The cell polarity proteins Boi1 and Boi2 direct an actin nucleation complex to sites of exocytosis in Saccharomyces cerevisiae

The directed transport of post-Golgi vesicles influences the shape of cells and forms diverse structures such as axons in animals, hyphal extensions in fungi or pollen tubes in plants. Polar growth is the result of a conserved and complex interplay between RhoGTPase-based signaling, cytoskeletal organization, and vesicular traffic and fusion. By growing exclusively through budding, yeast serves as a model organism to understand the molecular mechanisms behind polar growth (Bi and Park, 2012). The bud tip of yeast cells is the site of preferred exocytosis and hosts a complex assembly of proteins that tether secretory vesicles to the plasma membrane. These vesicles arrive on actin cables at the tip by a myosin type V (Myo2)-driven transport (Donovan and Bretscher, 2012, 2015). Myo2 is anchored to the vesicles by binding to the membrane-bound RabGTPase Sec4 and to Sec15, a member of the exocyst (Jin et al., 2011). The exocyst is an eight-membered complex belonging to the CATCHR family of protein receptors that adsorb vesicles to the membranes of their target organelles. The exocyst is bound to the vesicle and activated by Sec4_GTP (Guo et al., 1999). Upon arrival at the cortex of the tip, the exocyst attaches the vesicle to the membrane and initiates the formation of the docking complex (Boyd et al., 2004; Guo et al., 1999; He et al., 2007; Yue et al., 2017). The minimal bridge between vesicle and plasma membrane consists of plasma membrane-bound t-SNAREs, vesicle-bound v-SNAREs and the SM protein Sec1 (Hashizume et al., 2009; Morgera et al., 2012). Fusion is achieved by rearrangements of the docking complex that drive the two membranes into close apposition (Sudhof and Rothman, 2009).

The polar direction of the vesicular flow and the formation of a vesicle tethering and docking zone at its receiving are regulated by the small GTPase Cdc42. Control is exerted at least on three levels. First, linear actin cables that point to the bud tip are formed by the nucleation-promoting factor Bud6 in cooperation with the formin Bni1 (Amberg et al., 1997; Graziano et al., 2011; Moseley et al., 2004). Bud6 and Bni1 colocalize at the tip of the bud as members of the polarisome multi-protein complex (Amberg et al., 1997; Sheu et al., 1998). Full activation of this complex requires the binding of small RhoGTPases to relieve the auto-inhibition of Bni1 (Evangelista et al., 1997; Li and Higgs, 2003, 2005).

Second, Cdc42 binds to the exocyst members Sec3 and Exo70, and contributes directly to the localization and activation of the exocyst (Wu et al., 2010; Yamashita et al., 2010; Zhang et al., 2001). Finally, Cdc42_GTP binds to the scaffold protein Bem1 and the paralogous polarity proteins Boi1 and Boi2 (Bender et al., 1996). Bem1, and Boi1 and Boi2 (hereafter referred to as Boi1/2) form a complex at the cortex that recruits the Cdc42 guanine nucleotide exchange factor Cdc24, and several members of the exocyst complex (Bender et al., 1996; Kustermann et al., 2017; Liu and Novick, 2014). The C-terminal PH domains of Boi1/2 interact with phospholipids and Sec1 (Kustermann et al., 2017). These interactions were proposed to participate in vesicle tethering and the assembly of the docking complex. Consequently, the deletion of Boi1/2 leads to a massive accumulation of post-Golgi vesicle in the bud (Kustermann et al., 2017; Masgrau et al., 2017).

With vesicle diameters of 70–100 nm, each fusion alters the lipid and protein composition of the plasma membrane. The estimated consumption of one vesicle per 3 s should continuously dilute polarity and tethering factors at the tip by enlarging the membrane surface (Donovan and Bretscher, 2012). How a directional persistency of tip direct flow and vesicle fusion is maintained under these conditions remains an open question. Post-Golgi vesicles do not fuse with the plasma membrane immediately upon arrival, but stay immobile for a defined time at or close by the site of their prospective fusion. This dwell time depends, among other factors, on the vesicle-bound Myo2 and thus on the presence of an actin structure adjacent to the site of fusion (Donovan and Bretscher, 2015). It is thus possible that members of the tethering and docking machinery are linked to the cortex and might even stimulate the outgrowth of actin filament to attract the incoming vesicles to the sites of previous fusions.

Deletion of core components of the polarisome dissolves the focused distribution of Bud6 and Bni1, yet leaves actin filament formation and the delivery of vesicles less tip directed but otherwise still largely intact (Tcheperegine et al., 2005). This finding suggests the existence of additional factors that direct filament formation towards the membrane of the bud. We describe a complex consisting of Bud6, Bni1 and the vesicle-tethering factors Boi1 and Boi2, which initiates the formation of actin filaments at or close to the sites of vesicle fusion.

To identify alternative regulators of actin filament nucleation in the bud, we performed a systematic Split-Ubiquitin (Split-Ub) interaction analysis and screened Bni1 and Bud6 as C_ub–RUra3 fusions (CRU) against an array of 504 N_ub fusion proteins. The array was enriched in proteins involved in the actin cytoskeleton, vesicular trafficking and other activities of polar cell growth (Hruby et al., 2011; Johnsson and Varshavsky, 1994; Wittke et al., 1999). The Split-Ub analysis revealed specific binding partners for Bni1 or Bud6, and binding partners that interacted with both proteins (Fig. 1A; Fig. S1, Table S1). Among those were Boi1 and Boi2, two homologous proteins that perform overlapping functions in the tethering and fusion of post-Golgi vesicles at the plasma membrane (Kustermann et al., 2017; Masgrau et al., 2017). As vesicles travel along actin cables to the membrane, a better understanding of the interaction between Boi1/2 and the actin nucleation complex might reveal how cable formation might be regulated at sites of exocytosis. Boi1 and Boi2 were already shown to interact with Bud6 in vivo (Kustermann et al., 2017). As Boi1 and Boi2 perform their essential functions redundantly, we focused our molecular analysis predominantly on Boi1.

Split-Ub interaction analysis in strains lacking either BUD6 or BNI1 confirmed that both proteins interact independently of each other with Boi1, whereas N_ub fusions to the negative controls Tdh1, Guk1 and Kel1 showed no interactions (Fig. 1B–E). N_ub fusions to fragments of Boi1 or Boi2 localized the binding sites for Bud6 and Bni1 to their N-terminal 300 residues (Fig. 1C). Split-Ub analysis of CRU fusions to fragments of Bni1 placed the binding site for Boi1 within the N-terminal 854 residues of Bni1 and thus away from the C-terminally located binding site for Bud6 (Fig. 1D) (Moseley and Goode, 2005). N_ub fusions to fragments of Bud6 confined the Boi1 interaction site to the N-terminal 141 residues of Bud6 (Fig. 1E). We expressed the so-defined minimal binding fragments of Boi1 (1–203), Bud6 (1–141) and Bni1 (1–854) in Escherichia coli and could show by pull-down analysis that the interaction between Boi1 and Bni1, and that between Boi1 and Bud6, is direct (Fig. 1F). To obtain a mutation that disrupts the interaction to Bud6 without grossly disturbing the structure of Boi1, we further fine mapped the Bud6 interaction site on Boi1. A screen of N_ub fusions to different N-terminal fragments of Boi1 identified the linker region (residues 77–178) between the SRC homology 3 (SH3) domain and the sterile alpha motif (SAM) domain as an autonomous binding site for Bud6 (Fig. S2). A pull down of the purified His-tagged Bud6_1–141 with a glutathione S-transferase (GST) fusion to Boi_77–178 confirmed our analysis (Fig. S2). We deleted the Boi1-linker region in the yeast genome (boi1_Δ86–178) and tested the Boi1_Δ86–178CRU-expressing strain against the N_ub array, and the N_ub fusion to Boi1_Δ86–178 against Bud6CRU and Bni1CRU. Both analyses confirmed that Boi1_Δ86–178 had specifically lost its interaction with Bud6 (Fig. 1G,H). The interactions with other polarity proteins such as Bem1 and Sec1 remained unaffected by this deletion (Fig. 1G; Fig. S2). Importantly, the deletion did not detectably impair the interaction between Boi1 and Bni1 (Fig. 1H).

Overexpressing Boi1–mCherry leads to large unbudded cells in which the plasma membrane is decorated by Boi1–mCherry (Fig. 1J) (Bender et al., 1996). Visualizing the actin structures in these cells with Lifeact, or co-expressing GFP fusions to Bud6 or Bni1, shows that all three proteins relocate to the Boi1–mCherry-stained cortex. This observation provides the first indication that Boi1 might bind to the Bni1–Bud6 complex in its active actin filament-promoting conformation (Fig. 1I,J).

Our experiments suggest that Boi1/2 might guide the Bni1–Bud6 complex to generate actin filaments at sites of exocytosis. To corroborate our hypothesis we visualized the actin cytoskeleton in boi1Δboi2Δ cells expressing fragments of Boi1 of increasing length (Fig. 2A). The C-terminal PH domain of Boi1 is the minimal fragment that rescues the essential function of Boi1/2 during vesicle fusion (Kustermann et al., 2017). Cells expressing this domain (boi1_Δ414 boi2Δ) were significantly enriched in delocalized actin patches and appeared to contain fewer and thinner actin cables than the corresponding boi2Δ cells (Fig. 2A,B). This effect documents a clear impact of Boi1/2 on the actin cytoskeleton. We next correlated the presence of the Bud6 and Bni1 binding sites on the expressed Boi1 fragments with changes of the actin cytoskeleton in these cells. A deletion of the first 203 residues removes the major Bud6 binding site of Boi1 and strongly affects the binding to Bni1 (Fig. 1C). This deletion already reduced the percentage of actin cables and also increased the amount of delocalized actin patches (Fig. 2A,B). Additionally, removing the SAM domain of Boi1 (boi1_Δ299) and thus any residual interactions with Bni1 further lowered the amount of actin cables and enhanced the number of delocalized actin patches to similar levels to those found in cells expressing only boi1_Δ414 (Figs 1C and 2A,B). The impaired vesicle fusion of a boi1Δboi2Δ strain can be suppressed by overexpression of the t-SNARE Sso1 (Kustermann et al., 2017). The still-disorganized actin structure of this strain confirms that actin organization and vesicle docking are distinct activities of Boi1/2 (Fig. 2A,B). A GFP fusion to the RabGTPase Sec4 is a marker for post-Golgi vesicles, which become highly polarized in the growing bud (Jin et al., 2011). Truncating Boi1 from its N-terminus increasingly dissolved the tip-focused distribution of these vesicles (Fig. 2A,C). This observation further substantiates the role of Boi1/2 in actin organization as secretory vesicles strictly travel on actin cables (Pruyne et al., 1998).

The high density of actin patches interferes with the simultaneous detection of actin cables in the bud. We thus repeated the actin staining of boi2Δ cells carrying the minimally perturbed boi1_Δ86–178 allele after treatment with the actin patch inhibitor CK-666 (Hetrick et al., 2013). Structured illumination microscopy (SIM) of the actin cytoskeleton showed that the bud of boi1_Δ86–178boi2Δ cells was less densely filled with actin cables than that of boi2Δ cells (Fig. 2D). By applying the coefficient of variation (COV) as a quantitative measure of actin cable density, we could support the conclusion derived from the visual inspection of the cells (Fig. 2E) (Garabedian et al., 2018) (see Materials and Methods). Compared to wild-type and boi2Δ cells, boi1_Δ86–178 boi2Δ cells also displayed fewer actin cables crossing a virtual plane that was placed in the bud perpendicular to the polarity axis (Fig. 2E). In contrast, actin filament number and density were not changed in mother cells upon deletion of the Bud6 binding site in Boi1 (Fig. 2E). The polar localization of Bud6 was not affected in boi1_Δ86–178 boi2Δ cells (Fig. S3).

To measure the influence of the Boi1/2-induced actin structures on the movement and fusion of post-Golgi vesicles, we photobleached the buds of cells expressing GFP–Sec4 and compared the trajectories of individual vesicles entering the bud of wild-type, boi2Δ and boi1_Δ86–178 boi2Δ cells by time-lapse microscopy (Movie 1) (Donovan and Bretscher, 2015). Incoming vesicles in wild-type cells were often directly transported to the cell cortex, where they either tethered and subsequently fused, or moved along the cortex to the tip, where fusion occurred. Incoming vesicles in boi2Δ cells took longer to find their final destinations, and this was more pronounced in boi1_Δ86–178 boi2Δ cells (Movie 1). In particular, post-Golgi vesicles in boi1_Δ86–178 boi2Δ cells headed after their first contact with the cortex to a different region of the cortex or tumbled within the center of the bud. We occasionally observed an incoming vesicle that was redirected to the mother cell after shortly touching the cortex of the bud. Moreover, incoming vesicles seemed to reside longer at the neck before entering the bud (Movie 1).

To quantify the differences between the alleles, we measured the GFP–Sec4 fluorescence intensity of a small corridor adjacent to the plasma membrane of the bud and normalized it to the intensity of the whole bud (Fig. 3A). In wild-type cells, GFP–Sec4 was clearly restricted to the narrow zone beneath the plasma membrane. In boi1_Δ86–178 boi2Δ cells, GFP–Sec4 was more equally distributed throughout the bud. GFP–Sec4 was also slightly enriched at the bud neck of these cells (Fig. 3A).

Tracking individual post-Golgi vesicles is best achieved in medium-sized and large buds. To complement our tracking experiments, we observed GFP–Sec4 by SIM of fixed cells to look at the distribution of post-Golgi vesicles in small buds, where vesicular traffic is more tip directed. Vesicles stained the cortex in a very restricted zone at the tip of wild-type cells (Fig. 3B). This zone became slightly broader in boi2Δ cells (Fig. 3B). This trend continued in boi1_Δ86–178 boi2Δ cells in which the GFP–Sec4 staining also extended more toward the center of the bud and additionally appeared in small clusters at the bud neck (Fig. 3B).

Vesicles that reached their final destination at the cortex stayed there on average for 10.85 s before disappearing, most probably through fusion with the plasma membrane (Fig. 3C). The tethering time was significantly reduced to 7.7 s in boi1_Δ86–178 Δboi2 cells. As boi2Δ cells displayed a near-wild-type tethering time of 10.5 s, we conclude that the impaired interaction between Bud6 and Boi1_Δ86–178 causes the faster fusion of vesicles with the plasma membrane (Fig. 3C).

The Boi1/2-independent location of Bud6, and the presence of Boi1/2-independent actin structures at the cortex, prevent us from unequivocally concluding from the experiment in Fig. 1J that a Boi1/2–Bud6–Bni1 complex initiates actin filaments de novo (Fig. S3). By fusing Boi1 to Pex3, a membrane protein of the peroxisomes, we aimed to remove Boi1 from the known actin nucleation centers of the bud to study its activity in isolation at the membrane of the peroxisome (Fig. 4A) (Luo et al., 2014). Pex3_1–45–mCherry–Boi1 (Pex–Boi1) is efficiently targeted to the membrane of the peroxisomes (Fig. 4B). By co-expressing a GFP fusion to proteins involved in polar growth including known ligands of Boi1/2, we could show that Boi1 attracts all its tested binding partners to the peroxisome including Bni1 and members of the exocyst (Fig. 4C,E; Fig. S4). In contrast, polarity proteins not known to directly bind to Boi1/2 (Ste20, Cla4, Rga1, Rga2) were not enriched at the Boi1-labeled peroxisomes (Fig. 4E; Fig. S4) (Kustermann et al., 2017). GFP–Cdc42 partially relocated to Boi1-labeled peroxisomes, whereas a GFP fusion to the Cdc42- and Rac-interactive binding (CRIB) domain of Gic2 (Gic2 CRIB), a probe to sense the GTP-bound form Cdc42, remained exclusively at the bud tip (Fig. 4E; Fig. S4) (Brown et al., 1997). Although indicative, the experiment cannot definitely exclude the presence of active Cdc42 at the Boi1-labeled peroxisomes, as a detection by Gic2 CRIB not only requires Cdc42_GTP but also other features of the plasma membrane that might not be found at the peroxisome (Takahashi and Pryciak, 2007).

Actin staining revealed the establishment of an alternative axis of cell polarity in the Pex–Boi1-expressing cells (Fig. 4C,D). The actin cables of this alternative axis seem to emanate from peroxisomes located in the mother. To distinguish the contribution of the actin nucleation factors from the contributions of all other recruited proteins, we repeated the experiments with cells expressing peroxisome-targeted Boi1_Δ86–178 (Pex–Boi1_Δ86–178) lacking the binding site to Bud6. Accordingly, Bud6–GFP was no longer found at Pex–Boi1_Δ86–178-labeled peroxisomes. Bni1, Exo84, Sec3, Cdc24 and Bem1 bind Boi1 at a different site and consequently still colocalized with Pex–Boi1_Δ86–178-labelled peroxisomes (Figs 4C and 1H; Table S2). Actin cables in this strain were often less polarized towards the cell tip but did not any longer align towards the peroxisomes (Fig. 4C,D). A significant portion of cells still contained actin patches around peroxisomes (Fig. 4D).

We performed three additional experiments to better characterize the Boi1-generated actin structures around the peroxisomes. First, a GFP fusion to Bnr1, the bud neck-localized formin of Saccharomyces cerevisiae, is not enriched at Boi1-labeled peroxisomes (Fig. 5A,B). Second, a deletion of BNI1 abrogates the activity of Boi1 to initiate actin structures around peroxisomes (Fig. 5C,D). Finally, incubation with CK-666 does not impair but slightly increases the generation of the actin cables from Boi1-labeled peroxisomes (Fig. 5E,F). All three features of the peroxisomal actin structure confirm its similarity to the actin structure formed at the tip of the cells.

Deleting the region between the N-terminal SH3 domain and the SAM domain in Pex–Boi1_Δ86–178 impairs the interaction with Bud6, and might also affect other activities of Boi1 (Fig. 1). To further support the existence of functional Bud6–Boi1–Bni1 complex at Boi1-labeled peroxisomes, we compared the formation of actin cables in Pex–Boi1-expressing bud6Δ cells containing ectopic copies of GFP–Bud6 or GFP–Bud6_360–end. In contrast to the full-length protein, GFP–Bud6_360–end, lacking the Boi1 binding site (Fig. 1E) but still being able to bind to actin and Bni1, failed to support the formation of actin structures at the peroxisomes (Fig. 5G,H) (Tu et al., 2012).

The GFP fusion of the v-SNARE Snc1 (GFP–Snc1) and Sec4 were enriched at the peroxisomes of Pex–Boi1- but not of Pex–Boi1_Δ86–178-expressing cells (Fig. 6A,B). As both GFP fusions are attached to post-Golgi vesicles, their recruitment to the peroxisome indicates the reconstitution of an at least partially functional vesicle-tethering zone. Consequently, Boi1- but not Boi1_Δ86–178-labeled peroxisomes, when found in close apposition to the plasma membrane, often induce an outward bulging of the cell wall (Figs 4C, 5C,E,G and 6A).

Tracking of individual GFP–Sec4-labeled post-Golgi vesicles confirmed the formation of an alternative and functional polarity axis (Fig. 6C,D,E; Movie 2). Cells containing Boi1-decorated peroxisomes displayed a reduced flux of vesicles to the bud (Fig. 6C,D). The reduction was partially compensated by an increase in the fraction of vesicles that moved away from the bud towards the Boi1-labeled peroxisomes (Fig. 6C,E). This redirection of vesicular traffic was not seen in cells expressing Pex–Boi1_Δ86–178 (Fig. 6C-E). The measured directional vesicular traffic was strictly actin dependent (Fig. 6F).

In most eukaryotic organisms, post-Golgi vesicles arrive at the cell plasma membrane through a directed long-distance walk on microtubules or actin cables. The vesicles are then handed over to the actin filaments underlying the cortex (Hume et al., 2011; Porat-Shliom et al., 2013). The actin-bound vesicles are either kept on hold during regulated exocytosis or processed directly for docking and fusion. In budding yeast, post-Golgi vesicles are transported exclusively on actin cables to the plasma membrane of the bud (Pruyne et al., 1998). We propose that in budding yeast, similar to other eukaryotes, secretory vesicles switch from actin cables used for long-distance transport to cortical actin filaments that guide the vesicle to the docking and fusion zone. Our experiments point to Boi1/2 as contact sites for actin filament nucleation below the plasma membrane. Boi1/2 locate at the cortex and bind to Bud6 and the formin Bni1, two proteins that together form a potent actin nucleation and elongation complex (Graziano et al., 2011, 2013; Moseley and Goode, 2005). Abrogating the interaction between Boi1/2 and Bud6 reduces actin cable density in buds, increases the random movement of vesicles and shortens the tethering time of the bound vesicles. Furthermore, the artificial relocation of Boi1 to peroxisomes creates an alternative tethering zone at the peroxisome, including secretory vesicles and actin filaments that emanate from these sites. The zone is formed by the many binding partners of Boi1 and depends on the ability of Boi1 to recruit Bud6 and Bni1 to initiate actin cables (Fig. 7). The efficacy with which the peroxisome-tethered Boi1 competes with other factors in the cell for actin filament formation suggests that Boi1/2 not only anchors but also activates the Bni1–Bud6 complex upon binding. The architecture and stoichiometry of the Bud6–Boi1/2–Bni1 complex is unresolved. The cartoon in Fig. 1I is certainly an oversimplification as all three proteins form at least dimers, and Bni1 might also attach to Boi1 through its separate interaction with Bud6. The formation of a trimeric Bud6–Boi1–Bni1 complex is compatible with our experiments but other arrangements are at this point equally possible (Fig. 1I).

By focusing our analysis on the boi1_Δ86–178 allele, we tried to separate its influence on actin filament formation from the protein's two other main functions, the localization of the Bem1–Cdc24 complex and the formation of the tethering and docking complex (Bender et al., 1996; Kustermann et al., 2017). The former activity is located on a short binding motif in the middle of the sequence of Boi1, whereas the latter activity locates on the membrane- and Sec1-binding C-terminal PH domain, and the Exo84-, Sec3-binding N-terminal SH3 domain (Kustermann et al., 2017). We propose that the concentration of all three activities in one protein coordinates vesicle fusion with trafficking and enables the control of both activities through RhoGTPases (Fig. 7A). Linking vesicle tethering and fusion with actin nucleation might foster vesicle docking at sites where fusion has recently occurred and thus equip secretion with the processivity that is required for polarized growth. Two alternative non-exclusive models that could explain processivity are shown in Fig. 7A and B. Binding to Bem1–Cdc24 might channel the activated Cdc42 through the Cdc42_GTP-binding PH domain of Boi1/2 to the exocyst components Sec3 and Exo70, and to the formin Bni1. As a consequence, vesicles are not only tethered to the membrane but mark the sites where new actin filaments will be generated (Fig. 7A) (Adamo et al., 2001; Guo et al., 1999, 2001; He et al., 2007; Morgera et al., 2012; Yue et al., 2017). Alternatively, Boi1/2-containing receptor complexes might generate actin filaments at the membrane that are used as stable tracks for multiple vesicles.

The reduction in vesicle-tethering time through the dissolution of the Boi1–Bud6 complex seems counterintuitive, yet might indicate that the cortical actin in yeast, as in higher eukaryotes, not only directs movement to the membrane but also restricts and controls the fusion of the vesicles with the plasma membrane (Li et al., 2018; Meunier and Gutiérrez, 2016). It was shown that the connection between Myo2 and the vesicles has to be dissolved before fusion can occur (Donovan and Bretscher, 2015). A reduced subcortical actin network might anchor Myo2 less rigidly at the cortex and thereby increase the chance of premature vesicle fusion. Our hypothesis is supported by the phenotypes of cells carrying a BUD6 deletion. Here, the number of actin cables is reduced, but the velocity of post-Golgi vesicles, and the randomness of their movements, is increased. At the same time, exocytosis becomes less efficient in these cells (Jose et al., 2015).

Although budding yeast uses only actin structures for transport and docking, the significance of our findings is not restricted to these cells. Studies in neuroendocrine cells showed that the role of the cortical actin cytoskeleton is quite similar with respect to the coordination of exocytosis, where it also directs vesicular flow, and mediates docking and fusion (Chasserot-Golaz et al., 2005; Gabel et al., 2015). Upon stimulation, actin-associated proteins like the actin-bundling protein annexin A2 are targeted to the SNARE complex at the plasma membrane to reorganize the integrity of the cortical actin cytoskeleton and generate a vesicle fusion-promoting environment (Gabel et al., 2015; Umbrecht-Jenck et al., 2010).

The fission yeast Schizosaccharomyces pombe transports post-Golgi vesicles on microtubules. Deletion of its single Boi1/2 homolog Pop1 leads to the accumulation of secretory vesicles in the cytosol (Nakano et al., 2011). Pop1 was also shown to bind to the S. pombe formin For1. Disrupting this interaction disturbs the actin cytoskeleton (Rincón et al., 2009). Pop1 complements the essential function of Boi1/2 in vesicle fusion (Kustermann et al., 2017). These findings indicate that the molecules and mechanisms involved in the transfer of secretory vesicles from their long-distance carrier to cortical actin structures are quite conserved.

All yeast strains in this study are derivatives of the S. cerevisiae JD47 strain. Cells were incubated at 30°C in yeast extract peptone dextrose (YPD) or synthetic medium lacking specific amino acids, or complemented with antibiotics for selection. E. coli XL1 blue cells were used for plasmid amplification and grown at 37°C in lysogeny broth (LB) medium containing antibiotics. E. coli BL21 cells were used for protein production and were grown in LB or super broth (SB) medium at 37°C or 18°C.

Detailed lists of all primers, plasmids and strains from this study are provided in Tables S3, S4 and S5. Fusions of GFP or CRU to BUD6, BNI1, BOI1 or boi1_Δ86-178 were constructed by PCR amplification from genomic DNA of the respective C-terminal open-reading frames (ORFs) without stop codon as described (Dünkler et al., 2012; Wittke et al., 1999). The obtained DNA fragments were cloned via EagI and SalI restriction sites in front of the CRU, GFP or mCherry module on a pRS303, pRS304 or pRS306 vector (Sikorski and Hieter, 1989). For integration into the genome, the plasmids were linearized using a single restriction site within the C-terminal genomic DNA sequence. Successful integration was verified by PCR of single yeast colonies with diagnostic primer combinations using a forward primer annealing in the target ORF, but upstream of the linearization site, and a reverse primer annealing in the C-terminal module. Gene deletions were obtained by replacing the ORF with an antibiotic resistance cassette through single-step homologous recombination as described (Janke et al., 2004). Genomic N_ub fusions were obtained as described (Hruby et al., 2011). Generation of yeast centromeric plasmids containing N_ub fusion proteins included initial PCR amplification of indicated fragments from genomic DNA containing a SalI restriction site in the forward primer and Acc65I restriction site in the reverse primer, digestion and ligation into the plasmid N_ub-empty kanMX4.

Fragments of BUD6, BNI1 or BOI1 were expressed as GST or 6xHis fusions in E. coli BL21. GST fusions were obtained by amplification of the respective fragments from genomic yeast DNA using primers containing NcoI or EcoRI restriction sites. The PCR fragments were cloned in-frame behind GST in the plasmid pGex6P1 or pGex2T (GE Healthcare, Buckinghamshire, UK). For 6xHis-tagged fragments, a PCR of the respective fragment from genomic DNA using primers containing SfiI restriction sites was performed, and the product was inserted in-frame downstream of a 6xHis tag into the pAC plasmid (Schneider et al., 2013). The chimeric Pex3_1–45–mCherry pRS306 plasmid was adapted from Luo et al. (2014). BOI1 or boi1_Δ86–178 were amplified from genomic DNA and inserted in-frame behind the mCherry tag using BamHI or SalI restriction sites.

Genomic integration of the boi1_Δ86–178 allele was performed by ‘delitto perfetto’ methodology or CRISPR-Cas9 (Laughery et al., 2015; Storici and Resnick, 2006). The successful deletion and exchange of amino acids were confirmed by sequencing of single-colony PCRs. A detailed description of the construction of all plasmids can be obtained upon request.

Overnight cultures of E. coli BL21 cells were diluted to an optical density at a wavelength of 600 nm (OD₆₀₀) of 0.3, and incubated at 37°C in LB or SB medium to an OD₆₀₀ of 0.8 before protein synthesis was induced by the addition of isopropyl β-d-1-thiogalactopyranoside (IPTG). Protein expression conditions were optimized for each expression construct (Table S6). Cell pellets were stored after induction at −80°C.

Cell pellets were resuspended in 1× PBS or 1× HBSEP (pH 7.4, 10 mM Hepes, 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20) containing 1× protease inhibitor cocktail (Roche Diagnostics, Penzberg, Germany), incubated for 20 min with 1 mg/ml lysozyme on ice, and subsequently subjected to sonification for 2×4 min with a Bandelin Sonapuls HD 2070 (Reichmann Industrieservice, Hagen, Germany). Lysates were spun down at 40,000 g for 10 min at 4°C. Supernatants were transferred either directly to the binding assay or used for further purification.

All incubation steps were carried out under rotation in the cold room. Extracts of GST or GST fusion proteins were incubated for 0.5–1 h with glutathione-coated sepharose beads (GE Healthcare, Freiburg, Germany) equilibrated in PBS (Boi1–Bni1) or HBSEP (Boi1–Bud6). Beads were washed twice and incubated with 0.1 mg/ml bovine serum albumin (BSA) (Boi1–Bni1) (Sigma Chemicals, St Louis, MO) for 30 min, before the beads were treated with either purified 2 µM 6xHis Bud6_1–141, or extract of 6xHis Boi1_1–300 in the presence of 0.1 mg/ml BSA for 1 h. Beads were washed 3× with HBSEP or PBS before eluting the bound protein with 1× GST elution buffer (pH 8.0, 50 mM Tris, 20 mM reduced glutathione). Protein eluates were separated by SDS–PAGE and stained with Coomassie Brilliant Blue or with anti-His antibody after transfer onto a nitrocellulose membrane (Sigma-Aldrich, Steinheim, Germany; 1:5000).

Western blots were quantified with ImageJ. A detailed step-by-step procedure is provided in https://di.uq.edu.au/community-and-alumni/sparq-ed/sparq-ed-services/using-imagej-quantify-blots. Briefly, the histogram of the intensities of each band of a western blot was used to calculate the area under the curve (AUC), which correlates to the size and brightness of each band. The AUC of each band was normalized to the AUC of the input band to quantitatively compare the amount of bound 6xHis-tagged fusion protein in each lane of the gel.

For purification of 6xHis–Bud6_1–141, cell pellets were extracted as above in 1× IMAC binding buffer (pH 7.5, 300 mM NaCl, 50 mM KH₂PO₄, 20 mM imidazole). Purification was achieved by immobilized metal affinity purification followed by size exclusion chromatography on an ÄktaPurifier Chromatography System (GE Healthcare, Buckinghamshire, UK).

The final protein concentration was determined with a NanoDrop ND-1000 spectral photometer (Peqlab, Erlangen, Germany) at 280 nm excitation, and based on a calculated excitation coefficient of 8.5 mM⁻¹ cm⁻¹ and a molecular mass of 17.7 kDa (www.expasy.org). The purified protein was used directly for pull-down analysis or stored at −20°C.

For Split-Ub array analysis, a library of 533 α-strains each expressing a different N_ub fusion were mated with a Bni1CRU-, Bud6CRU-, Boi1CRU- or Boi1_Δ86–178CRU-expressing α-strain. Diploids were transferred as independent quadruplets on synthetic defined (SD) medium containing 1 mg/ml 5-fluoroorotic acid (5-FOA), and different concentrations of copper to adjust the expression of the N_ub fusions (Fig. S1) (Dünkler et al., 2012). For individual Split-Ub interaction analysis, CRU- and N_ub-expressing strains were mated or co-expressed in haploid cells, and spotted onto medium containing 1 mg/ml 5-FOA and different concentrations of copper in four 10-fold serial dilutions starting from an OD₆₀₀ of 1. Growth at 30°C was recorded every day for 2–5 days.

Fluorescence microscopy was performed on an Axio Observer Z.1 spinning-disc confocal microscope (Zeiss, Göttingen, Germany) containing a switchable Evolve512 EMCCD (Photometrics, Tucson, AZ), or an Axiocam Mrm camera (Zeiss). The microscope was also equipped with a Plan-Apochromat 100×/1.4 NA oil differential interference contrast (DIC) objective, and 488, 561 and 635 nm diode lasers (Zeiss). Images were recorded with the Zen2 software (Zeiss) and analyzed with FIJI [version 2.0.0-rc-69; National Institutes of Health (NIH) (Schindelin et al., 2012)]. Alternatively, time-lapse microscopy was performed with a DeltaVision system (GE Healthcare, Freiburg, Germany) provided with an Olympus IX71 wide-field microscope (Olympus, Hamburg, Germany). This microscope contained a CoolSNAP HQ2-ICX285 or a Cascade II 512 EMCCD camera (Photometrics), a 100× UPlanSApo 100×/1.4 NA oil ∞/0.17/FN26.5 objective (Olympus, Münster, Germany), a steady-state heating chamber and a Photofluor LM-75 halogen lamp (89 NORTH, ChromaTechnology, Williston, ND).

Secretory vesicles in the bud tip were observed with an iMIC microscope (Photonics, Pitsfield, MA) equipped with an Andor-Clara camera (Type Clara DR 328G-C01-SIL CCD), an oligochrome, a light-emitting diode (LED) camera with excitation filters FF01-340/26-25, FF01-387/11-25 and FF01-470/22-25, and a 60× oil Olympus ApoN 60×/1.490 NA oil ∞/0.13–0.19 objective. The microscope was controlled with the software Live Acquisition (v.2.6.0.34) (FEI Munich, Gräfelfing, Germany). Images were taken in five z sections (Δz=0.5 µm) with an excitation time of 80 ms and a laser intensity of 90%.

Imaging setup was as described in Rüthnick et al. (2017). In brief, Alexa Fluor 488–phalloidin-stained cells were suspended in PBS and imaged with an N-SIM system (Nikon, Tokyo, Japan) equipped with a total internal reflection fluorescence Apochromat 100×/1.49 NA oil immersion objective and a single-photon detection electron-multiplying charge-coupled device camera (iXon3 DU-897E; Andor Technology, Belfast, UK) using a 488 nm laser for excitation with an emission bandpass filter of 520/45. Reconstructions were performed with the image analysis software (Nikon). Images were taken in five z sections.

All microscopy files were processed and analyzed with FIJI [version 2.0.0-rc-69; NIH (Schindelin et al., 2012)]. Images were acquired as five to 14 z-stacks and analyzed either in single layers or as projections of single layers.

The vesicle distribution in the bud of cells expressing GFP–Sec4 was determined from SIM images with the FIJI tool ‘plot profile’ by setting a ruler from the bud neck to the bud tip. The relative length was divided into 5% steps from 0% at the bud neck to 100% at the tip. Along this line (thickness, 0.16 µm=5 pixels), the mean fluorescence intensity was determined and the background subtracted. The mean intensities along this line represent relative intensities normalized to the highest mean intensity. Ten cells with bud lengths between 2 µm and 2.5 µm were analyzed for each genotype.

Images were stacked to maximum projections. Background subtraction for the determination of actin cable numbers and actin cable staining density in bud and mother was performed with the ‘running ball’ method in FIJI with a radius of 20 pixels (=0.64 µm). To subtract the background for determination of the vesicle distribution in the bud, the highest intensity value in a vesicle-free cytosolic area of the mother was subtracted from all intensities in the bud.

Exponential-grown cells were fixed for 10 min by adding 3.7% formaldehyde to the medium. Cells were resuspended in 3.7% formaldehyde (in 100 mM KH₂PO₄) and incubated for 1 h before buffer was exchanged to 1 µM ethanolamine (in 100 mM KH₂PO₄) for 10 min. Cells were washed twice with PBS, and incubated with 66 nM Alexa-fluorophore-conjugated phalloidin (Thermo Fisher Scientific, Waltham, MA) for 30 min or overnight at 4°C. Actin patches were removed by addition of 100 µM CK-666 (Merck, Darmstadt, Germany) to the cell medium at 30°C, 10 min before cells were fixed. Samples that were imaged by SIM were prepared with buffers that were filtrated with a 0.22 µm syringe filter (#99722; TPP, Trasadingen, Switzerland) prior to usage.

Exponential-growing cells were embedded between a coverslip and a custom-made glass slide (Glassbläserei, Universität Ulm, Ulm, Germany) into a 3.8% agarose gel (containing 1× SD medium). Cells were imaged at room temperature (∼22–26°C). To keep conditions constant among different genotypes, we measured the corresponding cells in a defined order and reversed this order during the repetition. To follow individual vesicles in the bud, images (five z-stacks, 0.5 µm per stack) were taken every second. Each time-lapse series included four images that were acquired before, and 100 images acquired after, bleaching the bud of the cell. Tethering time was determined as the time interval between the arrival of a vesicle at the cortex and the time when it disappears. Only vesicles were taken into account that stayed fixed for more than 4 s at the cortex and were visible in the middle three layers.

To quantify the vesicle distribution of incoming vesicles in the bud over the whole time course of 100 images, the middle three layers were stacked to maximum projections. Maximum projections of the consecutive 100 images were further maximum projected into a single image. The mean intensity of a corridor below the plasma membrane of the bud was determined using the ‘segmented line’ tool in FIJI with a line width of 3 pixels (=0.326 µm), and normalized to the mean intensity of the whole bud. Both mean intensity values were background subtracted before calculating the ratio.

Yeast cells expressing GFP–Sec4 from a centromeric plasmid under a methionine adjustable P_MET17 promoter were grown in selective medium lacking methionine. Exponentially grown cells were spun down and resuspended in selective medium. Then, 3.1 µl of the cell suspension was embedded between a glass microscopy slide and a cover slip. Microscopy was performed at room temperature. Images were taken with a confocal microscope over a time course of 35 s using a single z-stack and acquiring images every 100 ms (intensity of the 488 nm laser, 20%; excitation time, 50 ms). Before the start of each time lapse, an image of the green and red channel was taken to reconstruct the position of peroxisomes in the trajectory plots for post-analysis. The bud of the cells and vesicle-dense regions (in Pex–Boi1) in the cortex of the mother were subsequently bleached to increase the signal-to-noise ratio of single vesicles in the mother and daughter cell.

Time-lapse images were analyzed with the MOSAIC suite plugin for Fiji/ImageJ (Sbalzarini and Koumoutsakos, 2005; Schindelin et al., 2012). To observe and follow vesicles over time, we adjusted the following parameters, whereas all other variables were left in default mode: for particle detection: particle radius, 0.133 µm; percentile (r), 0.3–1.5 (value was adjusted and vesicle detection was verified in randomly chosen images of the time lapse by using the ‘preview detected’ function). For particle linking: link range, five images; dynamics, Brownian; displacement, 1.33 µm/100 ms.

Directionality analysis based on microscopy coordinate measurements of trajectories was performed in R statistical package (R Core Team, 2013). For each cell, we first fitted a linear function delineating the mother cell and the bud. This function was later used to select trajectories originating from the mother cell for downstream directionality analyses. For each trajectory measured over a time span within a given cell, we calculated two Euclidean distance measurements: (1) distance between two subsequent time points, and (2) absolute Euclidean distance (displacement) from the original position. We then took the ratio of the net displacement to the cumulative distance travelled, with a maximum attainable upper limit value of 1 representing a perfectly directional trajectory moving in a straight line. For ease of interpretation, we took the log of this ratio, setting the maximum value to 0, so all variations of movement would fall in the negative range. In the final step, we measured area under the curve (AUC) of the calculated ratios along the entire length of the trajectory. To test the statistical significance of the directional movement against a random motion, we generated a null hypothesis distribution using a Monte Carlo simulation. To achieve this, coordinate measurements of each trajectory were first randomized and Euclidean and AUC measurements were calculated by following the steps discussed above. This was iterated 1000 times for each trajectory to finally generate the null distribution. Significance test of the empirical AUC was performed using the formula P=(r+1)/(N+1), where N is the total number of iterations (1000), r is the number of events in the simulation model that yielded AUC values greater than the empirical or observed AUC, and P is the calculated P-value (Davison and Hinkley, 1997). The decision on whether a given trajectory was moving either towards or away from the bud was made strictly based on its final destination as determined by the linear function delimiting the mother cell and the bud. For final percentage of directional trajectories calculations, we excluded short trajectories with lengths less or equal to 9. Statistical comparisons of directionality measurements among genotypes was performed using the non-parametric Kruskal–Wallis one-way ANOVA followed by Dunn's pairwise post-hoc test.

To quantify the number of vesicles moving towards peroxisomes, we overlaid a plot containing vesicle destinations (generated with R) with an image of both fluorescent channels taken right before time-lapse microscopy. Peroxisomes at the mother cortex are mainly immobile, allowing us to combine still images with subsequent time-lapse analysis (Fagarasanu et al., 2009). The directionality plot contained the destinations of vesicle trajectories that were calculated to move directional but away from the bud (not ending in the bud), and the destinations of all trajectories. The latter were required to overlay and correctly fit the proportions of the raw microscopy image with the directionality plot in order to locate the peroxisomes on each plot. We counted the number of trajectories moving away from the bud that colocalized or directly neighboring peroxisomal sites within each cell of a given allele.

We thank the Nikon Imaging Facility of the University of Heidelberg for granting access to SIM and its analysis; the Institute for General Physiology at Ulm University for providing introduction and access to the iMIC microscope; Ute Nussbaumer, Steffi Timmermann and Nicole Schmid for strain construction; and Drs Judith Müller, Alexander Dünkler and Reinhild Rösler for sharing constructs, advice and strains.