The cell polarity proteins Boi1p and Boi2p stimulate vesicle fusion at the plasma membrane of yeast cells

The fusion of a secretory vesicle with the plasma membrane (PM) can be conceptually divided into the four discrete steps of vesicle trafficking, membrane tethering, docking and the actual fusion of the two membranes. Each step is thought to require the elaborate and coordinated activities of a suite of proteins whose identities, specific roles and mechanisms of actions are still the subject of intense study (Rizo and Südhof, 2012; Südhof and Rothman, 2009).

In many eukaryotic cells, vesicle fusion does not occur evenly over the entire PM but rather occurs only at certain areas to promote active cell expansion and the preferred release of signaling and extracellular matrix molecules. In the budding yeast Saccharomyces cerevisiae, polar growth is achieved by directing post-Golgi vesicles to the tip of the cell. The RhoGTPase Cdc42p, the master regulator of polar growth in yeast and other eukaryotes, is at the top of the cascade of proteins that initiates and maintains the polarity of the cell (Bi and Park, 2012). Active Cdc42p (Cdc42_GTP) is concentrated preferentially at the membrane of the growing tip and enables polarized secretion by at least two interdependent mechanisms. Cdc42_GTP activates the yeast formin Bni1p to stimulate the outgrowth of actin cables from this site (Evangelista et al., 1997). Post-Golgi vesicles then ride on these cables towards the cell tip (Donovan and Bretscher, 2012, 2015), where Cdc42_GTP binds and possibly activates Exo70p and Sec3p, two members of the exocyst protein complex (Adamo et al., 2001; Baek et al., 2010; Guo et al., 2001; Yamashita et al., 2010; Zhang et al., 2001). The exocyst belongs to the CATCHR family of complexes that tether incoming vesicles to their target membranes (Heider et al., 2016; Yu and Hughson, 2010). Its eight subunits Sec3p, Sec5p, Sec6p, Sec8p, Sec10p, Sec15p, Exo70p and Exo84p were first discovered in yeast and later found to perform similar conserved functions during exocytosis in other eukaryotes as well (TerBush and Novick, 1995; TerBush et al., 1996; Yu and Hughson, 2010). Sec10p and Sec15p are suggested to make contact with the membrane of the post-Golgi vesicle whereas Sec3p and Exo70p bind active Cdc42p and phospholipids at the PM (Boyd et al., 2004; Guo et al., 1999; He et al., 2007; Picco et al., 2017; Wiederkehr et al., 2003; Wu et al., 2008, 2010; Yamashita et al., 2010; Zhang et al., 2001).

After vesicle tethering is accomplished, the soluble Sec1p–Munc18 (SM) protein Sec1p stimulates the formation of an inter-membrane docking complex between the membrane-bound v- and t-SNAREs (soluble N-ethylmaleimide-sensitive-factor attachment receptor) Snc1/2p, Sso1/2p and Sec9p (Hashizume et al., 2009; Morgera et al., 2012). A rearrangement of the components of this docking complex, and/or the release of inhibitory factors subsequently initiates membrane fusion (Südhof and Rothman, 2009).

In addition to its role in vesicle trafficking and tethering, Cdc42_GTP might stimulate the assembly of a protein scaffold that organizes the sequence of the individual steps of exocytosis at the cortex of the bud (Liu and Novick, 2014). Defining the composition, functions and architecture of this cortical domain is challenging as it probably forms only transiently during the cell cycle and will coexist with other protein assemblies that simultaneously perform the many other activities at this location.

The homologous proteins Boi1p and Boi2p (Boi1/2p) of S. cerevisiae are localized below the bud tip and were identified as subunits of a polarity complex including Cdc42p, the GEF for Cdc42p, Cdc24p and the scaffold protein Bem1p (Bender et al., 1996). As Pob1p, the Boi1/2p homologue in Schizosaccharomyces pombe, was functionally linked to secretion, Boi1p and Boi2p are potentially important members of a Cdc42_GTP-induced active area of secretion at the cell tip (Nakano et al., 2011). Here, we characterize Boi1p and Boi2p as essential scaffold proteins for exocytosis that assemble Cdc42_GTP, Bem1p, Cdc24p, the exocyst and Sec1p into one complex at the PM. By activating Cdc42p and binding directly to the exocyst and Sec1p, this newly described complex supports a focused association of the vesicles with the PM and might stimulate the formation of the docking complex.

Boi1p and Boi2p consist of an N-terminal SH3 domain (SH3_Boi1; SH3_Boi2) a central SAM domain and a C-terminal PH domain (PH_Boi1, PH_Boi2; Fig. 1A) (Bender et al., 1996). An extended linker region between the SAM and the PH domain contains the PxxP motif (P) that binds to the second SH3 domain of Bem1p (SH3_2_Bem1) (Bender et al., 1996). Sporulation analysis of a heterozygous BOI1/Δboi1; BOI2/Δboi2 strain confirms that the presence of either BOI1 or BOI2 is necessary for cell survival (Fig. 1B). The Δboi1Δboi2 strain can be rescued by the expression of fragments of the proteins harboring either PH_Boi1 or PH_Boi2 (Fig. 1C) (Bender et al., 1996). To investigate the essential cellular functions of Boi1/2p we created a Δboi2 P_GAL1 BOI1 strain where the shift to glucose medium killed the cells by repressing the expression of Boi1p (Fig. 1D). The overexpression of Boi1p is not toxic in cells lacking Boi2p (Fig. 1D) (Bender et al., 1996). We examined Δboi2 P_GAL1 BOI1 cells by transmission electron microscopy (TEM) 12 h after the shift from galactose- to glucose-containing medium. Δboi2 cells depleted of Boi1p show a massive accumulation of cytosolic vesicles in their buds (Fig. 1E,F). The vesicles display a characteristic bilayer structure with variable diameters of 90-100 nm (Fig. 1E,G). The phenotype resembles the accumulation of post-Golgi vesicles in the late sec mutants (Novick et al., 1980). To gain independent evidence of the nature of these vesicles, we repeated the shut-off experiment with a Δboi2 P_GAL1BOI1 strain that additionally expressed a mCherry fusion to Sec4p, the yeast Rab GTPase and marker for post-Golgi vesicles (Guo et al., 1999). Cells depleted of Boi1/2p show a more intense mCherry–Sec4p staining at the bud tip than cells expressing a single copy of Boi1p (Fig. 1H,I). In contrast to wild-type cells, the mCherry signals in Δboi2 P_GAL1BOI1 cells are not confined to the cell tip but trail toward the mother cell (Fig. 1H,I). Occasionally we observed a bending of the bud. mCherry–Sec4p was concentrated below the new tip (Fig. 1H). We conclude that the mis-localization of mCherry–Sec4p mirrors the accumulation of vesicles observed in TEM images of Δboi2 P_GAL1BOI1 cells.

PH_Boi1 and PH_Boi2 bind to lipids and active Cdc42p (Bender et al., 1996; Hallett et al., 2002). Neither of these activities is specifically linked to the accumulation of post-Golgi vesicles. We thus assumed that PH_Boi1/2 might contact additional ligands that are more directly connected to exocytosis. To identify this hypothetical ligand, we searched for new interaction partners of Boi1/2p by a systematic split-ubiquitin (split-Ub) interaction screen using an array of 389 yeast cells, each expressing the N-terminal half of Ubiquitin (N_ub) as a fusion to a different yeast protein (Fig. 2A; Fig. S1) (Hruby et al., 2011; Johnsson and Varshavsky, 1994; Müller and Johnsson, 2008; Wittke et al., 1999). Cells containing the N_ub fusions were mated with a yeast strain carrying the protein under study as an N-terminal fusion to the C-terminal half of Ubiquitin (C_ub). C_ub is C-terminally extended by an N-degron (R) followed by the uracil-synthesizing enzyme Ura3p (CRU). A diploid strain carrying a pair of interacting N_ub and C_ub fusions will reconstitute a native-like Ub. The cleaved RUra3 is subsequently degraded and the cells will survive on medium containing 5-Fluoro-orotic acid (5-FOA). 5-FOA kills all other cells with remaining Ura3p activity. The array is enriched in N_ub fusions to genes known to be involved in polar growth, secretion, the cytoskeleton and stress responses (Dünkler et al., 2012; Hruby et al., 2011). As the relevant hits should play essential roles in secretion and should equally interact with both Boi1p and Boi2p in a PH-domain dependent manner, we compared the interaction profiles of the CRU fusions of full-length Boi1p and Boi2p with the profiles of their fragments lacking the PH domains (Boi1_1-733–CRU and Boi2_1-760–CRU) (Table 1; Fig. S1). N_ub–Sec1p was the only fusion to an essential component of the exocytosis pathway whose affinity to Boi1–CRU and Boi2–CRU substantially decreases upon deletion of their PH domains (Fig. 2A, Table 1; Fig. S1). The Split-Ub analysis of the CRU fusion to the isolated PH domain of Boi1p (PH_Boi1–CRU) suggests a direct binding between PH_Boi1 and Sec1p or Cdc42p, whereas the newly found interaction between Boi1p and Bud6p is restricted to regions N-terminal to PH_Boi1 (Fig. 2B, Table 1). We next reconstituted the PH_Boi1/2–Sec1p complexes with bacterially expressed fusion proteins. Enriched 6×His–Sec1p is specifically precipitated by GST–PH_Boi1 and GST–PH_Boi2 (Fig. 2C). We used Split-Ub fluorophore fluorophore analysis (SPLIFF) to determine the dynamics of the Boi1p–Sec1p interaction in living cells. SPLIFF is a recent modification of the Split-Ub technique where the C_ub is sandwiched between auto-fluorescent mCherry and GFP (CCG) (Moreno et al., 2013). Upon the interaction-induced reassociation with a N_ub fusion, the GFP is cleaved off and rapidly degraded. The subsequent local increase in the ratio of red to green fluorescence indicates where and when the interaction between both proteins took place. Yeast cells expressing Boi1–mCherry–C_ub–GFP (Boi1CCG) were mated under the microscope with cells expressing N_ub–Sec1p. The N_ub-induced conversion of Boi1CCG to Boi1CC was subsequently recorded by fluorescence microscopy during the first cell cycle of the generated diploid cells (Fig. 2D) (Dünkler et al., 2015; Moreno et al., 2013). The interaction occurs immediately after cell fusion, later at the incipient bud site and during early bud growth at the tip of the cell. The fast conversion of Boi1CCG is considerably slowed down during the growth of larger buds and during mitosis and cytokinesis (Fig. 2D). The measured interactions are specific, as a N_ub fusion to Ptc1p, a protein that does not bind Boi1p, converts Boi1CCG to levels that are hardly above background (Fig. 2D).

Our analysis suggests that PH_Boi1/2 might stimulate vesicle fusion by recruiting Sec1p to a phospholipid- and Cdc42_GTP-enriched patch of the PM. If correct, increasing concentrations of Sec1p might compensate for the depletion of Boi1/2p. We inserted the methionine-repressible P_Met17 promoter in front of the genomic loci of SEC1, of selected members of the exocyst, and of SEC4. The P_Met17 promoter is repressed at high methionine concentrations, moderately active at 70 µM and fully active in medium lacking methionine. Of all tested proteins, only the overexpression of Sec1p rescues the growth of the Δboi2 P_GAL1BOI1 cells on glucose medium (Fig. 3A). Similarly, a plasmid-borne GFP fusion to SEC1 under control of the P_Met17 promoter enables growth of Δboi2 P_GAL1BOI1 cells in medium containing glucose but no methionine (Fig. 3B). TEM analysis of these cells shows a clear reduction in the amount of accumulated vesicles when compared with the same cells but grown in the presence of methionine, or when compared with Δboi2 P_GAL1BOI1 cells expressing a GFP-containing plasmid without SEC1 (Fig. 3C). Quantifying the fluorescence intensity of mCherry–Sec4p indirectly confirms that overexpression of GFP–Sec1p partially suppresses vesicle accumulation upon depletion of Boi1/2p (Fig. 3D). Overexpression of Sec1p does not, however, reduce the occurrence of bent buds that are reproducibly encountered in Δboi2 P_GAL1BOI1 cells (Fig. 3E). This effect is not encountered in other secretion-defective mutants because sec1-1 cells, for example, never display bent buds at either the restrictive (37°C) or non-restrictive temperature (30°C; Fig. 3E) (Novick and Schekman, 1979).

Sec1p associates with the SNARE complex and supports its assembly and function during vesicle fusion (Morgera et al., 2012; Hashizume et al., 2009). Boi1/2p might stimulate the formation of the SNARE complex by presenting Sec1p to one of its subunits or to a certain intermediate that accumulates during the assembly process. The hypothesis predicts that, similar to observations with Sec1p, an increase in cellular concentration of this subunit/intermediate alleviates the loss of Boi1/2p. Indeed, overexpression of the t-SNARE Sso1p but not of the v-SNARE Snc1p supports growth of Δboi2 P_GAL1BOI1 cells on glucose (Fig. 3F).

Mutations in PH_Boi1/2 that are known to interfere with phospholipid binding disrupt the essential functions of the proteins (PH_Boi1KKTK, see also Fig. 4E; Fig. S2A) (Hallett et al., 2002). To find out whether binding to Cdc42 is equally important, we first searched for residues in PH_Boi1 that are required for complex formation with active Cdc42p. Homology modeling of the complex between PH_Boi1 and Cdc42p using the solved structure of the PH domain of the human Exo84 bound to human Rall_GTP as template identified residues at position R827, L829 and T894 on PH_Boi1/2 as potentially important for binding active Cdc42p (https://honiglab.c2b2.columbia.edu/PrePPI/) (Fig. 4A) (Jin et al., 2005). We replaced all three residues with alanines and could demonstrate that a GST fusion of the resulting triple mutant PH_Boi1RLT no longer measurably interacts with the constitutively active Cdc42_G12V (Fig. 4B). However, GST–PH_Boi1RLT, in contrast to GST–PH_Boi1KKTK, still co-sedimented with phospholipid vesicles (Fig. 4C; Fig. S2A), and precipitated purified 6×His–Sec1p with comparable efficiency as the GST fusions to wild-type PH_Boi1, PH_Boi2 or PH_Boi1KKTK (Fig. 4D) (Hallett et al., 2002). GFP–Boi1_RLT but not GFP–Boi1_KKTK supports growth of Δboi2 P_GAL1BOI1 cells on glucose (Fig. 4E). As both proteins were expressed at comparable levels, we conclude that binding Cdc42p in contrast to binding phospholipids is not essential for the role of Boi1/2p in exocytosis (Fig. S2B).

Sec1p overexpression suppressed the effects of Boi1/2p depletion (Fig. 3). This observation strongly suggests that the interaction between Sec1p and PH_Boi1/2 is functionally relevant. To further test this hypothesis we searched for Boi1/2p homologues that might still bind lipid and Cdc42p but not Sec1p of S. cerevisiae. Sequence homologues of Boi1/2p can be found in closely related yeasts, but also in S. pombe and even in fungi, such as Aspergillus nidulans (Nakano et al., 2011). Alignment of the PH domains of the Boi proteins from S. cerevisiae (PH_Boi1), S. pombe (PH_SpBoi) and A. nidulans (PH_AnBoi) identifies the conserved lipid-binding motif and the binding motif for Cdc42p in the sequences of both species (Fig. 5A). However, although detected at comparable levels in protein extracts, only Δboi2 P_GAL1BOI1 cells expressing PH_SpBoi, but not PH_AnBoi, survived on glucose-containing medium (Fig. 5B). Co-precipitation experiments prove that both PH domains show similar binding to phospholipid and Cdc42_G12V (Fig. 5C,D; Fig. S3A,B). We thus tested whether binding to Sec1p might distinguish the complementing PH_SpBoi from the non-complementing PH_AnBoi. Pull-down experiments showed a significantly reduced affinity of GST-PH_AnBoi for 6×His–Sec1p, whereas the complementing PH_SpBoi seems to bind 6×His–Sec1p slightly more strongly than PH_Boi1 (Fig. 5E, see also Fig. S3C,D).

The hammerhead morphologies occasionally displayed by Δboi2 P_GAL1BOI1 cells after switch to glucose suggests that Boi1/2p keep the area of exocytosis focused to the front end of the cells (Fig. 1H). In agreement, the depletion of Boi1/2p shifted the distribution of GFP–Sec1p from its polar location at the bud tip more towards the sides of the bud, to the tips of bent buds, or occasionally into mother cells (Fig. 6A,B). To find out whether Boi1/2p reciprocally require a functional Sec1p for their correct cortical localization, we shifted sec1-1ts cells co-expressing mCherry–Sec4p and Boi1–GFP to their restrictive temperature (Novick and Schekman, 1979). Although the accumulation of mCherry–Sec4p becomes clearly apparent after the temperature shift, a significant fraction of Boi1–GFP remained attached to the cortex (Fig. 6C). Latrunculin A treatment disrupts the actin cytoskeleton of yeast cells and thus the traffic of post-Golgi vesicles to the bud (Ayscough et al., 1997). As the drug did not measurably affect the cortical localization of Boi1–GFP or Boi2–GFP (Fig. 6D,E), we conclude that Boi1/2p do not reach the PM through vesicular traffic but instead are part of the structure that receives and processes vesicles at the cortex.

Boi1/2p bind Bem1p, and through Bem1p, indirectly bind Cdc24p (Table 1) (McCusker et al., 2007). The interaction sites for Bem1p are located N-terminally to the PH domains of Boi1/2p (Fig. 1A). The observation that the expression of isolated PH_Boi1/2 or Boi1_RTK can rescue the loss of Boi1/2p questions the functional significance of the Boi1/2p-mediated connection between exocytosis and Cdc42p-generated polarity. To uncover any potential redundancy that might exist within this connection, we created deletions of the N-terminal 131 or 216 residues of BEM1 (bem1_Δ1-131, bem1_Δ1-216) in a Δboi1Δboi2 strain ectopically expressing PH_Boi1. These deletions do not cause a severe growth defect in cells expressing the full-length Boi2p but are not tolerated in the strain harboring only PH_Boi1 (Fig. 7A). The effect is specific for the bem1 alleles as the removal of some of the other newly discovered binding partners of Boi1/2p had no measurable effect on the growth of the cells expressing only PH_Boi1 (Fig. 7A). Deletion of the first 138 residues removes the N-terminal SH3 domain of Bem1p that binds to the exocyst subunit Sec15p (France et al., 2006). The lack of this domain in otherwise wild-type cells only slightly impairs Sec15p targeting and exocyst complex formation (France et al., 2006). Accordingly, the strong effect of bem1_Δ1-131 and consequently of bem1_Δ1-216 on the growth of PH_boi1 cells reveals a functional redundancy between BEM1 and BOI1/BOI2 that might be explained by a direct physical connection between the exocyst and Boi1/2p. We thus screened members of the exocyst except SEC10 as CRU fusions against our N_ub array and discovered interaction signals between N_ub–Boi2p and all tested subunits of the exocyst. N_ub–Boi1p interacts with a smaller subset of the exocyst proteins in our Split-Ub analysis (Fig. 7B; Table S1, Fig. S4). Tonikian et al. (2009) predicted for SH3_Boi1 a binding motif in the sequence of Exo84p, and for SH3_Boi1 and SH3_Boi2, a common binding motif in the sequence of Sec3p. Indeed, a N_ub fusion to a fragment of Boi1p that lacks SH3_Boi1 (N_ub–Boi1_70-980) failed to interact with Exo84CRU, whereas the Split-Ub-based interaction signal between Boi2p and Exo84p depends on its PH domain but not on its SH3 domain (Fig. 7C). N_ub–Boi1p interacted in an SH3-dependent manner with the CRU fusions to peptides harboring the prospective binding sites of Sec3p or Exo84p (Fig. 7C). In contrast, N_ub-Boi2p did not interact, or if so, only very weakly, with the same peptides (Fig. 7C). Pull-down analysis with the bacterially expressed and enriched GST fusions to SH3_Boi1 or SH3_Boi2 confirms that SH3_Boi1 binds directly whereas SH3_Boi2 does not bind to either of the peptides (Fig. 7D; Fig. S5A). The measured K_d of approximately 1 µM for the SH3_Boi1–6×His–Exo84_121-141–AGT complex is typical for the weak affinities by which SH3 domains bind their targets (Fig. S5D) (Gorelik and Davidson, 2012).

Using GST–SH3_Boi1 as bait, we enriched from yeast extracts the direct binding partners Sec3p and Exo84p and the exocyst subunits Sec5p and Sec15p (Fig. 7E; Fig. S5B,C). As the latter do not display consensus sites for SH3_Boi1, we assume that Boi1/2p bind indirectly to Sec5p and Sec15p. This strongly suggests that Boi1p associates with the intact exocyst complex (Tonikian et al., 2009). The measured interactions between Boi1p and the exocyst are indirectly confirmed by the co-localization of Exo84–mCherry with Boi1–GFP throughout the cell cycle (Fig. 7F).

Our experiments identify Boi1/Boi2p as central scaffold proteins that bind Cdc42_GTP, its GEF, phospholipid, the exocyst and Sec1p at the PM below the tip of the cell. The binding sites for the ligands could be mapped to different domains of Boi1/2p or to different interfaces of the same domains. Although not proven, a co-existence of the different ligands in one protein complex is thus likely. The known activities of its members and its location at the cell cortex suggest that this novel protein complex plays an important role in reinforcing the maintenance of polarity through the fusion of post-Golgi vesicles at restricted sites of the PM. Indeed, depletion of Boi1/2p leads to the accumulation of post-Golgi vesicles in the buds of the cells.

The PH domains of Boi1/2p carry the essential activities of the proteins and target them to the cell tip (Hallett et al., 2002). Mutations in PH_Boi that reduce lipid binding disturb the polar distribution of the proteins and interfere with their functions (Bender et al., 1996; Hallett et al., 2002). Unexpectedly, the expression of the PH domain of A. nidulans does not rescue yeast cells lacking Boi1/2p despite displaying the conserved lipid binding signature and binding equally well to lipid and Cdc42_GTP as its yeast counterparts. However, PH_AnBoi displays a much weaker affinity to Sec1p, suggesting that the newly discovered interaction between PH_Boi1/2 and Sec1p performs an important step during exocytosis. Overexpression of Sec1p increases the amount of assembled SNARE complexes in vivo (Wiederkehr et al., 2004). It was therefore revealing to discover that a rise in Sec1p concentration also partially compensates for the loss of Boi1/2p. This led us to the conclusion that Boi1/2p assist in a critical step during the ordered formation of the SNARE complex. As the t-SNARE Sso1p also suppresses the depletion of Boi1/2p we argue that this critical step involves chaperoning the Sso1p–Sec1p interaction (Fig. S6). Intriguingly, the N-terminal PH domain of Sec3p binds the active conformation of Sso1/2p and promotes its assembly into the Sec9p–Sso1/2p dimer (Yue et al., 2017). Boi1/2p interact directly with Sec3p and all three proteins bind phospholipids and active Cdc42p via their PH domain (Yamashita et al., 2010). Consequently, this small interaction network will align the two PH domain-specific ligands, Sso1/2p and Sec1p, into very close proximity to promote their assembly into a Sec9p–Sso1/2p–Sec1p trimer (Fig. S6). How and at which step the incorporation of the v-SNARE Snc1/2p occurs is not clear, especially as Snc1/2p are thought to intervene at multiple stages of this process (Carr et al., 1999; Carr and Rizo, 2010; Hashizume et al., 2009; Morgera et al., 2012). Although experimentally supported by others and us, the proposed assembly line remains hypothetical unless in vivo data on the efficiency and kinetics of SNARE complex formation become available (Yue et al., 2017).

Vesicle fusion most likely stimulates the hydrolysis of Cdc42_GTP to dissociate the exocyst from the membrane. In addition, it dilutes membrane-bound polarity factors and Cdc42_GTP through the incorporation of new membranes. The observed robustness by which polar growth is maintained is thus difficult to reconcile with a static connection between the exocyst and Cdc42_GTP. Cdc24p, the GEF for Cdc42p, binds constitutively to Bem1p and through Bem1p to Boi1/2p (Table 1) (McCusker et al., 2007). As part of the exocyst–Boi1/2p–Sec1p complex, Cdc24p is thus placed next to the site of vesicle tethering and fusion (Fig. S6). This coupling could rapidly restore depleted pools of Cdc42_GTP and regain the fusion competence of the PM at this site. The design of this CDC42_GTP-producing tethering complex guides the exocyst and Sec1p to the area of highest Cdc42p concentration and conversely attracts the source of active Cdc42p to sites where vesicle tethering and fusion occurs. Both effects might reinforce each other to contribute to a stable axis of growth.

Yeast cells display at least two populations of post-Golgi-vesicles that can be distinguished by their cargo (Harsay and Bretscher, 1995). It was shown in a parallel study that the depletion of Boi1/2p led to the specific accumulation of Bgl2p-containing post-Golgi vesicles that carry material for the new plasma membrane and the cell wall (Masgrau et al., 2017, preprint). By specifically receiving these vesicles at the cortex, Boi1/2p might be thus part of the protein network that underlies the connection between exocytosis and cellular polarity (Adamo et al., 2001). Our work further suggests that Boi1/2p play an instructive part in assembling this network through binding to the exocyst, Bem1p and Cdc42p (Table 1). Boi1p, Boi2p, and Bem1p share physical and negative genetic interactions with multiple subunits of the exocyst as well as with each other (Costanzo et al., 2010, 2016; France et al., 2006; Liu and Novick, 2014). The non-overlapping binding sites allowed us to propose that Boi1p, Boi2p and Bem1p can be found in one complex with the exocyst (Fig. S6). Accordingly, the growth of a PH_Boi1-only expressing strain is dependent on the first SH3 domain of Bem1p that binds to Sec15p. The genetic interactions thus complement our protein interaction network and locate the polarity factors Boi1/2p and Bem1p at the center of a cortical sub-domain that organizes polar secretion and cooperates directly in vesicle tethering and docking. The network is held together by a system of SH3 domain interactions that connects the exocyst subunits Sec15p, Exo84p and Sec3p with Boi1p and Bem1p, and Bem1p with Boi1p and Boi2p (Fig. S6). With a K_d of approximately 1 µM, the single interactions between SH3_Boi1 and Exo84p, and between SH3_2_Bem1 and Boi1p or Boi2p are weak (Gorelik and Davidson, 2012). However, Boi1p and Boi2p interact with each other and with themselves, SH3_1_Bem1 interacts with Sec15p, and all three proteins bind to membranes and active Cdc42p. Once combined in a network, each single weak interaction might synergistically contribute to a robust connection between exocytosis and the establishment and maintenance of polarity (Fig. S6) (Gorelik and Davidson, 2012).

Gene fusions with the coding sequences for the N-terminal 35 residues of ubiquitin (N_ub), the C-terminal 41 residues of ubiquitin (C_ub), GFP carrying a S65T exchange, mCherry, a 9×MYC or the Cherry–C_ub–GFP (CCG) module were performed as described (Hruby et al., 2011; Dünkler et al., 2012; Moreno et al., 2013; Neller et al., 2015). Specifically, BOI1-GFP, BOI1CRU or BOI1CCG were constructed by genomic in-frame insertion of the GFP, CRU or CCG module behind the genomic BOI1 ORF. In brief, a PCR fragment of 840 bp of the 3′ end of BOI1 without stop codon and containing an EagI and a SalI restriction site at the 3′ and 5′ ends, respectively, was cloned in front of the GFP, CRU or CCG module of a pRS303, pRS306, or pRS304 vector (Sikorski and Hieter, 1989). The obtained vectors were linearized using a single EcoRI site in the BOI1 sequence, and transformed into yeast cells. Colony PCR with diagnostic primer combinations was used to verify the successful integration. C-terminal 9×MYC fusions of exocyst subunits, EXO84-mCHERRY, and all C-terminal CRU fusions have been generated accordingly. The native genomic promoter sequence was replaced by P_Met17, P_Met17-GFP, or P_GAL1 through recombination with a PCR fragment generated from the pYM-N35, pYM-N37 or pYM-N22 vectors, and primers containing sequences identical to the respective genomic location at their 5′ ends (Janke et al., 2004). To express CRU fusions from centromeric plasmids, the ORFs of the respective genes or gene fragments were cloned between P_Met17 and the coding sequence for CRU on the vector CRU pRS313 using EagI and SalI restriction sites. N_ub fusions expressed from a centromeric plasmid were obtained by cloning the desired sequence in frame behind the N_ub coding sequence of the plasmid pCup1-Nui-HA kanMX. To fuse mCHERRY in front of SEC4, the ORF of SEC4 was cloned behind the sequence of P_CUP1-mCHERRY located on pCup-mCHERRY pRS313 using SalI and BamHI restriction sites. GFP or mCherry fusions to SEC1, SNC1 and SSO1 were created by inserting the ORF of the respective genes behind the P_Met17-GFP/mCherry module on a pRS315 vector using EagI and SalI restriction sites. GST fusions were obtained by placing the ORF of the respective gene or gene fragment in-frame behind the E. coli GST sequence on the pGEX-2T plasmid (GE Healthcare, Freiburg, Germany) using BamHI and EcoRI restriction sites. Fusions to the human O6-Alkyl-DNA transferase (SNAP-tag, New England Biolabs, Beverly, MA) were expressed from plasmid pAGT-Xpress, a pET-15b derivative (Schneider et al., 2013). Gene fragments were inserted in-frame into a multi-cloning site located between the upstream 6×His-tag coding sequence and the downstream SNAP-tag coding sequence. The 6×His-tag fusions were obtained by placing the ORF of the respective gene or gene fragment behind the E. coli 6×His-tag sequence on the pAC plasmid (Schneider et al., 2013). Lists of plasmids and yeast strains used in this study can be found in Tables S2 and S3. Plasmid maps are available upon request.

Genetic manipulations of yeast and cultivation in different media followed standard protocols (Guthrie and Fink, 1991). Medium for Split-Ub interaction tests contained 1 mg/ml 5-fluoro-orotic acid (5-FOA, Thermo Fisher Scientific, Waltham, MA, USA). All yeast strains used in this study were derivatives of JD47, a segregant from a cross of the strains YPH500 and BBY45 (Dohmen et al., 1995). Gene deletions were performed by one-step PCR-based homologous recombinations using pFA6a-hphNT1, pFA6a- natNT2, pFA6a-kanMX6 and pFA6a-CmLEU2 as templates (Bähler et al., 1998; Janke et al., 2004).

Large-scale split-Ub assays were performed using a library of 389 different yeast α-strains expressing N_ub that were mated with a single a-strain expressing the CRU fusion. Matings and replica plating on different media were performed with a RoToR HDA stamping robot (Singer Instruments, Somerset, UK) as described (Hruby et al., 2011; Neller et al., 2015). To individually measure interactions between single pairs of CRU and N_ub fusion proteins, JD47 cells expressing the respective CRU fusion protein were mated with JD53 cells expressing the respective N_ub fusion protein. The resulting diploid cells co-expressing both fusion proteins were spotted onto medium containing or lacking 1 mg/ml FOA in 10-fold serial dilutions from OD₆₀₀=1 to 0.0001. The plates were grown at 30°C and recorded every day for 4 to 5 days (Eckert and Johnsson, 2003).

Exponentially grown yeast cell cultures were pelleted and resuspended in yeast extraction buffer (50 mM HEPES, 150 mM NaCl, 1 mM EDTA) with 1× protease inhibitor cocktail (Roche Diagnostics, Penzberg, Germany). Cells were lysed by vortexing them together with glass beads (3-fold amount of glass beads and extraction buffer to pellet weight) 12 times for 1 minute interrupted by short incubations on ice. These yeast cell extracts were clarified by centrifugation at 16,000 g for 20 min at 4°C. Extracts were probed with anti-Pho8p (D3A10, Thermo Fisher Scientific, Waltham, MA, USA; dilution 1:100) or anti-MYC antibodies, followed by goat anti-mouse IgG (Sigma-Aldrich, Steinheim, Germany; dilution 1:5000).

All proteins were expressed in E.coli cells (BL21, Amersham, Freiburg, Germany). GST and all PH domains fused to GST were expressed at 30°C for 5 h in LB medium, 1 mM IPTG. GST fusions to SH3 domains were expressed at 18°C in SB medium with 0.2 mM IPTG overnight. 6×His–Sec1p was expressed overnight at 18°C in SB medium, 0.2 mM IPTG. 6×His–Cdc42_G12V was expressed for 5 h at 30°C in LB medium, 1 mM IPTG. 6×His–Exo84_121-141–AGT, 6×His–Sec3_481-521–AGT and 6×His–AGT were expressed at 30°C in LB medium, 1 mM IPTG. Cells were pelleted, washed once with 1×PBS, resuspended in 1×PBS containing protease inhibitor cocktail (Roche Diagnostics, Penzberg, Germany) and lysed by lysozyme treatment (1 mg/ml, 30 min on ice), followed by sonication with a Bandelin Sonapuls HD 2070 (Reichmann Industrieservice, Hagen, Germany). Extracts were clarified by centrifugation at 40,000 g for 10 min at 4°C. 6×His–Sec1p and 6×His–Cdc42_G12V were purified by metal affinity (HisTrap, GE healthcare, Freiburg, Germany) followed by size exclusion chromatography (Superdex200, GE healthcare) in PBS or Tris buffer (50 mM Tris-HCl, 50 mM NaCl, 10 mM MgCl₂, pH 8). 6×His–Exo84_121-141–AGT was purified by metal affinity- followed by anion exchange chromatography (ResourceQ, GE healthcare, Freiburg, Germany) in Tris buffer (50 mM Tris-HCl, 50 mM NaCl, pH 7.5). 6×His–Sec3_481-521–AGT was directly used after metal affinity chromatography and buffer exchange to PBS through PD10 column chromatography (GE Healthcare, Freiburg, Germany). AGT was purified by metal affinity chromatography and preparative size exclusion chromatography. Before use, the monomeric and aggregate-free state of all purified 6×His-tagged fusion proteins were verified through analytical size exclusion chromatography (Superdex600, GE Healthcare, Freiburg, Germany) in PBS.

GST-tagged proteins were immobilized on glutathione–Sepharose beads (GE Healthcare, Freiburg, Germany) directly from E. coli extracts. After incubation for 1 h at 4°C with either yeast extracts or purified proteins under rotation at 4°C, the beads were washed three times with the respective buffer. Bound material was eluted with GST elution buffer (50 mM Tris, 20 mM reduced glutathione) and analyzed by SDS-PAGE followed by Coomassie Blue staining and immunoblotting with anti-His (Sigma-Aldrich, Steinheim, Germany; dilution 1:5000) or anti-MYC antibodies (Hruby et al., 2011).

1,2-Dipalmitoyl-xn-glycero-3-phosphoserine (DPPS; Echelon Biosciences, Salt Lake City, UT, USA) was dissolved in chloroform at a concentration of 10 mg/ml. For liposome preparation, lipids were dried down under a stream of nitrogen gas. From the resulting lipid film, a suspension of vesicles was generated using bath-type sonication in PBS. For spin-down experiments lipids (final concentration: 0.25 mg/ml) and purified GST fusion proteins (final concentration: 0.25 μM) were mixed in a centrifuge tube and centrifuged for 1 h at 500,000 g at 4°C. Supernatants and pellets were boiled in Laemmli buffer and analyzed by western blotting with anti-GST antibodies (Sigma-Aldrich, Steinheim, Germany; dilution 1:5000).

Δboi2 P_GAL1BOI1 cells (with or without additional plasmids) were grown for 10 or 12 h in liquid SD medium containing either galactose or glucose. Cells were washed three times with PBS and fixed with 2% glutaraldehyde (1 h) followed by 2% osmium tetroxide (ChemPur GmbH, Karlsruhe, Germany). After dehydration in increasing ethanol concentrations (30%, 50%, 70% and 90%) cells were incubated with uranyl acetate (Merck, Darmstadt, Germany) for 30 min. After three wash steps with ethanol, samples were incubated in increasing concentrations of Spurr resin dissolved in propanol (2:1, 1:2, 1:3 for 1 h each step) (Polysciences Inc., Warminster, PA, USA). Then, the samples were incubated in Spurr resin overnight. Resin polymerization occurred at 60°C for 48 h. Ultrathin sections (about 70 nm) were cut with a Leica Ultracut UCT ultramicrotome using a diamond knife (Diatome, Biel, Switzerland). Sections were analyzed with a Jeol 1400 TEM (Jeol, Tokyo, Japan) and the images were digitally recorded with a Veleta camera (Olympus, Münster, Germany).

Fluorescence microscopic images were generated with an Axio Observer spinning-disc confocal microscope (Zeiss, Göttingen, Germany) equipped with an Evolve512 EMCCD camera (Photometrics, Tucson, USA), a Plan-Apochromat 63×/1.4 oil DIC objective, and 488 nm and 561 nm diode lasers (Zeiss, Göttingen, Germany). Images were analyzed with the ZEN2012 software package (Zeiss). SPLIFF analysis was performed with a DeltaVision system (GE Healthcare, Freiburg, Germany) provided with an Olympus IX71 microscope (Olympus, Münster, Germany) equipped with a CoolSNAP HQ²-ICX285 or a Cascade II 512 EMCCD camera (Photometrics, Tucson, USA), a 100× UPlanSApo 100×1.4 Oil ∞/0.17/FN26.5 objective (Olympus, Münster, Germany), a steady-state heating chamber and a Photofluor LM-75 halogen lamp (Burlington, VT, USA).

Yeast cells were prepared for microscopy as previously described (Schneider et al., 2013). In brief, an overnight culture (grown in liquid SD medium) was diluted 1:15 in 3 ml fresh SD medium and incubated for 3 to 5 h to mid-log phase. Cells were centrifuged briefly and resuspended in 50 μl fresh medium. Then, 3 μl of this suspension was transferred to a microscope slide, covered with a glass coverslip and analyzed under the microscope. For time-lapse analysis, cells were immobilized with a coverslip on custom-designed glass slides containing solid SD medium with 1.8% agarose.

Microscopy files were analyzed and processed using ImageJ64 1.45 s (US National Institute of Health). All images were acquired as adapted z-series and projected to one layer. For the quantitative comparison of the mean fluorescent intensities in daughter and mother cells (I_D/I_M), randomly selected areas around the cells were measured as background and subtracted from the respective intracellular intensities.

mCherry-Sec4p fluorescence intensity profiles were measured using the ImageJ plot profile tool. All fluorescence quantifications were performed with z-overlays of seven stacks.

Exponentially growing yeast cells (200 μl) were treated for 12 min under shaking with 2 μl of 10 mM latrunculin A (Sigma-Aldrich GmbH, Steinheim, Germany) stock solution dissolved in DMSO. Afterwards, cells were fixed with a final concentration of 3.7% formaldehyde. Cells were washed in PBS and incubated with Alexa Fluor 594– Phalloidin (Thermo Fisher Scientific, Waltham, MA, USA) at a final concentration of 66 nM for 30 min at 4°C.

Tetrad analysis was performed as described (Neller et al., 2015). To prove the synthetic lethality between BOI1 and BOI2, deletions of each gene were generated in the haploid strains JD47 and JD53. Afterwards Δboi1 and Δboi2 cells were mated to generate heterozygous diploids. After sporulation, asci were dissected and single spores spotted onto YPD agar plates (MSM300; Singer Instruments, Somerset, UK) and grown for 4-6 days at 30°C. The allelic compositions of the spores were analyzed by exposing the cells to suitable selection media and by performing diagnostic PCRs.

To measure the K_d value of the 6×His–Exo84_121-141–AGT/6×His–SH3_Boi1 complex surface plasmon resonance measurements using a Biacore X100 system (GE Healthcare, Freiburg, Germany) were performed as described (Renz et al., 2013). In brief, purified 6×His–Exo84_121-141–AGT, covalently labeled with SNAP-Biotin (New England Biolabs) in HBSEP buffer [100 mM HEPES, 1.5 M NaCl, 30 mM EDTA, 0.5% (w/v) Tween20, pH 7.4] was captured on the surface of a CM5 chip (GE Healthcare, Freiburg, Germany) previously coated with an anti-biotin antibody (US Biologicals, Pittsburgh, PA, USA). The chip was titrated with increasing concentrations of 6×His–SH3_Boi1 in HBSEP buffer. The data were analyzed with the Biacore Evaluation Software (Version 1.1; GE Healthcare, Freiburg, Germany). The sensor chip was regenerated with 12 mM NaOH after each experiment.

Statistical data evaluations were performed using GraphPad Prism5. Student's t-tests were used to compare the percentage of cells of a certain phenotype. Mann–Whitney U-tests were applied to evaluate the significance of the differences between ratios of the mean fluorescent intensities in daughter and mother cells of different genotypes.

We thank Steffi Timmermanns, and Ute Nussbaumer for technical assistance. We thank Sebastian Heucke for performing the initial Boi1/2p interaction screens and Reinhardt Fischer (KIT, Karlsruhe, Germany) for A. nidulans genomic DNA.