The GAP activity of Msb3p and Msb4p for the Rab GTPase Sec4p is required for efficient exocytosis and actin organization

Polarization of cell growth is critical for generating distinct cellular domains and is ultimately responsible for the diversity of cell types, tissues, and organs. Thus, cell polarity is essential for development and differentiation in many organisms (Drubin and Nelson, 1996). In the budding yeast Saccharomyces cerevisiae, polarized cell growth is thought to occur in a hierarchal manner. At the beginning of the cell cycle, the small Rho GTPase Cdc42p and its regulators such as the guanine nucleotide–exchange factor (GEF) Cdc24p are clustered at a specific region of the cell cortex, marking the site for polarity establishment (Johnson, 1999). Cdc42p effectors, including the p21-activated kinases, Ste20p and Cla4p; the formin, Bni1p; and the structurally related proteins, Gic1p and Gic2p; are then recruited to the cortical site to polymerize and/or organize the actin cytoskeleton, including actin cables and actin patches, at the presumptive bud site (Pruyne and Bretscher, 2000b). Actin patches are thought to mediate endocytosis (Munn, 2000), whereas actin cables are thought to function as tracks along which post-Golgi vesicles are transported from the mother cell to the daughter cell (Pruyne and Bretscher, 2000a).

Secretion in eukaryotes occurs in multiple, sequential steps, each of which is controlled by a distinct Rab GTPase. In S. cerevisiae, the Rab GTPase Sec4p plays a central role in polarized secretion from Golgi to the plasma membrane and is thought to act by tethering secretory vesicles to the plasma membrane via its effector, the exocyst, a multisubunit protein complex (Guo et al., 1999). Like other Ras-family members, Sec4p cycles between an inactive GDP- and active GTP-bound state. This cycling is regulated by its GEF, Sec2p (Walch-Solimena et al., 1997), and presumably by its GAP(s), whose identity has not been determined.

Msb3p and Msb4p are a pair of structurally related proteins that localize to the sites of polarized growth (Bi et al., 2000). Overexpression of Msb3p or Msb4p suppresses cdc24-Ts and cdc42-Ts mutants, although the mechanism underlying this suppression is not clear. Both Msb3p and Msb4p have a Rab GAP domain and indeed display a GAP activity toward a number of Rab GTPases in vitro (Albert and Gallwitz, 1999, 2000). However, the in vivo Rab targets of Msb3p and Msb4p are not known. In this report, we present multiple lines of evidence to indicate that Msb3p and Msb4p regulate exocytosis by functioning as GAPs for Sec4p in vivo. In contrast to the general view that polarized actin cytoskeleton guides secretion to a specific cellular domain, we also present evidence to indicate that a primary defect in polarized secretion can cause defects in polarized actin organization. Thus, polarized actin organization and polarized secretion appear to reenforce each other.

Because Msb3p and Msb4p display a GAP activity toward a number of Rab GTPases in vitro, a substrate promiscuity that is common among most known Rab GAPs, we decided to define the in vivo Rab target(s) of Msb3p and Msb4p by the following approaches. First, we examined possible protein-trafficking defects in cells lacking both Msb3p and Msb4p. We found that msb3Δ msb4Δ cells accumulated a large number of vesicles at 24°C, whereas wild-type and single-mutant cells had none or very few vesicles (Fig. 1 A and Fig. 4 A). The vesicles were 90.00 ± 11.77 nm (n = 204) in diameter, falling within the range of 80–100 nm post-Golgi secretory vesicles (Novick and Schekman, 1979). In most cells, the vesicles appeared to be distributed randomly in the mother and the daughter (Fig. 1 A, bottom left). In a few cells, vesicles were concentrated in the daughter (Fig. 1 A, top right). These data suggest that Msb3p and Msb4p must share a function in secretion.

The second approach used to assess a possible secretory defect in msb3Δ msb4Δ cells was to monitor the secretion of invertase, a sucrose-metabolizing enzyme, and of Bgl2p, an endo-β-1,3-glucanase required for cell wall organization and biogenesis (Mrsa et al., 1993). When invertase secretion was followed over time at 24°C, the ratio of secreted invertase versus total invertase (external plus internal) was mildly, but consistently lower in msb3Δ msb4Δ cells than in wild-type cells, with the difference peaking around 45 min after induction (Fig. 1 B). However, the majority of invertase was secreted efficiently to the periplasmic region during the course of induction (Table I). In contrast, Bgl2p was accumulated in large quantity inside the msb3Δ msb4Δ cells at both 24°C and at 37°C (Fig. 1 C). As expected, Bgl2p accumulation in the late secretory mutant, sec6–4, was temperature dependent (Fig. 1 C). These data suggest that most vesicles accumulated in msb3Δ msb4Δ cells carry Bgl2p and a small fraction of vesicles carry invertase.

If Msb3p and Msb4p play a role in exocytosis, deletion or overexpression of MSB3 and MSB4 may display genetic interactions with some of the late secretory mutants. Indeed, deletion of MSB3 and MSB4 produced synthetic inhibitory effects on cell growth with sec3–2 and sec9–4 mutants at 30°C, but not with sec1–1, sec2–41, sec4–8, and sec6–4 mutants (Fig. 2 A). In addition, overexpression of Msb3p or Msb4p inhibited the growth of sec2–41 cells at 30°C, but not of any other late sec mutants, including sec1–1, sec3–2, sec4–8, sec5–24, sec6–4, sec8–9, sec9–4, sec10–2, and sec15–1 (Fig. 2 B). In contrast, overexpression of Gyp1p, a GAP for Ypt1p that is involved in ER to Golgi transport but also exhibits a GAP activity toward Sec4p in vitro (Du et al., 1998; Du and Novick, 2001; De Antoni et al., 2002), did not inhibit the growth of sec2–41 cells at 30°C (Fig. 2 B). These results suggest that Msb3p and Msb4p are involved in exocytosis and can antagonize the function of Sec2p, the known GEF for Sec4p.

Because Msb3p and Msb4p display a GAP activity toward Sec4p in vitro and appear to colocalize with Sec4p at the sites of polarized growth during the cell cycle, it seemed likely that Msb3p and Msb4p might participate in the regulation of exocytosis by functioning as GAPs for Sec4p in vivo. To test this hypothesis, we took advantage of the observation that a sec4-Q79L sec15–1 double mutant is inviable at 25°C (Walworth et al., 1992). The Q79L mutation shifts Sec4p toward its GTP-bound form by decreasing the intrinsic GTPase activity, but this Sec4p mutant is still responsive to GAP action (Walworth et al., 1992; Du et al., 1998). Sec15p, an effector of Sec4p, is thought to mediate the role of Sec4p in the assembly of the exocyst (Guo et al., 1999). We reasoned that if Msb3p and Msb4p are physiological GAPs for Sec4p, their overexpression might suppress the synthetic lethality between sec4-Q79L and sec15–1 by decreasing the level of GTP-bound Sec4p.

To examine this possibility, we constructed a sec4-Q79L sec15–1 double mutant harboring an URA3-marked plasmid carrying wild-type SEC4. A LEU2-marked multicopy plasmid carrying either MSB3 or MSB4 was transformed into the tester strain and assayed for its ability to replace the SEC4-containing plasmid by examining cell growth on plates containing 5FOA, a chemical that selects for cells that have lost the URA3-containing plasmid (Fig. 2 C). Multicopy MSB3, but not GYP1 or MSB4, was able to suppress the sec4-Q79L sec15–1 mutant, supporting the hypothesis that Msb3p functions as a GAP for Sec4p in vivo.

GAPs for Ras, Rho, and Rab GTPases all contain an invariant arginine residue (the “finger arginine”) in the catalytic domain that is critical for their GAP activities (Ahmadian et al., 1997; Albert et al., 1999). Because Msb3p and Msb4p contain an arginine residue (R282 in Msb3p and R200 in Msb4p) at the corresponding position, we decided to examine whether Msb3p and Msb4p function on Sec4p by a similar mechanism. In addition, we hoped that Msb3p and Msb4p mutants deficient in the GAP activity toward Sec4p might offer an opportunity to distinguish the role of Msb3p and Msb4p in secretion from their role in actin organization. For these reasons, we substituted the arginine residue in Msb3p and Msb4p for either phenylalanine or lysine and determined the properties of the mutant proteins.

To facilitate protein purification, Msb3p, Msb4p, and their derivatives were all tagged with six histidines at their COOH termini. The tagged Msb3p and the arginine mutants were expressed from a galactose-inducible promoter in yeast cells. Crude extracts containing induced Msb3p or its mutant forms were subjected to a filter assay with GTP-loaded Sec4p as a substrate. Only extracts from cells overexpressing wild-type Msb3p exhibited measurable GAP activity (unpublished data), even though the mutant proteins, Msb3p-R282F and Msb3p-R282K, and the wild-type protein were expressed at similar levels (Fig. 3 A). For a quantitative assay, Msb3p and its mutants were purified from induced cells by affinity chromatography and assayed for their GAP activities toward GTP-loaded Sec4p by an HPLC-based method. As shown in Fig. 3 B, in comparison to wild-type Msb3p, both arginine mutants showed a significantly reduced Sec4p-GAP activity.

Msb4p and its mutants, Msb4p-R200F and Msb4p-R200K, were purified from E. coli cells and tested with GTP-loaded Sec4p as a substrate (Fig. 3 C). Again, the arginine mutation led to a significant loss of GAP activity. These data suggest that Msb3p and Msb4p function as GAPs for Sec4p by an arginine finger-like mechanism.

Despite the drastic reduction in their GAP activity toward Sec4p, the arginine mutants of Msb3p and Msb4p were expressed at normal levels (Fig. 3 D) and localized to the sites of polarized growth like the wild-type proteins (Fig. 3 E, and unpublished data). These data suggest that a significant loss of the GAP activity of Msb3p and Msb4p does not compromise the molecular interactions required for their targeting to the growth sites.

To determine whether the GAP activity of Msb3p and Msb4p is required for their in vivo function, we developed two assays. The first assay is based on our previous observation that msb3Δ msb4Δ gic1Δ gic2Δ quadruple mutant is inviable with a loss-of-polarity phenotype (Bi et al., 2000). The second assay is based on our new observation that msb3Δ msb4Δ is synthetically lethal with cdc42–201, a newly isolated temperature-sensitive cdc42 allele (Zhang et al., 2001). These two tester strains were kept alive by introducing an URA3-marked plasmid that carries either wild-type GIC1 for the first assay or CDC42 for the second assay. HA-tagged Msb3p, Msb4p, and their arginine mutants expressed from a LEU2-marked, high-copy plasmid in the tester strains were assayed for their ability to replace the URA3-marked plasmids on SC-Leu+5FOA plates. Plasmids carrying the arginine mutants of MSB3 or MSB4 failed to replace the URA3-marked plasmids in both assays, in direct contrast to the plasmids carrying wild-type MSB3 or MSB4 (Fig. 3 F and unpublished data). These data indicate that the GAP activity of Msb3p and Msb4p is essential for their in vivo function(s).

Overexpression of Msb3p-R282K and Msb4p-R200K mutants also failed to inhibit the growth of sec2–41 cells at 32°C (Fig. 3 G), suggesting that the GAP activity of Msb3p and Msb4p is required for antagonizing the function of Sec2p. The arginine mutant of Msb3p also failed to suppress the synthetic lethality between sec4-Q79L and sec15–1 (Fig. 3 H), further supporting the notion that it is the GAP activity of Msb3p toward Sec4p, not merely the presence of Msb3p, that is responsible for the suppression.

To determine whether vesicle accumulation in msb3Δ msb4Δ cells was due to the absence of the proteins or the loss of their GAP activity, we examined two pairs of haploid strains, JGY184A (MSB3 msb4Δ) and JGY190A (msb3-R282K msb4Δ), and JGY51 (msb3Δ MSB4) and JGY127A (msb3Δ msb4-R200K). In most MSB3 msb4Δ or msb3Δ MSB4 cells, either no or just a few vesicles (usually <10 vesicles per cell section) were detected (Fig. 4 A, left). In contrast, most msb3-R282K msb4Δ or msb3Δ msb4-R200K cells accumulated a large number of vesicles similar to those observed in msb3Δ msb4Δ cells (Fig. 4 A, right). These data indicate that the loss of the GAP activity of Msb3p and Msb4p is responsible for vesicle accumulation.

Cells of msb3Δ msb4Δ strain are rounder in shape, heterogeneous in size, and have a partially disrupted actin cytoskeleton (Bi et al., 2000). Actin patches in these cells tended to delocalize into the mother side at early stages of the cell cycle when the patches should be predominantly concentrated in the buds (Fig. 4 B, compare columns 1 and 2). Actin cables were clearly present and largely well organized in msb3Δ msb4Δ cells. However, some cables appeared to be shorter, and sometimes misoriented in the mutant strain (Fig. 4 B, compare columns 1 and 2).

To determine whether the GAP activity of Msb3p and Msb4p is required for actin-patch organization, MSB3, msb3-R282K, MSB4, or msb4-R200K was integrated into the msb3Δ msb4Δ mutant at the msb3Δ or msb4Δ locus, respectively. The integrants with MSB3 or MSB4 showed normal cell morphology and actin-patch organization (Fig. 4 B, columns 3 and 5). Interestingly, the integrants with msb3-R282K or msb4-R200K displayed similar defects in cell morphology and actin-patch organization as the msb3Δ msb4Δ mutant did (Fig. 4 B, columns 4 and 6), suggesting that the GAP activity of Msb3p and Msb4p is required for actin-patch organization.

Multicopy MSB3 is known to suppress the growth defect of cdc42–1 cells at the nonpermissive temperature (Bi et al., 2000) (Fig. 5 A, top). We found that multicopy MSB3 also suppressed the growth defect of another cdc42-Ts allele, cdc42–201 (Fig. 5 A, bottom). In addition, the budding and the actin-organization defects in both cdc42-Ts mutants were largely suppressed by multicopy MSB3 (Fig. 5 B) (Table II). Interestingly, multicopy msb3-R282K failed to suppress both the budding and the actin-organization defects of the two cdc42-Ts mutants (Fig. 5, A and B) (Table II), which were not defective in secretion per se as indicated by EM studies (Fig. 5 C). These results suggest that the GAP activity of Msb3p is required for the suppression of the actin-organization defects in cdc42 mutants. Together with the results described in the previous section, these data raise an intriguing possibility that polarized secretion may be normally involved in modulating polarized actin organization.

The fact that the GAP activity of Msb3p is required to rescue both the secretory and the actin-patch-organization defects in msb3Δ msb4Δ cells raises the possibility that the actin-patch disorganization in msb3Δ msb4Δ cells might be a consequence of the fusion of mistargeted vesicles with the plasma membrane of the mother cells. To examine this possibility, we took advantage of mutants in TPM1 and TPM2, which encode two isoforms of tropomyosin that are required specifically for the formation of actin cables but not actin patches (Pruyne et al., 1998). When tropomyosins are conditionally inactivated, all actin cables are lost within one minute. As a result, secretory vesicles are no longer transported to the daughter cell, but instead fuse with the plasma membrane of the mother cell. After inactivation of tropomyosins for 30–60 min, actin patches become randomly distributed in both the mother and the daughter cells (Fig. 6, A and B) (Pruyne et al., 1998).

To test our hypothesis, we examined the distribution of actin patches in sec6–4 tpm1–2 tpm2Δ cells. At the restrictive temperature, secretory vesicles in this mutant are no longer delivered to the bud due to the loss of actin cables. In addition, vesicles in this mutant fail to fuse with the plasma membrane due to the inactivation of Sec6p, a component of the exocyst that is essential for vesicle tethering. We observed that, upon shifting to 36°C for 60 min, 51% of the triple mutant cells still displayed a polarized organization of actin patches in comparison to 88% of the tpm2Δ cells, 85% of the sec6–4 cells, and 0% of the tpm1–2 tpm2Δ cells (Fig. 6, A and B). These results suggest that the actin-patch disorganization in tpm1–2 tpm2Δ cells, and, by extrapolation, in msb3Δ msb4Δ cells, depends on the fusion of secretory vesicles in the mother cells with the plasma membrane.

One possible explanation for the polarized organization of actin patches in sec6–4 tpm1–2 tpm2Δ cells is that the lifespan of the “old patches” (existed before the temperature shift) in the buds of the small-budded cells is significantly increased. We measured the lifespan of actin patches in four different strains at two different temperatures, 20°C and 36°C, using Abp1p-GFP as a marker for the patches (Fig. 6, C and D). At 20°C, actin patches in all four strains displayed a similar lifespan, ∼16 s. At 36°C, the lifespan of actin patches in both tropomyosin mutants (tpm2Δ and tpm1–2 tpm2Δ) were ∼9 s. In contrast, the lifespan of actin patches in both mutants carrying sec6–4 (sec6 tpm1–2 tpm2Δ, and sec6–4) were ∼21 s. We also observed that the lifespan of actin patches in the mother and the daughter compartments of the same cell were virtually identical for all four strains at both temperatures. These data suggest that blocking exocytosis at 36°C increases the lifespan of actin patches, but this increase alone is not sufficient to explain the polarized actin-patch organization in the sec6 tpm1–2 tpm2Δ mutant.

The challenge for studying Rab GAPs is twofold. First, most Rab GAPs in yeast are not essential for cell viability. Cells lacking a single known Rab GAP in yeast, Gyp1p, Mdr1p/Gyp2p, Gyp5p, Gyp6p, Gyp7p, or Gyp8p, produce no obvious defects in protein trafficking (Strom et al., 1993; Vollmer and Gallwitz, 1995; Du et al., 1998; Albert and Gallwitz, 1999; Vollmer et al., 1999; Du and Novick, 2001; De Antoni et al., 2002). Second, all known Rab GAPs show substrate promiscuity in in vitro assays, and with the exception of Ypt1p-GAPs (Du and Novick, 2001; De Antoni et al., 2002), no evidence for their in vivo function has been obtained.

We took multiple approaches to assess whether Msb3p and Msb4p are Sec4p-specific GAPs in vivo. First, deletion of MSB3 and MSB4 together, more specifically, inactivation of the GAP activity of Msb3p and Msb4p caused a defect in exocytosis. Second, msb3Δ msb4Δ cells displayed synthetic interactions with late secretory mutants. Third, multicopy MSB3 suppressed the synthetic lethality of a sec4-Q79L sec15–1 mutant, presumably, by reducing the level of GTP-bound Sec4p. Fourth, multicopy MSB3 or MSB4 inhibited the growth of sec2–41 cells, but not of any other late sec mutants. Finally, among the 11 Rab GTPases in S. cerevisiae, only Sec4p shares a localization profile and mutant phenotype (post-Golgi vesicle accumulation and exocytosis defect) with Msb3p and Msb4p. We further demonstrate here that Msb3p and Msb4p most likely act on Sec4p by an arginine finger–like mechanism. Together, these results provide compelling arguments for the involvement of Msb3p and Msb4p in exocytosis by functioning as GAPs for Sec4p in vivo.

Deletion of MSB3 and MSB4, presumably leading to a higher level of Sec4p-GTP in the cell, caused significant accumulation of 100-nm vesicles and of the endoglucanase Bgl2p in the cell, but produced little effect on invertase secretion. A similar effect on invertase secretion is also observed in a strain carrying sec4-Q79L, in which Sec4p is predominantly in the GTP-bound form (Walworth et al., 1992). These data suggest that a defect in GTP hydrolysis by Sec4p affects secretion of different cargoes differentially, which is consistent with a previous report that there are at least two distinct populations of post-Golgi vesicles accumulated in late sec mutants: a minor population carries invertase, whereas the major population carries Bgl2p (Harsay and Bretscher, 1995). Similar differential effect on secretion is also observed in a cdc42-Ts mutant (Adamo et al., 2001). Because Msb3p, Msb4p, and Cdc42p are all involved in polarized growth and because Msb3p and Msb4p interact with Cdc42p genetically (Bi et al., 2000) and biochemically (unpublished data), it is possible that one role of Msb3p, Msb4p, and Cdc42p in polarized growth is to regulate secretion of Bgl2p, a cell wall–remodeling enzyme that is needed during cell-surface expansion.

Our EM studies indicate that in most msb3 msb4 mutant cells, vesicles are distributed randomly within the cell, and in a few cells, vesicles are preferentially localized to the daughter cell. This pattern of vesicle accumulation would be consistent with a defect in vesicle transport and/or tethering. Vesicle transport from Golgi membrane to bud tip requires the function of vesicle-associated Sec4p-GTP, whose formation is catalyzed by the vesicle-associated GEF Sec2p, presumably using the cytosolic pool of Sec4p-GDP as the substrate. Vesicle tethering requires the function of Sec4p-GTP and its effector, the exocyst (Guo et al., 1999). In the absence of Msb3p and Msb4p, Sec4p would be primarily in the GTP-bound form and less Sec4p-GDP would be recycled back to the cytosol. This raises the question as to how an alteration in the ratio of GTP-bound versus GDP-bound Sec4p results in vesicle accumulation. We imagine two major scenarios. First, the decreased recycling of Sec4p-GDP to the cytosol in msb3 msb4 mutant cells could be responsible for the vesicle accumulation. In this case, less Sec4p-GDP from the cytosol is recruited to the Golgi site to be converted to Sec4p-GTP on Golgi membranes and/or vesicles. Consequently, the efficiency of vesicle transport is compromised. The second scenario is that the increased amount of Sec4p-GTP in msb3 msb4 mutant cells could be responsible for the vesicle accumulation. In this case, the increased level of Sec4p-GTP might hold the exocyst in place for longer periods of time so that reduced recycling of exocyst components would cause a defect in additional rounds of vesicle tethering. It is also possible that the disassembly of the vesicle-tethering complex is a prerequisite for the formation of the trans-SNARE complexes between Golgi membrane–derived vesicles and the plasma membrane, which leads to vesicle fusion. Because the assembly of the exocyst depends on Sec4p-GTP (Guo et al., 1999), its disassembly may depend on the hydrolysis of Sec4p-bound GTP. Thus, a deficiency in GTP hydrolysis may cause a secretory defect by indirectly preventing efficient formation of the SNARE complexes.

To distinguish between the two scenarios, we reasoned that, if the Sec4p-GDP recycling is the key, multicopy wild-type SEC4 should suppress the morphological defect of the msb3Δ msb4Δ sec4-Q79L mutant, because more Sec4p-GDP would be generated in the cell due to the intrinsic GTPase activity of the wild-type Sec4p. In contrast, if the absolute amount of Sec4p-GTP is critical, multicopy SEC4 should exacerbate the morphological defect of the msb3Δ msb4Δ sec4-Q79L mutant, because more Sec4p-GTP would be produced in the cell due to the increased concentration of Sec4p protein. We found that multicopy SEC4 suppressed the morphological defect of the triple mutant reasonably well (unpublished data), thus favoring the first scenario. However, this result cannot rule out the second scenario with certainty.

The cellular locations of the GEF and the GAPs for Sec4p provide direct clues on how the activity of Sec4p, and hence the exocytosis, is spatially regulated (Fig. 7 A). Sec2p, the GEF for Sec4p, colocalizes with Sec4p on the secretory vesicles. This colocalization is thought to ensure Sec4p in its GTP-bound form, which is required for the transport of post-Golgi vesicles to the active growth sites (Walch-Solimena et al., 1997). In contrast, Msb3p and Msb4p, the GAPs for Sec4p, are mainly concentrated at the active growth sites in close association with the plasma membrane. This conclusion is based on cell-fractionation and localization studies on Msb3p and Msb4p (Bi et al., 2000) (unpublished data). Thus, Msb3p and Msb4p are well positioned to promote efficient recycling of Sec4p by facilitating the hydrolysis of Sec4p-bound GTP at the plasma membrane.

Our studies led to two interesting findings. First, the GAP activity of Msb3p and Msb4p is required for efficient exocytosis and actin-patch organization. Second, the GAP activity of Msb3p is required for the suppression of the actin-organization defects in cdc42–1 and cdc42–201 cells. These observations raise an intriguing question: could a defect in polarized secretion cause a defect in actin organization? In tpm1–2 tpm2Δ cells where actin cables are absent at 36°C, secretory vesicles fuse with any part of the plasma membrane, causing depolarized growth. Meanwhile, actin patches are gradually reorganized from the small buds to the entire cell cortex (Pruyne et al., 1998). However, when the actin cables and the vesicle-tethering/fusion process are simultaneously inactivated in the sec6–4 tpm1–2 tpm2Δ cells, actin patches remain in the small buds. This result suggests that the reorganization of actin patches in tpm1–2 tpm2Δ cells depends on ongoing exocytosis and, likely, endocytosis, both of which are defective in sec6–4 and all other late sec mutants (Riezman, 1985).

Actin-patch polarization in the small buds of the sec6–4 tpm1–2 tpm2Δ cells at 36°C can be explained by one of the three possibilities: patch motility is blocked; the lifespan of the patches is increased dramatically; and the putative patch-clustering factor and/or patch-assembly factor remains in the small bud in the absence of continuous exocytosis and endocytosis. Our data support the third possibility.

Patch motility is unlikely to be the answer. First, patches are highly labile structures with a life span of ∼10 s (Smith et al., 2001; Carlsson et al., 2002). Second, most patches display random motion and a few display directed motion. Third, patches have an average speed of 0.49 ± 0.30 μm/s. Together, most patches would have disassembled before they can cross the bud neck. Thus, it has been concluded that the organization of actin patches is due to the assembly of the patches at the sites of polarized growth (Smith et al., 2001). The lifespan of actin patches in different mutants cannot explain the observed phenotype either, because the lifespan of the patches in sec6–4 tpm1–2 tpm2Δ cells increases only twofold over that in tpm1–2 tpm2Δ cells at 36°C.

The patch polarization in sec6–4 tpm1–2 tpm2Δ cells and patch organization in general can be explained by assuming that secretory vesicles carry factors that are required for actin-patch clustering and/or assembly; and that local concentration of these factors depends on the balance of exocytosis and endocytosis (Fig. 7 B). During bud growth of wild-type cells, exocytosis prevails over endocytosis; these factors accumulate dynamically at the sites of polarized growth, leading to polarized actin-patch organization. In the tropomyosin mutants, these putative factors are deposited over the entire cell surface through ongoing depolarized exocytosis, and the high concentration of these factors at the “old buds” is eliminated through active endocytosis, which is not blocked in these mutants (Pruyne et al., 1998); thus, leading to the reorganization of actin patches. In the sec6–4 tpm1–2 tpm2Δ and sec6–4 mutants, exocytosis and endocytosis are blocked together, the high concentration of the putative factors in the old buds remains; thus, actin patches are still polarized, and are still going through their dynamic assembly and disassembly cycle. These putative factors could include polarity proteins such as Cdc42p and Rho1p, and actin-binding proteins such as Aip3p/Bud6p, because their accumulation at the active growth sites depends on intact secretory pathway (McCaffrey et al., 1991; Jin and Amberg, 2000; Wedlich-Soldner et al., 2003).

Our finding on the cause-effect relationship between vesicle fusion and the organization of actin patches has profound biological implications. Exocytosis and endocytosis are intimately coupled in many biological systems including S. cerevisiae (Riezman, 1985), Drosophila (Roos and Kelly, 1999), and mammals (Sudhof, 2000), but the mechanisms are unclear. In S. cerevisiae, many components of the actin patches are required for the internalization step of endocytosis (Munn, 2000); thus, our finding could provide a concrete means to spatially link exocytosis to endocytosis (Fig. 7, A and B): fusion of the exocytic vesicles with the plasma membranes leads to the clustering of actin patches at the fusion site, which mediate endocytosis to recycle the exocytic machinery. This finding also explains why the actin patches and the ends of the actin cables, which mediate exocytosis, are always in close proximity with each other (Karpova et al., 1998).

Yeast strains used in this study are listed in Table III. Standard culture media and genetic techniques were used (Guthrie and Fink, 1991). 1 mg/ml 5-fluroorotic acid (5FOA) (Angus Buffers and Biochemicals) was added to media to select for the loss of URA3-containing plasmids.

Plasmids used in this study include YEplac181 (2 μ, LEU2), pRS316 (CEN, URA3), pFA6a-kanMX6 (Longtine et al., 1998), pYES2-MSB3 (Albert and Gallwitz, 1999), pET22-MSB4 (Albert and Gallwitz, 2000), YEp181–3HA-MSB3 and YEp181–3HA-MSB4 (Bi et al., 2000), pCC904-GIC1 (2 μ, URA3) (Bi et al., 2000), pRS316-CDC42 (Bi and Pringle, 1996), and pRS423-SEC4 (2 μ, HIS3). Plasmids YEp181-SEC4 and pRS316-SEC4 were constructed by inserting a 1.5-kb BamHI-EcoRI fragment carrying SEC4 from pRS423-SEC4 into YEplac181 and pRS316 at the corresponding sites, respectively. GYP1 (from −561 to 2,310) was cloned into YEplac181 vector by plasmid gap-repair. Change of arginine to phenylalanine or lysine codon in MSB3 and MSB4 was achieved by using Quikchange site-directed mutagenesis kit (Stratagene). The Q79L mutation in SEC4 was introduced into YEp181-SEC4 by a PCR-based method. All intended mutations, including msb3-R282F, msb3-R282K, msb4-R200F, msb4-R200K, and sec4-Q79L, were confirmed by DNA sequencing.

To construct strain JGY10 for our first functional assay, the LEU2 gene marking the gic1-Δ1 deletion in YEF1563 (Bi et al., 2000) was replaced precisely with kanMX6 by a PCR-based method (Longtine et al., 1998). The resulting heterozygote (GIC1/gic1-Δ1::kanMX6) carrying plasmid pCC904-GIC1 was sporulated to generate JGY10. To construct JGY18 for our second assay, YEF1291 was crossed to YEF2258 carrying pRS316-CDC42. The resulting heterozygote was sporulated to generate JGY18.

To construct JGY28B (sec2–41), JGY30A (sec6–4), JGY31B (sec9–4), JGY32B (sec3–2), and JGY82B (sec15–1) strains, we crossed wild-type strains from our laboratory (YEF473A or YEF473B) with strains bearing these mutations in another genetic background and isolated segregants of desirable genotypes. Appropriate segregants from one of theses crosses were mated to generate JGY381 (sec6–4/sec6–4). Triple mutants JGY37B (sec3–2 msb3Δ msb4Δ), JGY39A (sec6–4 msb3Δ msb4Δ), and JGY40A (sec9–4 msb3Δ msb4Δ) were isolated from crosses between individual sec mutants and haploid msb3Δ msb4Δ cells.

To construct a strain carrying sec4-Q79L, one copy of SEC4 in YEF473 was replaced with KanMX6 by the PCR-mediated method (Longtine et al., 1998). The resulting heterozygote carrying pRS316-SEC4 was sporulated to generate JGY48A. The 1.5-kb BamHI-EcoRI fragment containing sec4-Q79L was transformed into JGY48A, and colonies that were 5FOA-resistant and G418-sensitive were selected, generating JGY73. JGY86A was a segregant from the cross between JGY82B (sec15–1) and JGY73 (sec4-Q79L) carrying pRS316-SEC4.

To construct strains JGY184 and JGY190, we integrated Tth111I-digested plasmids YIplac-3HA-MSB3 (Bi et al., 2000) and YIplac-3HA-MSB3-R282K at the Tth111I site (−401 position) upstream of the msb3Δ locus in YEF1264. The correctness of the transformants was confirmed by the linkage of HIS3 and URA3 markers in every single segregant and by the expression of HA-tagged proteins. Strains of opposite mating types with desired genotypes were crossed to generate JGY184 and JGY190. Similarly, strains JGY71 and JGY130 were constructed by integrating EcoNI-digested plasmids YIplac-3HA-MSB4 (Bi et al., 2000) and YIplac-3HA-MSB4-R200K at the EcoNI site (−639 position) upstream of the msb4Δ locus in YEF1264.

Production of His₆-tagged Msb3p and its arginine mutants from yeast, and His₆-tagged Msb4p and its arginine mutants from E. coli, and measurement of the GAP activity were all performed as described previously (Albert and Gallwitz, 1999; Albert and Gallwitz, 2000).

For immunoblotting, the mouse monoclonal anti-HA antibody HA.11 (Berkeley Antibody Company), or the rabbit anti-Isp42p polyclonal antibodies were used. Proteins were detected with the ECL Western blotting detection reagents.

For localization of HA-tagged Msb3p or Msb4p, yeast cells grown exponentially in SC-Leu media at 24°C were fixed by formaldehyde and processed for immunofluorescence microscopy as described by Pringle et al. (1991). Mouse anti-HA antibody HA.11 and the secondary Cy3-conjugated donkey anti–mouse IgG antibody were used. For visualizing the actin cytoskeleton, yeast cells were fixed with 4% formaldehyde for 1 h at 24°C and stained with rhodamine-phalloidin (Molecular Probes). DNA was stained with 1 μg/ml bisBenzimide (Sigma-Aldrich). Differential interference contrast (DIC) and fluorescence microscopy were performed using Nikon Microscope ECLIPSE E800 (Nikon Corporation) with a 60× plan apo objective. The images were acquired using Image-Pro Plus software (Media Cybernetics).

Yeast cells were grown in YPD media to early log phase at 24°C. For cdc42-Ts mutants, cells grown exponentially at 24°C were shifted to restrictive temperatures for 1 h, and were prefixed with 1.6% of glutaraldehyde in culture for 10 min. 10–20-A₆₀₀ units of cells were fixed in 1 ml fixative (2% glutaraldehyde in PBS buffer, pH 7.4) at 24°C for 30 min, which was followed by additional 30 min with fresh fixative. Cells were then spheroplasted and fixed with 1% glutaraldehyde (in PBS buffer, pH 7.4) at 4°C overnight. Spheroplasts were washed in 0.1 M cacodylate buffer and postfixed twice with ice-cold solution containing 0.5% OsO₄ and 0.8% potassium ferricyanide on ice for 10 min each time. Spheroplasts were then washed with dH₂O and incubated in 2% uranyl acetate at 24°C for 30 min in dark, which was followed by standard dehydration. Cell pellets were embedded in Spurr's resin. Thin sections were cut and processed for electron microscopy. Cells were viewed with a JEOL 1010 electron microscope and photographed at 80 kV. The size of vesicles was measured at 100,000× magnification.

Yeast cells were grown to exponential phase in YPD media containing 5% dextrose at 24°C. Approximately 1.5-A₆₀₀ units of cells were collected, washed twice with YPD media containing 0.1% dextrose, and resuspended in 1.5 ml of the same media. Secretion of invertase was induced for various time at 24°C. Cells from each time point were washed with and resuspended in 1 ml of ice-cold 10 mM NaN₃ solution. A₆₀₀ of this cell suspension was measured. 20 μl of cell suspension was directly applied to assay external (periplasmic) pool of invertase. For assaying the internal (intracellular) pool of invertase, 0.5 ml cell suspension was mixed with 0.5 ml 2× spheroplast cocktail mix (2.8 M sorbitol, 0.1 M Tris-HCl, 10 mM NaN₃, pH 7.5, 0.4% β-mercaptoethanol) and 50 μl 10 mg/ml lyticase to remove the cell wall. Spheroplasts were lysed in 0.5 ml 0.5% Triton X-100. The lysate was centrifuged at 14,000 rpm for 10 min at 4°C. 20 μl of supernatant was used to assay internal pool of invertase. The external and internal pools of invertase activity were assayed by following the protocol described by Adamo et al. (1999).

Yeast cells were grown to midlog phase at 24°C. Half of the culture was kept at 24°C and the other half was shifted to 37°C for 1 h. At the end of shift, NaN₃ and NaF were added to ∼20 mM each to all cultures. 25-A₆₀₀ units of cells from each culture were washed in 20 mM NaN₃/20 mM NaF twice and resuspended in 830 μl of spheroplasting solution (100 mM Tris-HCl, pH 7.5, 1.6 M sorbitol, 12 mM NaN₃, 0.1% β-mercaptoethanol, 200 μg/ml zymolyase 100-T). Spheroplasts were gently pelleted at 2,000 rpm for 5 min. The top 830 μl of the supernatant (external Bgl2p pool) was transferred to a new tube and 170 μl of 6× sample buffer (0.35 M Tris-HCl, pH 6.8, 30% glycerol, 10% SDS, 93 mg/ml DTT, 0.1 mg/ml bromphenol blue) was added to it, and the samples were boiled immediately for 10 min. The pellet (internal Bgl2p pool) was resuspended in 1 ml of 2× sample buffer and boiled for 10 min 50 μl of samples were separated by a 12.5% SDS-polyacrylamide gel, probed with a rabbit α-Bgl2p antibody, and detected by ECL.

Cdc42-Ts mutant cells harboring YEplac181-based plasmids were grown on SC-Leu plates at 24°C for 4 d. Cells were then scraped off plates and resuspended in 25 ml of 1 M sorbitol + 50% SC-Leu. Unbudded cells were enriched by repeated centrifugation at 800 rpm for 1 min until more than 90% of cells was unbudded. 5–10-A₆₀₀ units of the enriched cells were transferred into 40 ml SC-Leu media and incubated at restrictive temperatures until ∼50% cells from the positive samples became budded. Cells were then fixed with 4% formaldehyde at the restrictive temperatures for 15 min. For 24°C controls, the remaining enriched cells were fixed in 4% formaldehyde at 24°C for 1 h. Fixed cells were stained for F-actin and DNA.

To determine actin-patch organization, an URA3-marked integrative plasmid pASF125 carrying an NH₂-terminal tagged GFP-TUB1 was linearlized by StuI and integrated at the ura3 locus in yeast strains ABY971, ABY973, ABY999, and JGY381, respectively. The resulting strains were inoculated in 50 ml YPD media and incubated at 24°C until A₆₀₀ reached 0.2–0.4. Approximately 4-A₆₀₀ units of cells were diluted into 20 ml YPD and incubated in a water-bath shaker at 36°C. After incubation at 36°C for 30 min or 60 min, formaldehyde was added directly into cultures to 4% final concentration. Fixed cells were stained for F-actin. Only small-budded cells with a short spindle were scored for actin-patch organization.

To determine the lifespan of actin patches, plasmid pRB2139 carrying ABP1-GFP was transformed into strains ABY971, ABY973, ABY999, and JGY381, respectively. 5 μl of exponentially growing cells in SC-Ura medium at 20°C were spotted onto a flat agarose patch that contains the same medium with 2% agarose on a microscope slide, covered with a cover glass, and sealed with nail polish. Time-lapse series were obtained with a computer-controlled Nikon E800 microscope and with a digital camera (model 4742-95; Hamamatsu Photogenics), using the Image-Pro Plus software. 31 images of each cell expressing Abp1p-GFP were acquired in a single focal plane at 20°C and at 36°C using 0.5-s exposure and 2-s interval. For 36°C studies, sealed slides were preincubated in a moistened chamber at 36°C for 60 min, and quickly put under an objective lens heated to 36°C during the course of the time-lapse study. Images were analyzed using NIH ImageJ 1.29 software (http://rsb.info.nih.gov/). The lifespan of an Abp1p-GFP patch is defined as the period between the first frame that it appears and the first frame that it becomes completely invisible. Only budded cells, including the mother and daughter compartments, were analyzed.

We thank Drs. P. Brennwald, D. Pruyne, A. Bretscher, D. Amberg, and J. Pringle for providing antibody, plasmids, and yeast strains; U. Welscher-Altschäffel and N. Shah for expert technical assistance; and Drs. D. Lew, S. Zigmond, and M. Chou for critically reading the manuscript.

This work was supported by National Institutes of Health grants GM59216 to E. Bi. and GM61221 to C. Burd, the Max Planck Society, and grants to D. Gallwitz from the Deutsche Forschungsgemeinschaft and Fonds der Chemischen Industrie and Human Frontier Science Program.