The role of cell cycle–regulated expression in the localization of spatial landmark proteins in yeast

A central feature of cellular morphogenesis is cell polarization, which involves the asymmetric organization of the cytoskeleton, secretory system, and plasma membrane components along an appropriate axis (Drubin and Nelson, 1996). In the budding yeast Saccharomyces cerevisiae, such polarization allows asymmetric growth to form a bud, which becomes the daughter cell. An important feature of cell polarization is the selection of an axis. In S. cerevisiae, this is manifested in the selection of bud sites, which occurs in two distinct patterns depending on cell type (Hicks et al., 1977; Chant and Pringle, 1995). In the axial pattern, as seen in normal haploid cells, the daughter cell's first bud forms adjacent to the division site (as marked by the birth scar), and each subsequent bud forms adjacent to the immediately preceding bud site (as marked by the bud scar). This pattern appears to depend on a transient cortical marker that involves the Bud3p, Bud4p, and Axl2p/Bud10p proteins (Chant and Herskowitz, 1991; Chant and Pringle, 1995; Chant et al., 1995; Halme et al., 1996; Roemer et al., 1996a; Sanders and Herskowitz, 1996; Sanders et al., 1999; Lord et al., 2000); during each cell cycle, this marker is deposited at the mother-bud neck and then distributed to the division site on both mother and daughter cells.

In contrast, the bipolar pattern, as seen in normal diploid cells, appears to depend on persistent markers that are deposited at both the birth-scar-distal and birth-scar-proximal poles of the daughter cell, as well as at the division site on the mother cell (Chant and Pringle, 1995). These markers can direct bud formation to the marked site either in the next cell cycle or in a later one. A screen for mutants defective specifically in bipolar bud-site selection led to the identification of the BUD8 and BUD9 genes, whose mutant phenotypes suggest that they encode components of the markers at the distal and proximal poles of the daughter cell, respectively (Zahner et al., 1996; Harkins et al., 2001). The biochemical properties and localizations of Bud8p and Bud9p are consistent with this hypothesis. Both proteins appear to be integral membrane proteins of the plasma membrane, and each protein consists of a long, heavily glycosylated, NH₂-terminal extracytoplasmic domain followed, at the COOH terminus, by two apparent transmembrane domains bracketing a cytoplasmic domain of ∼40 amino acids. The Bud8p and Bud9p cytoplasmic domains are very similar to each other in sequence and presumably provide the recognition sites for the Rsr1p/Bud2p/Bud5p GTPase signaling module, which appears to transmit the positional information from the axial and bipolar cortical markers to the proteins responsible for cell polarization (Pringle et al., 1995; Roemer et al., 1996b; Chant, 1999; Park et al., 1999; Kang et al., 2001; Marston et al., 2001). As expected (Chant and Pringle, 1995; Amberg et al., 1997), Bud8p appears to be localized primarily to the presumptive bud site, the distal pole of the bud, and the distal pole of the daughter cell (Taheri et al., 2000; Harkins et al., 2001). In addition, although Bud9p has also been reported to localize to the distal pole (Taheri et al., 2000), other data suggest strongly that it actually localizes primarily to the bud side of the mother-bud neck and thence to the proximal pole of the daughter cell (Harkins et al., 2001; this study).

In this study, we investigated when and how the very similar Bud8p and Bud9p proteins are localized to the opposite poles of the cell. We found that Bud8p is delivered to the bud site early in the cell cycle, whereas Bud9p is delivered to the neck just before cell division. The timing and site of localization of each protein appear to depend primarily on when in the cell cycle its gene is transcribed.

In previous studies, although Bud8p and Bud9p each localized primarily to the expected pole (Introduction), each protein was also observed in some cells at the site expected for the other (Taheri et al., 2000; Harkins et al., 2001). It seemed likely that these unexpected features of Bud8p and Bud9p localization were artifacts of overexpression of the proteins and/or of expression at inappropriate times in the cell cycle. Thus, we reexamined the localization of these proteins using reagents and procedures that were improved in several respects. First, plasmids were constructed using a triple-mutant green fluorescent protein (GFP)^* variant that resulted in brighter and more stable signals. Second, because our previously used GFP–BUD8 construct contained only 290 bp of BUD8 upstream sequences, which might have truncated the promoter and thus caused mistimed and/or abnormal levels of expression, we constructed plasmids that contained additional BUD8 upstream sequence (a total of 1410 bp; see Supplemental materials and methods, available). (Although we showed that constructs with 290 bp of upstream sequence could rescue the bud8 mutant phenotype [Harkins et al., 2001], Mösch and Fink [1997] had identified a bud8 mutant with a transposon insertion 565 bp upstream of the start codon, suggesting that the BUD8 promoter might be ≥565 bp in length.) Third, observations were made using a sensitive CCD camera and appropriate image-processing software. Even with these procedures, we were unable to observe strong or consistent signals when the GFP fusion proteins were expressed from low-copy plasmids. However, using high-copy plasmids, we obtained results that were more consistent from cell to cell, simpler, and more easily reconciled with other evidence about the functions of Bud8p and Bud9p.

As in our previous studies, we observed GFP–Bud8p primarily at the tips of buds of various sizes and at the distal poles and nascent budding sites on unbudded cells (Fig. 1 A, cells 1–8). A detectable GFP-Bud8p signal was observed at the bud tip in 54% (n = 254) of budded cells, whereas a detectable signal was present at one pole in 44% (n = 234) of unbudded cells (61% of unbudded daughters; n = 200). Virtually no unbudded cells were observed with GFP–Bud8p signal at both poles, and the pole with signal was always the distal pole (Fig. 1 A, cells 5–8). Moreover, the unbudded cells with a GFP–Bud8p signal were essentially all newborn daughter cells (with no bud scars), a result that was expected given that daughter cells nearly always make their first buds at their distal poles (Chant and Pringle, 1995) and that the Bud8p at that pole appears to pass to the bud during bud emergence (Harkins et al., 2001). In addition, as observed previously, GFP–Bud8p was sometimes detected at the mother-bud neck (Fig. 1 A, cell 4). However, in contrast to our previous observations, such signal was now observed infrequently (13% of the 254 budded cells counted) and was always on the mother side, rather than the bud side, of the neck, consistent with the absence of unbudded daughter cells with GFP–Bud8p signal at their proximal poles. A signal was observed at the neck only in large-budded cells (∼25% of such cells), and was seen only in cells that also had a detectable patch of GFP–Bud8p at the bud tip.

As observed previously (Harkins et al., 2001), the diffuse patch of GFP–Bud8p at the distal pole of the newborn daughter cell became a tighter, brighter patch just before bud emergence at that site (Fig. 1, compare cell 8 to cells 5–7; and Fig. 2 C, cell b). It was difficult to tell whether the tighter patch represented coalesced older material, newly delivered material, or both. Because nearly all daughter cells make their first buds at their distal poles, this also made it difficult to answer the more general question of whether new Bud8p is normally delivered to the bud site at the time of bud emergence. Previously, we had observed that starved cells with no detectable GFP–Bud8p signal developed a tight patch of GFP–Bud8p at the presumptive bud site before bud emergence after refeeding (Harkins et al., 2001). We now extended these observations in several ways. First, after staining normal cells with Calcofluor, we could find cells with tight, bright GFP–Bud8p patches at the tips of small buds in various positions relative to the distal pole and on mother cells of various ages (Fig. 2 A). Similar observations (Fig. 2 B) were made using an rsr1 mutant strain, which buds in random locations (Bender and Pringle, 1989). Finally, in time-lapse experiments, we observed tight patches of GFP–Bud8p appearing approximately coincident with bud emergence at bud sites remote from the distal pole and/or on previously budded mother cells (Fig. 2 C, arrows). Taken together, the results suggest strongly that whatever the fate of the older Bud8p, new Bud8p is delivered to the nascent bud site shortly before or coincident with bud emergence.

As in our previous studies, we observed GFP–Bud9p at the bud side of the neck in budded cells and at the proximal poles of unbudded cells (Fig. 1 B). Only large-budded cells had detectable GFP–Bud9p signal at the neck, and signal was observed in 25% (n = 200) of such cells. (Much higher percentages were observed in experiments using synchronized cells, as described below). 46% (n = 228) of unbudded cells had detectable GFP–Bud9p at the proximal pole, and these cells were nearly all newborn daughter cells (Fig. 1 B). The signal at the proximal pole was still detectable in 29% (n = 204) of daughter cells during the first budding cycle (Fig. 1 B, *), but it was only rarely detected (<1%) in cells budding for the second or third time. In contrast to our previous observations (made using GFP–BUD9 expressed from the GAL promoter), in the present studies we rarely observed GFP–Bud9p at bud tips or the distal poles of unbudded cells.

The absence of detectable GFP–Bud9p at the neck in small-budded cells suggested that Bud9p is delivered to the neck late in the cell cycle. To investigate this issue further, we first examined exponentially growing cells that had been stained with DAPI to reveal their nuclear morphology. In cells that had not completed anaphase, we never observed GFP–Bud9p at the neck (Fig. 2 D, cells 1 and 5). In contrast, 58% (n = 207) of cells that had completed anaphase had detectable GFP–Bud9p at the neck (Fig. 2 D, cells 2–4, 6, and 7). We then examined cells that had been arrested by a temperature-sensitive mutation in CDC15, whose product is a component of the mitotic exit network that controls exit from mitosis and entry into cytokinesis (Morgan, 1999; Lee et al., 2001; McCollum and Gould, 2001; Menssen et al., 2001). The arrested cells displayed no GFP–Bud9p at their necks (Fig. 2 E). However, within 45 min of return to permissive temperature, 68% (n = 201) of the cells displayed GFP–Bud9p at the bud side of the neck (Fig. 2 F). In a parallel experiment, cdc15 cells expressing GFP–Bud8p showed a broad patch of signal at the distal pole of the bud during arrest at restrictive temperature (Fig. 2 G). Upon return to permissive temperature, no concentration of GFP–Bud8p was observed at the neck before division (Fig. 2 H), although, as expected, tight patches of GFP–Bud8p were observed at the nascent bud sites in the following cell cycle (Fig. 2 I). Thus, most or all new Bud9p appears to be delivered to the neck very late in the cell cycle, just before cytokinesis, and the delivery of Bud8p and Bud9p appears to be under different cell cycle control.

As integral-membrane proteins, Bud8p and Bud9p are presumably delivered to the plasma membrane in vesicles derived from the secretory system. Both early in the cell cycle (when Bud8p appears to be delivered) and late in the cell cycle (when Bud9p appears to be delivered), polarized delivery of secretory vesicles to the cell surface appears to depend largely on the actin cytoskeleton (Lew and Reed, 1995; Drubin and Nelson, 1996; Pruyne and Bretscher, 2000a, 2000b). Consistent with this model, we observed previously that GFP–Bud8p delivery to presumptive bud sites was abolished in cells whose F-actin was ablated by treatment with Latrunculin A (Lat A) (Harkins et al., 2001). Similarly, treatment with Lat A also abolished the delivery of GFP–Bud9p to the neck late in the cell cycle (Fig. 3 A; in the DMSO-only control, 59% [n = 200] of the cells showed GFP–Bud9p at the neck), indicating that this delivery is also actin dependent. In addition, experiments with a temperature-sensitive septin mutant indicated that delivery of Bud9p to the neck is also septin dependent (Fig. 3 B). In contrast, delivery of Bud8p to presumptive bud sites and bud tips appeared to be unaffected by disruption of the septin/scaffold (Harkins et al., 2001).

Although Bud8p and Bud9p are structurally similar proteins, it is possible that targeting signals within the polypeptides may cause (or contribute to) their distinct localizations. However, the data described above suggested that timing of expression might also be important for localization and function. Consistent with this possibility, microarray analyses have indicated that the BUD8 and BUD9 mRNAs accumulate at distinct times in the cell cycle (Cho et al., 1998; Spellman et al., 1998). To explore this possibility further, we first reanalyzed the BUD8 and BUD9 mRNA levels through the cell cycle using Northern blotting and two different methods of synchronization. In general agreement with the microarray analyses, we observed that BUD8 mRNA peaks in G2/M, coincident with CLB2 mRNA, whereas BUD9 mRNA peaks in late G1, coincident with CLN2 mRNA (Fig. 4 A). Although consistent with the hypothesis that differential timing of gene expression is important for the distinct functions of Bud8p and Bud9p, these results were also surprising in that each mRNA peaked long before the corresponding protein appears to be delivered to the cell surface.

To test directly whether the timing of gene expression is critical for the localization and/or function of Bud8p and Bud9p, we then constructed plasmids and strains in which an untagged or tagged copy of each gene was expressed from the promoter of the other gene. To confirm that exchanging the promoters really did alter the timing of gene expression, we examined RNA from a synchronized population of cells that had both a normal BUD8 locus and a P_BUD9HA-BUD8 construct in place of the BUD9 locus. As expected, the HA-BUD8 mRNA showed a peak in late G1 (Fig. 4 B), the time at which BUD9 mRNA normally peaks (see above).

We next examined the function of the chimeric genes by evaluating their abilities to rescue the bud8 and bud9 mutant phenotypes. As expected, the absence of proximal-pole budding in a bud9/bud9 strain (Fig. 5, A, panel 1, and B, panel 1) could be rescued by a plasmid containing a normal BUD9 gene (Fig. 5 B, panel 2; this budding pattern was essentially the same as that in wild-type cells [Harkins et al., 2001]). In contrast, neither low-copy (unpublished data) nor high-copy (Fig. 5 B, panel 3) BUD8 expressed from its own promoter could rescue the bud9/bud9 mutant phenotype. Remarkably, however, low-copy plasmids expressing either untagged or tagged BUD8 from the BUD9 promoter could rescue proximal-pole budding in the mutant strain (Fig. 5, A, panel 2, and B, panels 4 and 6). In contrast, a plasmid expressing BUD9 itself from the BUD8 promoter could not rescue the mutant phenotype (Fig. 5 B, panel 5). As expected if expression in late G1 is critical for BUD9 function, either BUD8 or BUD9 could partially rescue the phenotype of the bud9/bud9 strain when expressed from the CLN2 promoter but not when expressed from the CLB2 promoter (Fig. 5 B, panels 7–10).

Consistent with these tests of function, when GFP–Bud8p was expressed from the BUD9 promoter, signal was observed predominantly in the locations normal for Bud9p (Fig. 1 D), despite the potential complications arising from the need to use a high-copy plasmid in these experiments. In particular, 27% (n = 218) of large-budded cells (but no small-budded cells) had detectable GFP–Bud8p signal at the daughter side of the neck, whereas 67% (n = 239) of unbudded daughter cells had signal at the proximal pole. In addition, 15% (n = 410) of small-budded cells and 24% (n = 404) of large-budded cells had detectable GFP–Bud8p signal at the bud tip (Fig. 1 D, *). However, these signals were typically rather faint, and daughter cells with detectable signal at the distal pole were rare: just 2.5% (n = 440) of unbudded cells had signal detectable at both poles or at the distal pole only.

Given the results just described, it was surprising that the reciprocal experiments presented a more complicated picture. As expected, the absence of distal-pole budding in a bud8/bud8 strain (Fig. 6, A, panel 1, and B, panel 1) could be rescued by a plasmid containing a normal BUD8 gene (Fig. 6 B, panel 2; this budding pattern was nearly the same as that in wild-type cells [Harkins et al., 2001]). In contrast, neither low-copy (unpublished data) nor high-copy (Fig. 6 B, panel 3) BUD9 expressed from its own promoter could rescue the bud8/bud8 mutant phenotype. However, although a plasmid expressing BUD8 from the BUD9 promoter also rescued distal-pole budding only poorly (Fig. 6 B, panel 5), plasmids expressing BUD9 from the BUD8 promoter (with either the BUD8 or BUD9 downstream sequences) were only slightly more effective (Fig. 6, A, panel 2, and B, panels 4 and 6). Moreover, although as expected, a plasmid expressing BUD8 from the CLB2 promoter rescued distal-pole budding well (Fig. 6 B, panel 7), and a plasmid expressing BUD9 from the CLN2 promoter showed no rescue (Fig. 6 B, panel 10), a plasmid expressing BUD9 from the CLB2 promoter showed only partial rescue of distal-pole budding (Fig. 6 B, panel 9) and was actually somewhat less effective than a plasmid expressing BUD8 from the CLN2 promoter (Fig. 6 B, panel 8). Although these results are difficult to interpret in detail, they do suggest that effective Bud8p function requires some aspects of the Bud8p protein that are not fully shared by Bud9p.

Despite these complications, examination of cells expressing GFP–Bud9p from the BUD8 promoter confirmed the importance of the promoter in determining the localization of the protein. In particular, under these conditions, GFP–Bud9p was seen only at the tips of buds and at the distal poles of daughter cells (Fig. 1 C), and never in the normal Bud9p locations, consistent with the failure of P_BUD8BUD9 to rescue proximal-pole budding in a bud9/bud9 strain (see above). Moreover, the GFP–Bud9p signals seen upon expression from the BUD8 promoter were considerably weaker (and detectable in a smaller fraction of the cells) than for the other cases considered here, suggesting that the failure of P_BUD8BUD9 to effectively rescue distal-pole budding in a bud8/bud8 mutant strain might reflect inefficient delivery of Bud9p to the normal Bud8p location, poor stability once there, or both. Some support for this hypothesis was provided by the observation that in some cells expressing GFP–Bud9p from the BUD8 promoter, a signal was detected in internal vesicles (Fig. 1 C, asterisks).

In an attempt to determine what portion(s) of Bud8p are necessary to have effective Bud8p function, we constructed several genes expressing chimeric Bud8p/Bud9p proteins from the BUD8 promoter and tested their abilities to rescue distal-pole budding in a bud8/bud8 mutant strain. Interestingly, a protein comprised of the extracytoplasmic domain of Bud8p and the transmembrane and cytoplasmic domains of Bud9p rescued the bud8/bud8 mutant phenotype essentially as well as did full-length Bud8p (Fig. 7 A; compare Fig. 6 B, panel 2), whereas the reciprocal construct rescued no better than did the expression of full-length Bud9p from the BUD8 promoter (Fig. 7 B; compare Fig. 6 B, panel 4). Thus, the transmembrane and cytoplasmic domains of Bud8p and Bud9p appear to be interchangeable, and the special feature(s) of Bud8p appear to lie in the extracytoplasmic domain. Analysis of additional chimeras suggested that both the NH₂-terminal and COOH-terminal halves of the Bud8p extracytoplasmic domain contribute to Bud8p function (Fig. 7, C and D; compare and A and B, respectively) and that a region near the middle of the domain is important (Fig. 7 E) but not sufficient (Fig. 7 F) for effective Bud8p function.

Several lines of evidence suggest strongly that Bud8p and Bud9p play central roles in marking the distal and proximal poles of the daughter cell, respectively, as potential sites for budding in the bipolar pattern (Introduction). However, it should be noted that each spatial landmark may also involve other proteins; for example, it appears that Rax2p may also form part of the mark at the proximal pole (Chen et al., 2000). In this study, we investigated when and how the structurally similar Bud8p and Bud9p are delivered to the opposite poles of the cell. Although there remain caveats resulting from our inability to visualize the proteins consistently at normal levels of expression, our results indicate that Bud8p is normally delivered to the nascent bud site shortly before or coincident with bud emergence, whereas Bud9p is normally delivered to the bud side of the neck very late in the cell cycle, after activation of the mitotic exit network and just before cytokinesis. These are the times at which secretory vesicles are delivered to these sites by the actin cytoskeleton (Lew and Reed, 1995; Drubin and Nelson, 1996; Pruyne and Bretscher, 2000a,b), so it is not surprising that the localization of both Bud8p (Harkins et al., 2001) and Bud9p (this study) to the appropriate sites is actin dependent. In addition, because localization of proteins to the neck typically depends on the septins (Gladfelter et al., 2001), it is not surprising that the localization of Bud9p is also septin dependent. The mechanisms by which Bud9p and certain other neck-localized proteins (Gladfelter et al., 2001) become distributed asymmetrically on the septin scaffold remain to be elucidated.

The timing of gene expression appears to be a primary determinant of Bud8p and Bud9p localization and function: when Bud8p is expressed from the BUD9 promoter, it localizes as if it were Bud9p and appears fully competent to provide Bud9p (but not Bud8p) function. Moreover, when Bud9p is expressed from the BUD8 promoter, it localizes as if it were Bud8p and is unable to provide Bud9p function. However, under these conditions, the Bud9p localization signal appears weak, and the protein at the distal pole is only partially effective in providing Bud8p function. Thus, there appear to be features of the Bud8p polypeptide itself that are important for its efficient delivery, stability, and/or function at the distal pole. Although these features have not yet been defined in detail, experiments with chimeric proteins indicate that they lie primarily in the central and NH₂-terminal portions of the extracytoplasmic domain. Because it seems likely that the extracellular domains of Bud8p and Bud9p must interact with components of the cell wall in order to maintain the localization of the proteins, it is possible that the special feature(s) of Bud8p allow it to interact effectively with some (unknown) component of the wall at the distal pole.

Consistent with the evidence for a role of cell cycle–regulated expression, the BUD8 and BUD9 mRNAs show distinct peaks in abundance during the cell cycle (Cho et al., 1998; Spellman et al., 1998; this study). For several reasons, these peaks are likely to reflect cell cycle–regulated transcription. First, this is the most common mechanism for cell cycle regulation of transcript levels (Spellman et al., 1998). Second, consistent with the apparent times of transcription of the two genes, the BUD9 upstream sequences contain potential binding sites for the MBF and SBF transcription factors, whereas the BUD8 upstream sequences contain potential binding sites for the MCM1 + SFF transcription factor (Spellman et al., 1998). Finally, despite the complications that might result from using promoters of different inherent strengths, expression of either Bud8p or Bud9p from the CLN2 promoter, which is known to drive G1 transcription (Breeden, 1996; Spellman et al., 1998), can provide partial Bud9p function, whereas expression of either protein from the CLB2 promoter, which is known to drive G2/M transcription (Spellman et al., 1998), can provide partial Bud8p function.

However, these results also raise an important question: Why are the BUD8 and BUD9 mRNAs expressed so far in advance of when the corresponding proteins appear to arrive at the cell surface? The experiments involving block and release by a temperature-sensitive mutation affecting the mitotic exit network are particularly intriguing: in the cdc15-arrested cells, GFP–Bud9p is not detectable at the cell surface, and GFP–Bud8p is present only as a diffuse patch at the distal pole of the bud (as would be expected from its appearance in large-budded cells during an unperturbed cell cycle). Upon release by return to permissive temperature, GFP–Bud9p appears at the bud side of the neck before cell division, whereas no detectable new delivery of GFP–Bud8p to the cell surface occurs until the beginning of the next cell cycle, when it appears at the nascent bud site in the usual way. The several possible explanations for these observations include cell cycle–regulated delays in mRNA translation (conceivably coupled to a localization of the mRNAs themselves) or in the delivery of Bud8p- and Bud9p-containing secretory vesicles to the cell surface. We cannot yet discriminate between these (or perhaps other) explanations. In further studies of this issue, it will be necessary also to explain how starved cells (which are arrested in G0/G1) (Pringle and Hartwell, 1981) can deliver new Bud8p to the nascent bud site upon refeeding (Harkins et al., 2001) without passing through the stage of the cell cycle in which BUD8 appears to be transcribed.

A concern in these studies was whether a lag in the development of fluorescence by the GFP-tagged proteins might complicate the interpretation of the results. That is, it seemed possible that Bud8p and/or Bud9p might be translated and delivered to the plasma membrane immediately after the corresponding mRNA was expressed, but that the fluorescence of the tagged proteins might become detectable only later. However, such a model is difficult to reconcile with several features of the available data. First, to explain the appearance of new GFP–Bud8p at the nascent bud site just at the time of bud emergence, we would need to assume that protein synthesized in G2/M of one cell cycle could be delivered at that time to the precise spot that would be used for budding in the next cell cycle. This seems quite unlikely, particularly in a case like that of the random-budding rsr1 mutant. Second, the absence in the cdc15-arrested cells of any spot of GFP–Bud8p resembling those seen at new bud sites also appears to argue against the model, as does the appearance of GFP–Bud8p at the nascent bud site when starved cells are refed (Harkins et al., 2001). Third, the absence of GFP–Bud9p at the neck of cells arrested at the cdc15 block for 3–4 h (about twice the length of the normal cell cycle) is also difficult to reconcile with the model. Although this experiment could be misleading if the GFP has greater difficulty in achieving a fluorescent state at 37°C, we also observed that cells arrested in G2/M at 23°C by nocodazole and benomyl had no detectable GFP–Bud9p at the neck even after 3 h of incubation (unpublished data). Thus, it seems safe to conclude that there really is a substantial delay between the expression of the BUD8 and BUD9 mRNAs and the arrival of the proteins at the cell surface.

It has long been known that many yeast genes are expressed in a cell cycle–regulated manner (Johnston, 1990; Breeden, 1996), and recent microarray experiments have indicated that the total of such genes is in the range of 416 (Cho et al., 1998) to 800 (Spellman et al., 1998). However, the functional significance of this regulation has been tested in relatively few cases, and in many cases in which tests have been done, eliminating the cell cycle regulation of transcription has had little or no effect on cell viability or function (Johnston, 1990; Breeden, 1996; Igual et al., 1996; Detweiler and Li, 1997). Thus, BUD8 and BUD9 provide important examples of cases in which the timing of transcription does appear to be critical for protein localization and function. Another important example, and an interesting comparison to the case of BUD8 and BUD9, is provided by Axl2p/Bud10p, an integral-membrane component of the cortical marker used in axial budding (Halme et al., 1996; Roemer et al., 1996a). Altering the normal late-G1 expression of AXL2 results in a severe disruption of protein localization and function, apparently because expression of the AXL2 mRNA is quickly followed by translation and the delivery of vesicles containing the newly made Axl2p to wherever the general secretory vesicle traffic of the cell is directed at that time in the cell cycle (Lord et al., 2000). Nonetheless, the transport of Axl2p to the cell surface appears to involve distinctive features, because it is affected much more than general secretory traffic by mutations affecting the ER-membrane protein Erv14p (Powers and Barlowe, 1998). Another interesting comparison is provided by the proteins involved in mating-type switching: cell cycle regulation of transcription of the HO gene is part of the complex mechanism (involving also the asymmetric localization of the mRNA for the transcriptional repressor Ash1p) by which the HO protein is localized specifically to the mother cell (Bobola et al., 1996; Sil and Herskowitz, 1996; Long et al., 2001; Maxon and Herskowitz, 2001; and references cited therein). It will be interesting to see if there are many other examples in which protein localization (and hence function) depends on the timing of gene expression, and, if so, which types of mechanisms are involved.

Yeast strains used in this study are listed in Table I; construction of strains LSY305 and LSY388 is described below. Standard methods of yeast genetics were used (Guthrie and Fink, 1991; Gietz et al., 1992). Except where noted, cells were grown at 23°C on YM-P rich liquid medium, YPD rich solid medium, or synthetic complete (SC) medium lacking appropriate nutrients as needed to maintain plasmids (Lillie and Pringle, 1980; Guthrie and Fink, 1991). All media contained 2% glucose as carbon source. Cells to be examined for GFP fluorescence were grown in the dark to minimize photobleaching.

To synchronize cells as a population of unbudded, G1-phase cells, α factor (Sigma-Aldrich) was added to a final concentration of 30 ng/ml to an exponentially growing culture (∼10⁷ cells/ml), and incubation was continued for 90 min. For Northern blot analyses (see below), cells were collected by centrifugation at 650 g for 5 min at 23°C, resuspended at a similar density in fresh medium without α factor, and incubated further. At intervals, samples were collected by centrifugation at high speed in a table-top centrifuge for 2 min at 23°C, and the pellets were flash frozen in an ethanol/dry ice bath. To investigate the actin dependence of Bud9p localization, Lat A (Molecular Probes) was added to a final concentration of 100 μM from a 20-mM stock solution in DMSO. Lat A (or the same concentration of DMSO alone as a control) was added to the synchronized culture 70 min after the shift back to medium without α factor, and the cells were examined 20 min later. To investigate the septin dependence of Bud9p localization, the α factor–arrested cells were resuspended in fresh medium without α factor that had been prewarmed to 37°C, incubation was continued at 37°C, and cells were examined after 90 min.

To synchronize cells late in the cell cycle using the cdc15-2 temperature-sensitive mutation, a culture growing exponentially at 23°C was shifted to 37°C and incubated for 3 to 4 h, at which point ∼98% of the cells had large buds. To release cells from the arrest, they were collected by centrifugation at 650 g for 5 min and resuspended in fresh medium at 23°C.

Plasmids used in this study are listed in Table II and/or described in the Supplemental materials and methods (available). Escherichia coli strains DH12S and DH5α (Life Technologies) were used for plasmid maintenance by standard procedures (Sambrook et al., 1989). Standard methods of DNA manipulation were used (Sambrook et al., 1989; Ausubel et al., 1995). PCR was performed using Taq DNA polymerase (Promega). Other enzymes were purchased from New England Biolabs, and oligonucleotide primers (Integrated DNA Technologies) are listed in Table III (Supplemental materials and methods, available). DNA sequencing was performed by the UNC-Chapel Hill Automated DNA Sequencing Facility.

To construct a strain carrying a complete deletion of the RSR1 ORF, the PCR method (Baudin et al., 1993) was used to replace one wild-type RSR1 allele in YEF473. Primers LS174 and LS175 were used with plasmid pFA6a-His3MX6 (Wach et al., 1997; Longtine et al., 1998) as template. The presence of the deletion (named rsr1-Δ2) was confirmed by PCR (Longtine et al., 1998), and 2:2 segregation of the marker together with a random budding phenotype was observed. Appropriate segregants were mated to obtain the homozygous mutant strain LSY388.

To construct a strain expressing the HA-BUD8 coding region from the chromosomal BUD9 promoter, plasmid YIpP_BUD9HA-BUD8 was digested with HindIII (cutting the plasmid once, in the BUD9 upstream sequence) and transformed into bud9-Δ1 strain LSY87. A Ura⁺ transformant showing restoration of bipolar budding was designated LSY305. PCR using DNA from LSY305 as template and primers LS157 and LS156 (one upstream of the HindIII site in the BUD9 upstream sequences, the other within the BUD8 ORF) confirmed that LSY305 had the desired structure at the BUD9 locus.

Yeast total RNA was isolated by the hot-phenol extraction method as described by Leeds et al. (1991). For Northern blotting, samples of total RNA were denatured by formamide–formaldehyde treatment fractionated on 1% agarose/formaldehyde gels (Ausubel et al., 1995), and then transferred to Hybond-N+ membranes (Amersham Pharmacia Biotech). Hybridization of DNA probes to these RNA blots was essentially as described by Sambrook et al. (1989). To generate probes specific for BUD8, BUD9, CLN2, CLB2, HA, or ACT1, primer pairs LS93 and LS94, LS95 and LS96, LS112 and LS113, LS114 and LS115, LS158 and LS159, and LS183 and LS184, respectively, were used to amplify DNA fragments by PCR from templates YCpBUD8, YCpBUD9, pGAL-CLN2 (a gift from S. Reed via D. Lew [Duke University, Durham, NC]), pGAL-CLB2 (a gift from S. Reed via D. Lew), YCpHA-BUD9, and pRS316-GAL1-ACT1 (Liu et al., 1992), respectively. To allow normalization of the amounts of RNA per lane, a preliminary blot was probed for ACT1, and the signals were quantitated using a Storm 840 Phosphorimager and Imagequant 5.0 software (Molecular Dynamics).

To visualize bud scars and birth scars for the determination of budding patterns, exponentially growing cells were suspended in water containing 1 μg/ml Calcofluor as described by Zahner et al. (1996) and examined on a Nikon Microphot SA microscope using an Apo 60X/1.40 NA oil-immersion objective. To visualize GFP–Bud8p or GFP–Bud9p in living cells, cells were grown to exponential phase in the dark, collected by centrifugation at 23°C, and suspended in water or in water containing 1 μg/ml Calcofluor. To visualize nuclear DNA, the exponentially growing cells were stained by adding DAPI (Sigma-Aldrich) from a 1-mg/ml stock solution in water to a final concentration of 1 μg/ml and continuing to incubate the culture for 90 min before resuspending the cells in water for examination. To prepare slides for time-lapse microscopy, cells growing exponentially in SC-Leu medium were pipetted onto slabs of 2× SC medium (like SC but with all ingredients at twice normal concentration; 2% glucose as carbon source) containing 25% gelatin, and coverslips were applied and sealed with Valap (Salmon et al., 1998). All images for figures were collected using Metamorph software (Universal Imaging Corporation) with a Nikon Eclipse 600 FN microscope equipped with a Hamamatsu ORCA-2 CCD camera and an Apo 100X/1.40 NA oil-immersion objective. Exposure times were set automatically and averaged ∼6 s for GFP images except for those of cells expressing GFP–Bud9p from YEpP_BUD8GFP*–BUD9, which averaged ∼15 s. Exposure times for Calcofluor images averaged ∼0.2 sec.

The online supplement (available) contains the details of plasmid constructions and a table of oligonucleotide primers used.

We thank Amos McKenzie III, Mark Longtine, Dale Beach, David Brenner, Heidi Harkins, I-Ching Yu, Danny Lew, Bruce Futcher, Howard Bussey, and other members of our laboratories for encouragement, helpful discussions, and/or strains and plasmids.

This work was supported by National Institutes of Health Grant GM31006.