Bud4 mediates the cell-type-specific assembly of the axial landmark in budding yeast

In many different cell types, a variety of macromolecular structures, such as the contractile ring, are organized at specific subcellular locations. Organization of such structures and their positioning in a cell are crucial for cellular processes or function. During vegetative growth, cells of the budding yeast Saccharomyces cerevisiae choose a specific bud site depending on their cell type. Haploid a and α cells bud in the axial pattern in which both mother and daughter cells select new bud sites adjacent to their immediately preceding division sites. In contrast, diploid a/α cells bud in the bipolar pattern, in which mother cells choose a bud site adjacent to the division site or at the opposite pole and daughter cells choose a bud site at the pole distal to the division site (Chant and Herskowitz, 1991; Chant and Pringle, 1995; Freifelder, 1960; Hicks et al., 1977). These different budding patterns occur in response to cell-type-specific markers, which determine the axis of cell polarity during budding and ultimately the division plane. The Rsr1 GTPase module, which is composed of Rsr1 (also known as Bud1), its GTPase activating protein Bud2 and its GDP–GTP exchange factor (GEF) Bud5 (Bender, 1993; Bender and Pringle, 1989; Chant et al., 1991; Chant and Herskowitz, 1991; Park et al., 1993), conveys the cell-type-specific information to the downstream polarity establishment machinery including the Cdc42 GTPase (for a review, see Park and Bi, 2007).

In cells undergoing axial budding, a bud site is likely to be marked by a transient cortical marker that involves Bud3, Bud4, Axl1 and Axl2 (also known as Bud10) (Chant and Herskowitz, 1991; Chant et al., 1995; Chant and Pringle, 1995; Fujita et al., 1994; Halme et al., 1996; Lord et al., 2002; Roemer et al., 1996; Sanders and Herskowitz, 1996). The septins, a family of GTP-binding proteins (Cdc3, Cdc10, Cdc11, Cdc12 and Shs1/Sep7) that form a heteromeric protein complex and filaments (for reviews, see Longtine and Bi, 2003; Versele and Thorner, 2005), are also necessary for axial budding (Flescher et al., 1993). The bipolar pattern appears to depend on persistent cortical markers and is dependent on transmembrane proteins including Bud8, Bud9, Rax1 and Rax2 (Chen et al., 2000; Fujita et al., 2004; Harkins et al., 2001; Kang et al., 2004a; Zahner et al., 1996).

How is spatial information inherited from one cell division cycle to the next, allowing cells to exhibit specific budding patterns with such high fidelity? Because of the transient nature of the axial landmark, it is thought that a macromolecular complex that marks an axial bud site would assemble and disassemble in every cell cycle. Bud3 and Bud4 localize as a double ring encircling the mother–bud neck during and after the G2 phase and as a single ring at the division site after cytokinesis (Chant et al., 1995; Sanders and Herskowitz, 1996). Localization of Bud3 and Bud4 depends on septin integrity (Chant et al., 1995; Sanders and Herskowitz, 1996). Axl1 and Axl2 also localize as a double ring encircling the mother–bud neck prior to cytokinesis, and this double ring splits into two single rings after cytokinesis, although their localization patterns are different from each other before M phase (Halme et al., 1996; Lord et al., 2000; Roemer et al., 1996). These localization patterns of the axial-budding-specific proteins are consistent with the idea that a spatial cue from the division site directs the position of the subsequent bud site (Chant et al., 1995; Flescher et al., 1993; Snyder et al., 1991).

Previous studies had shown interesting structural features of the proteins specifically involved in the axial budding pattern but had left considerable uncertainly as to the nature of their roles in bud-site selection. Axl1 is expressed in a and α cells but not in a/α cells (Fujita et al., 1994). Although Axl1 has homology with the insulin-degrading enzyme family of endoproteases, its protease activity is required for processing of the mating pheromone a-factor precursor but not for bud-site selection (Adames et al., 1995). Axl2 is a Type I transmembrane glycoprotein with an N-terminal signal sequence and a central transmembrane domain (Halme et al., 1996; Roemer et al., 1996). Bud4 and Bud3 contain a putative GTP-binding motif (Sanders and Herskowitz, 1996) and a Rho GEF homology domain (also called a DH domain), respectively. Bud4 also shares sequence similarity to two anillin-related proteins in S. pombe, Mid1 and Mid2, which have homology to the PH (Pleckstrin homology) domain and a domain of unknown function, DUF1709 (Berlin et al., 2003; Tasto et al., 2003). However, it had not been known whether these putative domains are important for their roles in bud-site selection. How these axial-budding-specific proteins might assemble into a complex to mark a new bud site had also remained unclear.

Here we examined the assembly of the axial landmark by live cell imaging and interaction assays using strains that express all Bud proteins at their endogenous levels. Our results provide a detailed view of the interdependence of localization and interaction among the components of the axial landmark. We also show that Bud4 binds to GTP and that its GTP-binding is essential for the axial budding pattern at higher temperature, although less critical at 30°C. Our results suggest that Bud4 plays a key role in mediating the sequential assembly of the axial landmark and its cell-type-specific interaction with Bud5.

To investigate the functional interactions among the axial-budding-specific proteins, we first examined the interdependence of their localization by time-lapse microscopy. We examined localization of the functional GFP–Bud4 in cells, which express Spc42–mCherry (a component of the spindle pole body) as a cell cycle marker. As expected from a previous study (Sanders and Herskowitz, 1996), GFP–Bud4 localized to the mother–bud neck as a discrete double ring and then to the division site as a single ring after division (100%, n = 6; Fig. 1A,a) in wild-type cells. In bud3Δ cells, GFP–Bud4 initially localized to the mother–bud neck as in wild type, although the signal was slightly weaker (Fig. 1A,b). However, in bud3Δ cells undergoing cytokinesis, GFP–Bud4 appeared as fragmented rings, which were dispersed around the mother–bud neck region or near the periphery of the cell (66.7%, n = 6; Fig. 1A,b, time 67), while some cells showed a normal double ring (33.3%). Even when a loose double ring of GFP–Bud4 was visible, a 3D reconstruction revealed a fragmented ring structure in bud3Δ (Fig. 1B). Unlike wild-type cells, newly born bud3Δ cells mostly lacked a single ring structure of GFP–Bud4 at the division site (75%, n = 12; Fig. 1A,b 84).

We next determined whether Bud4 localization depends on Axl1 or Axl2. GFP–Bud4 localization in axl1 and axl2Δ mutants was indistinguishable from that in wild type (Fig. 1C). When many static images were analyzed, the integrated density of GFP–Bud4 in a fixed rectangle region of the mother–bud neck was similar in wild type and these mutants during M phase (Fig. 1D). When an approximately 50% larger rectangle region of the mother–bud neck was used for bud3Δ cells (due to the less tightly organized GFP–Bud4 localization in bud3Δ cells), the mean intensity of GFP–Bud4 was also not significantly lower in large-budded cells of bud3Δ than in wild type (Fig. 1D), suggesting that recruitment of Bud4 to the mother–bud neck is independent of these axial budding components.

We then asked whether localization of Bud3 is dependent on Bud4 by time-lapse microscopy. Bud3–CFP localized in a similar pattern as GFP–Bud4 in wild-type cells (Fig. 1E,a) but it failed to localize to the mother–bud neck in bud4Δ cells (Fig. 1E,b) (n = 6 for each strain). The integrated density of Bud3–CFP at the mother–bud neck was over six times higher in wild type than in bud4Δ during M phase, despite the elevation of the Bud3 protein level in bud4Δ (see Fig. 4A). These results thus suggest that Bud3 becomes de-localized in the absence of Bud4. Examining static images confirmed that loss of Bud3–CFP double rings in bud4Δ cells has an important consequence, since no unbudded cells showed the Bud3–CFP ring at the division site in bud4Δ cells (0%; n = 80), whereas 87% of the unbudded wild-type cells (n = 80) showed such rings. Taken together, these results suggest that Bud4 is necessary for localization of Bud3 while Bud3 is important for the integrity of Bud4 during/after cytokinesis.

As described above, the localization of Bud4 to the mother–bud neck does not depend on any specific axial-budding component. We wondered how localization of Axl1 and Axl2 might be affected in the absence of Bud4 or the other axial-budding-specific proteins. We determined localization of Axl2 using strains expressing Axl2–GFP (and Myo1–mCherry as an actomyosin ring marker). Although Axl2–GFP showed diffuse signals in the cytoplasm and the vacuolar lumen in some cells, it became concentrated at the sites of polarized growth, including the incipient bud site (7%; n = 100 unbudded cells), periphery of a small bud (98%; n = 50), mother–bud neck as a double ring (96%; n = 100 large budded cells) and the division site (90%; n = 100 unbudded cells) (Fig. 2; data not shown), consistent with previously reports (Halme et al., 1996; Roemer et al., 1996). This localization pattern was altered in bud4Δ cells, particularly during M and early G1. Although Axl2–GFP localized to the mother–bud neck in 85% of large budded cells (n = 250), less than 1% of cells had a double ring structure at the neck. Some cells also exhibited Axl2–GFP concentrated at the pole of mother cells in addition to the neck (14%) (Fig. 2e,g). Axl2–GFP at the neck often resided at the bud side in these cells and was not tightly organized as a ring structure or a splitting double ring during and after cytokinesis in bud4Δ (Fig. 2d–g), suggesting that the Axl2–GFP rings are not stable in these cells. As a consequence, only 7% of unbudded cells showed a single ring of Axl2–GFP at the division site in bud4Δ (n = 200) (Fig. 2, compare c and h). Axl2–GFP localized to an incipient bud site in unbudded bud4Δ cells that had the Myo1–mCherry patch (thus in late G1) and bud tips as in wild type (Fig. 2,i; data not shown).

In an axl1 mutant, Axl2 localization was mildly altered before and during cytokinesis (supplementary material Fig. S1). Although Axl2–GFP appeared at the pole of some mother cells and as a single ring at the mother–bud neck (15%, n = 140; supplementary material Fig. S1,a2), it appeared as a double ring in the majority of large-budded cells (69%; supplementary material Fig. S1,a1). Most unbudded cells also maintained a single ring of Axl2–GFP at the division site in the axl1 mutant (89%, n = 140; supplementary material Fig. S1,c1).

We next used time-lapse microscopy to examine localization of Axl1–GFP in axl2Δ and bud4Δ mutants, which also express Myo1–mCherry. While almost all wild-type cells exhibited Axl1–GFP signal at the neck (99%; n = 6) (Fig. 3a), Axl1–GFP failed to localize to the mother–bud neck almost completely in axl2Δ and bud4Δ mutants (n = 6 for each strain) (Fig. 3b,c). Quantification of Axl1–GFP in many static images of axl2Δ and bud4Δ cells confirmed almost no detectable fluorescence at the mother–bud neck and the division site during M and G1 (Fig. 3; see figure legend). Together with interaction data (see below), these results suggest that Axl1 and Axl2 assemble into the axial landmark through the interaction with Bud4 and that the association of Axl2 with Bud4 is necessary for the assembly of Axl1 into the axial landmark.

As described above, the localization of Bud3, Axl1 and Axl2 to the mother–bud neck depends on Bud4. These observations suggested that these proteins might assemble as an axial landmark through the interaction with Bud4. To test this idea, we performed immunoprecipitation experiments using anti-Bud4 antibodies and extracts from strains expressing Bud3 or Axl2 as a Myc-tagged protein at the endogenous level. Bud3–Myc co-precipitated with Bud4 from both haploid a and diploid a/α cells to a similar extent and also in the absence of Axl1 or Axl2 (Fig. 4A). These results suggest that the Bud3–Bud4 interaction is independent of cell type and is direct rather than being bridged by Axl1 or Axl2. We noticed that the Bud3 protein level was elevated in the axl1, axl2Δ or bud4Δ mutant (see input in Fig. 4A), but the reason is not clear at the present time. In a similar assay, Axl2–Myc was brought down with Bud4 efficiently in a cells, but less efficiently in a/α cells, even though these proteins are present at about equal levels in both cell types (Fig. 4B). The Bud4–Axl2 association was also less efficient in an axl1 mutant, and was almost completely absent in a bud3Δ mutant (Fig. 4B), suggesting that the Bud3–Bud4 interaction is a prerequisite for the association of Axl2 with Bud4.

We next determined whether Bud4 or Axl2 associates with Axl1 by a TAP (Tandem Affinity Purification) pull-down assay using the haploid AXL1-TAP strains. Bud4 associated with Axl1 efficiently in wild type but poorly in bud3Δ and axl2Δ mutants (Fig. 4C), consistent with the idea that Bud4 interacts with Axl1 after its association with Bud3 and Axl2. In contrast, little Axl2 was recovered with Axl1–TAP in a similar assay (Fig. 4D), suggesting that Axl1 and Axl2 did not associate with each other or that the interaction might be too transient to be easily detected.

Localization and interaction studies presented above suggested that Bud4 interacts with the axial-budding-specific proteins in a sequential manner to promote the assembly of the axial landmark. How might Bud4 perform such function? Bud4 has a putative GTP-binding motif that also overlaps with other putative domains at the C terminus (see Introduction; Fig. 5A). The GTP-binding motif of Bud4 does not belong to any specific subfamily of small GTPases (Sanders and Herskowitz, 1996), but resembles that of Gpa1 or septins (Fig. 5A). Using Bud4 purified as a TAP-tagged protein from yeast, we tested whether Bud4 binds guanine nucleotides. Bud4–TAP bound to [³H]GTP and [³H]GDP, although not as efficiently as Cdc42 (data not shown). Since the G1 box (‘P-loop’) is also found among ATP-binding proteins, we determined whether Bud4 binds specifically to GTP by a competition assay. When a 250 molar excess of non-radioactive ATP or GTP was pre-incubated with Bud4–TAP, GTP interfered with [³H]GTP loading onto Bud4 whereas ATP did not (Fig. 5B), suggesting that Bud4 specifically binds to GTP. To confirm that the putative GTP-binding motif of Bud4 is involved in GTP or GDP binding, we performed a GTP-binding assay using a truncated form of Bud4 (a.a. 1082–1447) fused to MBP (maltose-binding protein) and fluorescent GTP and GDP analogs (mant-GTP and mant-GDP). MBP–Bud4 bound to mant-GTP or mant-GDP, whereas MBP did not (Fig. 5C; supplementary material Fig. S2A). In contrast, a deletion of the eight residues of the G1 box (bud4ΔG1) significantly reduced its mant-GDP binding at room temperature (Fig. 5C). Taken together, these results indicate that Bud4 is indeed a GTP-binding protein.

To address whether GTP binding to Bud4 is important for axial budding, we first generated deletion mutations bud4ΔG1 and bud4ΔG2 that lack the G1 box (a.a. 1175–1182) and the G2–G4 boxes (a.a. 1344–1404), respectively. We also generated substitution mutations of the conserved residues in the GTP-binding motif, bud4^Q1367L and bud4^K1181N, which are expected to express Bud4 in the GTP and GDP-locked states in vivo, respectively (Fig. 5A). We then determined the budding pattern of a haploid (a) strain carrying each mutation at the BUD4 locus. While the bud4ΔG1 mutant exhibited a partial defect in axial budding at 30°C, it was severely defective in axial budding at 37°C (Fig. 5D). The bud4ΔG2 mutant was almost completely defective in the axial budding pattern, similar to bud4Δ (Fig. 5D), but its phenotype is likely due to multiple factors since its mant-GDP binding was not as defective as bud4ΔG1 (supplementary material Fig. S2A; see Discussion). Quantification from an immunoblot indicates that the Bud4ΔG1 protein level was about 93% of wild type (see Fig. 6A) whereas Bud4ΔG2 was about 67% of wild type (supplementary material Fig. S2C). Despite slight reduction of the protein level, Bud4ΔG1 localizes normally as wild type (see below), suggesting that the phenotype of bud4ΔG1 is unlikely to be due to any global structural change or instability of the protein. Taken together, these results indicate that GTP binding to Bud4 is required for normal axial budding. In contrast, bud4^Q1367L and bud4^K1181N exhibited little defect in the budding pattern (Fig. 5D). A simple interpretation of these results is that GTP binding to Bud4 has a structural role rather than a regulatory role (see Discussion).

We next asked how the bud4ΔG1 mutation affects the assembly of the axial landmark by immunoprecipitation with anti-Bud4 antibodies. Bud4ΔG1 was defective in association with Axl1–Myc and Axl2–Myc to different extents. When normalized based on the relative recovery of each protein and the amount of each protein present in the reaction, the association of Axl1 and Axl2 with Bud4ΔG1 was reduced to 6.1% and 34.1% of that with the wild-type Bud4, respectively (Fig. 6A). These results suggest that GTP binding to Bud4 is particularly important for its interaction with Axl1. Since Axl2 does not associate with Bud4 efficiently in the absence of Axl1 (see Fig. 4B), less association of Axl2 with Bud4ΔG1 than with wild-type Bud4 is likely to be a consequence of poor association of Axl1 with Bud4ΔG1.

We then addressed the same question by live cell imaging because of concerns about the overall stability of proteins during immunoprecipitation at a higher temperature. The localization pattern of GFP–Bud4ΔG1 was similar to that of GFP–Bud4, particularly during M and G1 at 30°C, although the localized signal of GFP–Bud4ΔG1 was weaker than that of wild type (Fig. 6B; supplementary material Fig. S3A). As expected, Axl1–mCherry co-localized with GFP–Bud4 to the mother–bud neck in large-budded cells (Fig. 6B,a) and to the division site in unbudded cells (Fig. 6B,b). In contrast, Axl1–mCherry co-localized with GFP–Bud4ΔG1 less efficiently to the mother–bud neck (Fig. 6B,e) and to the division site (Fig. 6B,f) at 30°C, as estimated by the percentage of cells with the localized signal and the signal intensity (Fig. 6B; see also legend to 6B). Instead, Axl1–mCherry appeared diffuse in the cytoplasm or localized at the bud tip of large-budded cells (Fig. 6B,g) or at the pole opposite to the division site (Fig. 6B,h) in addition to its normal localization sites at 30°C. Although the signals were slightly diminished, GFP–Bud4 and Axl1–mCherry still co-localized in most cells at 37°C (Fig. 6Bc,d). However, Axl1–mCherry appeared diffuse in most of large budded cells (Fig. 6B,i) and failed to concentrate at the division site (Fig. 6B,j) in the GFP-bud4ΔG1 mutant at 37°C, despite normal localization of GFP–Bud4ΔG1.

We also examined co-localization of Bud3 and Axl2 with GFP–Bud4ΔG1. While localized Bud3–mCherry slightly diminished in the GFP-bud4ΔG1 mutant compared to wild type, as did GFP–Bud4ΔG1 itself, the bud4ΔG1 mutation did not affect their co-localization (supplementary material Fig. S3B). Localization of Axl2–mCherry was mildly altered in the GFP-bud4ΔG1 mutant but to a much lower extent than Axl1 (data not shown). Taken together, these results suggest that GTP binding to Bud4 is necessary for stable association of Bud4 with Axl1 and thus the assembly of the axial landmark, accounting for the bud-site selection defect of the bud4ΔG1 mutant.

We wondered why the bud4ΔG2 mutant has such a severe defect in axial budding despite its relatively mild defect in GDP binding. We tested by immunoprecipitation with anti-Bud4 antibodies whether Bud4ΔG2 was defective in association with Axl1 or Axl2. When normalized, the association of Axl1 and Axl2 with Bud4ΔG2 was reduced to about 8% and 0.01% of that with the wild-type Bud4, respectively (supplementary material Fig. S2B). We then performed time-lapse imaging using a strain expressing GFP–Bud4ΔG2 and Myo1–mCherry. GFP–Bud4ΔG2 localized to the mother–bud neck until M but failed to segregate as an intact single ring after cytokinesis (n = 6; supplementary material Fig. S2D): the Bud4 ring at the mother side coalesced into a single patch, followed by the ring at the bud side (supplementary material Fig. S2D, time 47–60). The 3D reconstructed image indicated that GFP-Bud4ΔG2 at the bud side also failed to form a normal ring at time 47 (supplementary material Fig. S2D, right), suggesting that Bud4ΔG2 is defective in maintaining the integrity of Bud4, particularly after cytokinesis.

We reported previously that the cytoplasmic domain of Axl2 interacts with the Rsr1 GEF Bud5 (Kang et al., 2001). This observation supports the idea that the GEF links the axial landmark to the Rsr1 GTPase module for axial budding. However, the cell-type-specific interaction between Bud5 and Axl2 had been less clear, likely due to the overexpression of some of the proteins in the previous study. We thus explored this issue further by a TAP pull-down assay using strains expressing all proteins at their endogenous levels. Indeed, we found that Axl2 co-purified with Bud5–TAP from haploid a cells but not from diploid a/α cells. Axl2 also failed to associate with Bud5 in haploid axl1Δ, bud3Δ or bud4Δ cells (Fig. 7A). In a similar assay, Axl1 also co-purified with Bud5–TAP in wild-type cells but not in cells lacking any other axial components (Fig. 7B). Similarly, Bud4 co-purified with Bud5–TAP specifically in haploid cells but not in the absence of Bud3, Axl1 or Axl2 (Fig. 7C). Unlike these other axial components, Bud3 was brought down poorly with Bud5, but their association was dependent on Axl1 and Bud4 (Fig. 7D), Taken together, these results suggest that a complex including Axl1, Axl2 and Bud4 associates with Bud5 specifically in haploids and that Bud3 associates with Bud5 either very transiently or rather remotely through another axial component such as Bud4.

Our findings reported in this study suggest that Bud4 serves as a platform for the assembly of the axial landmark. Despite their similar localization patterns in wild-type cells, Bud4 localized normally in the absence of Bud3 until cytokinesis, whereas Bud3 was almost completely delocalized in the absence of Bud4. These observations and the close association of Bud4 with septins (Sanders and Herskowitz, 1996; Wloka et al., 2011; P.J.K. and H.-O.P., unpublished observation) suggest that Bud4 assembles with septins first and that Bud3 assembles into an axial landmark through the interaction with Bud4, while these proteins are interdependent during cytokinesis. We noticed that slowly migrating Bud3 bands were enriched by immunoprecipitation with anti-Bud4 antibodies (see Fig. 4A), suggesting that Bud4 associates preferentially with the modified Bud3. This observation supports the idea that Bud3 may assemble into an axial landmark efficiently after phosphorylation, as Bud3 is known to be phosphorylated by Cdc28–Clb2 kinase (Ubersax et al., 2003). Further investigation is necessary to test the idea. Our data presented here also suggest that Axl1 and Axl2 assemble into the axial landmark after Bud3 and that the association between Bud4 and Axl2 is likely to be a prerequisite for the interaction between Axl1 and Bud4. Thus, our data, together with previous studies, provide a temporal view of the ordered assembly of the axial landmark (Fig. 8).

The axial landmark interacts with the Rsr1 GTPase module through its GEF Bud5 (Kang et al., 2001) and then directs polarity establishment at a site adjacent to the previous division site (Kozminski et al., 2003; Park et al., 1997). At what point in the cell cycle is the axial landmark linked to the Rsr1 GTPase module? The physical association between the axial landmark and Bud5 suggests that this link is likely to occur after anaphase but before late G1 (i.e. before disassembly of the Bud3 and Bud4 rings in G1) (Fig. 8). This is consistent with the timing of Bud5 and Rsr1 localization to the division site after anaphase (Kang et al., 2001; Marston et al., 2001; Park et al., 2002) and Rsr1 dimerization, which also depends on Bud5 (Kang et al., 2010). Rsr1 becomes enriched in a patch at bud emergence, but this localization pattern in late G1 is observed even in cells budding randomly and thus is unlikely to be relevant to its role in bud-site selection. Similarly, localization of Axl2 in late G1 appears to be important for septin organization rather than bud-site selection (Gao et al., 2007). We found that Bud4 is required for the localization of Axl2 during M and early G1 but its localization in late G1 is independent of Bud4 (see Fig. 2). These observations are also consistent with the idea that Axl2 assembles into the axial landmark prior to late G1, while its localization in late G1 is irrelevant to its role in bud-site selection.

Previous studies had indicated that Bud5 interacts with the components of the axial or bipolar landmark including Axl2, Bud8 and Bud9 (Kang et al., 2004b; Kang et al., 2001; Krappmann et al., 2007). However, a key question about the cell-type-specificity of the budding pattern had remained unclear since these proteins are present in all cell types. How does Bud5 respond to one landmark but not to the other in one cell type? Our findings reported here have now clarified at least partly this outstanding issue. The cell-type-specific association of the axial landmark proteins with Bud5 would account for selection of an axial bud site specifically in a and α cells. Our data also highlight that all of these known components of the axial landmark are necessary for the proper assembly of the axial landmark and for their specific interaction with Bud5. In particular, we show that Axl1, which is expressed only in a and α cells (Fujita et al., 1994), is necessary for the association of Bud5 with Axl2 and Bud4. We thus propose that an active complex is formed in haploids when all of these axial-budding-specific proteins assemble together, allowing its interaction with Bud5 specifically in a and α cells but not in a/α cells.

A previous report has suggested that Bud4 contains a GTP-binding motif (Sanders and Herskowitz, 1996). We now provide evidence that Bud4 can bind specifically to guanine nucleotides. However, Bud4 was not as efficient as Cdc42 for GTP binding. It remains unclear whether Bud4 binds to GTP (or GDP) in 1∶1 molar ratio in vivo and whether it is a GTPase. Bud4 carries Ser at 1177 in the G1 box, similar to septin proteins (except Cdc3), Sec4 and Ypt1, which have very low intrinsic GTPase activity (Kabcenell et al., 1990; Versele and Thorner, 2004). In contrast, Ras or Rho family GTPases carry Gly at the analogous position, and alterations of this residue result in a decreased GTPase activity (Barbacid, 1987; McCormick, 1989). Indeed, Bud4 exhibited little GTP hydrolysis (C.-H.J. and H.-O.P., unpublished). Bud4^Q1367L and Bud4^K1181N, which are analogous to Ras^Q61L and Ras^K16N and are thus expected to be the GTP-locked state and GDP-locked or nucleotide-empty state, respectively, exhibited little defect in the axial budding pattern. A simple interpretation could be that GTP binding to Bud4 has a structural role rather than a regulatory role in the assembly of the axial landmark, but further studies are necessary to fully understand the potential regulation of GTP binding to Bud4.

Despite some uncertainly as to its regulatory mechanism, GTP binding to Bud4 seems to be critical for the interaction between Bud4 and Axl1 and thus the axial budding pattern. A bud4 mutant that is defective in GTP binding (bud4ΔG1) is also defective in the interaction with Axl1. Considering its defect in association with Axl1 (and partially with Axl2), it is surprising that axial budding of bud4ΔG1 is only partially defective at 30°C while being disrupted much more severely at 37°C. Although we do not have a clear explanation for this observation, one possibility is that its mild defect might be more easily compensated by itself or other proteins at a low temperature if Bud4 functions as an oligomer with itself or other proteins such as septins in vivo. While Bud4ΔG1 localizes in a similar pattern to Bud4, the localization of Axl1 in the bud4ΔG1 mutant is partially defective at 30°C but more severely defective at 37°C. This localization defect of Axl1 in the bud4ΔG1 mutant is well correlated with the defect in the axial budding pattern of bud4ΔG1 at different temperatures. Although Bud4ΔG2 caused a complete defect in axial budding even at 30°C, one caveat is that this deletion also covers a region that overlaps the putative PH domain. Furthermore, Bud4ΔG2 is only partially defective in mant-GDP binding, suggesting that its phenotype is unlikely solely due to the defect in GTP binding. Because of its normal localization to the mother–bud neck in M (despite its lower level compared to wild type), Bud4ΔG2 may not carry a global structural alteration. However, its localization defect observed during/after cytokinesis suggest that Bud4ΔG2 is not stably anchored at the division site. It is possible that Bud4ΔG2 becomes unstable after cytokinesis, perhaps due to the disruption of the PH domain. These observations suggest that the integrity of Bud4 at the division site is likely to be critical for the assembly of the axial landmark. However, further investigation is required to understand the exact role of the PH domain of Bud4 in bud-site selection.

While this study uncovers a unique role of Bud4 in the assembly of the axial landmark, it also raises several questions as discussed above. Our future studies will address these challenging questions to fully understand the mechanism by which a specific growth site is determined. Studies of bud-site selection will continue to provide an opportunity to uncover a paradigm of cell polarization and morphogenetic differentiation.

Standard methods of yeast genetics, DNA manipulation, and growth conditions were used (Ausubel et al., 1999; Guthrie and Fink, 1991) unless indicated otherwise. The strains and plasmids used in this study are listed in supplementary material Tables S1 and S2, and their description is provided below each table. All tagged versions of each gene used for localization and interaction assays were expressed from their own promoters and at their chromosomal loci (except Myo1–mCherry, which was expressed from a CEN plasmid) and functionally substituted each corresponding wild-type gene based on their bud-site selection and growth phenotypes.

The full-length Bud4 was expressed in yeast as a TAP-tagged protein, which carries a calmodulin-binding peptide (CBP), a TEV cleavage site, and two IgG binding domains of protein A (Ghaemmaghami et al., 2003). Bud4–TAP was purified from HPY908 as previously described (Ghaemmaghami et al., 2003; Rigaut et al., 1999). The GTP-binding motif of Bud4 (a.a. 1081–1447) was expressed as an MBP–fusion protein using plasmid pHP1384 and purified from a protease deficient E. coli strain NB42 (from the Herskowitz lab collection) as previously described (Kozminski et al., 2003; Park et al., 1997). Bud4 antiserum was prepared using GST (glutathione-S-transferease)-Bud4 (a.a. 1–398), which was expressed using pSS43 and purified from E. coli, as previously described (Sanders and Herskowitz, 1996). 0.5 mg GST–Bud4 was used for the initial injection of two rabbits and for subsequent boosts (Lampire Biological Laboratories, Pipersville, PA).

Yeast cells were grown to mid-log phase (OD₆₀₀ ∼1.0) in YPD at 30°C, and extracts were prepared as previously described (Kang et al., 2004a). Immunoprecipitation or IgG-agarose pull-down assays were carried out at 4°C essentially as previously described (Kozminski et al., 2003), except as noted. Typical recovery of Bud4 and its mutant forms by immunoprecipitation with anti-Bud4 antibodies was about the same and estimated at about 8 ∼ 10% of the input. Axl1–TAP and Bud5–TAP were purified by IgG-agarose pull-down assays, and their typical recovery was about 3 ∼ 15% of the input. After immunoprecipitation or pull-down assays, 44% of the precipitated proteins and 1% of input were subjected to SDS-PAGE. Epitope-tagged proteins were then detected using appropriate antibodies – anti-HA antibody HA11 (Covance Research Products, Denver, PA), anti-Myc antibody 9E10 (kindly provided by M. Bishop, University of California-San Francisco), and rabbit monoclonal anti-calmodulin binding protein (Upstate Cell Signaling Solutions, Temecula, CA). Bud4 and Axl2 were detected using polyclonal antibodies against Bud4 and Axl2 (kindly provided by M. Snyder, Stanford University), respectively. Protein bands were then detected with Alexa Fluor ^R 680 goat anti-rabbit IgG (Molecular Probes, Eugene, OR) or IRDye^R 800CW conjugated goat anti-mouse IgG (LI-COR Biosciences, Lincoln, Nebraska) secondary antibodies using the LI-COR Odyssey system (LI-COR Biosciences, Lincoln, Nebraska). The relative amount of proteins compared to an internal control on immunoblots was quantified using the software of the LI-COR Odyssey system.

GTP binding was measured by filter binding assays as previously described (Kozminski et al., 2003; Park et al., 1993). The affinity-purified Bud4–TAP was incubated with [8,5′-³H]GTP or [8,5′-³H]GDP (38.2 Ci/mmol, NEN Life Science Products) in the nucleotide loading buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM DTT, 10 mM EDTA, 10% glycerol) at 30°C for 30 min. The nucleotide loading mixture was diluted to 1 ml with wash buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM DTT, 10 mM MgCl₂, 10% glycerol) and then filtered through a nitrocellulose filter (0.2 µm pore-diameter, Millipore). After washing with 10 ml of wash buffer, the radioactivity retained on the filters was quantified by scintillation counting. For a competition experiment, Bud4–TAP protein was incubated with 250 molar excess ATP or GTP for an additional 30 min at room temperature prior to [³H]GTP loading, followed by a filter binding assay.

Binding of the fluorescent GTP and GDP analogs, mant-GTP and mant-GDP, was monitored at room temperature, as previously described (Kang et al., 2010) with slight modification. An equimolar concentration (typically 0.5∼1 µM, see figure legends) of MBP–Bud4 (a.a. 1081–1447), MBP–Bud4 mutant proteins, and MBP were loaded with a 150-fold molar excess of mant-GTP or mant-GDP (Molecular Probes, Eugene, OR) in nucleotide exchange buffer (50 mM Tris-HCl, pH 7.5, 20 mM EDTA, 5 mM DTT) at room temperature for 25 min. After proteins were immobilized on amylose–agarose beads, the unbound nucleotides were washed off with wash buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl₂, 5 mM DTT). The mant-GTP- and mant-GDP-loaded proteins were eluted from the beads using the elution buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl₂, 5 mM DTT, 10 mM maltose) and then subjected to fluorescence measurement using a Cary Eclipse fluorescence spectrophotometer (Varian, Palo Alto, CA). Emission spectra were collected from 400 to 600 nm following excitation at 360 nm, using the SCAN program according to the instrument's manual.

To determine budding patterns, bud scars were visualized by staining cells with Calcofluor as previously described (Pringle, 1991). For live-cell imaging, cells were grown at 30°C, unless indicated otherwise, in synthetic complete (SC) or SC-Ura media when selecting for the Myo1-mCherry plasmid. Cells were then concentrated by centrifugation and spotted on a slab of the same media containing 2% agarose. GFP-, CFP- and mCherry-fusion proteins were visualized in exponentially growing cells essentially as previously described (Kang et al., 2004a; Kang et al., 2001), using a Nikon E800 microscope (Nikon, Tokyo, Japan) fitted with a 100X oil-immersion objective (N.A. = 1.30) and FITC/GFP, CFP and mCherry/TexasRed filters from Chroma (Brattleboro, VT). A series of optical sections (0.3 µm intervals) were collected using Slidebook software (Intelligent Imaging Innovations, Denver, CO) with a Hamamatsu ORCA-2 CCD camera (Hamamatsu Photonics, Bridgewater, NJ) at room temperature (22–24°C). A maximum projection was created for most 2D images and, where indicated, deconvolution and 3D reconstruction were performed with Slidebook software. Quantification of fluorescence intensities was performed using a single section of each image with NIH ImageJ software. The integrated density at the bud neck region was calculated by subtracting the fluorescence intensity in the cytosol from the total intensity in an ImageJ-drawn polygon covering the mother–bud neck region, and was shown as mean ± s.d. Statistical significance was determined using Student's t-test. Where indicated, relative intensity of fluorescence was analyzed along a line in many cells (typically n = ∼40–50) and the representative cells are shown.

We thank M. Longtine, E. Bi, M. Snyder, S. Sanders, J. Chant, M.-N. Simon, C. Boone, J. K. Hood-DeGrenier, M. Bishop, T. Davis and A. Fujita for providing antibodies, strains and plasmids; J. Konz for his assistance with preparing Bud4 antibodies; and K. Kozminski for his comments on the manuscript.