, and Wsp1 but not other patch components associate with the Myo1 SH3–LCA fragment in our in vivo binding assay. In the absence of Bbc1, Vrp1 shows increased accumulation in patches, and Bbc1 shows increased accumulation in patches in the absence of Vrp1. Conversely, overexpression of Bbc1 reduces the levels of Vrp1, Wsp1 and Myo1 in patches, presumably by disrupting the interactions of Myo1 with Vrp1 and Wsp1. Most significantly, in the absence of Bbc1, Vrp1 and Wsp1 show an unusual splitting behavior where a portion of Vrp1 and Wsp1 remains with Myo1 at the base of invagination. Since this splitting behavior was reduced or eliminated in the absence of Vrp1 or the Myo1 SH3 domain, respectively, the splitting of Wsp1 and Vrp1 patches is likely due to enhanced interaction of Vrp1 and Wsp1 with Myo1 in the absence of competition from Bbc1. Future biochemical work is needed to verify this competition model with purified proteins, which is hindered by the large size and proline-rich nature of Bbc1.

Enhanced accumulation of Bbc1 in the absence of Vrp1 and vice versa strongly support our proposed competition mechanism. However, measured stoichiometry and an observed 40% increase in Bbc1 in patches in vrp1Δ suggest that at peak only ∼70 out of ∼360 Myo1 molecules are available for binding Bbc1 and Vrp1; 50 are occupied by Bbc1 and 20 by Vrp1. Then, in the absence of Bbc1, Vrp1 can access an additional 50 sites, increasing peak accumulation of Vrp1 over the normal number of 130 molecules (Sirotkin et al., 2010), in line with the observed ∼50% increase. The access to Myo1 tail may be limited by binding of other ligands or by intramolecular folding. The evidence for such intramolecular folding has been provided for amoeboid (Hwang et al., 2007; Jontes et al., 1998; Lee et al., 1999), mammalian (Stöffler and Bähler, 1998) and yeast (Grötsch et al., 2010) long-tailed class 1 myosins.

Competition for the Myo1 SH3 domain may also explain an unusual internalization of a portion of Myo1 observed in sla1Δ. We hypothesize that this behavior can be due to Myo1 binding to a proline-rich ligand, such as Wsp1 normally bound by the Sla1 SH3 domains.

In S. cerevisiae, Bbc1 and Sla1 both inhibit Las17 in vitro (Rodal et al., 2003) and bbc1Δ sla1Δ cells show enhanced actin patch assembly, sometimes resulting in the formation of actin comet tails (Kaksonen et al., 2005). In S. pombe, either bbc1Δ or sla1Δ alone had no or only minor effects on the extent of actin patch assembly but together resulted in a 2-fold increase in actin assembly, consistent with the proposed role of Bbc1 and Sla1 as NPF inhibitors. However, this increase of actin patch assembly in bbc1Δ sla1Δ cells was accompanied by ∼2–3-fold increased accumulation of all NPFs, raising the possibility that the enhanced actin assembly may be the result of a greater concentration of NPFs in patches rather than loss of inhibition of NPF activity. Alternatively, enhanced patch accumulation of Myo1 and Wsp1 in bbc1Δ sla1Δ cells can be explained by a hypothetical positive-feedback loop linking actin assembly and NPF accumulation.

Unlike S. pombe Wsp1, S. cerevisiae WASp Las17 remains at the base of the endocytic invagination and does not internalize with the patch (Kaksonen et al., 2003, 2005; Picco et al., 2015). However, in S. cerevisiae bbc1Δ cells, as in S. pombe, patches of the WASp homolog Las17 split and endocytic vesicles internalize deeper (Kaksonen et al., 2005). While this behavior has been attributed to possible interactions between Las17 and coat proteins in the absence of Bbc1 (Kaksonen et al., 2005), Las17 splitting can also be explained by competition of Wsp1 and Bbc1 for the myosin-1 tail, similar to the mechanism we propose for S. pombe Wsp1. A recent super-resolution microscopy study (Mund et al., 2018) revealed that Las17 localizes in a donut-like pattern surrounding the endocytic coat and, in bbc1Δ cells, spreads to a wider area normally occupied by the myosin-1 Myo5 and Bbc1, suggesting increased association of Las17 with Myo5 in the absence of Bbc1. This, coupled with a larger membrane surface area internalized in bbc1Δ cells (Picco et al., 2018), suggests that the internalizing portion of splitting Las17 patches may represent increased internalization of the Las17-coated membrane area that is not normally internalized, while the non-internalizing portion represents Las17 associated with Myo5 bound at the wider ring.

S. cerevisiae Bbc1 was also shown to recruit the Tda2–Aim21 complex, which regulates capping protein recruitment (Farrell et al., 2017; Shin et al., 2018). However, this role of Bbc1 has not yet been explored in S. pombe.

In S. pombe, as in other organisms, actin polymerization provides force for overcoming membrane tension during endocytosis (Basu et al., 2014). In wild-type cells, prior to initiation of an endocytic invagination, Wsp1 and Myo1 colocalize at endocytic patches where they interact directly and indirectly via Vrp1 (Carnahan and Gould, 2003; Sirotkin et al., 2005). Vrp1 activates Myo1 NPF activity (Sirotkin et al., 2005), while Bzz1 stimulates Wsp1 NPF activity (Arasada and Pollard, 2011). As the invagination forms, Wsp1 and Vrp1 internalize, while Myo1 and Bzz1 remain on the membrane at the base of the invagination (Sirotkin et al., 2005). One model of force generation suggests that this separation results in two separate zones of branched actin assembly that push off of each other (Arasada and Pollard, 2011; Arasada et al., 2018).

An alternative model proposes that most of the force is generated in one zone at the base of the emerging invagination (Picco et al., 2015). Several lines of evidence favor this latter model. First, Wsp1 is a much stronger NPF than Myo1, which needs stimulation by Vrp1 (Sirotkin et al., 2005). Second, Wsp1 reaches peak accumulation 2 s before the onset of invagination, when Wsp1 and Myo1 still colocalize in the same zone (Sirotkin et al., 2010). Further, movement of Wsp1–Vrp1 with the patch would separate Wsp1 and Myo1 from their respective activators, Bzz1 and Vrp1.

In addition to splitting of Wsp1 and Vrp1 patches, endocytic patch proteins internalize twice as far in bbc1Δ cells. We hypothesize that the prolonged localization of Wsp1 and Vrp1 at the base of the invagination leads to continued stimulation of Wsp1 by Bzz1, further stimulation of Myo1 NPF activity by Vrp1, and orientation of the growing actin filament ends towards the plasma membrane and near Myo1, where Myo1 can exert an additional force driving membrane invagination. We propose that this membrane-directed actin assembly is ultimately responsible for the increased distance traveled by endocytic structures observed in bbc1Δ cells either through elongated invaginations or increased propulsion of endocytic vesicles away from the membrane.

In summary, we have provided evidence that the SH3 domains of endocytic proteins compete for their ligands and this competition plays an important role in regulating actin assembly. These observations underscore the importance of future work towards dissecting the complex network of SH3 domain interactions at endocytic sites (Fernandez-Ballester et al., 2009; Tong et al., 2002; Tonikian et al., 2009) and the role of these interactions in orchestrating the precise assembly of proteins driving endocytic internalization.

S. pombe strains used in this study are listed in Table S2. LifeAct–mCherry was originally described in Huang et al. (2012). All other yeast strains were constructed by PCR-based gene tagging and genetic crosses. Deletions or fluorescent protein tags marked with selectable markers were introduced into endogenous chromosomal loci by homologous recombination of PCR-amplified gene tagging modules (Bähler et al., 1998; Wach et al., 1994). Amplified modules were transformed into cells by lithium acetate method (Keeney and Boeke, 1994) and successful integrants were isolated based on selectable markers and confirmed by PCR and sequencing.

To tag proteins at the C-termini with fluorescent protein tags, the stop codons of the targeted genes were replaced with gene tagging cassettes PCR-amplified from pFA6a-mGFP-kanMX6 (pJQW 85-4) for monomeric (m)GFP (Sirotkin et al., 2010; Wu et al., 2003), pFA6a-mCherry-kanMX6 (pKS390) or pFA6a-mCherry-natMX6 (pKS391) for mCherry (Snaith et al., 2005), and pFA6a-mNeonGreen-kanMX6 for mNeonGreen (Shaner et al., 2013; Willet et al., 2015). All C-terminally tagged proteins were expressed from endogenous loci under control of the native promoters.

Gene deletions were constructed by replacing gene open reading frames with the ura4⁺ cassette amplified from KS-ura4 (Bähler et al., 1998) for bbc1Δ::ura4⁺ and vrp1Δ::ura4⁺, kanMX6 cassette amplified from pFA6a-kanMX6 (Bähler et al., 1998) for bbc1Δ::kanMX6 and bzz1Δ::kanMX6, and natMX6 cassette amplified from pCR2.1-nat (Sato et al., 2005) for sla1Δ::natMX6.

The C-terminal truncations of Myo1 were constructed by replacing sequences encoding domains targeted for deletion with a stop codon followed by the ura4⁺ cassette amplified from KS-ura4 (Bähler et al., 1998). The unmarked myo1ΔTH2 allele was constructed by replacing the ura4⁺ cassette of myo1ΔTH2-SH3-LCA allele with the SH3-LCA-stop codon sequence amplified by PCR with primers flanked by 70 nt regions of homology to the end of TH1 domain and myo1⁺ 3′ UTR. The myo1ΔTH2 transformants were isolated by counterselection on EMM plates containing 2 mg/ml 5-fluoroorotic Acid (5-FOA) (Stark et al., 2013). The Myo1 domain boundaries were defined as previously described (Lee et al., 2000; Sirotkin et al., 2005): TH2, amino acids (aa) 967–1112; SH3, aa 1113–1163; L, aa 1164–1186; and CA, aa 1187–1217.

A gene expressing N-terminally tagged mCherry–Cdc15 from the native locus under control of endogenous Pcdc15 promoter was constructed by replacing the ura4⁺ cassette in a cdc15⁺/cdc15Δ::ura4⁺ diploid strain with the cassette consisting of 500 bp of the 5′ cdc15⁺ flanking region, the mCherry coding sequence, a GGGGSGGGGSG linker sequence, the cdc15⁺ coding sequence, and 500 bp of the 3′ cdc15⁺ flanking region. Haploid mCherry–Cdc15 integrants were isolated by 5-FOA selection and confirmed by sequencing.

Strains combining two or more deletion or tagged alleles were constructed by standard genetic crosses on malt extract (ME) plates and tetrad dissections on YES plates followed by screening for wanted gene combinations by replica plating onto appropriate selective plates, microscopy and PCR diagnostics. In some cases, myo1Δ and wsp1Δ cells were transformed with pUR-myo1⁺ and pUR-wsp1⁺ plasmids (Lee et al., 2000), respectively, which facilitate mating and are subsequently lost upon sporulation.

The Ppc89-Linker-mCherry cassette flanked by XhoI and NheI sites in pCR-BluntII-TOPO vector (Invitrogen) was created by In-Fusion cloning using ClonTech In-Fusion HD Cloning Kit (Takara Bio, USA) following the manufacturer's instructions. Briefly, a linear fragment encoding a start codon (bold), aa 261–783 of Ppc89, and a 5 aa GSGSG linker (italicized) was PCR-amplified from pREP81-ppc89 f5/r3 with the primers Ppc89F (5′-CCTTCACTCGAGATGACAATAGATGAAAACGGTAATGC-3′) and Ppc89R (5′-CTCGCCCTTGCTCACACCAGAACCAGAACCGCTAAAATCCTTACCAACGTTG-3′). The resulting Ppc89-Linker product was fused by recombination via the 15 nt homology regions (underlined in primer sequences) to the N-terminal sequence of previously cloned mCherry in pCR-BluntII-TOPO vector that was linearized by PCR amplification with the TmChF primer (5′-GTGAGGAAGGGGGAGGAG-3′) corresponding to mCherry aa 2-7 and TmChR primer (5′-CATCTCGAGTGAAGGGCGA-3′) corresponding to the start codon, XhoI site, and pCR-BluntII-TOPO MCS. The resulting Ppc89-Linker-mCherry fragment flanked by XhoI and NheI sites was cleaved from pCR-BluntII-TOPO vector and cloned in place of mGFP in the previously made pSGP573-3xPnmt1-mGFP-L-23A and -3A constructs (Siam et al., 2004) digested with XhoI and NheI. In the resulting constructs, sequences encoding the Myo1 TH2–SH3–LCA (23A) fragment (aa 967–1217) or the SH3-LCA (3A) fragment (aa 113–1217) are fused to the C-terminus of Ppc89-Linker-mCherry via a 7-aa linker with the sequence GASGTGS, and the expression is controlled by the thiamine-repressible 3×Pnmt1 promoter (Maundrell, 1990).

To make Ppc89-mCherry-LCA construct, the pSGP573-3×Pnmt1-Ppc89-L-mCherry-L-23A plasmid was linearized by PCR amplification with primers flanking and thereby removing the TH2–SH3–LCA sequence: the mChGSr primer (5′-GGATCCAGTACCGCTAGCA-3′) and SPBNot1d primer (5′-TGAGCGGCCGCTCTAGGT-3′). The Myo1 LCA sequence was amplified from genomic DNA with primers SPBm1LCAd (5′-AGCGGTACTGGATCCAAAGGTTCTACACCTCAAACCACC-3′) and SPBm1LCAr (5′-TAGAGCGGCCGCTCACCAATCTTCTTCTTCATCACT-3′). The Myo1 LCA sequence was fused in place of TH2–SH3–LCA sequence in linearized pSGP573-3xPnmt1-Ppc89-L-mCherry-L-23A plasmid by recombination via the 15 nt homology regions (underlined in primer sequence) using the In-Fusion HD Cloning Kit (Takara Bio).

Untagged Bbc1 overexpression constructs were made by subcloning the bbc1⁺ ORF into the XhoI and XmaI sites of the pREP3x vector (Forsburg, 1993). To N-terminally tag Bbc1 with mEGFP, the Bbc1 overexpression construct was linearized with XhoI and a fragment encoding mEGFP followed by a GGGGSGGGGSG linker fused in through Gibson cloning (Gibson et al., 2009).

All plasmid constructs were verified by sequencing and then electroporated into cells followed by selection on appropriate EMM plates containing 5 µg/ml thiamine and lacking uracil for pSGP573 constructs or leucine for pREP3x constructs.

Cells were grown to log phase at 25°C in YES medium. 4×10⁷ cells were resuspended in 1 ml YES and then serially diluted 10-fold four times. Then, 2 µl of each dilution were spotted onto YES plates and, for drug sensitivity assays, YES plates containing 0.25 µM latrunculin A (Cayman Chemical) or 10 µg/ml benomyl (Sigma). Plates were incubated at different temperatures as indicated on Fig. S1.

Except for Figs 1A–D, 2B,C and Fig. S2C, imaging was performed on a UltraView VoX Spinning Disc Confocal System (PerkinElmer) mounted on a Nikon Eclipse Ti-E microscope with a 100×/1.4 NA PlanApo objective, equipped with a Hamamatsu C9100-50 EMCCD camera, and controlled by Volocity (PerkinElmer) software. Stably integrated S. pombe strains were grown to exponential phase [optical density at 595 nm (OD₅₉₅)=0.2–1.0] in liquid YES medium (Sunrise) while shaking in the dark at 25°C over 2 days. Cells containing protein expression plasmids with the 3×Pnmt1 promoter were grown overnight at 25°C in liquid EMM medium (Sunrise) lacking uracil (for pSGP573 constructs) or leucine (for pREP3x constructs) and containing 5 µg/ml thiamine to suppress premature protein overexpression, washed next day three times with EMM lacking thiamine, and grown overnight at 25°C for 15–20 h in the absence of thiamine to induce protein expression in liquid EMM medium lacking uracil (for pSGP573 constructs) or leucine (for pREP3x constructs). For microscopy, cells from 0.5–1 ml of culture were collected by a brief 1-min centrifugation in a microfuge (2,000 g), washed with EMM, pelleted, and 5 µl of partly re-suspended pellet was placed onto a pad of 25% gelatin in EMM on a glass slide, covered by a coverslip and sealed with 1:1:1 vaseline:lanolin:paraffin (VALAP). Samples were imaged after a 5-min incubation to allow for partial depolarization of actin patches. For time-lapse imaging, single-color images were taken every second for 60–120 s and two-color images were taken every 2 s for 60–120 s. Z-series of images spanning the entire cell width were captured at 0.4 μm intervals.

Images in Fig. 1A,B, Figs S1C and S2C were acquired using a personal DeltaVision microscope system (GE Healthcare), which includes an Olympus IX71 microscope, 60×/1.42 NA PlanApo and 100×/1.40 NA UPlanSApo objectives, fixed- and live-cell filter wheels, a Photometrics CoolSnap HQ2 camera, and softWoRx imaging software. Images were acquired at 25–29°C. Images are maximum-intensity projections of z-sections spaced at 0.5 µm for Figs 1B and 1 µm for Fig. S2C that were deconvolved with 10 iterations. For images in Fig. 1A and Fig. S2C, 1 ml of log-phase cells grown overnight in YES at 25°C in a shaking water bath were collected by centrifugation (400 g for 1–4 min), resuspended in ∼10 µl YES, and spotted directly on glass slide. Time-lapse imaging in Fig. 1B was performed using an ONIX microfluidics perfusion system (CellASIC). Cells were grown and collected as for Fig. 1A except that 4×10⁷ cells were resuspended in 1 ml YES. Then, 50 µl of cell suspension was loaded into Y04C plates for 5 s at 8 psi, and YES liquid medium flowed into the chamber at 5 psi throughout imaging. Time-lapse images were obtained at 4-min intervals.

Images in Figs 1D, 2B,C were acquired with a spinning disk system, which includes a Zeiss Axiovert200 m microscope, Yokogawa CSU-22 scan head, 63×/1.46 NA PlanApochromat and 100×/1.40 NA PlanApo oil immersion objectives, and Hamamatsu ImageEM-X2 camera. For end-on imaging, cells grown to log-phase in EMM plus adenine, uracil and leucine, were loaded to vertical chambers in a 4% MAUL agarose pad and imaged every 2 min. To image patches in Fig. 2B,C, cells were grown to log-phase at 25°C in MAUL and then mounted onto a 2% MAUL agarose pad for imaging.

For visualization of nuclei and septa in Fig. S1C, cells were fixed with ice-cold 70% ethanol and incubated at 4°C for at least 30 min. After washing with potassium-free PBS, fixed cells were incubated in 1 mg/ml Methyl Blue (MB) for 30 min at room temperature. MB-stained cells were pelleted by centrifugation (400 g for 4 min), resuspended in PBS, and mixed with DAPI in 1:1 ratio immediately prior to imaging.

Image analysis was performed in ImageJ (National Institutes of Health). To make sure that different mutant backgrounds did not alter the expression levels of tagged proteins and that imaging conditions remained stable throughout an imaging session, for all strains within each dataset, we measured average background-subtracted whole-cell intensities, which correspond to the tagged protein expression levels. For each time series, we measured whole cell intensities of five cells, subtracted these values for either extracellular background or the intensities of untagged wild-type cells, and averaged the background-subtracted values at each time point. All strains within each dataset had statistically similar whole-cell intensities.

The intensities and positions of fluorescent protein-tagged markers in individual endocytic actin patches were manually tracked throughout lifetime of each patch using a circular region of interest (ROI) with a 10-pixel diameter. Mean fluorescence intensities of patches were subtracted for mean cytoplasmic intensities measured in cell areas away from patches, and distances traveled by patches from the original positions were calculated. Initiation of actin patch internalization was defined as the moment the patch shifted its position by at least 2 pixels (180 nm). Except for measurements of the number of molecules, time courses of background-subtracted intensities and distances from origin for individual patches were aligned to the time of peak intensity (time=0) and averaged at each time point. Where noted, fluorescence intensities for each condition were normalized to the average peak intensity in the wild-type (WT) cells.

For the measurements of the numbers of molecules, strains combing Fim1–mCherry and one of the mGFP-tagged patch markers were imaged in a single medial plane at the rate of 1 frame every 2 s. Whole-cell intensities and the intensities and positions of patches were measured as described above. For these measurements, the time courses of cytoplasmic background-subtracted intensities and distances from origin for individual patches were aligned to the time Fim1–mCherry patch started moving (time=0), corresponding to the initiation of patch internalization. The intensities and distance values were averaged at each time point and the fluorescence intensities were converted into the numbers of molecules using a calibration curve constructed based on the previously measured average peak numbers of molecules of End4, Pan1, Wsp1 and Myo1 in patches, reported in Fig. 7 in Sirotkin et al. (2010).

Colocalization between mGFP-tagged Bbc1 or Cdc15 and mCherry-tagged Fim1 was measured over lifetimes of Fim1 patches by visual inspection and confirmed by using the Plot Z function to detect the presence of mGFP and mCherry signals within a 10-pixel wide ROI; 30 patches were analyzed for each strain.

The percentage of patches that were splitting was measured by manually following the fates of at least 80 patches for each condition. Patch splitting events were identified visually and confirmed by measuring fluorescence intensities along line scans across patches (Fig. S6A).

Mass spectrometry analysis was performed as previously described in Roberts-Galbraith et al. (2009, 2010).

All fluorescence intensity and distance values are displayed as mean±s.d. The data are from single or, where indicated, multiple imaging experiments where all strains were imaged under identical conditions. The number of cells and patches analyzed are indicated in figure legends. The statistical significance of the data for datasets of only two conditions was tested by Student's t-test in Microsoft Excel. The statistical significance for datasets of more than two conditions was tested by one-way ANOVA with Tukey's post-hoc test in KaleidaGraph (Synergy Software). Both tests were two-tailed and the significance values are reported in figures and figure legends. The statistical significance of the patch splitting data was tested through a χ-squared analysis in Microsoft Excel.

We thank Ms Keely Macmillan for help with making Myo1 C-terminal truncations, Ms Elisabeth Barone for help with making strains and tracking patches for counting molecules, and the anonymous reviewers for helpful suggestions.